The Contribution of Melanopsin Signalling to Image-Forming Vision During Retinal Degeneration

The contribution of melanopsin signalling to image-forming vision during retinal degeneration

A thesis submitted to The University of Manchester for the degree of

Doctor of Philosophy

in the Faculty of Biology, Medicine and Health

2016

Christopher Anthony Procyk

SCHOOL OF BIOLOGICAL SCIENCES

Division of Neuroscience and Experimental Psychology

Table of Contents

Chapter 1: General Introduction ...... 18 1.1 The Vertebrate Retina ...... 19

1.1.1.1 Opsins ...... 21 1.1.1.2 Phototransduction ...... 22 1.1.1.3 Deactivation of the phototransduction cascade ...... 23 1.1.1.4 Chromophore regeneration ...... 24

1.1.2.1 Cone signalling through bipolar cells ...... 25 1.1.2.2 Rod signalling through bipolar cells ...... 25

1.1.4.1 Horizontal cells ...... 28 1.1.4.2 Amacrine cells ...... 29 1.2 Intrinsically Photosensitive Retinal Ganglion Cells ...... 30

1.2.5.1 Melanopsin phototransduction ...... 36 1.2.5.2 Melanopsin bi-stability and chromophore regeneration ...... 37 1.2.5.3 Electrophysiological responses to light ...... 38

1.2.7.1 Central targets of ipRGCs for non-image forming vision ...... 42

1.2.8.1 The Lateral Geniculate Nucleus of the mouse ...... 44 1.2.8.2 Anatomy of the mouse dLGN ...... 45 1.2.8.3 Electrophysiological responses in the mouse dLGN ...... 47 1.3 Melanopsin’s contribution to image forming vision ...... 48

1.4 Retinal degeneration ...... 52

1.4.1.1 Stages of degeneration in the rd1 retina ...... 53

1.5 The melanopsin system in retinal degeneration ...... 59 1.6 Aims and Objectives ...... 61

Chapter 2: General Methods ...... 63 2.1 Transgenic mouse models of retinal degeneration ...... 64

2.2 Presentation of Visual Stimuli ...... 69

2.3 Data collection and Recording methodologies ...... 70

2.3.1.1 The action potential...... 71 2.3.1.2 Initiation ...... 71 2.3.1.3 Propagation & Termination ...... 72

2.3.3.1 Brainbow ...... 74 2.3.3.2 Immunohistochemistry ...... 76 2.3.3.3 Fluorescence microscopy ...... 77 2.3.3.4 Confocal Microscopy ...... 78 2.4 Data Analysis ...... 79

Chapter 3: Identification and anatomical characterisation of all known ipRGC subtypes in the degenerate retina using a novel multi-colour analysis method ... 81 3.1 Abstract ...... 82 3.2 Introduction ...... 83 3.3 Methods ...... 85

3.3.6.1 Filtering images ...... 88

3.3.6.2 Principal component analysis and cluster isolation...... 89 3.3.6.3 Reconstructing individual cells ...... 90 3.3.6.4 Quantitative analysis of reconstructed cells...... 91 3.4 Results ...... 92 3.5 Discussion ...... 113 3.6 References ...... 122

Chapter 4: Spatial receptive fields in the retina and dorsal lateral geniculate nucleus (dLGN) of mice lacking rods and cones ...... 125 4.1 Abstract ...... 126 4.2 Introduction ...... 127 4.3 Methods ...... 129

4.4 Results ...... 134 4.5 Discussion ...... 144 4.6 References ...... 147

Chapter 5: Delayed maturation of ipRGCs projecting to the dorsal Lateral Geniculate Nucleus (dLGN) ...... 149 5.1 Abstract ...... 150 5.2 Introduction ...... 151 5.3 Methods ...... 153

5.4 Results ...... 158 5.5 Discussion ...... 172 5.6 References ...... 177

Chapter 6: General Discussion ...... 180 6.1 Implications ...... 182

6.2 General discussion about experimental techniques ...... 188

6.3 Future work...... 194

Chapter 7: General References ...... 200

Appendix A………………………………………………………………………………..…………………….222

Word Count = 70,293

Table of Figures

Chapter 1: General Introduction

Figure 1.1 The vertebrate retina ...... 19 Figure 1.2 Spectral sensitivity of visual photopigments in the mouse eye ...... 21 Figure 1.3 Phototransduction in rod and cone photoreceptors ...... 23 Figure 1.4 Spatial receptive field structure of an individual retinal ganglion cell ...... 27 Figure 1.5 Establishment of the centre-surround organisation ...... 29 Figure 1.6 Somato-dendritic profile of the five ipRGC subtypes ...... 32 Figure 1.7 Summary of ipRGC subtypes ...... 34 Figure 1.8 Phototransduction in ipRGCs ...... 37 Figure 1.9 Melanopsin-driven light responses in the mouse ...... 39 Figure 1.10 Synaptic inputs to ipRGCs in the mouse retina ...... 41 Figure 1.11 Projection targets of ipRGCs ...... 43 Figure 1.12 The Lateral Geniculate Nucleus of the mouse...... 45 Figure 1.13 Retinotopic organisation of the mouse visual system...... 46 Figure 1.14 Light response profiles in the mouse dLGN...... 48 Figure 1.15 Stages of retinal degeneration in the rd1 mouse – Phase I ...... 54 Figure 1.16 Stages of retinal degeneration in the rd1 mouse – Phase II ...... 56 Figure 1.17 Stages of retinal degeneration in the rd1 mouse – Phase III ...... 57

Chapter 2: General Methods

Figure 2.1 Gene construct of the Opn4Cre/+ mouse...... 66 Figure 2.2 Sequence structure of an individual loxP site ...... 67 Figure 2.3 Action of Cre-LoxP recombination ...... 68 Figure 2.4 Spectral sensitivity of photopigments in the rd/rd Opn1mwR mouse ...... 68 Figure 2.5 The action potential ...... 72 Figure 2.6 Extracellular multi-channel recording electrode ...... 73 Figure 2.7 Brainbow viral construct ...... 74 Figure 2.8 Brainbow recombination using mutant loxP sites...... 75 Figure 2.9 Colour variation from stochastic Brainbow expression ...... 77 Figure 2.10 Analysing electrophysiological data ...... 80

Chapter 3: Identifying and anatomical characterisation of all known ipRGC subtypes in the degenerate retina using novel multi-colour analysis techniques

Figure 3.1 Viral transduction of Brainbow virus in the rd/rd Opn4Cre/+ mouse ...... 93 Figure 3.2 Stochastic expression of Brainbow cassettes in the rd/rd Opn4Cre/+ mouse ...... 97 Figure 3.3 Processing retinal Brainbow images ...... 101 Figure 3.4 Modelling Colour and brightness with artificial images ...... 103 Figure 3.5 Modelling noise with artificial images ...... 105 Figure 3.6 Single cell isolation of naturalistic Brainbow images from the rd/rd Opn4Cre/+ retina ...... 107 Figure 3.7 Categorisation of the M1, M3 and M5 ipRGC subtypes in the rd/rd Opn4Cre/+ retina ...... 109 Figure 3.8 Categorisation of the M2 and M4 ipRGC subtypes in the rd/rd Opn4Cre/+ retina ...... 111 Figure 3.9 Anatomical characteristics of ipRGC subtypes from the degenerate and wildtype retina 112

Chapter 4: Spatial receptive fields in the retina and dorsal Lateral Geniculate Nucleus (dLGN) of mice lacking rods and cones

Figure 4.1 Identification of single units in the retina and dLGN of the rd/rd cl mouse ...... 132 Figure 4.2 Spatial receptive field properties of two representative ipRGCs recorded from the rd/rd cl retina ...... 135 Figure 4.3 Receptive field properties of ipRGC population recorded from the rd/rd cl retina ...... 136 Figure 4.4 Receptive field properties of two representative melanopsin driven neurones in the rd/rd cl dLGN ...... 138 Figure 4.5 Receptive field properties of melanopsin-driven responses in the rd/rd cl dLGN ...... 139 Figure 4.6 Retinotopic organisation of the rd/rd cl visual system ...... 142 Figure 4.7 Temporal fidelity of the melanopsin-driven light-responses in the rd/rd cl dLGN...... 143

Chapter 5: Delayed Maturation of ipRGCs projecting to the dorsal Lateral Geniculate Nucleus

Figure 5.1 Dark adapted light-responses in the rd1 Opn1mwR dLGN ...... 159 Figure 5.2 Contrast sensitivity in the rd1 dLGN ...... 161 Figure 5.3 Spatial receptive fields in the rd1 dLGN ...... 163 Figure 5.4 Design and validation of silent substitution stimuli ...... 165 Figure 5.5 Contribution of melanopsin signalling to the juvenile rd1 dLGN ...... 167 Figure 5.6 Contribution of melanopsin signalling to the adult dLGN and juvenile OPN ...... 169 Figure 5.7 Melanopsin expression in the juvenile Opn1mwR and rd1 Opn1mwR retina ...... 171

Chapter 6: General Discussion

Figure 6.1 Effects of isoflurane anaesthesia on light responses in the dLGN ...... 191

Abbreviations

AC - amacrine cell ATP - adenosine triphosphate AMPA - α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid BFP - blue fluorescent protein cGMP - cyclic guanosine monophosphate CGNA3 - cone cGMP-gated cation channel (CNG3 subunit) dLGN - dorsal Lateral Geniculate Nucleus DAG - diacylglycerol ERG - electroretinogram eYFP - enhanced Yellow Fluorescent Protein GABA - gamma-aminobutyric acid GCL - ganglion cell layer GDP - guanosine diphosphate GFP - green fluorescent protein Gnat1 - rod transducin alpha-subunit GPCR - G-protein coupled receptor G-protein - GTP binding protein GTP - guanosine triphosphate HC - horizontal cell IF - image forming IGL - inter geniculate leaflet INL - inner nuclear layer IP3 - inositol triphosphate IPL - inner plexiform layer ipRGC - intrinsically photosensitive retinal ganglion cell ISI - interstimulus interval KA - kainic acid LGN - lateral geniculate nucleus L-cones - long-wavelength sensitive cones M-cones - medium-wavelength sensitive cones mGluR - metabotropic glutamate receptor MetaII - metarhodopsin II mCherry - red fluorescent protein derived from the Discosoma anemone MEA - multi-electrode array mTeal - teal fluorescent protein

NDS - normal donkey serum NGS - normal goat serum NIF - non-image forming NMDA - N-methyl-D-aspartate acid ONL - outer nuclear layer OPL - outer plexiform layer OPN - olivary pretectal nucleus PBS - phosphate buffered saline PBS-X - phosphate buffer with triton X PC - principal component PCA - principal component analysis PDE - phosphodiesterase PDEβ6 - phosphodiesterase enzyme (β6 subunit) PFA - paraformaldehyde PLC - phospholipase C PLR - pupillary light reflex PO - posterior thalamic nuclear group pSON - peri-supraoptic nucleus rd1 - retinal degeneration 1 (mouse model) rd/rd cl - rodless-coneless (mouse model) RGC - retinal ganglion cell RPE - retinal pigment epithelium SC - superior colliculus SCN - suprachiasmatic nucleus S-cones - short-wavelength sensitive cones TRP - transient receptor potential vLGN - lateral geniculate nucleus, ventral region VPM - ventral posteriomedial thalamic nucleus vSPZ - subparaventricular zone, ventral region WT - wildtype

Abstract

In the mammalian retina, a small population of retinal ganglion cells are intrinsically photosensitive due to the expression of the photopigment melanopsin. These intrinsically photosensitive retinal ganglion cells (ipRGCs) integrate their own intrinsic light response with that of rod and cone photoreceptors to drive a variety of physiological and behavioural responses to light. Recently, however, a subset of these cells have been shown to project to the dorsal Lateral Geniculate Nucleus (dLGN) of the visual thalamus, where they directly contribute to visual perception. In the case of retinal degenerations (the most common being retinitis pigmentosa which affects up to 1:2000 people worldwide), the death of the rod and cone photoreceptors results in complete visual blindness with no available treatment. At least some ipRGCs survive retinal degeneration and can continue signalling light information to the dLGN, suggesting that these cells could support some form of visual perception. However, to-date little is known about this projection during retinal degeneration. Thus, characterising its anatomy and physiology is key to determining the quality of visual information that is conveyed to the dLGN during retinal degeneration and what prevents these cells supporting behaviourally relevant vision.

A subset of ipRGCs target the dLGN and continue signalling light information even at advanced stages of retinal degeneration. However, it is unknown whether all ipRGC subtypes survive following the death of rod and cone photoreceptors, and whether they retain normal dendritic architecture following reorganisation of the remnant neural retina. We set out to answer these questions using the multi-colour labelling technique Brainbow. In doing so, we design and describe a unique methodology and toolset, based on Principal Component Analysis (PCA), to analyse 3-Dimensional (3D) multi-colour images. We then demonstrate its utility by identifying, isolating and reconstructing the 3D morphology of individual ipRGCs from a population of labelled cells in the degenerate retina and quantitatively characterise their dendritic architecture. The results indicate that all known ipRGC subtypes are resilient to the effects of outer photoreceptor degeneration.

Melanopsin responses in the dLGN have been shown to support global brightness perception in mice with advanced retinal degeneration. However, to-date, it is unknown whether these cells can encode spatial information. Using in-vitro and in-vivo electrophysiological recordings from mice in advanced stages of retinal degeneration, we demonstrate for the first time that ipRGCs in the retina, and their target neurones in the dLGN, possess discrete spatial receptive fields. These receptive fields are large and lack a centre-surround organisation. The retinotopic organisation of these cells’ projections would suggest they could support spatial vision. However the poor temporal resolution of the deafferented melanopsin response is the most significant limitation precluding melanopsin signalling from supporting behaviourally relevant vision under naturalistic viewing conditions.

Considering these temporal limitations, we finally investigated if melanopsin could contribute to visual perception at earlier stages of degeneration which is more representative of clinical conditions in humans. Here, vision can rely on both the intrinsic melanopsin-driven light response and residual cone function. Using silent substitution in combination with in-vivo electrophysiological recordings from the dLGN of mice in early-stage degeneration, we identify a number of cone-driven responses which could support normal visual function. However, we were unable to detect a significant and robust contribution of melanopsin signalling to these residual light-responses using our silent substitution stimuli in both retinally degenerate and wildtype mice at these age. However, we did find a significant contribution to the Olivary Pretectal Nucleus (OPN) of visually intact mice at equivalent ages, and to the adult dLGN. Supported by anatomical data, this suggests that there is a specific temporal delay in the maturation of ipRGCs which project to the dLGN during development of the visual system.

Declaration

I, Christopher Procyk, declare that no portion of the work referred to in the thesis has been submitted in support of an application for another degree or qualification of this or any other university or other institute of learning.

i. The author of this thesis (including any appendices and/or schedules to this thesis) owns certain copyright or related rights in it (the “Copyright”) and s/he has given The University of Manchester certain rights to use such Copyright, including for administrative purposes. ii. Copies of this thesis, either in full or in extracts and whether in hard or electronic copy, may be made only in accordance with the Copyright, Designs and Patents Act 1988 (as amended) and regulations issued under it or, where appropriate, in accordance with licensing agreements which the University has from time to time. This page must form part of any such copies made. iii. The ownership of certain Copyright, patents, designs, trademarks and other intellectual property (the “Intellectual Property”) and any reproductions of copyright works in the thesis, for example graphs and tables (“Reproductions”), which may be described in this thesis, may not be owned by the author and may be owned by third parties. Such Intellectual Property and Reproductions cannot and must not be made available for use without the prior written permission of the owner(s) of the relevant Intellectual Property and/or Reproductions. iv. Further information on the conditions under which disclosure, publication and commercialisation of this thesis, the Copyright and any Intellectual Property University IP Policy (see http://documents.manchester.ac.uk/display.aspx?DocID=24420), in any relevant Thesis restriction declarations deposited in the University Library, The University Library’s regulations (see http://www.library.manchester.ac.uk/about/regulations/) and in The University’s policy on Presentation of Theses

Contributions

Chapter 3: Dr Nina Milosavljevic assisted with intravitreal injections of viral vectors. Dr Egor Zindy (Senior Research Officer in the Faculty of Life Sciences, University of Manchester) wrote and implemented the software for colour analysis following discussions with myself, Dr Nina Milosavljevic and Professor Robert Lucas. Microscopes and bio-imaging software (IMARIS; Filament Tracer) was purchased by the University of Manchester on behalf of the Bio-imaging department in the Faculty of Biology, Medicine and Health.

Chapter 4: Dr Cyril Eleftheriou assisted with in-vitro electrophysiological experiments. Dr Riccardo Storchi and Dr Timothy Brown wrote MatLab scripts for control of visual stimuli during in-vivo electrophysiological experiments. Dr Annette Allen and Dr Riccardo Storchi provided MatLab scripts to aid in analysis of electrophysiological data.

Chapter 5: Dr Frank Martial built the projector system which allowed for the presentation of silent substitution stimuli for in-vivo electrophysiological experiments. Dr Annette Allen designed and wrote scripts to present silent substitution stimuli for electrophysiological experiments and kindly provided a subset of electrophysiological recordings from the dLGN of adult Opn1mwR mice.

Acknowledgements

First and foremost, I would like to express my sincere gratitude to my supervisor Professor Robert Lucas. His support and guidance through every stage of my PhD has been invaluable. I appreciate the time and energy he has put into this project and feel very grateful to have worked in his Lab. I would also like to thank my co-supervisor Dr Timothy Brown and my academic advisor Dr Ken Grieve for their advice and support throughout my PhD.

I must also thank all members of the Lucas laboratory, past and present, for their invaluable knowledge and support over the past four years. They make the Lucas laboratory an enjoyable environment for scientific learning and growth and I have learnt so much from them over the past four years. In particular, I would like to thank Dr Annette Allen and Dr Nina Milosavljevic who have always shown great enthusiasm and support for this project and have taught me many of the techniques which form the basis of this work. I would also like to thank Jonathon Wynne for his help and support in generating and breeding of numerous transgenic mouse lines used in these investigations.

I would like to extend a special thanks to Dr Egor Zindy for his tireless dedication and enthusiasm towards this project over the last four years. I would also like to thank the bio-imaging department for their teaching of numerous microscopy techniques and for helpful discussions during the establishment and implementation of the Brainbow project, and the BSF for their animal care and maintenance. I would also like to take this moment to thank the European Research Council for funding this project.

Finally I must thank my family: Mama, Tatuś, Mysz and Babcia for their love and support throughout these four years without which I wouldn’t be where I am today. My Final thanks go to Claire. I will always be grateful for the constant belief, love and support she provided me with throughout my PhD and I could not have achieved this without her by my side.

This thesis is dedicated to my family and Claire.

Preface

Humans are extremely visual oriented animals and we rely primarily on our visual perception in order to interact with the world. In doing so, we have evolved a complex visual system which allows us to detect, process and interpret the constant stream of light information arriving from the environment. The detection of light occurs in the retina, a specialised multi-layered tissue in the eye, where a number of cell types are employed to detect and encode numerous features of visual stimuli including brightness, colour, spatial structure, direction and motion, all on behaviourally relevant timescales. This information is then subsequently relayed to visual centres in the brain for further visual processing in order to co-ordinate an appropriate behavioural response.

However, as with other sensory systems, the retina is susceptible to trauma, disease and age-related degeneration. Inherited retinal degenerations, such as retinitis pigmentosa, are the most common of these and occur at an incidence of between 1 in 4,000 (Strettoi, 2015, Delyfer et al., 2004, Hartong et al., 2006). They are often a result of mutations in the light-sensitive cells of the retina that cause compromised photoreceptor function and subsequent cell death (Hartong et al., 2006). As a consequence, the retina undergoes complex stages of cell death and re-organisation in response to the loss of visual input (Jones and Marc, 2005, Strettoi, 2015). This inevitably results in progressive visual blindness with currently no available treatments or cure.

Over the last fifteen years, a newly discovered light-sensitive photopigment termed melanopsin has been found to express in a small population of cells in the inner retina, rendering them intrinsically sensitive to light (Berson et al., 2002, Hattar et al., 2002, Lucas et al., 2003a). These cells have been shown to modulate a variety of non-image forming (NIF) physiological and behavioural responses to light (Milosavljevic et al., 2016, Lucas et al., 2001a, Berson et al., 2002, Hattar et al., 2006). Recently, a newly identified subset of these cells have been shown to directly project to image-forming (IF) brain structures (Ecker et al., 2010, Brown et al., 2010) where they have been shown to contribute to visual perception (Storchi et al., 2015, Schmidt et al., 2014, Davis et al., 2015). At least some of these cells survive retinal degeneration (Vugler et al., 2008, Lin and Peng, 2013) and continue signalling light-information to the brain even at advanced stages of the disease (Brown et al., 2010). This would suggest that these cells could support some aspect of visual perception in those patients afflicted with retinal degeneration.

However, visual perception cannot rely upon melanopsin signalling alone. In this thesis we set out to investigate what the quality of visual information ipRGCs can convey to the image-forming visual system during retinal degeneration, and look to determine what precludes melanopsin from supporting behaviourally relevant vision. We look specifically to identify the survival and structural integrity of ipRGC subtypes which project to the dLGN during advanced stages of degeneration, what

the spatio-temporal resolution of melanopsin signalling is originating from these cells in both the retina and dLGN at advanced stages of the disease using in-vivo and in-vitro electrophysiological recordings, and whether the additional activation of melanopsin in combination with cone excitation during early stages of progressive degeneration can contribute significant visual information to the dLGN. In order to address these questions, an overview of the rodent visual system in health and disease are described below, with emphasis on cell function, neurochemistry and network properties, in order to provide a comprehensive review of the literature to date.

Chapter 1 General Introduction

Chapter 1: General Introduction

Chapter 1 General Introduction

1.1 The Vertebrate Retina

Figure 1.1 The vertebrate retina. Schematic diagram of a cross section of the vertebrate retina illustrating the major cells types involved in visual processing. The outer nuclear layer consists of the outer segments of rod (black) and cone (green and purple) photoreceptors and their cell bodies. They form synapses with bipolar cells (BC; red) and horizontal cells (HC; violet) in the outer plexiform layer (OPL). The inner nuclear layer (INL) contains the cell bodies of bipolar cells and amacrine cells (AC; yellow). These cells make synapse with Retinal Ganglion Cells (RGCs; dark blue) in the inner plexiform layer (IPL), which is subdivided into two sublaminae. The OFF-sublamina receives input from OFF BCs and synapses with OFF RGCs. In the ON-sublamina, ON BCs synapse with ON RGCs. The cell bodies of RGCs and intrinsically photosensitive retinal ganglion cells (ipRGCs; light blue) are found in the Ganglion cell layer (GCL) where their axons form the optic nerve to the brain. Müller cells (MC; dark green) are specialised glia which span the entire length of the retina. The retinal pigment epithelium (RPE), a thin layer of pigmented tissue, is found at the back of the ONL adjacent to the photoreceptors. Diagram adapted from (Soto & Kerschensteiner., 2015).

Chapter 1 General Introduction

The retina is a specialised multi-layered tissue at the back of the vertebrate eye whose primary role is the conversion of photons of light into electrical signals to provide an internal representation of the external environment. A number of neuronal subtypes are employed in the retinal circuity in order to transmit and process the constant stream of complex visual information arriving from the environment (Figure 1.1). These can be described by two pathways: the vertical pathway and the horizontal pathway. The vertical pathway involves the conversion of light to electrical signals in the rod and cone photoreceptors. These cells synapse with bipolar cells that subsequently relay this information to retinal ganglion cells. The retinal ganglion cells are the only output cells of the retina whose major target for image forming vision is the dorsal lateral geniculate nucleus (dLGN) of the visual thalamus. The horizontal pathway of the retina comprises horizontal cells and amacrine cells which provide modulatory input to the vertical pathway for early visual processing.

Rod and cone photoreceptors Rods and cones are the light-sensitive cells of the mammalian retina. Their primary role is the transduction of photons of light into graded electrical signals. These two photoreceptors exhibit molecular, morphological, and electrophysiological differences which allow the retina to respond to a vast array of visual stimuli including colour (wavelength) and brightness (irradiance). In the mouse retina, there are approximately four million photoreceptors of which 97% are rods, with the remaining 3% comprising of cone photoreceptors (Jeon et al., 1998, Carterdawson and Lavail, 1979). Rod and cone photoreceptors are distributed throughout the outer retina where they demonstrate spatial variation with retinal eccentricity which allows the entire retina to continue responding to light from across the visual scene (Ahnelt and Kolb, 2000, Ortin-Martinez et al., 2014, Wang et al., 2011).

Both rods and cones comprise of an outer-segment, an inner segment and a cell body with an axon terminal (Carterdawson and Lavail, 1979). It is in the outer-segments, situated at the back of the retina, where photons of light are transduced into an electrical signal due to the high density expression of light-sensitive proteins termed opsins. These opsins are expressed in the plasma membrane of specially modified discs. In rods, these discs are found within the cell and are completely detached from the plasma membrane, whilst in cones these discs are continuous with the plasma membrane and form invaginations (Carterdawson and Lavail, 1979).

Rods demonstrate high sensitivity to light due to the expression of the photopigment rhodopsin. They therefore dominate vision under low light intensities - scotopic vision. This is why nocturnal species, such as the mouse, possess a significantly higher proportion of rod photoreceptors compared to cones (Jeon et al., 1998). As light levels increase, the response of rod photoreceptors can begin to saturate. This is compensated for by the cone photoreceptors which demonstrate much lower sensitivity to light but can continue signalling at higher light intensities - photopic vision. In

Chapter 1 General Introduction

addition, mice express two types of cone photoreceptors - short wavelength and medium wavelength sensitive cones (Applebury et al., 2000). It is the relative activation of these two classes of cone photoreceptor which allow for the extraction of colour from the visual scene. This variety of opsins expressed in different photoreceptor subtypes allows the retina to continue signalling visual information over a wide range of background light intensities and across broad spectrum illumination.

1.1.1.1 Opsins Opsins belong to the super family of G Protein coupled receptors (GPCRs) (Palczewski, 2006, Fredriksson et al., 2003). They form a light-sensitive photopigment when covalently bound to a retinaldehyde chromophore, and as such, can confer light sensitivity to cells in which they are expressed. Following absorption of light by the retinaldehyde chromophore, conformational changes occur in the structure of the opsin protein which subsequently results in changes in intracellular signalling cascades (Palczewski et al., 2000). Rod and cone opsins express different amino acids sequences in close proximity to the region which binds the retinaldehyde chromophore. This results in a unique molecular relationship between the opsin and the chromophore which defines their sensitivity to different wavelengths of light (Palczewski et al., 2000). This relationship is termed the absorption spectrum and describes the wavelength of light to which the photopigment maximally responds (λmax). All mammalian opsins utilise a Vitamin A based chromophore and therefore show the same probability distribution of photon capture which is simply wavelength-shifted across the visible spectrum (Figure 1.2).

Figure 1.2 Spectral sensitivity of visual photopigments in the mouse eye Wildtype mice express four distinct opsins in the retina. These opsins covalently bind a retinaldehyde chromophore to form a functional light- sensitive photopigment. Small changes in the amino acids surrounding the binding pocket of the retinaldehyde chromophore result in these photopigments demonstrating sensitivity to different wavelengths of light in the visible spectrum: S-cone opsin (λmax = 390nm; violet), Melanopsin (λmax = 480nm; blue), Rod opsin (λmax = 498nm; black) and M-cone opsins (λmax = 508nm; green). λmax values are taken from (Fu & Yau, 2007) and (Lucas et al., 2001a) and applied to the visual pigment template described by Govardovskii et al. (Govardovskii et al., 2000)

Chapter 1 General Introduction

Rod photoreceptors exclusively express the photopigment rhodopsin which has a λmax of 498nm (Okawa and Sampath, 2007, Bridges, 1959). They are extremely sensitive, such that a single photon of light is enough to result in a 1mV change in membrane potential (Hecht et al., 1942). This permits vision under low light intensities when the irradiance (the total number of photons falling on the retina, per unit area, per second) is low. Mice are dichromats as they only express two types of cone opsin: S-cone opsin (λmax = 360nm; violet light) and M-cone opsin (λmax = 505nm; green light) (Calderone and Jacobs, 1995, Fu and Yau, 2007, Sun et al., 1997). Humans and primates additionally express a third cone opsin termed L-cones (λmax = 552nm; red light) which is sensitive to long wavelength red light (Merbs and Nathans, 1992). The expression of different cone opsins allow mammals to extract colour vision due to the relative activation of cone opsins with different spectral sensitivities. Interestingly, mice co-express both M and S-cone opsin within the same photoreceptor (Szel et al., 1992, Rohlich et al., 1994) which would limit their ability to discriminate between different colours. However, there is a dorso-ventral gradient in the expression level of the different opsins, with M-cone opsin expression decreasing from dorsal to ventral retina (Applebury et al., 2000, Haverkamp et al., 2005, Nikonov et al., 2006). Approximately 5% of cone photoreceptors in the peripheral retina exclusively express the S-Cone opsin (Hoon et al., 2014). A fourth photopigment termed melanopsin is expressed in a small proportion of retinal ganglion cells in the inner retina showing sensitivity to blue light (λmax = 480nm) (Lucas et al., 2001a, Berson et al., 2002).

1.1.1.2 Phototransduction Activation of both rod and cone opsins follows a well-documented intracellular signalling cascade (Yau and Hardie, 2009, Pugh and Lamb, 1993) and is illustrated in Figure 1.3. Under dark conditions, the outer segments of rod and cone photoreceptors continuously create cGMP. This maintains the open state of cGMP-gated cation channels which permit the steady flow of sodium and calcium ions into the cell. Simultaneously, there is a continuous outward flow of potassium ions via non-gated potassium channels. Equilibrium is maintained due to the active transport of these ions via sodium/potassium pumps which span the plasma membrane (Cervetto et al., 1989, Schnetkamp, 2004). This ionic balance is termed “the dark current” and maintains photoreceptors at a slightly depolarised resting membrane potential of -40mV (Hagins et al., 1970).

Absorption of a photon of light by rod and cone opsins results in the isomerisation of the bound retinaldehyde chromophore from its 11-cis-retinal configuration to all-trans-retinal. Structural re- arrangements occur in the trans-membrane domains to accommodate the conformational change of the photopigment to the excited state called Metarhodopsin II (Farrens et al., 1996, Farahbakhsh et al., 1993, Palczewski et al., 2000). To stabilise the newly-generated Meta II intermediate, a rotation of Helix VI exposes a site for binding the g-protein transducin (an evolutionary relative of Gαi) on Helix VIII. The binding of transducin facilitates its activation by exchanging GDP for GTP on the alpha subunit and liberating the G-protein into two active subunits: Gαt and Gβɣ (Kuhn et al., 1981, Okada

Chapter 1 General Introduction

and Palczewski, 2001). Activated Gαt stimulates cGMP phosphodiesterases which hydrolyse cGMP to 5’GMP. The result of this is closure of cGMP-gated cation channels (Fesenko et al., 1985), a graded hyperpolarisation of the cell (Haynes and Yau, 1985) and consequently a reduction in the tonic release of glutamate from the photoreceptor.

Figure 1.3 Phototransduction in rod and cone photoreceptors (1) In the dark state, cGAP (turquoise) constitutively catalyses the conversion of GTP (dark blue) into cGMP (red). (2) cGMP maintains the open state of cGMP-gated cation channels permitting the influx of sodium ions and establishing the “dark current”. (3) The absorption of a photon by the photopigment (black) results in conformational changes and the subsequent activation of the G-protein Transducin by exchanging GDP for GTP. (4) Activated Transducin is liberated into its α and βγ forms. (5) Activated Gαt binds to the cGMP phosphodiesterase (green) which facilitates the conversion of cGMP to 5’cGMP. (6) This results in the reduction of cGMP and the closure of the cGMP-gated cation channels (light blue) resulting in cellular hyperpolarisation. Image based on data from (Yau & Hardie., 2009).

1.1.1.3 Deactivation of the phototransduction cascade It is vitally important that this signalling cascade is terminated rapidly in order to respond to the continuous flow of visual information reaching the retina. There are a number of points in the phototransduction cascade which maintain the rapid and efficient deactivation of the light-response. Firstly, the Metarhodopsin II state of the opsin complex is thermally unstable and decays through a series of distinct photo-intermediates which releases the chromophore yielding free all-trans-retinal and the opsin protein (Palczewski, 2006). This is further accelerated by the protein rhodopsin kinase which phosphorylates Metarhodopsin II (Chen et al., 1999) recruiting the binding of arrestin (Xu et al., 1997) and preventing further activation of Transducin (Kuhn and Wilden, 1987). Transducin is

Chapter 1 General Introduction

also deactivated by its own intrinsic GTPase activity whereby the bound GTP is replaced by GDP (Arshavsky and Bownds, 1992, Pages et al., 1992, Arshavsky and Wensel, 2013) and is further accelerated by regulators of G-protein signalling (Chen et al., 2000). This results in its dissociation from the active Gαt subunit, which in turn reduces phosphodiesterase activity. Consequently, the intracellular levels of cGMP increase allowing for cGMP-gated cation channels to re-open so the photoreceptor can depolarise to its resting membrane potential of approximately -40mV.

1.1.1.4 Chromophore regeneration Rod and cone opsins form a stable photopigment when binding 11-cis-retinal. The absorption of a photon of light and the isomerisation of 11-cis to the all-trans-retinal isoform results in instability of the photopigment and the subsequent disassociation of the chromophore from the opsin protein. Therefore, rod and cone photoreceptors need a continuous fresh supply of the 11-cis-retinaldehyde isoform in order to retain their ability to respond to light. This is achieved by a complex metabolic pathway which is shared between the photoreceptors and the retinal pigment epithelium (RPE) - a thin layer of pigmented tissue found adjacent to the photoreceptors at the back of the retina (Kiser et al., 2012, McBee et al., 2001, Wang and Kefalov, 2011). In brief, all-trans-retinal is converted to all-trans-retinol within the photoreceptors. Once in this isoform, it is transported to the RPE where the isomerase Retinal Pigment Epithelium 65 (RPE65) converts the chromophore back to its 11-cis- retinal configuration through a series of enzymatic steps known as the visual cycle. It is then transported back to the photoreceptor outer segments where it can bind to an opsin protein to form a new functional photopigment (Palczewski et al., 1999, Saari, 2000, McBee et al., 2001). More recently, Müller cells have also been shown to provide a supportive role in chromophore regeneration of cone photoreceptors under daylight conditions (Kaylor et al., 2013, Wang and Kefalov, 2009, Mata et al., 2002, Tang et al., 2013).

Bipolar cells Bipolar cells are interneurons which are strategically positioned to receive direct glutamatergic input from the rod and cone photoreceptors in the outer plexiform layer, and to transmit this light information to the retinal ganglion cells in the inner retina. However, their function is not to simply transfer photic information, but to establish a parallel processing pathway through which early visual processing can occur. There are at least 11 known types of cone bipolar cell (Ghosh et al., 2004), which are categorised as either ON or OFF (Waessle et al., 2009), whilst rods synapse with one dedicated type of ON rod bipolar cell in the mouse retina (Masland, 2001).

Chapter 1 General Introduction

1.1.2.1 Cone signalling through bipolar cells Whether a bipolar cell is described as ON or OFF is determined by the glutamate receptor expressed on its post-synaptic terminal and the consequential electrophysiological response following photoreceptor activation (Thoreson and Witkovsky, 1999). ON bipolar cells express a specific class of metabotropic glutamate receptor (mGluR6) which conjugates to the Gαo G-protein pathway (Vardi et al., 2000, Slaughter and Awatramani, 2002, Dhingra et al., 2002). Accordingly, light-stimulation results in reduced glutamate release from the photoreceptors, dis-inhibition of the ON bipolar cell and a graded cellular depolarisation (Sharma et al., 2005). This is known as a ‘sign inverting’ response as when the photo-receptor is hyperpolarised in response to light, the ON bipolar cell is depolarised.

By contrast, OFF bipolar cells express ionotropic glutamate receptors (iGluRs) on their dendrites, namely α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) and Kainate (KA) receptors (Thoreson and Witkovsky, 1999, Haverkamp et al., 2001, Borghuis et al., 2014). A reduction in light- evoked glutamate release from photoreceptors results in reduced excitability of the OFF bipolar cell and its graded hyperpolarisation. This is therefore known as a ‘sign conserving’ response as both the photoreceptor and OFF bipolar cell are hyperpolarised in response to increases in light. AMPA and Kainate receptors can be differentially expressed on different OFF bipolar cell subtypes giving rise to unique temporal signalling kinetics and adaptation properties in different bipolar cell subclasses (DeVries, 2000, Lindstrom et al., 2014). Individual ON and OFF cone bipolar cells receive input from multiple rod and cone photoreceptors. These two parallel processing pathways allows the retina to discriminate between increments and decrements in light from across the visual field.

1.1.2.2 Rod signalling through bipolar cells It is believed that the rod photoreceptive system developed after the cone system, and consequently rods signalling evolved to utilise the existing cone circuitry to convey light information in the retina (Okano et al., 1992). Under scotopic conditions, rods relay their information to a single dedicated rod bipolar cell class that synapses with AII amacrine cells (Sharpe and Stockman, 1999). These amacrine cells can make a direct chemical synapses with OFF cone bipolar cells (Menger et al., 1998, Bloomfield and Dacheux, 2001) or electrical synapses with ON cone bipolar cells (Smith et al., 1986, Sterling et al., 1988, Sharpe and Stockman, 1999) before reaching RGCs. Although rhodopsin has intrinsically high sensitivity to light (Okawa and Sampath, 2007, Bridges, 1959), this indirect pathway maximises vision under dim lighting conditions as rod responses are summated from across a large area of the retina (Masland, 2001, Smith and Vardi, 1995). This does however come at the expense of lower visual acuity. Under photopic conditions this system is bypassed and rod photoreceptors transmit their light-information via electrical synapses with Horizontal cells. These directly influence cone bipolar cell activity before being relayed to the retinal ganglion cell population (Volgyi et al., 2004, Bloomfield and Dacheux, 2001).

Chapter 1 General Introduction

Retinal ganglion cells Bipolar cells synapse with retinal ganglion cells in the inner plexiform layer (IPL) of the retina. The IPL is divided into the ON sublamina and the OFF sublamina (Figure 1.1), thus maintaining spatial as well as functional differentiation between the ON and OFF pathways. ON bipolar cells synapse with ON RGCs in the ON plexus of the IPL, whilst OFF bipolar cells synapse with OFF RGCs in the OFF plexus of the IPL (Nelson et al., 1978).

Retinal ganglion cells express ionotropic glutamate receptors on their post synaptic dendrites which when activated, result in cellular depolarisation (Diamond and Copenhagen, 1995, Cohen and Miller, 1994, Lin et al., 2002). In this sense, RGCs are unique as they are the only cells in the retina that respond to light by changes in action potential generation, rather than graded changes in membrane potential (Lukasiewicz, 2005). This supports the primary function of RGCs which is to relay light information to retinorecipient nuclei in the brain. Their long unmyelinated axons are found in the inner most regions of the retina and converge at the optic disc to form the optic nerve (Purves et al., 2001). The optic nerve projects to a number of brain structures involved in image forming (IF) vision such as the dorsal Lateral Geniculate Nucleus (dLGN) (Reese, 1988, Piscopo et al., 2013, Grubb and Thompson, 2003), or to non-image forming (NIF) structures including the Suprachiasmatic nucleus (SCN) (Warren et al., 2003, Johnson et al., 1988) and the olivary pretectal nucleus (OPN) (Baver et al., 2008, Hattar et al., 2006).

However, secondary to this, an astonishing level of visual processing already occurs in the retina due to the convergence of light information onto the 22 different subtypes of RGCs, and their integration with the retinal circuitry (Voelgyi et al., 2009, Farrow and Masland, 2011, Coombs et al., 2006). The differences in their morphology and wiring are mirrored by the diverse array of computational functions they perform including orientation selectivity (Chen et al., 2014), chromatic sensitivity (Ekesten and Gouras, 2005) and direction and motion selectivity (Vaney et al., 2012, Demb, 2007, Weng et al., 2005) to name a few. One of the most important computations that arise in the early visual system is the generation of spatial receptive fields. A spatial receptive field is defined as the volume of visual space which, when a light stimulus is present, changes a retinal ganglion cell’s firing pattern (Purves et al., 2001). Receptive fields are formed by the complex and unique integration of light information from rod and cone photoreceptors with Horizontal cells, which is subsequently relayed via bipolar cells to an individual RGC. The spatial receptive field structure of an individual retinal ganglion cell comprises of two components: the centre and the surround (Kuffler, 1953, Baylor et al., 1971). RGCs can therefore be defined as ON-centre, whereby light falling on the centre increases the cell’s firing pattern whilst light falling on the periphery reduces signalling (Figure 1.4). Conversely, OFF-centre RGCs inhibit their signalling when light falls on the centre of the receptive field, whilst light falling on the periphery excites the cell. In both cases, if light covers the entirety of the receptive field, the output of the cell can cancel out.

Chapter 1 General Introduction

Figure 1.4 Spatial receptive field structure of an individual retinal ganglion cell (A) (Top) Receptive field structure of a single ON RGC demonstrating the centre (white circle) and the surround (grey circle) when no stimulus is present on the receptive field. (Below) Activity bar of the retinal ganglion shows low frequency spontaneous depolarisations. (B) Light stimulus (yellow circle) falling partially on the centre and the surround results in a small increase in firing activity of the RGC as both the ON centre and OFF surround are partially stimulated. (C) Light falling exclusively on the centre component of the receptive field drives a large increase in firing activity. (D) Light falling on both the centre and the surround results in only a modest increase in firing activity above baseline as both responses can cancel each other out. Diagram adapted from (Purves et al., 2001)

The size of the centre component of a spatial receptive field is the limiting factor of spatial acuity. It is defined by the spatial extent of photon capture by rod and cone photoreceptors and the convergence of this signal onto an individual ganglion cell. Primates have high acuity vision due to their fovea - a region of high cone concentration where cone photoreceptors synapse with an individual bipolar cell and onto a single RGC (Masland, 2001, Perry and Cowey, 1985). Rodents, however, do not possess a fovea but evenly express their photoreceptors in the outer retina in order to efficiently sample the entire visual scene. It has been proposed that this is advantageous for mammals such as mice as it allows them to detect movement across the visual field in order to avoid predation (Jeffrey et al., 2011, Huberman and Niell, 2011).

Chapter 1 General Introduction

Laterally positioned Cells Whilst the primary role of the vertical pathway in the retina is the conversion and transfer of visual information to the visual processing structures in the brain, the horizontal pathway can be thought of as providing a modulatory role in shaping this information. Two major cell types, the horizontal cells and the amacrine cells perform these functions by modulating the signal transfer at the synapses of photoreceptors to bipolar cells, and bipolar cells to retinal ganglion cells, respectively (Figure 1.1).

1.1.4.1 Horizontal cells Horizontal cells represent the first tier of lateral processing in the retina. Rodent retinas are believed to employ one type of horizontal cell whose long processes synapse with photoreceptor terminals and the dendrites of bipolar cells, in the outer plexiform layer (OPL) (Reese et al., 2005, Raven et al., 2005, Peichl and Gonzalezsoriano, 1994). These cells are vital in the establishment of the surround antagonism of spatial receptive fields (McMahon et al., 2004, Mangel, 1991, Kuffler, 1953, Kaneko, 1970, Thoreson and Mangel, 2012) by predominantly expressing the excitatory ionotropic receptor AMPA on their dendrites (Haverkamp et al., 2001, Yang, 2004). Stimulating the centre of a retinal ganglion cells receptive field results in hyperpolarisation of the photoreceptors, the subsequent graded depolarisation of ON bipolar cells, and an increase in action potential generation by ON RGCs. However, the light-induced reduction in glutamate release from photoreceptive cells also hyperpolarises horizontal cells (Figure 1.5). The consequence of this is that the signal transfer between photoreceptors and bipolar cells more distal to the centre (the surround) are inhibited whilst the centre is strongly excited (Dacey et al., 2000, Baylor et al., 1971, Thoreson and Mangel, 2012, Verweij et al., 1996). By contrast, OFF bipolar cells respond in the opposite manner and possess an inhibitory centre with an excitatory surround. The creation of this surround organisation is retained through the bipolar cell network and retinal ganglion cells in both the ON and OFF pathways. This mechanism enhances the relative contrast between the edges of objects which appear adjacent to each other in the visual field and forms the basis of high acuity spatial vision.

Chapter 1 General Introduction

Figure 1.5 Establishment of the centre-surround organisation Schematic cross sectional diagram of the retina illustrating the synaptic wiring of rod and cone photoreceptors in the centre region (black & green) and the antagonistic surround region (greyed out) of a ganglion cells (RGC; dark blue) receptive field. Light-induced hyperpolarisation of photoreceptors in the centre region result in a graded depolarisation of the ON Bipolar cell (ON BC) and the consequential depolarisation of the ON retinal ganglion cell. The light-induced hyperpolarisation of photoreceptors also hyperpolarises horizontal cells (HC; violet), whose contacts with surrounding photoreceptors (greyed out) will result in a reduction in transmission at the photoreceptor to bipolar cell synapse in the surround region (Bipolar cells are not shown in the surround region for clarity). Diagram modified from (Soto & Kerschensteiner., 2015) and based on data from (Dacey et al., 2000) and (Thoresen & Mangel., 2012).

1.1.4.2 Amacrine cells Amacrine cells represent the second tier of horizontal modulation in the inner plexiform layer. They contact both bipolar cell terminals and the dendrites of RGCs in order to modulate bipolar cell to RGC transmission using predominantly inhibitory neurotransmitters (Crooks and Kolb, 1992). There are more than 50 Amacrine cell types which can be broadly categorised into three groupings based on their morphology, stratification and neurochemistry: narrow-field, medium-field and wide-field (Werblin, 2011, MacNeil and Masland, 1998, Kolb et al., 1981).

Narrow-field amacrine cells provide local glycinergic inhibition in the retina (Menger et al., 1998). Their processes extend up to 100µm laterally in the IPL and modulate both ON and OFF visual pathways (Chen et al., 2010). Medium-field amacrine cells provide a local GABAergic inhibition due to their processes extending up to 300µm laterally in the IPL (Chen et al., 2010). Similar to narrow-

Chapter 1 General Introduction

field amacrine cells, the medium-field also modulate both ON and OFF pathways but have a more controlled stratification pattern in the IPL (MacNeil and Masland, 1998). Wide-field amacrine cells are poly-axonal and provide a wide GABAergic inhibition with processes that extend up to 1mm laterally in the IPL. Individual wide-field amacrine cells exclusively stratify in a single stratum of the IPL, but as a population have been shown to innervate all known strata (MacNeil and Masland, 1998, Volgyi et al., 2001). The spatial extent of their processes allows for a more global signal to be transmitted in the IPL, which is further enhanced by their electrical coupling to other amacrine cells and retinal ganglion cells (Volgyi et al., 2001).

With such a large variety of amacrine cell-types in the retina, it is perhaps unsurprising that they have been shown to fulfil numerous and vital functions in regulating and shaping the visual information that will be transmitted from RGCs to visual centres in the brain. These functions range from motion detection (Baccus et al., 2008), direction selectivity (Fried et al., 2002, Zhou and Lee, 2008, Yang and Masland, 1994), looming signals (Muench et al., 2009) to contrast gain control (Zaghloul et al., 2007).

1.2 Intrinsically Photosensitive Retinal Ganglion Cells Discovery The classical rod and cone photoreceptive pathway is well-established and was thought to account for all physiological light responses in mammals. However, this was challenged when it was discovered that mice that suffer severe retinal degeneration of the rod and cone photoreceptors (rd/rd and rd/rd cl) retain the ability to entrain their circadian rhythm to the light-dark cycle (Foster et al., 1991, Foster et al., 1993), inhibit melatonin secretion in response to light onset (Lucas et al., 1999) and retain a pupillary light reflex (Lucas et al., 2001a). Similar observations were made in human patients with outer retinal degeneration (Czeisler et al., 1995, Klerman et al., 2002). This suggested that rods and cones were not solely responsible for mediating all physiological responses to light. The severe outer degeneration suffered in these models indicated that a third ocular photopigment may be expressed in the inner retina and was supported by evidence which showed that enucleating the mouse abolishes all physiological and behavioural responses to light (Freedman et al., 1999).

Melanopsin was initially discovered in the photo-sensitive dermal melanophores of the Xenopus laevis (Provencio et al., 1998). Resolving the sequence of the Melanopsin gene (Opn4) facilitated the identification of melanopsin expression in a subset of RGCs in both mice and primates by in situ hybridisation (Provencio et al., 2000). It had long been established that the SCN was the central circadian clock in mammals functioning to regulate sleep propensity, appetite, locomotor activity, hormone secretions (among other physiological behaviours) in response to changes in the light/dark

Chapter 1 General Introduction

cycle - see (Reppert and Weaver, 2002) for review. Consequently, Berson and colleagues injected a retrograde tracer into the SCN and identified a small subset of RGCs that formed the Retinohypothalamic tract (RHT) (Berson et al., 2002). Electrophysiological recordings from these labelled RGCs demonstrated a depolarising response to light even when synaptic input from rods and cones was pharmacologically blocked (Berson et al., 2002). It was later identified that these light-responsive RGCs and melanopsin-expressing RGCs were one and the same (Hattar et al., 2002).

Confirming melanopsin as the putative photopigment arose from work on both cell lines and transgenic mouse models. The heterologous expression of mammalian melanopsin into cultured cells in-vitro induces depolarising currents in cells that were previously not light-sensitive (Melyan et al., 2005, Qiu et al., 2005). These light responses demonstrate a similar response profile as the intrinsic melanopsin response in ipRGCs when synaptic input from rod and cone photoreceptors is pharmacologically inhibited (Berson et al., 2002). Anatomical tracing has also provided concrete evidence of melanopsin expression in the mouse retina. The Opn4tau/LacZ mouse expresses the β- galactosidase enzyme fused to a signal sequence for the axonal protein tau and therefore acts as a reporter for melanopsin expressing RGCs (Hattar et al., 2002). Homozygous Opn4tau/LacZ mice lose their intrinsic photosensitivity due to loss of melanopsin expression and exhibit attenuated non- image forming responses to light (Lucas et al., 2003b). Crossing these mice with aged rd/rd cl mice, which have a complete loss of rod and cone photoreceptors, abolishes all image forming and non- image forming responses to light (Lucas et al., 2003b). Selective ablation of ipRGCs also results in attenuated non-image forming behavioural responses to light (Panda et al., 2003, Goez et al., 2008, Hatori et al., 2008).

Classification Since the initial discovery of melanopsin in the mouse retina, five distinct subtypes of ipRGC have been identified which account for approximately 5% of the total RGC population (Berson et al., 2010, Ecker et al., 2010, Estevez et al., 2012). These cells can be characterised by differences in their somato-dendritic profiles (Figure 1.6), intrinsic photosensitivity, electrophysiological responses to light and central projection targets (Berson et al., 2010; Ecker et al., 2010; Estevez et al., 2012; Schmidt et al., 2011a) and are described below.

Chapter 1 General Introduction

Figure 1.6 Somato-dendritic profile of the five ipRGC subtypes Dendrites of M1 cells stratify in the OFF sublamina of the inner plexiform layer (IPL) and make contacts with both ON bipolar cells (ON BC) and OFF bipolar cells (OFF BC). M2, M4 and M5 cells stratify in the ON sublamina and receive inputs exclusively from ON Bipolar cells. M3 cells bistratify in both the ON and OFF sublamina but receive input only from ON Bipolar cells. M4 cells have the largest soma size followed by M2 cells. M1, M3 and M5 cells have a comparatively smaller soma size. Imaged based on data from (Schmidt et al., 2011).

M1-type ipRGCs

Opn4tau/LacZ knock-in mice were initially used to identify and investigate the distribution of ipRGCs in the mouse retina and their projection targets in the brain (Hattar et al., 2006). However, it is now known that the Opn4tau/LacZ mouse only expresses β-galactosidase at detectable levels in approximately 50% of cells that also express melanopsin (Baver et al., 2008, Hatori et al., 2008). These retinal ganglion cells which co-express β-galactosidase and melanopsin are termed M1 ipRGCs and are the best documented subtype. M1 cells are characterised morphologically by a relatively small soma size of approximately 15µm and a sparsely branching dendritic field ranging from 275- 365µm (Berson et al., 2010, Schmidt et al., 2011a). These dendritic fields overlap to form a ‘photoreceptive net’ which spans the entire retina. Their dendrites arborize in the outer (OFF) plexus of the IPL, yet receive synaptic input from ON bipolar cells and only weak synaptic from OFF bipolar cells (Wong et al., 2007). M1 ipRGCs express melanopsin at a density of ̴3 molecules per µm-2 which is believed to be much higher than other subtypes based on immunohistochemical analysis (Do et al., 2009, Ecker et al., 2010). Accordingly, M1 ipRGCs exhibit the largest intrinsic photo-sensitivity producing large and sustained depolarising currents in response to light, even when isolated from rod and cone photoreceptor input (Schmidt and Kofuji, 2009). Due to the large photo-currents generated by M1 ipRGCs, they are prone to depolarising block which results in transient low frequency action potential fire in response to a sudden and large increase in light (Wong et al., 2007). Anatomical tracing has revealed that M1 ipRGCs innervate numerous central brain targets associated with the circadian system including the suprachiasmatic nucleus (SCN), the shell of the Olivary Pretectal Nucleus (OPN shell), the intergeniculate leaflet (IGL), the lateral habenula (LHb),

Chapter 1 General Introduction

the medial amygdala, the ventral subparaventricular zone (vSPZ), the peri-supraoptic nucleus and the ventral lateral geniculate nucleus (vLGN) (Baver et al., 2008, Hattar et al., 2006, Lucas et al., 2003a).

M2-type ipRGCs

The remaining 50% of ipRGCs that exhibit positive anti-melanopsin antibody labelling but are undetectable under β-galactosidase staining are termed M2 ipRGCs. These are distinguishable from M1 cells by both anatomical and electrophysiological differences. M2 ipRGCs have a slightly larger soma size of approximately 16µm and a larger more orderly dendritic field spanning approximately 250-300µm (Ecker et al., 2010). Their dendrites arborize and form synapses with ON bipolar cells in the outer (ON) plexus of the IPL (Schmidt et al., 2011b). They demonstrate weaker histological labelling for the melanopsin protein, suggesting that melanopsin is expressed at lower levels in M2- type than M1-type ipRGCs (Baver et al., 2008, Hatori et al., 2008). This is supported by electrophysiological recordings which demonstrate that M2-type ipRGCs demonstrate 10-fold lower intrinsic light-response when exposed to the equivalent light intensity as M1-type cells (Schmidt and Kofuji, 2009). The exact density of melanopsin expression in these cells is currently unknown. M2 ipRGCs receive extensive synaptic input from ON bipolar cells and amacrine cells which suggests that they are strongly integrated with, and hence influenced by, photic input arising from the classical photoreceptive pathway (Jusuf et al., 2007, Schmidt and Kofuji, 2010). Their projections weakly innervate the SCN but strongly innervate the core region of the OPN (Baver et al., 2008).

M3, M4 and M5-type ipRGCs

The emergence of cre-lox mouse lines such as Opn4Cre knock-in mice have provided much higher sensitivity when identifying ipRGCs than was available from the Opn4tau/LacZ mouse or melanopsin anti-body labelling alone (Ecker et al., 2010). This has aided the identification of a further three ipRGC sub-types; M3, M4 and M5 (Ecker et al., 2010, Estevez et al., 2012, Schmidt and Kofuji, 2011). These cells were previously un-identifiable using anti-body labelling due to their presumed weak expression of the melanopsin photopigment (Berson et al., 2010, Hattar et al., 2006, Ecker et al., 2010). Nonetheless, all three exhibit weak intrinsic light-responses when isolated from rod and cone input (Estevez et al., 2012, Ecker et al., 2010, Schmidt and Kofuji, 2011).

M3 cells are sparsely expressed in the murine retina and account for between 10% (Berson et al., 2010) and 26% (Schmidt et al., 2008) of the total ipRGC population. They are predominantly identified by their dendritic bi-stratification into both the inner and outer plexus of the IPL. These cells weakly express melanopsin and their dendritic fields do not appear to overlap, so do not tile the entire retina as observed with the M1 and M2-types (Berson et al., 2010, Provencio et al., 2002).

Chapter 1 General Introduction

For this reason, their projection targets and physiological function remains elusive (Schmidt and Kofuji, 2011). One suggested function of M3-type is that their bi-stratification allows for the transfer of light-information between M1 and M2-type ipRGCs (Berson et al., 2010).

M4 cells are distinguishable from other ipRGCs due to their large soma size of approximately 21µm and large complex dendritic field ranging from 300-440µm which stratify exclusively in the inner (ON) sublamina of the IPL (Ecker et al., 2010, Estevez et al., 2012). The expression of melanopsin cannot be confirmed by typical immunofluorescence with the anti-melanopsin antibody UF006, due to its presumed low expression, and is in agreement with its extremely weak intrinsic response to light (Ecker et al., 2010). However, expression of the melanopsin protein has since been confirmed using Tyramide-amplified immunofluorescence (Estevez et al., 2012). M5 cells in contrast have a smaller soma size and a smaller, symmetrical and bushier dendritic field which arborizes in the (ON) plexus of the IPL (Ecker et al., 2010). Much less is known about this subtype due its recent identification in the mouse retina. Both M4 and M5-subtypes are presumed to innervate image- forming structures including the dLGN and the superior colliculus (Ecker et al., 2010, Brown et al., 2010). In addition, an M6 subtype has recently been reported in the mouse. It shows a complex dendritic architecture with dendrites that regularly cross the ON-OFF sublaminae but exclusively stratify in the ON sublamina of the IPL. However, to date there has been no published data regarding this proposed subtype and therefore there has been no characterisation of its somato-dendritic architecture or intrinsic responses to light (Quattrochi et al., 2013).

Overall, much less information is known about the M3, M4 and M5 subtypes compared to the M1 and M2 subtypes. Collectively, literature often refers to M1 ipRGCs and non-M1 ipRGCs (M2, M3, M4 and M5) as current information surrounding M3, M4 and M5 ipRGCs suggests similarities between them and M2 ipRGCs in regards to somato-dendritic fields and electrophysiological responses to light when isolated from the outer retina (Berson et al., 2010, Estevez et al., 2012, Ecker et al., 2010). A summary table of the subtype specific characteristics of individual ipRGCs is described in Figure 1.7.

Figure 1.7 Summary of ipRGC subtypes Anatomical and electrophysiological characteristics of the five known subtypes of ipRGC in the Opn4Cre mouse retina. ipRGCs are readily characterised anatomically by differences in soma size, dendritic stratification, dendritic field size and complexity of dendritic architecture. They also exhibit differences in the magnitude of their intrinsic light-response and sensitivity. Data compiled from (Schmidt et al., 2011), (Ecker et al., 2010) and (Estevez et al., 2012).

Chapter 1 General Introduction

Melanopsin expression in ipRGCs Melanopsin is encoded in mammals by the Opn4 gene. Two splice variants of this gene are expressed in mice: a 521 amino acid protein (Opn4L) and a 466 amino acid protein (Opn4S) (Pires et al., 2009). M1 ipRGCs express both isoforms of the Opn4 gene whilst M2 cells express only the Opn4L isoform (Pires et al., 2009). The Opn4S isoform is also expressed at 40 times the concentration of Opn4L. The first 454 amino acids of these isoforms are identical with alterations in amino acid sequence occurring at the C-terminal region. Variations of this C-terminal region in other GPCRs govern receptor kinetics, rates of desensitisation, G-protein coupling and constitutive activity (Minneman, 2001). Accordingly, the differences in the temporal pattern of expression during development (Hughes et al., 2012b), their differential cell-type specific expression in the adult retina and differences in their C-terminal length (Pires et al., 2009) is thought to be an important feature which accounts for the different signalling kinetics of the five ipRGC subtypes (Schmidt et al., 2011a, Schmidt et al., 2011b). To date, there have been no investigations into the isoforms expressed in M3, M4, and M5 subtypes, likely due to their more recent identification and extremely weak expression of the melanopsin protein (Ecker et al., 2010). Once transcribed, melanopsin is transported to the plasma membrane where it is expressed in the soma, dendrites and axons of ipRGCs in order to maximise the surface area for photon capture (Hattar et al., 2002).

Spectral sensitivity of melanopsin The first action spectrum for mammalian melanopsin utilised the pupillary light reflex (PLR) as a reporter of its spectral sensitivity. It found that the then unknown inner retinal photopigment met the spectral response properties of a Vitamin A based chromophore and formed a functional photopigment which was maximally sensitive to 479nm photo-stimulation (Lucas et al., 2001a). Subsequent action spectra used a variety of different physiological and behavioural responses to light including circadian phase-shifting of locomotor activity (Hattar et al., 2003), the cone electroretinogram (Hankins and Lucas, 2002), and direct electrophysiological recordings of ipRGCs (Berson et al., 2002). All these methods supported the findings that these light responses are mediated by a photopigment with a λmax of approximately 480nm. The spectral identity of the established murine photoreceptors: rods (λmax = 498nm) and cones (S-cone: λmax = 360nm; M-cone:

λmax = 505nm) (Fu and Yau, 2007) do not match the peak physiological or behavioural responses elicited by 480nm stimulation and therefore they were believed to be mediated by the newly identified melanopsin photopigment (Lucas et al., 2001a). However, this was not unequivocally confirmed until it was demonstrated that heterologous expression of the melanopsin protein in Neura-2a (Melyan et al., 2005), Xenopus oocytes (Panda et al., 2005) and HEK293-TRPC3 cells (Qiu et al., 2005) conferred light sensitivity. Photo-stimulating these cells resulted in large and sustained depolarising currents similar to that observed in ipRGCs in-vivo (Berson et al., 2002).

Chapter 1 General Introduction

Physiological responses of ipRGCs 1.2.5.1 Melanopsin phototransduction The intrinsic light-response of ipRGCs is indicative of melanopsin utilising an invertebrate phototransduction cascade. This is supported by cDNA analysis that demonstrates that Xenopus melanopsin has significantly higher sequence similarity to squid rhodopsin (39%) than it does to mammalian rod and cone opsins (30%) (Provencio et al., 1998). Furthermore, the presence of a tyrosine residue in the counterion position (E113) and elongations in the third cytoplasmic loop and C-terminus are also highly suggestive of a strictly regulated invertebrate-like photopigment (Provencio et al., 1998, Provencio et al., 2000).

A wealth of pharmacological (Panda et al., 2005, Qiu et al., 2005) and electrophysiological recordings (Berson et al., 2002, Dacey et al., 2005), conducted both in-vitro and in-vivo, also indicate that melanopsin employs an invertebrate-like signalling cascade. The application of Pertussis toxin (PTX) specifically inhibits the interaction of Gαi and Gαt G-proteins with their corresponding GPCR. Therefore PTX inhibits rod and cone signalling pathways. However, the application of PTX onto melanopsin-expressing cells does not affect light-induced depolarisations in these cells (Graham et al., 2008, Panda et al., 2005). In addition, antibodies specifically raised against the Gq G-protein result in an attenuation of the light-response but this is not affected when utilising Gi or Gt antibodies (Panda et al., 2005). Pharmacological blockade of the Gi/o pathway using NF023 also has no effect on the melanopsin-dependent response (Melyan et al., 2005), whilst application of the competitive antagonist of Gαq (GPant-2a) inhibits the response in HEK293-TRPC3 cells (Graham et al., 2008, Panda et al., 2005). Further downstream in the proposed signalling cascade, PLC inhibitors have been shown to block melanopsin dependent light-induced depolarisations (Panda et al., 2005,

Qiu et al., 2005), supporting the hypothesis that photo-stimulation of melanopsin activates the Gq / phosphatidylinositol signalling pathway (Graham et al., 2008, Hughes et al., 2012a).

Thus, a hypothesised photo-transduction mechanism for melanopsin has been put forward and has been recently reviewed (Hughes et al., 2012a) and is illustrated in Figure 1.8. In brief, stimulation of the melanopsin photopigment photo-isomerises 11-cis-retinal to all-trans-retinal (Walker et al.,

2008). The consequential liberation of Gαq subunits activates the Phospholipase C isoform PLCβ4 which cleaves the membrane-bound protein phosphatidylinositol 4,5-bisphosphate (PIP2) into two products: free cytoplasmic inositol 1,4,5-trisphosphate (IP3) and membrane bound 1,2-diacylglycerol

(DAG). IP3 diffuses through the cytoplasm where it acts on IP3 receptors expressed on the membrane of the endoplasmic reticulum (ER), stimulating the release of intracellular calcium stores. Meanwhile, DAG remains bound to the plasma membrane where it activates Protein Kinase C (PKC) through interactions with released calcium ions. PKC acts through various signalling pathways to open transient receptor potential (TRP) channels, most notably TRPC6 and TRPC7 in M1 ipRGCs (Hartwick et al., 2007, Sekaran et al., 2007, Warren et al., 2006). These channels are non-selective cation channels. The resultant influx of cations subsequently raises membrane potential and

Chapter 1 General Introduction

activates voltage-gated cation channels (Graham et al., 2008, Hartwick et al., 2007) which result in neuronal depolarisation and action potential generation. However, more recently it has been demonstrated that mice lacking functional Gq isoforms still retain melanopsin-driven behavioural responses to light which indicates that melanopsin may additionally employ other signalling pathways (Chew et al., 2014).

Figure 1.8 Phototransduction in ipRGCs (1) The absorption of light by the melanopsin photopigment (black) results in conformation changes and the subsequent activation and liberation of the Gαq G-protein. (2) Liberated Gαq binds and activates PLCB4 (green). (3) Activated PLCB4 cleaves PIP2 (red) into two products, membrane bound DAG (red; rectangle) and free IP3 (red; circle). (4) Liberated IP3 activates IP3 receptors on the Endoplasmic Reticulum resulting in the release of calcium ions (yellow). (5) Simultaneously, membrane-bound DAG activates PKC with the aid of released calcium ions. (6) Activated PKC acts through other intermediates to open Transient receptor channels (blue) and allow for the influx of Calcium (gold) and sodium (navy) ions. Image based on data collected from (Hughes et al., 2012a).

1.2.5.2 Melanopsin bi-stability and chromophore regeneration Rod and cone photoreceptors use a complex metabolic pathway involving the retinal pigment epithelium in order to recycle all-trans retinal back to the 11-cis retinal isoform (Kiser et al., 2012, McBee et al., 2001, Wang and Kefalov, 2011). It would seem inefficient for ipRGCs, which are found far away from the RPE, to use this same method. It has therefore been proposed that melanopsin may form a bi-stable photopigment with its retinaldehyde chromophore (Fu et al., 2005, Melyan et al., 2005, Mure et al., 2009). Bi-stable photopigments are thermally stable after photon absorption and therefore do not release the all-trans-retinal isoform following light-induced activation. Instead, the subsequent absorption of additional photons can regenerate the 11-cis-retinal isoform without

Chapter 1 General Introduction

the requirement of an enzymatic recycling pathway. This is a common feature shared by many invertebrate photopigments (Hillman et al., 1983).

A number of studies have implicated melanopsin’s bi-stability, however this is yet to be conclusively proved. For example, the pupillary light reflex (PLR) only reaches a maximum stable constriction a few minutes after broad spectrum light-stimulation which is characteristic of the establishment of a photo-equilibrium between 11-cis and all-trans-retinal isoforms (Mure et al., 2009). Two absorption spectra for melanopsin binding 11-cis-retinal and all-trans-retinal have been proposed with peak sensitivities at 481nm and 587nm, respectively (Mure et al., 2009). Additionally, ipRGCs in RPE65-/- knockout mice (a key enzyme in the retinaldehyde recycling cascade in the RPE) demonstrate 20-40 fold lower firing rates. Normal light responses are restored by the addition of further all-trans-retinal which indicates that ipRGCs utilise a mechanism independent of the RPE-mediated visual cycle (Fu et al., 2005). More recently, it has also been suggested that melanopsin may form a tri-stable photopigment; here, stimulation with broad-spectrum light results in the melanopsin photopigment existing in one of two inactive states or one signalling state (Emanuel and Do, 2015). A photo- equilibrium is created which allows for melanopsin to sustain long term signalling in response to broad wavelength light (Do et al., 2009, Wong, 2012).

As Müller cells span the length of the retina (Figure 1.1) and are involved in the cone specific visual cycle (Wang and Kefalov, 2011), they have also been proposed to play a role in chromophore regeneration for melanopsin (Lucas, 2006). However there is currently no direct evidence that melanopsin uses this method. Nonetheless, the potential bi-stable characteristics of melanopsin would be well suited to provide the sustained depolarising responses observed in ipRGCs which help them support their roles in non-image forming vision (Lucas et al., 2001a, Lucas et al., 2001b).

1.2.5.3 Electrophysiological responses to light The electrophysiological response of melanopsin activation exhibits key identifiable features which distinguish them from light responses derived from rod and cone photoreceptors. The intrinsic melanopsin light response exhibits a modest activation threshold with light responses observable at irradiances above 1012 photons cm-2 s-1 (Berson et al., 2002, Tu et al., 2005, Do et al., 2009). In response to a light step above this threshold, the intrinsic melanopsin response demonstrates a slow increase in firing rate which reaches a steady plateau over the course of seconds. This firing rate is sustained throughout the duration of the light stimulus and for tens of seconds to minutes after its termination (Berson et al., 2002, Dacey et al., 2005, Tu et al., 2005, Wong et al., 2007) (Figure 1.9A). The intrinsic melanopsin response exhibits a long response latency (Berson et al., 2002) with absorption of a single photon resulting in up to an 8 second delay before the cell reaches its peak firing rate (Do et al., 2009). This long onset latency under dim lighting allows for temporal integration of light so that photon absorption has a summative effect on membrane potential and

Chapter 1 General Introduction

will result in a higher probability of action potential generation. Accordingly, this response latency is inversely proportional to the strength of the stimulus, with the saturated melanopsin response reaching peak within a few hundred milliseconds (Lucas et al., 2001a). Thus, although the intrinsic melanopsin response demonstrates poor sensitivity they exhibit a monotonic relationship with increasing light intensity. This is reflected in the proportional increase of their firing rate in response to increasing irradiance (Berson et al., 2002, Warren et al., 2003).

These poor physiological responses to light are in stark contrast to the intrinsic melanopsin response which additionally integrates light information from rod and cone photoreceptors. Under these conditions an ipRGC’s light response demonstrates a brisk transient increase in firing rate which peaks within 200ms before relaxing to a steady state firing rate throughout the duration of the light pulse (Figure 1.9B). Furthermore, the activation threshold for the intrinsic melanopsin response is approximately six orders of magnitude lower under these conditions (Wong et al., 2007).

Figure 1.9 Melanopsin-driven light responses in the mouse. (A) Representative schematic diagrams of the intrinsic light-response of an ipRGC when presented with a full field light step (indicated by the light monitor at bottom). Following light-onset, the firing rate of deafferented ipRGCs builds up slowly over time before reaching its peak. This peak response is maintained until light-offset at which point the firing rate slowly decreases over the course of tens of seconds to minutes. (B) The integrated light response of the intrinsic melanopsin signalling and extrinsic synaptic input from rods and cones demonstrates an initial transient increase in firing rate, which relaxes to a steady state plateau which is maintained throughout the course of the light-step. The onset and offset kinetics are tightly coupled to both light onset and light offset, respectively. Schematic diagrams based on data collected by (Hattar et al., 2002) and (Brown et al., 2010).

A recent study by Do et al., 2009 looked to investigate the signalling properties of the intrinsic melanopsin response. The authors identified that the melanopsin photopigment is expressed at a very low density of approximately 3 molecules per µm-2 (Do et al., 2009) compared to rods and cones which have their pigments arranged efficiently into stacks at a density of ̴25,000 molecules per µm-2. This poor density of expression means ipRGCs are poor at catching photons, rather than

Chapter 1 General Introduction

possessing an inefficient phototransduction mechanism (Do et al., 2009). This low pigment density does however permit maximum transmission of light through to the outer retina with minimal interference, in order to retain the high spatio-temporal resolution that rods and cones require for high visual acuity. ipRGCs subsequently receive strong synaptic inputs back from the outer retina (Perez-Leon et al., 2006, Wong et al., 2007) which compensates for their poor photon capture. As such, an ipRGCs membrane potential continuously operates near the threshold for action potential generation such that single photon can result in a 1mV change in membrane potential, and as a consequence, result in action potential generation (Do et al., 2009).

Inputs to ipRGCs ipRGCs are intrinsically photosensitive and can respond to light in isolation of any synaptic input due to the expression of the melanopsin photopigment (Berson et al., 2002, Hattar et al., 2002, Lucas et al., 2001a). However, anatomical (Belenky et al., 2003, Ostergaard et al., 2007) and electrophysiological (Wong et al., 2007, Do et al., 2009) evidence demonstrates that ipRGCs receive extensive synaptic input from both bipolar cells and amacrine cells and so integrate their intrinsic light-response with those from the outer retina under normal physiological conditions (Figure 1.10).

These inputs have been readily identified by pharmacology coupled with in-vitro electrophysiological recordings. In the dark, ipRGCs exhibit spontaneous depolarisations due to the synaptic input from bipolar and amacrine cells maintaining their membrane potential near threshold (Perez-Leon et al., 2006, Wong et al., 2007, Do et al., 2009). Pharmacologically blocking synaptic input from Amacrine cells results in an increase in the frequency of these spontaneous depolarisations (Wong et al., 2007) indicating that Amacrine cells provide the inhibitory input to ipRGCs. Upon presentation of a single light pulse whilst amacrine cells are under pharmacological blockade, ipRGCs respond with the characteristic intrinsic-melanopsin light response - a long response latency, large sustained depolarisations and persistence of firing after stimulus termination. However, both the transient and sustained components of the light response are more robust under these conditions and support the conclusions that amacrine cells provide inhibitory input to ipRGCs (Wong et al., 2007). As previously mentioned, recording from ipRGCs under normal physiological conditions where rod and cone signalling is permitted demonstrates that light some six orders of magnitude lower can initiate the melanopsin-driven light-response (Wong et al., 2007). These responses are more tonic displaying brief action potential fire with faster temporal kinetics. This indicates that the rod and cone photoreceptors provide excitatory input via bipolar cells, and modulate the firing activity of ipRGCs at lower light intensities.

Chapter 1 General Introduction

Figure 1.10 Synaptic inputs to ipRGCs in the mouse retina. Schematic diagram illustrating the synaptic modulations of ipRGC activity by bipolar cells and amacrine cells in the mouse retina. ON bipolar cells provide excitatory input (green crosses) from glutamatergic transmission in response to light. OFF Bipolar cells provide excitatory input at the removal of the light. Amacrine cells provide inhibitory input (red minus sign) via GABAergic and Glycinergic transmission in addition to modulatory Dopamine release. Diagram adapted from (Wong et al., 2007).

The neurotransmitter systems which modulate ipRGC signalling have also been characterised by pharmacological and electrophysiological investigations in both rodents and primates. In mice, an ON-response is observed during the initial presentation of the light-step when pharmacologically blocking Amacrine cell input (Wong et al., 2007). The addition of the metabotropic agonist L-AP4 has been shown to reduce the firing rate of ipRGCs (Schmidt and Kofuji, 2010) indicating that ON-bipolar cells provide the excitatory input by release of glutamate from their presynaptic terminal. By contrast, the presence of GABA and Glycine in synaptic terminals of Amacrine cells (Pourcho, 1996, Crooks and Kolb, 1992, Menger et al., 1998) and the fact that their pharmacological blockade increases baseline firing activity of ipRGCs (Wong et al., 2007) suggests these cells provide the inhibitory input to ipRGCs.

A fourth neurotransmitter has also been shown to influence melanopsin signalling. ipRGCs receive synaptic input from dopaminergic (DA) expressing amacrine cells (Ostergaard et al., 2007, Vugler et al., 2007). Dopamine has been shown to attenuate the intrinsic light response by acting on D1-type receptors expressed on the dendrites of ipRGCs (Van Hook et al., 2012). This action has been speculated to fulfil a light-adaptation role where it is capable of modulating the sensitivity of ipRGCs to light during the transition between scotopic and photopic vision (Witkovsky, 2004). However, a juxtaposing role for retinal Dopamine is its action on D2-type receptors expressed on the dendrites of ipRGCs which act to down-regulate the expression of melanopsin mRNA in response to light

Chapter 1 General Introduction

(Sakamoto et al., 2005). Consequently, the role of Dopamine on ipRGCs is still an area of ongoing research. Whilst the major pathways in the retina use chemical synapses, there is evidence that electrical synapses provide a second mechanism of communication (Bloomfield and Volgyi, 2009). Tracer studies of ipRGCs in the wildtype retina demonstrated that they make electrical synapses with wide-field GABAergic amacrine cells (Muller et al., 2010); however, the authors found no evidence of direct communication between ipRGCs or heterologous communication with other RGC subtypes. This suggests that ipRGCs do not directly communicate with each other by this mechanism.

Parallel investigations in primates support these findings and provide further insight into the contribution of different cell types. Work by Jusuf et al., 2007 investigated the synaptic connections of ipRGCs in the marmoset retina. The authors demonstrated that ipRGCs that stratified in the ON sublamina of the IPL (the non-M1 type ipRGCs) made synaptic connections with cone OFF-bipolar cells which predominantly receive input from M- and L-cones (Jusuf et al., 2007). Contrastingly, ipRGCs which stratify in the OFF sublamina (the M1 ipRGCs) make synaptic connections with Rod bipolar cells and S-cone bipolar cells. These findings are in agreement with the work of Dacey et al., 2005 who report that ipRGCs in the primate retina can signal irradiance and colour to the dLGN due to synaptic input from OFF S-cone bipolar cells which act antagonistically to ON L- and M-cone bipolar cells (Dacey et al., 2005).

Outputs of ipRGCs 1.2.7.1 Central targets of ipRGCs for non-image forming vision The initial discovery of melanopsin and the subsequent research into its functional significance in mammalian physiology predominantly focussed on its role in non-image forming behaviours for two reasons. Firstly, the use of transgenic mice such as melanopsin knockout (Opn4-/-) and rodless/coneless (rd/rd cl) mice demonstrated the contribution of melanopsin signalling to non- image forming behaviours through impairment of circadian entrainment (Panda et al., 2002, Ruby et al., 2002, Foster et al., 1993) and the PLR (Lucas et al., 2001a, Lucas et al., 2003b). Secondly, histology on the Opn4tau/LacZ mouse demonstrated that ipRGCs specifically innervated brain structures involved in non-image forming responses to light (Hattar et al., 2006, McNeill et al., 2011) before it was identified that this method did not label all ipRGC subtypes (Ecker et al., 2010, Brown et al., 2010). These structures included the SCN (the master circadian clock), the OPN shell (the relay centre for the PLR) and the intergeniculate leaflet (IGL) whose primary role is to integrate photic and non-photic information to synchronise circadian rhythms (Hattar et al., 2006, Baver et al., 2008). Additionally, ipRGCs project to other non-image forming nuclei including the lateral habenula, the medial amygdala, the ventral subparaventricular zone, the supraoptic nucleus and the ventral lateral geniculate nucleus (vLGN) (Hattar et al., 2006, Baver et al., 2008) (Figure 1.11).

Chapter 1 General Introduction

Figure 1.11 Projection targets of ipRGCs Top: Schematic diagram of a sagittal projection of the mouse brain illustrating the central projection targets of ipRGCs in the Opn4Cre mouse. Regions in blue highlight innervations to non-image forming brain structures whilst regions in green illustrate more recently identified projections to image forming nuclei. Bottom: Abbreviations and functions of these innervated brain regions and their suspected subtype specificity. Image adapted from (Hattar et al., 2006) with subtype specific data additionally compiled from (Baver et al., 2008), (Ecker et al., 2010) and (Schmidt et al., 2011).

Chapter 1 General Introduction

Central targets of ipRGCs in image forming vision The development of Opn4Cre mice revealed a much wider array of ipRGC retinofugal projections than was initially appreciated. This new repertoire of non-M1 ipRGC target nuclei had been expanded to include the dorsal lateral geniculate nucleus (dLGN) and the superior colliculus (SC) (Brown et al., 2010, Ecker et al., 2010) - two structures strongly associated with the image forming visual system. The dLGN has a well-established role in processing and relaying visual information to the primary visual cortex, whilst the superior colliculus is involved in eye movements and co-ordinating goal orientated behaviours by integrating sensory and motor information (Grubb and Thompson, 2003, May, 2006, Wang et al., 2010, Denman and Contreras, 2016). These newly found ipRGC projections have fuelled a resurgence into understanding the capacity and extent to which the melanopsin system contributes to image forming vision. Growing anatomical and electrophysiological evidence from the dLGN, in addition to behavioural experiments, suggests that the melanopsin system is not mutually exclusive from that of the rod and cone system, and that integrated light-information from both these photoreceptive systems contributes to visual perception (Allen et al., 2014, Brown et al., 2010, Brown et al., 2012, Storchi et al., 2015).

1.2.8.1 The Lateral Geniculate Nucleus of the mouse The major projection target of RGCs in the mouse is the Lateral Geniculate Nucleus (LGN) of the visual thalamus (Grubb and Thompson, 2003). Approximately 97% of the RGCs arising from each eye cross the optic chiasm and innervate the contralateral LGN (Hubener, 2003, Reese, 1988). The LGN is a kidney bean shaped nucleus found in the visual thalamus, just ventral of the hippocampus (Figure 1.12A). Its function is to relay visual information from the retinal ganglion cell population to brain structures involved in both image forming and non-image forming vision. Accordingly, it can be subdivided into three anatomically and functionally distinct regions, the dLGN (dorsal Lateral Geniculate nucleus), the IGL (Inter geniculate leaflet) and the vLGN (ventral Lateral Geniculate nucleus) (Figure 1.12B).

Chapter 1 General Introduction

Figure 1.12 The Lateral Geniculate Nucleus of the mouse. (A) Schematic diagram of a coronal section from the mouse brain atlas (AP = -2.10mm; Paxinos and Watsons “The mouse atlas in stereotaxic coordinates”) highlighting the anatomical position of the Lateral Geniculate Nucleus (LGN) in blue. (B) Magnified schematic diagram of the blue region from (A) showing the structure of the mouse LGN, which consists of three sub regions: the ventral LGN (vLGN), the intergeniculate leaflet (IGL) and the dorsal LGN (dLGN). The dLGN can be further subdivided into an inner core and outer shell (dashed lines) due to the retinotopic organisation of the projections it receives from both eyes.

1.2.8.2 Anatomy of the mouse dLGN The dorsal LGN is the major retinorecipient of RGC projections for image forming vision in the mouse (Huberman and Niell, 2011, Grubb and Thompson, 2003). It can be further subdivided into an inner core and outer shell based upon the region specific targeting of RGCs from the retina (Reese, 1988). The inner core maps the contralateral region of the visual field by receiving strong projections from the contralateral retina and weakly from the temporal regions of the ipsilateral retina. The outer shell receives projections exclusively from the contralateral retina (Reese, 1988). The decussation of RGC axons at the optic chiasm, and innervation pattern in the dLGN, suggest a region corresponding to approximately 60° of visual angle in front of the mouse is mapped by both eyes (Sefton and Dreher, 1995). This specific innervation pattern also results in the creation of a retinotopic map of visual space (Piscopo et al., 2013, Hubener, 2003). This retinotopic organisation preserves spatial relationships between salient objects in the visual scene as stimuli which appear adjacent to each other in the environment are processed by adjacent populations of neurones in the dLGN (Figure 1.13). This anatomical organisation, in addition to more recent electrophysiological evidence in rodents (Grieve, 2005, Howarth et al., 2014), has demonstrated that binocular integration arises as early as the dLGN in mice. In primates this does not occur until the visual cortex (Casagrande and Boyd, 1996) and highlights the difference in the structural organisation of the dLGN in mice and non- human primates.

Chapter 1 General Introduction

Figure 1.13 Retinotopic organisation of the mouse visual system. The left and right visual fields and their representations in the retina, dLGN and Visual cortex are coloured in red and green, respectively. 97% of RGC projections (dark red and dark green) cross the optic chiasm and innervate the contralateral dLGN in a retinotopic manner. A small number of RGCs from temporal retina of each eye (light red and light green) project to the inner core of the ipsilateral dLGN. These fibres monitor the small binocular zone which encompasses approximately 60° directly in front of the mouse (between regions 2 & 4). LGN neurones then project to the visual cortex, which is also represented in a retinotopic order. Numbers in the visual field represent the retinotopic organisation of the visual cortex. Note the binocular zone is represented in both hemispheres in V1. Diagram adapted from (Hubener, 2003).

This inner core and outer shell structure is significantly different to the laminar structure found in the primate dLGN. Consequently, the well-established primate (P, M and K) or feline (X, Y and Z) parallel processing pathways are yet to be convincingly identified in the mouse based upon their known electrophysiological characteristics in other species (Grubb and Thompson, 2003, Denman and Contreras, 2016). However, cells in the mouse dLGN can be identified morphologically and suggest that there are functional differences in the information conveyed (Krahe et al., 2011).

In addition to receiving sensory information directly from the retina, the dLGN also receives a number of reciprocal projections from other visual processing structures including the striate cortex and superior colliculus (Hubener, 2003, Grubb et al., 2003). Neurones in the dLGN send their projections to numerous brain structures, of which the major recipient is the visual cortex (Huberman and Niell, 2011, Gao et al., 2010, Cruz-Martin et al., 2014, Niell and Stryker, 2010). Thalamo-cortical projections are also arranged retinotopically with cortical cells exhibiting a number of complex features including orientation, speed and direction selectivity (Gao et al., 2010, Bonin et al., 2011, Schuett et al., 2002a, Smith and Haeusser, 2010).

Chapter 1 General Introduction

1.2.8.3 Electrophysiological responses in the mouse dLGN Although the mouse has become a widely-appreciated model to study the visual system, there have been relatively few studies which have characterised visual responses in the mouse dLGN. Nonetheless, light-responses in the dLGN can be readily classified as exhibiting either a transient or sustained response to a light-step (Brown et al., 2010, Piscopo et al., 2013)(Figure 1.14). Transient cells show an initial increase at light onset (Transient-ON), at light offset (Transient-OFF) or at both light onset and light offset (Transient ON-OFF). However, the firing rate of these cells quickly returns to baseline during and after the initial change in irradiance. On the contrary, Sustained-ON cells show an initial transient increase in firing rate at light onset which relaxes to a plateau and is maintained throughout the duration of the light-pulse. Sustained-OFF cells show a brisk increase in firing rate at light onset which then relaxes and remains significantly below baseline firing throughout the duration of the light-pulse. These sustained cells can also show a small transient OFF response at the termination of the stimulus. This variation in these light-responses represents the convergence of individual retinal ganglion cells onto an individual LGN neurone. It is believed that on average four retinal ganglion cells synapse with an individual dLGN neurone in the mouse dLGN (Chen and Regehr, 2000).

Chapter 1 General Introduction

Figure 1.14 Light response profiles in the mouse dLGN. The electrophysiological response of LGN neurones to light can be broadly divided into two groups, Transient or Sustained. Transient cells show a brisk increase in firing rate at light onset (Transient-ON) or at light offset (Transient-OFF) but quickly return to baseline after. Transient ON-OFF cells show a brief increase in firing rate at both light onset and offset. Sustained cells can be categorised as Sustained-ON, whereby they show an initial transient increase in firing rate which then relaxes to a steady firing rate that is maintained throughout the light-step. At light-offset there can sometimes be a small second transient response. Sustained-OFF cells show a decrease in firing rate at light onset which is maintained throughout the duration of the light-step. At light-offset there can sometimes be a small transient-OFF response. Schematic diagrams based on data collected by (Brown et al., 2010)

The spatial and temporal characteristics of these light-responses have also been studied. The spatial receptive fields of dLGN neurones are concentric circles which exhibit the antagonistic centre- surround similar to that described in the retina (Grubb and Thompson, 2003). These cells possess either ON or OFF-centre receptive fields and typically have a receptive field diameter of between 2- 10°, although some cells have been reported to possess significantly larger receptive fields up to 27° in diameter (Grubb and Thompson, 2003). LGN neurones in the mouse exhibit poor spatial acuity compared to other mammals due to their small eyes and poor optics. Accordingly, LGN neurones possess spatial frequency cut-off between 0.5-0.6 cycles/° and an average spatial frequency of 0.2 cycles/° (Grubb and Thompson, 2003) with a preferred temporal frequency of approximately 4Hz. Populations of neurones in the dLGN also exhibit orientation and direction selectivity (Piscopo et al., 2013).

1.3 Melanopsin’s contribution to image forming vision

The discovery that a subset of ipRGCs project to the dLGN raised a number of questions about what the role of melanopsin may be in this nucleus. Melanopsin had always been viewed as being ill- suited for image forming vision due to its slow signalling kinetics and its strong modulation from synaptic inputs from rod and cone photoreceptors. Nonetheless, a number of studies have since been investigating how melanopsin signalling influences more conventional aspects of image forming vision using molecular, electrophysiological and behavioural techniques (Brown et al., 2010, Brown et al., 2012, Storchi et al., 2015, Allen et al., 2014, Schmidt et al., 2014, Ecker et al., 2010).

ipRGCs during development Spontaneous waves of synchronised depolarisations are observed throughout the retina of the developing system. These so called “retinal waves” are critical in the development of the mature visual system by targeting, refining and pruning RGC projections in the dLGN (Chen and Regehr, 2000) and the formation of ocular dominance columns in the primary visual cortex (V1) (Wong, 1999). However, during the time-frame within which retinal waves are observed, the rod and cone system is not fully developed or light-sensitive until at least post-natal day 10 (P10) (Rich et al., 1997). For this reason, retinal waves were presumed to be solely generated internally by cholinergic transmission originating from starburst amacrine cells (Zheng et al., 2006). Since the discovery of

Chapter 1 General Introduction

melanopsin expressing ipRGCs this view has been somewhat revised. Whilst the rod and cone system is absent until P10, M1-type ipRGCs can be detected as early as P0 (Tu et al., 2005). Although mice maintain closed eyes until full development of the visual system (approximately P12), light- stimulation of juvenile mice (P4-P7) increases burst activity of retinal waves by up to 50%. This does not occur in Opn4-/- knockout mice (Renna et al., 2011).

Thus, whilst light is not essential for retinal wave propagation it clearly has the capacity to modulate this system. This was demonstrated anatomically by rearing Opn4-/- and wild type mice in bright light for 7 days, after which a retrograde tracer was injected into the eye to map the projections of RGCs. Histological analysis confirmed that Opn4-/- mice exhibited less segregated retino-geniculate afferents, implicating the involvement of ipRGCs in the specific targeting and arborisation RGC terminals in this nucleus (Renna et al., 2011). These projections to the dLGN suggest a role in visual system development, and are supported by negative phototaxis observed in juvenile wild type mice (P6 - P9) but not Opn4-/- knockouts (Johnson et al., 2010).

Another notable difference during development of the mouse retina is the temporal expression of the melanopsin gene. Opn4S increases rapidly between P0 and P3, whilst Opn4L remains at extremely low concentrations until around P14 (Hughes et al., 2012b). As aforementioned, Opn4S is expressed in M1-type ipRGCs along with Opn4L (Pires et al., 2009). This renders M1 cells photosensitive from birth, whilst M2-type ipRGCs become photosensitive two weeks after birth (Hughes et al., 2012b). Interestingly, the increase in melanopsin expression in non-M1 ipRGCs coincides with the development of rods, cones and the image forming visual system around P10 (Tu et al., 2005) and supported the hypothesis that non-M1 cells may exhibit a more closely related function with the image forming visual system.

Electrophysiological responses Arguably, the strongest role of melanopsin’s involvement in image-forming vision arises from electrophysiological recordings from the dLGN. This is most clearly demonstrated in aged rd/rd cl mice who have complete loss of the rod and cone photoreceptors, but the inner retina remains functional and intact (Brown et al., 2010, Carterdawson et al., 1978). Accordingly, stimulation with blue light (460nm) results in widespread light-responses in approximately 40% of neurones in the dLGN (Brown et al., 2010). These cells demonstrate the slow temporal kinetics and sustained firing activity associated with the intrinsic melanopsin light response. Moreover, optical imaging methods have revealed these same slow increases in activity following light stimulation in the visual cortex and other cortical structures involved in visual processing (Brown et al., 2010).

One of the complications of determining melanopsin’s contribution to visual responses in the intact retina is the similarity in the spectral sensitivity of rod opsin, melanopsin and M-cone opsins (Jacobs

Chapter 1 General Introduction

et al., 2007). One way to circumvent this problem is to use Red cone knock-in (Opn1mwR) mice. These mice have a fully intact visual system and normal projection patterns, however the M-cone opsin is replaced with human L-cone opsin (Smallwood et al., 2003). Therefore cones which did express M-cone opsin show a spectral shift from green light to red light (Lall et al., 2010). One can take advantage of the unique spectral sensitivity of the different opsins and create metameric pairs of stimuli which provide contrast for certain photopigments, whilst keeping others isoluminant (silent). Therefore, this mouse model allows one to more readily tease apart the contribution of individual photoreceptors to light responses based upon their individual spectral sensitivity in-vivo (Allen and Lucas, 2016, Brown et al., 2010).

Accordingly, Brown et al., 2010 used this method and presented bright blue light-steps (460nm) to Opn1mwR mice and found light responsive LGN neurones responded in two distinct ways: transient or sustained. Transient cells demonstrate brief spiking activity in response to the initial presentation of a stimulus but rapidly decay to baseline soon after. Sustained cells exhibit an increase in firing activity in response to light presentation which remains elevated throughout duration of the stimulus and for up to 10 seconds after its absence. Next, these mice were presented with longer wavelength red light (655nm), which was indistinguishable for cones. Whilst there was no effect on the transient population of cells, the sustained populations demonstrated significantly higher sensitivity when presented with the shorter wavelength light which matched the spectral sensitivity of the melanopsin photopigment (Brown et al., 2010). The sustained population of cells responded within 200ms of the presentation of the blue light stimulus, and took up to 2 seconds to reach a maximum sustained firing rate with no decline throughout the 60 second light-step. The slow temporal kinetics of the sustained component of the light response was reminiscent of the known sluggish kinetics of melanopsin signalling described in the retina (Berson et al., 2002, Wong et al., 2007). Furthermore, melanopsin-deficient mice (Opn4-/-) have a significant attenuation of the sustained component of the response when compared to rd/rd cl and Opn1mwR mice, who retain functional expression of the melanopsin photopigment (Brown et al., 2010).

This data demonstrated that melanopsin contributes significantly to approximately 40% of neurones in the dLGN which display the sustained-ON phenotype. However an unresolved question remained - why do ipRGCs, which account for less than 5% of the total RGC population in the retina, make such a disproportionate contribution to light responses in the dLGN and what function they fulfil? A number of subsequent studies have begun to investigate the contribution melanopsin signalling makes to the image-forming visual system using in-vitro and in-vivo electrophysiology.

In the non-image forming system, the fundamental role of melanopsin is to encode environmental irradiance. Indeed, this is no different in the projection to the dLGN. The firing rate of both the transient and sustained components of light-responses in the dLGN are proportional to the irradiance signal, demonstrating that melanopsin is crucial in encoding the intensity of visual stimuli

Chapter 1 General Introduction

(Brown et al., 2010). Taking this concept further, a recent study demonstrated that as background irradiance increases, comparable to that which occurs at the night-time to dawn transition, individual LGN cells demonstrate an increased background firing rate that originates from ipRGCs (Storchi et al., 2015). When high-frequency visual stimuli (designed to preferentially excite rod and cone photoreceptors and mimic naturalistic vision) were super-imposed onto this increasing background, dLGN cells showed an increase in the amplitude, and an improvement in the reliability of the transient component of the light-response. This suggests that not only are ipRGCs encoding fundamental irradiance information but that melanopsin signalling to the dLGN also helps increase the signal:noise ratio of visual responses (Storchi et al., 2015). In agreement with this, the activation of melanopsin has been shown to increase feature selectivity such that the visual system becomes tuned for finer spatial resolution and faster motion during daylight conditions when melanopsin signalling would be most active (Allen et al., 2014). In the mouse retina, the M4 ipRGC, which projects to the dLGN, has also been shown to possess spatial receptive fields with antagonistic centre-surround organisation (Ecker et al., 2010). A direct melanopsin projection to the dLGN has also been identified in non-human primates which transmits colour opponent information (Dacey et al., 2005), but this has yet to be found in rodents and other mammalian species.

Behavioural responses Whilst electrophysiological recordings have been critical in understanding the contribution of melanopsin signalling to the dLGN, they are limited by the inability to determine their behavioural significance. Thus, a number of behavioural investigations have also been conducted which compliment these electrophysiological findings.

In a Y swim maze task, rd/rd cl mice can learn to swim towards a light target over a dark target to reach an escape platform after a long period of training, indicating these mice can perceive brightness (Brown et al., 2012). This finding has also been demonstrated using Opn1mwR mice using a similar paradigm, where the escape platform was associated with a green target and the incorrect lane, a red target. Mice rapidly learnt to swim towards the green target to which the melanopsin photopigment is more sensitive. To confirm that this behaviour was not due to any form of colour discrimination driven by cones, the green light was replaced with a 4 times dimmer red light. This resulted in the mice swimming towards the now perceived “brighter” red light without any further training (Brown et al., 2012). Parallel investigations supporting this brightness perception have also been conducted in human subjects. Participants underwent a two-interval alternative forced-choice test where they had to judge the perceived brightness of three metameric stimuli (melanopsin contrast: -11%, 0% and 11%) compared to a test stimulus of equal spectral composition whose radiance varied between trials. Subjects identified equal brightness when the test radiance was lower than the -11% contrast and at higher test radiances than the 11%, indicating higher

Chapter 1 General Introduction

melanopsin excitation correlates with greater perceived brightness (Brown et al., 2012). In a different study, optokinetic tracking and a visual navigation task were used to show that non-M1 type ipRGCs are critical in influencing the ON alpha Ganglion cell population in the retina in the mouse. In doing so, they aid in establishing the contrast detection threshold under different irradiance levels (Schmidt et al., 2014). To further confirm this, mice which had an inducible deletion of the melanopsin photopigment showed severe deficits in their contrast sensitivity function.

Many of these early behavioural investigations into the role of melanopsin in image forming vision used full-field uniform stimuli. However, a vital function of mammalian vision is their ability to resolve spatial information about their environment. Gnat1-/-; Cnga3-/- mice (which lack function rod and cone signalling cascades but express melanopsin) were placed in a similar Y maze behavioural paradigm as described above. However, this time a sinusoidal modulated grating (0.12 cycles/°) was presented down one arm of the maze under which a submerged escape platform was present, whilst down the other, a uniform grey stimulus of equal luminance was presented with no escape platform (Ecker et al., 2010). Although it took twice as long as their wild type controls, these melanopsin-only mice learnt to swim to the patterned stimulus to escape the maze to an extent that was greater than the criterion performance of 70% (Ecker et al., 2010). This behaviour was supported by the up-regulation of the neuronal metabolic marker c-Fos in the visual cortex after exposure to the pattern stimulus but not the uniform stimulus of equal irradiance, supporting the increased activity found in V1 by Brown et al., 2010. Finally, Gnat1-/-; Cnga3-/-; Opn4-/- mice (which lack rod and cone signalling and the melanopsin photopigment) were then investigated in the same paradigm to confirm the contribution of the melanopsin system to this pattern vision (Ecker et al., 2010). These mice did not reach criterion performance in the Y maze task and did not demonstrate up-regulation of c-Fos in the visual cortex confirming melanopsin’s contribution to these responses. Combined, this data suggests that ipRGCs are not simply encoding the global irradiance of the visual scene, but can discriminate the distribution of it across the visual field. However, this has since been questioned as it has been shown that the Gnat1-/-; Cnga3-/- mouse model retains some residual rod function which could account for these behavioural and histological observations (Allen et al., 2010, Hughes et al., 2016).

1.4 Retinal degeneration

As with other sensory systems, the retina is susceptible to disease, injury and cell death. Retinal degenerations encompass a wide range of diseases including retinitis pigmentosa, retinal dystrophy and age-related macular degeneration. Retinitis pigmentosa is the most common of these occurring at an incidence of between 1:2000 and 1:4000 (Strettoi, 2015, Delyfer et al., 2004, Hartong et al., 2006). They can result from a number of possible mutations which often occur in the rhodopsin protein itself (Hartong et al., 2006) and typically render rod photoreceptors incapable of photo-

Chapter 1 General Introduction

transduction, eventually leading to rod, and subsequent cone, death. This results in impaired vision, and in advanced stages leads to complete visual blindness with no available cure.

The rd1 mouse model of retinitis pigmentosa The rd1 mouse is one of the most widely used mouse models of retinitis pigmentosa (Soto and Kerschensteiner, 2015). This mouse strain carries a nonsense mutation on exon 7 of the PDEβ6 gene (Pittler and Baehr, 1991). This gene codes for the β-subunit of the cGMP phosphodiesterase which is expressed in the outer segments of rod cells. The effect of this mutation is that the PDEβ6 protein is expressed in a truncated form and therefore the animal is unable to produce a functional PDEαβ heteromer (Bowes et al., 1990). As such, cGMP cannot be broken down and begins to accumulate in rod cells where it becomes toxic (Farber and Lolley, 1974) and subsequently results in progressive rod cell death (Lolley, 1994). This is a rapid form of retinal degeneration such that by post-natal day 17 the rod cell population has undergone 99% cell death (Carterdawson et al., 1978). Approximately 50% of cones undergo cell death between post-natal day 17-30 with only 1% of the cone population surviving by 18 months of age (Carterdawson et al., 1978). The rd1 retina responds to the loss of sensory input by remodelling, initially at the cellular level but progressing to full scale re- organisation of the neural retina (Jones et al., 2003, Jones and Marc, 2005, Strettoi et al., 2003). The effect of photoreceptor death in rd1 mice is well documented and follows three main stages.

1.4.1.1 Stages of degeneration in the rd1 retina Phase 1

Phase 1 of rd1 retinal degeneration can be described by the elevated levels of cGMP resulting in rod photoreceptor stress (Figure 1.15). Although rods appear anatomically normal at birth, at no age can light responses originating from these cells be observed (Strettoi et al., 2002). Rods quickly appear shorter as their outer segments become phagocytosed by the RPE (Marc et al., 2003, Carterdawson et al., 1978). The synaptic contacts between rods and bipolar cells begin to deconstruct whilst concurrently sprouting new terminals which bypass their normal horizontal cell and bipolar cell targets, with some reaching as far as the ganglion cell layer (Marc et al., 2003) . Some horizontal cells also begin to show early signs of hypertrophy as they retract their dendrites from the rod photoreceptor to bipolar cells synapse. Towards the end of phase 1, rod cells retract their aberrant neurites which essentially isolates the rod bipolar cell network from photoreceptor input. As a consequence of this, some rod bipolar cells can also be observed to retract their dendrites and begin internalising their respective glutamate receptors (Strettoi et al., 2002, Strettoi et al., 2003).

Chapter 1 General Introduction

Figure 1.15 Stages of retinal degeneration in the rd1 mouse – Phase I As a result of the PDEβ6 mutation there is a toxic accumulation of cGMP in the outer segments of rod photoreceptors resulting in photoreceptor stress. Rod cells (black) in the ONL appear to shrink as they shed their outer segments. They begin sending aberrant neurite growths into the INL, IPL and GCL. This increase in toxicity results in rapid cell death of the rod photoreceptors by P17. Consequently, horizontal cells (HC) start retracting their projections and some rod bipolar cells (BC) can begin to show abnormal dendritic morphology in the OPL. The cone population appears anatomically normal at early stages but exhibit compromised responses to light during this stage. Some cone photoreceptors may also send aberrant neurite growths towards the GCL towards the end of stage 1. Diagram modified from (Soto & Kerschensteiner., 2015).

Phase 2

Phase 2 describes photoreceptor death (Figure 1.16). Here, rod cells have retracted their aberrant neurites and rod cell death ensues. This process is complete by post-natal day 17 (P17). Cone photoreceptors subsequently become stressed and start shedding their outer segments (Lin et al., 2009) and show abnormal retinal anatomy (Fei, 2002). Although they still demonstrate a compromised signalling capacity (Stasheff, 2008), cone also undergo cell death, most likely due to the loss of neurotrophic survival signals from dying rods (Mohand-Said et al., 1998) and the stress of the surrounding micro-environment. Only a monolayer of nuclei are observed in the photoreceptor

Chapter 1 General Introduction

layer by P30 (LaVail et al., 1997) however no light-responses can be identified by ERG (Strettoi et al., 2003) or MEA recordings (Stasheff, 2008) at this age. The pathological hallmark of this stage is the large scale activation and invasion of microglia into the sub-retinal space in order to remove the cellular debris of dead and dying photoreceptors (Strettoi and Pignatelli, 2000). Müller cells become hypertrophic as they extend their processes to form the glial seal which segregates the neural retina from the remaining photoreceptor layer, RPE and choroid (Strettoi et al., 2002, Strettoi et al., 2003). The RPE, essential for the recycling of the retinaldehyde chromophore, begins to invade the sub- retinal space and may additionally contribute to the increased stress of surviving photoreceptors (Jones and Marc, 2005, Marc et al., 2003).

Towards the end of phase 2, early stage retinal remodelling begins to occur in cells that had received direct input from the photoreceptors. Mirroring the temporal domain of photoreceptor degeneration, surviving rod bipolar cells demonstrate a reduction in the number of dendrites, whilst creating temporary synapses with horizontal cells and residual cone photoreceptors (Strettoi and Pignatelli, 2000, Peng et al., 2000). Rod bipolar cells continue internalising their respective glutamate receptors in response to the loss of tonic glutamatergic input from the outer retina. Their continued isolation can result in some cells undergoing cell death at this stage, shortly followed by the cone bipolar cell population (Strettoi and Pignatelli, 2000, Strettoi et al., 2003).

Chapter 1 General Introduction

Figure 1.16 Stages of retinal degeneration in the rd1 mouse – Phase II. The complete death of the rod photoreceptors has occurred and subsequently the cone photoreceptors (green) become stressed. They too now begin shedding their outer segments and begin to undergo cell death, but at a slower rate than the rod population. Isolated cone photoreceptors can survive at this stage. There is a large increase in microglia activation (grey cells) in the outer nuclear layer (ONL) as they remove cellular debris from the dead and dying photoreceptors. Muller cells (MC; dark green) form a glial seal, which acts as a barrier that isolates the sub- retinal space from the remaining neurones in the inner retina. Bipolar cells (BC; red) begin to show abnormal dendritic morphology in the OPL and some rod bipolar cells begin to show signs of cell death. The RPE begins to degenerate, and in some cases, invades the sub-retinal space. Diagram modified from (Soto & Kerschensteiner., 2015).

Phase 3

Phase 3 comprises global remodelling of the remnant neural retina (Figure 1.17). This occurs relatively late in the rd1 mouse compared to other models of retinal degeneration (Marc et al., 2003). It has been suggested that this is because small islands of cone photoreceptors can survive long after the death of the rod photoreceptor population (Ogilvie et al., 1997, Lin et al., 2009) and therefore provides a focal point of neurite growth and remodelling (Jones et al., 2003). The loss of

Chapter 1 General Introduction

excitatory synaptic inputs from the outer retina results in the isolation of remnant cell types in the inner retina, which now consist of surviving bipolar cells, amacrine cells and retinal ganglion cells (Strettoi et al., 2002, Strettoi et al., 2003). All of these cell types can send neurite growths towards the glial seal in the outer plexiform layer in an attempt to seek out surviving excitatory inputs (Jones and Marc, 2005). In doing so, they create incorrect synaptic contacts with other surviving inner retinal neurones. The retinal ganglion cell population are the most resilient of these remaining cells (Margolis et al., 2008, Lin and Peng, 2013) but can begin to show abnormal stratification in latter stages of the disease (Anderson et al., 2016). Regardless of cell type, there is a progressive increase in the death of remaining retinal cells with time as many fail to find these connections.

Figure 1.17 Stages of retinal degeneration in the rd1 mouse – Phase III Complete loss of both rod and cone photoreceptors has occurred at this stage. Rod bipolar cells undergo significant cell death and cone bipolar cells begin to show abnormalities in their dendritic architecture. Surviving neurones in the inner retina, notably bipolar cells, amacrine cells and some retinal ganglion cells can be seen to send projections towards the glial seal and create ectopic synapses with other surviving neurones of the inner retina. There is continued and progressive cell death of all cell-types with age. Diagram modified from (Soto & Kerschensteiner., 2015).

Chapter 1 General Introduction

Electrophysiological properties of the rd1 retina These anatomical hallmarks of degeneration and remodelling are reflected in electrophysiological recordings from the rd1 retina. ERG recordings demonstrate that at no age can rod-driven light- responses be identified (Strettoi et al., 2003). Whilst photopic ERGs are detectable, the amplitude of the α wave, which represents the hyperpolarisation of the photoreceptor in response to light, is significantly reduced (Strettoi et al., 2003). The β wave, which corresponds to the graded depolarisation of the downstream bipolar cells, is also reduced in amplitude and shows significant temporal deficits compared to wildtype controls at P14. This could reflect both the reduced drive from the remaining cone photoreceptor population but also the internalisation of the post-synaptic glutamate receptors on bipolar cells. Furthermore, the β wave also shows an abnormally long duration when compared to age-matched controls (Strettoi et al., 2003) suggesting problems with cellular metabolism and internal control of cellular signalling. Light responses fade rapidly in these animals and cannot be observed using ERG recordings by P16 (Strettoi et al., 2003) or MEA recordings by P28 (Stasheff et al., 2011). This correlates with anatomical data which demonstrates a lack of detectable nuclei in the outer nuclear layer of the rd1 retina by approximately four weeks of age (Carterdawson et al., 1978).

Two major electrophysiological hallmarks can be identified in the remaining retinal circuitry of the rd1 mouse: hyperactivity and rhythmicity of baseline firing. In a juxtaposing response to the loss of sensory input, the RGCs in the rd1 retina become hyper-excitable (Ivanova et al., 2015, Stasheff et al., 2011). The origin of this increase in baseline firing rate is currently unknown but is believed to reflect the re-organisation of the remaining inner retinal neurones as they seek out new synaptic contacts (Goo et al., 2016). Secondly, the rd1 retina demonstrates a robust 10Hz oscillation in firing rate which increases with age (Menzler and Zeck, 2011, Margolis et al., 2008). It has been proposed that the loss of outer photoreceptor input results in the AII Amacrine cell population being tonically hyperpolarised at rest (Borowska et al., 2011, Trenholm et al., 2012). These oscillations are believed to be spread via gap junctions through the AII-ON bipolar cell network which results in correlated burst firing across heterologous populations of RGCs (Choi et al., 2014, Menzler and Zeck, 2011, Ivanova et al., 2015, Barrett et al., 2015). This pathophysiological state of the rd1 retina is an increasing area of research when attempting to restore light perception (Barrett et al., 2015). Recently, it has been found that these oscillations can be dampened pharmacologically by application of gap junction blockers such as Meclofenamic acid (MFA) (Barrett et al., 2015) and Flupirtine (Toychiev et al., 2013).

Electrophysiological properties of image-forming pathways in rd1 mice Many of the anatomical and electrophysiological studies from rd1 mice and other retinally degenerate models have focused on the retina, as this is the primary source of degeneration. Few

Chapter 1 General Introduction

electrophysiological studies have investigated downstream targets of retinal ganglion cells in order to identify the effects degeneration has on this region. Early tracer studies have shown there is no gross differences in the projection from retina to superior colliculus or striate cortex in the rd1 mouse (Drager and Hubel, 1978). Electrophysiological recordings from these image-forming brain regions revealed that cells show the same hyper-excitability and rhythmic burst firing that is observed in the retina (Drager and Hubel, 1978, Ivanova et al., 2015). Spatial receptive fields have also been recorded from the tectum of rd1 mice at P18 and were found to possess an average receptive field diameter of 11.5° with a normal retinotopic organisation (Drager and Hubel, 1978). However, by P24 no receptive fields could be recorded from the central 100° of the visual field. Although weak receptive field responses could still be recorded which originated from the peripheral retina, they gradually faded up to P132. This decay likely reflects the progressive loss of the cone photoreceptor population with age.

1.5 The melanopsin system in retinal degeneration

Whilst the loss of rod and cone photoreceptors abolishes the conventional visual pathway in advanced retinal degeneration; rd1 mouse models retain an impaired ability to synchronise their body clocks to the light-dark cycles (Freedman et al., 1999) and exhibit a compromised yet functional PLR at high irradiances (Lucas et al., 2001a). This retained sensitivity to light is mediated by surviving ipRGCs in the inner retina. Whilst melanopsin driven light-responses to full-field light-steps can be recorded from the retina (Davis et al., 2015) and the dLGN (Brown et al., 2010) of rd/rd cl mice, there has been little direct investigation into the anatomical or physiological effects retinal degeneration has on this cell type and its projection to the dLGN.

Regardless of the early onset of outer photoreceptor degeneration in the rd1 mouse, the ipRGC population demonstrates normal retinal distribution and stratification patterns during development (Ruggiero et al., 2009). However, the number of ipRGCs in the mature visual system is higher in the rd1 mouse indicating the outer retinal photoreceptors may be important regulators of programmed cell death during visual system development. Further investigations into the integrity of the ipRGC population in the aged rd/rd cl retina identified that there is a 40% decrease in the ipRGC population of mice between with increasing age (Semo et al., 2003). This was not significantly different to age matched controls and therefore suggests there is a natural decline in the ipRGC population with age and that the loss of outer retinal photoreceptors does not exacerbate this effect.

Following this, a paper by Vugler et al. 2008 quantified ipRGCs, their dendritic processes and how these change during retinal degeneration in the RCS dystrophic rat. They investigated two time points: 15 weeks and 14 months after birth. They found no significant difference between dystrophic and non-dystrophic rodents at 15 weeks, however there was a ≈ 22% loss of the ipRGC population in both, supporting the observations of Semo et al., 2003 (Vugler et al., 2008). However, at 14 months

Chapter 1 General Introduction

of age, dystrophic rodents exhibited distinct molecular and anatomical abnormalities compared to those at 15 weeks, suggesting remodelling of at least some of the remaining ipRGC population. The total ipRGC population had decreased by ≈ 34%, signifying further ipRGC death with age. This decline in the ipRGC population was contrasted by an increase in the stability of their dendritic processes. These abnormal ipRGCs expressed elevated levels of Map1b (a protein associated with neuronal development and regeneration) and reduced levels of α-acetylated tubulin (a protein associated with neuronal stabilisation) (Vugler et al., 2008). The changes in concentration of these two biochemical markers are supported by the observed aberrant neurite growths which can occur as a form of retinal remodelling in other RGC classes during advanced stages of degeneration (Jones and Marc, 2005). Moreover, the axons of these surviving neurones exhibit irregular deflections into the inner retina. Whilst this data is supported by studies in other rodent models (Esquiva et al., 2013), investigations in the rd1 mouse suggest that the ipRGC population survives retinal degeneration with broadly normal retinal anatomy (Lin and Peng, 2013).

These anatomical studies have relied on use of the melanopsin antibody UF006 (Lin and Peng, 2013, Vugler et al., 2008). Since the discovery of the M3, M4 and M5-subtypes in visually intact mice, it has become apparent that labelling from this antibody is restricted to the M1 and M2 subtypes only (Ecker et al., 2010, Estevez et al., 2012). Accordingly, the survival and integrity of the more recently identified ipRGC subtypes is unknown in the degenerate retina. Electrophysiological investigations have recorded melanopsin driven light-responses in the dLGN of rd1 mice to uniform light steps (Brown et al., 2010) indicating that at least some of these more recently identified subtypes may survive retinal degeneration. In agreement with this, behavioural experiments in these animals have additionally demonstrated that they could exhibit crude brightness perception (Brown et al., 2012). These data are supported by a behavioural study from a human subject who suffered from severe outer retinal degeneration but demonstrated the retention of crude visual awareness, such that he could readily distinguish between short and long wavelengths of light (Zaidi et al., 2007). An action spectra which used the PLR as a reporter found that the peak sensitivity of this response occurred at 480nm, matching the known peak sensitivity of the melanopsin photopigment.

Chapter 1 General Introduction

1.6 Aims and Objectives

Over the last decade, it has become evident that the retina is a more complex tissue than previously appreciated. In addition to the rod and cone system providing light information for image forming and non-image forming responses to light, the retina employs a small subset of RGCs that express the photopigment melanopsin. These cells are intrinsically photosensitive, but also integrate light information from rod and cone photoreceptors for a variety of physiological and behavioural functions. The original stance that ipRGCs function solely as irradiance detectors for non-image forming homeostatic responses to light is consequently under review. The use of transgenic mice, most notably rd/rd cl, Opn4-/-, Opn4tau/LacZ and Opn4Cre mice, has facilitated the increasing appreciation of the retinal wiring, electrophysiological properties, projections and functions of the five distinct ipRGC subtypes, and have helped demonstrate that rods, cones and melanopsin form a complementary and synergistic photoreceptive system. The now five distinct subtypes of ipRGC are capable of modulating and integrating photic information with evidence supporting an extended role of the melanopsin system in contributing to visual perception.

During retinal degeneration, the retina undergoes a complex but predictable pattern of photoreceptor death and neuronal re-organisation. Whilst the rod and cone photoreceptors are extremely susceptible to this, the ipRGC population appear comparatively resilient to these effects. This indicates that ipRGCs could support residual visual function in patients afflicted with inherited retinal degenerations. However, the anatomical and functional integrity of those ipRGCs which project to the dLGN are unknown. Whilst melanopsin driven light-responses can be recorded from this brain structure, and degenerate mice demonstrate rudimentary brightness discrimination, a functional projection from retina to dLGN has never been directly investigated. If anatomically intact, what is the quality of visual information that ipRGCs can convey to the image forming visual system and what prevents them from supporting visual perception? This thesis aims to investigate the anatomy and physiology of the melanopsin projection from retina to dLGN during retinal degeneration in order to answer these outstanding questions from the literature. This thesis is presented in alternative format where Chapters 3-5 are presented in the form of published papers. Chapter 4 has been published in the Journal of Neurophysiology.

In Chapter 3, I set out to investigate whether all known ipRGC subtypes survive retinal degeneration in the rd1 retina. I take an anatomical approach to answer this question using the established multi- colour labelling technique termed Brainbow. In order to analyse this data, I need to design and describe a new approach to multi-colour analysis of 3D images which identifies, isolates and reconstructs individual cells from a population based on their unique colour identity. The 3D reconstruction of individual cells allows for the quantitative characterisation of their dendritic architecture to determine whether the ipRGC population is effected by neuronal remodelling following outer retinal degeneration.

Chapter 1 General Introduction

In Chapter 4, I set out to investigate whether the projection from retina to dLGN is anatomically and functionally intact at advanced stages of retinal degeneration, where any residual vision would have to rely on the intrinsic melanopsin light response. Using in-vitro and in-vivo electrophysiology, in combination with anatomical techniques, I look to investigate the spatio-temporal resolution of melanopsin-driven responses in the retina and dLGN in order to determine whether these cells could support some form of spatial vision in advanced stages of retinal degeneration.

Finally, in Chapter 5 I investigate earlier stages of retinal degeneration where visual function can rely on both residual cone photoreceptors as well as melanopsin signalling, and is more reflective of progressive stages of inherited retinal degenerations found in humans. Accordingly, I generate a new transgenic mouse model of retinal degeneration which allows me to use photoreceptor silent substitution in combination with in-vivo electrophysiological recordings from the dLGN. This technique allows me to tease apart the contributions of individual photoreceptors to residual light responses in the dLGN of mice with progressive retinal degeneration.

Chapter 2 General Methods

Chapter 2: General Methods

Chapter 2 General Methods

In order to collect the data demonstrated in this thesis, a number of different transgenic animal models were used. Some of these are already established models of retinal degeneration (Chang et al., 2002), whilst others were specifically generated for this thesis by crossing these retinally degenerate mice with other colonies of established mice in the Lucas laboratory. In addition, a number of different techniques including in-vitro and in-vivo electrophysiology, anatomy and imaging were used in order to answer the questions raised in this thesis. Descriptions of these new transgenic mouse models, and an explanation of the major recording methodologies and analysis techniques are described below.

2.1 Transgenic mouse models of retinal degeneration Animal Housing All animals used in this thesis were generated and bred at the University of Manchester with the help of Jonathon Wynne, and cared for by staff in the Biological Services Facility (BSF). Mice were housed under a 12:12 light/dark cycle, at a temperature of 22°C with food and water available ad libitum. All scientific procedures conformed to the requirements of the United Kingdom Animals (Scientific Procedures) act 1986, and were approved by the Home Office under the personal license 40/9905 and the project license 70/8918.

The C3H rd/rd cl mouse In chapter 4 I utilise the already established C3H rd/rd cl mouse - a model of advanced retinal degeneration. This mouse possesses a naturally occurring nonsense mutation on exon 7 of the PDEβ6 gene. This gene codes for the β-subunit of the cGMP phosphodiesterase which is expressed in the outer segments of rod cells (Bowes et al., 1990). The effect of this mutation is such that cGMP cannot be broken down and begins to accumulate in rod cells. As this occurs it becomes toxic and subsequently results in rod cell death. This is a rapid form of retinal degeneration such that by post- natal day 17 the rod cell population has undergone 95% cell death and by post-natal day 30 very few nuclei are detectable (Carterdawson et al., 1978). As described previously, cone cells naturally begin to degenerate as a result of rod photoreceptor death. However, to increase the efficiency of this cone death, these photoreceptors are additionally targeted by the cl transgene that drives expression of the diphtheria toxin alpha subunit (Lucas et al., 1999, Freedman et al., 1999). This specifically targets cone photoreceptors for cytoxic lesion. This aggressive model of retinal degeneration develops a normal visual system and projections; however rods and cones both degenerate rapidly post-natally such that by 80 days, ipRGCs are the sole surviving photoreceptive cells in the retina.

Chapter 2 General Methods

Generating new transgenic mouse models of retinal degeneration An important technical aspect which arose during the work on the C3H rd/rd cl mice was the increasing awareness that these mice are notoriously poor breeders. This resulted in slow throughput of experimental animals and influenced the decision on which mice to use for subsequent experiments. I therefore turned my attention to the C57 rd1 mouse which is commercially available from Jackson laboratories (C57BL/6J-Pde6brd1-2J/J; Stock #: 004766). This mouse possesses a chemically inducible mutation on exon 16 of the PDEβ6 gene. This mutation results in a similar phenotype as the C3H rd1 mouse, the toxic accumulation of cGMP resulting in rapid rod, and subsequent cone, degeneration (Stasheff et al., 2011). The transgenic mice used in Chapters 3 and 5 were derived from crossing this C57 rd1 mouse with other transgenic strains already established in the Lucas laboratory.

The C57 rd/rd OPN4Cre/+ mouse In Chapter 3 I take an anatomical approach to investigate the structure and architecture of ipRGCs in the retina during retinal degeneration. I specifically label ipRGCs in the retina using the established multi-colour labelling technique termed Brainbow (Cai et al., 2013). In order to achieve this, I created a new double transgenic mouse model, the C57 rd/rd OPN4Cre/+ mouse. This mouse is homozygous for the rd1 mutation and was crossed with a colony of OPN4Cre/+ mice, which were kindly donated by Sammer Hatter (John Hopkins University) to the Lucas laboratory and first described by (Ecker et al., 2010). In our mouse model, one of the alleles encodes the gene for mouse melanopsin, whilst the other allele encodes the bacterial enzyme Cre-recombinase (Figure 2.1). Therefore, in a heterozygote animal, cells which express melanopsin (ipRGCs) will also express Cre- recombinase. Importantly, the Opn4 gene is haplosufficient and therefore only one allele encoding mouse melanopsin is required to produce normal physiological function (Lucas et al., 2003b). An advantage of this generation method was that these degenerate animals were derived directly from their wildtype controls, thus minimising any phenotypic differences observed in anatomy that is not due to the degenerative state. I specifically chose use of heterozygous Opn4Cre/+ mice for experimental purposes as it has been previously shown that ipRGCs play an important role in the targeting and pruning of RGC projections during visual system development (Renna et al., 2011).

Chapter 2 General Methods

Figure 2.1 Gene construct of the Opn4Cre/+ mouse Heterozygous Opn4Cre/+ mice possess one allele encoding the melanopsin protein (Opn4; blue) downstream of the melanopsin promotor (grey). The second allele encodes the gene for the Cre-recombinase enzyme (orange) in addition to other regulatory elements (β globulin intron, NLS, pgk-neoR) downstream of the melanopsin promotor. Thus, ipRGCs which express functional melanopsin will additionally express the enzyme Cre-recombinase. Image adapted from (Ecker et al., 2010)

Cre-recombinase is a bacterial enzyme that was originally discovered in the P1 Bacteriophage (Hoess et al., 1984, Sternberg and Hamilton, 1981). Its function is to catalyse the recombination of DNA when a gene of interest is flanked by two LoxP (Locus of Crossing over P1) sites. The Cre-loxP recombination system is readily used in many fields of biology, particularly in neuroscience, in order to achieve targeted cell-specific gene expression (Sauer and Henderson, 1988, Kos, 2004).

When a gene of interest flanked by two LoxP sites it is said to be ‘floxed’. A single LoxP site is a 34 base pair (bp) sequence that consists of an 8bp core region flanked by two 13 bp inverted repeat sequences (left element and right element; Figure 2.2). In the presence of a pair of loxP sites, Cre- recombinase forms a complex tetrameric structure where each monomer binds to one of the inverted repeat sequences on each LoxP sites (Guo et al., 1997). The relative orientation of the core sequences between the pair of loxP sites, which is usually denoted by the direction of the triangle in schematic diagrams, defines the recombination event which can take place. The possible recombination events include deletion, insertion, inversion and translocation of DNA sequences (Sauer, 1998). For the purposes of this thesis I will describe two such events which provide a background to our use of the Cre-loxP system: deletion and inversion.

Chapter 2 General Methods

Figure 2.2 Sequence structure of an individual loxP site. A single loxP site is a 34 a base pair sequence that consist of an 8 base pair core region flanked by two 13 base pair inverted repeat sequences. These inverted repeat sequences are termed the left element (5’ end) and the right element (3’ end). The left and right element can each bind a single monomer of Cre-recombinase. The relative orientation of the core region of two loxP sites flanking the gene of interest defines the recombination event that will take place.

If one wants controlled deletion of a gene of interest from the genome, it is designed to be floxed by a pair of loxP sites which face in the same parallel direction. Cre-recombinase will recognise the orientation of the pair of core sequences, excise the gene of interest and delete it from the genome, together with one loxP site (Figure 2.3; Deletion). Gene inversion is often used in order to specifically switch on or switch off a gene of interest. In our case, the gene of interest is in the anti-sense direction (3’-5’) and floxed by a pair of loxP sites which face towards each other. When in this anti- sense direction, the cells transcriptional machinery is unable to transcribe the gene of interest and therefore it is effectively silenced. However, in the presence of Cre-recombinase, the gene of interest is excised, inverted and re-integrated into the genome and is now oriented in the sense (5’- 3’) direction. Here, the cells transcriptional machinery can now begin transcription and translation of the gene of interest which will result in expression of the protein product (Figure 2.3; Inversion).

Chapter 2 General Methods

Figure 2.3 Action of Cre-LoxP recombination Deletion: A gene of interest (blue) is floxed by a pair of loxP sites in a parallel direction (green triangles). Cre-recombinase (orange) binds to the inverted repeat sequences on each pair and recognises the parallel direction of the core sequence resulting excision of the gene from the genome whilst leaving a single loxP site. Inversion: A gene of interest (blue) is encoded in the inverse (3’-5’) direction downstream of its promotor. Due to this orientation, the cells transcriptional machinery (purple) is unable to transcribe or translate the protein product of this gene. The gene is floxed by a pair of LoxP sites orientated in an anti-parallel direction (green triangles). Thus, in the presence of Cre-recombinase (orange) the gene of interest is excised and inverted into the (5’ - 3’) direction. The cells transcriptional machinery now recognises the start codon on the gene of interest and can begin transcription and translation of the protein product. Image based on data from (Guo et al., 1997).

The C57 rd/rd Opn1mwR In Chapter 5 I create the C57 rd/rd Opn1mwR mouse. Here, I crossed our commercially available C57 rd1 mouse with an already established colony of Opn1mwR mice from the Lucas Laboratory. The Opn1mwR mouse expresses the human red cone opsin (L-Cone opsin) in place of the naturally occurring medium wavelength sensitive cone opsin (M-Cone opsin) (Smallwood et al., 2003). The effect of this is such that M-cone opsin expressing photoreceptors demonstrate a shift in spectral sensitivity from 508nm to 556nm (λmax = 556nm; red light), a change of approximately 50nm (Figure 2.4). As this mouse is homozygous for the rd1 mutation, there is a rapid death of the rod photoreceptors and the subsequent death of the cone photoreceptor population. However, unlike the C3H rd/rd cl model, these mice do not express the cl transgene and therefore cone photoreceptors demonstrate a slower rate of degeneration.

Figure 2.4 Spectral sensitivity of photopigments in the rd/rd Opn1mwR mouse. (A) Wildtype mice express S- cone opsin (λmax = 390; purple), Melanopsin (λmax = 480nm; blue), Rod opsin (λmax = 498nm; black) and M-cone R opsins (λmax = 508nm; green). (B) The Opn1mw mouse possess the same photoreceptors as the wildtype retina however the M-cone opsin (λmax = 508nm; green) has been replaced by the human L-cone opsin (λmax = 556nm; red) shifting the spectral sensitivity of these photoreceptors towards longer wavelength light. (C) rd/rd Opn1mwR mice only express three functional photoreceptors in the retina: S-cones, L-Cones and melanopsin. 1 Rod photoreceptors are rendered functionless from birth due to the rd mutation and rapidly degenerate. λmax values are taken from (Fu & Yau, 2007), (Lucas et al., 2001a) and (Smallwood et al., 2003) and applied to the visual pigment template described by Govardovskii et al., (Govardovskii et al., 2000).

Chapter 2 General Methods

2.2 Presentation of Visual Stimuli

In chapter 4 and 5 I provide carefully calibrated visual stimuli to anaesthetised mice and record these responses to light using in-vivo electrophysiology. In Chapter 4, I use the C3H rd/rd cl mice which only expresses melanopsin as a functional photopigment at the ages employed in these experiments. Therefore, whilst the irradiance measurement of the stimuli was of vital importance in order to determine the effective Michaelson contrast for the melanopsin photopigment, I did not have to account for the contribution of the other murine photoreceptors to the light response in these animals. In Chapter 5, the rd1 Opn1mwR mouse possesses three functional photopigments: S- cone opsin, L-cone opsin and melanopsin. Therefore, I could design carefully calibrated stimuli which allow me to control for the contribution of each photoreceptor to the light response using silent substitution (Estevez and Spekreijse, 1982). The introduction of the L-Cone opsin in these mice allows me to maximise the effective contrast for each functionally expressed photopigment. A brief description of calibration and the silent substitution method is described below.

Calibrating visual stimuli Calibration of visual stimuli was achieved by recording the spectral power densities for the Television in Chapter 4, and each of the four LEDs and the laser input for the projector system in Chapter 5. All stimuli were measured at the corneal plane at the approximate distance of the mouse eye from the light-source during the experiment using a calibrated spectroradiometer (Bentham instruments Ltd., UK) in 1nm increments between 300-800nm. The effective photon flux for each photopigment was calculated by weighting the measured spectral irradiance with the absorption spectra for each photopigment as estimated from the pigment spectral efficiency function (Govardovskii et al., 2000). For in-vivo measurements, we additionally accounted for spectral lens transmission (Jacobs et al., 2007).

Silent substitution The generation of the C57 rd/rd Opn1mwR model in Chapter 5 is with the aim of being able to tease apart the contribution of different photopigments to visual responses in the dLGN in-vivo. To achieve this, I adopt a method from psychophysics termed ‘photoreceptor silent substitution’, (Estevez and Spekreijse, 1982) where one can selectively modulate the output of certain photoreceptor subclasses whilst keeping others constant (silent) (Allen and Lucas, 2016, Allen et al., 2014). The principle of silent substitution is possible due to the response properties of individual photopigments to light. Opsins respond to light as a function (f) of wavelength (λ) and irradiance (I). Therefore, different combinations of wavelength and intensity can produce the same

Chapter 2 General Methods

electrophysiological output for an individual photoreceptor. These different spectra, which are indistinguishable for an individual photoreceptor class, are termed metamers.

Photoreceptor output = 푓(λ , I)

The transition between two metamers will only be silent for a particular photoreceptor class that expresses a particular opsin. The mouse retina expresses four spectrally distinct opsins and therefore this same spectral transition will provide a Michaelson contrast for the remaining photoreceptor subtypes. This would manifest itself as a graded change in membrane potential for those photoreceptors. Thus, the importance of using the red-cone knock-in mouse in these silent substitution experiments is most clearly highlighted when looking at the difference in spectral sensitivity of the murine photoreceptors and calculating the effective Michaelson contrast for each Photopigment (Figure 2.4). Wildtype mice express S-cone opsins, melanopsin, rod opsin and M-cone opsins. With the exemption of the S-cone opsin, these photopigments demonstrate peak sensitivity to a very narrow range of the visible spectrum (approximately 28nm) which limits the magnitude of the calculated Michaelson contrast one can deliver to a particular photoreceptor, whilst being isoluminant for others. Shifting the spectral sensitivity of the M-cones to longer wavelength light allows for a larger contrast to be generated. This is especially important for the intrinsic melanopsin light-response which is known to demonstrate low sensitivity and small amplitude responses in response to light steps from background (Davis et al., 2015, Do et al., 2009, Berson et al., 2002).

2.3 Data collection and Recording methodologies Electrophysiological Recordings The data collected in Chapters 4 and 5 primarily consist of in-vivo extracellular multichannel electrophysiological recordings from ipRGC-recipient nuclei in anaesthetised C3H rd/rd cl and C57 rd/rd Opn1mwR mice. Multichannel extracellular electrophysiological recordings allow for simultaneous live recording of spiking activity from a large population of neurones from outside the cell. One can therefore record from a large area of a target nucleus, without compromising the cells integrity, therefore maximising the duration of the experiment and the number of cells recorded. A detailed description of the recording methodology is described in Chapter 4 and 5 however a brief explanation of the principle of the electrophysiological technique is outlined below.

Chapter 2 General Methods

2.3.1.1 The action potential The action potential is the means by which neurones communicate. The initiation of an action potential can occur via a number of different routes. Sensory neurones, such as rods and cones, directly respond to environmental stimuli due to the expression of sensory receptors (opsins) that respond to changes in light. The neurones recorded in this thesis receive direct communication from other neurones and can do so via electrical or chemical synapses. At an electrical synapse, two neurones become directly connected by the expression of connexins across their plasma membrane. This allows for the direct flow of ions between cells and the propagation of the action potential (Bloomfield and Volgyi, 2009). At a chemical synapse, neurotransmitters are released in vesicles from the pre-synaptic neurones terminal and bind to specific receptors on the post-synaptic neurone. Depending on the neurotransmitter released and the receptors expressed, different changes in membrane potential can occur due to downstream signalling cascades. A brief description of the action potential is described below to provide a background to the recording methodologies and data analysis.

2.3.1.2 Initiation A neurones resting membrane potential normally ranges between -60 and -80mV. This negative membrane potential is primarily driven by the Na+/K+ pump which exchanges three sodium ions out of the cell for every two potassium ions in. This occurs against their concentration gradients and thus uses energy from Adenosine Tri-phosphate (ATP). Many ion channels expressed in a neurones plasma membrane also possess selective permeability to potassium ions and as such there is a constant flow of potassium out of the cell down its electrochemical gradient providing a constant source of ions for the Na+/K+ pump to maintain this negative intracellular membrane potential.

When an excitatory neurotransmitter is released from the pre-synaptic terminal it binds to its respective post-synaptic receptor. This results in the initial opening of sodium channels and allows for sodium ions to travel down their electrochemical gradient into the cell. This causes a localised region of depolarisation in the cell which is termed the Excitatory Post Synaptic Potential (EPSP). If enough neurotransmitter is released from the pre-synaptic terminals, and enough EPSP’s arrive at the same time (temporal summation), the summative effect of these depolarising current increases the membrane potential of the postsynaptic cell allowing it to reach threshold and fire an action potential (Adapted from Purves et al., 2001).

Chapter 2 General Methods

Figure 2.5 The action potential. Schematic diagram illustrating the change in membrane potential of a neurone over the course of an action potential. The neurone is initially at rest (-70mV; black dotted line). Release of neurotransmitter from the pre-synaptic neurone results in an increase in membrane potential until the cell reaches threshold (-55mV; Orange solid line) and triggers an action potential. This increase in membrane potential results in the opening of VGSC’s and the subsequent influx of sodium ions into the cell resulting in depolarisation of the cell membrane. Once at approximately +40mV, VGSCs close and VGPCs open allowing potassium ions to leave the cell and help repolarise the neurone. VGSCs remain inactive for a short period of time after closure and the neurone enters its refractory period preventing another action potential being generated within 1ms. Diagram adapted from (Purves et al., 2001).

2.3.1.3 Propagation & Termination Once a neurone has reached threshold it will fire an action potential and hence it is described as an “all or nothing” response (Figure 2.5). The localised increase in the cells membrane potential results in the opening of Voltage Gated Sodium Channels (VGSCs). These VGSC’s are found along the length of the cell and its axon and result in these localised depolarisations spreading passively down the length of the axon towards the synaptic terminal, which further raises the neurones membrane potential. Once the membrane potential of the neurone reaches approximately +40mV, the VGSCs close and enter an inactive state whereby they can no longer open, regardless of the membrane potential of the neurone. This is vital in ensuring the action potential is propagated in one direction towards the synaptic terminal. As the membrane potential begins to reduce due to the reduction of Sodium influx, Voltage-sensitive Potassium Channels open allowing Potassium efflux which helps the repolarisation of the neurone. During this process, there is a slight hyperpolarisation of the membrane potential below resting potential which is termed the refractory period. During this time the neurone is incapable of firing another action potential regardless of stimulus intensity. This whole process from initiation to termination takes approximately one millisecond. (Adapted from Purves et al., 2001)

Chapter 2 General Methods

Recording extracellular electrophysiological activity The changes in ion concentration that occur across the neurones membrane during an action potential are mirrored spatio-temporally by changes in the local extracellular ion concentrations. Accordingly, placement of an extracellular recording electrode into a brain nucleus of interest allows each silicon recording site (or channel) of a recording electrode to detect the activity of a population of neurones in close proximity. In this thesis, all data is recorded using a Neuronexus probe (A4X8-5 mm-50-200-413; Neuronexus, MI, USA) consisting of four shanks (spaced 200µm apart), each with eight recordings sites (spaced 50µm apart) (Figure 2.6). When recording from a large nucleus such as the dLGN, this allows for wide coverage across its medial-lateral extent when the electrode is inserted in the coronal plane.

Figure 2.6 Extracellular multi-channel recording electrode Image of a Neuronexus probe (A4X8-5mm-50-200- 413) used for multichannel extracellular in-vivo electrophysiological recordings and magnified schematic of the organisation of the silicon recording channels showing 4 shanks spaced 200µm apart. Each shank consists of 8 silicon recording sites (black circles) spaced 50µm apart, each with an area of 413µm2.

A Recorder64 recorder system (Plexon TX, USA) was used to acquire electrophysiological signals. The signal was amplified (x3000) and digitised at 40 kHz. This data was saved as time-stamped waveforms. The population activity of neurones that are recorded from an individual electrode site is termed multi-unit data. In order to resolve the response characteristics of a single cell (unit) over time, this data was spike sorted and will be described in further detail in the data analysis section in this Chapter.

Chapter 2 General Methods

Anatomy and Microscopy 2.3.3.1 Brainbow In chapter 3 I take an anatomical approach in order to label, isolate and re-create individual ipRGCs in the retina of rd1 mice. To achieve this I use our newly generated rd/rd Opn4Cre/+ mice in combination with the established technique of “Brainbow”. The Brainbow technique relies on Cre- LoxP recombination in order to stochastically express a combination of up to four fluorescent proteins providing each cell with a unique colour identity (Cai et al., 2013, Livet et al., 2007).

To achieve this, I deliver a commercially available floxed Brainbow virus (Vector Core, University of Pennsylvania, USA) via injections of two adeno-associated viral vectors (AAV9; Figure 2.7) into the vitreous cavity of rd/rd Opn4Cre/+ mice. Each virus encodes two fluorophores: the first virus encodes the blue fluorescent protein (BFP; peak emission = 456nm) and enhanced yellow fluorescent protein (eYFP; peak emission = 514nm) whilst the second virus encodes the mCherry (peak emission = 619nm) and mTeal (peak emission = 492nm) fluorophores. The genes for these fluorescent proteins are encoded in an antisense orientation (3’ - 5’) and are each flanked by two pairs of mutant loxP sites orientated in antiparallel directions (Cai et al., 2013).

Figure 2.7 Brainbow viral construct. AAV Brainbow constructs for two Brainbow viruses. Virus 1: TagBFP (dark blue arrow) and eYFP (yellow arrow); Virus 2: mCherry (red arrow) and mTFP (light blue arrow). Genes encoding these fluorophores are encoded in the anti-sense direction (denoted by direction of arrow; 3’-5’) between mutant loxP sites (light and dark green triangles; light green = mutant element, dark green = wild type element) orientated in antiparallel directions. Regulatory elements including EF1α (elongation factor 1α gene); W (woodchuck hepatitis virus post-transcriptional regulatory element); pA (polyadenylation sequence) are added to provide higher transduction efficiency and viral stability. 5’ITR and 3’ITR are the 5’ Intron and 3’ Intron, respectively. Diagram adapted from (Cai et al., 2013).

As described earlier, a single loxP site consists of a core region flanked by two inverted repeat sequences (left and right elements). The loxP sites used in this viral construct are mutated as one of the elements of an individual loxP site contains a mutation in the inverted repeat sequence. These mutated loxP sites are orientated in antiparallel directions which will result in inversion of the gene that they flank. However, these inversion events are constrained such that they can only occur between loxP sites where the right element of one loxP site is compatible with the left element from another loxP site.

Chapter 2 General Methods

During an inversion event, both right elements from a pair of loxP sites exchange positions on the construct before re-integrating into the genome. Thus, new recombinant loxP sites can be created after an inversion event: one loxP site with both elements mutated (double mutant), one with both elements containing the wildtype sequence (wildtype loxP) and one with either the right or left element containing a mutation (Figure 2.8; Inversion). The double mutant (light green) has a very low affinity for Cre-recombinase and therefore the inversion of the gene of interest immediately downstream of it is locked in place, preventing any further recombination events from occurring (Cai et al., 2013). However, these other newly created loxP sites in the construct can still serve as substrates for Cre-recombinase if the gene of interest lies between compatible pairs. Due to the initial design of the viral construct, compatible pairs now face in the same parallel direction and can therefore result in deletion of the second gene (Figure 2.8; Deletion).

Figure 2.8 Brainbow recombination using mutant loxP sites Schematic diagram of one Brainbow construct with mutant portions (light green) and wildtype portions (dark green) of loxP sites orientated towards each other and the possible recombination events. (1) One set of compatible loxP sites exist in an antiparallel direction flanking the mCherry gene allowing for gene inversion. (2) Newly created double mutant loxP site (light green triangle) now exists upstream of mCherry preventing further recombination events. A second newly created pair of compatible loxP sites exist between the gene encoding mTFP, and is orientated in a parallel direction resulting in deletion. (3) A second pair of compatible loxP sites exists in the original construct which allows for inversion of the whole sequence (both mCherry and mTeal). (4) This sequence inversion creates a new pair of loxP sites which flank the mCherry gene in a parallel direction resulting in deletion. Image adapted from (Cai et al., 2013).

Therefore, following intravitreal injections, ipRGCs which express Cre-recombinase, and are additionally transduced by the virus can induce random inversion and deletion events, and hence

Chapter 2 General Methods

stochastic expression of up to four fluorophores. The floxed mechanism results in high cellular specificity and a minimal chance of false positives because it relies on cell-specific expression of Cre- recombinase and the antisense orientation of a gene of interest that can only be transcribed once inverted by Cre-recombinase. The stochastic element of this expression is enhanced not only due to the random nature of Cre-recombinase activity at any one of the eight pairs of mutant loxP sites in each viral particle, but also the viral transduction efficiency in each cell and the transcriptional and translational ability of individual ipRGCs.

2.3.3.2 Immunohistochemistry The intrinsic fluorescence of many fluorophores is often weak and prone to bleaching under prolonged light-exposure. This is especially true for those with a peak spectral sensitivity to shorter wavelength light such as BFP and mTeal (Cai et al., 2013). I therefore perform immunofluorescence antibody staining on our samples in order to amplify the signal. Here, one uses a primary antibody which has been raised against the protein wishing to be labelled (the antigen) to which it demonstrates extremely high specificity. A secondary antibody, which is then conjugated to a more stable fluorophore, is then applied which exhibits high specificity to the primary antibody. A number of conjugated secondary antibodies can bind to an individual primary antibody resulting in signal amplification. Furthermore, these conjugated secondary antibodies are available in a variety of emission wavelengths and show high photo-stability. Therefore, in addition to signal amplification, they also provide a means by which to increase the distance between excitation and emission spectra of individual fluorophores. With regards to Brainbow, where up to four fluorophores may be imaged, this allows for peak emission spectra to be spread across the entire visible spectrum, resulting in reduced spectral cross talk during image acquisition. One can then image these cells on a fluorescent or confocal microscope, which effectively measures the relative concentration of each fluorophore and represents this as a value in each colour channel ranging from 0 - 65,535 in a 16-bit monochrome image (Figure 2.9).

Chapter 2 General Methods

Figure 2.9 Colour variation from stochastic Brainbow expression Seven possible colours can be produced due to the stochastic nature of the recombination events for three of the pseudocoloured fluorophores: White, Yellow, purple, red, cyan, green and blue. Each row describes the possible recombination events after immunohistochemical labelling for three fluorophores: mCherry (Alexa 546, red), eYFP (Alexa 448; green) and mTeal (Alexa 647; blue). Colour of arrow denotes the pseudo colour assigned to each fluorophore during image acquisition. Arrows facing right describe a successful recombination event where inversion and expression of the gene has occurred. Black arrows pointing left denote no recombination event or a deletion for that fluorophore. On the right of each row shows the possible colour outcome of the differential expression of fluorophores which underwent a successful recombination event. Note this possible colour outcome is underrepresented as it does not account for the relative contribution of each individual fluorophore to the spectral identity of the cell. Diagram adapted from (Cai et al., 2013).

2.3.3.3 Fluorescence microscopy A number of microscopes have been used to capture the fluorescent images presented in this thesis. The most common microscope used is a wide field Leica DM2500 microscope which uses a CoolLED pE300 LED light source filtered through a Chroma L5 ET (k) / Chroma Y3 ET (k) and a Chroma A4 ET (k) filter set and captures images using a Leica DFC365 FX camera. I additionally use a Leica stereo fluorescence microscope (M205 FA) using an apochromatic 20.5:1 zoom and captured using a Coolsnap ES camera (Photometrics). Specific band pass filter sets for FITC and Texas red were used to prevent bleed through from one channel to the next. Both these microscopes are excellent at

Chapter 2 General Methods

capturing fluorescence from a target region in a tissue of interest. However, a major limitation of wide field fluorescent microscopy is that the entire specimen is illuminated by the excitation source whilst only wishing to image a small portion of the tissue in question. This can result in rapid bleaching of the fluorophore over repeated exposures. Additionally, all emitted light is captured by the photodetector and camera even if it is from outside of the focal plane. These results in an increase in image blur with thicker tissue specimens. This is further exacerbated by light refracting as it passes into, and out of, the tissue. This reduces both the intensity of the incident light which can reach the fluorophore and the detectable fluorescence, with increasing tissue depth.

2.3.3.4 Confocal Microscopy One way to overcome these limitations of conventional wide field fluorescent microscopes is to use laser scanning confocal microscopy. In Chapter 5, I use a LEICA SP5 AOBS inverted tandem head confocal microscope as it provides a number of distinct advantages over conventional wide-field microscopy. Firstly, this system can use up to five individual lasers which possess discrete excitation wavelengths (Blue diode = 405nm; Argon = 458nm, 476nm, 488nm, 496nm and 514nm; Green HeNe = 543nm, Orange HeNe = 594nm and Red HeNe = 633nm) permitting narrow band-pass specificity to excite individual fluorophores. Fluorophores which overlap in their excitation or emission spectra can be imaged sequentially in order to minimise cross-talk and bleed through between channels providing high quality colour information. The use of lasers provides point illumination, where only a very small region of the specimen is excited at any one time. This greatly reduces the amount of fluorophore bleaching providing the laser power is used at their minimum requirement. Additionally, a pinhole is placed in front of the photodetector which prevents any fluorescence which is not emitted from the focal plane reaching the detector. The combination of these two permits the capture of extremely high resolution images with a superior signal to noise ratio and high specificity to individual fluorophores.

Another major advantage of confocal microscopy is the ability to change the depth of the focal plane in small increments allowing one to image up to 100µm deep into a tissue sample. This is further enhanced with the aid of a motorised stage which allows for tile scanning in XY co-ordinate space. Individual tiles can be imaged at high magnification and then be stitched together to form one large high resolution image from the region of interest. Using new and ever improving software one can then re-create 3-Dimensional images of labelled biological tissue with high spatial and colour information. The disadvantage of this technique is that this high spatial and colour resolution comes at the expense of the time it takes to image large areas of biological tissue and the size of the data collected for analysis.

Chapter 2 General Methods

2.4 Data Analysis

The data collected in this thesis relies on Principal Component Analysis (PCA) in order to analyse our colour data in Chapter 3 and our electrophysiological data in Chapter 4 and 5. PCA analysis is a well- established statistical technique in neuroscience to analyse electrophysiological data, and for this purpose we employ commercially available software (Offline Sorter, Plexon Inc, USA) to perform the analytics. In Chapter 3, however, we develop a novel approach to colour analysis using PCA on 3D multi-colour images (based around three pseudo colour channels - red, green and blue). A brief overview of the concept that lies behind Principal Component Analysis is therefore described below.

Large data sets with multiple variables pose significant analytical and computation problems when attempting to identify and quantify possible relationships between its various dimensions. PCA is a statistical approach frequently used in various disciplines to reduce a large number of dimensions to something essential and meaningful (Jolliffe, 2002). It achieves this by applying a linear transformation to the data to produce a new set of dimensions (principal components) which can be visualised on orthogonal axis and ranked by decreasing variance.

Analysing extracellular electrophysiological activity Isolating single units in electrophysiological recordings is typically achieved using PCA in software such as Offline Sorter (Plexon, TX, USA). Here, each spike is represented as a dot in principal component space (Figure 2.10A) and one can compare different parameters of the spike waveform which are unique to a specific cell’s firing pattern including the amplitude and shape of the waveform (Figure 2.10B). One can also compare the inter-spike interval (ISI) as individual units cannot produce more than one spike per millisecond due to the defined temporal kinetics of the action potential (Figure 2.10C). One can therefore cluster these spikes based on their similarities and differences. Each cluster of dots represents the spikes of a single unit over time. The time of each spike for individual units can then be exported to Neuroexplorer (Nex Technologies, MA, USA) and MATLAB (The Mathworks Inc., MA, USA), where custom written programmes from members of the Lucas laboratory are used to analyse the responses of these cells to a variety of different visual stimuli. These programs allow for the construction of Peristimulus time histograms (PSTH; Figure 2.10D) and analysis on the change in firing rates which are directly related to the time (event) at which the stimulus occurred.

Chapter 2 General Methods

Figure 2.10 Analysing electrophysiological data (A) Representative scatter plots of spike waveforms plotted in a 2D representation of principal component space (X axis = PC1; Y axis = PC2) shows two readily identifiable single units (Unit A = Yellow; Unit B = Green) from a single recording site. For clarity, multi-unit data is not plotted. (B) The mean ± standard deviation of the spike waveforms of Unit A and B are plotted to the right showing the difference in spike amplitude and shape of the two waveforms. We use Manova F and P-values as an indicator of the degree of cluster separation (High F values and low P values demonstrate well isolated units). (C) Log Inter spike Interval histograms of Unit A (yellow), Unit B (green) and remaining multi-unit data (Grey). Isolated single units show clear sharp peaks at discrete intervals greater than 1ms, indicative of spikes fired in bursts. Multi-unit waveforms which show activity less than 1ms indicate noise in the recording (% indicates percentage of total spikes which occur less than 1ms apart). (D) Average Peristimulus time histograms (PSTH) show the light-evoked responses of the single units isolated in (A) and the comparative multi-unit data (white bar indicates duration of light-pulse). Unit A demonstrates a sustained-ON response and Unit B a sustained-OFF response when presented with 20 repeats of a 10s full-field light step from darkness and a 50s ISI (Irradiance =2.50x1014 photons cm-2 s-1). Remaining multi-unit data shows the response of a population of cells which cannot be isolated but still demonstrate a Transient ON-OFF response to the light-step. (Note differences in baseline firing and peak firing rates between single and multi-unit data).

PCA analysis of colour In chapter 3 we design a custom written analysis software which uses PCA analysis to analyse 3D multi-colour images. Here, unlike electrophysiological data where the PCA analysis describes the relationship between variables of the spike waveform, we use it to describe the relationship of the contribution of each colour channel (Red, Green and Blue) to individual voxels in a 3 Dimensional colour image. Based on the Brainbow technique, each transduced cell should be represented by a unique colour relationship and therefore be more readily identifiable after PCA analysis. A more in depth description and demonstration of PCA analysis of colour is described in Chapter 3.

Chapter 3 Identifying ipRGCs in the degenerate retina

Chapter 3: Identification and anatomical characterisation of all known ipRGC subtypes in the degenerate retina using a novel multi-colour analysis method

Chapter 3 Identifying ipRGCs in the degenerate retina

3.1 Abstract

Following outer photoreceptor degeneration, the death of the rod and cone photoreceptors results in global re-modelling of the remnant neural retina. Intrinsically photosensitive retinal ganglion cells (ipRGCs) survive retinal degeneration even at advanced stages of the disease and act as the only method by which light information can reach the brain. These cells support a variety of non-image forming physiological and behavioural responses to light including circadian photo-entrainment and the pupillary light reflect (PLR) even in the absence of rod and cone photoreceptors. This diversity in function is not only reflected in their numerous projection patterns to different brain nuclei, but also in their retinal anatomy. Accordingly, five ipRGC subtypes are in the mouse retina, of which at least one projects to the dorsal Lateral Geniculate Nucleus (dLGN) and therefore could provide a viable route through which light information can still reach visual processing structures in order to support vision. While 2 of the ipRGC sub-types have been previously shown to survive structurally intact in advanced retinal degeneration, the anatomy of the other 3 types (which includes those innervating the dLGN) in such conditions remains unknown. We therefore set out describe the anatomy of all five subtypes of ipRGC in retina of rd/rd mice (a model of advanced retinal degeneration). We designed a novel analytical approach using principal component analysis (PCA) to aid in the isolation and extraction of voxels which correspond to singles cells from retinas treated with the multi-colour labelling technique Brainbow. We first validated this method using artificial images before continuing to demonstrate its utility by using it to reconstruct images of individual ipRGCs from the retina of rd/rd Opn4Cre/+ mice transfected with a floxed Brainbow virus. Following reconstruction of their 3D morphology we are able to compare the soma sizes and the extent, complexity and location within the inner-plexiform layer of dendritic fields with previous descriptions of ipRGC anatomy in the wildtype retina. We find ipRGCs whose characteristics match those reported for all 5 subtypes of ipRGC in the wildtype retina. This indicates that ipRGCs survive outer retinal degeneration intact, at least at the anatomical level.

Chapter 3 Identifying ipRGCs in the degenerate retina

3.2 Introduction

A small population of retinal ganglion cells are intrinsically photosensitive due to the expression of the protein melanopsin (Lucas et al., 2003b). In advanced retinal degeneration, loss of rod and cone photoreceptors renders melanopsin-expressing intrinsically photosensitive retinal ganglion cells (ipRGCs) as the only source of light information. The intrinsic light-response of ipRGCs can drive a variety of non-image forming responses such as circadian photo-entrainment (Lucas et al., 2001b) and the pupillary light reflex (Lucas et al., 2001a) by their projections to the Suprachiasmatic Nucleus (SCN) and the Olivary Pretectal Nucleus (OPN), respectively. More recently, these cells have also been found to project to a number of image-forming regions in visually intact mice including the dorsal LGN and superior colliculus (Ecker et al., 2010) where they have been found to support a number of roles in image forming vision including irradiance coding (Brown et al., 2010), contrast detection (Schmidt et al., 2014) and increasing the signal to noise ratio of fast visual responses (Storchi et al., 2015).

This diversity of projection targets and physiological roles driven by melanopsin signalling is reflected in the five distinct anatomical subtypes of ipRGCs identified in the mouse retina (M1-M5) (Schmidt et al., 2011a, Schmidt et al., 2011b, Ecker et al., 2010, Estevez et al., 2012). Previous data has demonstrated that the overall number of ipRGCs develop normally in the rd1 mouse, a model of retinal degeneration (Ruggiero et al., 2009), and that their total number is not affected by the loss of the outer retinal photoreceptors (Semo et al., 2003). Furthermore, following re-organisation of the neural retina in response to outer photoreceptor degeneration (Jones and Marc, 2005, Strettoi et al., 2003), at least two subtypes of ipRGC survive with broadly normal retinal anatomy (Vugler et al., 2008, Lin and Peng, 2013). However, these data have relied on immunohistochemical labelling of the melanopsin protein, which has since been found to only identify the M1 and M2 subtypes. The more recent discovery of the M3-M5 subtypes has been facilitated by the use of the OPN4Cre mouse line which demonstrates far superior sensitivity in detecting melanopsin expressing cells and their projections (Ecker et al., 2010, Brown et al., 2010). Therefore, an outstanding question from the literature is whether all ipRGC subtypes survive following outer retinal photoreceptor degeneration, and whether they retain normal retinal architecture following reorganisation of the remnant neural retina. This question has particular importance for any potential role of melanopsin in perceptual vision, as projections to the dLGN are dominated by M4 and M5 type ipRGCs.

Chapter 3 Identifying ipRGCs in the degenerate retina

One approach to answer this question is with the transgenic multicolour labelling technique termed Brainbow (Cai et al., 2013). Brainbow uses Cre-loxP recombination to label individual neurones with up to four spectrally distinct fluorophores (Cai et al., 2013, Livet et al., 2007). Hundreds of hues can be generated from the stochastic and combinatorial expression of these four fluorescent proteins, thus providing each labelled cell with unique spectral identity. This diversity of colour provides a powerful tool for studying network mapping and neural connections and has already shown great promise in a variety of non-neuronal and neuronal tissue types ranging from drosophila (Hampel et al., 2011, Boulina et al., 2013, Hadjieconomou et al., 2011), zebrafish (Pan et al., 2013, Gupta and Poss, 2012) and mice (Hammer et al., 2015). However, Brainbow has not reached its full potential as it is yet to yield quantifiable data regarding the morphology or connectivity of individual cells from a population of labelled neurones. This has been in part due to the lack of a robust and reliable colour analysis method in bio-imaging (Weissman and Pan, 2015). Current attempts to analyse Brainbow labelled tissue have predominantly relied on comparing the RGB ratio values from a subset of pixels from simple structured portions of a neuron such as axons or terminals (Hammer et al., 2015). However, this approach primarily relies on identifying the spatial continuity of the cell upon visual inspection and only uses the colour ratio to try and support this assignment. In addition, these data are analysed on a 2D projection of the image. Here, each pixel is an average of the RGB value of all voxels in the Z stack which can result in great variation in the unique colour identity of a labelled cell on a pixel by pixel basis. There have been no published attempts to quantify the number of Brainbow labelled cells in a population and none have looked to isolate those pixels that correspond to a single cell in order to reconstruct their structural morphology in 3D using the stochastic nature of Brainbow expression.

Here, we address this methodological deficit by designing a new approach based upon principal component analysis (PCA) of 3 Dimensional multi-colour images. We characterise the performance of this analysis method using artificial images and continue to demonstrate its utility on images captured from retinas of rd/rd OPN4Cre/+ mice following intraocular injection of a floxed Brainbow virus. In doing so, we isolate and reconstruct the 3 Dimensional morphology of all known subtypes of ipRGC in the degenerate retina and provide a comprehensive analysis of these cells’ dendritic architecture. We find that these ipRGCs from the degenerate retina match the dendritic morphology reported from the wildtype retina, indicating that outer retinal degeneration does not have a significant impact on the anatomy of these cell types within the retina.

Chapter 3 Identifying ipRGCs in the degenerate retina

3.3 Methods Animal housing Mice were bred at the University of Manchester and housed under a 12:12 light/dark cycle, with food and water available ad libitum. rd/rd Opn4Cre/+ were created in house by crossing an established colony of Opn4Cre/+ mice with commercially available C57 rd1 mice (Stock #: 000659; Jackson Laboratories) - see chapter 2 for a more in-depth description of mouse generation. All mice used in these experiments were greater than 4 months old at the time of injection. All procedures conformed to requirements of the UK Animals (Scientific Procedures) Act, 1986.

Intravitreal injections Five rd/rd OPN4Cre/+ mice (>4 months old) were anaesthetised with an intraperitoneal injection of Ketamine (100mg/kg) and Xylazine (10mg/kg). Once anaesthetised, one drop of 1% Tropicamide (Sigma Aldrich, UK) and 2.5% Phenylephrine (Sigma Aldrich, UK) was topically applied to the left eye in order to fully dilate the pupil. A fine Hamilton needle (35G; 8mm) fitted to a 5µl Hamilton glass syringe was then passed through the equator of the sclera and into the vitreous cavity whilst being careful to avoid the lens. The injection consisted of 1.0µl of floxed Brainbow virus (AAV9.hEF1a.lox.TagBFP.lox.eYFP.lox.WPRE.hGH-InvBYF & AV9.hEF1a.lox.mCherry.lox.mTFP1.lox.WPRE.hGH-InvCheTF; 1013 genomic particles/ml; Penn Vector Core, USA) with a 0.5µl mixture of Heparinase III (250 units/ml; Sigma Aldrich, UK) and Hyaluronan lyase (250 units/ml; Sigma Aldrich, UK) (Cehajic-Kapetanovic et al., 2011). In one mouse, we injected 2µl of each virus to ensure maximum transduction efficiency for anatomical tracing. The injection was conducted slowly over the course of one minute to prevent a rapid increase in intra-ocular pressure and subsequent leaking of the virus. Once complete, the needle was gently removed and the eye checked to ensure there was no damage to either the retina or the lens. A topical analgesic (0.25% Bupivacaine; Sigma Aldrich, UK) was applied to the injected eye. Mice were allowed to recover on a homeothermic heat mat (Harvard Apparatus, Kent, UK) before being returned to the colony rooms in the BSF in a fresh cage with mash and water available at floor level. Mice were checked for subsequent days to ensure there no signs of infection or inflammation in the injected eye.

Chapter 3 Identifying ipRGCs in the degenerate retina

Tissue preparation After a 4-6 week incubation period, mice were overdosed with an intraperitoneal injection of Urethane (30% w/v; Sigma Aldrich, UK) and transcardially perfused with 0.1M PBS followed by 4% methanol-free Paraformaldehyde (4% mf-PFA; Sigma Aldrich, UK). Immediately following perfusion, mice were enucleated and the brains were removed and both post-fixed in 4% methanol-free PFA overnight at 4°C. The retina from the injected eye was carefully dissected from the eye cup and stored in 4% PFA before undergoing immunohistochemistry as a free floating retinal wholemount. For retinal sections, the eye cup was stored in 4% PFA before being cryoprotected in a 30% sucrose solution overnight. The eye cup was then immersed into optical cutting temperature (OCT) media (Thermoscientific, Cheshire, UK) and flash frozen on ice-cold Isopentane. Retinal sections were cut to a thickness of 18µm on a cryostat (Microm HM560s) and mounted onto slides. Slides were stored at -80°C until undergoing immunohistochemistry. For anatomical tracing, brain sections were also post-fixed overnight in 4% methanol-free PFA before being cryoprotected in a 30% sucrose solution. Following this, they were cut to a thickness of 50µm on a cryostat (Microm HM560s) in either the coronal or sagittal plain and stored in 0.1M PBS at 4°C until undergoing immunohistochemistry as free floating sections.

Immunocytochemistry Both retinal wholemounts and brain sections underwent immunohistochemistry for the eYFP, mCherry and mTeal proteins as free floating sections, whilst immunohistochemistry for retinal sections was performed on the slide. Tissue was initially permeabilised in a 1% TritonX-100 solution in PBS (1% PBS-X; PBS - Phosphate-buffered saline; 3x10 minutes) before being blocked in a mixture of 5% donkey serum (Sigma Aldrich, UK) and 5% goat serum (Sigma Aldrich, UK) in 0.2% PBS-X for 3 hours at room temperature. Tissue was subsequently incubated with primary antibodies against eYFP (chicken anti-GFP; 1:500; Kerafast) and mCherry (rabbit anti-mCherry; 1:500; Kerafast) and mTeal (rat anti-TFP; 1:500; Kerafast) for 3 days at 4°C. After this time, sections were washed thoroughly in 0.2% PBS-X before being incubated in their respective secondary antibodies: eYFP (Alexa-488 conjugated goat anti-Chicken; 1:200; Life technologies), mCherry (Alexa-546 conjugated donkey anti-rabbit; 1:200; Life Technologies) and mTeal (Alexa-647 conjugated donkey anti- rat; 1:200; Life technologies) for 12 Hours at 4°C. Sections were then thoroughly washed in

0.2% PBS-X before undergoing one final wash in dH20.

Chapter 3 Identifying ipRGCs in the degenerate retina

Wholemount retinas were placed onto glass slides with the ganglion cell layer facing up. The retina was cut into a Maltese cross motif before being mounted with Prolong Gold (Invitrogen, UK), coverslipped and left to dry. Brain sections were mounted sequentially from the order they were cut (rostral to caudal or lateral to medial) and mounted in the same orientation throughout. Once dry, sections were mounted with Prolong Gold (Invitrogen, UK), coverslipped and left to dry overnight in the dark at room temperature.

Imaging Both retinal and brain sections, in addition to wholemount retinas were initially imaged on a Leica M165 FC fluorescent stereo Microscope using the FITC and Texas red filter sets in order to sample the extent of viral transfection of the eYFP and mCherry labelled fluorophores. Wholemount retinas were imaged at high magnification (x20) and stitched together using the MosaicJ plugin in ImageJ in order to confirm labelling of soma, dendrites and axons. Brain sections were imaged using the 16.5:1 optical zoom at a range of magnifications which encompassed the entire nucleus under consideration.

3 Dimensional multi-colour images were acquired using a Leica TCS SP5 AOBS inverted confocal microscope which was equipped with a x63 x/ 0.50 Plan Fluotar objective. The confocal settings were as follows: pinhole 1 airy unit, scan speed 1000Hz bidirectional, format 1024 x 1024. Images were collected using the following detection mirror settings; 493-520nm; 566-625nm; 638-750nm using the 488nm and 564nm and 637nm laser lines, respectively. The power of individual laser lines was adjusted for each experiment in order to maximise the brightness range in each colour channel whilst preventing saturation. All images were acquired at a 16-bit colour depth via three photomultiplier tube (PMT) detectors and pseudocoloured during image acquisition. In order to eliminate cross-talk between colour channels images were collected sequentially. For acquiring 3D optical stacks, Z-depth was maintained at 1µm for both retinas and brain sections. To gain high magnification images over a wide area of target tissue, we utilised the tile-scan function in the confocal software (Leica AS). This uses a motorised X, Y stage during image acquisition. Individual tiles are imaged x63 magnification through the Z-stack. These individual tiles were stitched back together in X, Y and Z dimensions following image acquisition in order to represent a large region of the target tissue with high spatial resolution.

Chapter 3 Identifying ipRGCs in the degenerate retina

Multi-colour Image Analysis Our approach to multi-colour image analysis of Brainbow images can be divided into three major components: Pre-filtering of the image; Principal Component Analysis (PCA) and cluster isolation; and Post-filtering and cell reconstruction. Pre-filtering includes background noise reduction and exclusion of voxels that possess an RGB value in only a single colour channel and would not be deemed a unique identifying feature of a Brainbow- labelled cell during cluster identification. The PCA transformation of the RGB value of each voxel in the image is a standard statistical technique that creates, from a linear combination of the measured variables, a new set of orthogonal dimensions (‘Principle Components’) that are ranked by decreasing variance. The data is, in effect, rescaled, rotated and reflected such that, seen through the dimensions of its principle components, it reveals the internal structure of the data in a way that best explains its variance. This provides the most informative way with which to view a given ROI’s colour palette. This procedure, in one step, optimizes the ‘viewing angle’ of the data and increases the confidence in achieving reliable cluster isolation of voxels which correspond to a single cell. Finally, the post-filtering step allows for smoothing of pixels which correspond to individual isolated cells during spatial reconstruction and which may have been lost due to natural variation in fluorophore expression or noise during image acquisition. This provides far greater accuracy in the automated reconstruction of the 3D structure of the cell in order to gain quantitative structural information about the cell. These concepts are described in further detail below.

3.3.6.1 Filtering images All processing of Brainbow images was performed on a 64-bit desktop computer running Windows 7 (Microsoft). This computer possessed a 2GHz Intel Xeon processor (E5 2650; 16 cores) with 64GB of RAM and used an NVIDIA Graphics card (GTX 1070, Geforce). Raw Lif files of retinal tile scans were opened in IMARIS 8.3 (Bitplane, Zurich). ROIs were identified as regions of a wholemount retina which exhibited dense labelling of cells and demonstrated both variations in colour and brightness. These regions were subsequently cropped in X and Y (ensuring the entirety of the dendritic field of a single ipRGC remained in the field of view) and also in Z (ensuring the ganglion cell layer and both the ON and OFF sublaminae of the inner plexiform layer remained).

Chapter 3 Identifying ipRGCs in the degenerate retina

Following ROI identification, voxels were sent to python using the IMARIS XT function and analysed using the python library NumPy (van der Walt et al., 2011). A custom written programme, written in C by Dr Egor Zindy, filtered the ROI using the application programming interface (API) OpenMP (Open Multi-processing). This allowed for distribution of the filtering calculations of our image to be spread across all 16 cores of the computer’s processor, thus increasing the speed of calculations and spreading the computational demand of our large data sets. ROIs were filtered using a number of different parameters. Firstly, we used an undecimated multi-dimension wavelet transform (a trous wavelet) to remove high frequency noise from our images, based on the separable linear 3x3x3 kernel. This was applied to each of our Red, Green and Blue colour channels. We secondly compensated for voxel anisotropy in the Z axis by performing a nearest neighbour interpolation in 3D. We additionally applied a low pass filter to the voxel values in each of our R, G and B values in order to remove dim voxels from our image which corresponded to background noise and auto-fluorescence from the tissue. This was achieved by setting a threshold as a percentage of the full scale (0-65,535 in a 16-bit colour image) in each of the Red, Green and Blue channels. The cleaned image was subsequently sent back to IMARIS for 3D reconstruction and viewing to confirm adequate thresholding of the image.

3.3.6.2 Principal component analysis and cluster isolation The input variables for the 3D PCA analysis consisted of the RGB vectors whose elements are the three recorded colour channels from the confocal microscope from our ROI following the pre-filtering process. Each Voxel of our filtered ROI were once more sent back to Python, where the NumPy library was used to perform a principal component analysis on all voxels of the filtered image which had a value greater than 1. These new voxel values were plotted in principal component space and viewed as either a 3-Dimensional plot (PC1 vs. PC2 vs. PC3) or 3, interactive 2D plots (PC1 vs. PC2, PC1 vs. PC3 and PC2 vs. PC3) using the matplotlib python library (Hunter, 2007). In order to allow the programme to be interactive, we plotted a maximum of 400,000 random pixels in each of our 2D comparisons of principal component space as we found this was suitable for reliable extraction, whilst maintaining a practical refresh rate for graph generation and live updates of the 2D projection. In our 3D representations of principal component space, all voxels of the image are plotted.

Chapter 3 Identifying ipRGCs in the degenerate retina

In order to isolate voxels that formed clusters, we generated a graphical user interface (GUI) using the TkInter Python library within our 2D principal component plots. Here, we generated a 2 Dimensional 3-point polyline which was defined by an interactive handle point which extended to both the calculated black point and white point in principal component space. Adjustments for length and thresholding either side of this handle point, as well as the width of the polygon at the handle point were introduced as interactive scales arbitrarily ranging from 0-200. This allowed for real-time adjustments to the dimensions of the polygon in principal component space. To aid in single cell isolation, we additionally generated a live 2D maximum projection of voxels which fell within the constraints of our polygon in each of our three comparisons of principal component space. A colour bar was also generated which was calculated by the average RGB value of voxels at the user defined handle point.

To reconstruct a cell and account for image noise, we define a width around the polyline and only consider the pixels within the area of the polygon in PC1 vs. PC2, PC1 vs. PC3 and PC2 vs. PC3. The handle point defines the maximum length of the cluster from the black point. For high intensity voxels, we could consider the thresholding and length of a second segment of the polyline ranging from the handle point to saturation, where the handle point is now placed at the point of inflection in principal component space. We utilised the 2D projection and colour bar to aid in the confidence of our cluster isolation. All parameters for the analysis were saved as html files for each ROI. These parameters included the co-ordinates of the black point, the white point and the handle in principal component space, the length of the polyline, the width of the polyline at the handle point in addition to the forward and backwards calculations for the principal component analysis. We could additionally extract our input image, our output image and both the 3D and 2D projections of our principal component space with our isolation parameters.

3.3.6.3 Reconstructing individual cells Following isolation of an individual cluster the spatially indexed voxels were sent back to IMARIS 8.3 for 3D reconstruction and further processing. During this transfer, we once more used an undecimated wavelet transform in order to smooth over voxels which may not have been isolated due to the non-uniform expression of the fluorophore across the length of the cell. For this task, we found that JPEG2000 kernel performs best by preserving fine structures of dendrites and interpolating across these missing voxels. Spatially

Chapter 3 Identifying ipRGCs in the degenerate retina

reconstructed voxels which corresponded to an isolated cluster were stored in IMARIS as a new channel whose colour was defined by the average RGB voxel value at the handle point in principal component space.

We used the commercially available plugin filament tracer (IMARIS, Bitplane, Zurich) to trace the morphology of voxels representing a single cell in our new channel. Using the creation wizard, we utilised the automatic (no-loop) detection parameters to trace the 3 Dimensional structure of our voxels isolated from the analysis. Contrast settings and seeding values were set independently for individual isolated cells. Following this automated creation, the traced cell was manually inspected to ensure correct tracing and branch points were created by overlaying both the 3D isolated cell, and the original filtered RGB image. The “remove disconnected segments” function in addition to the “disconnect segments” and “join segments” function was used to improve the quality of the automated method and account for any discrepancies between the isolated cell and its reconstruction.

3.3.6.4 Quantitative analysis of reconstructed cells The dendrite beginning point was defined as the centre of the soma when overlaid onto the reconstructed cell. For visualisation purposes, the dendrite thickness was set to 2µm and therefore does not reflect the true thickness of individual cells. Images of the spatially reconstructed pixels from a cluster and their corresponding filament tracer were taken as 2 dimensional snapshots in both the X, Y and X, Z dimensions in IMARIS. Quantitative statistics for dendritic length, branch points and Sholl analysis were calculated from the reconstructed cell in filament tracer. Statistical values were exported and stored in Microsoft Excel 2007. Soma size and dendritic field size was calculated from the 2D projection of the isolated channel in ImageJ (Schindelin et al., 2012) and confirmed on the 2D projection of the reconstructed cell. For an individually isolated cell, we traced the outline using a minimal convex polygon enclosing on the cell body or the tips of the dendrites. We measured the area of this polygon in ImageJ and calculated the diameter of a circle of equal area as previously described (Berson et al., 2010, Schmidt and Kofuji, 2009). All graphs were generated in Graphpad Prism 7.

Chapter 3 Identifying ipRGCs in the degenerate retina

3.4 Results We first set out to confirm that our two Brainbow viruses could be used to label all subtypes of ipRGCs in the retina of the rd/rd Opn4Cre/+ mouse following intravitreal injection. Four weeks post-injection, expression of our Brainbow cassettes was confirmed by visualising strong immuno-labelling of the eYFP and mCherry proteins. This labelling was found throughout the retinal ganglion cell layer and across the entire retina (Figure 3.1A). As ipRGCs show subtype specific differences in their somato-dendritic profile we next set out to confirm that our labelled cells reflected these differences. Firstly, we measured the soma size of 362 labelled cells from 3 rd/rd Opn4Cre/+ retinas and found these to range from 6.51 - 33.62µm (14.9 ± 0.2; mean ± S.E.M), matching the distribution pattern for the five known ipRGC subtypes reported in the wildtype retina (Berson et al., 2010); Figure 3.1B). This suggested there was no strong bias in transduction efficiency of our virus across ipRGC subtypes. However, to further interrogate this, we next labelled retinal sections as different ipRGC subtypes are known to show dendritic stratification into different sublayers of the inner plexiform layer. Accordingly we found strong labelling of our Brainbow cassettes in the soma of cells whose dendrites extended into both the ON and OFF sublamina of the inner plexiform layer (Figure 3.1C).

In addition to differences in the retina, ipRGCs also demonstrate subtype-specific differences in their central projection targets for image forming (IF) and non-image forming (NIF) vision (Hattar et al., 2006, Ecker et al., 2010). Accordingly, we turned our attention to the brain and identified strong labelling of our Brainbow cassettes in axons innervating multiple NIF regions including the SCN (suprachiasmatic nuclei; Figure 3.1D), which receives strong input from M1 and weak input from M2 subtypes, and to the OPN (Olivary Pretectal nucleus; Figure 3.1E) which receives input from predominantly M2 but also M1 cells (Baver et al., 2008). We additionally found strong labelling in two IF brain nuclei, the Lateral Geniculate Nucleus (LGN; Figure 3.1F) and Superior colliculus (SC; Figure 3.1G). Labelling in the LGN encompassed both the ventral and dorsal portions of this nucleus (vLGN; dLGN) and the Intergeniculate leaflet (IGL). The dLGN receives projections exclusively from the M4 subtype (Ecker et al., 2010), whilst the IGL and vLGN are known to receive dense input from the M1 and M2 subtypes (Hattar et al., 2006). We found strong innervations to at least four of the superficial layers of the SC, a region involved in sensory processing and visual-motor control.

Chapter 3 Identifying ipRGCs in the degenerate retina

Figure 3.1 Viral transduction of Brainbow virus in the rd/rd Opn4Cre/+ mouse (A) Following intravitreal injection of two Brainbow viruses, immunohistochemical staining revealed transgene expression of the eYFP and mCherry fluorophores (in monochrome) in the soma and dendrites of cells across the ganglion cell layer. (B) Distribution of soma sizes from 362 mCherry and eYFP labelled cells from 3 rd/rd Opn4Cre/+ retinas. Lines above indicate the expected ranges from the known ipRGC subtypes in the wildtype Opn4Cre/+ retina. (C) Representative retinal section from an rd/rd Opn4Cre/+ mouse shows labelling of cells in the ganglion cell layer whose dendrites extend into both the ON and OFF sublamina of the inner plexiform layer (note the lack of observable nuclei in the outer nuclear layer in the DAPI counter stain reflecting loss of the rod and cone photoreceptors in advanced retinal degeneration). Strong subtype specific projections to non-image forming brain nuclei including (D) the SCN and (E) the OPN, in addition to image forming structures including (F) the LGN and (G) the SC. SCN = suprachiasmatic nuclei; OPN = Olivary Pretectal Nucleus; LGN Lateral geniculate nucleus; SC = Superior Colliculus. Scale Bars = 100µm in (D), (E) and (F); 500µm in (G).

Chapter 3 Identifying ipRGCs in the degenerate retina

We next looked to confirm that we could drive stochastic expression of our four fluorophores (tagBFP, mTeal, eYFP and mCherry) in order to maximise the colour variation between individually labelled ipRGCs. The stochastic expression of the Brainbow cassettes is achieved due to the viral construct encoding each fluorophore in an anti-sense direction whilst being flanked by mutant or wildtype loxP sites (Figure 3.2A). The action of Cre- recombinase at any of these pairs of loxP sites, depending on their orientation, can result in stochastic inversion and deletion events of up to four of these fluorescent proteins in each genomic particle delivered (Cai et al., 2013) - see Chapter 2 for further description of the stochastic recombination events. However, throughout our investigations we found we could not readily identify a strong signal originating from the tagBFP fluorophore, even after immunohistochemical amplification, and as such we did not use it for further investigations. We therefore used our three remaining fluorophores; eYFP (labelled with alexa-488 and pseudocoloured green during image acquisition), mCherry (labelled with alexa-546 and pseudocoloured red during image acquisition) and mTeal (labelled with alexa-647 and pseudocoloured blue during image acquisition). Using these three pseudo colours we could generate up to 7 unique colours with the possibility of hundreds of different hues depending upon the relative concentration of each expressed fluorophore in an individual cell (Figure 3.2B)

In order to maximise our colour variation and resolution, we turned to confocal microscopy. In addition to possessing fine control of excitation wavelengths and emission windows, this allowed us to obtain high quality spatial information in the X, Y and Z dimensions allowing for 3D reconstruction of dendritic architecture. Using this microscopy technique, we found that we could reliably image the strong stochastic labelling of our three labelled fluorophores in the rd/rd Opn4Cre/+ retina (Figure 3.2C). However, in order to reconstruct ipRGCs and their dendritic architecture in the retina we would need to retain this spatial resolution but image over a much larger area to ensure we encompassed the entire dendritic field of individual cells, which could be in excess of 500µm in diameter. We therefore took advantage of the tile scanning function on the Leica SP5 confocal microscope which uses a calibrated motor stage during image acquisition. Here individual tiles are imaged at a high magnification and are re-constructed following image acquisition, permitting high magnification images of a large area of tissue. Although constrained by the memory of the acquisition computer, we found we could not only readily image the stochastic expression of our Brainbow cassettes over large regions of the retina (Figure

Chapter 3 Identifying ipRGCs in the degenerate retina

3.2D) and optic nerve (Figure 3.2E) but also found this method to work in ipRGC projection targets including the SCN (Figure 3.2F) and the LGN of rd/rd Opn4Cre/+ mice (Figure 3.2G).

Chapter 3 Identifying ipRGCs in the degenerate retina

Figure 3.2 Stochastic expression of Brainbow cassettes in the rd/rd Opn4Cre/+ mouse (A) AAV Brainbow constructs for two Brainbow viruses. Virus 1 TagBFP (dark blue arrow) and eYFP (yellow arrow); Virus 2: mCherry (red arrow) and mTFP (light blue arrow). Genes encoding these four fluorophores are encoded in the anti-sense direction (denoted by direction of arrow; 3’-5’) between recombinant loxP sites (light and dark green portions of triangles; light green = mutant element, dark green = wild type element) in orientation determined by the position of the triangles. (B) Each row describes the possible recombination events after immunohistochemical labelling for three fluorophores: mCherry (Alexa 546, red), eYFP (Alexa 448; green) and mTeal (Alexa 647; blue). Colour of arrow denotes the pseudo colour assigned to each fluorophore during image acquisition. Arrows facing right describe a successful inversion event resulting in gene expression. Black arrows facing left denote no recombination event or a deletion event and hence no expression. (Right) the possible colour outcome of the combinatorial expression of fluorophores which underwent a successful recombination event. Seven possible colours can be produced due to the stochastic nature of Brainbow expression for three of the pseudocoloured fluorophores: white, yellow, purple, red, cyan, green and blue. Note this possible colour outcome does not account for the relative contribution of each individual fluorophore to the colour identity of the cell. (C) Representative maximal projection of a confocal image of an individual tile (magnification = x63) of stochastically labelled ipRGCs in the rd/rd Opn4Cre/+ retina (scale bar = 20µm). The tile-scanning function on the Confocal microscope allows for imaging of each fluorophore in an individual colour channel across a wide region of (D) the retina, (E) the optic nerve, (F) the SCN and (G) the LGN. The merged image demonstrates stochastic expression of Brainbow cassettes. Scale Bars are indicated for each image. Scale bars = 50µm in (D) and (E); 100µm in (F) and (G). (Higher spatial and colour resolution images are available in the digital version of this thesis).

Now that we had confirmed we could drive strong stochastic expression, we looked to devise an analysis method which could isolate an individual ipRGC based on its unique colour identity. To date, there have been no published attempts to analyse multi-colour images, such as those generated by Brainbow labelling, with the aim of isolating, extracting and spatially reconstructing single cells in 3D. Those which have been published rely on the visual inspection of spatially structured portion of a cell (such as an axon) and comparing the ratio between the Red, Green and Blue colour channels for a subset of pixels which make up that cell in order to support the assignment (Hammer et al., 2015). Furthermore, these approaches are conducted on 2D images of 3D acquired data. This not only loses the 3D structure of the biological specimen being analysed but also reveals substantial variability in colour ratios across pixels from a single cell as the RGB ratio of each voxel is averaged across the Z-stack (Hammer et al., 2015). As a result of this variability, it is harder to identify the group of pixels from a single cell from the population across an image in which several cells are labelled. In order to maximise the possibility of achieving this goal, we looked to take a novel approach by using principal component analysis (PCA) to define the most informative way of describing colour variation in 3 Dimensional multi-colour images. This statistical approach has not been used before to analyse Brainbow images but would provide the most informative method with which to view the relationship between any linear combination of the R, G and B values in principal component space, and therefore allow us to more reliably identify single cells based on their combinatorial expression of our three fluorophores. This approach should not only improve the reliability

Chapter 3 Identifying ipRGCs in the degenerate retina

in identifying the entirety of a single cell based on its unique colour identity, but additionally extracts those voxels that constitute it and reconstruct its morphology in 3D. This allows a researcher, for the first time, to quantitatively describe and characterise the 3 dimensional structure of a single cell following labelling of a population.

In order to gain structural information about the dendritic architecture of ipRGCs in the retina, we had to image at high magnification (x63) and at high resolution (voxel dimensions = 0.24 x 0.24 x 1um in X, Y and Z, respectively). At this resolution, the quantity of information generated made it unfeasible to image the entire wholemount retina. Instead, we concentrated on regions of dense stochastic labelling, which most often occurred around the optic disk and surrounding retina, to which intravitreal injections were targeted (Figure 3.3A). We imaged three such regions from three rd/rd Opn4Cre/+ mice which encompassed approximately 2150 x 2150 x 40µm in the X, Y and Z dimensions, respectively, and which would be used for analysis. However, the data size of these individual images was far too large to be analysed at once using PCA and therefore we cropped regions of interest (ROIs) on which we performed our PCA analysis (Figure 3.3B).

We took one of these representative ROIs from a retinal wholemount which demonstrated both high colour variation and changes in brightness. However, it was clear that these natural images possessed a high level of background noise which could be attributable to a number of different sources. A first source of noise is biological in origin and relates to the non-uniform expression of the fluorophore throughout the length of the cell (often most notable in the soma) which would manifest itself as variation from the mean colour of the cell on a voxel by voxel basis. Although our analysis of 3D images should minimise the colour variation within a single cell by analysing each voxel independently, it is of course possible that two different coloured cells can be imaged within the same voxel where portions of the cell (most likely dendrites) closely overlap in space (<1µm apart in Z). Under this situation the voxel would encode the weighted average of the RGB value from both labelled neurons resulting in a shift from the mean RGB value of that voxel from the rest of the cell. A second source of noise arises during image acquisition whereby autofluorescence from the retinal tissue and shot noise from the confocal microscope can introduce both a dim baseline activity into the background of each colour channel as well as a speckled appearance, respectively. In addition to this, there is a loss of laser penetration, and the ability to detect the emitted fluorescent signal, with increasing image

Chapter 3 Identifying ipRGCs in the degenerate retina

depth. As this is also a wavelength dependent effect there would be a shift in the RGB ratio with increasing imaging depth (Figure 3.3B).

As such we looked for a method to filter this noise. Thus, we ran our ROIs through a custom written python code which allowed for removal of a percentage of the total range (0- 65,535 in a 16-bit image) in each of the R, G and B channels. This filtering method not only removed the dimmest pixels from the entire range in each colour channel, but was also designed to remove both high frequency noise and pixels which were represented by only a single colour channel and therefore would not be defined as uniquely attributable to any single cell (Figure 3.3C). We next performed a PCA on this cleaned ROI and identified that our voxels were distributed across principal component space and appeared to form clusters around lines originating from a single point in principal component space. The mean of individual clusters resembled a vector with both direction and magnitude; however these clusters also demonstrated variation in their width which represented a deviation from this mean vector on a voxel by voxel basis (Figure 3.3D).

Chapter 3 Identifying ipRGCs in the degenerate retina

100

Chapter 3 Identifying ipRGCs in the degenerate retina

Figure 3.3 Processing retinal Brainbow images (A) Maximum projection of a confocal image representing stochastic labelling of ipRGCs from a tile-scanned region of the rd/rd Opn4Cre/+ retina (Dimensions = 1683µm x 1777µm x 40µm in X, Y & Z, respectively) centred around the optic disk (scale bar = 100µm). (B) Regions of interest (ROIs; white dashed box in (A)) were identified as labelled areas of the wholemount retina which expressed high variations in both colour and brightness and also encompassed the entire dendritic field of at least one uniquely labelled ipRGC (orange cell in the centre of the image; scale bar = 50µm). (C) ROIs were filtered using a custom written python code to remove high frequency noise, dim pixels and pixels which were represented in only a single colour channel whilst compensating for anisotropy in the 3D image. (D) Following principal component analysis of the filtered ROI in (C), voxels were plotted in a 3D representation of principal component space (PC1 vs. PC2 vs. PC3) and shows voxels which make up individual clusters diverging from the calculated black point. (Higher spatial and colour resolution images are available in the digital version of this thesis).

A PCA transformation of a dataset is unique based on the values it encompasses. Therefore, differences in colour values of individual images would result in differences in the identifiable clusters in PC space. Thus, we next looked to model our data in order to understand how the changes in colour and brightness generated by our stochastic Brainbow labelling would manifest themselves in principal component space, in order to devise a method by which to define clusters. We initially generated three bars which represented the three pseudo colour channels used during image acquisition: red (mCherry), green (eYFP) and blue (mTeal) (Figure 3.4A). As the Brainbow labelling method allows for a variation of brightness in each of these colour channels, we generated each bar to range from darkness (0; black) at its edges to saturation (65,535 in a 16-bit image) in the centre. Performing a PCA on this image we found these pixels cluster into lines which diverge from a point of origin before reaching a point of inflection where they converge back to a single point in 3-Dimensional principal component space (Figure 3.4B). The colour indexing of the pixels which made up these individual lines suggested they each represented an individual bar from our original input image. To confirm this, we looked to extract the pixels which represented distinct regions of these clusters in order to understand what characteristics of the original input image they represented. Due to the computational demands of isolating and extracting this data in 3D, we instead visualised our 3D principal component spaces as 3, 2-dimensional comparisons for the isolation and extraction process. Accordingly, we generated three plots which compared PC1 vs. PC2, PC1 vs. PC3 and PC2 vs. PC3 (Figure 3.4C).

We identified that the point of divergence and convergence in principal component space could be calculated for each image during the principal component transformation and that they represented black (‘black point’) and saturated pixels (‘white point’), respectively. In

101

Chapter 3 Identifying ipRGCs in the degenerate retina

each principal component comparison, voxels clustered into three lines which extended from this black point to the white point via a point of inflection. Voxels which clustered in one principal component comparison also clustered in the other two. However, individual clusters were not always separable by their position in principal component space in a single paired comparison when viewing our 2D projections, but were separable in the other two comparisons. This is most clearly demonstrated by the blue and green clusters which are inseparable in PC1 vs. PC2, however readily split in PC1 vs. PC3 and PC2 vs. PC3 (Figure 3.4C).

We therefore looked to generate an extraction tool which reflected the shape of these clusters. We generated two rectangles, one which extended from the black point to the point of inflection, and the other from the point of inflection to the white point. An interactive “handle point” was used to define the point of inflection at which the two rectangles would meet and reflect the three-point polyline created by the clusters in principal component space. Voxels which lay within the boundaries of these rectangles would define the cluster which would be extracted. However, as individual clusters could overlap in an individual comparison of principal component space, we applied a constraint so that only voxels which were present within the boundaries of the extraction bars in each individual 2D comparison of principal component space would be extracted for reconstruction. As at least one of the principal components was always present in each of the other two 2D, adjusting the position of the handle point in one comparison resulted in the relative change in the position of that principal component in the other two comparisons due to their shared co-ordinate system. Importantly, indexing of the spatial position of the voxel in the original image and its RGB colour value was maintained throughout the principal component transformation.

Using this isolation and extraction tool, we first looked to extract the region of the green cluster from the black point to the point of inflection. Isolation and subsequent extraction of these pixels demonstrated that this region in principal component space corresponded to the dimmer pixels in the image as observed by the spatial reconstruction of the outer edges of the green bar (Figure 3.4D). Subsequently, we placed our extraction bar from the point of inflection to the white point and found that this corresponded to the brighter pixels in the centre of the green bar (Figure 3.4E). This indicated that the colour of a cluster is represented by its vector direction whilst its brightness is defined by its magnitude (‘length’) in principal component space. The point of inflection therefore represents the

102

Chapter 3 Identifying ipRGCs in the degenerate retina

brightness at which pixels from this cluster are most widely separated from the rest of the population in this principal component space.

Figure 3.4 Modelling Colour and brightness with artificial images (A) Artificial image of three coloured bars (red, green and blue which range from black at the edges to saturation in the centre) reflects the maximal brightness range of each of the three pseudo colours used during image acquisition. (B) Following principal component analysis of the image in (A), voxels were plotted in a 3D representation of principal component space (PC1 vs. PC2 vs. PC3) which shows three distinct clusters ranging from the black point to the white point, via a point of inflection, in principal component space. (C) The handle point (interactive red square) was placed at the point of inflection and extraction and spatial reconstruction of voxels which occupy the region between the black point and point of inflection of the green cluster represent dim pixels on the outer edges of the green bar from the input image. (D) Extracting and reconstruction of voxels which occupy the region from the handle point to the white point of the green cluster represent bright pixels from the centre of the green bar in the input image. (E) The entire cluster from the black point to the white point must be extracted in order to reconstruct the entire brightness range of a single colour.

103

Chapter 3 Identifying ipRGCs in the degenerate retina

We next set out to determine how the analysis method might function with the composite colours (defined by co-expression of up to three fluorophores at differing concentrations), found in our transfected retinas. To this end, we generated an artificial image containing four coloured bars: purple (1.0, 0.0, 1.0, 0.8), yellow (0.8, 0.8, 0.1, 0.5), cyan (0.2, 1.0, 1.0, 0.8) and orange (1.0, 0.3, 0.0, 1.0) in a Red, Green, Blue, Intensity colour format where each parameter ranges from 0-1 (Figure 3.5A). During acquisition of genuine retinal images, we identified that individual colour channels contribute their own noise to the image, and this was often strongest in the green and red channels. We therefore introduced a high level of Gaussian noise into the green colour channel (20 standard deviations from the mean) of our simulated image and a lower level of noise into the red and blue colour channel (5 standard deviations from the mean). Performing PCA on this artificial image identified four clusters in 3 Dimensional principal component space which related to the four individual bars from our input image (Figure 3.5B). These clusters again diverged from the point of origin, however the extent to which each cluster reached the convergence point (its ‘length’) varied according to the brightness input. This can be most clearly observed when comparing the cluster corresponding to the bright orange bar, which is saturated in the centre, with that for the dim yellow bar, which is comparatively shorter and does not reach a point of inflection. Thus, voxels above a certain brightness level for an individual colour cluster around a 3-point polyline, whilst voxels which make up dimmer colours cluster around a single straight line from the origin which does not reach a point of inflection. As the PCA is unique based on the variables of the input image, the extent to which this happens would be unique for each individual input image. The introduction of high level noise into the green channel resulted in an increase in the width of clusters to which the green colour channel contributes. Accordingly, the clusters representing the cyan, orange and yellow bars became much wider, whilst the cluster representing the purple bar, of which the red and blue colour channel primarily contribute is comparatively narrow in the 3D principle component space (apparent in the plot for PC2 vs. PC3 in Figure 3.5C).

Whilst both the length and width of individual clusters could vary depending on the brightness and noise, respectively, this did not prevent the extraction of voxels corresponding to an individual bar. As an example, voxels making up the cyan bar were still readily isolated and extracted under these conditions; however the quantity of background noise which is extracted also increased (Figure 3.5C). We therefore introduced two further user defined inputs in order to avoid this problem. Firstly, we introduced a thresholding function to exclude dim pixels near the black point and saturated pixels near the white

104

Chapter 3 Identifying ipRGCs in the degenerate retina

point, which cannot be well separated in principal component space. Secondly, we exchanged our two rectangles for a polygon, whose width tapered as it tended towards the black point and white point. Using these new parameters, we found we could readily isolate a large proportion of the pixels which represented our signal (the cyan bar) with minimal background noise (Figure 3.5D).

Figure 3.5 Modelling noise with artificial images (A) Artificial images of four coloured bars to which each of the red, green and blue colour channels contribute (Purple: 1.0, 0.0, 1.0, 0.8), Yellow (0.8, 0.8, 0.1, 0.5), Cyan (0.2, 1.0, 1.0, 0.8) and Orange (1.0, 0.3, 0.0, 1.0) in a Red, Green, Blue, Intensity colour format where each parameter ranges from 0-1. High level Gaussian noise (20 standard deviations from the mean) is added to the green channel and low level Gaussian noise (5 standard deviations) is added to both the red and blue channels. (B) Following principal component analysis of the image in (A) voxels were plotted in a 3D representation of principal component space (PC1 vs. PC2 vs. PC3) and shows four distinct clusters originating from the black point, with each cluster showing variation in length and width representing changes in brightness and noise, respectively. (C) Representative example of extraction of voxels which constitute the bright cyan cluster and their spatial reconstruction identify the cyan bar in the image with high level background noise. (D) Swapping the extraction bar for a polygon whose width is the same at the handle point but which tapers towards the black and white point results in spatial reconstruction of the cyan bar with minimal background noise.

105

Chapter 3 Identifying ipRGCs in the degenerate retina

We then applied these methods to the filtered ROI from our natural Brainbow image from the rd/rd Opn4Cre/+ retina (Figure 3.6A). Although filtered ROIs could now be run through our PCA analysis, they still contained hundreds of millions of pixels. It was far too computationally demanding to plot all of these pixels in the 2D principal component spaces while also allowing for interactive identification of clusters. Therefore, whilst every voxel within the filtered ROI contributed to the principal component analysis and was assigned a location in 3D principle component space; user definition of upper and lower thresholds, length and width for the polygons to define clusters was undertaken on principle component projections for a random sample of 400,000 voxels. As an additional aide to this process, the user was given access to a 2D projection of all the voxels contained within the extraction polygon (including those outside the randomly selected 400,000) on the ROI. This provided live feedback of whether voxels falling within a proposed cluster defined an individual ipRGC.

Using this approach, we were able to reliably isolate 3 ipRGCs in the rd/rd retina from 10 identifiable cell bodies in our representative filtered ROI (Figure 3.6A). We demonstrate the isolation in each of the 3 2D comparisons and the 2D projection for the pink (Figure 3.6B), green (Figure 3.6C) and orange cells (Figure 3.6D). Whilst we could additionally isolate a purple cell which corresponded to a defined cluster in principal component space, we found this cell contained multiple axons and faint dendrites in the 2D projection suggesting contamination from other cells with a similar colour (Figure 3.6E). Additionally, we found that the readily identifiable blue/purple cluster actually encompassed cell bodies and dendrites originating from at least two separate cells, suggesting these cells could not be separated in principal component space based on colour alone (Figure 3.6F).

Combined, these data demonstrated that our analysis method could reliably isolate, extract and spatially reconstruct individual ipRGCs that had been both uniquely and strongly labelled with Brainbow technology based on the parameters described on our modelled data. However, this could not be reliably achieved without additional reference to the location of extracted voxels in anatomical space to ensure that they comprised a single identifiable cell. Thus, in some instances, what appeared to be an individual clusters based upon principle component space could contain voxels originating from multiple cells. This likely reflects the similar expression levels of cells which express only two fluorophores (in this example purple).

106

Chapter 3 Identifying ipRGCs in the degenerate retina

Figure 3.6 Single cell isolation of naturalistic Brainbow images from the rd/rd Opn4Cre/+ retina (A) Clusters of voxels are identifiable in the three 2D representations of principal component space; PC1 vs PC2 (left), PC1 vs. PC3 (middle) and PC2 vs. PC3 (right) from the filtered ROI. Cluster isolation of the (B) Pink cluster, (C) Green cluster and (D) orange demonstrates single cell isolation from the ROI in (A) following their 2D spatial reconstruction (note differences in the direction, length, width and thresholding of the extraction tool between these three isolated clusters in PC space). (E) Isolation of the lilac cluster demonstrates one cell in the 2D spatial reconstruction, however faint dendrites and portions of axons from other cells with a similar colour can also be observed in the spatial reconstructed 2D projection. (F) Isolation of the blue / purple cluster demonstrates that voxels from this cluster represent somas and dendrites from at least two cells whose colour identity cannot be separated in principal component space. (Higher spatial and colour resolution images are available in the digital version of this thesis).

107

Chapter 3 Identifying ipRGCs in the degenerate retina

We finally looked to implement our colour analysis method by assessing the morphology of individual ipRGCs in the rd/rd retina to determine the impact of outer retinal degeneration. Using the principal component analysis method we developed, we were able to reliably isolate and re-construct 33 uniquely labelled ipRGCs out of a total of 355 cells expressing one or more of the fluorescent proteins from 3 rd/rd retinas, corresponding to an approximate 10% success rate. Isolated voxel clusters were sent back from our isolation software to IMARIS where they could be visualised in 3D. We used the commercially available software filament tracer (IMARIS, Bitplane; Zurich) to reconstruct the 3D morphology and architecture of these clusters. In doing so, we obtained a number of statistics about this population including soma size, dendritic field size, dendritic length, the number of branch points, and complexity based upon 3D Sholl analysis.

In the wildtype retina, the five different ipRGC subtypes exhibit differences in their dendritic stratification, dendritic field size and complexity of their dendritic architecture (Berson et al., 2010, Estevez et al., 2012, Schmidt et al., 2011b). Accordingly, we could utilise our 3D re-constructions from filament tracer and these statistics in order to categorise our cells. We therefore first set out to categorise our cells based on their stratification. We identified six cells that stratify into the inner plexiform layer. A representative example of one of these isolated cells and its reconstruction in filament tracer are shown in both X, Y dimensions and X, Z dimensions in Figure 3.7A. This population had an average soma size of 13.1 ± 0.5µm and an average dendritic field size of 305.4 ± 28µm. Based on anatomy from the wildtype retina, these ipRGCs would be classified as the M1 subtype. Additionally, one isolated cell from this population had its soma in the inner nuclear layer matching the known description of displaced M1 cells (dM1) (Figure 3.7B). We identified two cells that stratified in both the ON and OFF sublaminae of the inner plexiform layer (Figure 3.7C). The dendritic architecture of these bistratifying cells matched that described for the M3 subtype with a large dendritic field size 277.54 ± 12 µm and soma size 16.9 ± 0.5 µm.

All other known ipRGC subtypes exclusively stratify in the ON sublamina of the inner plexiform layer. To help categorise the remaining 24 ipRGCs from the rd/rd retina, we looked at their dendritic field size, which is reported to vary between M2, M4 and M5 subtypes. We identified some cells with a relatively small yet complex dendritic field size (Figure 3.7D). This population had a mean soma size of 14.59 ± 0.7 and a mean dendritic

108

Chapter 3 Identifying ipRGCs in the degenerate retina

field size of 189.9 ± 11µm. This small and complex population of cells matches the description of the recently reported M5 ipRGC in the wildtype retina (Ecker et al., 2010).

Figure 3.7 Categorisation of the M1, M3 and M5 ipRGC subtypes in the rd/rd Opn4Cre/+ retina. (A) 3D reconstruction of an isolated cluster shows a representative M1 cell in (top) X, Y and (bottom) Y, Z dimensions, whose dendrites stratify exclusively in the OFF sublamina of the IPL. The filament tracer reconstruction of this isolated cluster is shown to the right in both X, Y and Y, Z dimensions. (B) A representative Displaced M1 (dM1) cell also stratifies exclusively in the OFF sublamina but has its soma displaced to the inner nuclear layer. (C) 3D reconstruction of voxels, and their filament tracer reconstruction, shows an M3 cell that stratifies in both the ON and OFF sublamina of the IPL. (D) 3D reconstruction of an isolated cluster and its filament tracer reconstruction show a representative M5 cell whose dendrites exclusively stratify in the ON sublamina of the IPL but displays a small and complex dendritic field. (E) Soma size vs. dendritic field size of M1 (blue; n = 6), M3 (black; n = 2) and M5 (red; n = 7) of reconstructed cells from the rd/rd Opn4Cre/+ retinas. (F) Dendritic field size

109

Chapter 3 Identifying ipRGCs in the degenerate retina

vs. total number of branch points, as calculated from filament tracer, of M1, M3 and M5 cells. (G) Sholl analysis of reconstructed M1 and M5 ipRGCs plots the number of dendritic crossing (Mean ± SEM) with increasing retinal eccentricity shows M5 cells possess a smaller but more complex dendritic field architecture. Scale Bar = 50µm in (A), (B), (C) & (D). (Higher spatial and colour resolution images are available in the digital version of this thesis).

To confirm the categorisations of these three cell types, we plotted a number of dendritic field characteristics that are known to differ between M1, M3 and M5 types in the wildtype retina. Firstly, we plotted soma size against dendritic field size and could clearly identify our two M3 bi-stratifying cells as having a larger soma size than M1 or M5 subtypes (Figure 3.7E). M1 and M5 cells are also known to exhibit differences in complexity of dendritic architecture. We found that in our sample, cells categorised as M1 or M5 by other methods also were separated in this dimension (Mean ± SEM number of branch points; M1 cells = 11.3 ± 2.0 and M5 cells = 36.8 ± 2.5; Unpaired T-test, p = <0.0001) (Figure 3.7F). This was more apparent in Sholl analysis of the M1 and M5 cell population which shows the number of crossings as a function of soma eccentricity (Figure 3.7G). For the M5 population, the Mean ± SEM number of crossings peaked at 25.8 ± 1.6 at a distance of approximately 60μm from the soma and dropped to nearly 0 by 150μm, reflecting the small dendritic field size of this cell type. By contrast, the M1 population had a much sparser dendritic morphology, with the average number of crossings peaking at 9.7 ± 1.8 at a distance of approximately 60μm from the soma but not reaching zero till greater than 220μm. We did not plot Sholl analysis for M3 cells due to the small sample size and their known variability on a cell by cell basis (Schmidt and Kofuji, 2011).

M2 and M4 ipRGCs are significantly harder to categorise as they both stratify in the ON sublamina of the inner plexiform layer and exhibit significant overlap in their dendritic field size in the wildtype retina. However, M4 cells can be somewhat distinguished from M2 cells as they possess a much larger soma size and on average slightly larger and more complex dendritic architecture (Estevez et al., 2012). We therefore first plotted our remaining cells in terms of these first two parameters and found that they fell into two discrete categories (Figure 3.8A). To confirm this categorisation we next plotted the dendritic field size of these cells against the total number of branch points. In doing so, we identified that the suggested M4 population of cells had a significantly more complex dendritic architecture (Mean ± SEM branch points = 37.9 ± 2.3) than that of M2 cells (27.5 ± 1.8) (Figure 3.8B; p = 0.005; Unpaired T-test). Performing Sholl analysis on these two populations confirmed this difference in complexity (Figure 3.8C). A representative isolated cell from the M2

110

Chapter 3 Identifying ipRGCs in the degenerate retina

population (Figure 3.8D) and the M4 population (Figure 3.8E) are displayed below alongside their reconstruction in both X, Y and X, Z dimensions.

Figure 3.8 Categorisation of the M2 and M4 ipRGC subtypes in the rd/rd Opn4Cre/+ retina (A) Soma size vs. dendritic field size of M2 (green; n = 9) and M4 (orange; n = 9) demonstrates that M4 cells have a larger soma size than the M2 population. (B) Dendritic field size vs. total number of branch points demonstrates that M4 cells have a slightly larger and more complex dendritic field (M4 = 37.9 ± 2.3, M2 = 27.5 ± 1.8; p = 0.005; Unpaired T-test). (C) Sholl analysis of reconstructed M2 and M4 ipRGCs plotting the number of dendritic branches (Mean ± SEM) with increasing retinal eccentricity confirms M4 cells possess a more complex dendritic field than M2 cells. 3D reconstruction of an isolated cluster showing (D) a representative M2 ipRGC and (E) a representative M4 ipRGC with their corresponding filament tracer reconstruction in both X, Y and Y, Z dimensions. Scale bar = Scale Bar = 50µm in (D) & (E). (Higher spatial and colour resolution images are available in the digital version of this thesis).

We finally compared these anatomical measures for the five ipRGC subtypes in this study of the rd/rd OPN4Cre/+ retina (Figure 3.9A) with those taken from the literature for visually intact Opn4Cre mice (Figure 3.9B). Based on our recorded parameters, we find that the soma sizes, dendritic field sizes and total number of branch points (an indicator of dendritic complexity) to be within the range of those previously reported from the wildtype retina. An exception to this is that of the M3 cells which showed a smaller dendritic field size and

111

Chapter 3 Identifying ipRGCs in the degenerate retina

likely reflects the low number of cells isolated and their great variation in the wildtype retina. We do however find that the total dendritic length of our cells shows more variation. We identify that M1 and M2 cells to have a slightly larger dendritic length then previously reported, whilst M3 and M4 cells are slightly shorter.

Figure 3.9 Anatomical characteristics of ipRGC subtypes from the degenerate and wildtype retina. Population data (Mean ± SEM) for soma size, dendritic field size, total dendritic length and total number of branch points collected from (A) reconstructed Brainbow cells from 3 rd/rd Opn4Cre/+ retinas and (B) from published literature in visually intact Opn4Cre/+ mice. Wildtype values previously reported by (Schmidt et al., 2011), (Ecker et al., 2010) and (Estevez et al., 2012).

112

Chapter 3 Identifying ipRGCs in the degenerate retina

3.5 Discussion Following the death of rod and cone photoreceptors in retinal degeneration, ipRGCs provide the only source of light information reaching the brain. In the healthy retina, these cells influence a number of physiological and behavioural responses to light which are reflected in the distinct anatomical differences between the five subtypes of ipRGC in the retina (Berson et al., 2010, Schmidt et al., 2011a, Estevez et al., 2012). Their survival in retinal degeneration is therefore of vital importance in order to continue supporting behavioural responses to light at a time when a large proportion of retinal cell types are undergoing re-modelling following outer retinal degeneration (Jones and Marc, 2005, Marc et al., 2007). Whilst the M1 and M2 subtype are known to survive retinal degeneration with a broadly normal retinal anatomy (Vugler et al., 2008, Lin and Peng, 2013), there have been no direct investigations into the survival and anatomy of the more recently identified M3- M5 subtypes. At least one of these subtypes, the M4 cell, is known to project to image forming regions including the dLGN and superior colliculus (Ecker et al., 2010, Brown et al., 2010) and therefore provide a viable route through which light-information can still reach visual processing structures in the brain. We therefore set out to identify and quantitatively characterise the dendritic architecture of the 5 known ipRGCs subtypes in the retina of aged rd/rd Opn4Cre/+ mice using the multi-colour labelling technique Brainbow. In doing so, we additionally design a novel approach to analyse multi-colour images using principal component analysis in which we not only identify cells based on their unique colour, but for the first time isolate and reconstruct their 3D structural morphology allowing for a comprehensive analysis of their dendritic architecture.

Limitations of current multi-colour image analysis methods

A principal aim of neuroscience is to understand the network connectivity of populations of cells with single cell resolution. Colour provides an excellent method to represent and visualise individual cells in a population and as a result, there has been an explosion in the number of methods now available to label populations of neurones with a unique combination of fluorescent proteins (Cai et al., 2013, Livet et al., 2007, Nern et al., 2015). However, in comparison, the ability to quantitatively analyse these data sets has lagged behind. Current approaches have employed 2D projections of colour images acquired in 3D and used the ratio between the R,G and B values to assign portions of a biological structure as belonging to the same cell (Hammer et al., 2015). However, that strategy has a

113

Chapter 3 Identifying ipRGCs in the degenerate retina

significant drawback as it relies on the projection of a 3D image into 2D. This distorts the true colour value of individual voxels as they become averaged over the entire Z-stack range. This only enhances the non-uniformity of fluorophore expression that can occur along the length of a cell. Other approaches have looked to transform RGB values into different colour spaces such as HSV (Hue Saturation Lightness) in order to more reliably identify the unique colour identity of an individual cell however these methods are yet to yield any significant or published results from histological images (Weissman and Pan, 2015).

Novel approaches to multi-colour image analysis

We therefore took the approach of using principal component analysis (PCA). PCA is a mathematical transformation of a multi-variable dataset, which is subsequently ranked according to the relationship between variables which possess the highest variance. In our data, our variables consist of our three fluorophores and are represented by RGB colour values, each ranging from 0 - 65,535 in our 16-bit images. By observing these relationships in principal component space, we effectively rescaled and rotated this new linear combination of variables in order to provide the most informative method by which to view the data and therefore more reliably identify clusters which represent an individually labelled cell (Figure 3.6). This statistical approach is used throughout neuroscience, most notably in the analysis of electrophysiological data, and therefore we set out to design a colour analysis method analogous this. Using this approach on artificial images used to model Brainbow, we identified that individual colours are represented by the direction of individual lines in 3 Dimensional principal component space to which the voxels of a particular image cluster (Figure 3.4). Changes in brightness and noise alter the length and width of these clusters respectively (Figure 3.5). However, thresholding for both dim and saturating pixels allowed us to reliably identify individual spatial structures in both our modelled data and naturalistic Brainbow images based solely on their unique colour identity.

However, we must consider that our artificial images on which we modelled our analysis technique do introduce two artefacts which are not present when analysing histological images. Firstly, the coloured bars in these images are generated with a resolution of 20 pixels in width. Therefore, coloured bars which possess a large brightness range appear to

114

Chapter 3 Identifying ipRGCs in the degenerate retina

have a periodic pattern in PCA space compared to dim coloured bars which do not reach saturation as there is a larger spread of the same number of discrete data points. This can be rectified by increasing the resolution of the artificial image or by introducing Gaussian noise into the image. Secondly, we find in our artificial images there is a point of inflection in clusters which reflects the point at which one of the colour channels reaches saturation before the other two. These two unsaturated colour channels can still increase in brightness depending on the stochastic expression of the fluorophores in question. If this occurs they tend back towards white (or saturation) with increasing brightness resulting in a change in vector direction. This inflection point is not often reached in histological images due to the stochastic nature of fluorophore expression and the careful setting of imaging parameters and laser lines to be just below saturation in order to maximise the full range of values in a 16-bit image. To further increase the accuracy of our modelled data we may also consider generating overlapping horizontal and vertical bars of stochastic colours in 3D in order to more readily replicate the scenario where there may be the colour contribution from two neurones to an individual voxels RGB value.

The principal component analysis method could provide a comprehensive method to isolate individual cells based on unique colour labelling. It is therefore important to carefully consider the efficacy of our cluster isolation and whether it does indeed result in identification of the dendritic architecture originating from a single ipRGC. The importance of this is highlighted in the demonstration of our representative ROI (Figure 3.6A). Whilst we do successfully isolate a number of single cells (Figure 3.6B, C & D), we also find that some clusters contain voxels from axons, somas and dendrites of multiple labelled cells with similar colour (Figure 3.6E & F). Therefore to maximise the accuracy of our single cell identification and reduce the chance of false positives we introduce a number of steps in both our imaging and analysis software to maximise our confidence in single cell isolation.

Firstly, we find that brightly labelled cells are more readily isolated with minimal noise as the voxels which represent them are more widely distributed in principle component space at large distances from the black point (Figure 3.4 & 3.5). Accordingly, maximising the brightness range during image acquisition is vital to ensuring reliable isolation and driving stronger colour variation. Secondly, image pre-processing is vital. Initial filtering of raw ROIs removes high frequency noise which represents auto-fluorescence from the tissue and makes up a significant proportion of voxels in our 3D images (Figure 3.3B & C). Removal of these voxels allows for clearer distinctions of individual clusters in principal component

115

Chapter 3 Identifying ipRGCs in the degenerate retina

space, especially near the black point. Furthermore, removal of pixels which are only present in a single colour channel is an effective method to threshold out cells which have not been labelled by multiple fluorophores, making it most unlikely that they have a uniquely identifiable colour.

The separation of our clusters is most readily visualised in 3D representations of PC space (Figure 3.3D). However, due to computational limitations this was not achievable and thus we used our 3 2-Dimensional comparisons to visualise our data. Nonetheless, it can be clearly identified that individual clusters are often more separable in one comparison of the principal components than another (Figure 3.6). Our ability to view all three of these comparisons at once, and the constraint that only voxels which appear in our extraction polygon in all three of these comparisons are included in the cluster therefore minimises the chance of extracting voxels from clusters which may overlap in one of these comparisons. This aspect is further aided by our ability to have a live view of the 2- Dimensional spatial reconstruction of the voxels defined by our extraction polygon which could be inspected prior to cell extraction. In addition, the use of our colour bar provides a qualitative online confirmation that the average colour of a cluster and the colour of the cell in the original image are in agreement and therefore maximising the extracted number of voxels which relate to a single cell.

A unique aspect of this work was that we not only looked to identify the unique colour ratios of these cells, but to also extract pixels which made up an individual cluster in order to recreate their 3D morphology. This provides an extra spatial dimension when confirming single cell cluster isolation which has not been achievable when using current approaches to colour analysis. Consequently, our final confirmation is in the reconstruction of the isolated voxels of the individual cluster itself. Any voxels which may encode the weighted RGB value of two neurones whose dendrites overlap may not be extracted due to a shift in the position of those voxels in PC space compared to the average for the entire cell. This would result in a punctate loss of voxels across regions of overlap in the spatially reconstructed cell. However, our post-filtering methods smooths over these voxels which may have been lost during the isolation process, aiding in more accurate cellular reconstruction and tracing. In addition, individual cells, by their very nature, are continuous structures. Therefore, discontinuities which could arise from portions of dendrites or axons of different cells with a similar colour palette can be readily assigned as noise during the reconstruction process if they do not show continuity with the isolated cell. A challenge for

116

Chapter 3 Identifying ipRGCs in the degenerate retina

this approach is when it returns two groups of dendrites each comprised of continuously connected voxels, but not connected to one another. Do these represent different elements of the same cell or different cells? The only unambiguous way to determine that they are the same cell is to find a common branch point from which they diverge. Achieving that might require imaging and processing of ROIs which encompass a larger region of tissue, especially if one is working in the brain where projections can span many millimetres from one brain region to another.

The observation that pixels from more than one cell sometimes reside in a single cluster indicates that there is room for improvement in the method to ensure that a single cluster in principal component space corresponds to a uniquely labelled individual cell. The nature of PC analysis allows for large data sets with multiple variables to be considered. One way in which we could significantly increase our reliability is by increasing the number of fluorophores contributing to the analysis. We currently utilise three variables which are our three fluorescent proteins (mCherry, eYFP and mTeal). Driving strong expression of the tagBFP fluorophore would add a fourth dimension to this colour analysis which would further increase the colour diversity of all labelled cells per unit area (Cai et al., 2013) and should manifest itself as better and more reliable clustering in principal component space.

Improving our multi-colour analysis method

Using our software we were able to reliably isolate and reconstruct 32 single ipRGCs from 335 cells expressing one or more fluorescent proteins 3 rd/rd OPN4Cre/+ retinas. This represents a conservative 10% success rate from all labelled cells in the population as this does not account for cells removed from the analysis during the pre-filtering process. Nonetheless, this is far a superior success rate compared to other approaches currently in use to reconstruct the entire 3D structure of a single cell. Existing strategies to characterise the dendritic architecture of ipRGCs in the retina rely on intracellular injections of dye on a cell by cell basis following electrophysiological recordings which is both time consuming and low throughput (Estevez et al., 2012, Ecker et al., 2010, Schmidt and Kofuji, 2009). Furthermore, for more recent studies which have looked to trace an ipRGC projections to the brain, this is limited to a single cell per animal (Fernandez et al., 2016). Thus, our approach also has ethical implications by vastly improving both the quantity and quality of data collected from a single animal. Nonetheless, there may be scope to increase this

117

Chapter 3 Identifying ipRGCs in the degenerate retina

efficiency further. At present we have extracted approximately 11 cells per retina, however, collection of this data set has only concentrated on the central region of the retina where we regularly found strong labelling. Therefore there are also technological modifications which would improve both segmentation quality and quantity. Optimising the excitation of the three expressed fluorophores in each colour channel across the entire 16-bit range, whilst preventing saturation across the entire retina, would drive maximal colour diversity and hence increase the yield of uniquely identifiable cells. Furthermore, as is evident in figure 3.3D, we found signal in the green channel (eYFP expression), to be consistently weaker and often not co-expressed with cells which expressed mTeal (blue) and mCherry (red). The eYFP fluorophores is encoded on a different virus compared to the mTeal and mCherry fluorophores (Figure 3.2A); indicating co-transfection of the same cell by the two viruses is submaximal. This manifests itself by many labelled cells being biased towards the colour purple (the combined expression of mCherry and mTeal) or just green (just eYFP expression). One possible method to overcome this would be to introduce a larger volume of the virus encoding eYFP in order to increase the chance of co-transfection and drive stronger eYFP expression. The additional labelling of a fourth fluorophore to the analysis (tagBFP) and sparser labelling of ipRGCs across the entire retina would further increase the efficiency of this analysis technique.

Identification of ipRGC subtypes in the rd1 retina

Analysis of our 32 reconstructed cells identified populations that could be categorised into the known subtypes of ipRGC (M1-M5) based on descriptions of their anatomy in the wildtype retina (Berson et al., 2010, Ecker et al., 2010, Estevez et al., 2012). On a subtype specific basis, we find that the soma size, dendritic field size and complexity of the M1 cells isolated in these experiments fall well within the range reported from wildtype OPN4Cre mice (Schmidt et al., 2011b, Estevez et al., 2012, Ecker et al., 2010) and consistent with previous reports that this cell type remain structurally intact following retinal degeneration in rodents (Vugler et al., 2008, Lin and Peng, 2013). One previous report does however identify the dendritic field size of M1 cells as being significantly smaller in the rd1 mouse than in wildtype controls (Damiani et al., 2012). The authors suggest this is due to developmental problems following early onset retinal degeneration, however this is in conflict with other reports that ipRGC number and architecture develop normally in the rd1 mouse (Ruggiero et al., 2009). We did not find such an effect on dendritic field size in our

118

Chapter 3 Identifying ipRGCs in the degenerate retina

identification of the M1 population, or any other ipRGC population, and suggest this could reflect differences in the ages of mice and methodology employed between the two studies. The soma size, dendritic field size and complexity of our M2 population was also in agreement with those reported in the wildtype retina (Berson et al., 2010, Estevez et al., 2012) and those labelled with the anti-melanopsin antibody which is known to label both M1 and M2 subtypes in a previous study of the degenerate retina (Vugler et al., 2008). The identification of the structural integrity of these two cell types additionally acts as an internal control for the accuracy of our isolation and reconstruction of the more recently identified, and therefore less well characterised, ipRGC cell types.

Although rare, we did successfully identify two bi-stratifying cells that relate to the M3 subtype. Whilst we do find variations in the dendritic field size and total dendritic length of these cells compared to the wildtype retina (Figures 3.9A & B), one must be careful when interpreting these data given the very small sample size. There is great variation in anatomical parameters across M3 ipRGCs (Schmidt and Kofuji, 2011). For this reason, it is difficult to draw accurate comparisons until a larger population of M3 cells are isolated and reconstructed from the rd/rd retina. We additionally identify the M5 subtype in the rd/rd retina. This is the most recently identified melanopsin subtype and data remains scarce regarding the quantification of its dendritic architecture and projection targets in visually intact mice (Estevez et al., 2012). Nonetheless we identified a population of cells which possessed small and complex dendritic fields which matched those reported from the wildtype retina.

Finally, our identification of the M4 ipRGCs is consistent with anatomical data (Figure 3.1F&G) and electrophysiological recordings indicating ipRGC input to the dLGN of retinally degenerate mice (Brown et al., 2010), as this projection is thought to arise from M4 subtype. We find both the dendritic field size and complexity of these cells to be in agreement with those reported from the wildtype retina (Estevez et al., 2012) suggesting that this subtype, associated with the image-forming visual system, retains its normal dendritic architecture following loss of outer retinal photoreceptors. This is of importance when considering any potential role melanopsin could play in supporting residual visual function in those afflicted with retinal degeneration and the and current gene therapy strategies employed to restore vision.

M4 ipRGCs possess one of the largest dendritic field sizes ranging from 250µm to 440µm in diameter in the wildtype retina (Estevez et al., 2012, Ecker et al., 2010), and therefore our

119

Chapter 3 Identifying ipRGCs in the degenerate retina

reports are on the lower end of this range in the rd1 mouse, with the largest M4 cell measuring 375µm (Figure 3.8A & B). One possibility is that this difference in dendritic field size reflects the global remodelling of other retinal cell types following outer photoreceptor degeneration (Marc et al., 2007, Jones and Marc, 2005, Strettoi et al., 2003). M4 ipRGCs project to the dLGN, the relay station for image forming vision (Ecker et al., 2010, Brown et al., 2010), and therefore this cell is likely to be strongly influenced by outer photoreceptor input in the wildtype retina. Accordingly, it exhibits a comparatively weak intrinsic light response (Estevez et al., 2012). It is therefore plausable that the loss of this strong rod and cone signal in advanced stages of degeneration could make M4 ipRGCs comparatively more susceptible to changes in their dendritic field than has been observed in other RGC types (Lin and Peng, 2013). However, our data confirming normal dendritic architecture from other ipRGC subtypes, in addition to those from other reports from the rd1 retina (Vugler et al., 2008, Lin and Peng, 2013) indicate that the ipRGC population remains comparatively resilient to this effect. One possibility is that because ipRGCs do not not solely rely on rod and cone photoreception for signalling, but also their own intrinsic light response, they are resistant to the fate that other retinal ganglion cells have been observed to suffer (Jones et al., 2003).

An alternative explanation for the difference in dendritic field size in M4 cells described here for rd/rd mice and elsewhere for wildtype mice is the technical limitation in the size of the ROI which we could process. To ensure accurate reconstruction of cells, we only isolated cells where we could guarantee the entire dendritic field was present in a single ROI. We therefore may have biased our identification and reconstruction to cells from this population who had dendritic field sizes that could be encompassed in one ROI. This limitation can be overcome by increasing the processing power and the memory of our current analysis computer to permit larger ROIs to be analysed. Alternatively, we could run the PCA on multiple ROIs which surround an isolatable cell, but whose dendritic field may extend past that of our ability to process in a single image. We could then use both the colour and spatial relationship between regions of that cell from different sections for reconstruction. One caveat to this approach is that a PCA transformation is unique to the variables of the input image. We could overcome this by using the same parameters (Eigen values) calculated for the original PCA transformation in the surrounding ROIs. If the dendrites of this cell are present in these surrounding sections they should fall within the same vector direction in principal component space. Success could be subsequently

120

Chapter 3 Identifying ipRGCs in the degenerate retina

confirmed during spatial reconstruction of these ROIs. This may also prove a fruitful approach to use for neuronal tracing across and between brain regions.

Whilst we find that our measurements for soma size, dendritic field size and overall complexity of ipRGC subtypes reconstructed here compared to that reported from the wildtype retina (Estevez et al., 2012, Ecker et al., 2010), we do find differences in the total dendritic length of ipRGCs. This likely reflects the higher resolution of our 3D reconstructions which allow for a more accurate description of the length of a cell in all 3 dimensions, which is not currently achieved by tracing 2D projections. However, this can not be unequivocally confirmed until performing our analysis on retinas from visually intact Opn4Cre/+ mice.

In the healthy visual system, ipRGCs modulate a number of aspects of physiology and behaviour and this is reflected in the variation of the somato-dendritic profile of ipRGC subtypes and their numerous projection targets. Following loss of the outer retinal photoreceptors, ipRGCs act as the sole source of light information reaching the brain and therefore our identification of all five subtypes and their normal dendritic architecture complements current literature on the M1 and M2 subtypes surrounding their survival (Lin and Peng, 2013; Vugler et al., 2008), whilst also expanding our understanding of the M3, M4 and M5 subtypes. Our anatomical identification in the retina further compliments electrophysiological (Brown et al., 2010) and behavioural studies (Lucas et al., 2001b, Panda et al., 2003) that this heterogeneous cell type provides an important route through which light can still affect physiology in those afflicted with retinal degeneration, most significantly for the image-forming visual system (Brown et al., 2012, Brown et al., 2010). In doing so, we additionally develop a novel analysis tool which compliments the growing use of multi-colour labelling techniques in neuroscience. These labelling tools are used in species from drosophila to mice, for wide ranging questions from cell lineage tracking and neuronal tracing in health and disease, within and between brain nuclei. Our novel approach therefore has significant implications to provide high-throughput quantifiable data to answer wide ranging scientific question throughout the biological sciences.

121

Chapter 3 Identifying ipRGCs in the degenerate retina

3.6 References

Baver, S. B., G. E. Pickard, P. J. Sollars, and G. E. Pickard, 2008, Two types of melanopsin retinal ganglion cell differentially innervate the hypothalamic suprachiasmatic nucleus and the olivary pretectal nucleus: European Journal of Neuroscience, v. 27. Berson, D. M., A. M. Castrucci, and I. Provencio, 2010, Morphology and Mosaics of Melanopsin-Expressing Retinal Ganglion Cell Types in Mice: Journal of Comparative Neurology, v. 518. Boulina, M., H. Samarajeewa, J. D. Baker, M. D. Kim, and A. Chiba, 2013, Live imaging of multicolor-labeled cells in Drosophila: Development, v. 140, p. 1605-1613. Brown, T. M., C. Gias, M. Hatori, S. R. Keding, M. a. Semo, P. J. Coffey, J. Gigg, H. D. Piggins, S. Panda, and R. J. Lucas, 2010, Melanopsin Contributions to Irradiance Coding in the Thalamo-Cortical Visual System: Plos Biology, v. 8. Brown, T. M., S.-i. Tsujimura, A. E. Allen, J. Wynne, R. Bedford, G. Vickery, A. Vugler, and R. J. Lucas, 2012, Melanopsin-Based Brightness Discrimination in Mice and Humans: Current Biology, v. 22. Cai, D., K. B. Cohen, T. Luo, J. W. Lichtman, and J. R. Sanes, 2013, Improved tools for the Brainbow toolbox: Nature Methods, v. 10, p. 540-+. Damiani, D., E. Novelli, F. Mazzoni, and E. Strettoi, 2012, Undersized dendritic arborizations in retinal ganglion cells of the rd1 mutant mouse: A paradigm of early onset photoreceptor degeneration: Journal of Comparative Neurology, v. 520, p. 1406-1423. Ecker, J. L., O. N. Dumitrescu, K. Y. Wong, N. M. Alam, S.-K. Chen, T. LeGates, J. M. Renna, G. T. Prusky, D. M. Berson, and S. Hattar, 2010, Melanopsin-Expressing Retinal Ganglion- Cell Photoreceptors: Cellular Diversity and Role in Pattern Vision: Neuron, v. 67. Estevez, M. E., P. M. Fogerson, M. C. Ilardi, B. G. Borghuis, E. Chan, S. Weng, O. N. Auferkorte, J. B. Demb, and D. M. Berson, 2012, Form and function of the m4 cell, an intrinsically photosensitive retinal ganglion cell type contributing to geniculocortical vision: The Journal of neuroscience : the official journal of the Society for Neuroscience, v. 32. Fernandez, D. C., Y. T. Chang, S. Hattar, and S. K. Chen, 2016, Architecture of retinal projections to the central circadian pacemaker: Proceedings of the National Academy of Sciences of the United States of America, v. 113, p. 6047-6052. Gupta, V., and K. D. Poss, 2012, Clonally dominant cardiomyocytes direct heart morphogenesis: Nature, v. 484, p. 479-U102. Hadjieconomou, D., S. Rotkopf, C. Alexandre, D. M. Bell, B. J. Dickson, and I. Salecker, 2011, Flybow: genetic multicolor cell labeling for neural circuit analysis in Drosophila melanogaster: Nature Methods, v. 8, p. 260-U111. Hammer, S., A. Monavarfeshani, T. Lemon, J. M. Su, and M. A. Fox, 2015, Multiple Retinal Axons Converge onto Relay Cells in the Adult Mouse Thalamus: Cell Reports, v. 12, p. 1575-1583. Hampel, S., P. Chung, C. E. McKellar, D. Hall, L. L. Looger, and J. H. Simpson, 2011, Drosophila Brainbow: a recombinase-based fluorescence labeling technique to subdivide neural expression patterns: Nature Methods, v. 8, p. 253-U102.

122

Chapter 3 Identifying ipRGCs in the degenerate retina

Hattar, S., M. Kumar, A. Park, P. Tong, J. Tung, K. W. Yau, and D. M. Berson, 2006, Central projections of melanopsin-expressing retinal ganglion cells in the mouse: Journal of Comparative Neurology, v. 497. Hunter, J. D., 2007, Matplotlib: A 2D graphics environment: Computing in Science & Engineering, v. 9, p. 90-95. Jones, B. W., and R. E. Marc, 2005, Retinal remodeling during retinal degeneration: Experimental Eye Research, v. 81, p. 123-137. Jones, B. W., C. B. Watt, J. M. Frederick, W. Baehr, C. K. Chen, E. M. Levine, A. H. Milam, M. M. Lavail, and R. E. Marc, 2003, Retinal remodeling triggered by photoreceptor degenerations: Journal of Comparative Neurology, v. 464, p. 1-16. Lin, B., and E. B. Peng, 2013, Retinal Ganglion Cells are Resistant to Photoreceptor Loss in Retinal Degeneration: Plos One, v. 8. Livet, J., T. A. Weissman, H. N. Kang, R. W. Draft, J. Lu, R. A. Bennis, J. R. Sanes, and J. W. Lichtman, 2007, Transgenic strategies for combinatorial expression of fluorescent proteins in the nervous system: Nature, v. 450, p. 56-+. Lucas, R. J., R. H. Douglas, and R. G. Foster, 2001a, Characterization of an ocular photopigment capable of driving pupillary constriction in mice: Nature Neuroscience, v. 4. Lucas, R. J., M. S. Freedman, D. Lupi, M. Munoz, Z. K. David-Gray, and R. G. Foster, 2001b, Identifying the photoreceptive inputs to the mammalian circadian system using transgenic and retinally degenerate mice: Behavioural Brain Research, v. 125. Lucas, R. J., S. Hattar, M. Takao, D. M. Berson, R. G. Foster, and K. W. Yau, 2003, Diminished pupillary light reflex at high irradiances in melanopsin-knockout mice: Science, v. 299. Marc, R. E., B. W. Jones, J. R. Anderson, K. Kinard, D. W. Marshak, J. H. Wilson, T. Wensel, and R. J. Lucas, 2007, Neural reprogramming in retinal degeneration: Investigative Ophthalmology & Visual Science, v. 48, p. 3364-3371. Nern, A., B. D. Pfeiffer, and G. M. Rubin, 2015, Optimized tools for multicolor stochastic labeling reveal diverse stereotyped cell arrangements in the fly visual system: Proceedings of the National Academy of Sciences of the United States of America, v. 112, p. E2967-E2976. Pan, Y. A., T. Freundlich, T. A. Weissman, D. Schoppik, X. C. Wang, S. Zimmerman, B. Ciruna, J. R. Sanes, J. W. Lichtman, and A. F. Schier, 2013, Zebrabow: multispectral cell labeling for cell tracing and lineage analysis in zebrafish: Development, v. 140, p. 2835-2846. Panda, S., I. Provencio, D. C. Tu, S. S. Pires, M. D. Rollag, A. M. Castrucci, M. T. Pletcher, T. K. Sato, T. Wiltshire, M. Andahazy, S. A. Kay, R. N. Van Gelder, and J. B. Hogenesch, 2003, Melanopsin is required for non-image-forming photic responses in blind mice: Science, v. 301. Ruggiero, L., C. N. Allen, R. L. Brown, and D. W. Robinson, 2009, The development of melanopsin-containing retinal ganglion cells in mice with early retinal degeneration: European Journal of Neuroscience, v. 29, p. 359-367. Schindelin, J., I. Arganda-Carreras, E. Frise, V. Kaynig, M. Longair, T. Pietzsch, S. Preibisch, C. Rueden, S. Saalfeld, B. Schmid, J. Y. Tinevez, D. J. White, V. Hartenstein, K. Eliceiri, P. Tomancak, and A. Cardona, 2012, Fiji: an open-source platform for biological-image analysis: Nature Methods, v. 9, p. 676-682.

123

Chapter 3 Identifying ipRGCs in the degenerate retina

Schmidt, T. M., N. M. Alam, S. Chen, P. Kofuji, W. Li, G. T. Prusky, and S. Hattar, 2014, A Role for Melanopsin in Alpha Retinal Ganglion Cells and Contrast Detection: Neuron, v. 82, p. 781-788. Schmidt, T. M., S.-K. Chen, and S. Hattar, 2011a, Intrinsically photosensitive retinal ganglion cells: many subtypes, diverse functions: Trends in Neurosciences, v. 34. Schmidt, T. M., M. T. H. Do, D. Dacey, R. Lucas, S. Hattar, and A. Matynia, 2011b, Melanopsin-Positive Intrinsically Photosensitive Retinal Ganglion Cells: From Form to Function: Journal of Neuroscience, v. 31, p. 16094-16101. Schmidt, T. M., and P. Kofuji, 2009, Functional and Morphological Differences among Intrinsically Photosensitive Retinal Ganglion Cells: Journal of Neuroscience, v. 29. Schmidt, T. M., and P. Kofuji, 2011, Structure and Function of Bistratified Intrinsically Photosensitive Retinal Ganglion Cells in the Mouse: Journal of Comparative Neurology, v. 519, p. 1492-1504. Semo, M., D. Lupi, S. N. Peirson, J. N. Butler, and R. G. Foster, 2003, Light-induced c-fos in melanopsin retinal ganglion cells of young and aged rodless/coneless (rd/rd cl) mice: European Journal of Neuroscience, v. 18. Storchi, R., N. Milosavljevic, C. G. Eleftheriou, F. P. Martial, P. Orlowska-Feuer, R. A. Bedford, T. M. Brown, M. A. Montemurro, R. S. Petersen, and R. J. Lucas, 2015, Melanopsin-driven increases in maintained activity enhance thalamic visual response reliability across a simulated dawn: Proceedings of the National Academy of Sciences of the United States of America, v. 112, p. E5734-E5743. Strettoi, E., V. Pignatelli, C. Rossi, V. Porciatti, and B. Falsini, 2003, Remodeling of second- order neurons in the retina of rd/rd mutant mice: Vision Research, v. 43, p. 867-877. van der Walt, S., S. C. Colbert, and G. Varoquaux, 2011, The NumPy Array: A Structure for Efficient Numerical Computation: Computing in Science & Engineering, v. 13, p. 22-30. Vugler, A. A., M. Semo, A. Joseph, and G. Jeffery, 2008, Survival and remodeling of melanopsin cells during retinal dystrophy: Visual Neuroscience, v. 25. Weissman, T. A., and Y. A. Pan, 2015, Brainbow: New Resources and Emerging Biological Applications for Multicolor Genetic Labeling and Analysis: Genetics, v. 199, p. 293-306.

124

Chapter 4 Spatial vision in advanced retinal degeneration

Chapter 4: Spatial receptive fields in the retina and dorsal lateral geniculate nucleus (dLGN) of mice lacking rods and cones

This chapter has been published in the Journal of Neurophysiology

125

Chapter 4 Spatial vision in advanced retinal degeneration

4.1 Abstract

In advanced retinal degeneration loss of rods and cones leaves melanopsin-expressing intrinsically photosensitive retinal ganglion cells (ipRGCs) as the only source of visual information. ipRGCs drive non-image forming responses (e.g. circadian photoentrainment) under such conditions but despite projecting to the primary visual thalamus (dorsal lateral geniculate nucleus; dLGN), do not support form vision. We wished to determine what precludes ipRGCs supporting spatial discrimination following photoreceptor loss, using a mouse model (rd/rd cl) lacking rods and cones. Using multielectrode arrays we find that both retinal ganglion cells and neurones in the dLGN of this animal have clearly delineated spatial receptive fields. In the retina, they are typically symmetrical, lack inhibitory surrounds, and have diameters in the range 10-30° of visual space. Receptive fields in the dLGN were larger (diameters typically 30-70°), but matched the retinotopic map of the mouse dLGN. Injections of a neuroanatomical tracer (cholera toxin β subunit) into the dLGN confirmed that retinotopic order of ganglion cell projections to the dLGN and thalamic projections to the cortex is at least superficially intact in rd/rd cl mice. However, as previously reported for deafferented ipRGCs, onset and offset of light responses have long latencies in the rd/rd cl retina and dLGN. Accordingly, dLGN neurons failed to track dynamic changes in light intensity in this animal. Our data reveal that ipRGCs can convey spatial information in advanced retinal degeneration and identifies their poor temporal fidelity as the major limitation in their ability to provide information about spatial patterns under natural viewing conditions.

126

Chapter 4 Spatial vision in advanced retinal degeneration

4.2 Introduction

In mammals with advanced retinal degeneration, melanopsin-expressing ipRGCs survive to support non-image forming (NIF) light-responses such as circadian photo-entrainment and the pupillary light reflex (PLR) (Freedman et al., 1999, Lucas et al., 2003b, Berson et al., 2002). These NIF responses rely upon measuring background light levels rather than detecting spatial patterns. However, melanopsin is not restricted to those retinal ganglion cells (termed M1 ipRGCs) innervating NIF brain centres, but is expressed also in several classes whose anatomy and physiology imply a role in form vision (Ecker et al., 2010, Brown et al., 2010, Schmidt et al., 2014). Spatial receptive fields have been described for ipRGCs in the rodent retina and are shown, in some cases, to include centre-surround antagonism (Wong et al., 2007, Estevez et al., 2012), with their projection patterns encompassing the primary visual thalamus (Ecker et al., 2010, Brown et al., 2010).

These findings imply that ipRGCs could support form vision in some patients with outer retinal degeneration. However, at present there has been no direct demonstration that ipRGCs contribute to spatial information in the dLGN. In fact, many more dLGN neurones show evidence of melanopsin influence than can be accounted for by a simple one-to-one input from ipRGCs (Brown et al., 2010). A possible explanation for this is that central pathways sample across many ipRGCs to provide individual dLGN neurones with a signal of mean light intensity across the visual scene. Determining the spatial extent of visual information provided by ipRGCs to individual dLGN neurones would therefore help to resolve this issue as well.

In assessing the potential relevance of ipRGCs to form vision during progressive retinal degeneration, a second question is how these photoreceptors are impacted by outer retinal dystrophy. ipRGCs certainly survive, with broadly normal anatomy, following severe outer retinal dystrophy (Semo et al., 2003, Vugler et al., 2008). It remains unclear, however, whether their functional receptive fields are degraded by the impact such conditions have on the remaining inner retinal circuitry (Marc et al., 2007, Strettoi and Pignatelli, 2000). Similarly, ipRGCs drive light responses in the dLGN in advanced retinal degeneration, but we do not know whether this signal allows individual dLGN neurons to respond to light from particular points in visual space. Finally, it is uncertain whether the projection of ipRGCs (or indeed the ganglion cell population in general) to the dLGN retains retinotopic order under these conditions.

Here, we set out to address these questions by recording melanopsin-evoked electrophysiological responses to spatially structured stimuli in the retina and dLGN of retinally degenerate mice. The model we chose for this work, C3H rd/rd cl mice, carries the rd1 loss of function mutation in the gene encoding the phosphodiesterase β6 subunit (PDEB6) that abolishes rod phototransduction and causes rod, and subsequent cone, cell death. Surviving cones are additionally targeted by the cl transgene that drives expression of the diphtheria toxin alpha subunit in this cell type. In these animals, rods and cones degenerate rapidly post-natally such that, by 80 days, ipRGCs are the sole

127

Chapter 4 Spatial vision in advanced retinal degeneration

surviving photoreceptors (Freedman et al., 1999, Lucas et al., 1999, Lucas et al., 2001b). Here, we find that ipRGCs retain spatial receptive fields in the retina, and retinotopic order to their projection to the dLGN in this animal. Within the dLGN, we find neurones with functional receptive fields over a range of spatial scales. These data indicate that the remaining ipRGC photoreceptors have the fundamental ability to provide spatial information in advanced retinal degeneration. However, our data further demonstrate that the poor temporal fidelity of the ipRGC light response following rod and cone loss represents a substantial limitation to employing this capacity to support form vision.

128

Chapter 4 Spatial vision in advanced retinal degeneration

4.3 Methods In-vitro electrophysiology Five male rd/rd cl mice were sacrificed by cervical dislocation and immediately enucleated. Retinal isolation was performed in carboxygenated (95% O2 / 5% CO2) aCSF (artificial cerebro-spinal fluid, concentration in mM: 118 NaCl, 25 NaHCO3, 1 NaH2PO4, 3 KCl, 1 MgCl2, 2 CaCl2, 10 C6H12O6, 0.5 L- Glutamine; Sigma Aldritch, UK). The retina was incised 4 times in a Maltese cross motif and mounted onto a 256 channel Multi Electrode Array (256MEA200/30iR-ITO; Multichannel Systems GmbH, Germany) with the ganglion cell layer facing down onto the electrodes. A Cyclopore membrane filter (5μm pores; Whatman, USA) held the retina in place whilst being weighed down by a stainless steel anchor (~0.75g) bearing a framework of parallel polyimide coated fused silica capillaries (TSP320450, Polymicro Technologies, USA). Electrophysiological signals were acquired using MC_Rack software (Multi Channel Systems, Germany) through a USB-MEA256 amplifier (for 256 channel recordings; Multi Channel Systems, Reutlingen, Germany). Recordings were made at 25 kHz sampling frequency during the acquisition of electrophysiological activity. To preserve physiological conditions, the tissue was perfused with carboxygenated aCSF at 2.2ml/minute and maintained at 32˚C (TC01 controller; Multi Channel Systems, Reutlingen, Germany).

Light stimuli were projected onto the retina’s ganglion cell layer from below. Full-field light stimuli (melanopsin irradiance = 4.21x1014 photons/cm2/s) were delivered by a custom written LabView program instructing an Arduino (Arduino Due, Italy) to control a Phlatlight LED. Spatiotemporal stimuli were delivered as 5° or 10° vertical or horizontal bars (which spanned 150µm or 300µm on the retina, respectively) for 60 seconds (7.91x1013 melanopsin photons/cm2/s), with a 180 second inter-stimulus interval (2.25x1012 melanopsin photons/cm2/s) by a custom written Python script (PsychoPy) instructing an Arduino and a polarising LCD projector system (HoloEye Photonics AG, Germany).

In-vivo electrophysiology Eight adult C3H rd/rd cl mice (5 male and 3 female; 80-400 days old) were administered with an initial dose of 0.125% Chlorprothixene Hydrochloride (0.5mg/kg; Sigma-Aldrich, UK) prior to being anaesthetised with 2% Isoflurane in O2. Mice were mounted onto a bespoke stereotaxic frame (SG- 4N-S, Narishige, Japan) which was fixed onto a ‘lazy Susan’ (RBB12A, Thorlabs, Germany). Isoflurane anaesthesia (0.4 - 1.0% maintenance) was administered via a nose cone (GM-4, Narishige, Japan) and body temperature was maintained at 37°C with a homeothermic blanket (Harvard Apparatus, Kent, UK). An incision to expose the skull surface was made and a small hole (~1 mm diameter) drilled 2.3 mm posterior and 2.3 mm lateral to the bregma, targeting the dorsal LGN. The pupil, contralateral to the craniotomy, was dilated with topical atropine sulphate (1% w/v; Sigma-Aldrich, UK) and the cornea kept moist with mineral oil. The ipsilateral eye remained covered with blackout

129

Chapter 4 Spatial vision in advanced retinal degeneration

material throughout the experiment. A recording probe (A4X8-5 mm-50-200-413; Neuronexus, MI, USA) consisting of four shanks (spaced 200 µm apart), each with eight recordings sites (spaced 50 µm apart) was then positioned centrally on the exposed surface in the coronal plane, and lowered to a depth of 2.5 - 3.3mm using a fluid filled micromanipulator (MO-10; Narishige, Japan). Once the recording probe was in position mice were dark adapted for 30 minutes, allowing neuronal activity to stabilise after probe insertion. Neural signals were acquired using a Recorder64 system (Plexon Inc, TX; USA). Signals were amplified x3000, high-pass filtered at 300 Hz and digitized at 40 kHz. Multiunit activity (spikes with amplitudes >50µV) were saved as time-stamped waveforms and analysed offline (see data analysis).

Six rd/rd cl mice were used for in-vivo receptive field mapping. Light stimuli were generated in MATLAB (The Mathworks Inc., MA, USA) and controlled by a laptop running PsychoPy V2.6 (Peirce, 2008). Stimuli were presented via a 46-inch back-lit LED TV monitor (DS46LO4, Dynascan Technology Inc., CA; USA). The monitor was at a fixed distance of 30cm away from the eye contralateral to the recording probe and angled 20° towards the frame. Light responsive cells were confirmed by the presentation of a 60 second uniform white stimulus (melanopsin irradiance = 8.72x1014 photons/cm2/s) repeated 10 times, with a 120 second inter-stimulus interval. For receptive field mapping, stimuli consisted of a 60 second pre-stimulus dark screen before the 60 second presentation of a vertical blue stripe (melanopsin irradiance = 2.56 x 1014 photons/cm2/s) spanning 10° of the mouse’ visual field. This was followed by a 120 second inter-stimulus interval. The mouse was then rotated in 10° increments and the light protocol repeated such that the stimulus appeared to move from 150° behind the mouse to 30° into the contralateral visual field. This completed one full rotation, with each rotation repeated a minimum of 4 times. Mice were otherwise kept in complete darkness.

When recording responses to temporal modulations in irradiance from two rd/rd cl mice, light stimuli were generated using a bespoke light source (Cairn Research, Faversham, UK) consisting of a blue LED (λmax = 460nm) which passed through neutral density filters (Cairn Research) and focused onto a 5mm diameter circle of opal diffusing glass (Edmund Optics Inc., York, UK) placed <1mm from the eye contralateral to the recording probe. Placing the diffuser so it was centred to the midpoint of the eye meant that light would be distributed evenly across the retina. Programmes written in LabVIEW (Version 8; National Instruments) controlled a National Instruments card (USB-6229) which permitted sinusoidal modulations of the blue LED. This provided a 92% Michaelson contrast for melanopsin (mean irradiance = 1.24x1015photons/cm2/s) at four temporal frequencies - 10Hz, 1Hz, 0.1Hz and 0.01Hz. All light measurements were measured using a calibrated spectroradiometer (Bentham Instruments, Reading, UK). Effective photon flux for the melanopsin photopigment was determined using the calculated spectra and visual pigment template described by Govardovskii et al. (Govardovskii et al., 2000).

130

Chapter 4 Spatial vision in advanced retinal degeneration

Tracer Injections

2 rd/rd cl mice (434-450 days old) were anaesthetised with 1% isoflurane in O2 and a micropipette (NanojectII; Drummond Scientific, UK) fitted with a pulled glass capillary (20μm diameter tip) was used to inject Cholera Toxin β subunit conjugated to either Alexa 488 or Alexa 555 (Invitrogen; Cat #: C-34775 or C-34776) into the medial (2.4mm posterior and 2.1mm lateral to the bregma) and lateral (2.4mm posterior and 2.3mm lateral to the bregma) regions of the dorsal LGN. A trail of 50nl injections were administered in 50µm steps spanning the entire dorsal LGN. Brains and retinas were collected between 5-7 days after injection.

Histology At the end of electrophysiological experiments mice were transcardially perfused with 0.1M phosphate buffered saline (PBS) followed by 4% paraformaldehyde. The brain was removed, post- fixed overnight, cryoprotected with 30% sucrose and sectioned at 100µm on a freezing sledge microtome. Sections were mounted with DPX (Sigma-Aldrich, UK) and coverslipped. Electrode placement in the dLGN was confirmed by visualisation of a fluorescence dye (Cell Tracker CM-DiI; Invitrogen Ltd. Paisley, UK) applied to the probe prior to recording. Images were collected on an Olympus BX51 upright microscope using a 4x/ 0.30 Plan Fln, and captured using a Coolsnap ES camera (Photometrics) through MetaVue Software (Molecular Devices). Specific band pass filters set for DAPI, FITC and Texas red prevented bleed through of channels. Following tracer studies, mice were killed with an overdose of 30% urethane before being transcardially perfused and brains processed as above. Additionally, both eyes were removed and the retinas immediately dissected in the dark in order to maintain their correct dorsal-ventral and nasal-temporal axis. Retinas were left overnight in 4% PFA before being mounted onto glass slides and cover slipped with Prolong Diamond antifade mountant (Life Technologies, UK).

Data analysis Multichannel, multiunit recordings were analysed in Offline Sorter (Plexon Inc, TX; USA). Following removal of cross-channel artefacts, principal component based sorting was used to discriminate single units, identifiable as a distinct cluster of spikes in principal component space (Figure 4.1A & D) with a clear refractory period (>1ms) in their interspike interval distribution (Figure 4.1C & F). Following spike sorting, data was exported to Neuroexplorer (Nex technologies, MA, USA) and MATLAB R2013a (The Mathworks Inc., MA, USA) for construction of peristimulus histograms and further analysis. Light responsive units were identified as those where the peristimulus average showed a clear peak that exceeded the 99% confidence limits estimated from a Poisson distribution derived from the prestimulus spike counts.

131

Chapter 4 Spatial vision in advanced retinal degeneration

Figure 4.1 Identification of single units in the retina and dLGN of the rd/rd cl mouse. Representative scatter plots of spike waveforms plotted in a 3D representation of principal component space (X axis = PC1; Y axis = PC2; Z axis = PC3) for (A) a retinal and (D) a dLGN recording. A single unit is readily identified in both preparations (single unit = yellow, unsorted waveforms = grey). We use Manova F and P-values as an indicator of the degree of cluster separation (retina: F = 27.7; p = 9.6-13; LGN: F = 7.3; P = 0.006). (B, E) Mean Spike waveform for the isolated single units (black trace) and unsorted waveforms (grey trace) from the units identified in A and D. These two units have a spike amplitude of 192µV and 152µV, respectively. Log inter-spike interval (ISI) for (C) retinal and (F) dLGN recordings display the single unit in black and unsorted waveforms in grey. Isolated single units show clear sharp peaks at discrete intervals greater than 1ms, indicative of spikes fired in bursts. The unsorted waveforms show activity across a much wider range of ISIs which is indicative of tonic firing of multiple cells which could not be isolated. The average (mean±S.E.M) spike waveform for the 28 single dLGN units = 82.9µV ± 3.75 and spike width at the half maximum of the valley = 0.19s ± 0.003.

132

Chapter 4 Spatial vision in advanced retinal degeneration

Receptive field diameters were analysed by subtracting the average change in firing rate during the 60 second pre-stimulus baseline from the average firing rate during the 60 second light-pulse to calculate the average change in firing rate. This was normalised to the peak change during each rotation, for each light-responsive channel, to account for changes in the absolute amplitude of the light response over the long duration of the experiment. Some preparations had periods in which light responses disappeared for a while, to account for this effect, we excluded from analysis any repeat at which there was no response to light at any spatial location. A Gaussian function was applied in Graphpad Prism 6 (GraphPad software Inc.) from which the receptive field centre and size could be determined. A 90% confidence limit (±1.64 standard deviations from the mean) was applied to the best-fit Gaussian in order to calculate receptive field size.

To generate the retinotopic map, DiI tracts from individual mice were used to determine the placement and depth of the final position of the recording electrode from which the position of individual recording channels could be identified. The mean of the Gaussian fit for multi-unit activity (as described above) was correlated with the anatomical position of the corresponding light- responsive channel in the dLGN. This was mapped onto a common LGN template at -2.06mm posterior from the bregma and a 2D heat map was generated by a custom written programme in MATLAB. Temporal modulations in irradiance were analysed using a custom written chi square based periodogram programme in MATLAB with a significance value of 0.001.

133

Chapter 4 Spatial vision in advanced retinal degeneration

4.4 Results

We first sought to characterise the receptive field of individual ipRGCs in the rd/rd cl retina using in- vitro electrophysiology. We presented a bar of polychromatic light occupying 5 or 10 degrees of visual space in 10 or 20 locations spanning the multi-electrode array (which itself covered 3mm2 of the retina or 100 degrees of visual space in each direction). While the majority of recording sites across the multielectrode arrays failed to record responses to this stimulus in this preparation, we found a small number at which reliable changes in firing were apparent for at least one spatial location of the bar (64 out of 1280 multiunit traces from 5 rd/rd cl preparations). Spike sorting of these data resolved 31 consistently responsive single units. A couple of representative response profiles are presented in Figure 4.2. All recorded responses were excitatory (average peak change in firing rate = 4.15 spikes/s ± 0.56). As previously reported for melanopsin driven responses, the rd/rd cl light responses was characterised by sluggish kinetics; with a long latency to respond, followed by a gradual increase in firing throughout the duration of the light pulse (mean time to peak = 43.3s ± 3.3), and a slow return to baseline after lights off (Figure 4.2A).

134

Chapter 4 Spatial vision in advanced retinal degeneration

Figure 4.2 Spatial receptive field properties of two representative ipRGCs recorded from the rd/rd cl retina (A) Peristimulus time histogram (PSTH) demonstrating the response of a representative light responsive ganglion cell to the presentation of a 60 second, 10 degree vertical bar, in its preferred spatial position over four repeats (time bin = 1s). (B) Trial bin count (yellow bar represents lights on at 0s and off at 60s) demonstrating the reproducibility of the melanopsin light-response to spatial bar position over 4 repeats of the receptive field mapping protocol (time bin = 5s; colour code = spikes/s) and (C) the best fit Gaussian (R2 = 0.993) of the relationship between normalised response amplitude and spatial position of the vertical bar which reveals a receptive field diameter of 13.3° (dotted vertical line denotes edge of calculated receptive field size). (D) Trial bin count showing the reproducibility of a second representative unit mapped in 2 dimensions using horizontal (top) and vertical (bottom) 10 degree bars (time bin = 5s; colour code = spikes/s) over 4 repeats spanning 100° of visual angle. (E) The best fit Gaussian function of the relationship between normalised response amplitude and spatial position of the vertical bar (top panel; RF size = 38.2°; R2 = 0.80) and horizontal bar (bottom panel; RF size = 33.6°; R2 = 0.99). (F) A 2 Dimensional heat map (colour code = response amplitude in delta firing rate normalised to maximum response to horizontal and vertical bar presentation) displaying the circular receptive field of this unit when mapped on a 100 degree region of visual space. In all cases stimulus location is described as distance in both μm and ° of visual space from the edge of the array.

We found that in all cases RGC responses were strongly dependent upon the spatial location of light presentation. This is apparent in the relationship between response amplitude and position of the light bar in the two representative units presented in Figure 4.2B & D. In all 31 units, the peak response was elicited by bars away of the edge of the MEA, allowing us to approximate the receptive field by fitting a Gaussian function to at least one limb (Figure 4.2C & E). Receptive field diameter was estimated by applying a 90% confidence limit (±1.64 standard deviations from the mean) as measured from a best fit Gaussian applied to the relationship between bar position and the normalised response amplitude. This demonstrated that these receptive fields ranged in size from 8° to 50° (mean receptive field diameter = 22.3° ± 1.9) of visual angle; although most fell between 15° and 25° (Figure 4.3A). 18 of these units were mapped in two dimensions and we found that for these cells receptive fields were roughly circular, with differences in receptive field diameter estimates in each dimension typically being no more than 8% of the mean (Figure 4.2F). In order to confirm that our estimates of receptive field size were not greatly overestimated by our choice of 10° bar widths, three of the single units were mapped using 5° bars. We found that these three units fell well within the range of cells mapped with a 10° bar, again showing roughly circular receptive fields with diameters of 21.3°, 15.0° and 29.8° of visual angle.

Although our study did not set out to systematically assess the impact of ageing on receptive field structure, we did include a range of ages in this work. A comparison of receptive field sizes between relatively young (3-6 month) and older (greater than 6 month) mice indicated no very dramatic change in this parameter (Figure 4.3A). Given the fragile nature of the rd/rd cl retina it was not always possible to keep track of its orientation with respect to dorso-ventral and temporo-nasal axes during dissection. However, this was achieved in three preparations. The active recording sites from these preparations were biased towards the ventral retina (which should have allowed us to sample from all ipRGC types (Hughes et al., 2013) but across this region the receptive field centre of light responsive units appeared fairly evenly distributed and there was no strong relationship between receptive field size and retinal eccentricity (Figure 4.3B).

135

Chapter 4 Spatial vision in advanced retinal degeneration

Figure 4.3 Receptive field properties of ipRGC population recorded from the rd/rd cl retina (A) Population data displaying receptive field size of 31 single units recorded from 5 rd/rd cl retinas. Receptive field diameter ranged from 8.8 - 49.9° of visual angle (n = 31 units; mean±STD receptive field size = 22.3° ± 10.4). (B) Plot of receptive field location for 18 single units projected onto map of retinal surface from 3 rd/rd cl mice. Lines depict the extent of receptive fields for individual neurones in horizontal axis (in μm of retinal surface); scale bar and indication of retinal orientation are provided to the right.

Some ipRGCs send axon collateral back into the retina (Joo et al., 2013). If these were to drive spiking activity in target neurones, we might record melanopsin evoked responses in locations very distant from that of light exposure. In fact we always found that the strongest light responses occured when the light bar fell above the electrode from which the activity of that unit was recorded.

136

Chapter 4 Spatial vision in advanced retinal degeneration

We next turned our attention to the dLGN. We first identified melanopsin driven responses by presenting a full field white light pulse to isoflurane anaesthatised rd/rd cl mice. In common with previous reports we found slow but sustained increases in firing in response to this stimulus at around 37% of channels (71/192 light-responsive channels). We next applied a receptive field mapping protocol analagous to that used for the retina in-vitro. In brief, 10° vertical bars were presented for 60 seconds at 19 locations spanning 180° of the visual field with an inter-stimulus interval of 240 seconds. In pilot experiments we found that such very long timescales were required to ensure robust responses. However, this meant that each repeat of the receptive field mapping protocol lasted 2hrs 15 mins.

Of these 71 light-responsive channels, 58 multi unit traces within the LGN responded reliabiliy to at least one spatial location of the bar across multiple repeats. From the 58 channels showing reproducible responses to the spatial stimuli, we isolated 28 single units for detailed receptive field analysis. The response of two such units is shown in Figure 4.4. The response of these cells when the bar was in the centre of their receptive field had the poor temporal resolution expected for melanopsin-driven activity (n=28 cells; average time to peak firing = 32.5 ± 2.0 seconds; Figure 4.4A). In all cases, cells responded to at least two spatial locations of the bar on multiple repeats (Figure 4.4B), allowing us to approximate receptive fields by fitting Gaussian functions to the relationship between response amplitude and bar position (Figure 4.4C). We confirmed that large receptive fields (such as that in Figure 4.4C; bottom panel) were not an artefact of poor spike sorting by calculating the mean spike waveform evoked during bar presentation at both the centre (above bar position 100°) and edges (above bar position 50° & 150°) of the receptive field and found these to be consistent throughout our analysis.

137

Chapter 4 Spatial vision in advanced retinal degeneration

Figure 4.4 Receptive field properties of two representative melanopsin driven neurones in the rd/rd cl dLGN (A) Peristimulus time histograms (PSTH) demonstrating the average melanopsin light-response of two representative units with a narrow (top) or wide (bottom) receptive field in the dLGN when presented with a 60 second 10° vertical bar in its preferred spatial location (time bin = 1s). (B) Trial bin counts demonstrating the reproducibility of this melanopsin response to spatial bar position over 4 repeats of the receptive field mapping protocol (time bin = 5s; colour code = spikes/s). (C) The best fit Gaussian of the relationship between normalised response amplitude and spatial position of the vertical bar (top panel, R2 = 0.98; bottom panel R2 = 0.80) demonstrating a small and large receptive field size of 29° and 89° respectively. The average spike waveform evoked during the 60s bar presentation is shown for bars at the edge of the large receptive field (bar position 50° and 150°) as well as the peak response (bar position 100°).

Across the population of responsive units receptive field sizes ranged from 18.2° - 92.6° of visual angle (Figure 4.5A). Receptive field centres were distributed across a wide range of bar positions (Figure 4.5B). Across this range, there was no obvious relationship between the location and size of the receptive fields. Interestingly, none of the units had receptive field centres in the most nasal 60° sampled (although given the large receptive field size that is not to say that we failed to record responses for bars at these locations). One possible explanation for this is that our electrode placements targeting the central dLGN had failed to adequately sample very rostral portions of the nucleus receiving strongest input from the temporal retina.

138

Chapter 4 Spatial vision in advanced retinal degeneration

Figure 4.5 Receptive field properties of melanopsin-driven responses in the rd/rd cl dLGN (A) The distribution of receptive field widths in the horizontal axis for 28 single units recorded from the dLGN of 6 rd/rd cl mice. Receptive field diameter ranged from 18.2 - 92.6° of visual angle (mean±STD receptive field size = 56.6 ± 22.0°). (B) The relationship between the diameter (in horizontal axis) and location in visual space of the receptive fields from 28 single units. Single units recorded from each of the 6 different rd/rd cl mice are presented in a single colour. The location of receptive field centres is provided in ° of visual angle, using a reference of 0° for directly in front of the mouse. Note that few receptive fields are centred in the nasal part of the visual field and that there is no clear correlation between receptive field position and receptive field size.

The implication would be that the remaining visual input to the rd/rd cl dLGN retains some retinotopic order. To explicitly investigate this possibility, we used histological techniques to estimate the location of the electrodes from which these responses had been recorded (Figure 4.6A). In order to maximise the sample size for this analysis, we turned to multiunit activity, which allowed inclusion of responses from units that could not be unambiguously spike sorted. Correlating the receptive field centre with the anatomical location of the light-responsive channel revealed a smooth gradient from nasal to temporal visual field across a ventro-medial to dorso-lateral axis of the dLGN (Figure 4.6B). This relationship exists both within individual rd/rd cl mice and when combining the single or multi-unit receptive fields of all 6 animals, and is consistent with previous

139

Chapter 4 Spatial vision in advanced retinal degeneration

reports for visually intact animals. We continued to demonstrate the retention of retinotopic order by injecting cholera toxin β-subunit conjugated to either Alexa 488 or Alexa 555 into the medial and lateral portions of the dLGN, respectively (Figure 4.6C). Here, from two injected mice we find the retrograde transport of these tracers preferentially labels distinct portions of the temporal and nasal retina as predicted given the retinotopic organisation of the mouse dLGN (Figure 4.6D). We also found Alexa 488 labelling in the most temporal regions of the ipsilateral retina, which is in agreement with the injection site of this tracer spanning the binocular zone of the mouse dLGN (data not shown). In one mouse, we also found that these two tracers preferentially labelled distinct locations of the medial and lateral regions of V1 and in V2M indicating that retinotopic order remains at least partially intact following retinal degeneration (Figure 4.6E & F).

140

Chapter 4 Spatial vision in advanced retinal degeneration

141

Chapter 4 Spatial vision in advanced retinal degeneration

Figure 4.6 Retinotopic organisation of the rd/rd cl visual system (A) Reconstructed positions of the recording sites superimposed upon schematic representatives of the dLGN extent in coronal sections through the mouse brain atlas at -1.82mm & -2.18mm relative to the bregma. Reconstruction was based on DiI labelling of the electrode shank. Coloured dots (orange) depict recording sites from which visually responsive multiunit traces were recorded. Brain schematics based on Paxinos and Franklin Mouse Atlas (Paxinos and Franklin, 2004), solid lines depict boundaries of dorsal and ventral LGN, dotted line the intergeniculate leaflet. (B) A 2 Dimensional heat map correlating anatomical position of electrodes in the dLGN with position of receptive fields in the visual scene for multiunit activity recorded at that site. A gradient from ventro-medial to dorso-lateral dLGN matches the progression from nasal to ventral visual fields (n = 58 multiunit traces within the LGN from 6 rd/rd cl mice). (C; top panel) Cholera toxin β-subunit injections into the medial (Alexa488; green) and lateral (Alexa555; red) regions of the dLGN and (C; bottom panel) overlay of injection pattern onto mouse atlas to illustrate the extent of cholera toxin labelling in the dorsal LGN. (D) Wholemount retina contralateral to the injected dLGN shown in (C) demonstrating distinct regions of labelling with Cholera toxin β-subunit conjugated to Alexa 488 in the nasal retina and Alexa555 in the temporal retina. (D; bottom panel) Magnified image of Cholera toxin labelled retinal ganglion cell bodies from nasal (green highlighted box in D; top panel) and temporal (red highlighted box in D; top panel) retina. (E; left) Regions of V1 and V2ML (3.5mm posterior to the bregma) in the ipsilateral hemisphere to the dLGN injection site demonstrate distinct regions of labelling with both Cholera Toxin β- subunit conjugates and (E; right) magnified image highlighting labelled cell bodies in lateral and medial V2M by Cholera toxin β-subunit conjugates Alexa488 and 555, respectively. (Higher spatial and colour resolution images are available in the digital version of this thesis)

The retention of spatial order in the visual information reaching the dLGN of rd/rd cl mice raises the possibility of melanopsin-driven light responses in this animal supporting spatial discrimination. However, two aspects of the responses we record would substantially reduce the utility of this information under natural viewing conditions. Firstly, we find that the magnitude and timing of melanopsin-driven responses varies substantially between presentations. This is apparent in the activity of a representative unit across multiple repeats of a 60 second full field pulse over half an hour (Figure 4.7A). A second, most substantial problem is their very poor temporal fidelity. Increases in firing rate build up and decay very slowly and therefore do not track light increments and decrements very accurately (Brown et al., 2010) ; Figure 4.4A and 4.7 A & B). Moreover, in order to record the data presented here we have had to include very long inter-stimulus intervals, indicating very gradual recovery from prior light exposure. To further explore the implication of these characteristics for the ability of melanopsin to track dynamic changes in light intensity in these retinally degenerate animals we finally recorded the response to full field sinusoidal modulations (92% Michelson contrast) over a wide range of temporal frequencies - 10Hz, 1Hz, 0.1Hz and 0.01Hz. Across 22 light response channels (or 16 single units isolated from them), we find no significant modulation in their firing rate (p > 0.001; chi square periodogram) which demonstrated that these cells could not track sinusoidal modulations in irradiance at any of the temporal frequencies investigated.

142

Chapter 4 Spatial vision in advanced retinal degeneration

Figure 4.7 Temporal fidelity of the melanopsin-driven light-responses in the rd/rd cl dLGN Trial bin count (A) and corresponding peristimulus time histogram (PSTH; B) of a representative melanopsin-driven light response in the rd/rd cl dLGN (time bin = 5s; colour code = spikes/s) reveals poor temporal resolution and great variation in response latency, peak response amplitude and recovery time over 10 repeats of a 60s light pulse (inter- stimulus interval = 120s; time bin = 5s). (C) Mean±SEM firing rate for a representative light responsive unit across sinusoidal modulations in irradiance (1.24x1015 photons /cm2/s) at four different temporal frequencies - 10Hz, 1Hz, 0.1Hz and 0.01Hz (100 repeats; 20 repeats for 0.01Hz; time bin = 100 data points per period for each temporal frequency; Chi square periodogram p=1.00). Data are double plotted, with the associated change in irradiance shown with the solid line above. No unit showed a significant modulation in firing at any frequency (Chi square periodogram p>0.001). (D) This is also apparent in the mean (±SEM) firing rate across all 22 multiunit traces from the LGN of 2 rd/rd cl mice exposed to sinusoidal modulations at these temporal frequencies (p>0.001).

143

Chapter 4 Spatial vision in advanced retinal degeneration

4.5 Discussion

Using in-vitro and in-vivo extracellular electrophysiological recordings, we mapped spatial receptive fields in the retina and dLGN of rd/rd cl mice. In each method, we found that receptive fields are large but clearly defined and cover the mouse visual field. Furthermore, these responses adopt a functional retinotopic organisation in the dLGN. This argues that low resolution spatial discrimination could be possible from melanopsin-driven light-responses in the rd/rd cl mouse. However, the very poor temporal fidelity of melanopsin-driven light response in this model represents a substantial barrier to realising that potential.

In the intact retina, spatial receptive fields have been described for individual ipRGCs in the rat and mouse (Wong et al., 2007); (Zhao et al., 2014a, Estevez et al., 2012). Receptive field diameter was found to vary between ipRGC subtypes, being around 500µm for M1 cells, and ranging from 200 - 500µm for non-M1 types. These values are in agreement with the anatomical estimates of dendritic field diameters of ipRGCs (149 - 477µm) in wildtype retinae (Schmidt and Kofuji, 2009, Berson et al., 2010, Ecker et al., 2010, Estevez et al., 2012). Of interest is that the non-M1 ipRGCs, which are believed to target the dLGN, were found to possess centre-surround receptive fields whilst M1 ipRGCs, which predominantly target NIF centres, do not (Zhao et al., 2014a). Given that surround antagonism originates in the circuitry linking rods/cones to ganglion cells, it is unsurprising that we did not observe centre-surround elements of receptive fields in either the retina or dLGN of rd/rd cl mice. This would though, impact the quality of spatial information provided by ipRGCs in advanced retinal degeneration.

In other respects, our data are broadly consistent with data from wild type mice. In our experiments in the rd/rd cl retina we find receptive field diameter was commonly between 15 - 25° of visual angle, corresponding to between 450 - 750µm in diameter on the retina (mean±SEM diameter = 668µm ± 56). This is consistent with estimates of ipRGC dendritic field size and morphological evidence that outer retinal degeneration does not greatly affect the integrity of the ipRGC population (Semo et al., 2003, Vugler et al., 2008). However, a small proportion of retinal ganglion cells (6/31) possessed receptive fields which were more than double that predicted for a single ipRGC with even the largest dendritic field. What could result in these larger diameter receptive fields? One possibility is that a melanopsin-driven light response propagates through gap junctions creating a localised spatial network of excitability in the retina. ipRGCs make gap junction connections with displaced Amacrine cells in the Ganglion cell layer (Bramley et al., 2011, Schmidt et al., 2008, Sekaran et al., 2003) which can relay this information to a heterologous population of retinal ganglion cells (Muller et al., 2010). This is supported by strongly correlated firing of RGCs via gap junctions which has been reported in the rd1 retina (Toychiev et al., 2013, Borowska et al., 2011). If that were the case then we might expect factors that modulate gap junction coupling in the

144

Chapter 4 Spatial vision in advanced retinal degeneration

inner retina to further improve the quality of spatial information provided by ipRGCs in advanced retinal degeneration.

We found that the receptive fields of melanopsin-driven LGN neurones are substantially larger than those recorded from ipRGCs in the retina. In the rd/rd cl, an LGN neurone’s receptive field size would be defined by the number of ipRGCs from which it receives input (directly or indirectly) and the dendritic field of each of these ipRGCs. The latter parameter would vary between subtypes of ipRGC. Nevertheless, a simple comparison between the average receptive field diameter of light responsive units in the rd/rd cl retina (diameter = 22°; area = 380°2) and dLGN diameter = 56° area = 2463°2) suggests that each light responsive dLGN unit is influenced on average by at least six light responsive retinal ganglion cells with non-overlapping receptive fields. However, there must be great variation in this value on the individual unit level. Thus, if the dLGN unit with the largest receptive field (diameter = 92°; area = 6500°2 if symmetrical) received its input from those retinal units with the smallest receptive fields (diameter = 9°; area = 64°2) then it must be downstream of at least 100 ganglion cells (even assuming that the ganglion cells influencing it tile the retina perfectly). Conversely, we find dLGN units whose receptive field diameters are within the range of those found in the retina indicating that they could be downstream from a single light responsive retinal ganglion cell. Current literature reports that between one and four RGCs synapse with a single neurone in the wildtype LGN (Chen and Regehr, 2000, Hooks and Chen, 2008). Our data do not directly address the question of whether this arrangement is altered in retinal degeneration. Although, many of our light responsive dLGN units clearly lie downstream from more than four RGCs, this need not reflect direct synaptic input, as within the timescales of response there is ample opportunity for the visual information reaching them to be routed via other neurons in the thalamus or elsewhere in the brain.

An important determinant of the potential contribution of ipRGCs to spatial vision is whether their projections match the retinotopic order of other ganglion cell terminals in the dLGN. Estevez et al (Estevez et al., 2012), reported that a tracer injected into the medial dLGN in wild type mice labelled ipRGCs in the dorsal retina, indicating that they do. Our functional characterisation of visual responses in the dLGN is consistent with this view. Thus, we find that the distribution of receptive fields in visual space across the dLGN of rd/rd cl mice matches that recently described for conventional RGC input in wild types (Piscopo et al., 2013). This is supported by our Cholera toxin tracer study which demonstrates that at least some order is retained in a heterologous population of RGCs projecting from the retina to the dLGN in this model when compared to the known retinotopic organisation of the wildtype dLGN (Piscopo et al., 2013) and visual cortex (Schuett et al., 2002b). To our knowledge, ours is the first report of functional retinotopy in the dLGN following retinal degeneration. The fact that it is broadly retained is encouraging for the potential of therapies currently under development aimed at restoring vision in patients with advanced retinal degeneration. These are largely focussed on providing visual information to the surviving retinal

145

Chapter 4 Spatial vision in advanced retinal degeneration

ganglion cells and will rely upon those targeted neurons making appropriate connections within the brain.

Although the receptive field sizes of neurons in the rd/rd cl dLGN are large, the fact that they exist and adopt a retinotopic order implies that they could support low acuity spatial vision. In fact, although there is evidence that melanopsin can support visual discrimination in advanced retinal degeneration (Zaidi et al., 2007), rd/rd (and rd/rd cl) mice perform extremely poorly in even the most rudimentary spatial navigation tasks (Brown et al., 2012, Thyagarajan et al., 2010). The very poor temporal fidelity of the melanopsin driven response in this genotype provides a straightforward explanation for that finding. The very long latency for changes in firing following both appearance and disappearance of light bars here are consistent with many previous reports of the light response of deafferented ipRGCs (Berson et al., 2002, Do et al., 2009). To explore how that might impact their ability to track dynamic modulations in local luminance, we presented full field sinusoidal oscillations in light intensity over a wide range of temporal frequencies. We were unable to find any units in the rd/rd cl dLGN that convincingly tracked sinusoidal oscillations over any temporal frequency. One interesting question is whether this is a fundamental limit to melanopsin’s capacity or, in part, a consequence of retinal degeneration. The kinetics of the melanopsin single photon response are poor (Do et al., 2009) but neither onset nor decay latencies approach those observed here for dLGN responses in rd/rd cl mice. The reasons for this discrepancy are unclear. However, kinetics of the melanopsin driven responses in animals with an intact visual system more closely match those of the single photon response, suggesting that the very sluggish activity observed here may be a consequence of rod/cone loss. In this context, our data revealing discrete receptive fields and retinotopic order to melanopsin-driven responses in the dLGN has a couple of interesting implications. Firstly, the fundamental capacity to provide spatial information these data reveal might be translated into a genuine contribution to pattern vision in intact animals and at earlier stages of degeneration. Secondly, any method of addressing the poor temporal kinetics of the melanopsin-driven response in retinal degeneration might represent a viable strategy for restoring crude spatial discrimination.

146

Chapter 4 Spatial vision in advanced retinal degeneration

4.6 References

Berson, D. M., A. M. Castrucci, and I. Provencio, 2010, Morphology and Mosaics of Melanopsin- Expressing Retinal Ganglion Cell Types in Mice: Journal of Comparative Neurology, v. 518. Berson, D. M., F. A. Dunn, and M. Takao, 2002, Phototransduction by retinal ganglion cells that set the circadian clock: Science, v. 295. Borowska, J., S. Trenholm, and G. B. Awatramani, 2011, An Intrinsic Neural Oscillator in the Degenerating Mouse Retina: Journal of Neuroscience, v. 31, p. 5000-5012. Bramley, J. R., E. M. Wiles, P. J. Sollars, and G. E. Pickard, 2011, Carbenoxolone Blocks the Light- Evoked Rise in Intracellular Calcium in Isolated Melanopsin Ganglion Cell Photoreceptors: Plos One, v. 6. Brown, T. M., C. Gias, M. Hatori, S. R. Keding, M. a. Semo, P. J. Coffey, J. Gigg, H. D. Piggins, S. Panda, and R. J. Lucas, 2010, Melanopsin Contributions to Irradiance Coding in the Thalamo-Cortical Visual System: Plos Biology, v. 8. Brown, T. M., S.-i. Tsujimura, A. E. Allen, J. Wynne, R. Bedford, G. Vickery, A. Vugler, and R. J. Lucas, 2012, Melanopsin-Based Brightness Discrimination in Mice and Humans: Current Biology, v. 22. Chen, C. F., and W. G. Regehr, 2000, Developmental remodeling of the retinogeniculate synapse: Neuron, v. 28, p. 955-966. Do, M. T. H., S. H. Kang, T. Xue, H. Zhong, H.-W. Liao, D. E. Bergles, and K.-W. Yau, 2009, Photon capture and signalling by melanopsin retinal ganglion cells: Nature, v. 457. Ecker, J. L., O. N. Dumitrescu, K. Y. Wong, N. M. Alam, S.-K. Chen, T. LeGates, J. M. Renna, G. T. Prusky, D. M. Berson, and S. Hattar, 2010, Melanopsin-Expressing Retinal Ganglion-Cell Photoreceptors: Cellular Diversity and Role in Pattern Vision: Neuron, v. 67. Estevez, M. E., P. M. Fogerson, M. C. Ilardi, B. G. Borghuis, E. Chan, S. Weng, O. N. Auferkorte, J. B. Demb, and D. M. Berson, 2012, Form and function of the m4 cell, an intrinsically photosensitive retinal ganglion cell type contributing to geniculocortical vision: The Journal of neuroscience : the official journal of the Society for Neuroscience, v. 32. Freedman, M. S., R. J. Lucas, B. Soni, M. von Schantz, M. Munoz, Z. David-Gray, and R. Foster, 1999, Regulation of mammalian circadian behavior by non-rod, non-cone, ocular photoreceptors: Science, v. 284. Govardovskii, V. I., N. Fyhrquist, T. Reuter, D. G. Kuzmin, and K. Donner, 2000, In search of the visual pigment template: Visual Neuroscience, v. 17, p. 509-528. Hooks, B. M., and C. Chen, 2008, Vision triggers an experience-dependent sensitive period at the retinogeniculate synapse: Journal of Neuroscience, v. 28, p. 4807-4817. Hughes, S., T. S. Watson, R. G. Foster, S. N. Peirson, and M. W. Hankins, 2013, Nonuniform Distribution and Spectral Tuning of Photosensitive Retinal Ganglion Cells of the Mouse Retina: Current Biology, v. 23, p. 1696-1701. Joo, H. R., B. B. Peterson, D. M. Dacey, S. Hattar, and S.-K. Chen, 2013, Recurrent axon collaterals of intrinsically photosensitive retinal ganglion cells: Visual Neuroscience, v. 30, p. 175-182. Lucas, R. J., M. S. Freedman, D. Lupi, M. Munoz, Z. K. David-Gray, and R. G. Foster, 2001, Identifying the photoreceptive inputs to the mammalian circadian system using transgenic and retinally degenerate mice: Behavioural Brain Research, v. 125. Lucas, R. J., M. S. Freedman, M. Munoz, J. M. Garcia-Fernandez, and R. G. Foster, 1999, Regulation of the mammalian pineal by non-rod, non-cone, ocular photoreceptors: Science, v. 284. Lucas, R. J., S. Hattar, M. Takao, D. M. Berson, R. G. Foster, and K. W. Yau, 2003, Diminished pupillary light reflex at high irradiances in melanopsin-knockout mice: Science, v. 299.

147

Chapter 4 Spatial vision in advanced retinal degeneration

Marc, R. E., B. W. Jones, J. R. Anderson, K. Kinard, D. W. Marshak, J. H. Wilson, T. Wensel, and R. J. Lucas, 2007, Neural reprogramming in retinal degeneration: Investigative Ophthalmology & Visual Science, v. 48, p. 3364-3371. Muller, L. P. d. S., M. T. H. Do, K.-W. Yau, S. He, and W. H. Baldridge, 2010, Tracer Coupling of Intrinsically Photosensitive Retinal Ganglion Cells to Amacrine Cells in the Mouse Retina: Journal of Comparative Neurology, v. 518. Paxinos, G., and K. Franklin, 2004, The Mouse Brain in Stereotaxic Coordinates (2nd Edition), Gulf Professional Publishing. Peirce, J. W., 2008, Generating Stimuli for Neuroscience Using PsychoPy: Frontiers in neuroinformatics, v. 2, p. 10-10. Piscopo, D. M., R. N. El-Danaf, A. D. Huberman, and C. M. Niell, 2013, Diverse Visual Features Encoded in Mouse Lateral Geniculate Nucleus: Journal of Neuroscience, v. 33, p. 4642-4656. Schmidt, T. M., N. M. Alam, S. Chen, P. Kofuji, W. Li, G. T. Prusky, and S. Hattar, 2014, A Role for Melanopsin in Alpha Retinal Ganglion Cells and Contrast Detection: Neuron, v. 82, p. 781-788. Schmidt, T. M., and P. Kofuji, 2009, Functional and Morphological Differences among Intrinsically Photosensitive Retinal Ganglion Cells: Journal of Neuroscience, v. 29. Schmidt, T. M., K. Taniguchi, and P. Kofuji, 2008, Intrinsic and extrinsic light responses in melanopsin-expressing ganglion cells during mouse development: Journal of Neurophysiology, v. 100. Schuett, S., T. Bonhoeffer, and M. Hubener, 2002, Mapping retinotopic structure in mouse visual cortex with optical imaging: Journal of Neuroscience, v. 22, p. 6549-6559. Sekaran, S., R. G. Foster, R. J. Lucas, and M. W. Hankins, 2003, Calcium imaging reveals a network of intrinsically light-sensitive inner-retinal neurons: Current Biology, v. 13. Semo, M., D. Lupi, S. N. Peirson, J. N. Butler, and R. G. Foster, 2003, Light-induced c-fos in melanopsin retinal ganglion cells of young and aged rodless/coneless (rd/rd cl) mice: European Journal of Neuroscience, v. 18. Strettoi, E., and V. Pignatelli, 2000, Modifications of retinal neurons in a mouse model of retinitis pigmentosa: Proceedings of the National Academy of Sciences of the United States of America, v. 97, p. 11020-11025. Thyagarajan, S., M. van Wyk, K. Lehmann, S. Lowel, G. P. Feng, and H. Wassle, 2010, Visual Function in Mice with Photoreceptor Degeneration and Transgenic Expression of Channelrhodopsin 2 in Ganglion Cells: Journal of Neuroscience, v. 30, p. 8745-8758. Toychiev, A. H., E. Ivanova, C. W. Yee, and B. T. Sagdullaev, 2013, Block of Gap Junctions Eliminates Aberrant Activity and Restores Light Responses during Retinal Degeneration: Journal of Neuroscience, v. 33, p. 13972-13977. Vugler, A. A., M. Semo, A. Joseph, and G. Jeffery, 2008, Survival and remodeling of melanopsin cells during retinal dystrophy: Visual Neuroscience, v. 25. Wong, K. Y., F. A. Dunn, D. M. Graham, and D. M. Berson, 2007, Synaptic influences on rat ganglion- cell photoreceptors: Journal of Physiology-London, v. 582. Zaidi, F. H., J. T. Hull, S. N. Peirson, K. Wulff, D. Aeschbach, J. J. Gooley, G. C. Brainard, K. Gregory- Evans, J. F. Rizzo, III, C. A. Czeisler, R. G. Foster, M. J. Moseley, and S. W. Lockley, 2007, Short- wavelength light sensitivity of circadian, pupillary, and visual awareness in humans lacking an outer retina: Current Biology, v. 17, p. 2122-2128. Zhao, X., B. K. Stafford, A. L. Godin, W. M. King, and K. Y. Wong, 2014, Photoresponse diversity among the five types of intrinsically photosensitive retinal ganglion cells: Journal of Physiology- London, v. 592, p. 1619-1636.

148

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

Chapter 5: Delayed maturation of ipRGCs projecting to the dorsal Lateral Geniculate Nucleus (dLGN)

149

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

5.1 Abstract

During advanced retinal degeneration, the progressive loss of the rod photoreceptors, followed by the slower decline of the cone photoreceptor population, leaves intrinsically photosensitive retinal ganglion cells (ipRGCs) as the only available source of visual information. ipRGCs survive outer retinal degeneration and continue signalling light information to the dorsal Lateral Geniculate Nucleus (dLGN) but exhibit poor spatio-temporal resolution in the absence of rod and cone photoreceptor input. However, isolated pockets of cones can survive in the outer retina and therefore during early stage retinal degeneration vision can rely on both residual cone photoreceptors and melanopsin to provide visual perception. Here, we combine extracellular multichannel in-vivo electrophysiological recordings and photoreceptor silent substitution to investigate the contribution of cones and melanopsin to light-responses in the dLGN of mice with early stage retinal degeneration.

In doing so, we identify widespread cone-driven light responses in the dLGN under both light and dark adapted conditions that show significant impairment in both their amplitude and latency. We continue to demonstrate that LGN neurones in rd1 mice retain contrast sensitivity over a modest contrast range and possess spatial receptive fields which are not dissimilar to that of their age matched controls. We find under these experimental conditions, there is at best a modest contribution of melanopsin signalling to these residual light-responses using our silent substitution stimuli in both the retinally degenerate and wildtype mice at the ages employed in this study. However we do find a significant and robust contribution of melanopsin signalling to the OPN at equivalently aged mice, and to the adult dLGN. Supported by anatomical data, this suggests a specific temporal delay in the maturation of ipRGCs which project to the dLGN; however we cannot rule out that this is an experimental artefact of studying visual responses in juvenile mice.

150

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

5.2 Introduction

Inherited retinal degenerations, such as retinitis pigmentosa, are the most common cause of visual blindness in humans owing to the progressive loss of rod and cone photoreceptors. A subset of retinal ganglion cells expresses the photopigment melanopsin rendering them intrinsically photosensitive (Berson et al., 2002, Hattar et al., 2002) but also receive extensive input from rod and cone photoreceptors (Wong et al., 2007, Perez-Leon et al., 2006). These melanopsin expressing RGCs (ipRGCs) survive in advanced retinal degeneration with broadly normal retinal anatomy (Vugler et al., 2008) and support non image-forming (NIF) responses such as the pupillary light reflex (Lucas et al., 2003b) and circadian entrainment (Freedman et al., 1999). However, melanopsin is also expressed in a subset of ipRGCs which project to the dLGN (Ecker et al., 2010). These cells can continue signalling to a large proportion of dLGN neurones at advanced stages of degeneration when no rods and cones are present (Brown et al., 2010), but exhibit poor spatio-temporal resolution (Procyk et al., 2015). These characteristics of melanopsin signalling are at least in part a result of retinal degeneration and preclude the ability of deafferented ipRGCs providing behaviourally relevant spatial vision.

However, at early stages of degeneration, vision can rely upon melanopsin as well as residual cone activity. In the rd1 mouse model of retinitis pigmentosa, the retina undergoes well-defined stages of cell death and re-organisation (Jones and Marc, 2005). The rod photoreceptor population initially undergoes rapid cell death by Post-natal day 18 (P18) and this is shortly followed by the decline of the cone photoreceptor population by around P30. However, isolated pockets of cones can survive into the later stages of degeneration (Jimenez et al., 1996, Ogilvie et al., 1997), mirroring some human conditions. These residual cone-driven responses show deficits in their synaptic connectivity and physiology even at this early stage (Strettoi et al., 2002, Stasheff, 2008). Nonetheless, this begs the question as to whether at these earlier stages of vision loss; can melanopsin make a more significant contribution to spatial vision?

One of the difficulties when investigating the contribution of cone and melanopsin to residual light- responses in the dLGN is the spectral sensitivity of the different photopigments expressed in the mouse retina. To circumvent this problem, we created a new double transgenic mouse model which we term the C57 rd1 Opn1mwR mouse. This mouse carries the rd1 loss of function mutation in the gene encoding the phosphodiesterase β6 subunit (PDEβ6) that abolishes rod phototransduction and causes rod, and subsequent cone, cell death. Additionally, these mice have the human L-cone opsin knocked-in place of their natural occurring medium wavelength sensitive (MWS) opsin, red-shifting the spectral sensitivity of these cone photoreceptors (Smallwood et al., 2003). Using extracellular multi-channel in-vivo electrophysiology, this mouse model allows us to take a psychophysics approach of photoreceptor silent substitution so that we can provide contrast to individual photopigments whilst keeping others silent (Allen and Lucas, 2016). This allows us to tease apart the

151

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

contribution of melanopsin and residual cones to light-responses in the dLGN during early stage retinal degeneration.

In doing so, we find a variety of light-responsive units throughout the dLGN up to 5 weeks of age which respond under both dark and light adapted conditions. These cells respond to moderate contrasts and possess spatial receptive fields; however both the temporal kinetics and amplitude of these residual light-responses are significantly impaired as previously reported in the rd1 retina. We continue to look at the contribution of melanopsin phototransduction to light responses in the dLGN of 4 week old mice and find there are a low proportion of cells compared to the adult which demonstrated the sustained phenotype associated with melanopsin signalling. We find under these experimental conditions, there is at best a modest contribution of melanopsin signalling in either degenerate or wild type mice at these young ages. As the stimuli used to probe melanopsin responses drive robust responses in the adult dLGN, and in the Olivary Pretectal Nucleus (OPN) of 4 week old mice, we conclude that there is a specific developmental delay in the appearance of melanopsin input to the dLGN, however we cannot rule out that this is an experimental artefact of studying light-responses in juvenile mice.

152

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

5.3 Methods Animal Housing Mice were bred at the University of Manchester and housed under a 12:12 light/dark cycle, with food and water available ad libitum. rd1 Opn1mwR were created in house by crossing an established colony of Opn1mwR mice with commercially available C57 rd1 mice (Stock Number 000659; Jackson Laboratories). Note that Opn1mwR refers to the transgenic allele originally generated by Smallwood et al. (2003), and termed simply ‘R’ by them. For all electrophysiological experiments, Opn1mwR and rd1 Opn1mwR were used between 3 - 5 weeks of age, except for adult recordings from the dLGN of Opn1mwR mice where mice were between 3-4 months of age. All procedures conformed to requirements of the UK Animals (Scientific Procedures) Act, 1986.

In-vivo electrophysiology Six juvenile C57 rd/rd Opn1mwR mice, eight juvenile Opn1mwR (all 3-5 weeks old) and 5 adult Opn1mwR (3-4 months old) were administered with 20% Urethane (1.6mg/kg; i.p.). Once anaesthetised, mice were mounted onto a bespoke stereotaxic frame (SG-4N-S, Narishige, Japan) which was fixed onto a ‘lazy Susan’ (RBB12A, Thorlabs, Germany). Core body temperature was maintained at 37°C with a homeothermic blanket (Harvard Apparatus, Kent, UK).

An incision to expose the skull surface was made and a small hole (~1 mm diameter) drilled 2.2 mm posterior and 2.2 mm lateral to the bregma, targeting the dorsal LGN. For OPN recordings, the drill site was made 2.8mm posterior to the bregma and 0.8mm lateral of the midline. A recording probe (A4X8-5 mm-50-200-413; Neuronexus, MI, USA) consisting of four shanks (spaced 200 µm apart), each with eight recordings sites (spaced 50 µm apart) was then positioned centrally on the exposed surface in the coronal plane, and lowered to a depth of 2.5 - 3.3mm to target the dLGN, and 2.0 - 2.5mm to target the OPN, using a fluid filled micromanipulator (MO-10; Narishige, Japan). Once the recording probe was in position mice were dark adapted for 30 minutes, allowing neuronal activity to stabilise after probe insertion. The pupil, contralateral to the craniotomy, was dilated with topical atropine sulphate (1% w/v; Sigma-Aldrich, UK) and the cornea kept moist with mineral oil. The ipsilateral eye remained covered with blackout material throughout the entire experiment. Neural signals were acquired using a Recorder64 system (Plexon Inc, TX; USA). Signals were amplified x3000, high-pass filtered at 300 Hz and digitized at 40 kHz. Multiunit activity (spikes with amplitudes >50µV) were saved as time-stamped waveforms and analysed offline (see data analysis).

153

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

Presentation of visual stimuli Light stimuli were generated in MATLAB (The Mathworks Inc., MA, USA) and controlled by a laptop running PsychoPy V2.6 (Peirce, 2008). Light stimuli were presented via a commercially available projection system which had been modified so that each of the R, G and B channels was a combination of up to five, independently controlled wavelengths (λmax = 405, 455, 525, 561, 630nm). This allowed us to present patterned stimuli that only present spatial/temporal contrast for particular photopigments. As such we could create three metameric pairs of stimuli: “all photoreceptor” (S-cone Opsin = 51%, melanopsin = 52%, rod opsin = 34%, L-cone Opsin, 47%); “Mel- less” (S-cone opsin = 49%, melanopsin = <5%, rod opsin = 30% L-cone opsin = 51%); and “Mel- isolating” (S-cone opsin = 1%, melanopsin = 48%, rod opsin = 5% , L-cone opsin =-2%). All light measurements were measured using a calibrated spectroradiometer (Bentham Instruments, Reading, UK). Effective photon flux for each photopigment was determined using the calculated spectra and visual pigment template described by Govardovskii et al. (Govardovskii et al., 2000).

Visual Stimuli Dark adapted responses: At the beginning of each experiment we presented 200ms full field flashes (irradiance = 2.50x1014 photons cm-2 s-1) from darkness with a 1 second inter-stimulus interval (ISI) for 50 repeats. We additionally presented 10s light-steps from darkness to the same irradiance with an ISI of 50 second over 20 repeats to identify the sustained component of light-responses.

Contrast sensitivity: Full field 1s light-steps, with a 5 second inter stimulus interval, were presented at eight increasing cone contrasts (1%, 2%, 5%, 16%, 20%, 30%, 40% and 50%) from a light adapted background (irradiance = 2.64x1014 photons cm-2 s-1). Each sequence was repeated 20 times in an interleaved manner using the “all-photoreceptor” stimulus settings.

Receptive field mapping: Vertical bars (occupying ~13° of the visual field; irradiance 1.04x1014 photons cm-2 s-1) from a background (irradiance =1.55x1013 photons cm-2 s-1) were used to map receptive fields of dLGN neurons. Vertical bars were presented for 250ms in a pseudorandom order in 13 (overlapping) spatial locations (4.5° separation in bar position; ISI = 1.25 seconds). The spectra used for these spatial stimuli did not elicit significant responses in the rd1 population. However, as these mice do not possess functional rods, we were able to generate a new spectral transition starting from lower background irradiance. This allowed us to present bars with a larger calculated Michaelson contrast for both S- and L-cone opsins. Spatial receptive fields for rd1 mice were mapped under these new settings.

Silent Substitution Steps: We initially presented full field transitions (4Hz) between our three pairs of silent substitution stimuli: all-photoreceptor, mel-less and mel-isolating. The stimulus spectra was adjusted every 50 repeats in order to validate each stimulus condition and ensure we did not

154

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

observe transient responses to the “mel-isolating” steps. Following this, full field 10s light steps from a light adapted background were presented 20 times with a 50 second inter-stimulus interval under the all-photoreceptor, mel-less and mel-isolating stimulus conditions. Stimuli were presented in a pseudorandom order in order to determine the contribution of activating both cones and melanopsin together, cones in isolation, and melanopsin in isolation.

Data Analysis Offline, neural waveforms were processed using Offline Sorter (version 2.8.8; Plexon Inc. USA). Cross-channel artefacts were identified and removed, and then each channel analysed separately. For each channel, single-unit spikes were detected and categorised based on the spike waveform via a principal component analysis, whereby distinct clusters of spikes were readily identifiable and showed a clear refractory period in their interspike interval distribution (>1ms). Single-units and unclassified waveforms were also combined to form multi-unit data for each channel. Single and multi-unit data were subsequently sent to NeuroExplorer (version 4.032; Nex Technologies, MA, USA) and MATLAB R2010a (The Mathworks Inc.) to further analyse changes in firing rate of single units in response to different visual stimuli.

Identification of light responses: In the dark adapted state, responses were classed as light- responsive if the firing rate during stimulus presentation exceeded 2 standard deviations of the mean baseline firing rate prior to light exposure. Presentation or the 10s light-step under the dark adapted state allowed us to categorise cells based on their light-response profile. Accordingly, single units were defined as Transient ON if they demonstrated significant change in firing rate after light- onset which quickly returned to baseline during the light pulse. Transient ON-OFF cells also showed an initial increase in firing rate at light onset before quickly returning to baseline, however showed a second significant increase in firing immediately after light-offset. Sustained-ON and Sustained-OFF cells were categorised if a significant increase or decrease in firing rate was maintained for more than 5 seconds of a 10 second light-step, respectively. Under light-adapted conditions, single units were categorised based on their response to the “all-photoreceptor” condition.

Contrast Sensitivity Analysis: Single units were checked to ensure that the firing rate at the maximum cone contrast (50%) demonstrated a significant change in firing rate which was greater than 2 standard deviations above the pre-stimulus baseline. If this criterion was met, the response of that unit at the seven lower contrasts was used for analysis regardless of whether it crossed the confidence interval. Contrast sensitivity curves were calculated by subtracting the pre-stimulus baseline from the average firing rate over the first 500ms of the 1 second light step.

155

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

Spatial Receptive field analysis: Single units were initially checked to ensure that they demonstrated a significant change in firing rate to at least one bar position over 90 repeats of the stimulus sequence. The spatial receptive field size was estimated by fitting a 2-Dimensional Gaussian fit (R2 value greater than 0.7) to the relationship between response amplitude and bar position in Graphpad Prism 6 (GraphPad software Inc.). The receptive field size for individual cells was described as 1 standard deviation of the best-fit Gaussian.

Silent Substitution Analysis: Single units were first classified as sustained or transient based on their response to a 10s light-step under the all photoreceptor condition. Single units were classified as sustained if they maintained their change in firing rate greater than two standard deviations above baseline for more than 5s over the course of the 10s light step. Comparisons between the average firing rates over the last 5s of the light pulse in the all photoreceptor and mel-less conditions in both genotypes was used to determine the contribution of melanopsin signalling to the dLGN and were analysed using 2 Way ANOVA. The contribution of the melanopsin to the mel-isolating condition was determined by calculating the change in firing rate of the last 5 seconds of the light step compared to the pre-stimulus baseline.

Tissue preparation Following electrophysiological recordings, mice were transcardially perfused with 0.9% saline followed by cold 4% methanol-free paraformaldehyde (Sigma Aldrich; UK). The brain was removed and post-fixed overnight in 4% paraformaldehyde, prior to cryoprotection for 24 hours in 30% sucrose in 0.1M PBS. 100μm coronal sections were cut using a sledge microtome (, mounted onto glass slides and cover slips were applied using Vectashield (Vector Laboratories, Inc.). Electrode placement in the dLGN was confirmed by visualisation of a fluorescence dye (Cell Tracker CM-DiI; Invitrogen Ltd. Paisley, UK) applied to the probe prior to recording. Images were collected on an Olympus BX51 upright microscope using a 4x/ 0.30 Plan Fln, and captured using a Coolsnap ES camera (Photometrics) through MetaVue Software (Molecular Devices). Specific band pass filters set for DAPI, FITC and Texas red prevented bleed through of channels.

Immunohistochemistry In some experiments, the mouse was also enucleated following transcardial perfusion in order to label retinas for the melanopsin protein. The eye cup was post-fixed in 4% methanol-free PFA (Sigma Aldrich; UK) overnight before cryoprotection for 24 hours in 30% sucrose solution. Retinas were sectioned on a cryostat (CM3050 Series; Leica) to a thickness of 20µm. Immunohistochemistry for the melanopsin protein was performed on the slide for retinal sections. In brief, sections were initially permeabilized with 1% PBS-X for one hour before being blocked in 10% donkey serum

156

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

(Sigma Aldrich, UK) for 3 hours at room temperature. Sections were subsequently incubated in primary antibody (rabbit anti-melanopsin; 1:2000; UF006; ABCAM) with 2.5% donkey serum in 1% PBS-X for 2 hours at room temperature after which they were washed thoroughly in 1% PBS-X. Sections were then incubated in a secondary antibody of Alexa-546 conjugated donkey anti-rabbit (1:200; Life technologies; UK) for 2 Hours at room temperature after which they were once more thoroughly washed in 1% PBS-X before undergoing one final wash in distilled water. Once dry, retinal sections were mounted with Prolong Gold Diamond anti-fade media containing DAPI (Invitrogen; UK), coverslipped, and kept in the dark until imaging. For retinal wholemounts, the retina was dissected from the eye cup and processed free floating. The same protocol was used, however, the primary antibody solution was left to incubate overnight at room temperature and secondary antibody mixture was incubated overnight at 4°C. Retinal wholemounts were mounted with Prolong Gold Diamond anti-fade media (Invitrogen; UK) without DAPI.

Imaging Retinal wholemounts were imaged on a Leica upright stereoflourescence microscope (M205 FA; Leica) at x40 magnification under a Texas Red filter and stitched together in 2-Dimensions using the MosaicJ plugin in FIJI (Schindelin et al., 2012) . Images of retinal sections were collected on a Leica TCS SP5 AOBS inverted confocal microscope using a 20x / 0.50 Plan Fluotar objective (2x confocal zoom). The confocal settings were as follows: pinhole 1 airy unit, scan speed 1000Hz bidirectional, format 1024x1024. Images were collected using the following detection mirror settings; DAPI 494- 530nm and Texas red 602-665nm using the 405nm (100%) and 546nm (100%) and 633nm (100%) laser lines respectively. To eliminate cross-talk between channels, the images were collected sequentially. During 3D acquisition of optical stacks the confocal software was used to determine the optimal number of Z sections to be 1µm. Only the maximum intensity projections of these 3D stacks are shown in the results.

157

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

5.4 Results

We first set out to identify residual cone driven responses in the dLGN of young rd1 Opn1mwR mice using extracellular multi-channel recording electrodes (Figure 5.1A). We presented full field 200ms flashes (2.50x1014 photons cm-2 s-1) from darkness and spike sorting this data resolved 157 light- responsive units from eight Opn1mwR mice and 84 light-responsive units from seven rd1 Opn1mwR mice. The average number of light-responsive units per electrode placement in both genotypes was similar (Figure 5.1B; unpaired T-test = 0.32), however the range of light-responsive units per experiment was larger. We therefore next plotted the number of light-responses as a function of age and found that in the rd1 animals there was a significant decrease in light-responses with increasing post-natal age (slope = -1.35; p = 0.003; Figure 5.1C). Nonetheless, during the ages recorded here, reconstruction of the anatomical position of light-responsive channels demonstrated that light- responses could be observed across the entire extent of the dorsal LGN (Figure 5.1D).

We next presented 10s full field steps (irradiance = 2.50x1014 photons cm-2 s-1) from darkness and could categorise light-responses into four groups: Transient ON cells, Transient ON-OFF cells, Sustained ON cells and Sustained OFF cells. Transient-ON cells show an initial increase in firing rate at light onset but quickly return to baseline throughout the duration of the stimulus (Figure 5.1E; top row). Transient ON-OFF cells show a transient increase in firing at both light onset and offset (Figure 5.1E; second row). Sustained ON cells demonstrate an initial increase in firing rate at light onset and remained elevated above baseline throughout the duration of the light stimulus (Figure 5.1E; third row). Conversely, Sustained OFF cells show a reduction in firing rate throughout the duration of the light stimulus (Figure 5.1E; bottom row).

In wildtype Opn1mwR mice we found 24% of light-responsive units demonstrated a transient-ON response to light onset whilst 23% demonstrated a transient ON-OFF response. Sustained-ON cells accounted for nearly half of all light-responsive units (45%) whilst sustained-OFF cells were much rarer only accounting for 8% of the total population. Comparatively, light-responses in rd1 Opn1mwR mice were far more transient in nature, with 65% of light-responsive units demonstrating a transient ON response to light onset whilst a further 15% demonstrated a Transient ON-OFF response. The remaining 20% of cells demonstrated a sustained ON response. We did not find sustained OFF responses in the rd1 population. We first set out to examine the ON transient component of these responses from the Transient ON, Transient ON-OFF and Sustained ON populations. We found that the amplitude of this response in the Opn1mwR population was significantly larger (11.79±0.67 Spikes/s) compared to rd1 Opn1mwR mice (10.24±0.97 Spikes/s) (Figure 5.1F; p = 0.0312; unpaired T- test). Furthermore, the time to reach peak response was significantly longer in the rd1 Opn1mwR population (207.4ms ± 4.6) compared to age matched controls (158.1ms ± 4.9) (Figure 5.1G; p<0.0001; unpaired T-test). We finally looked at the power spectrum density PSD upon presentation of these 10s light steps. We found that this population of light-responsive cells demonstrated a

158

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

robust oscillation peaking at 31.3±0.39Hz in Opn1mwR mice (Figure 5.1H). We found no significant peaks in the power spectrum in rd1 Opn1mwR mice between 0 and 50 Hz.

Figure 5.1 Dark adapted light-responses in the rd1 Opn1mwR dLGN. (A) Representative image of DiI labelled electrode tract from an rd1 Opn1mwR mouse used to confirm electrode placement in the dLGN (dotted line). (B) Mean ± S.E.M number of light-responsive single units were found to be similar between rd1 Opn1mwR (10.75 ± 1.9) and Opn1mwR mice (13.09 ± 1.3; p = 0.32, Unpaired T-test). (C) Plotting the number of light-responsive units per electrode placement as a function of age demonstrated a significant decrease in light-responsive units with age in the rd1 Opn1mwR population (slope = -1.35; p = 0.003) compared to Opn1mwR mice (slope = 0.67; p = 0.06). (D) Representative anatomical reconstruction of light-responsive channels found throughout dorso- ventral and medial-lateral extent of the rd1 Opn1mwR dLGN using a 4x8 configuration extracellular multichannel recording electrode. (E) Single unit light-responses could be categorised as Transient or Sustained in

159

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

response to a 10s light-step (irradiance =2.50x1014 photons cm-2 s-1) from darkness. Transient cells could be further subdivided in transient ON and transient ON-OFF responses to light whilst sustained cells demonstrated a sustained ON or sustained OFF response to light. (% indicates percentage of cells of the total population that demonstrate each light-response profile). (F) Histogram of peak response amplitude to the ON component of the light-response demonstrates that single units in the Opn1mwR dLGN showed significantly larger response (11.79 ± 0.6 Spikes/s) compared to the rd1 Opn1mwR population (10.24 ± 0.9 Spikes/s; p = 0.03, unpaired T- test). (G) Time to peak response was significantly smaller for the Opn1mwR population (158.1 ± 4.9ms) compared to the rd1 Opn1mwR population (207.4 ± 4.5ms; p < 0.0001, unpaired T-test). (H) Normalised Power Spectrum Density (PSD) of light-responsive units during a 10s light pulse demonstrates that a robust peak can be identified in the Opn1mwR population (26.7 ± 0.4 Hz) but no discernible peak in the in the rd1 Opn1mwR population.

This data confirmed that we could record light-responses to stimuli of infinite contrast (from darkness) in rd1 mice. We next wanted to further interrogate these fast cone-driven responses under more naturalistic light-adapted conditions. Accordingly, we presented 1 second light-steps from a background (irradiance = 2.64x1014 photons cm-2 s-1) with eight increasing cone contrasts ranging from 1 - 50%. We identified 61 units that showed a significant change in firing rate following light onset at the highest contrast in the Opn1mwR population and 54 units from the rd1 Opn1mwR population. The mean ± S.E.M PSTH at each contrast are shown in Figure 5.2A. Plotting the average change in firing rate over the first 500ms after light onset demonstrated that the rd1 Opn1mwR mice had a significantly reduced change in firing rate across this contrast range for all but the lowest two cone contrasts (Figure 5.2B). Normalising these changes in firing rate to the maximum response amplitude in each genotype demonstrated that these cells do retain contrast sensitivity similar to that of their Opn1mwR controls as both data can be fit by the same dose-response curve (Figure 5.2C; R2 = 0.94; p = 0.34, F-test).

160

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

Figure 5.2 Contrast sensitivity in the rd1 dLGN (A) Mean ± S.E.M. peristimulus time histograms (PSTH) of light- responsive units in the dLGN of Opn1mwR (green trace, n = 61 units) and rd1 Opn1mwR (black trace; n= 54 units) in response to 20 repeats of a 1 second light-step at eight increasing cone contrasts (1%, 2%, 5%, 16%, 20%, 30%, 40% and 50%) from a light-adapted background (irradiance = 2.64x1014 photons cm-2 s-1; time bin = 0.01s; inter stimulus interval = 5 seconds; Scale bar = 5 Spikes/s). (B) Mean ± S.E.M. change in firing rate over the first 500ms of the light-step plotted as a function of average cone (S-cone opsin and L-cone opsin) contrast. Opn1mwR showed a significantly larger amplitude response at all but the lowest two contrasts (2-Way ANOVA: * p < 0.05; ** p < 0.005; *** p = 0.001). (C) Normalised peak response amplitude of the data in (B) plotted as a function of average cone (S-cone opsin and L-cone opsin) contrast demonstrates that both populations are best fit by a single dose-response curve (F-test = 0.798; R2 = 0.94)

We next asked whether spatial receptive field structure of individual dLGN units was impacted in rd1 Opn1mwR mice. We presented a single vertical bar of polychromatic light spanning 13° of visual angle upon a projected background at 13 different overlapping spatial locations (4.5° separation in

161

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

bar position) on the azimuth in a pseudorandom order. In order to facilitate comparisons between genotypes we initially aimed to present bars visible only to cones as the rd1 mice lacked functional rods (Figure 5.3A). However, in practice we found we were unable to reliably detect responses in the rd1 Opn1mwR mice using spatial stimuli at the maximum cone contrast available for this method.

However, as these mice do not possess functional rod photoreceptors at these ages, we could change the spectral composition of the background irradiance and therefore generate a bar which produced a larger Michaelson contrast for cones (Figure 5.3B). We determined the contrast sensitivity under these new settings similar to that described above for full field stimuli and found we could elicit significantly larger amplitude responses at a maximum average cone contrast of 70% (Figure 5.3C). Consequently, rd1 mice were mapped under these settings. This approach was not appropriate for Opn1mwR mice as the new background allowed for a rod contrast and therefore the receptive field mapping in this genotype was undertaken using the original cone isolating stimuli.

Using these two protocols, we isolated 38 single units from Opn1mwR mice and 48 single units from rd1 Opn1mwR mice that showed a significant increase in firing rate above baseline (>2 standard deviations) to at least one bar position over the average of 90 presentations. The response of two such units is shown in Figure 5.3D. For all single units we applied a best fit Gaussian to the relationship between bar position and response amplitude (R2 > 0.7; mean = 0.87 for rd1 Opn1mwR and Opn1mwR mice). A representative Gaussian fit of the two units from Figure 5.3D is shown in Figure 5.3E. Receptive field size could be estimated by calculating 1 standard deviation from the mean as measured from a best fit Gaussian. This demonstrated the mean receptive field size of the Opn1mwR population to be 12.17° ± 0.5 and for the rd1 Opn1mwR population to be 9.96° ± 0.3 (Figure 5.3F; p = 0.0005; unpaired T-test). We continued by investigating other characteristics of responses to spatial stimuli and found that LGN neurones in the rd1 population showed significantly smaller amplitude (7.02 ± 0.8 Spikes/s; Figure 5.3G) and longer latency (177.9ms ± 5.4; Figure 5.3H) responses then their wildtype controls (amplitude = 10.1 ± 1.2 Spikes/s (p= 0.031); latency = 112.3ms ± 4.46 (p <0.0001, unpaired T-test)), even under these higher cone contrast settings.

162

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

Figure 5.3 Spatial receptive fields in the rd1 dLGN. (A) The effective photon flux of the background and bar stimulus, with calculated Michaelson contrast, for each photopigment expressed in the OPN1mwR retina (L- Cone opsin = 47%; S-Cone opsin = 51%; Rod opsin 34%; & melanopsin = 51%) (B) The effective photon flux of the background and bar stimulus used for the rd1 mice demonstrating the larger Michaelson contrast for cones (S-cone opsin = 73%; L-cone opsin = 67%; & Melanopsin = 91%). As these mice lack functional rod photoreceptors the rod contrast (94%) does not contribute to the light-response (C) Full field 1s contrast steps were presented to generate a contrast sensitivity curve to the same population of rd1 LGN units (grey dotted line) and demonstrated this could elicit a significantly larger amplitude response at the highest cone contrast producible (2 Way Anova, p = 0.04) (D) Representative 2 Dimensional heat map (colour code = response amplitude in delta firing rate) for a single unit in the dLGN of an OPN1mwR (top) and rd1 OPN1mwR (bottom) when presented with a 250ms vertical overlapping bars spanning 13° of visual angle (at 4.5° resolution) in a pseudorandom order over 90 presentations. (E) A best-fit 2-dimensional Gaussian was applied to the relationship between mean ± S.E.M of the response amplitude and bar position of the two single units in (D). (F) Mean ± S.E.M of receptive field size for all light-responsive units were significantly larger in the OPN1mwR mouse (12.17° ± 0.5; n = 38 units; green bar) compared to the rd1 OPN1mwR mouse (9.96° ± 0.3; n = 48; black bar; unpaired T-test: p= 0.0005). (G) Mean ± S.E.M for peak response amplitude of the same population of cells demonstrates OPN1mwR mice have a significantly larger response amplitude (10.1 ± 1.2 Spikes/s) compared to rd1 OPN1mwR (7.02 ± 0.8 Spikes/s; unpaired T-test = 0.03). (H) Mean ± S.E.M of response latency demonstrates a significantly slower time to peak response for the rd1 OPN1mwR mice (177.9ms ± 5.4) compared to OPN1mwR mice (112.3ms ± 4.46; unpaired T-test: p <0.0001)

163

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

We next wished to investigate whether melanopsin could make a significant contribution to the visual information reaching dLGN in these young mice. As the mouse retina expresses four different photopigments with well-defined spectral sensitivities, we were able to design three carefully calibrated metamers that, when you transition between any two of these, allows for selective modulation of different photopigments (Figure 5.4A)

Transition from a background of spectrum A (green trace) to spectrum C (orange trace) we term “all photoreceptor” as it provides a positive contrast for all murine photoreceptors in the Opn1mwR retina (S-Cone: 51%; L-Cone: 47%; Rod: 34% and Melanopsin: 51%). Transition between Spectrum B (pink trace) and Spectrum C (orange trace) we term “mel-less” as they provide equivalent contrast for S-Cones (50%), L-Cone (49%) and Rods (30%); however provides a very low contrast for melanopsin (< 5%). Finally, transition between spectrum A (green trace) and spectrum B (pink trace) is silent for S-Cones (<1%), L-Cones (-3%) and Rods (5%) but provides an equivalent contrast for melanopsin (49%) as in the “all photoreceptor” condition. A full table of the spectral irradiance and the calculated Michaelson contrast for each photopigment during spectral transitions is shown in Figure 5.4B.

To confirm these calibrated stimuli are indeed silent for individual photopigments, we presented 50 repeats of full field 4Hz flashes under “all photoreceptor”, “mel-less” and “mel-isolating” conditions at the beginning of each experiment. An example of a single unit from an Opn1mwR and rd1 Opn1mwR mouse is shown in Figure 5.4C. Our “all photoreceptor” and “mel-less” stimuli drove equivalent responses in the dLGN when presented at this temporal frequency but there was no response to the “mel-isolating” condition, as this is outside the known temporal resolution of melanopsin phototransduction. However, this allowed us to confirm online that these stimuli settings were indeed rod silent. An example PSTH of the rasters in Figure 4.4C is plotted for each stimulus condition (Figure 5.4D).

164

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

Figure 5.4 Design and validation of silent substitution stimuli. (A) The output of the four LEDs (peak emissions = 405nm, 455nm, 525nm, 630nm) and a laser (peak emission = 561nm) were used to produce three spectra (Spectrum A = green trace, Spectrum B = pink trace, Spectrum C = orange trace). Transition between any two of these spectra was calculated to produce a Michaelson contrast for individual photopigments whilst keeping others silent in the Opn1mwR retina. Transition between Spectra A and Spectra C we term “all photoreceptor” as it provides contrast for rod opsin, cone opsin and melanopsin. Transition between Spectra B and Spectra C we term “mel-less” as it provides the same contrast for rod and cone photoreceptors as the “all photoreceptor” condition but has a low melanopsin contrast. Finally, transition between Spectra A and spectra B we term “Mel- isolating” as it drives the same contrast for melanopsin as the “all photoreceptor” condition whilst keeping rod and cone signalling silent. (B; left) The effective photon flux for each photopigment in the Opn1mwR retina (L- Cone opsin, S-Cone opsin, Rod opsin and melanopsin) when presented with Spectrum A, B and C. (B; right) Michaelson contrast calculated for L-Cone opsin, S-Cone opsin, rod opsin and melanopsin for transitions in the all photoreceptor, Mel-less and Mel-isolating condition. (C) Representative raster plot of a single light- responsive LGN unit from an OPN1mwR (left) and an rd1 OPN1mwR (right) mouse to 50 repeats of a 250ms flash under each of the three silent substitution conditions (time bin = 0.01s). (D) Peristimulus time histograms (PSTH) demonstrating the average light-response of the two units from (C) in response to 50 presentations of the “all photoreceptor” (black trace), “mel-less” (red trace) and “Mel-isolating” (blue trace) conditions. Data shown is baseline subtracted (time bin = 0.01s)

165

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

The stimuli presented thus far focus on the fast visual responses provided by cone photoreceptors. We therefore turned our attention to longer duration steps where the activation of melanopsin is known to be more influential. Accordingly, we used our silent substitution stimuli and provided 10s full-field steps from a background under the “all photoreceptor”, “mel-less” and “mel-isolating” conditions. Activation of melanopsin is known to contribute to the sustained phenotype in the dLGN and therefore we could categorise our cells as being Transient or Sustained under the “all photoreceptor” condition. Our transient population therefore provided us with an internal negative control for our silent substitution stimuli in this experimental paradigm. We found 11 units with a sustained OFF phenotype in the Opn1mwR population but none in the rd1 population and therefore could not make a valuable comparison.

We initially looked at the rd1 Opn1mwR population and found, similarly to the dark adapted state, that a large proportion of units (50/64) showed a transient response to the 10s “all photoreceptor” condition with the remaining 14/64 cells showing a sustained phenotype. Comparing the average firing rate over the last five seconds of the light step of the sustained population demonstrated that, although there was a consistent difference in firing rate between the “all photoreceptor” condition (2.13 ± 0.4 Spikes/s) and the “mel-less” condition (1.4 ± 0.3 Spikes/s), this was not significant (Figure 5.5A; p = 0.09.; 2-way ANOVA). In the Opn1mwR population, we found a significant reduction in the proportion of sustained ON units (20/68) in response to the “all photoreceptor” condition compared to the dark condition (71/157). In these 20 sustained units, we did not observe a significant difference between the average firing rate in the “all photoreceptor” condition (1.99 ± 0.5 Spikes/s) compared to the “mel-less” condition (1.62 ± 0.3 Spikes/s; p = 0.68) (Figure 5.5B). In addition, none of these sustained units showed a significant firing rate above baseline in response to the “mel- isolating” condition in either the rd1 Opn1mwR (0.08 ± 0.1 Spikes/s; p = 0.56; Figure 5.5C) or Opn1mwR mice (0.06 ± 0.1 Spikes/s; p = 0.63; Figure 5.5D). These comparisons are summarised in Figures 5.5E & F.

As expected, the transient cell population in both genotypes showed no significant change in their firing rates between the “all photoreceptor” or “mel-less” condition during the last 5s of the light step (Figure 5.5G; p = 0.14, paired T-test; Figure 5.5H, p =0.149), nor a response to the “mel- isolating” condition (Figures 5.5I & J; p = 0.85 and p = 0.47, paired T-test). The transient cell population are summarised in Figure 5.5K & L.

166

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

Figure 5.5 Contribution of melanopsin signalling to the juvenile rd1 dLGN PSTH (mean ± S.E.M) of sustained cells from (A) rd1 Opn1mwR mice (n= 14 single units) and (B) Opn1mwR mice (n= 20 single units) in response to a 10s light step under the all photoreceptor (black traces) and mel-less (red traces) stimulus conditions (light on at time = 0; time bin = 0.25s). (C) Average response amplitude (change in firing rate) during the last 5s of the light-pulse was not significantly different between stimulus conditions (p = 0.09; 2 Way ANOVA) or between genotypes (p = 0.81; 2 Way ANOVA). PSTH (mean ± S.E.M) of the same sustained population of cells in the (D) rd1 Opn1mwR mice and (E) Opn1mwR mice in response to the mel-isolating stimulus (blue trace). (F) Response amplitude over the last 5s of the stimulus was not significantly different to baseline (p = 0.56 and p = 0.63, paired T-test; for rd1 Opn1mwR mice and Opn1mwR mice, respectively. These were not significantly different between genotypes (p = 0.88; unpaired T-test). PSTH (mean ± S.E.M) of the transient cell population in (G) the rd1 Opn1mwR mice (n = 50 single units) and (H) Opn1mwR mice (n = 71 single units) in response to the all photoreceptor (black trace) and mel-less (red trace) stimulus conditions. (I) Average response amplitude during the last 5s of the light-pulse was not significantly different between stimulus conditions (p = 0.15; 2 Way ANOVA) or between genotypes (p = 0.35; 2 Way ANOVA). PSTH (mean ± S.E.M) of the same sustained population of cells in the (J) rd1 Opn1mwR mice and (K) Opn1mwR mice in response to the mel-isolating stimulus (blue trace). (L) Response amplitude over the last 5s of the stimulus was not significantly different to baseline (p = 0.85 and p = 0.47, paired T-test) for rd1 Opn1mwR mice and Opn1mwR mice, respectively, or between genotypes (p = 0.89, unpaired T-test)

167

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

This lack of robust melanopsin activity in the dLGN of these young mice led us to first question whether the melanopsin system had fully developed. We therefore sought to confirm that when at full maturity, ipRGCs which project the dLGN do indeed contribute significant visual information. We therefore recorded from the dLGN of five, four month old Opn1mwR mice. We once more applied our silent substitution stimuli and could again categorise these cells as Transient (174/255) or Sustained-ON (81/255). However, here we identified 28 single units which demonstrated a significant increase in their firing rate (1.19±0.5 Spikes/s) compared to baseline in response to the “mel-isolating” stimulus (Figure 5.6A; p = 0.0004; paired T-test). These responding units all demonstrated a sustained ON phenotype under the “all photoreceptor” and “mel-less” conditions (Figure 5.6B). Accordingly, these cells showed a significant difference in their firing rates over the last five seconds (0.61±0.2 Spikes/s) between “all photoreceptor” condition (3.10±0.5 Spikes/s) and “mel-less” condition (2.49±0.4 Spikes/s; Figure 5.6C).

We next hypothesised that there could be a difference in the maturity of ipRGCs which project to IF and NIF brain regions. To investigate this further, we recorded responses in the Olivary Pretectal Nucleus (OPN; Figure 5.6D) which is strongly innervated by M1 type ipRGCs, from four week old Opn1mwR mice. Here, we recorded 44 light-responsive single units from three Opn1mwR animals. When categorising these cells we found a far higher proportion of sustained cells in the OPN (21/44) compared to the dLGN of equivalently aged mice. The remaining cells (23/44) showed a transient ON-OFF response to a 10s light step under the “all-photoreceptor” condition (Figure 5.6E). The average firing rate over the last 5 seconds of these cells was 4.21 ± 0.6 Spikes/s. Under the “mel- less” condition these same cells demonstrated significantly lower average firing rate (2.62 ± 0.6 Spikes/s; p = 0.02, paired T-test) over the same time period suggesting a measurable melanopsin contribution (Figure 5.6F). We therefore additionally applied our melanopsin isolating stimulus and found a modest but significant increase in firing rate (0.53 ± 0.1 Spikes/s) of these sustained ON categorised cells (Figure 5.6G; p=0.0004; paired T-test). These results are summarised in Figure 5.6H. We found that this phenomenon held true for the sustained population but not the transient population. Accordingly, these cells showed no significant changes during “all photoreceptor” and “Mel-isolating” conditions (Figure 5.6I), nor could we stimulate any “mel-isolating” responses (Figure 5.6J). The results for the transient population are summarised in Figure 5.6K.

168

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

Figure 5.6 Contribution of melanopsin signalling to the adult dLGN and juvenile OPN. (A) PSTH (mean ± S.E.M) of 28/255 single units from the dLGN of five adult Opn1mwR mice (age > 4 months) which demonstrated a significant increase in firing rate in response to the mel-isolating stimulus (1.19±0.5 Spikes/s; blue trace; p = 0.0004, paired T-test). (B) PSTH (mean ± S.E.M) of these 28 single units demonstrated a significant difference in firing rate (ΔFR = 0.6 ± 0.2 Spikes/s; p = 0.014, paired T-test) between the all photoreceptor (black trace) and mel-less (red trace) stimulus conditions. (C) Average response amplitude (change in firing rate) during the last 5s of each stimulus condition is quantified for the sustained population. (D) Representative DiI tracts delineate the final resting position of the 4x8 recording electrode in the Olivary Pretectal Nucleus (dotted line) in a 4- week old Opn1mwR mouse. (E) 21/44 light-responsive units (n = 3 mice) showed a sustained ON response with the remaining units 23/44 demonstrating a transient response to a 10s light-step under the all-photoreceptor condition. (F) The sustained population showed a significant difference in firing rate over the last 5s of the light step between all photoreceptor and mel-less conditions (ΔFR = 1.58 ± 0.6 Spikes/s; p = 0.02, paired T-test). (G) PSTH (mean ± S.E.M) of these same sustained cells in response to the mel-isolating demonstrate a significant increase in firing rate during the light-step compared to baseline (ΔFR = 0.53 ± 0.1 Spikes/s; p=0.0004; paired T- test). These data for the sustained population are summarised in panel (H). The transient population showed no significant difference in firing rate during the (I) all photoreceptor and mel-less conditions, or (J) the mel- isolating condition. These data are summarised in panel (K).

169

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

The reduced fraction of sustained responses identified in the young dLGN of both genotypes compared to the adult, and that a significant melanopsin contribution could be identified in the OPN of young Opn1mwR mice led us to investigate the degree of melanopsin expression in the retina at these equivalent ages. We therefore first labelled a wholemount retina of a four week old rd1 Opn1mwR mouse with the melanopsin antibody UF006, which is known to label both isoforms of the melanopsin protein (Hughes et al., 2012b). Here, we found strong labelling of somas and dendrites of retinal ganglion cells across the entire retina with labelled cells demonstrating the stereotype “photoreceptive net” described for ipRGCs (Figure 5.7A).

We subsequently labelled retinal sections from two four week old Opn1mwR and two four week old rd1 Opn1mwR mice in order to gain more information about the architecture of these cells. Again, we found strong labelling for the melanopsin protein in the soma and dendrites in both genotypes; however the labelled dendrites were restricted to the outer plexiform layer, indicative of M1 type ipRGCs (Figures 5.7B & C). Across these two retinas each from two mice we found no evidence of cells which showed M2 like architecture with dendritic stratification in the inner plexiform layer, nor convincing examples of labelled dendrites in the inner plexiform layer. To confirm that our antibody was indeed capable of labelling both subtypes in the Opn1mwR model, and to support our electrophysiological recordings from the adult dLGN, we next labelled retinal sections from two four month old Opn1mwR mice and found substantially more labelling of both M1 type and M2 type ipRGCs as apparent by the presence of dendrites in both the inner and outer plexiform layers (Figure 5.7D).

170

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

Figure 5.7 Melanopsin expression in the juvenile Opn1mwR and rd1 Opn1mwR retina. (A) Representative wholemount retina from a four-week old rd1 Opn1mwR mouse labelled with the anti-melanopsin antibody UF006 shows strong labelling for the melanopsin protein in retinal ganglion cells that are uniformly distributed across the retina (Scale bar = 500µm). Representative maximum projection confocal images of UF006 labelled retinal sections from (B) a four week old Opn1mwR mouse and (C) an rd1 Opn1mwR mouse. Both genotypes show strong labelling for the melanopsin protein in somas and dendrites however labelled dendrites are restricted to the outer plexiform layer. Note the number of nuclei under DAPI labelling in the outer nuclear layer of rd1 Opn1mwR retina is markedly reduced representing the loss of the rod photoreceptor population. (D) Representative UF006 labelling for the melanopsin protein an adult (4 month old) Opn1mwR mouse demonstrates labelling for the melanopsin protein in both inner and outer plexiform layers representing labelling of both M1 and M2-type ipRGCs. Scale bar for images B, C and D = 50µm.

171

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

5.5 Discussion

During retinal degeneration, the rod and cone photoreceptors undergo rapid cell death leaving intrinsically photosensitive retinal ganglion cells (ipRGCs) as the only light-responsive retinal cells at advanced stages. These cells project to the dLGN (Ecker et al., 2010) but demonstrate poor spatio- temporal signalling characteristics when isolated from the outer retinal photoreceptors (Brown et al., 2010, Procyk et al., 2015). During early stages of degeneration, pockets of cone photoreceptors survive (Jimenez et al., 1996) and can continue signalling light-information in the retina (Stasheff, 2008). Here, using in-vivo electrophysiology and photoreceptor silent substitution, we set out to investigate the extent to which melanopsin and cones each make significant contribution to residual visual responses in the dLGN during early stage retinal degeneration.

In doing so, we find that during early stage retinal degeneration cone-driven responses can be elicited under both light and dark adapted conditions. LGN neurones in rd1 mice exhibit a number of light-response profiles which match those found in age-matched controls, however they demonstrate a significant impairment in both the amplitude and latency of these residual light- responses. We continue to show that rd1 mice retain sensitivity over a moderate contrast range and possess spatial receptive fields during early stage retinal degeneration which is not dissimilar to the age-matched controls. However, we find at ages employed during these experiments, the melanopsin signal is at best modest in the dLGN of both Opn1mwR and rd1 Opn1mwR mice using our silent substitution stimuli. Presenting these same stimuli to the OPN of equivalently aged Opn1mwR mice, and to the dLGN of adult Opn1mwR mice, identifies a robust contribution of melanopsin to a proportion of sustained light-responses. This suggests a significant temporal delay in the development of melanopsin signalling to the dLGN.

Whilst cone responses can be identified by their rapid onset and more transient nature in response to light, identifying the contribution of melanopsin, whose response is apparent over the course of seconds, is significantly more complex. Therefore, an important application of our work was to use the silent substitution approach in order to tease apart the contribution of cone and melanopsin signalling in a fully intact animal during early stage retinal degeneration. Presentation of 10-second light-steps designed to activate both cone opsins and melanopsin identified only a small proportion of LGN cells in the rd1 mouse which demonstrated a sustained-ON phenotype, normally associated with melanopsin signalling (Brown et al., 2010). Accordingly, we did not identify any LGN units that received a significant and robust contribution originating from melanopsin signalling in combination with cone excitation, nor did we identify any units which received a significant melanopsin-driven response in isolation of cone excitation. However this interpretation is made with caution due to the low number of sustained units identified during our electrophysiological recordings making this statistical comparison underpowered. Nonetheless, these findings could not be explained by the

172

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

pathology of rd1 mice as we also failed to detect a significant and robust melanopsin contribution in our electrophysiological recordings from age-matched Opn1mwR controls.

Whilst our silent substitution stimuli could elicit robust responses from cone photoreceptors, it is possible that they were not adequate to strongly excite the melanopsin photopigment which is known to exhibit poor sensitivity (Do et al., 2009) and drive only modest changes in firing rate in the dLGN under light-adapted conditions (Davis et al., 2015). However, the melanopsin contrast presented in our “all photoreceptor” and “mel-isolating” stimuli (50% melanopsin contrast) is above those previously reported to elicit melanopsin responses in the adult dLGN to full field stimuli (>32% melanopsin contrast). We therefore looked to confirm that our stimuli were suitable for evoking melanopsin responses in the dLGN of adult Opn1mwR mice. Here, we identified a proportion of sustained-ON cells (34% of the sustained population) whose response was significantly influenced by melanopsin signalling.

These data suggest that there may be a deficit in melanopsin signalling to the dLGN at the ages employed in these set of experiments. There have been reports of differences in temporal expression of the two known isoforms of mammalian melanopsin (OPN4L and OPN4S) during visual system development, however these both reach adult concentrations by post-natal day 14 (P14) (Hughes et al., 2012b). Furthermore, anatomical studies have also confirmed the expression of melanopsin in rd1 mice at these young ages (Ruggiero et al., 2009). On this basis, our electrophysiological recordings in the dLGN should be outside of this period of ipRGC maturation.

However, different concentrations and isoforms of the melanopsin protein are expressed in different ipRGC subtypes. These subtypes project to different brain nuclei involved in a number of different IF and NIF functions (Baver et al., 2008, Hattar et al., 2006, Ecker et al., 2010). We therefore turned our attention to the retina to determine the degree of melanopsin expression in different ipRGC subtypes at these ages. Immuno-histochemical labelling of retinas of from 4 week old rd1 Opn1mwR and Opn1mwR revealed widespread melanopsin expression in both genotypes but that this was restricted to M1-type ipRGCs, based on their exclusive dendritic stratification in the outer plexiform layer. One of the major projection targets of M1 ipRGCs is the Olivary Pretectal Nucleus (OPN) (Hattar et al., 2002, Hattar et al., 2006) which is responsible for control of the pupillary light reflex. We therefore once more turned to our silent substitution stimuli whilst recording from this nucleus in 4 week old Opn1mwR mice. Similar to the adult dLGN, we found a large proportion of sustained responses (47%), a subset of which whose response was significantly influenced by melanopsin signalling.

Put together, our data indicate that at the time which we recorded visual responses in the dLGN, the contribution of M1 ipRGCs was both measurable and significant, whilst the anatomical and functional identification of other subtypes (M2-5) were significantly harder to detect. It is specifically the M4-subtype of ipRGC which has been more recently shown to directly project to the dLGN

173

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

(Estevez et al., 2012) and consequently much less is known about this subtype compared to M1 and M2 cells (Schmidt et al., 2011a). M4 ipRGCs certainly express melanopsin at extremely low concentrations and is detectable after Tyramide amplified immunofluorescence (Estevez et al., 2012). This subtype would therefore perhaps have been less readily identified using the UF006 antibody employed in this study, but which has previously been shown to co-label with cells that express both OPN4L and OPN4S isoforms of melanopsin (Hughes et al., 2012b). This raises the question of what isoform of melanopsin is expressed in M4-type ipRGCs, and therefore whether it also exhibits a different temporal pattern of expression during visual system development. Nonetheless, the differences in ipRGC subtype expression of the melanopsin protein and electrophysiological recordings in both young and old Opn1mwR mice supports a hypothesis that maturation of different ipRGC subtypes may well occur at different points during visual system development.

It is also important to note, that at these ages from which we recorded these data, the visual system is still undergoing a critical period of development. At P21, the ON-OFF pathways in the retina are still not fully segregated (Tian and Copenhagen, 2003) and the LGN is still undergoing pruning and refinement at the retino-geniculate synapse (Chen and Regehr, 2000, Hooks and Chen, 2008). It is therefore also reasonable to suggest that recordings at these developmental stages may preclude our ability to either elicit or detect any low amplitude melanopsin signalling to the visual thalamus.

Whilst we were able to detect at best a modest contribution of melanopsin signalling to light- responses in the young dLGN of rd1 mice, we were more readily able to identify cone-driven light- responses. We therefore first sought to characterise these light-responses under dark-adapted conditions and found that we could readily identify light-responses throughout dorsal LGN in the rd1 mouse. The number of individual light-responsive units per electrode placement was consistent between genotypes indicating that a significant proportion of LGN units still retain cone-inputs from the retina, even after loss of the rod photoreceptor population. In agreement with previous retinal recordings, we found these cone driven light-responses could be readily elicited up to approximately four weeks of age (Stasheff, 2008), however they showed a rapid decline between P18 and P33.

We could categorise these cone responses based on their response to a 10-second light-step from darkness, and found a proportional shift in the light-response profile of single units in the dLGN between the two genotypes. Most striking was the transient nature of responses in the rd1 mouse compared to controls, most likely reflecting the loss of the rod photoreceptors, which make up 97% of the photoreceptors in the mouse retina (Jeon et al., 1998). Furthermore, we failed to identify a single sustained-OFF response in our dLGN recordings. Of course, one must bear in mind that we presented steps from darkness and therefore a sustained-OFF response will only be identified if a cell has a baseline firing rate at which a clear reduction in firing can be observed. Nonetheless, these sustained-OFF responses were also absent during our silent substitution stimuli in light-adapted

174

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

conditions in the rd1 dLGN, but were present in both dark and light-adapted conditions in the Opn1mwR population. This could indicate specific deficits in the cone OFF-pathway which may arise earlier than the ON-pathway during C57 rd1 degeneration.

The physiology of these residual light-responses is also in agreement with previous recordings from the young rd1 retina (Stasheff, 2008, Strettoi et al., 2002). We find both a significant reduction in the amplitude of the light-response and an increase in the time to peak response. This likely reflects the reduced population of photoreceptors influencing an individual RGC’s activity resulting in fewer evoked spikes for the same given irradiance. Consequently, it takes longer for an RGC to reach threshold for action potential generation following light-onset.

We next wanted to determine the quality of light-information reaching the dLGN under more naturalistic light-adapted conditions. Indeed a large fraction of single units in the rd1 dLGN continued responding under light-adaptation; however the reduction in peak response amplitude was exacerbated under these conditions and is reflected in the contrast sensitivity curve generated for these mice. The amplitude of these residual light-adapted responses was on average 0.41 ± 0.04 of the wildtype light-response for each effective cone contrast investigated. This data is supported by photopic ERG recordings that have demonstrated that both α-waves and b-waves of rd1 mice are significantly reduced and delayed as early as at P14, indicating specific dysfunction to the cone signalling pathways in the retina at this age (Strettoi et al., 2002). Nonetheless, these LGN units still demonstrate contrast sensitivity which is not dissimilar to that of their wildtype counterparts; implying cone-opsin signalling remains functional in at least some photoreceptor outer-segments at this age.

The reduced amplitude of light-adapted responses was further highlighted by our inability to record responses to spatially structured stimuli at the highest cone contrast which had elicited robust responses to uniform stimuli. However, under higher cone contrast settings we were able elicit significant responses to spatial stimuli which demonstrated that individual LGN neurones in the rd1 mouse possess receptive fields with a mean diameter of 9.96° ± 0.3 ° and is within the range of those previously reported in the dLGN of adult wildtype mouse (2-10°) (Grubb and Thompson, 2003). This indicates that the outer retinal circuitry linking remaining cones, horizontal cells and bipolar cells to be at least superficially intact for those LGN neurones for which we could record spatial receptive fields. However, the receptive fields reported here for rd1 mouse were approximately 3° smaller than in the Opn1mwR mice. These smaller receptive fields were not an artefact of analysis due to the reduced light-response amplitude as we found no correlation between response-amplitude and variance across the two populations (data not shown). This indicates that this difference may be biological in origin. One possibility is that at the ages employed here, the complete death of the rod photoreceptor population results in the loss of a large proportion of photoreceptor input, thus

175

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

reducing the spatial extent of photon capture by photoreceptors which contribute to the activity of an individual retinal ganglion cells activity.

Our recordings from the dLGN of both young Opn1mwR and rd1 Opn1mwR mice suggest there are subtype specific differences in the development of ipRGC subtypes which project to the dLGN. This further highlights an already outstanding question from the literature, which is whether or not ipRGCs should be considered a heterologous population of Retinal Ganglion cells with subtype specific characteristics and roles in both IF and NIF vision. A clear limitation of our investigation into the contribution of melanopsin signalling to the dLGN during retinal degeneration was the use of the C57 rd1 mouse model. One method to study melanopsin during this early stage degeneration would be make use of other more slowly degenerating models, such as the C57 rd10 mouse (Chang et al., 2002) where rod cell death is not complete until around P30 (Chang et al., 2000) and cone bipolar cells show normal morphology and architecture up to approximately P45 (Gargini et al., 2007). This would more likely ensure full maturation of the murine visual system (Hooks and Chen, 2007) and provide a much wider epoch from which to record residual cone responses in conjunction with a fully mature melanopsin system. Additionally, this would confirm that the lack of robust melanopsin activity seen in the juvenile dLGN compared to the adult is not a result of the difficulty in studying visual responses in young mice.

Our electrophysiological recordings of cone-driven responses in the dLGN of young rd1 mice also has great impact for gene therapy strategies attempting to restore vision to those afflicted with retinal degeneration. At these early stages of retinal degeneration, the presence of robust light-responses under both dark and light-adapted conditions indicate that not only is the remaining retinal circuitry at least superficially intact in the retina, but that RGCs can still convey this residual visual information to the dLGN, long after the rod population has undergone cell death. This would imply that early intervention during retinal degeneration, prior to significant retinal remodelling, would provide the best chance of success not only for basic light perception, but also for more complex vision in detecting spatial structure under naturalistic light-adapted states.

176

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

5.6 References

Allen, A. E., and R. J. Lucas, 2016, Using Silent Substitution to Track the Mesopic Transition From Rod- to Cone-Based Vision in Mice: Investigative Ophthalmology & Visual Science, v. 57, p. 276-287. Baver, S. B., G. E. Pickard, P. J. Sollars, and G. E. Pickard, 2008, Two types of melanopsin retinal ganglion cell differentially innervate the hypothalamic suprachiasmatic nucleus and the olivary pretectal nucleus: European Journal of Neuroscience, v. 27. Berson, D. M., F. A. Dunn, and M. Takao, 2002, Phototransduction by retinal ganglion cells that set the circadian clock: Science, v. 295. Brown, T. M., C. Gias, M. Hatori, S. R. Keding, M. a. Semo, P. J. Coffey, J. Gigg, H. D. Piggins, S. Panda, and R. J. Lucas, 2010, Melanopsin Contributions to Irradiance Coding in the Thalamo-Cortical Visual System: Plos Biology, v. 8. Chang, B., N. L. Hawes, R. E. Hurd, M. T. Davisson, S. Nusinowitz, and J. R. Heckenlively, 2000, A new mouse retinal degeneration (rd10) caused by a missense mutation in exon 13 of the beta-subunit of rod phosphodiesterase gene: Investigative Ophthalmology & Visual Science, v. 41, p. S533- S533. Chang, B., N. L. Hawes, R. E. Hurd, M. T. Davisson, S. Nusinowitz, and J. R. Heckenlively, 2002, Retinal degeneration mutants in the mouse: Vision Research, v. 42. Chen, C. F., and W. G. Regehr, 2000, Developmental remodeling of the retinogeniculate synapse: Neuron, v. 28, p. 955-966. Davis, K. E., C. G. Eleftheriou, A. E. Allen, C. A. Procyk, and R. J. Lucas, 2015, Melanopsin-Derived Visual Responses under Light Adapted Conditions in the Mouse dLGN: PloS one, v. 10, p. e0123424-e0123424. Do, M. T. H., S. H. Kang, T. Xue, H. Zhong, H.-W. Liao, D. E. Bergles, and K.-W. Yau, 2009, Photon capture and signalling by melanopsin retinal ganglion cells: Nature, v. 457. Ecker, J. L., O. N. Dumitrescu, K. Y. Wong, N. M. Alam, S.-K. Chen, T. LeGates, J. M. Renna, G. T. Prusky, D. M. Berson, and S. Hattar, 2010, Melanopsin-Expressing Retinal Ganglion-Cell Photoreceptors: Cellular Diversity and Role in Pattern Vision: Neuron, v. 67. Estevez, M. E., P. M. Fogerson, M. C. Ilardi, B. G. Borghuis, E. Chan, S. Weng, O. N. Auferkorte, J. B. Demb, and D. M. Berson, 2012, Form and function of the m4 cell, an intrinsically photosensitive retinal ganglion cell type contributing to geniculocortical vision: The Journal of neuroscience : the official journal of the Society for Neuroscience, v. 32. Freedman, M. S., R. J. Lucas, B. Soni, M. von Schantz, M. Munoz, Z. David-Gray, and R. Foster, 1999, Regulation of mammalian circadian behavior by non-rod, non-cone, ocular photoreceptors: Science, v. 284. Gargini, C., E. Terzibasi, F. Mazzoni, and E. Strettoi, 2007, Retinal organization in the retinal degeneration 10 (rd10) mutant mouse: A morphological and ERG study: Journal of Comparative Neurology, v. 500, p. 222-238. Govardovskii, V. I., N. Fyhrquist, T. Reuter, D. G. Kuzmin, and K. Donner, 2000, In search of the visual pigment template: Visual Neuroscience, v. 17, p. 509-528. Grubb, M. S., and I. D. Thompson, 2003, Quantitative characterization of visual response properties in the mouse dorsal lateral geniculate nucleus: Journal of Neurophysiology, v. 90. Hattar, S., M. Kumar, A. Park, P. Tong, J. Tung, K. W. Yau, and D. M. Berson, 2006, Central projections of melanopsin-expressing retinal ganglion cells in the mouse: Journal of Comparative Neurology, v. 497. Hattar, S., H. W. Liao, M. Takao, D. M. Berson, and K. W. Yau, 2002, Melanopsin-containing retinal. ganglion cells: Architecture, projections, and intrinsic photosensitivity: Science, v. 295.

177

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

Hooks, B. M., and C. Chen, 2008, Vision triggers an experience-dependent sensitive period at the retinogeniculate synapse: Journal of Neuroscience, v. 28, p. 4807-4817. Hooks, B. M., and C. F. Chen, 2007, Critical periods in the visual system: Changing views for a model of experience-dependent plasticity: Neuron, v. 56, p. 312-326. Hughes, S., L. Welsh, C. Katti, I. Gonzalez-Menendez, M. Turton, S. Halford, S. Sekaran, S. N. Peirson, M. W. Hankins, and R. G. Foster, 2012, Differential Expression of Melanopsin Isoforms Opn4L and Opn4S during Postnatal Development of the Mouse Retina: Plos One, v. 7. Jeon, C. J., E. Strettoi, and R. H. Masland, 1998, The major cell populations of the mouse retina: Journal of Neuroscience, v. 18. Jimenez, A. J., J. M. GarciaFernandez, B. Gonzalez, and R. G. Foster, 1996, The spatio-temporal pattern of photoreceptor degeneration in the aged rd/rd mouse retina: Cell and Tissue Research, v. 284, p. 193-202. Jones, B. W., and R. E. Marc, 2005, Retinal remodeling during retinal degeneration: Experimental Eye Research, v. 81, p. 123-137. Lucas, R. J., S. Hattar, M. Takao, D. M. Berson, R. G. Foster, and K. W. Yau, 2003, Diminished pupillary light reflex at high irradiances in melanopsin-knockout mice: Science, v. 299. Ogilvie, J. M., T. Tenkova, J. M. Lett, J. Speck, M. Landgraf, and M. S. Silverman, 1997, Age-related distribution of cones and ON-bipolar cells in the rd mouse retina: Current Eye Research, v. 16, p. 244-251. Peirce, J. W., 2008, Generating Stimuli for Neuroscience Using PsychoPy: Frontiers in neuroinformatics, v. 2, p. 10-10. Perez-Leon, J. A., E. J. Warren, C. N. Allen, D. W. Robinson, and R. L. Brown, 2006, Synaptic inputs to retinal ganglion cells that set the circadian clock: European Journal of Neuroscience, v. 24. Procyk, C. A., C. G. Eleftheriou, R. Storchi, A. E. Allen, N. Milosavljevic, T. M. Brown, and R. J. Lucas, 2015, Spatial receptive fields in the retina and dorsal lateral geniculate nucleus of mice lacking rods and cones: Journal of Neurophysiology, v. 114, p. 1321-1330. Ruggiero, L., C. N. Allen, R. L. Brown, and D. W. Robinson, 2009, The development of melanopsin- containing retinal ganglion cells in mice with early retinal degeneration: European Journal of Neuroscience, v. 29, p. 359-367. Schindelin, J., I. Arganda-Carreras, E. Frise, V. Kaynig, M. Longair, T. Pietzsch, S. Preibisch, C. Rueden, S. Saalfeld, B. Schmid, J. Y. Tinevez, D. J. White, V. Hartenstein, K. Eliceiri, P. Tomancak, and A. Cardona, 2012, Fiji: an open-source platform for biological-image analysis: Nature Methods, v. 9, p. 676-682. Schmidt, T. M., S.-K. Chen, and S. Hattar, 2011, Intrinsically photosensitive retinal ganglion cells: many subtypes, diverse functions: Trends in Neurosciences, v. 34. Smallwood, P. M., B. P. Olveczky, G. L. Williams, G. H. Jacobs, B. E. Reese, M. Meister, and J. Nathans, 2003, Genetically engineered mice with an additional class of cone photoreceptors: Implications for the evolution of color vision: Proceedings of the National Academy of Sciences of the United States of America, v. 100, p. 11706-11711. Stasheff, S. F., 2008, Emergence of sustained spontaneous hyperactivity and temporary preservation of OFF responses in ganglion cells of the retinal degeneration (rd1) mouse: Journal of Neurophysiology, v. 99, p. 1408-1421. Strettoi, E., V. Porciatti, B. Falsini, V. Pignatelli, and C. Rossi, 2002, Morphological and functional abnormalities in the inner retina of the rd/rd mouse: Journal of Neuroscience, v. 22, p. 5492-5504. Tian, N., and D. R. Copenhagen, 2003, Visual stimulation is required for refinement of ON and OFF pathways in postnatal retina: Neuron, v. 39, p. 85-96. Vugler, A. A., M. Semo, A. Joseph, and G. Jeffery, 2008, Survival and remodeling of melanopsin cells during retinal dystrophy: Visual Neuroscience, v. 25.

178

Chapter 5 Delayed maturation of ipRGCs projecting to the dLGN

Wong, K. Y., F. A. Dunn, D. M. Graham, and D. M. Berson, 2007, Synaptic influences on rat ganglion- cell photoreceptors: Journal of Physiology-London, v. 582, p. 279-296.

179

Chapter 6 General Discussion

Chapter 6: General Discussion

180

Chapter 6 General Discussion

The aim of this thesis was to investigate the quality of visual information that melanopsin signalling can contribute to the dLGN during retinal degeneration and to determine what precludes its ability to support behaviourally relevant vision. Throughout this work I use a number of anatomical and electrophysiological techniques to address these outstanding questions from the literature and, in so doing, reveal a number of original findings which have significant implications in both the field of melanopsin signalling and retinal degeneration.

In Chapter 3 I demonstrate for the first time that all known subtypes of ipRGC survive in advanced stages of retinal degeneration and that their dendritic architecture is similar to that of their wildtype counterparts, indicating they are resilient to the effects of retinal remodelling following loss of the outer retinal photoreceptors. As a consequence of this, I introduce the design and successful application of a pioneering image analysis methodology based on principal component analysis of 3D multi-colour images following Brainbow labelling. This approach has the potential to transform the analytical study of multi-colour images in bio-imaging by allowing, for the first time, the extraction of quantitative data with single cell resolution from these labelling methods.

In Chapter 4, I demonstrate that ipRGCs in the retina and their projection targets in the dLGN possess discrete spatial receptive fields even in advanced stages of degeneration and these are defined by the dendritic field of individual ipRGCs in the retina. I confirm that ipRGC projections are retinotopically organised in the dLGN of rd/rd cl mice using both anatomical and electrophysiological techniques. I show that the temporal resolution and reliability of the deafferented melanopsin response poses the most significant limitation that preclude melanopsin signalling being employed for behaviourally relevant spatial vision.

Finally in Chapter 5, I continued investigating whether melanopsin signalling could contribute more significant information to image-forming vision at progressive stages of retinal degeneration where some residual cone function remains, and is more representative of human conditions of inherited retinal degenerations. Using photoreceptor silent substitution, in combination with in-vivo electrophysiology, I characterised a number of residual cone driven light-responses in the dLGN of these animals under both light and dark adapted conditions including spatial receptive fields and contrast sensitivity. The fact that I failed to record significant melanopsin responses in either the degenerate or wildtype animals, but could in other ipRGC-recipient nuclei and in the adult dLGN, strongly indicates a specific temporal delay in the maturation of melanopsin signalling in the M4 projection during development of the visual system. This suggests that better understanding of the heterogeneity of the individual ipRGC subtypes is required to provide greater insight into their varied and diverse roles in mammalian physiology.

A discussion follows below which describes the implications of these findings in the context of published literature, a critical evaluation of the experimental strategies used in these investigations and possible avenues of future research which stem from the work presented in this thesis.

181

Chapter 6 General Discussion

6.1 Implications Melanopsin signalling to the dLGN in retinal degeneration The initial discovery of ipRGCs suggested they were exclusively involved in non-image forming homeostatic responses to light, and as such, a wealth of data exists concerning its role in aspects of physiology such as circadian photo-entrainment (Lucas et al., 2001b, Berson et al., 2002, Panda et al., 2002, Ruby et al., 2002) and the pupillary light reflex (Lucas et al., 2003b, Lall et al., 2010, Lucas et al., 2001a). However, more recently, a further three subtypes of ipRGC have been identified, of which at least one projects directly to the dLGN (Estevez et al., 2012, Ecker et al., 2010, Schmidt and Kofuji, 2011, Brown et al., 2010). Accordingly, a number of publications have recently demonstrated the various contributions that melanopsin signalling makes to visual perception in visually intact animals (Brown et al., 2012, Schmidt et al., 2014, Storchi et al., 2015, Allen et al., 2014).

Retinitis pigmentosa and associated inherited retinal degenerations result in the irreversible death of rod and cone photoreceptors and are the single largest cause of visual blindness in humans (Hartong et al., 2006). In response to the loss of the outer retinal photoreceptors the surviving retinal cell types undergo significant remodelling (Jones and Marc, 2005, Strettoi et al., 2002). Current approaches to restore vision to these patients involve gene therapy strategies which seek to re-introduce light-sensitive proteins to the remaining retinal cells (Cehajic-Kapetanovic et al., 2015, Gaub et al., 2015). However, at least some of the ipRGC population survives retinal degeneration (Vugler et al., 2008, Lin and Peng, 2013) and continue signalling light information to the brain (Brown et al., 2010, Brown et al., 2012) suggesting they could support some form of visual perception. However, to date, there has been no direct investigations into whether melanopsin signalling could support such complex visual behaviour in degeneration.

This thesis categorically shows that the M4 and M5 ipRGC subtype, which are known to directly project to the dLGN, survive retinal degeneration with normal retinal anatomy and therefore provides a viable route through which an endogenous light-sensitive protein could still provide visual information to the brain. To date, the survival of this subtype has only been suggested by electrophysiological recordings from the dLGN (Brown et al., 2010) and by way of behavioural experiments in rd/rd cl mice (Brown et al., 2010). These data indicated that the melanopsin signal could encode global information about environmental irradiance and could therefore support some form of basic brightness perception in animals with advanced retinal degeneration. However, if the intrinsic melanopsin light response were to support vision in retinal degeneration, it would rely on both the survival of an anatomical and functional projection from the retina to the dLGN, and the capacity to encode and relay more complex features of the visual scene that would allow mammals to both perceive and interact with their environment.

In the healthy visual system, high acuity vision is achieved by the spatial extent of photon capture by rod and cone photoreceptors in the outer retina and their complex wiring with horizontal cells and

182

Chapter 6 General Discussion

bipolar cells before being relayed to an individual retinal ganglion cell (Van Wyk et al., 2009, Baylor et al., 1971). M4 cells have been shown to possess large spatial receptive fields in the intact retina which exhibit an antagonistic centre-surround organisation (Estevez et al., 2012). However, the functioning of this inner retinal circuitry becomes impaired in retinal degeneration (Marc et al., 2007, Strettoi and Pignatelli, 2000). Therefore, if ipRGCs could be employed to support spatial navigation, an important question is whether they possess spatial receptive fields under such conditions. Following the loss of rod and cone photoreceptors, the spatial receptive field would be defined by the spatial extent of photon capture by individual ipRGC. As the melanopsin photopigment is expressed throughout the plasma membrane in its soma and dendrites (Hattar et al., 2002) this would be defined by its dendritic field size. Thus, in chapter 3, our identification that the M4 and M5 subtypes in particular not only survives retinal degeneration, but also demonstrate broadly normal retinal architecture, provides the first evidence that these subtypes could support such behaviour in degeneration.

In Chapter 4, we directly investigate this using in-vitro electrophysiology and demonstrate for the first time that ipRGCs in the retina of degenerate mice possess spatial receptive fields whose size is in agreement with the dendritic field sizes we report in Chapter 3. However, for this information to be employed by the animal, it would require both an anatomically and functionally intact projection from retina to dLGN. Thus, our in-vivo electrophysiological recordings from the dLGN of rd/rd cl mice demonstrate that this connection is not only anatomically intact, since ipRGCs innervate the dLGN in a retinotopic manner, but that this spatial information is conveyed to the dLGN even in advanced stages of degeneration. We do however find that spatial receptive fields in the dLGN to be on average twice as big as those recorded in the retina which suggests the visual acuity of melanopsin- based vision would be poor. Furthermore, although we did not directly set out to investigate this, we did not find any ipRGCs in the retina, or neurones in the dLGN, which exhibited an antagonistic centre-surround organisation as has been previously reported in the wildtype retina (Estevez et al., 2012). This is perhaps unsurprising as the surround organisation is established upstream of ipRGCs in the circuitry linking photoreceptors with horizontal cells and this becomes compromised following outer photoreceptor loss (Strettoi et al., 2003, Jones et al., 2003). Nonetheless, this would further limit the spatial acuity of melanopsin-based spatial vision.

In some respects, the data presented in this thesis is supported by behavioural evidence from mice that only functionally express melanopsin in the intact retina (GNAT1-/- CGNA3-/-) - something previously suggesting that these animals could discriminate patterns in a behavioural water maze task (Ecker et al., 2010). However, our data indicate that melanopsin-driven receptive fields in the dLGN are much larger (30-70°) than the visual acuity demonstrated by these melanopsin only mice (0.16 cycles/°; approximately 12° of visual angle). This comparison must however be made with caution. Firstly, this could reflect differences between the two mice models used in these studies. In our study, we use the rd/rd cl mouse whose retinas undergo significant remodelling following death

183

Chapter 6 General Discussion

of rod and cone photoreceptors (Jones and Marc, 2005, Marc et al., 2007, Strettoi, 2015, Carterdawson et al., 1978). GNAT1-/- CGNA3-/- mice are rod and cone functionless, but their photoreceptors do not degenerate and retinal cells show normal anatomy and synaptic wiring (Hattar et al., 2003, Calvert et al., 2000, Biel et al., 1999). Secondly, it has more recently been shown that these mice may also not be relying solely on melanopsin signalling but also residual rod function (Allen et al., 2010, Hughes et al., 2016). This could readily explain the higher visual acuity of responses in those mice which are more closely in agreement with the size of spatial receptive fields originating from rod and cone photoreceptors in the mouse dLGN (Grubb and Thompson, 2003).

Most significantly in this work we highlight the poor temporal resolution of the deafferented melanopsin response in the degenerate retina. In rd/rd cl mice, we find that the intrinsic melanopsin response to spatially structured stimuli shows a long onset latency taking tens of seconds to reach peak firing, even in response to stimuli of infinite contrast (Chapter 4; Figure 4.4). This is significantly worse than can be explained by the single photon response observed in the wildtype retina (Do et al., 2009) and is likely a result of the altered synaptic input following remodelling in advanced stages of degeneration (Jones and Marc, 2005, Marc et al., 2003, Strettoi et al., 2002). As such, ipRGCs must rely on their poor photon capture in order to respond to light, which results in a long integration time and an increased delay before a change in firing rate is observed. However, not only are the temporal kinetics of ipRGC signalling poor in these animals, but they are also unreliable. A long interval is required between repetitive stimulations in order to maintain light responsiveness - something which has been previously observed in the rd/rd cl retina (Davis et al., 2015) and dLGN (Brown et al., 2010). As humans are diurnal mammals, being most active in the day, they would need to rely on any melanopsin-based vision at a range of temporal frequencies and background light levels. Accordingly, we investigated how ipRGCs would respond when challenged to such dynamic changes in irradiance. This not only confirmed the poor temporal resolution of the intrinsic melanopsin-light response in degeneration, but their poor sensitivity under naturalistic lighting conditions (Chapter 4; Figure 4.7). Thus, our identification of spatial receptive fields in the retina and dLGN of animals with advanced retinal degeneration would suggest that they could employ melanopsin signalling for some form of crude spatial navigation, however the temporal resolution of the deafferented melanopsin response would be incompatible with the movement and speed at which mammals explore their environment.

Whilst in this thesis we primarily focus on the spatio-temporal aspects of the intrinsic melanopsin response, it is important to consider what the quality of other aspects of melanopsin-based vision could be. As a single photopigment, melanopsin could not support any form of colour discrimination and therefore any vision would be monochromatic, with sensitivity to short wavelength light. Whilst ipRGCs have been show to exhibit motion selectivity, this is heavily reliant on extrinsic signals originating from rod and cone photoreceptors, and there has been no evidence that the intrinsic melanopsin response can support such complex feature extraction by itself (Zhao et al., 2014b).

184

Chapter 6 General Discussion

Furthermore, ipRGCs are not yet known to have any role in direction selectivity (Zhao et al., 2014b) which would also severely limit the behavioural relevance of this melanopsin-based vision in degeneration. Considering these factors, the data presented in Chapters 3 and 4 may explain why in some human cases of retinal degenerations patients report crude visual awareness, but cannot employ this for complex visual perception (Zaidi et al., 2007).

The data from rd/rd cl mice collected in Chapter 4 reflects advanced stages of retinal degeneration where melanopsin is the only surviving photopigment. However, in human cases of inherited retinal degenerations symptoms of vision loss do not begin to appear until early adolescence, and there is a progressive loss of visual function with age (Tsujikawa et al., 2008, Delyfer et al., 2004, Hartong et al., 2006). Under these conditions, image-forming vision could rely on both residual cone signalling and melanopsin. Thus, our electrophysiological recordings from Chapter 5 sought to replicate this progressive degenerative state in order to investigate whether melanopsin could make a more significant contribution to image-forming vision during this period, and whether the presence of residual cone activity could at least partially rescue the temporal deficits of the deafferented melanopsin response during advanced stages of the disease.

In the process, we identified a number of cone-driven light responses under dark and light-adapted conditions which support previous recordings of residual cone activity from the rd1 retina (Stasheff, 2008, Strettoi et al., 2003). These residual cone-driven responses provide spatial information, contrast sensitivity and faster temporal signalling, which although compromised in the temporal domain could support visual perception. However, our investigations into whether the additional activation of melanopsin could improve the quality of this information were inconclusive. We identify that the intrinsic melanopsin light response to the dLGN is undetectable at ages when there is still some residual cone activity in the rd1 retina. This was not a specific effect to the degenerate animal since we also failed to identify this response in visually intact age-matched controls using both anatomical and electrophysiological techniques. We did however readily identify melanopsin- driven responses in different subtype specific nuclei in young visually intact mice, and in the adult dLGN. This would indicate there is a specific temporal delay in the signalling and maturation of the M4 projection, similar to what has been reported previously between the M1 and M2 subtypes (Hughes et al., 2012b). The purpose of this temporal difference is currently unknown - however it highlights the requirement to understand both the anatomy and physiology of individual ipRGC subtypes in far greater detail, rather than as a heterogeneous population of cells.

Remnant neural networks in degeneration Throughout our anatomical and electrophysiological investigations in the retina and dLGN of the rd1 mouse, we document and characterise network properties which provide invaluable knowledge for those investigating strategies to cure blindness. Current gene therapy strategies look to re-introduce

185

Chapter 6 General Discussion

light-sensitive proteins into remaining retinal cells in order to restore some visual function (Cehajic- Kapetanovic et al., 2015, Gaub et al., 2015, Mace et al., 2015). However, in order for these strategies to work, they must rely upon the functional integrity of cell types in the retina to which the light- sensitive proteins will be delivered, and for the correct functional connections to be made in downstream visual processing structures. To date, many of these approaches seek to investigate the efficacy of these strategies in the retina (Lagali et al., 2008, Thyagarajan et al., 2010, Barrett et al., 2015) with only more recent publications recording activity in downstream visual processing structures in the brain using in-vivo electrophysiology (Cehajic-Kapetanovic et al., 2015), optical imaging (van Wyk et al., 2015) and recording visually evoked potentials from V1 (Gaub et al., 2015). Thus, although we highlight the temporal limitations of melanopsin signalling, we do prove that a functional projection from retina to dLGN is anatomically and physiologically intact. Furthermore, it innervates this nucleus in a retinotopic manner which is not dissimilar to that of visually intact animals (Piscopo et al., 2013). This indicates that projections to the dLGN appear resilient to the effects of retinal degeneration, and that restoration of light responses using these gene therapy strategies can reach higher visual processing structures unimpaired and have the potential to support some form of spatial vision. Our cholera toxin tracer study in Chapter 4 expands this finding to a heterogeneous population of retinal ganglion cells which indicates that this spatial organisation is superficially retained throughout the visual projection.

Melanopsin signalling in healthy vision A unique aspect of the data presented in this thesis is that recordings from the degenerate mouse also provides insights into melanopsin signalling in isolation of rod and cone photoreceptors. This is technically challenging to achieve in visually intact animals due to the spectral sensitivity of the photopigments expressed in the mouse retina (Fu and Yau, 2007). Whilst the silent substitution approach provides one method to address this (Allen and Lucas, 2016, Brown et al., 2012), our identification of the retinotopic organisation of ipRGC projections to the dLGN indicates a functional significance to this retino-geniculate wiring which is unlikely to be unique to the degenerative state. The presence of melanopsin-driven spatial receptive fields in both the retina and dLGN of degenerate animals suggests that spatially defined melanopsin signals may be employed by the healthy visual system to provide a coarse map of brightness that complements the high spatio- temporal visual information provided by rod and cone photoreceptors. In support of this hypothesis, unpublished data from the Lucas laboratory has demonstrated melanopsin-driven spatial receptive fields in the dLGN of Opn1mw1 mice, using the melanopsin isolating stimuli described in Chapter 5 (Dr Annette Allen; personal communication). However, direct quantitative comparisons of the spatio-temporal characteristics of these spatial receptive fields are challenging as one cannot account for any differences which may occur due to the pathophysiological state of the retina and dLGN in rd1 mice compared to visually intact mice.

186

Chapter 6 General Discussion

Multi-colour image analysis The structure-function relationship of neurones underpins their classification and provides significant insights into how the nervous system integrates information from sensory systems to coordinates precise and appropriate physiological and behavioural outputs. Accordingly, a fundamental aim of neuroscience is to be able to analyse and characterise a large population of neurones whilst maintaining single cell resolution. There are few technologies which have been developed in recent years which have attempted to even begin answering this question. Scanning Electron Microscopy (SEM) uses a beam of electrons to image the 3D structure of biological tissue specimens stained with heavy metals. Whilst this allows for high resolution reconstruction of small cellular regions in 3D, the spatial extent of this imaging is limited and therefore precludes the ability to recreate an entire cell and its projections throughout the brain (Briggman and Denk, 2006). A second approach is single cell labelling. This is not only technically challenging to achieve in-vivo, but is often limited to only a few cells due to the limited number of fluorescent markers available (Fernandez et al., 2016, Wu et al., 2000). This method therefore cannot provide information about a population of cells in an individual animal. The third and most promising approach is to uniquely label individual neurones from a population using multi-colour labelling techniques (Livet et al., 2007). The most documented of these techniques is Brainbow (Cai et al., 2013) which we utilise in Chapter 3, and is only one of a growing assortment of multi-colour labelling methods including spaghetti monster (Viswanathan et al., 2015), MAGIC (Loulier et al., 2014) and CLoNe (Garcia- Moreno et al., 2014). However, these multi-colour labelling techniques to suffer from a major shortcoming in that there has been no analytical method by which to identify, isolate and reconstruct the morphology of an individual cell in 3D (Weissman and Pan, 2015).

We overcome this significant methodological limitation in Chapter 3. We not only describe a method by which to reliably identify the entirety of a single cell based on its unique colour identity, but additionally extract those voxels that constitute it and reconstruct its morphology in 3D. This allows a researcher, for the first time, to quantitatively describe and characterise the 3 dimensional structure of a single cell following labelling of a population, as we do, by describing the dendritic field size, dendritic length, the number of dendritic branch points and the complexity of ipRGCs in the rd1 retina. An important aspect of this work is that not only do we reconstruct our cells in 3D; but each voxel is indexed with its true spatial position within the original image of the biological tissue, thus maintaining the spatial relationships of individual cells. As demonstrated in Chapter 3, we can readily isolate numerous cells from an individual image using our PCA approach (Chapter 3; Figure 3.6). This allows a researcher to further characterise the connectivity and interactions of neuronal networks across the brain whilst retaining this crucial aspect of single cell resolution. Our analysis therefore not only complements the vast array of multi-colour labelling techniques currently in use, but allows them to realise their full potential.

187

Chapter 6 General Discussion

A second impact of our work is its accessibility and versatility to the wider scientific community. Many of the alternative approaches currently employed to reconstruct single cells rely on expensive microscopy techniques which only work on small tissue regions, or by labelling a single cell in an individual animal. Both these approaches are time consuming, with low throughput and are expensive. Here, we have established a pioneering method which employs a commercially available virus, a conventional confocal microscope (readily available in many research institutions), along with custom written software which can be integrated with a number of different bio-imaging software packages. The ability to label hundreds of neurones and isolate and recreate multiple cells from a single animal additionally addresses ethical issues surrounding the use of animals in research by maximising data collection during in-vivo work. With an ever-growing variety of Cre-driven mouse lines to compliment these multi-colour labelling techniques, this software offers enormous opportunities to answer a wide variety of scientific questions ranging from cell lineage tracking (Garcia-Moreno et al., 2014, Gomez-Nicola et al., 2014, Loulier et al., 2014) neuronal connectivity (Hammer et al., 2015, Lakadamyali et al., 2012) and could be enhanced to answer questions regarding neuronal architecture in health and disease (as we begin to demonstrate in chapter 3) and single cell tracing of populations of neurones within and across entire brain regions.

6.2 General discussion about experimental techniques

Transgenic mouse models of degeneration Throughout this study we rely on the C3H rd/rd cl and C57 rd1 mice models of retinal degeneration. The rd1 mouse is one of the most studied mouse models of retinal degeneration due to the wealth of data available on the effect of the PDEβ6 mutation on rod cells (Farber and Lolley, 1974, Bowes et al., 1990), the consequential stages of cell death and retinal re-organisation (Jones and Marc, 2005, Marc et al., 2007, Strettoi et al., 2002) and the electrophysiological properties of the remnant neural retina (Choi et al., 2014, Ivanova et al., 2015, Stasheff, 2008). The rd1 mouse has additionally been crucial in identifying the contribution of melanopsin signalling in isolation of rod and cone photoreception (Brown et al., 2010, Lucas et al., 2001b, Freedman et al., 1999) which is challenging to achieve in visually intact animals without techniques such as silent substitution.

As with any animal model, one must consider a number of limitations when interpreting data recorded from them. First is a developmental aspect. The murine visual system does not reach full maturation until approximately P30 in wildtype mice (Hooks and Chen, 2007) with many aspects of development also reliant on sensory input - the so called critical period (Wiesel and Hubel, 1963, Hooks and Chen, 2008). However, many human conditions of inherited retinal degenerations occur in post-adolescent life (Delyfer et al., 2004, Hartong et al., 2006) once the visual system has reached

188

Chapter 6 General Discussion

full maturity. Therefore, with over 97% of murine photoreceptors rendered non-functional from birth in the rd1 mouse (Strettoi et al., 2003), this begs the question of what aspects of visual function may not fully develop in these animals?

A second aspect to consider arises from the variety of transgenic models of retinal degeneration which are now available (Chang et al., 2002). The degeneration and re-organisation of retinal circuitry is a common phenomenon observed across a number of different rodent models; however the speed and specificity varies greatly and has been shown to be dependent on the specific mutation which caused it (Chang et al., 2002, Jones and Marc, 2005). This is most clearly highlighted in a publication which examined the differences between the C3H rd1 mouse and the C57 rd1 mouse (Stasheff, 2008) both of which we employ in our investigations. Whilst the rd1 mutation is naturally occurring in the C3H mouse strain, it is chemically induced in the C57 strain. This results in the point mutation occurring on a different exon of the PDEβ6 gene. This difference in strain and position of the mutation result in significant phenotypic differences in regards to both the rate of degeneration of specific retinal pathways and the electrophysiological properties of the degenerating retina (Stasheff, 2008).

This raises the question of whether the current variety of degenerative models should be considered a heterologous population which show specific alliance with an analogous human condition, and indicates that great care must be taken so that the model of degeneration utilised is applicable to the scientific question at hand. After some review, the studies presented in this thesis may benefit from use of the rd10 mouse model of retinitis pigmentosa. The rd10 mouse possesses a missense mutation on exon 13 of the PDEβ6 gene (Chang et al., 2000) that results in interrupted photoreceptor signalling and consequential death of the rod and cone photoreceptors. However, this degeneration is significantly slower than that which occurs in rd1 mice (Gargini et al., 2007). Rod photoreceptors do not show abnormalities until approximately P16 in the rd10 mouse (Pennesi et al., 2012) and both light and dark adapted ERGs can be recorded until P50 (Chang et al., 2007). The reason for this later onset is believed to be that the mutation in the PDEβ6 gene in rd10 mouse does not render the cGMP phosphodiesterase functionless, but that its function is compromised, showing lower affinity for cGMP, and consequently a slower rate of catalysation (Chang et al., 2007). Accordingly, relatively normal visual function can occur at the early stages of life, but the concentration of cGMP accumulates over time in the rod outer segments, eventually become toxic and trigging photoreceptor degeneration and subsequent retinal remodelling.

As these mice are born with both functional rod and cone photoreceptors this would allow for a more natural development of the visual system before degeneration of the outer retina is complete (Chang et al., 2007, Pennesi et al., 2012, Gargini et al., 2007, Stasheff et al., 2011) and would therefore be more reflective of cases of inherited retinal degenerations observed in humans

189

Chapter 6 General Discussion

(Tsujikawa et al., 2008, Delyfer et al., 2004). The slower rate of degeneration would additionally allow for a wider epoch within which to conduct electrophysiological recordings.

Secondly, the colony of Opn4Cre mice requires further discussion. These mice were used to generate our rd/rd Opn4Cre/+ animals in Chapter 3 and were to act as a control for the structural integrity of ipRGCs in the degenerate retina using our new analysis. However, due to their original design for experiments by Sammer Hatter, they encode a floxed eYFP reporter for Cre-recombinase expression (Ecker et al., 2010). Therefore, in our heterozygote animals, eYFP was additionally expressed in ipRGCs. This was a significant problem for our Brainbow experiments in Chapter 3 as it resulted in a uniform baseline expression of eYFP in the somas of ipRGCs. This fluorescent signal could not be separated from the stochastic expression of the eYFP fluorophore encoded in our Brainbow virus. Interestingly, this did not affect our experiments in rd/rd Opn4Cre/+ mice as we believe the floxed eYFP reporter was likely bred out of the colony during creation of this new transgenic model. As there are no commercially available Opn4Cre mice which do not contain this reporter, the most pragmatic approach would be to set up a breeding strategy between homozygous Opn4Cre/Cre mice and C57 wildtype mice in an attempt to breed out the floxed eYFP from the genome. However, this does have ethical considerations due to the number of animals generated which may not be experimentally valuable.

Anaesthesia All the in-vivo electrophysiological data demonstrated in this thesis is collected from isoflurane or urethane anaesthetised mice. The slow temporal kinetics of the deafferented melanopsin light response in degenerate animals (Brown et al., 2010) and the long inter-stimulus interval required to ensure continued light-responsiveness to repeated stimulus presentations (Chapter 4) resulted in long duration experiments, often in excess of twelve hours in our rd/rd cl experiments. We therefore initially chose the inhalation anaesthetic isoflurane as it provides rapid induction, the ability to readily adjust the concentration of anaesthesia delivered, whilst also having minimal contraindications which would prevent premature death (Eger, 1981, Szczesny et al., 2004). However, at the beginning of these investigations it became clear that light responses in the dLGN were extremely sensitive to the depth of isoflurane anaesthesia (Figure 6.1). Most notably, the sustained component of the light-response, which melanopsin is known to influence (Brown et al., 2010) was most susceptible to these changes.

190

Chapter 6 General Discussion

Figure 6.1 Effects of isoflurane anaesthesia on light responses in the dLGN Representative light-responsive neurone from the dLGN of an Opn1mwR mouse shows the effect of increasing isoflurane concentration (labelled on the right) on neuronal activity in response to 10 repeats of a 20 second uniform light pulse (total irradiance = 8.72x1014 photons/cm2/s). Note the large reduction in both background firing rate and sustained activity in response to increasing isoflurane concentrations (0.6 – 2.0%), and how the activity returns when the concentration is returned to 0.75%.

Whilst this clearly indicates that isoflurane could influence visual responses in the dLGN, we took precautions against this during our recordings. Firstly we ensured the animal had both a stable breathing rate and spiking activity prior to any single repeat of the receptive field mapping protocol. Secondly, we only made small changes to isoflurane concentration if there was a significant change in the animal’s physiology (<0.2% increments). If a large change had to be made, that repeat was removed from the analysis and the entire rotation was restarted once the animal was stabilised. One method to obtain a more accurate description of the stability of isoflurane-induced anaesthesia would be to record physiological measurements such as heart rate, blood pressure and oxygen saturation. Alternatively, using a low dose combination of different anaesthetics may provide a more stable preparation for future experiments.

In Chapter 5, I use the injectable anaesthetic Urethane. As an injectable anaesthetic, following its administration one cannot control for the effect of the drug on physiology. Nonetheless, urethane is used in a number of electrophysiological recordings, including visual studies (Brown et al., 2010, Storchi et al., 2015, Howarth et al., 2014), where it provides stable anaesthesia with minimal effects on eye movements (Nair et al., 2011) or the autonomic and cardiovascular systems (Hara and Harris, 2002, Janssen et al., 2004, Maggi and Meli, 1986). Nonetheless, the anaesthetic state is physiologically unnatural and therefore the brain activity under these conditions is unlikely to reflect the true activity of awake, freely moving and behaving animals (Niell and Stryker, 2010, Maimon,

191

Chapter 6 General Discussion

2011). Thus, an attractive alternative to using anaesthesia would be to implant chronic recording electrodes into our retinally degenerate animals and perform similar recordings in either head-fixed or freely-moving mice. Not only does this overcome the requirement for use of anaesthetics, but allows for greater insight into the significance of these light-responses by correlating electrophysiological activity with behavioural output. An example of its potential use will be discussed later in this chapter.

Confirming electrode placement A third aspect which warrants discussion is the use of DII labelling. More prominent in the degenerate animals, was the difficulty in identifying light-responses in the first electrode placement. This resulted in multiple electrode placements prior to the final recording position and therefore multiple DII tracts could be present in the brain tissue upon inspection. This did sometimes make it problematic to identify the final resting place from which electrophysiological recordings were taken. This was of great importance when anatomically reconstructing individual electrode sites in order to confirm retinotopic organisation (Chapter 4; Figure 4.5A). Furthermore, DII is capable of travelling laterally through the lipid bilayer of cells following fixation (Godement et al., 1987), thus reducing the accuracy in identifying the exact resting position of the electrode. One approach to overcome this could be to use electrical lesions through the electrode contacts at the end of the experiment. This would result in a more accurate and uniquely identifiable marker of the final recording position which would additionally account for any tissue shrinkage during the fixation process. One consideration is that with time, electrical lesioning can result in damage to individual channels on the recording electrode.

Handling large data volumes Throughout the design and application of our multi-colour analysis software in Chapter 3, one of the biggest limitations we faced was computer processing power. The significant progress in microscopy techniques and image acquisition allowed us to collect 3D images with high spatial resolution, in multiple colour channels and over a wide volume of biological tissue. The consequence of this was finding a method by which to process and analyse these vast files sizes, often in excess of 20 GB when imaging the retina, and even larger when imaging ipRGC projection targets in the brain. Whilst our programming methods were designed to maximise the use of the computers resources (Chapter 3; Methods), analysing this wealth of 3 dimensional spatial and colour information became increasingly problematic as the processing power and available memory on computers which were licensed to use IMARIS were limited.

192

Chapter 6 General Discussion

Our approach in the development of the Brainbow analysis was to additionally develop a method that would be readily usable for any research institution. As part of the university we also have access to the Campus wider server system which possesses 4TB of RAM and far superior processing power. Whilst this would involve some re-programming of our current software to create a stand- alone version, this would not only improve the speed of analysis, it would additionally open up further avenues for improvement in both the imaging methodology and analysis, which are introduced below.

Firstly, whilst confocal microscopy provides sufficient images, one of the most challenging aspects of Brainbow imaging is maximising both the colour diversity and brightness range during image acquisition over such a large region of biological tissue. This is difficult to achieve as the laser power and magnification factor are inversely related. One method to overcome this could be to use other imaging techniques such as light-sheet microscopy. This technique allows one to image far larger tissue specimens by presenting sheets of light perpendicular to the objective (Reynaud et al., 2015). Rather than imaging at high magnification, one can image with high spatial resolution by increasing the pixilation under a lower magnification, thereby maintaining large regions of the tissue in the field of view. This microscopy technique maintains both the high spatial and colour resolution of confocal microscopy whilst additionally increasing imaging depth and speed, and reducing the rate of fluorophore bleaching. When looking to use this technique in the brain, it would allow for one to image deeper into tissue and therefore remove the requirement to section the brain prior to imaging. Combining this microscopy technique with novel methods to clear tissue such as CLARITY (Chung and Deisseroth, 2013) is an attractive proposition as it would further increase imaging depth for neuronal tracing over an entire nucleus, such as the dLGN.

Secondly, in regards to the data analysis, further computer power would allow us to readily analyse far larger areas of biological tissue (ROIs). This would be vital when applying this technology to answering questions in the mammalian brain as discussed above. It would additionally allow us to design our analysis to incorporate the fourth fluorophore (tagBFP) in the Brainbow construct (Chapter 3; Figure 3.2A) and would vastly increase our reliability and success of single cell isolation (Chapter 3; Discussion). It would however result in the generation of a fourth colour channel during image acquisition. This would increase the current data sizes by approximately a third and further highlights the requirement for greater computer power in order to analyse these rich data sets.

193

Chapter 6 General Discussion

6.3 Future work

Identifying and improving temporal deficits in the rd1 retina The poor temporal delay of the deafferented melanopsin response is the most significant limitation to melanopsin supporting behaviourally relevant visual perception in advanced retinal degeneration. Trying to elucidate the cause of this, and whether one can rectify this limitation, would increase the practicality of melanopsin signalling being employed for spatial navigation.

In the healthy visual system, the intrinsic melanopsin response is comparatively slower than that of rod and cone photoreceptors due to the low density expression of the melanopsin photopigment in ipRGCs (Do et al., 2009). However, this is compensated for by synaptic input arriving from the outer retina which maintains an ipRGCs resting membrane potential just below threshold for action potential generation (Zhao et al., 2014a, Wong et al., 2007). Thus, ipRGCs are in a “primed” state and readily depolarise in response to light (Wong et al., 2007, Do et al., 2009). However, following outer retinal degeneration there is a global re-organisation of the remnant neural retina with both the anatomy and physiology of inputs to the Ganglion cell layer significantly altered (Choi et al., 2014, Menzler and Zeck, 2011, Stasheff et al., 2011). Thus, one possibility is that this pathophysiological state of the rd1 retina influences the normal electrophysiological properties of an ipRGCs membrane potential such that they are no longer in this primed state at rest. As such, a higher irradiance is required to elicit a deafferented melanopsin-driven response in the rd1 retina. Alternatively, if the stimulus intensity is low, then an ipRGC requires a longer time over which to integrate this light information before the membrane potential can reach threshold to fire an action potential (Do et al., 2009).

A second possibility for the poor temporal kinetics of the deafferented melanopsin response in degeneration is related to the debated method by which melanopsin regenerates its retinaldehyde chromophore (Lucas, 2006, Mure et al., 2009). Rod and cone photoreceptors utilise a metabolic pathway shared between themselves and the retinal pigment epithelium (RPE) in the outer retina (Kiser et al., 2012, McBee et al., 2001). Müller cells have also been implicated in supporting chromophore regeneration for cones in daylight conditions (Kaylor et al., 2013). However, following outer photoreceptor degeneration, both the RPE and Müller cells show abnormal molecular and anatomical phenotypes (Jones and Marc, 2005). In numerous models of degeneration, the RPE invades the sub-retinal space as it begins to degenerate (Marc et al., 2003). This could significantly impact the supply of 11-cis-retinaldehyde available to the melanopsin photopigment under long- duration light exposure or repetitive stimulations. This hypothesis is supported by a recent study which found that removal of the RPE during MEA recordings from wildtype retinas significantly attenuates the melanopsin driven light response (Zhao et al., 2016). Additionally, Müller cells become hypertrophic in phase II of retinal degeneration and form a glial seal which segregates the

194

Chapter 6 General Discussion

sub-retinal space from surviving retinal neurones (Jones and Marc, 2005). The striking anatomical and physiological differences of Müller cells in the rd1 retina compared to that of visually intact mice (Chua et al., 2013) would likely compromise any potential supportive role they may perform for ipRGCs, whilst also interfering with the transfer of recycled chromophore from the remaining RPE.

Both these questions could be investigated using in-vitro electrophysiology. Patch clamping of individual ipRGCs in the rd1 retina would provide great insight into the physiological properties of these cells in advanced stages of retinal degeneration, as has been described previously in the wildtype retina (Zhao et al., 2014a). This technique would also allow for one to identify which ion currents, and hence receptors, may be specifically implicated in the pathophysiological state. One could then use a variety of pharmacological agents, which are known to specifically modulate individual receptors in the retina, in an attempt to restore the normal resting membrane potential of ipRGCs and determine whether this rescues some of the temporal deficits of the intrinsic melanopsin light response in degeneration. With respect to chromophore regeneration, one can readily apply different concentrations of the retinaldehyde chromophore to the aCSF (artificial cerebrospinal fluid) in order to determine how this affects the reliability and kinetics of the deafferented melanopsin response to repeated light stimulation.

Whilst understanding why the temporal resolution of these responses is significantly altered is of vital importance, it may be the case that the deafferented melanopsin response does not have the capacity to support behaviourally relevant vision at advanced stages of retinal degeneration. In the non-image forming system rods, cones and melanopsin form a synergistic photoreceptive system which complement each other in the intensity, spectral and temporal domains (Lall et al., 2010, McDougal and Gamlin, 2010, Lucas et al., 2012). This formed our original hypothesis which spurred our investigation in Chapter 5, and we therefore looked at earlier stages of retinal degeneration whereby residual cone-driven light-responses may compensate for the poor temporal kinetics of the intrinsic melanopsin light-response, and as a consequence, could provide more significant visual information to the dLGN. This is an unanswered question from this thesis and requires further investigation. The most pragmatic approach to resolving this outstanding question would be to create an rd10 Opn1mwR mouse line which would allow one to use the same silent substitution paradigms described in Chapter 5, in combination with both in-vitro and in-vivo electrophysiology.

However, there are currently no means by which to prevent the secondary degeneration of cone photoreceptors following rod cell death (Jones and Marc, 2005, Marc et al., 2003, Strettoi, 2015). Therefore any contribution melanopsin may make during this progressive stage would likely be short lived. Accordingly, a final approach may be to try and artificially increase the temporal sensitivity of melanopsin signalling. One publication has indicated that there is a significant decrease in the expression levels of melanopsin mRNA in degenerate rats (Sakamoto et al., 2004). Therefore a simple method to approach this would be to target ipRGCs with the introduction of more

195

Chapter 6 General Discussion

melanopsin, via intravitreal delivery of viral vectors, for example, in our rd/rd Opn4Cre/+ mouse line. The aim of this would be to increase melanopsin expression in ipRGCs and therefore improve the rate of photon capture and, in turn, the speed of action potential generation (Do et al., 2009). Melanopsin has previously been introduced to a heterogeneous population of retinal ganglion cells in rd1 mice with promise (Lin et al., 2008). However, one must consider whether this would be a fruitful approach. The deafferented melanopsin response exhibits both slow onset and offset kinetics to light (Chapter 4). The introduction of further melanopsin may decrease the latency of the onset kinetics, but would likely have no effect, if not exacerbate, the duration of the offset kinetics (Lin et al., 2008).

More recently, a number of other visually responsive photopigments have restored visual responses to the rd1 retina which demonstrate far superior temporal kinetics than melanopsin (Cehajic- Kapetanovic et al., 2015, van Wyk et al., 2015, Mace et al., 2015). Thus, one possible approach may be to target the delivery of gene therapy strategies to cells which signal via ipRGCs, as this may result in a reduced onset kinetics for the intrinsic melanopsin response by driving more rapid changes in membrane potential. This combinatorial strategy may also increase the possibility of recapturing some of the diverse and pivotal roles these two functionally distinct photoreceptors possess in visual perception within the healthy retina. Furthermore, the identification of spatial receptive fields and contrast sensitivity in the dLGN during progressive stages of retinal degeneration (Chapter 5) establishes that some retinal circuits upstream of ipRGCs survive these early stages. Therefore, early intervention using these gene therapy strategies may provide the best chance of recapturing the natural visual state (Busskamp et al., 2010).

Improving the quality of intrinsic melanopsin signalling in the rd1 retina If the temporal resolution of the deafferented melanopsin response could be recovered, a number of obstacles would still need to be overcome if melanopsin signalling could be employed for spatial vision following retinal degeneration. Firstly, both the M4 and M5 ipRGCs demonstrate weak intrinsic light-responses (Estevez et al., 2012, Zhao et al., 2014a). As a consequence, they demonstrate a modest response amplitude in both the retina and dLGN under light adapted conditions (Davis et al., 2015, Estevez et al., 2012). The ability to detect these responses in rd1 mice is further complicated due to the increased baseline firing rate and oscillatory activity of the degenerating retina (Ivanova et al., 2015, Borowska et al., 2011, Trenholm et al., 2012). A number of pharmacological agents are currently employed which have been shown to dampen these effects in rd1 mice, the most notable of which is Meclofenamic acid (MFA) (Trenholm et al., 2012, Toychiev et al., 2013). MFA blocks gap junctions and therefore would inhibit action potential propagation via this route. Application of this drug in-vitro during electrophysiological recordings may therefore greatly improve the signal:noise ratio of the deafferented melanopsin response to light.

196

Chapter 6 General Discussion

The use of these drugs also has interesting implications for the spatial acuity of melanopsin-based vision. The up-regulation of gap junctions in the rd1 retina would allow for the deafferented melanopsin light-response to be shared between heterologous populations of retinal ganglion cells in the inner retina. This hypothesis is not only supported by previous investigations using calcium imaging in the rd/rd cl retina (Sekaran et al., 2003) but also unpublished data from the Lucas laboratory which demonstrates that there is a far higher number of melanopsin driven light- responsive units in the rd/rd cl retina than predicted based on the known proportion of ipRGCs in the mouse (Dr Cyril Eleftheriou; personal communication). One could therefore envisage that in our in-vitro recordings from Chapter 4, a single ipRGCs intrinsic response to spatially structured stimuli could be propagated via these inner retinal circuits to a heterologous population of RGCs. As these cells project in a retinotopic manner to the dLGN (Chapter 4), this may result in an individual neurone in the dLGN inadvertently monitoring a far larger region of the visual field then a single ipRGCs dendritic field permits. In support of this, we identify that on average, the spatial receptive field of an individual dLGN neurone is twice as large as that in the retina. This suggests up to four cells with a melanopsin-driven light response synapse with a single LGN neurone in the rd/rd cl mouse. In the wildtype mouse, on average four heterologous RGCs are believed to synapse with a single dLGN neurone (Hammer et al., 2015, Chen and Regehr, 2000).

Using multi-electrode array (MEA) recordings from the rd1 retina, one could map the spatial receptive fields of ipRGCs before and after application of MFA to determine if it reduces the size of the melanopsin-driven spatial receptive field. If successful, the demonstration of this in-vivo using electrophysiological recordings from the dLGN may greatly increase the spatial acuity of melanopsin based vision in advanced stages of retinal degeneration. An attractive prospect of the use of MFA is that it is a non-steroidal anti-inflammatory which is approved by the FDA, and therefore the success of this drug would provide a fast route through to clinical trials.

Melanopsin signalling in other visual processing structures The data presented in this thesis primarily focuses on the sensory input of ipRGCs and the early visual processing that occurs at the first synapse outside of the retina. However, the mammalian visual system is extremely complex and relies on both bottom-up sensory inputs from the retina and top-down interpretation from other visual processing structures such as the superior colliculus and visual cortex (Huberman and Niell, 2011). The superior colliculus has vital roles in integrating sensory and motor information in order to co-ordinate goal orientated behaviours (May, 2006, Feinberg and Meister, 2015). A direct projection of M4 cells from retina to Superior colliculus has been recently identified (Brown et al., 2010, Ecker et al., 2010) and appears to survive retinal degeneration (Chapter 3; Figure 3.1G). As rd/rd cl mice have been previously shown to demonstrate brightness perception in behavioural paradigms (Brown et al., 2012), and considering the function of this

197

Chapter 6 General Discussion

nucleus, it would be of great interest to characterise light-responses in the superior colliculus to provide greater insights into how this light-information is integrated to provide motor output.

Neurones of the dLGN also project to the striate cortex where they form the thalamo-cortical pathway. The striate cortex is spatially organised and is involved in more complex visual processing by integrating and sharing information with at least nine other accessory cortical structures (Huberman and Niell, 2011, Wang and Burkhalter, 2007, Marshel et al., 2011). Accordingly, it would also be important to identify if melanopsin-driven light-responses are transmitted to the striate cortex. This is currently unknown in the healthy visual system. Whilst light responses which exhibit the slow temporal kinetics of melanopsin phototransduction have been demonstrated in the striate cortex of degenerate mice using optical imaging (Brown et al., 2010) electrophysiological recordings are yet to identify any significant contribution from melanopsin in either visually intact or degenerate mice (Dr Kate Davies; personal communication). An additional level of complexity exists in retinally degenerate mice when trying to answer this question. The cortex, in comparison to the thalamus, is an extremely plastic structure which can readily undergo synaptic re-wiring, especially in the absence of sensory input (Antonini et al., 1999, Burton, 2003). Although our data in Chapter 4 suggest that a spatially organised anatomical projection survives, characterising what visual information it can encode and relay would greatly influence the extent to which melanopsin alone could support more complex visual behaviours.

A second question is that if the deafferented melanopsin response does reach these visual processing structures, is the melanopsin signal both interpretable and meaningful to the degenerate animal? One approach to answer these outstanding questions raised above would be to implant a chronic recording electrode into the superior colliculus or striate cortex and record melanopsin- driven responses in freely moving rd/rd cl animals and wildtype controls. These awake recording techniques have become increasingly popular, especially in the visual system as they allow for one to correlate electrophysiological responses with a behavioural output (Okun et al., 2016, Frenkel et al., 2006). For example, one could present a variety of analytical stimuli to head-fixed mice, such as drifting gratings or inverting chequer boards in order to assess the spatio-temporal resolution of these more complex responses under more naturalistic conditions, without the complications of anaesthesia described earlier. Alternatively, one could use comprehensive behavioural paradigms such as presentation of naturalistic movies or spatial navigation tasks in freely moving mice in order to gain greater insight into how any spatial light perception can be translated into a behavioural output, and over what timescales.

Structure function relationships of ipRGCs in retinorecipient nuclei The structure of ipRGC subtypes is well characterised in the retina of visually intact rodents (Schmidt et al., 2011b, Estevez et al., 2012, Ecker et al., 2010), and in Chapter 3 we provide a significant

198

Chapter 6 General Discussion

update to the literature to those found in the degenerate retina. Whilst we identify that the projection to the dLGN remains superficially intact through anatomical and electrophysiological recordings in this work, nothing has been determined about the structural morphology of an individual ipRGC and its connectivity in the dLGN. What is the spatial extent of a single ipRGC projection in the dLGN? Does one ipRGC communicate with a single dLGN neurone which would provide high spatial acuity? Alternatively, does a single ipRGC influence multiple cells to provide a more global irradiance signal as has been traditionally thought? Importantly relating to the questions asked in this thesis; is this connectivity altered in the degenerative state? The development of our novel colour analysis technique in Chapter 3 provides a method by which we could begin to characterise the anatomy of these projections in both health and disease in order to begin answering some of these questions.

Throughout this thesis there is an underlying theme of the structure function relationships of the five distinct ipRGC subtypes in the retina and the vast array of physiological and behavioural responses to light which they modulate (Brown et al., 2010, Schmidt et al., 2014, Hattar et al., 2006, Milosavljevic et al., 2016, Allen et al., 2014, Berson et al., 2002, Lucas et al., 2001a). Whilst these cells are characterised in the retina, little is known about their morphology in their respective projection targets (Fernandez et al., 2016). Using our multi-colour analysis software, we can begin to answer a significant outstanding question from the literature by characterising the morphology of individual ipRGC projections in retinorecipient nuclei and pairing them with a specific subtype in the retina. Do different ipRGC subtypes exhibit different morphologies in the same nucleus? Does a single ipRGC innervate multiple nuclei and exhibit different morphologies for differing physiological responses to light? These data would further our understanding of the importance of melanopsin signalling to the brain, how light information is integrated within different retinorecipient nuclei and therefore more readily identify the role and function of individual ipRGCs with subtype specific resolution.

199

Chapter 7 General References

Chapter 7: General References

This Chapter contains the references from Chapters 1, 2 and 6. References for Chapters 3-5 are included at the end of their respective Chapter.

AHNELT, P. K. & KOLB, H. 2000. The mammalian photoreceptor mosaic-adaptive design. Progress in Retinal and Eye Research, 19, 711-777. ALLEN, A. E., CAMERON, M. A., BROWN, T. M., VUGLER, A. A. & LUCAS, R. J. 2010. Visual Responses in Mice Lacking Critical Components of All Known Retinal Phototransduction Cascades. Plos One, 5. ALLEN, A. E. & LUCAS, R. J. 2016. Using Silent Substitution to Track the Mesopic Transition From Rod- to Cone-Based Vision in Mice. Investigative Ophthalmology & Visual Science, 57, 276-287. ALLEN, A. E., STORCHI, R., MARTIAL, F. P., PETERSEN, R. S., MONTEMURRO, M. A., BROWN, T. M. & LUCAS, R. J. 2014. Melanopsin-Driven Light Adaptation in Mouse Vision. Current Biology, 24, 2481-2490. ANDERSON, E. E., GREFERATH, U. & FLETCHER, E. L. 2016. Changes in morphology of retinal ganglion cells with eccentricity in retinal degeneration. Cell and Tissue Research, 364, 263-271. ANTONINI, A., FAGIOLINI, M. & STRYKER, M. P. 1999. Anatomical correlates of functional plasticity in mouse visual cortex. Journal of Neuroscience, 19, 4388-4406. APPLEBURY, M. L., ANTOCH, M. P., BAXTER, L. C., CHUN, L. L. Y., FALK, J. D., FARHANGFAR, F., KAGE, K., KRZYSTOLIK, M. G., LYASS, L. A. & ROBBINS, J. T. 2000. The murine cone photoreceptor: A single cone type expresses both S and M opsins with retinal spatial patterning. Neuron, 27. ARSHAVSKY, V. Y. & BOWNDS, M. D. 1992. REGULATION OF DEACTIVATION OF PHOTORECEPTOR G-PROTEIN BY ITS TARGET ENZYME AND CGMP. Nature, 357, 416-417. ARSHAVSKY, V. Y. & WENSEL, T. G. 2013. Timing Is Everything: GTPase Regulation in Phototransduction. Investigative Ophthalmology & Visual Science, 54, 7725-7733. BACCUS, S. A., OELVECZKY, B. P., MANU, M. & MEISTER, M. 2008. A retinal circuit that computes object motion. Journal of Neuroscience, 28, 6807-6817. BARRETT, J. M., DEGENAAR, P. & SERNAGOR, E. 2015. Blockade of pathological retinal ganglion cell hyperactivity improves optogenetically evoked light responses in rd1 mice. Frontiers in Cellular Neuroscience, 9. BAVER, S. B., PICKARD, G. E., SOLLARS, P. J. & PICKARD, G. E. 2008. Two types of melanopsin retinal ganglion cell differentially innervate the hypothalamic suprachiasmatic nucleus and the olivary pretectal nucleus. European Journal of Neuroscience, 27. BAYLOR, D. A., FUORTES, M. G. F. & OBRYAN, P. M. 1971. RECEPTIVE FIELDS OF CONES IN RETINA OF TURTLE. Journal of Physiology-London, 214, 265-&. BELENKY, M. A., SMERASKI, C. A., PROVENCIO, I., SOLLARS, P. J. & PICKARD, G. E. 2003. Melanopsin retinal ganglion cells receive bipolar and amacrine cell synapses. Journal of Comparative Neurology, 460.

200

Chapter 7 General References

BERSON, D. M., CASTRUCCI, A. M. & PROVENCIO, I. 2010. Morphology and Mosaics of Melanopsin-Expressing Retinal Ganglion Cell Types in Mice. Journal of Comparative Neurology, 518. BERSON, D. M., DUNN, F. A. & TAKAO, M. 2002. Phototransduction by retinal ganglion cells that set the circadian clock. Science, 295. BIEL, M., SEELIGER, M., PFEIFER, A., KOHLER, K., GERSTNER, A., LUDWIG, A., JAISSLE, G., FAUSER, S., ZRENNER, E. & HOFMANN, F. 1999. Selective loss of cone function in mice lacking the cyclic nucleotide-gated channel CNG3. Proceedings of the National Academy of Sciences of the United States of America, 96, 7553-7557. BLOOMFIELD, S. A. & DACHEUX, R. F. 2001. Rod vision: Pathways and processing in the mammalian retina. Progress in Retinal and Eye Research, 20. BLOOMFIELD, S. A. & VOLGYI, B. 2009. The diverse functional roles and regulation of neuronal gap junctions in the retina. Nature Reviews Neuroscience, 10, 495-506. BONIN, V., HISTED, M. H., YURGENSON, S. & REID, R. C. 2011. Local Diversity and Fine-Scale Organization of Receptive Fields in Mouse Visual Cortex. Journal of Neuroscience, 31, 18506-18521. BORGHUIS, B. G., LOOGER, L. L., TOMITA, S. & DEMB, J. B. 2014. Kainate Receptors Mediate Signaling in Both Transient and Sustained OFF Bipolar Cell Pathways in Mouse Retina. Journal of Neuroscience, 34, 6128-6139. BOROWSKA, J., TRENHOLM, S. & AWATRAMANI, G. B. 2011. An Intrinsic Neural Oscillator in the Degenerating Mouse Retina. Journal of Neuroscience, 31, 5000-5012. BOULINA, M., SAMARAJEEWA, H., BAKER, J. D., KIM, M. D. & CHIBA, A. 2013. Live imaging of multicolor-labeled cells in Drosophila. Development, 140, 1605-1613. BOWES, C., LI, T. S., DANCIGER, M., BAXTER, L. C., APPLEBURY, M. L. & FARBER, D. B. 1990. RETINAL DEGENERATION IN THE RD MOUSE IS CAUSED BY A DEFECT IN THE BETA- SUBUNIT OF ROD CGMP-PHOSPHODIESTERASE. Nature, 347, 677-680. BRAMLEY, J. R., WILES, E. M., SOLLARS, P. J. & PICKARD, G. E. 2011. Carbenoxolone Blocks the Light-Evoked Rise in Intracellular Calcium in Isolated Melanopsin Ganglion Cell Photoreceptors. Plos One, 6. BRIDGES, C. D. B. 1959. VISUAL PIGMENTS OF SOME COMMON LABORATORY MAMMALS. Nature, 184, 1727-1728. BRIGGMAN, K. L. & DENK, W. 2006. Towards neural circuit reconstruction with volume electron microscopy techniques. Current Opinion in Neurobiology, 16, 562-570. BROWN, T. M., GIAS, C., HATORI, M., KEDING, S. R., SEMO, M. A., COFFEY, P. J., GIGG, J., PIGGINS, H. D., PANDA, S. & LUCAS, R. J. 2010. Melanopsin Contributions to Irradiance Coding in the Thalamo-Cortical Visual System. Plos Biology, 8. BROWN, T. M., TSUJIMURA, S.-I., ALLEN, A. E., WYNNE, J., BEDFORD, R., VICKERY, G., VUGLER, A. & LUCAS, R. J. 2012. Melanopsin-Based Brightness Discrimination in Mice and Humans. Current Biology, 22. BURTON, H. 2003. Visual cortex activity in early and late blind people. Journal of Neuroscience, 23, 4005-4011. BUSSKAMP, V., DUEBEL, J., BALYA, D., FRADOT, M., VINEY, T. J., SIEGERT, S., GRONER, A. C., CABUY, E., FORSTER, V., SEELIGER, M., BIEL, M., HUMPHRIES, P., PAQUES, M., MOHAND-SAID, S., TRONO, D., DEISSEROTH, K., SAHEL, J. A., PICAUD, S. & ROSKA, B. 2010. Genetic Reactivation of Cone Photoreceptors Restores Visual Responses in Retinitis Pigmentosa. Science, 329, 413-417. CAI, D., COHEN, K. B., LUO, T., LICHTMAN, J. W. & SANES, J. R. 2013. Improved tools for the Brainbow toolbox. Nature Methods, 10, 540-+. CALDERONE, J. B. & JACOBS, G. H. 1995. REGIONAL VARIATIONS IN THE RELATIVE SENSITIVITY TO UV-LIGHT IN THE MOUSE RETINA. Visual Neuroscience, 12, 463- 468.

201

Chapter 7 General References

CALVERT, P. D., KRASNOPEROVA, N. V., LYUBARSKY, A. L., ISAYAMA, T., NICOLO, M., KOSARAS, B., WONG, G., GANNON, K. S., MARGOLSKEE, R. F., SIDMAN, R. L., PUGH, E. N., MAKINO, C. L. & LEM, J. 2000. Phototransduction in transgenic mice after targeted deletion of the rod transducin alpha-subunit. Proceedings of the National Academy of Sciences of the United States of America, 97, 13913-13918. CARTERDAWSON, L. D. & LAVAIL, M. M. 1979. RODS AND CONES IN THE MOUSE RETINA .1. STRUCTURAL-ANALYSIS USING LIGHT AND ELECTRON-MICROSCOPY. Journal of Comparative Neurology, 188, 245-262. CARTERDAWSON, L. D., LAVAIL, M. M. & SIDMAN, R. L. 1978. DIFFERENTIAL EFFECT OF RD MUTATION ON RODS AND CONES IN MOUSE RETINA. Investigative Ophthalmology & Visual Science, 17, 489-498. CASAGRANDE, V. A. & BOYD, J. D. 1996. The neural architecture of binocular vision. Eye, 10, 153-160. CEHAJIC-KAPETANOVIC, J., ELEFTHERIOU, C., ALLEN, A. E., MILOSAVLJEVIC, N., PIENAAR, A., BEDFORD, R., DAVIS, K. E., BISHOP, P. N. & LUCAS, R. J. 2015. Restoration of Vision with Ectopic Expression of Human Rod Opsin. Current Biology, 25, 2111-2122. CEHAJIC-KAPETANOVIC, J., LE GOFF, M. M., ALLEN, A., LUCAS, R. J. & BISHOP, P. N. 2011. Glycosidic enzymes enhance retinal transduction following intravitreal delivery of AAV2. Molecular Vision, 17, 1771-1783. CERVETTO, L., LAGNADO, L., PERRY, R. J., ROBINSON, D. W. & MCNAUGHTON, P. A. 1989. EXTRUSION OF CALCIUM FROM ROD OUTER SEGMENTS IS DRIVEN BY BOTH SODIUM AND POTASSIUM GRADIENTS. Nature, 337, 740-743. CHANG, B., HAWES, N. L., HURD, R. E., DAVISSON, M. T., NUSINOWITZ, S. & HECKENLIVELY, J. R. 2000. A new mouse retinal degeneration (rd10) caused by a missense mutation in exon 13 of the beta-subunit of rod phosphodiesterase gene. Investigative Ophthalmology & Visual Science, 41, S533-S533. CHANG, B., HAWES, N. L., HURD, R. E., DAVISSON, M. T., NUSINOWITZ, S. & HECKENLIVELY, J. R. 2002. Retinal degeneration mutants in the mouse. Vision Research, 42. CHANG, B., HAWES, N. L., PARDUE, M. T., GERMAN, A. M., HURD, R. E., DAVISSON, M. T., NUSINOWITZ, S., RENGARAJAN, K., BOYD, A. P., SIDNEY, S. S., PHILLIPS, M. J., STEWART, R. E., CHAUDHURY, R., NICKERSON, J. M., HECKENLIVELY, J. R. & BOATRIGHT, J. H. 2007. Two mouse retinal degenerations caused by missense mutations in the beta-subunit of rod cGMP phosphodiesterase gene. Vision Research, 47, 624-633. CHEN, C. F. & REGEHR, W. G. 2000. Developmental remodeling of the retinogeniculate synapse. Neuron, 28, 955-966. CHEN, C. K., BURNS, M. E., HE, W., WENSEL, T. G., BAYLOR, D. A. & SIMON, M. I. 2000. Slowed recovery of rod photoresponse in mice lacking the GTPase accelerating protein RGS9-1. Nature, 403, 557-560. CHEN, C. K., BURNS, M. E., SPENCER, M., NIEMI, G. A., CHEN, J., HURLEY, J. B., BAYLOR, D. A. & SIMON, M. I. 1999. Abnormal photoresponses and light-induced apoptosis in rods lacking rhodopsin kinase. Proceedings of the National Academy of Sciences of the United States of America, 96, 3718-3722. CHEN, H., LIU, X. & TIAN, N. 2014. Subtype-dependent postnatal development of direction- and orientation-selective retinal ganglion cells in mice. Journal of Neurophysiology, 112, 2092-2101. CHEN, X., HSUEH, H. A., GREENBERG, K. & WERBLIN, F. S. 2010. Three Forms of Spatial Temporal Feedforward Inhibition Are Common to Different Ganglion Cell Types in Rabbit Retina. Journal of Neurophysiology, 103, 2618-2632. CHEW, K. S., SCHMIDT, T. M., RUPP, A. C., KOFUJI, P. & TRIMARCHI, J. M. 2014. Loss of G(q11) Genes Does Not Abolish Melanopsin Phototransduction. Plos One, 9.

202

Chapter 7 General References

CHOI, H., ZHANG, L., CEMBROWSKI, M. S., SABOTTKE, C. F., MARKOWITZ, A. L., BUTTS, D. A., KATH, W. L., SINGER, J. H. & RIECKE, H. 2014. Intrinsic bursting of AII amacrine cells underlies oscillations in the rd1 mouse retina. Journal of Neurophysiology, 112, 1491-1504. CHUA, J., NIVISON-SMITH, L., FLETCHER, E. L., TRENHOLM, S., AWATRAMANI, G. B. & KALLONIATIS, M. 2013. Early remodeling of muller cells in the rd/rd mouse model of retinal dystrophy. Journal of Comparative Neurology, 521, 2439-2453. CHUNG, K. & DEISSEROTH, K. 2013. CLARITY for mapping the nervous system. Nature Methods, 10, 508-513. COHEN, E. D. & MILLER, R. F. 1994. THE ROLE OF NMDA AND NON-NMDA EXCITATORY AMINO-ACID RECEPTORS IN THE FUNCTIONAL-ORGANIZATION OF PRIMATE RETINAL GANGLION-CELLS. Visual Neuroscience, 11, 317-332. COOMBS, J., VAN DER LIST, D., WANG, G. Y. & CHALUPA, L. M. 2006. Morphological properties of mouse retinal ganglion cells. Neuroscience, 140, 123-136. CROOKS, J. & KOLB, H. 1992. LOCALIZATION OF GABA, GLYCINE, GLUTAMATE AND TYROSINE-HYDROXYLASE IN THE HUMAN RETINA. Journal of Comparative Neurology, 315, 287-302. CRUZ-MARTIN, A., EL-DANAF, R. N., OSAKADA, F., SRIRAM, B., DHANDE, O. S., NGUYEN, P. L., CALLAWAY, E. M., GHOSH, A. & HUBERMAN, A. D. 2014. A dedicated circuit links direction-selective retinal ganglion cells to the primary visual cortex. Nature, 507, 358-+. CZEISLER, C. A., SHANAHAN, T. L., KLERMAN, E. B., MARTENS, H., BROTMAN, D. J., EMENS, J. S., KLEIN, T. & RIZZO, J. F. 1995. SUPPRESSION OF MELATONIN SECRETION IN SOME BLIND PATIENTS BY EXPOSURE TO BRIGHT LIGHT. New England Journal of Medicine, 332. DACEY, D., PACKER, O. S., DILLER, L., BRAINARD, D., PETERSON, B. & LEE, B. 2000. Center surround receptive field structure of cone bipolar cells in primate retina. Vision Research, 40, 1801-1811. DACEY, D. M., LIAO, H. W., PETERSON, B. B., ROBINSON, F. R., SMITH, V. C., POKORNY, J., YAU, K. W. & GAMLIN, P. D. 2005. Melanopsin-expressing ganglion cells in primate retina signal colour and irradiance and project to the LGN. Nature, 433. DAMIANI, D., NOVELLI, E., MAZZONI, F. & STRETTOI, E. 2012. Undersized dendritic arborizations in retinal ganglion cells of the rd1 mutant mouse: A paradigm of early onset photoreceptor degeneration. Journal of Comparative Neurology, 520, 1406- 1423. DAVIS, K. E., ELEFTHERIOU, C. G., ALLEN, A. E., PROCYK, C. A. & LUCAS, R. J. 2015. Melanopsin-Derived Visual Responses under Light Adapted Conditions in the Mouse dLGN. PloS one, 10, e0123424-e0123424. DELYFER, M. N., LEVEILLARD, T., MOHAND-SAID, S., HICKS, D., PICAUD, S. & SAHEL, J. 2004. Inherited retinal degenerations: therapeutic prospects. Biology of the Cell, 96, 261- 269. DEMB, J. B. 2007. Cellular mechanisms for direction selectivity in the retina. Neuron, 55, 179-186. DENMAN, D. J. & CONTRERAS, D. 2016. On Parallel Streams through the Mouse Dorsal Lateral Geniculate Nucleus. Frontiers in Neural Circuits, 10. DEVRIES, S. H. 2000. Bipolar cells use kainate and AMPA receptors to filter visual information into separate channels. Neuron, 28, 847-856. DHINGRA, A., JIANG, M. S., WANG, T. L., LYUBARSKY, A., SAVCHENKO, A., BAR-YEHUDA, T., STERLING, P., BIRNBAUMER, L. & VARDI, N. 2002. Light response of retinal ON bipolar cells requires a specific splice variant of G alpha(o). Journal of Neuroscience, 22, 4878-4884.

203

Chapter 7 General References

DIAMOND, J. S. & COPENHAGEN, D. R. 1995. THE RELATIONSHIP BETWEEN LIGHT-EVOKED SYNAPTIC EXCITATION AND SPIKING BEHAVIOR OF SALAMANDER RETINAL GANGLION-CELLS. Journal of Physiology-London, 487, 711-725. DO, M. T. H., KANG, S. H., XUE, T., ZHONG, H., LIAO, H.-W., BERGLES, D. E. & YAU, K.-W. 2009. Photon capture and signalling by melanopsin retinal ganglion cells. Nature, 457. DRAGER, U. C. & HUBEL, D. H. 1978. STUDIES OF VISUAL FUNCTION AND ITS DECAY IN MICE WITH HEREDITARY RETINAL DEGENERATION. Journal of Comparative Neurology, 180, 85-114. ECKER, J. L., DUMITRESCU, O. N., WONG, K. Y., ALAM, N. M., CHEN, S.-K., LEGATES, T., RENNA, J. M., PRUSKY, G. T., BERSON, D. M. & HATTAR, S. 2010. Melanopsin- Expressing Retinal Ganglion-Cell Photoreceptors: Cellular Diversity and Role in Pattern Vision. Neuron, 67. EGER, E. I. 1981. ISOFLURANE - A REVIEW. Anesthesiology, 55, 559-576. EKESTEN, B. & GOURAS, P. 2005. Cone and rod inputs to murine retinal ganglion cells: Evidence of cone opsin specific channels. Visual Neuroscience, 22, 893-903. EMANUEL, A. J. & DO, M. T. H. 2015. Melanopsin Tristability for Sustained and Broadband Phototransduction. Neuron, 85, 1043-1055. ESQUIVA, G., LAX, P. & CUENCA, N. 2013. Impairment of Intrinsically Photosensitive Retinal Ganglion Cells Associated With Late Stages of Retinal Degeneration. Investigative Ophthalmology & Visual Science, 54, 4605-4618. ESTEVEZ, M. E., FOGERSON, P. M., ILARDI, M. C., BORGHUIS, B. G., CHAN, E., WENG, S., AUFERKORTE, O. N., DEMB, J. B. & BERSON, D. M. 2012. Form and function of the m4 cell, an intrinsically photosensitive retinal ganglion cell type contributing to geniculocortical vision. The Journal of neuroscience : the official journal of the Society for Neuroscience, 32. ESTEVEZ, O. & SPEKREIJSE, H. 1982. THE SILENT SUBSTITUTION METHOD IN VISUAL RESEARCH. Vision Research, 22, 681-691. FARAHBAKHSH, Z. T., HIDEG, K. & HUBBELL, W. L. 1993. PHOTOACTIVATED CONFORMATIONAL-CHANGES IN RHODOPSIN - A TIME-RESOLVED SPIN-LABEL STUDY. Science, 262, 1416-1419. FARBER, D. B. & LOLLEY, R. N. 1974. Cyclic Guanosine Monophosphate: Elevation in degeneration photoreceptor cells of the C3H mouse retina. Science, 186, 449-451. FARRENS, D. L., ALTENBACH, C., YANG, K., HUBBELL, W. L. & KHORANA, H. G. 1996. Requirement of rigid-body motion of transmembrane helices for light activation of rhodopsin. Science, 274, 768-770. FARROW, K. & MASLAND, R. H. 2011. Physiological clustering of visual channels in the mouse retina. Journal of Neurophysiology, 105, 1516-1530. FEI, Y. J. 2002. Cone neurite sprouting: An early onset abnormality of the cone photoreceptors in the retinal degeneration mouse. Molecular Vision, 8, 306-314. FEINBERG, E. H. & MEISTER, M. 2015. Orientation columns in the mouse superior colliculus. Nature, 519, 229-+. FERNANDEZ, D. C., CHANG, Y. T., HATTAR, S. & CHEN, S. K. 2016. Architecture of retinal projections to the central circadian pacemaker. Proceedings of the National Academy of Sciences of the United States of America, 113, 6047-6052. FESENKO, E. E., KOLESNIKOV, S. S. & LYUBARSKY, A. L. 1985. INDUCTION BY CYCLIC-GMP OF CATIONIC CONDUCTANCE IN PLASMA-MEMBRANE OF RETINAL ROD OUTER SEGMENT. Nature, 313, 310-313. FOSTER, R. G., ARGAMASO, S., COLEMAN, S., COLWELL, C. S., LEDERMAN, A. & PROVENCIO, I. 1993. PHOTORECEPTORS REGULATING CIRCADIAN BEHAVIOR - A MOUSE MODEL. Journal of Biological Rhythms, 8, S17-S23.

204

Chapter 7 General References

FOSTER, R. G., PROVENCIO, I., HUDSON, D., FISKE, S., DEGRIP, W. & MENAKER, M. 1991. CIRCADIAN PHOTORECEPTION IN THE RETINALLY DEGENERATE MOUSE (RD/RD). Journal of Comparative Physiology a-Sensory Neural and Behavioral Physiology, 169. FREDRIKSSON, R., LAGERSTROM, M. C., LUNDIN, L. G. & SCHIOTH, H. B. 2003. The G- protein-coupled receptors in the human genome form five main families. Phylogenetic analysis, paralogon groups, and fingerprints. Molecular Pharmacology, 63, 1256-1272. FREEDMAN, M. S., LUCAS, R. J., SONI, B., VON SCHANTZ, M., MUNOZ, M., DAVID-GRAY, Z. & FOSTER, R. 1999. Regulation of mammalian circadian behavior by non-rod, non- cone, ocular photoreceptors. Science, 284. FRENKEL, M. Y., SAWTELL, N. B., DIOGO, A. C. M., YOON, B. J., NEVE, R. L. & BEAR, M. F. 2006. Instructive effect of visual experience in mouse visual cortex. Neuron, 51, 339-349. FRIED, S. I., MUNCH, T. A. & WERBLIN, F. S. 2002. Mechanisms and circuitry underlying directional selectivity in the retina. Nature, 420, 411-414. FU, Y. & YAU, K.-W. 2007. Phototransduction in mouse rods and cones. Pflugers Archiv- European Journal of Physiology, 454. FU, Y. B., ZHONG, H. N., WANG, M. H. H., LUO, D. G., LIAO, H. W., MAEDA, H., HATTAR, S., FRISHMAN, L. J. & YAU, K. W. 2005. Intrinsically photosensitive retinal ganglion cells detect light with a vitamin A-based photopigment, melanopsin. Proceedings of the National Academy of Sciences of the United States of America, 102. GAO, E., DEANGELIS, G. C. & BURKHALTER, A. 2010. Parallel Input Channels to Mouse Primary Visual Cortex. Journal of Neuroscience, 30. GARCIA-MORENO, F., VASISTHA, N. A., BEGBIE, J. & MOLNAR, Z. 2014. CLoNe is a new method to target single progenitors and study their progeny in mouse and chick. Development, 141, 1589-1598. GARGINI, C., TERZIBASI, E., MAZZONI, F. & STRETTOI, E. 2007. Retinal organization in the retinal degeneration 10 (rd10) mutant mouse: A morphological and ERG study. Journal of Comparative Neurology, 500, 222-238. GAUB, B. M., BERRY, M. H., HOLT, A. E., ISACOFF, E. Y. & FLANNERY, J. G. 2015. Optogenetic Vision Restoration Using Rhodopsin for Enhanced Sensitivity. Molecular Therapy, 23, 1562-1571. GHOSH, K. K., BUJAN, S., HAVERKAMP, S., FEIGENSPAN, A. & WASSLE, H. 2004. Types of bipolar cells in the mouse retina (vol 469, pg 70, 2004). Journal of Comparative Neurology, 476, 202-203. GODEMENT, P., VANSELOW, J., THANOS, S. & BONHOEFFER, F. 1987. A STUDY IN DEVELOPING VISUAL SYSTEMS WITH A NEW METHOD OF STAINING NEURONS AND THEIR PROCESSES IN FIXED TISSUE. Development, 101, 697-713. GOEZ, D., STUDHOLME, K., LAPPI, D. A., ROLLAG, M. D., PROVENCIO, I. & MORIN, L. P. 2008. Targeted Destruction of Photosensitive Retinal Ganglion Cells with a Saporin Conjugate Alters the Effects of Light on Mouse Circadian Rhythms. Plos One, 3. GOMEZ-NICOLA, D., RIECKEN, K., FEHSE, B. & PERRY, V. H. 2014. In-vivo RGB marking and multicolour single-cell tracking in the adult brain. Scientific Reports, 4. GOO, Y. S., PARK, D. J., AHN, J. R. & SENOK, S. S. 2016. Spontaneous Oscillatory Rhythms in the Degenerating Mouse Retina Modulate Retinal Ganglion Cell Responses to Electrical Stimulation. Frontiers in Cellular Neuroscience, 9. GOVARDOVSKII, V. I., FYHRQUIST, N., REUTER, T., KUZMIN, D. G. & DONNER, K. 2000. In search of the visual pigment template. Visual Neuroscience, 17, 509-528.

205

Chapter 7 General References

GRAHAM, D. M., WONG, K. Y., SHAPIRO, P., FREDERICK, C., PATTABIRAMAN, K. & BERSON, D. M. 2008. Melanopsin ganglion cells use a membrane-associated rhabdomeric phototransduction cascade. Journal of Neurophysiology, 99. GRIEVE, K. L. 2005. Binocular visual responses in cells of the rat dLGN. Journal of Physiology-London, 566. GRUBB, M. S., ROSSI, F. M., CHANGEUX, J. P. & THOMPSON, I. D. 2003. Abnormal functional organization in the dorsal lateral geniculate nucleus of mice lacking the beta 2 subunit of the nicotinic acetylcholine receptor. Neuron, 40, 1161-1172. GRUBB, M. S. & THOMPSON, I. D. 2003. Quantitative characterization of visual response properties in the mouse dorsal lateral geniculate nucleus. Journal of Neurophysiology, 90. GUO, F., GOPAUL, D. N. & VANDUYNE, G. D. 1997. Structure of Cre recombinase complexed with DNA in a site-specific recombination synapse. Nature, 389, 40-46. GUPTA, V. & POSS, K. D. 2012. Clonally dominant cardiomyocytes direct heart morphogenesis. Nature, 484, 479-U102. HADJIECONOMOU, D., ROTKOPF, S., ALEXANDRE, C., BELL, D. M., DICKSON, B. J. & SALECKER, I. 2011. Flybow: genetic multicolor cell labeling for neural circuit analysis in Drosophila melanogaster. Nature Methods, 8, 260-U111. HAGINS, W. A., PENN, R. D. & YOSHIKAMI, S. 1970. DARK CURRENT AND PHOTOCURRENT IN RETINAL RODS. Biophysical Journal, 10, 380-+. HAMMER, S., MONAVARFESHANI, A., LEMON, T., SU, J. M. & FOX, M. A. 2015. Multiple Retinal Axons Converge onto Relay Cells in the Adult Mouse Thalamus. Cell Reports, 12, 1575-1583. HAMPEL, S., CHUNG, P., MCKELLAR, C. E., HALL, D., LOOGER, L. L. & SIMPSON, J. H. 2011. Drosophila Brainbow: a recombinase-based fluorescence labeling technique to subdivide neural expression patterns. Nature Methods, 8, 253-U102. HANKINS, M. W. & LUCAS, R. J. 2002. The primary visual pathway in humans is regulated accordina to long-term light exposure through the action of a nonclassical photopigment. Current Biology, 12. HARA, K. & HARRIS, R. A. 2002. The anesthetic mechanism of urethane: The effects on rotransmitter-gated ion channels. Anesthesia and Analgesia, 94, 313-318. HARTONG, D. T., BERSON, E. L. & DRYJA, T. P. 2006. Retinitis pigmentosa. Lancet, 368, 1795-1809. HARTWICK, A. T. E., BRAMLEY, J. R., YU, J., STEVENS, K. T., ALLEN, C. N., BALDRIDGE, W. H., SOLLARS, P. J. & PICKARD, G. E. 2007. Light-evoked calcium responses of isolated melanopsin-expressing retinal ganglion cells. Journal of Neuroscience, 27. HATORI, M., LE, H., VOLLMERS, C., KEDING, S. R., TANAKA, N., SCHMEDT, C., JEGLA, T. & PANDA, S. 2008. Inducible Ablation of Melanopsin-Expressing Retinal Ganglion Cells Reveals Their Central Role in Non-Image Forming Visual Responses. Plos One, 3. HATTAR, S., KUMAR, M., PARK, A., TONG, P., TUNG, J., YAU, K. W. & BERSON, D. M. 2006. Central projections of melanopsin-expressing retinal ganglion cells in the mouse. Journal of Comparative Neurology, 497. HATTAR, S., LIAO, H. W., TAKAO, M., BERSON, D. M. & YAU, K. W. 2002. Melanopsin- containing retinal. ganglion cells: Architecture, projections, and intrinsic photosensitivity. Science, 295. HATTAR, S., LUCAS, R. J., MROSOVSKY, N., THOMPSON, S., DOUGLAS, R. H., HANKINS, M. W., LEM, J., BIEL, M., HOFMANN, F., FOSTER, R. G. & YAU, K. W. 2003. Melanopsin and rod-cone photoreceptive systems account for all major accessory visual functions in mice. Nature, 424.

206

Chapter 7 General References

HAVERKAMP, S., GRUNERT, U. & WASSLE, H. 2001. The synaptic architecture of AMPA receptors at the cone pedicle of the primate retina. Journal of Neuroscience, 21, 2488-2500. HAVERKAMP, S., WASSLE, H., DUEBEL, J., KUNER, T., AUGUSTINE, G. J., FENG, G. P. & EULER, T. 2005. The primordial, blue-cone color system of the mouse retina. Journal of Neuroscience, 25, 5438-5445. HAYNES, L. & YAU, K. W. 1985. CYCLIC GMP-SENSITIVE CONDUCTANCE IN OUTER SEGMENT MEMBRANE OF CATFISH CONES. Nature, 317, 61-64. HECHT, S., SHLAER, S. & PIRENNE, M. 1942. Energy, Quanta, and Vision. Journal of General Physiology, 25, 819-840. HILLMAN, P., HOCHSTEIN, S. & MINKE, B. 1983. TRANSDUCTION IN INVERTEBRATE PHOTORECEPTORS - ROLE OF PIGMENT BISTABILITY. Physiological Reviews, 63, 668-772. HOESS, R., ABREMSKI, K. & STERNBERG, N. 1984. THE NATURE OF THE INTERACTION OF THE P1 RECOMBINASE CRE WITH THE RECOMBINING SITE LOXP. Cold Spring Harbor Symposia on Quantitative Biology, 49, 761-768. HOOKS, B. M. & CHEN, C. 2008. Vision triggers an experience-dependent sensitive period at the retinogeniculate synapse. Journal of Neuroscience, 28, 4807-4817. HOOKS, B. M. & CHEN, C. F. 2007. Critical periods in the visual system: Changing views for a model of experience-dependent plasticity. Neuron, 56, 312-326. HOON, M., OKAWA, H., DELLA SANTINA, L. & WONG, R. O. L. 2014. Functional architecture of the retina: Development and disease. Progress in Retinal and Eye Research, 42, 44-84. HOWARTH, M., WALMSLEY, L. & BROWN, T. M. 2014. Binocular Integration in the Mouse Lateral Geniculate Nuclei. Current Biology, 24, 1241-1247. HUBENER, M. 2003. Mouse visual cortex. Current Opinion in Neurobiology, 13. HUBERMAN, A. D. & NIELL, C. M. 2011. What can mice tell us about how vision works? Trends in Neurosciences, 34, 464-473. HUGHES, S., HANKINS, M. W., FOSTER, R. G. & PEIRSON, S. N. 2012a. Melanopsin phototransduction: slowly emerging from the dark. Progress in brain research, 199. HUGHES, S., RODGERS, J., HICKEY, D., FOSTER, R. G., PEIRSON, S. N. & HANKINS, M. W. 2016. Characterisation of light responses in the retina of mice lacking principle components of rod, cone and melanopsin phototransduction signalling pathways. Scientific Reports, 6, 28086. HUGHES, S., WATSON, T. S., FOSTER, R. G., PEIRSON, S. N. & HANKINS, M. W. 2013. Nonuniform Distribution and Spectral Tuning of Photosensitive Retinal Ganglion Cells of the Mouse Retina. Current Biology, 23, 1696-1701. HUGHES, S., WELSH, L., KATTI, C., GONZALEZ-MENENDEZ, I., TURTON, M., HALFORD, S., SEKARAN, S., PEIRSON, S. N., HANKINS, M. W. & FOSTER, R. G. 2012b. Differential Expression of Melanopsin Isoforms Opn4L and Opn4S during Postnatal Development of the Mouse Retina. Plos One, 7. HUNTER, J. D. 2007. Matplotlib: A 2D graphics environment. Computing in Science & Engineering, 9, 90-95. IVANOVA , E., YEE, C. W., BALDONI, R. & SAGDULLAEV, B. T. 2015. Aberrant activity in retinal degeneration impairs central visual processing and relies on Cx36-containing gap junctions. Experimental Eye Research, 10. JACOBS, G. H., WILLIAMS, G. A., CAHILL, H. & NATHANS, J. 2007. Emergence of novel color vision in mice engineered to express a human cone photopigment. Science, 315, 1723-1725.

207

Chapter 7 General References

JANSSEN, B. J. A., DE CELLE, T., DEBETS, J. J. M., BROUNS, A. E., CALLAHAN, M. F. & SMITH, T. L. 2004. Effects of anesthetics on systemic hemodynamics in mice. American Journal of Physiology-Heart and Circulatory Physiology, 287, H1618-H1624. JEFFREY, B. G., MCGILL, T. J., HALEY, T. L., MORGANS, C. W. & DUVOISIN, R. M. 2011. Anatomical, Physiological, and Behavioral Analysis of Rodent Vision. Animal Models of Behavioral Analysis, 50. JEON, C. J., STRETTOI, E. & MASLAND, R. H. 1998. The major cell populations of the mouse retina. Journal of Neuroscience, 18. JIMENEZ, A. J., GARCIAFERNANDEZ, J. M., GONZALEZ, B. & FOSTER, R. G. 1996. The spatio- temporal pattern of photoreceptor degeneration in the aged rd/rd mouse retina. Cell and Tissue Research, 284, 193-202. JOHNSON, J., WU, V., DONOVAN, M., MAJUMDAR, S., RENTERIA, R. C., PORCO, T., VAN GELDER, R. N. & COPENHAGEN, D. R. 2010. Melanopsin-dependent light avoidance in neonatal mice. Proceedings of the National Academy of Sciences of the United States of America, 107. JOHNSON, R. F., MOORE, R. Y. & MORIN, L. P. 1988. LOSS OF ENTRAINMENT AND ANATOMICAL PLASTICITY AFTER LESIONS OF THE HAMSTER RETINOHYPOTHALAMIC TRACT. Brain Research, 460, 297-313. JOLLIFFE, I. T. 2002. Principal Component Analysis, Springer-Verlag New York. JONES, B. W. & MARC, R. E. 2005. Retinal remodeling during retinal degeneration. Experimental Eye Research, 81, 123-137. JONES, B. W., WATT, C. B., FREDERICK, J. M., BAEHR, W., CHEN, C. K., LEVINE, E. M., MILAM, A. H., LAVAIL, M. M. & MARC, R. E. 2003. Retinal remodeling triggered by photoreceptor degenerations. Journal of Comparative Neurology, 464, 1-16. JOO, H. R., PETERSON, B. B., DACEY, D. M., HATTAR, S. & CHEN, S.-K. 2013. Recurrent axon collaterals of intrinsically photosensitive retinal ganglion cells. Visual Neuroscience, 30, 175-182. JUSUF, P. R., LEE, S. C. S., HANNIBAL, J. & GRUENERT, U. 2007. Characterization and synaptic connectivity of melanopsin-containing ganglion cells in the primate retina. European Journal of Neuroscience, 26. KANEKO, A. 1970. PHYSIOLOGICAL AND MORPHOLOGICAL IDENTIFICATION OF HORIZONTAL, BIPOLAR AND AMACRINE CELLS IN GOLDFISH RETINA. Journal of Physiology-London, 207, 623-&. KAYLOR, J. J., YUAN, Q., COOK, J., SARFARE, S., MAKSHANOFF, J., MIU, A., KIM, A., KIM, P., HABIB, S., ROYBAL, C. N., XU, T. Z., NUSINOWITZ, S. & TRAVIS, G. H. 2013. Identification of DES1 as a vitamin A isomerase in Muller glial cells of the retina. Nature Chemical Biology, 9, 30-+. KISER, P. D., GOLCZAK, M., MAEDA, A. & PALCZEWSKI, K. 2012. Key enzymes of the retinoid (visual) cycle in vertebrate retina. Biochimica Et Biophysica Acta-Molecular and Cell Biology of Lipids, 1821, 137-151. KLERMAN, E. B., SHANAHAN, T. L., BROTMAN, D. J., RIMMER, D. W., EMENS, J. S., RIZZO, J. F. & CZEISLER, C. A. 2002. Photic resetting of the human circadian pacemaker in the absence of conscious vision. Journal of Biological Rhythms, 17, 548-555. KOLB, H., NELSON, R. & MARIANI, A. 1981. AMACRINE CELLS, BIPOLAR CELLS AND GANGLION-CELLS OF THE CAT RETINA - A GOLGI-STUDY. Vision Research, 21, 1081- 1114. KOS, C. H. 2004. Cre/loxP system for generating tissue-specific knockout mouse models. Nutrition Reviews, 62, 243-246. KRAHE, T. E., EL-DANAF, R. N., DILGER, E. K., HENDERSON, S. C. & GUIDO, W. 2011. Morphologically Distinct Classes of Relay Cells Exhibit Regional Preferences in the Dorsal Lateral Geniculate Nucleus of the Mouse. Journal of Neuroscience, 31.

208

Chapter 7 General References

KUFFLER, S. W. 1953. DISCHARGE PATTERNS AND FUNCTIONAL ORGANIZATION OF MAMMALIAN RETINA. Journal of Neurophysiology, 16, 37-68. KUHN, H., BENNETT, N., MICHELVILLAZ, M. & CHABRE, M. 1981. INTERACTIONS BETWEEN PHOTO-EXCITED RHODOPSIN AND GTP-BINDING PROTEIN - KINETIC AND STOICHIOMETRIC ANALYSES FROM LIGHT-SCATTERING CHANGES. Proceedings of the National Academy of Sciences of the United States of America-Biological Sciences, 78, 6873-6877. KUHN, H. & WILDEN, U. 1987. DEACTIVATION OF PHOTOACTIVATED RHODOPSIN BY RHODOPSIN-KINASE AND ARRESTIN. Journal of Receptor Research, 7, 283-298. LAGALI, P. S., BALYA, D., AWATRAMANI, G. B., MUNCH, T. A., KIM, D. S., BUSSKAMP, V., CEPKO, C. L. & ROSKA, B. 2008. Light-activated channels targeted to ON bipolar cells restore visual function in retinal degeneration. Nature Neuroscience, 11, 667- 675. LAKADAMYALI, M., BABCOCK, H., BATES, M., ZHUANG, X. W. & LICHTMAN, J. 2012. 3D Multicolor Super-Resolution Imaging Offers Improved Accuracy in Neuron Tracing. Plos One, 7. LALL, G. S., REVELL, V. L., MOMIJI, H., AL ENEZI, J., ALTIMUS, C. M., GUELER, A. D., AGUILAR, C., CAMERON, M. A., ALLENDER, S., HANKINS, M. W. & LUCAS, R. J. 2010. Distinct Contributions of Rod, Cone, and Melanopsin Photoreceptors to Encoding Irradiance. Neuron, 66. LAVAIL, M. M., MATTHES, M. T., YASUMURA, D. & STEINBERG, R. H. 1997. Variability in rate of cone degeneration in the retinal degeneration (rd/rd) mouse. Experimental Eye Research, 65, 45-50. LIN, B., KOIZUMI, A., TANAKA, N., PANDA, S. & MASLAND, R. H. 2008. Restoration of visual function in retinal degeneration mice by ectopic expression of melanopsin. Proceedings of the National Academy of Sciences of the United States of America, 105. LIN, B., MARTIN, P. R. & GRUNERT, U. 2002. Expression and distribution of ionotropic glutamate receptor subunits on parasol ganglion cells in the primate retina. Visual Neuroscience, 19, 453-465. LIN, B., MASLAND, R. H. & STRETTOI, E. 2009. Remodeling of cone photoreceptor cells after rod degeneration in rd mice. Experimental Eye Research, 88, 589-599. LIN, B. & PENG, E. B. 2013. Retinal Ganglion Cells are Resistant to Photoreceptor Loss in Retinal Degeneration. Plos One, 8. LINDSTROM, S. H., RYAN, D. G., SHI, J. & DEVRIES, S. H. 2014. Kainate receptor subunit diversity underlying response diversity in retinal Off bipolar cells. Journal of Physiology-London, 592, 1457-1477. LIVET, J., WEISSMAN, T. A., KANG, H. N., DRAFT, R. W., LU, J., BENNIS, R. A., SANES, J. R. & LICHTMAN, J. W. 2007. Transgenic strategies for combinatorial expression of fluorescent proteins in the nervous system. Nature, 450, 56-+. LOLLEY, R. N. 1994. THE RD GENE DEFECT TRIGGERS PROGRAMMED ROD CELL-DEATH - THE PROCTOR-LECTURE. Investigative Ophthalmology & Visual Science, 35, 4182-4191. LOULIER, K., BARRY, R., MAHOU, P., LE FRANC, Y., SUPATTO, W., MATHO, K. S., IENG, S. H., FOUQUET, S., DUPIN, E., BENOSMAN, R., CHEDOTAL, A., BEAUREPAIRE, E., MORIN, X. & LIVET, J. 2014. Multiplex Cell and Lineage Tracking with Combinatorial Labels. Neuron, 81, 505-520. LUCAS, R. J. 2006. Chromophore regeneration: Melanopsin does its own thing. Proceedings of the National Academy of Sciences of the United States of America, 103. LUCAS, R. J., DOUGLAS, R. H. & FOSTER, R. G. 2001a. Characterization of an ocular photopigment capable of driving pupillary constriction in mice. Nature Neuroscience, 4.

209

Chapter 7 General References

LUCAS, R. J., FREEDMAN, M. S., LUPI, D., MUNOZ, M., DAVID-GRAY, Z. K. & FOSTER, R. G. 2001b. Identifying the photoreceptive inputs to the mammalian circadian system using transgenic and retinally degenerate mice. Behavioural Brain Research, 125. LUCAS, R. J., FREEDMAN, M. S., MUNOZ, M., GARCIA-FERNANDEZ, J. M. & FOSTER, R. G. 1999. Regulation of the mammalian pineal by non-rod, non-cone, ocular photoreceptors. Science, 284. LUCAS, R. J., HATTAR, S., TAKAO, M., BERSON, D. M., FOSTER, R. G. & YAU, K. W. 2003a. Diminished pupillary light reflex at high irradiances in melanopsin-knockout mice. Science, 299, 245-247. LUCAS, R. J., HATTAR, S., TAKAO, M., BERSON, D. M., FOSTER, R. G. & YAU, K. W. 2003b. Diminished pupillary light reflex at high irradiances in melanopsin-knockout mice. Science, 299. LUCAS, R. J., LALL, G. S., ALLEN, A. E. & BROWN, T. M. 2012. How rod, cone, and melanopsin photoreceptors come together to enlighten the mammalian circadian clock. Neurobiology of Circadian Timing, 199, 1-18. LUKASIEWICZ, P. D. 2005. Synaptic mechanisms that shape visual signaling at the inner retina. Development, Dynamics and Pathology of Neuronal Networks: from Molecules to Functional Circuits, 147, 205-218. MACE, E., CAPLETTE, R., MARRE, O., SENGUPTA, A., CHAFFIOL, A., BARBE, P., DESROSIERS, M., BAMBERG, E., SAHEL, J. A., PICAUD, S., DUEBEL, J. & DALKARA, D. 2015. Targeting Channelrhodopsin-2 to ON-bipolar Cells With Vitreally Administered AAV Restores ON and OFF Visual Responses in Blind Mice. Molecular Therapy, 23, 7-16. MACNEIL, M. A. & MASLAND, R. H. 1998. Extreme diversity among amacrine cells: Implications for function. Neuron, 20. MAGGI, C. A. & MELI, A. 1986. SUITABILITY OF URETHANE ANESTHESIA FOR PHYSIOPHARMACOLOGICAL INVESTIGATIONS IN VARIOUS SYSTEMS .2. CARDIOVASCULAR-SYSTEM. Experientia, 42, 292-297. MAIMON, G. 2011. Modulation of visual physiology by behavioral state in monkeys, mice, and flies. Current Opinion in Neurobiology, 21, 559-564. MANGEL, S. C. 1991. ANALYSIS OF THE HORIZONTAL CELL CONTRIBUTION TO THE RECEPTIVE-FIELD SURROUND OF GANGLION-CELLS IN THE RABBIT RETINA. Journal of Physiology-London, 442, 211-234. MARC, R. E., JONES, B. W., ANDERSON, J. R., KINARD, K., MARSHAK, D. W., WILSON, J. H., WENSEL, T. & LUCAS, R. J. 2007. Neural reprogramming in retinal degeneration. Investigative Ophthalmology & Visual Science, 48, 3364-3371. MARC, R. E., JONES, B. W., WATT, C. B. & STRETTOI, E. 2003. Neural remodeling in retinal degeneration. Progress in Retinal and Eye Research, 22, 607-655. MARGOLIS, D. J., NEWKIRK, G., EULER, T. & DETWILER, P. B. 2008. Functional stability of retinal ganglion cells after degeneration-induced changes in synaptic input. Journal of Neuroscience, 28, 6526-6536. MARSHEL, J. H., GARRETT, M. E., NAUHAUS, I. & CALLAWAY, E. M. 2011. Functional Specialization of Seven Mouse Visual Cortical Areas. Neuron, 72, 1040-1054. MASLAND, R. H. 2001. The fundamental plan of the retina. Nature Neuroscience, 4. MATA, N. L., RADU, R. A., CLEMMONS, R. S. & TRAVIS, G. H. 2002. Isomerization and oxidation of vitamin a in cone-dominant retinas: A novel pathway for visual- pigment regeneration in daylight. Neuron, 36, 69-80. MAY, P. J. 2006. The mammalian superior colliculus: laminar structure and connections. Neuroanatomy of the Oculomotor System, 151, 321-378. MCBEE, J. K., PALCZEWSKI, K., BAEHR, W. & PEPPERBERG, D. R. 2001. Confronting complexity: the interlink of phototransduction and retinoid metabolism in the vertebrate retina. Progress in Retinal and Eye Research, 20, 469-529.

210

Chapter 7 General References

MCDOUGAL, D. H. & GAMLIN, P. D. 2010. The influence of intrinsically-photosensitive retinal ganglion cells on the spectral sensitivity and response dynamics of the human pupillary light reflex. Vision Research, 50, 72-87. MCMAHON, M. J., PACKER, O. S. & DACEY, D. M. 2004. The classical receptive field surround of primate parasol ganglion cells is mediated primarily by a non- GABAergic pathway. Journal of Neuroscience, 24, 3736-3745. MCNEILL, D. S., SHEELY, C. J., ECKER, J. L., BADEA, T. C., MORHARDT, D., GUIDO, W. & HATTAR, S. 2011. Development of melanopsin-based irradiance detecting circuitry. Neural Development, 6. MELYAN, Z., TARTTELIN, E. E., BELLINGHAM, J., LUCAS, R. J. & HANKINS, M. W. 2005. Addition of human melanopsin renders mammalian cells photoresponsive. Nature, 433, 741-745. MENGER, N., POW, D. V. & WASSLE, H. 1998. Glycinergic amacrine cells of the rat retina. Journal of Comparative Neurology, 401, 34-46. MENZLER, J. & ZECK, G. 2011. Network Oscillations in Rod-Degenerated Mouse Retinas. Journal of Neuroscience, 31, 2280-2291. MERBS, S. L. & NATHANS, J. 1992. ABSORPTION-SPECTRA OF HUMAN CONE PIGMENTS. Nature, 356, 433-435. MILOSAVLJEVIC, N., CEJAHIC-KAPETANOVIC, PROCYK, C. & LUCAS, R. 2016. Chemogenetic Activation of Melanopsin Retinal Ganglion Cells Induces Signatures of Arousal and / or Anxiety in Mice. Current Biology, 26, 2358–2363,. MINNEMAN, K. P. 2001. Splice variants of G protein-coupled receptors. Molecular interventions, 1. MOHAND-SAID, S., DEUDON-COMBE, A., HICKS, D., SIMONUTTI, M., FORSTER, V., FINTZ, A. C., LEVEILLARD, T., DREYFUS, H. & SAHEL, J. A. 1998. Normal retina releases a diffusible factor stimulating cone survival in the retinal degeneration mouse. Proceedings of the National Academy of Sciences of the United States of America, 95, 8357-8362. MUENCH, T. A., DA SILVEIRA, R. A., SIEGERT, S., VINEY, T. J., AWATRAMANI, G. B. & ROSKA, B. 2009. Approach sensitivity in the retina processed by a multifunctional neural circuit. Nature Neuroscience, 12, 1308-U135. MULLER, L. P. D. S., DO, M. T. H., YAU, K.-W., HE, S. & BALDRIDGE, W. H. 2010. Tracer Coupling of Intrinsically Photosensitive Retinal Ganglion Cells to Amacrine Cells in the Mouse Retina. Journal of Comparative Neurology, 518. MURE, L. S., CORNUT, P.-L., RIEUX, C., DROUYER, E., DENIS, P., GRONFIER, C. & COOPER, H. M. 2009. Melanopsin Bistability: A Fly's Eye Technology in the Human Retina. Plos One, 4. NAIR, G., KIM, M., NAGAOKA, T., OLSON, D. E., THULE, P. M., PARDUE, M. T. & DUONG, T. Q. 2011. Effects of common anesthetics on eye movement and electroretinogram. Documenta Ophthalmologica, 122, 163-176. NELSON, R., FAMIGLIETTI, E. V. & KOLB, H. 1978. INTRACELLULAR STAINING REVEALS DIFFERENT LEVELS OF STRATIFICATION FOR ON-CENTER AND OFF-CENTER GANGLION-CELLS IN CAT RETINA. Journal of Neurophysiology, 41. NERN, A., PFEIFFER, B. D. & RUBIN, G. M. 2015. Optimized tools for multicolor stochastic labeling reveal diverse stereotyped cell arrangements in the fly visual system. Proceedings of the National Academy of Sciences of the United States of America, 112, E2967-E2976. NIELL, C. M. & STRYKER, M. P. 2010. Modulation of Visual Responses by Behavioral State in Mouse Visual Cortex. Neuron, 65, 472-479.

211

Chapter 7 General References

NIKONOV, S. S., KHOLODENKO, R., LEM, J. & PUGH, E. N. 2006. Physiological features of the S- and M-cone photoreceptors of wild-type mice from single-cell recordings. Journal of General Physiology, 127, 359-374. OGILVIE, J. M., TENKOVA, T., LETT, J. M., SPECK, J., LANDGRAF, M. & SILVERMAN, M. S. 1997. Age-related distribution of cones and ON-bipolar cells in the rd mouse retina. Current Eye Research, 16, 244-251. OKADA, T. & PALCZEWSKI, K. 2001. Crystal structure of rhodopsin: implications for vision and beyond. Current Opinion in Structural Biology, 11, 420-426. OKANO, T., KOJIMA, D., FUKADA, Y., SHICHIDA, Y. & YOSHIZAWA, T. 1992. PRIMARY STRUCTURES OF CHICKEN CONE VISUAL PIGMENTS - VERTEBRATE RHODOPSINS HAVE EVOLVED OUT OF CONE VISUAL PIGMENTS. Proceedings of the National Academy of Sciences of the United States of America, 89, 5932-5936. OKAWA, H. & SAMPATH, A. P. 2007. Optimization of single-photon response transmission at the rod-to-rod bipolar synapse. Physiology, 22, 279-286. OKUN, M., LAK, A., CARANDINI, M. & HARRIS, K. D. 2016. Long Term Recordings with Immobile Silicon Probes in the Mouse Cortex. Plos One, 11. ORTIN-MARTINEZ, A., NADAL-NICOLAS, F. M., JIMENEZ-LOPEZ, M., ALBURQUERQUE-BEJAR, J. J., NIETO-LOPEZ, L., GARCIA-AYUSO, D., VILLEGAS-PEREZ, M. P., VIDAL-SANZ, M. & AGUDO-BARRIUSO, M. 2014. Number and Distribution of Mouse Retinal Cone Photoreceptors: Differences between an Albino (Swiss) and a Pigmented (C57/BL6) Strain. Plos One, 9. OSTERGAARD, J., HANNIBAL, J. & FAHRENKRUG, J. 2007. Synaptic contact between melanopsin-containing retinal ganglion cells and rod bipolar cells. Investigative Ophthalmology & Visual Science, 48. PAGES, F., DETERRE, P. & PFISTER, C. 1992. ENHANCED GTPASE ACTIVITY OF TRANSDUCIN WHEN BOUND TO CGMP PHOSPHODIESTERASE IN BOVINE RETINAL RODS. Journal of Biological Chemistry, 267, 22018-22021. PALCZEWSKI, K. 2006. G protein-coupled receptor rhodopsin. Annual Review of Biochemistry, 75, 743-767. PALCZEWSKI, K., KUMASAKA, T., HORI, T., BEHNKE, C. A., MOTOSHIMA, H., FOX, B. A., LE TRONG, I., TELLER, D. C., OKADA, T., STENKAMP, R. E., YAMAMOTO, M. & MIYANO, M. 2000. Crystal structure of rhodopsin: A G protein-coupled receptor. Science, 289, 739-745. PALCZEWSKI, K., VAN HOOSER, J. P., GARWIN, G. G., CHEN, J., LIOU, G. I. & SAARI, J. C. 1999. Kinetics of visual pigment regeneration in excised mouse eyes and in mice with a targeted disruption of the gene encoding interphotoreceptor retinoid- binding protein or arrestin. Biochemistry, 38, 12012-12019. PAN, Y. A., FREUNDLICH, T., WEISSMAN, T. A., SCHOPPIK, D., WANG, X. C., ZIMMERMAN, S., CIRUNA, B., SANES, J. R., LICHTMAN, J. W. & SCHIER, A. F. 2013. Zebrabow: multispectral cell labeling for cell tracing and lineage analysis in zebrafish. Development, 140, 2835-2846. PANDA, S., NAYAK, S. K., CAMPO, B., WALKER, J. R., HOGENESCH, J. B. & JEGLA, T. 2005. Illumination of the melanopsin signaling pathway. Science, 307. PANDA, S., PROVENCIO, I., TU, D. C., PIRES, S. S., ROLLAG, M. D., CASTRUCCI, A. M., PLETCHER, M. T., SATO, T. K., WILTSHIRE, T., ANDAHAZY, M., KAY, S. A., VAN GELDER, R. N. & HOGENESCH, J. B. 2003. Melanopsin is required for non-image- forming photic responses in blind mice. Science, 301. PANDA, S., SATO, T. K., CASTRUCCI, A. M., ROLLAG, M. D., DEGRIP, W. J., HOGENESCH, J. B., PROVENCIO, I. & KAY, S. A. 2002. Melanopsin (Opn4) requirement for normal light- induced circadian phase shifting. Science, 298.

212

Chapter 7 General References

PAXINOS, G. & FRANKLIN, K. 2004. The Mouse Brain in Stereotaxic Coordinates (2nd Edition), Gulf Professional Publishing. PEICHL, L. & GONZALEZSORIANO, J. 1994. MORPHOLOGICAL TYPES OF HORIZONTAL CELL IN RODENT RETINAE - A COMPARISON OF RAT, MOUSE, GERBIL, AND GUINEA-PIG. Visual Neuroscience, 11, 501-517. PEIRCE, J. W. 2008. Generating Stimuli for Neuroscience Using PsychoPy. Frontiers in neuroinformatics, 2, 10-10. PENG, Y. W., HAO, Y., PETTERS, R. M. & WONG, F. 2000. Ectopic synaptogenesis in the mammalian retina caused by rod photoreceptor-specific mutations. Nature Neuroscience, 3, 1121-1127. PENNESI, M. E., MICHAELS, K. V., MAGEE, S. S., MARICLE, A., DAVIN, S. P., GARG, A. K., GALE, M. J., TU, D. C., WEN, Y. Q., ERKER, L. R. & FRANCIS, P. J. 2012. Long-Term Characterization of Retinal Degeneration in rd1 and rd10 Mice Using Spectral Domain Optical Coherence Tomography. Investigative Ophthalmology & Visual Science, 53, 4644-4656. PEREZ-LEON, J. A., WARREN, E. J., ALLEN, C. N., ROBINSON, D. W. & BROWN, R. L. 2006. Synaptic inputs to retinal ganglion cells that set the circadian clock. European Journal of Neuroscience, 24. PERRY, V. H. & COWEY, A. 1985. THE GANGLION-CELL AND CONE DISTRIBUTIONS IN THE MONKEYS RETINA - IMPLICATIONS FOR CENTRAL MAGNIFICATION FACTORS. Vision Research, 25, 1795-&. PIRES, S. S., HUGHES, S., TURTON, M., MELYAN, Z., PEIRSON, S. N., ZHENG, L., KOSMAOGLOU, M., BELLINGHAM, J., CHEETHAM, M. E., LUCAS, R. J., FOSTER, R. G., HANKINS, M. W. & HALFORD, S. 2009. Differential Expression of Two Distinct Functional Isoforms of Melanopsin (Opn4) in the Mammalian Retina. Journal of Neuroscience, 29. PISCOPO, D. M., EL-DANAF, R. N., HUBERMAN, A. D. & NIELL, C. M. 2013. Diverse Visual Features Encoded in Mouse Lateral Geniculate Nucleus. Journal of Neuroscience, 33, 4642-4656. PITTLER, S. J. & BAEHR, W. 1991. IDENTIFICATION OF A NONSENSE MUTATION IN THE ROD PHOTORECEPTOR CGMP PHOSPHODIESTERASE BETA-SUBUNIT GENE OF THE RD MOUSE. Proceedings of the National Academy of Sciences of the United States of America, 88, 8322-8326. POURCHO, R. G. 1996. Neurotransmitters in the retina. Current Eye Research, 15. PROCYK, C. A., ELEFTHERIOU, C. G., STORCHI, R., ALLEN, A. E., MILOSAVLJEVIC, N., BROWN, T. M. & LUCAS, R. J. 2015. Spatial receptive fields in the retina and dorsal lateral geniculate nucleus of mice lacking rods and cones. Journal of Neurophysiology, 114, 1321-1330. PROVENCIO, I., JIANG, G. S., DE GRIP, W. J., HAYES, W. P. & ROLLAG, M. D. 1998. Melanopsin: An opsin in melanophores, brain, and eye. Proceedings of the National Academy of Sciences of the United States of America, 95. PROVENCIO, I., RODRIGUEZ, I. R., JIANG, G. S., HAYES, W. P., MOREIRA, E. F. & ROLLAG, M. D. 2000. A novel human opsin in the inner retina. Journal of Neuroscience, 20, 600- 605. PROVENCIO, I., ROLLAG, M. D. & CASTRUCCI, A. M. 2002. Anatomy: Photoreceptive net in the mammalian retina - This mesh of cells may explain how some blind mice can still tell day from night. Nature, 415, 493-493. PUGH, E. N. & LAMB, T. D. 1993. AMPLIFICATION AND KINETICS OF THE ACTIVATION STEPS IN PHOTOTRANSDUCTION. Biochimica Et Biophysica Acta, 1141, 111-149.

213

Chapter 7 General References

PURVES, D., AUGUSTINE, G., FITSPATRICK, D., KATZ, L., LAMANTIA, A., O MCNAMARA, J. & WILLIAMS, S. 2001. Neuroscience, 2nd Edition, Sunderland (MA), Sinauer Associates. QIU, X. D., KUMBALASIRI, T., CARLSON, S. M., WONG, K. Y., KRISHNA, V., PROVENCIO, I. & BERSON, D. M. 2005. Induction of photosensitivity by heterologous expression of melanopsin. Nature, 433, 745-749. QUATTROCHI, L. E., ESTEVEZ, M., KIM, I. & BERSON, D. Structure, function and outputs of a novel type of ganglion-cell photoreceptor. Society for Neuroscience, 2013 San Diego. Program number: 737.16/JJ8. RAVEN, M. A., STAGG, S. B. & REESE, B. E. 2005. Regularity and packing of the horizontal cell mosaic in different strains of mice. Visual Neuroscience, 22, 461-468. REESE, B. E. 1988. 'Hidden lamination' in the dorsal lateral geniculate nucleus: the functional organization of this thalamic region in the rat. Brain research, 472. REESE, B. E., RAVEN, M. A. & STAGG, S. B. 2005. Afferents and homotypic neighbors regulate horizontal cell morphology, connectivity, and retinal coverage. Journal of Neuroscience, 25, 2167-2175. RENNA, J. M., WENG, S. & BERSON, D. M. 2011. Light acts through melanopsin to alter retinal waves and segregation of retinogeniculate afferents. Nature Neuroscience, 14. REPPERT, S. M. & WEAVER, D. R. 2002. Coordination of circadian timing in mammals. Nature, 418. REYNAUD, E. G., PEYCHL, J., HUISKEN, J. & TOMANCAK, P. 2015. Guide to light-sheet microscopy for adventurous biologists. Nature Methods, 12, 30-34. RICH, K. A., ZHAN, Y. T. & BLANKS, J. C. 1997. Migration and synaptogenesis of cone photoreceptors in the developing mouse retina. Journal of Comparative Neurology, 388. ROHLICH, P., VANVEEN, T. & SZEL, A. 1994. 2 DIFFERENT VISUAL PIGMENTS IN ONE RETINAL CONE CELL. Neuron, 13, 1159-1166. RUBY, N. F., BRENNAN, T. J., XIE, X. M., CAO, V., FRANKEN, P., HELLER, H. C. & O'HARA, B. F. 2002. Role of melanopsin in circadian responses to light. Science, 298. RUGGIERO, L., ALLEN, C. N., BROWN, R. L. & ROBINSON, D. W. 2009. The development of melanopsin-containing retinal ganglion cells in mice with early retinal degeneration. European Journal of Neuroscience, 29, 359-367. SAARI, J. C. 2000. Biochemistry of visual pigment regeneration - The Friedenwald Lecture. Investigative Ophthalmology & Visual Science, 41, 337-348. SAKAMOTO, K., LIU, C. M., KASAMATSU, M., POZDEYEV, N. V., IUVONE, P. M. & TOSINI, G. 2005. Dopamine regulates melanopsin mRNA expression in intrinsically photosensitive retinal ganglion cells. European Journal of Neuroscience, 22. SAKAMOTO, K., LIU, C. M. & TOSINI, G. 2004. Classical photoreceptors regulate melanopsin mRNA levels in the rat retina. Journal of Neuroscience, 24. SAUER, B. 1998. Inducible gene targeting in mice using the Cre/lox system. Methods-a Companion to Methods in Enzymology, 14, 381-392. SAUER, B. & HENDERSON, N. 1988. SITE-SPECIFIC DNA RECOMBINATION IN MAMMALIAN- CELLS BY THE CRE RECOMBINASE OF BACTERIOPHAGE-P1. Proceedings of the National Academy of Sciences of the United States of America, 85, 5166-5170. SCHINDELIN, J., ARGANDA-CARRERAS, I., FRISE, E., KAYNIG, V., LONGAIR, M., PIETZSCH, T., PREIBISCH, S., RUEDEN, C., SAALFELD, S., SCHMID, B., TINEVEZ, J. Y., WHITE, D. J., HARTENSTEIN, V., ELICEIRI, K., TOMANCAK, P. & CARDONA, A. 2012. Fiji: an open- source platform for biological-image analysis. Nature Methods, 9, 676-682.

214

Chapter 7 General References

SCHMIDT, T. M., ALAM, N. M., CHEN, S., KOFUJI, P., LI, W., PRUSKY, G. T. & HATTAR, S. 2014. A Role for Melanopsin in Alpha Retinal Ganglion Cells and Contrast Detection. Neuron, 82, 781-788. SCHMIDT, T. M., CHEN, S.-K. & HATTAR, S. 2011a. Intrinsically photosensitive retinal ganglion cells: many subtypes, diverse functions. Trends in Neurosciences, 34. SCHMIDT, T. M., DO, M. T. H., DACEY, D., LUCAS, R., HATTAR, S. & MATYNIA, A. 2011b. Melanopsin-Positive Intrinsically Photosensitive Retinal Ganglion Cells: From Form to Function. Journal of Neuroscience, 31, 16094-16101. SCHMIDT, T. M. & KOFUJI, P. 2009. Functional and Morphological Differences among Intrinsically Photosensitive Retinal Ganglion Cells. Journal of Neuroscience, 29. SCHMIDT, T. M. & KOFUJI, P. 2010. Differential Cone Pathway Influence on Intrinsically Photosensitive Retinal Ganglion Cell Subtypes. Journal of Neuroscience, 30. SCHMIDT, T. M. & KOFUJI, P. 2011. Structure and Function of Bistratified Intrinsically Photosensitive Retinal Ganglion Cells in the Mouse. Journal of Comparative Neurology, 519, 1492-1504. SCHMIDT, T. M., TANIGUCHI, K. & KOFUJI, P. 2008. Intrinsic and extrinsic light responses in melanopsin-expressing ganglion cells during mouse development. Journal of Neurophysiology, 100. SCHNETKAMP, P. P. M. 2004. The SLC24 Na+/Ca2+-K+ exchanger family: vision and beyond. Pflugers Archiv-European Journal of Physiology, 447, 683-688. SCHUETT, S., BONHOEFFER, T. & HUBENER, M. 2002a. Mapping retinotopic structure in mouse visual cortex with optical imaging. Journal of Neuroscience, 22. SCHUETT, S., BONHOEFFER, T. & HUBENER, M. 2002b. Mapping retinotopic structure in mouse visual cortex with optical imaging. Journal of Neuroscience, 22, 6549-6559. SEFTON, A. J. & DREHER, B. 1995. Visual system. The rat nervous system, Second edition. SEKARAN, S., FOSTER, R. G., LUCAS, R. J. & HANKINS, M. W. 2003. Calcium imaging reveals a network of intrinsically light-sensitive inner-retinal neurons. Current Biology, 13. SEKARAN, S., LALL, G. S., RALPHS, K. L., WOLSTENHOLME, A. J., LUCAS, R. J., FOSTER, R. G. & HANKINS, M. W. 2007. 2-aminoethoxydiphenylborane is an acute inhibitor of directly photosensitive retinal ganglion cell activity in vitro and in vivo. Journal of Neuroscience, 27. SEMO, M., LUPI, D., PEIRSON, S. N., BUTLER, J. N. & FOSTER, R. G. 2003. Light-induced c-fos in melanopsin retinal ganglion cells of young and aged rodless/coneless (rd/rd cl) mice. European Journal of Neuroscience, 18. SHARMA, S., BALL, S. L. & PEACHEY, N. S. 2005. Pharmacological studies of the mouse cone electroretinogram. Visual Neuroscience, 22, 631-636. SHARPE, L. T. & STOCKMAN, A. 1999. Rod pathways: the importance of seeing nothing. Trends in Neurosciences, 22, 497-504. SLAUGHTER, M. M. & AWATRAMANI, G. B. 2002. On bipolar cells: Following in the footsteps of phototransduction. Photoreceptors and Calcium, 514. SMALLWOOD, P. M., OLVECZKY, B. P., WILLIAMS, G. L., JACOBS, G. H., REESE, B. E., MEISTER, M. & NATHANS, J. 2003. Genetically engineered mice with an additional class of cone photoreceptors: Implications for the evolution of color vision. Proceedings of the National Academy of Sciences of the United States of America, 100, 11706-11711. SMITH, R. G., FREED, M. A. & STERLING, P. 1986. MICROCIRCUITRY OF THE DARK-ADAPTED CAT RETINA - FUNCTIONAL ARCHITECTURE OF THE ROD CONE NETWORK. Journal of Neuroscience, 6, 3505-3517. SMITH, R. G. & VARDI, N. 1995. SIMULATION OF THE AII AMACRINE CELL OF MAMMALIAN RETINA - FUNCTIONAL CONSEQUENCES OF ELECTRICAL COUPLING AND REGENERATIVE MEMBRANE-PROPERTIES. Visual Neuroscience, 12, 851-860.

215

Chapter 7 General References

SMITH, S. L. & HAEUSSER, M. 2010. Parallel processing of visual space by neighboring neurons in mouse visual cortex. Nature Neuroscience, 13, 1144-U149. SOTO, F. & KERSCHENSTEINER, D. 2015. Synaptic remodeling of neuronal circuits in early retinal degeneration. Frontiers in Cellular Neuroscience, 9. STASHEFF, S. F. 2008. Emergence of sustained spontaneous hyperactivity and temporary preservation of OFF responses in ganglion cells of the retinal degeneration (rd1) mouse. Journal of Neurophysiology, 99, 1408-1421. STASHEFF, S. F., SHANKAR, M. & ANDREWS, M. P. 2011. Developmental time course distinguishes changes in spontaneous and light-evoked retinal ganglion cell activity in rd1 and rd10 mice. Journal of Neurophysiology, 105, 3002-3009. STERLING, P., FREED, M. A. & SMITH, R. G. 1988. ARCHITECTURE OF ROD AND CONE CIRCUITS TO THE ON-BETA-GANGLION CELL. Journal of Neuroscience, 8, 623-642. STERNBERG, N. & HAMILTON, D. 1981. BACTERIOPHAGE-P1 SITE-SPECIFIC RECOMBINATION .1. RECOMBINATION BETWEEN LOXP SITES. Journal of Molecular Biology, 150, 467-486. STORCHI, R., MILOSAVLJEVIC, N., ELEFTHERIOU, C. G., MARTIAL, F. P., ORLOWSKA-FEUER, P., BEDFORD, R. A., BROWN, T. M., MONTEMURRO, M. A., PETERSEN, R. S. & LUCAS, R. J. 2015. Melanopsin-driven increases in maintained activity enhance thalamic visual response reliability across a simulated dawn. Proceedings of the National Academy of Sciences of the United States of America, 112, E5734-E5743. STRETTOI, E. 2015. A Survey of Retinal Remodeling. Frontiers in Cellular Neuroscience, 9. STRETTOI, E. & PIGNATELLI, V. 2000. Modifications of retinal neurons in a mouse model of retinitis pigmentosa. Proceedings of the National Academy of Sciences of the United States of America, 97, 11020-11025. STRETTOI, E., PIGNATELLI, V., ROSSI, C., PORCIATTI, V. & FALSINI, B. 2003. Remodeling of second-order neurons in the retina of rd/rd mutant mice. Vision Research, 43, 867- 877. STRETTOI, E., PORCIATTI, V., FALSINI, B., PIGNATELLI, V. & ROSSI, C. 2002. Morphological and functional abnormalities in the inner retina of the rd/rd mouse. Journal of Neuroscience, 22, 5492-5504. SUN, H., MACKE, J. P. & NATHANS, J. 1997. Mechanisms of spectra tuning in the mouse green cone pigment. Proceedings of the National Academy of Sciences of the United States of America, 94, 8860-8865. SZCZESNY, G., VEIHELMANN, A., MASSBERG, S., NOLTE, D. & MESSMER, K. 2004. Long-term anaesthesia using inhalatory isoflurane in different strains of mice-the haemodynamic effects. Laboratory Animals, 38, 64-69. SZEL, A., ROHLICH, P., CAFFE, A. R., JULIUSSON, B., AGUIRRE, G. & VANVEEN, T. 1992. UNIQUE TOPOGRAPHIC SEPARATION OF 2 SPECTRAL CLASSES OF CONES IN THE MOUSE RETINA. Journal of Comparative Neurology, 325, 327-342. TANG, P. H., KONO, M., KOUTALOS, Y., ABLONCZY, Z. & CROUCH, R. K. 2013. New insights into retinoid metabolism and cycling within the retina. Progress in Retinal and Eye Research, 32, 48-63. THORESON, W. B. & MANGEL, S. C. 2012. Lateral interactions in the outer retina. Progress in Retinal and Eye Research, 31, 407-441. THORESON, W. B. & WITKOVSKY, P. 1999. Glutamate receptors and circuits in the vertebrate retina. Progress in Retinal and Eye Research, 18. THYAGARAJAN, S., VAN WYK, M., LEHMANN, K., LOWEL, S., FENG, G. P. & WASSLE, H. 2010. Visual Function in Mice with Photoreceptor Degeneration and Transgenic Expression of Channelrhodopsin 2 in Ganglion Cells. Journal of Neuroscience, 30, 8745-8758.

216

Chapter 7 General References

TIAN, N. & COPENHAGEN, D. R. 2003. Visual stimulation is required for refinement of ON and OFF pathways in postnatal retina. Neuron, 39, 85-96. TOYCHIEV, A. H., IVANOVA, E., YEE, C. W. & SAGDULLAEV, B. T. 2013. Block of Gap Junctions Eliminates Aberrant Activity and Restores Light Responses during Retinal Degeneration. Journal of Neuroscience, 33, 13972-13977. TRENHOLM, S., BOROWSKA, J., ZHANG, J., HOGGARTH, A., JOHNSON, K., BARNES, S., LEWIS, T. J. & AWATRAMANI, G. B. 2012. Intrinsic oscillatory activity arising within the electrically coupled AII amacrine-ON cone bipolar cell network is driven by voltage- gated Na plus channels. Journal of Physiology-London, 590, 2501-2517. TSUJIKAWA, M., WADA, Y., SUKEGAWA, M., SAWA, M., GOMI, F., NISHIDA, K. & TANO, Y. 2008. Age at onset curves of retinitis pigmentosa. Archives of Ophthalmology, 126, 337-340. TU, D. C., ZHANG, D. Y., DEMAS, J., SLUTSKY, E. B., PROVENCIO, I., HOLY, T. E. & VAN GELDER, R. N. 2005. Physiologic diversity and development of intrinsically photosensitive retinal ganglion cells. Neuron, 48. VAN DER WALT, S., COLBERT, S. C. & VAROQUAUX, G. 2011. The NumPy Array: A Structure for Efficient Numerical Computation. Computing in Science & Engineering, 13, 22- 30. VAN HOOK, M. J., WONG, K. Y. & BERSON, D. M. 2012. Dopaminergic modulation of ganglion-cell photoreceptors in rat. European Journal of Neuroscience, 35. VAN WYK, M., PIELECKA-FORTUNA, J., LOWEL, S. & KLEINLOGEL, S. 2015. Restoring the ON Switch in Blind Retinas: Opto-mGluR6, a Next-Generation, Cell-Tailored Optogenetic Tool. Plos Biology, 13. VAN WYK, M., WAESSLE, H. & TAYLOR, W. R. 2009. Receptive field properties of ON- and OFF-ganglion cells in the mouse retina. Visual Neuroscience, 26. VANEY, D. I., SIVYER, B. & TAYLOR, W. R. 2012. Direction selectivity in the retina: symmetry and asymmetry in structure and function. Nature Reviews Neuroscience, 13, 194- 208. VARDI, N., DUVOISIN, R., WU, G. & STERLING, P. 2000. Localization of mGluR6 to dendrites of ON bipolar cells in primate retina. Journal of Comparative Neurology, 423. VERWEIJ, J., KAMERMANS, M. & SPEKREIJSE, H. 1996. Horizontal cells feed back to cones by shifting the cone calcium-current activation range. Vision Research, 36, 3943-3953. VISWANATHAN, S., WILLIAMS, M. E., BLOSS, E. B., STASEVICH, T. J., SPEER, C. M., NERN, A., PFEIFFER, B. D., HOOKS, B. M., LI, W. P., ENGLISH, B. P., TIAN, T., HENRY, G. L., MACKLIN, J. J., PATEL, R., GERFEN, C. R., ZHUANG, X. W., WANG, Y. L., RUBIN, G. M. & LOOGER, L. L. 2015. High-performance probes for light and electron microscopy. Nature Methods, 12, 568-+. VOELGYI, B., CHHEDA, S. & BLOOMFIELD, S. A. 2009. Tracer Coupling Patterns of the Ganglion Cell Subtypes in the Mouse Retina. Journal of Comparative Neurology, 512, 664-687. VOLGYI, B., DEANS, M. R., PAUL, D. L. & BLOOMFIELD, S. A. 2004. Convergence and segregation of the multiple rod pathways in mammalian retina. Journal of Neuroscience, 24, 11182-11192. VOLGYI, B., XIN, D. Y., AMARILLO, Y. & BLOOMFIELD, S. A. 2001. Morphology and physiology of the polyaxonal amacrine cells in the rabbit retina. Journal of Comparative Neurology, 440, 109-125. VUGLER, A. A., REDGRAVE, P., SEMO, M. A., LAWRENCE, J., GREENWOOD, J. & COFFEY, P. J. 2007. Dopamine neurones form a discrete plexus with melanopsin cells in normal and degenerating retina. Experimental Neurology, 205. VUGLER, A. A., SEMO, M., JOSEPH, A. & JEFFERY, G. 2008. Survival and remodeling of melanopsin cells during retinal dystrophy. Visual Neuroscience, 25.

217

Chapter 7 General References

WAESSLE, H., PULLER, C., MUELLER, F. & HAVERKAMP, S. 2009. Cone Contacts, Mosaics, and Territories of Bipolar Cells in the Mouse Retina. Journal of Neuroscience, 29. WALKER, M. T., BROWN, R. L., CRONIN, T. W. & ROBINSON, P. R. 2008. Photochemistry of retinal chromophore in mouse melanopsin. Proceedings of the National Academy of Sciences of the United States of America, 105. WANG, J. S. & KEFALOV, V. J. 2009. An Alternative Pathway Mediates the Mouse and Human Cone Visual Cycle. Current Biology, 19, 1665-1669. WANG, J. S. & KEFALOV, V. J. 2011. The Cone-specific visual cycle. Progress in Retinal and Eye Research, 30, 115-128. WANG, L., SARNAIK, R., RANGARAJAN, K., LIU, X. & CANG, J. 2010. Visual Receptive Field Properties of Neurons in the Superficial Superior Colliculus of the Mouse. Journal of Neuroscience, 30. WANG, Q. X. & BURKHALTER, A. 2007. Area map of mouse visual cortex. Journal of Comparative Neurology, 502, 339-357. WANG, Y. V., WEICK, M. & DEMB, J. B. 2011. Spectral and Temporal Sensitivity of Cone- Mediated Responses in Mouse Retinal Ganglion Cells. Journal of Neuroscience, 31, 7670-7681. WARREN, E. J., ALLEN, C. N., BROWN, R. L. & ROBINSON, D. W. 2003. Intrinsic light responses of retinal ganglion cells projecting to the circadian system. European Journal of Neuroscience, 17. WARREN, E. J., ALLEN, C. N., BROWN, R. L. & ROBINSON, D. W. 2006. The light-activated signaling pathway in SCN-projecting rat retinal ganglion cells. European Journal of Neuroscience, 23. WEISSMAN, T. A. & PAN, Y. A. 2015. Brainbow: New Resources and Emerging Biological Applications for Multicolor Genetic Labeling and Analysis. Genetics, 199, 293-306. WENG, S. J., SUN, W. Z. & HE, S. G. 2005. Identification of ON-OFF direction-selective ganglion cells in the mouse retina. Journal of Physiology-London, 562, 915-923. WERBLIN, F. S. 2011. The retinal hypercircuit: a repeating synaptic interactive motif underlying visual function. Journal of Physiology-London, 589, 3691-3702. WIESEL, T. N. & HUBEL, D. H. 1963. EFFECTS OF VISUAL DEPRIVATION ON MORPHOLOGY AND PHYSIOLOGY OF CELLS IN CATS LATERAL GENICULATE BODY. Journal of Neurophysiology, 26, 978-&. WITKOVSKY, P. 2004. Dopamine and retinal function. Documenta Ophthalmologica, 108. WONG, K. Y. 2012. A Retinal Ganglion Cell That Can Signal Irradiance Continuously for 10 Hours. Journal of Neuroscience, 32. WONG, K. Y., DUNN, F. A., GRAHAM, D. M. & BERSON, D. M. 2007. Synaptic influences on rat ganglion-cell photoreceptors. Journal of Physiology-London, 582, 279-296. WONG, R. O. L. 1999. Retinal waves and visual system development. Annual Review of Neuroscience, 22. WU, Y., RICHARD, S. & PARENT, A. 2000. The organization of the striatal output system: a single-cell juxtacellular labeling study in the rat. Neuroscience Research, 38, 49-62. XU, J., DODD, R. L., MAKINO, C. L., SIMON, M. I., BAYLOR, D. A. & CHEN, J. 1997. Prolonged photoresponses in transgenic mouse rods lacking arrestin. Nature, 389, 505-509. YANG, G. & MASLAND, R. H. 1994. RECEPTIVE-FIELDS AND DENDRITIC STRUCTURE OF DIRECTIONALLY SELECTIVE RETINAL GANGLION-CELLS. Journal of Neuroscience, 14, 5267-5280. YANG, X. L. 2004. Characterization of receptors for glutamate and GABA in retinal neurons. Progress in Neurobiology, 73. YAU, K.-W. & HARDIE, R. C. 2009. Phototransduction Motifs and Variations. Cell, 139, 246- 264.

218

Chapter 7 General References

ZAGHLOUL, K. A., MANOOKIN, M. B., BORGHUIS, B. G., BOAHEN, K. & DEMB, J. B. 2007. Functional circuitry for peripheral suppression in mammalian Y-type retinal ganglion cells. Journal of Neurophysiology, 97, 4327-4340. ZAIDI, F. H., HULL, J. T., PEIRSON, S. N., WULFF, K., AESCHBACH, D., GOOLEY, J. J., BRAINARD, G. C., GREGORY-EVANS, K., RIZZO, J. F., III, CZEISLER, C. A., FOSTER, R. G., MOSELEY, M. J. & LOCKLEY, S. W. 2007. Short-wavelength light sensitivity of circadian, pupillary, and visual awareness in humans lacking an outer retina. Current Biology, 17, 2122-2128. ZHAO, X., STAFFORD, B. K., GODIN, A. L., KING, W. M. & WONG, K. Y. 2014a. Photoresponse diversity among the five types of intrinsically photosensitive retinal ganglion cells. Journal of Physiology-London, 592, 1619-1636. ZHAO, X. W., PACK, W., KHAN, N. W. & WONG, K. Y. 2016. Prolonged Inner Retinal Photoreception Depends on the Visual Retinoid Cycle. Journal of Neuroscience, 36, 4209-4217. ZHAO, X. W., STAFFORD, B. K., GODIN, A. L., KING, W. M. & WONG, K. Y. 2014b. Photoresponse diversity among the five types of intrinsically photosensitive retinal ganglion cells. Journal of Physiology-London, 592, 1619-1636. ZHENG, J. J., LEE, S. & ZHOU, Z. J. 2006. A transient network of intrinsically bursting starburst cells underlies the generation of retinal waves. Nature Neuroscience, 9. ZHOU, Z. J. & LEE, S. 2008. Synaptic physiology of direction selectivity in the retina. Journal of Physiology-London, 586.

219

Appendix

Appendix A

Copy of Chapter 4 as published in the Journal of Neurophysiology

Procyk, C.A., Eleftheriou, C.G, Storchi, R., Allen, A.E., Milosavljevic, N., Brown, T and Lucas, R,J. (2015). Spatial receptive fields in the retina and dorsal lateral geniculate nucleus of mice lacking rods and cones. Neurophysiology 114: 1321–1330, 2015

220

J Neurophysiol 114: 1321–1330, 2015. First published June 18, 2015; doi:10.1152/jn.00368.2015.

Spatial receptive ﬁelds in the retina and dorsal lateral geniculate nucleus of mice lacking rods and cones

Christopher A. Procyk, Cyril G. Eleftheriou, Riccardo Storchi, Annette E. Allen, Nina Milosavljevic, Timothy M. Brown, and Robert J. Lucas Faculty of Life Sciences, University of Manchester, Manchester, United Kingdom

Submitted 15 April 2015; accepted in ﬁnal form 12 June 2015

Procyk CA, Eleftheriou CG, Storchi R, Allen AE, Milosavljevic 2012; Wong et al. 2007), with their projection patterns encom- N, Brown TM, Lucas RJ. Spatial receptive fields in the retina and passing the primary visual thalamus (Brown et al. 2010; Ecker dorsal lateral geniculate nucleus of mice lacking rods and cones. J et al. 2010). Neurophysiol 114: 1321–1330, 2015. First published June 18, 2015; Downloaded from doi:10.1152/jn.00368.2015.—In advanced retinal degeneration loss of These findings imply that ipRGCs could support form vision rods and cones leaves melanopsin-expressing intrinsically photosen- in some patients with outer retinal degeneration. However, at sitive retinal ganglion cells (ipRGCs) as the only source of visual present there has been no direct demonstration that ipRGCs information. ipRGCs drive non-image-forming responses (e.g., circa- contribute to spatial information in the dorsal lateral geniculate dian photoentrainment) under such conditions but, despite projecting nucleus (dLGN). In fact, many more dLGN neurons show to the primary visual thalamus [dorsal lateral geniculate nucleus evidence of melanopsin influence than can be accounted for by

(dLGN)], do not support form vision. We wished to determine what a simple one-to-one input from ipRGCs (Brown et al. 2010). A http://jn.physiology.org/ precludes ipRGCs supporting spatial discrimination after photoreceptor loss, using a mouse model (rd/rd cl) lacking rods and cones. Using possible explanation for this is that central pathways sample multielectrode arrays, we found that both RGCs and neurons in the across many ipRGCs to provide individual dLGN neurons with dLGN of this animal have clearly delineated spatial receptive fields. In a signal of mean light intensity across the visual scene. Deter- the retina, they are typically symmetrical, lack inhibitory surrounds, mining the spatial extent of visual information provided by and have diameters in the range of 10–30° of visual space. Receptive ipRGCs to individual dLGN neurons would therefore help to fields in the dLGN were larger (diameters typically 30–70°) but resolve this issue as well. matched the retinotopic map of the mouse dLGN. Injections of a In assessing the potential relevance of ipRGCs to form neuroanatomical tracer (cholera toxin ␤-subunit) into the dLGN confirmed that retinotopic order of ganglion cell projections to the vision during progressive retinal degeneration, a second ques- by 10.220.33.4 on March 3, 2017 dLGN and thalamic projections to the cortex is at least superficially tion is how these photoreceptors are impacted by outer retinal intact in rd/rd cl mice. However, as previously reported for deaffer- dystrophy. ipRGCs certainly survive, with broadly normal ented ipRGCs, onset and offset of light responses have long latencies anatomy, after severe outer retinal dystrophy (Semo et al. in the rd/rd cl retina and dLGN. Accordingly, dLGN neurons failed to 2003; Vugler et al. 2008). It remains unclear, however, whether track dynamic changes in light intensity in this animal. Our data reveal their functional receptive fields are degraded by the impact that ipRGCs can convey spatial information in advanced retinal such conditions have on the remaining inner retinal circuitry degeneration and identify their poor temporal fidelity as the major (Marc et al. 2007; Strettoi and Pignatelli 2000). Similarly, limitation in their ability to provide information about spatial patterns ipRGCs drive light responses in the dLGN in advanced retinal under natural viewing conditions. degeneration, but we do not know whether this signal allows melanopsin; retinal degeneration; spatial receptive fields; dorsal lat- individual dLGN neurons to respond to light from particular eral geniculate nucleus points in visual space. Finally, it is uncertain whether the projection of ipRGCs (or indeed the ganglion cell population in general) to the dLGN retains retinotopic order under these IN MAMMALS with advanced retinal degeneration, melanopsin- conditions. expressing intrinsically photosensitive retinal ganglion cells Here we set out to address these questions by recording (ipRGCs) survive to support non-image-forming (NIF) light melanopsin-evoked electrophysiological responses to spatially responses such as circadian photoentrainment and the pupillary structured stimuli in the retina and dLGN of retinally degen- light reflex (PLR) (Berson et al. 2002; Freedman et al. 1999; erate mice. The model we chose for this work, C3H rd/rd cl Lucas et al. 2003). These NIF responses rely on measuring mice, carries the rd1 loss-of-function mutation in the gene background light levels rather than detecting spatial patterns. encoding the phosphodiesterase ␤6-subunit (PdeB6) that abol- However, melanopsin is not restricted to those RGCs (termed ishes rod phototransduction and causes rod, and subsequent M1 ipRGCs) innervating NIF brain centers but is expressed cone, cell death. Surviving cones are additionally targeted by also in several classes whose anatomy and physiology imply a the cl transgene that drives expression of the diphtheria toxin role in form vision (Brown et al. 2010; Ecker et al. 2010; ␣-subunit in this cell type. In these animals, rods and cones Schmidt et al. 2014). Spatial receptive fields have been de- degenerate rapidly postnatally such that, by 80 days, ipRGCs scribed for ipRGCs in the rodent retina and are shown, in some are the sole surviving photoreceptors (Freedman et al. 1999; cases, to include center-surround antagonism (Estevez et al. Lucas et al. 1999, 2001). Here, we found that ipRGCs retain Address for reprint requests and other correspondence: R. J. Lucas, Faculty spatial receptive fields in the retina and retinotopic order to of Life Sciences, Univ. of Manchester, Oxford Road, Manchester, M13 9PT, their projection to the dLGN in this animal. Within the dLGN, UK (e-mail: [email protected]). we found neurons with functional receptive fields over a range www.jn.org 0022-3077/15 Copyright © 2015 the American Physiological Society 1321 1322 SPATIAL VISION IN ADVANCED RETINAL DEGENERATION of spatial scales. These data indicate that the remaining ipRGC 40 kHz. Multiunit activity (spikes with amplitudes Ͼ 50 ␮V) was photoreceptors have the fundamental ability to provide spatial saved as time-stamped waveforms and analyzed off-line (see Data information in advanced retinal degeneration. However, our analysis). data further demonstrate that the poor temporal fidelity of the Six rd/rd cl mice were used for in vivo receptive field mapping. ipRGC light response after rod and cone loss represents a Light stimuli were generated in MATLAB (The MathWorks) and substantial limitation to employing this capacity to support controlled by a laptop running PsychoPy V2.6 (Peirce 2008). Stimuli were presented via a 46-in. back-lit LED TV monitor (DS46LO4; form vision. Dynascan Technology). The monitor was at a fixed distance of 30 cm away from the eye contralateral to the recording probe and angled 20° MATERIALS AND METHODS toward the frame. Light-responsive cells were confirmed by the pre- In vitro electrophysiology. All animal experiments were covered by sentation of a 60-s uniform white stimulus (melanopsin a license awarded by the UK Home Office under the terms of the UK irradiance ϭ 8.72 ϫ 1014 photons·cmϪ2·sϪ1) repeated 10 times, with Animals (Scientific Procedures) Act (1986). Five male rd/rd cl mice a 120-s interstimulus interval. For receptive field mapping, stimuli were killed by cervical dislocation and immediately enucleated. Ret- consisted of a 60-s prestimulus dark screen before the 60-s presenta- ϭ ϫ 14 inal isolation was performed in carboxygenated (95% O2-5% CO2) tion of a vertical blue stripe (melanopsin irradiance 2.56 10 artificial cerebrospinal fluid (aCSF; concentration in mM: 118 NaCl, photons·cmϪ2·sϪ1) spanning 10° of the mouse’s visual field. This was

25 NaHCO3, 1 NaH2PO4, 3 KCl, 1 MgCl2, 2 CaCl2,10C6H12O6, 0.5 followed by a 120-s interstimulus interval. The mouse was then Downloaded from L-glutamine; Sigma-Aldrich). The retina was incised four times in a rotated in 10° increments and the light protocol repeated such that the Maltese cross motif and mounted onto a 256-channel Multi Electrode stimulus appeared to move from 150° behind the mouse to 30° into the Array (256MEA200/30iR-ITO; Multi Channel Systems, Reutlingen, contralateral visual field. This completed one full rotation, with each Germany) with the ganglion cell layer facing down onto the elec- rotation repeated a minimum of four times. Mice were otherwise kept trodes. A Cyclopore membrane filter (5-␮m pores; Whatman) held the in complete darkness. retina in place while being weighed down by a stainless steel anchor When responses to temporal modulations in irradiance from two (ϳ0.75 g) bearing a framework of parallel polyimide-coated fused rd/rd cl mice were being recorded, light stimuli were generated with silica capillaries (TSP320450, Polymicro Technologies). Electrophys- a bespoke light source (Cairn Research, Faversham, UK) consisting of http://jn.physiology.org/ ␭ ϭ iological signals were acquired with MC_Rack software (Multi Chan- a blue LED ( max 460 nm) that passed through neutral-density nel Systems) through a USB-MEA256 amplifier (for 256-channel filters (Cairn Research) and focused onto a 5-mm-diameter circle of recordings; Multi Channel Systems). Recordings were made at 25 kHz opal diffusing glass (Edmund Optics, York, UK) placed Ͻ1 mm from sampling frequency during the acquisition of electrophysiological the eye contralateral to the recording probe. Placing the diffuser so it activity. To preserve physiological conditions, the tissue was perfused was centered to the midpoint of the eye meant that light would be with carboxygenated aCSF at 2.2 ml/min and maintained at 32°C distributed evenly across the retina. Programs written in LabVIEW (TC01 controller; Multi Channel Systems). (version 8) controlled a National Instruments card (USB-6229) that Light stimuli were projected onto the retina’s ganglion cell layer permitted sinusoidal modulations of the blue LED. This provided a from below. Full-field light stimuli (melanopsin irradiance ϭ 4.21 ϫ 92% Michaelson contrast for melanopsin (mean irradiance ϭ 1.24 ϫ by 10.220.33.4 on March 3, 2017 1014 photons·cmϪ2·sϪ1) were delivered by a custom-written Lab- 1015 photons·cmϪ2·sϪ1) at four temporal frequencies—10 Hz, 1 Hz, VIEW (National Instruments) program instructing an Arduino (Ar- 0.1 Hz, and 0.01 Hz. duino Due) to control a Phlatlight LED. Spatiotemporal stimuli were All light measurements were measured with a calibrated spectro- delivered as 5° or 10° vertical or horizontal bars (which spanned 150 radiometer (Bentham Instruments, Reading, UK). Effective photon ␮mor300␮m on the retina, respectively) for 60 s (7.91 ϫ 1013 flux for the melanopsin photopigment was determined with the cal- melanopsin photons·cmϪ2·sϪ1), with a 180-s interstimulus interval culated spectra and visual pigment template described by Govardo- (2.25 ϫ 1012 melanopsin photons·cmϪ2·sϪ1), by a custom-written vskii et al. (2000). Python script (PsychoPy) instructing an Arduino and a polarizing Tracer injections. Two rd/rd cl mice (434–450 days old) were

LCD projector system (HoloEye Photonics). anesthetized with 1% isoflurane in O2, and a micropipette (NanojectII; In vivo electrophysiology. Eight adult C3H rd/rd cl mice (5 male Drummond Scientific) fitted with a pulled glass capillary (20-␮m- and 3 female; 80–400 days old) were administered an initial dose of diameter tip) was used to inject cholera toxin ␤-subunit conjugated to 0.125% chlorprothixene hydrochloride (0.5 mg/kg; Sigma-Aldrich) either Alexa 488 or Alexa 555 (Invitrogen; catalog no. C-34775 or prior to being anesthetized with 2% isoflurane in O2. Mice were C-34776) into the medial (2.4 mm posterior and 2.1 mm lateral to mounted onto a bespoke stereotaxic frame (SG-4N-S; Narishige) that bregma) and lateral (2.4 mm posterior and 2.3 mm lateral to bregma) was fixed onto a “lazy Susan” (RBB12A; Thorlabs). Isoflurane anes- regions of the dLGN. A trail of 50-nl injections was administered in thesia (0.4–1.0% maintenance) was administered via a nose cone 50-␮m steps spanning the entire dLGN. Brains and retinas were (GM-4; Narishige), and body temperature was maintained at 37°C collected between 5 and 7 days after injection. with a homeothermic blanket (Harvard Apparatus, Edenbridge, UK). Histology. At the end of electrophysiological experiments mice An incision to expose the skull surface was made and a small hole were transcardially perfused with 0.1 M phosphate-buffered saline (ϳ1-mm diameter) drilled 2.3 mm posterior and 2.3 mm lateral to the (PBS) followed by 4% paraformaldehyde (PFA). The brain was bregma, targeting the dLGN. The pupil contralateral to the craniotomy removed, postfixed overnight, cryoprotected with 30% sucrose, and was dilated with topical atropine sulfate (1% wt/vol; Sigma-Aldrich) sectioned at 100 ␮m on a freezing sledge microtome. Sections were and the cornea kept moist with mineral oil. The ipsilateral eye mounted with DPX (Sigma-Aldrich) and coverslipped. Electrode remained covered with blackout material throughout the experiment. placement in the dLGN was confirmed by visualization of a fluores- A recording probe (A4X8-5 mm-50-200-413; Neuronexus) consisting cent dye (Cell Tracker CM-DiI; Invitrogen, Paisley, UK) applied to of four shanks (spaced 200 ␮m apart), each with eight recording sites the probe prior to recording. Images were collected on an Olympus (spaced 50 ␮m apart), was then positioned centrally on the exposed BX51 upright microscope using a ϫ4/0.30 Plan Fln and captured with surface in the coronal plane and lowered to a depth of 2.5–3.3 mm a Coolsnap ES camera (Photometrics) through MetaVue Software with a fluid-filled micromanipulator (MO-10; Narishige). Once the (Molecular Devices). Specific band-pass filters set for DAPI, FITC, recording probe was in position mice were dark adapted for 30 min, and Texas red prevented bleed-through of channels. After tracer allowing neuronal activity to stabilize after probe insertion. Neural studies, mice were killed with an overdose of 30% urethane before signals were acquired with a Recorder64 system (Plexon). Signals they were transcardially perfused and brains processed as above. were amplified 3,000ϫ, high-pass filtered at 300 Hz, and digitized at Additionally, both eyes were removed and the retinas immediately

J Neurophysiol • doi:10.1152/jn.00368.2015 • www.jn.org SPATIAL VISION IN ADVANCED RETINAL DEGENERATION 1323 dissected in the dark in order to maintain their correct dorsal-ventral Receptive field diameters were analyzed by subtracting the average and nasal-temporal axis. Retinas were left overnight in 4% PFA change in firing rate during the 60-s prestimulus baseline from the before being mounted onto glass slides and coverslipped with Prolong average firing rate during the 60-s light pulse to calculate the average Diamond antifade mountant (Life Technologies). change in firing rate. This was normalized to the peak change during Data analysis. Multichannel, multiunit recordings were analyzed each rotation, for each light-responsive channel, to account for in Offline Sorter (Plexon). After removal of cross-channel artifacts, changes in the absolute amplitude of the light response over the long principal component-based sorting was used to discriminate single duration of the experiment. Some preparations had periods in which units, identifiable as a distinct cluster of spikes in principal light responses disappeared for a while; to account for this effect, we component space (Fig. 1, A and D) with a clear refractory period excluded from analysis any repeat at which there was no response to (Ͼ1 ms) in their interspike interval distribution (Fig. 1, C and F). light at any spatial location. A Gaussian function was applied in After spike sorting, data were exported to NeuroExplorer (Nex GraphPad Prism 6 (GraphPad Software) from which the receptive Technologies) and MATLAB R2013a (The MathWorks) for con- field center and size could be determined. A 90% confidence limit struction of peristimulus histograms and further analysis. Light- (Ϯ1.64 SDs from the mean) was applied to the best-fit Gaussian in responsive units were identified as those in which the peristimulus order to calculate receptive field size. average showed a clear peak that exceeded the 99% confidence To generate the retinotopic map, DiI tracts from individual mice limits estimated from a Poisson distribution derived from the pre- were used to determine the placement and depth of the final stimulus spike counts. position of the recording electrode from which the position of Downloaded from http://jn.physiology.org/

Fig. 1. Identification of single units in the retina and dorsal lateral geniculate nucleus (dLGN) of the rd/rd cl mouse. A and D: representative scatterplots of spike waveforms plotted in a 3-dimensional representation of by 10.220.33.4 on March 3, 2017 principal component space (x-axis ϭ PC1; y-axis ϭ PC2; z-axis ϭ PC3) for a retinal (A) and a dLGN (D) recording. A single unit is readily identified in both preparations (single unit ϭ yellow, unsorted waveforms ϭ gray). We use MANOVA F and P values as an indicator of the degree of cluster separation (retina: F ϭ 27.7, P ϭ 9.6Ϫ13; LGN: F ϭ 7.3, P ϭ 0.006). B and E: mean spike waveform for the isolated single units and unsorted waveforms from the units identified in A and D. These 2 units have a spike amplitude of 192 ␮V and 152 ␮V, respectively. C and F: log interspike interval (ISI) for retinal (C) and dLGN (F) recordings display the single unit in black and unsorted waveforms in gray. Isolated single units show clear sharp peaks at discrete intervals Ͼ1 ms, indicative of spikes fired in bursts. The unsorted waveforms show activity across a much wider range of ISIs that is indicative of tonic firing of multiple cells that could not be isolated. Average (mean Ϯ SE) spike waveform for the 28 single dLGN units ϭ 82.9 Ϯ 3.75 ␮V, and spike width at the half-maximum of the valley ϭ 0.19 Ϯ 0.003 s.

J Neurophysiol • doi:10.1152/jn.00368.2015 • www.jn.org 1324 SPATIAL VISION IN ADVANCED RETINAL DEGENERATION individual recording channels could be identified. The mean of the of recording sites across the multielectrode arrays failed to Gaussian fit for multiunit activity (as described above) was corre- record responses to this stimulus in this preparation, we found lated with the anatomical position of the corresponding light- a small number at which reliable changes in firing were responsive channel in the dLGN. This was mapped onto a common LGN template at Ϫ2.06 mm posterior to bregma, and a two- apparent for at least one spatial location of the bar (64 of 1,280 dimensional heat map was generated by a custom-written program multiunit traces from 5 rd/rd cl preparations). Spike sorting of in MATLAB. Temporal modulations in irradiance were analyzed these data resolved 31 consistently responsive single units. A with a custom-written ␹2-based periodogram program in MATLAB couple of representative response profiles are presented in Fig. with a significance value of 0.001. 2. All recorded responses were excitatory (average peak change in firing rate ϭ 4.15 Ϯ 0.56 spikes/s). As previously reported for melanopsin-driven responses, the rd/rd cl light RESULTS responses were characterized by sluggish kinetics, with a long We first sought to characterize the receptive field of indi- latency to respond, followed by a gradual increase in firing vidual ipRGCs in the rd/rd cl retina by in vitro electrophysi- throughout the duration of the light pulse (mean time to ology. We presented a bar of polychromatic light occupying 5° peak ϭ 43.3 Ϯ 3.3 s) and a slow return to baseline after lights or 10° of visual space in 10 or 20 locations spanning the off. multielectrode array (which itself covered 3 mm2 of the retina We found that in all cases RGC responses were strongly Downloaded from or 100° of visual space in each direction). While the majority dependent upon the spatial location of light presentation. This http://jn.physiology.org/ by 10.220.33.4 on March 3, 2017

Fig. 2. Receptive field properties of 2 representative intrinsically photosensitive retinal ganglion cells (ipRGCs) recorded from the rd/rd cl retina. A: peristimulus time histogram (PSTH) demonstrating the response of a representative light-responsive ganglion cell to the presentation of a 60-s, 10° vertical bar in its preferred spatial position over 4 repeats (time bin ϭ 1 s). B: trial bin count (yellow bar in A represents lights on at 0 s and off at 60 s) demonstrating the reproducibility of the melanopsin light response to spatial bar position over 4 repeats of the receptive field mapping protocol (time bin ϭ 5 s; color code ϭ spikes/s). C: the best-fit Gaussian (R2 ϭ 0.993) of the relationship between normalized response amplitude and spatial position of the vertical bar, which reveals a receptive field diameter of 13.3° (dashed vertical lines denote edge of calculated receptive field size). D: trial bin count showing the reproducibility of a second representative unit mapped in 2 dimensions with horizontal (top) and vertical (bottom) 10° bars (time bin ϭ 5 s; color code ϭ spikes/s) over 4 repeats spanning 100° of visual angle. E: the best-fit Gaussian function of the relationship between normalized response amplitude and spatial position of the vertical bar (top; RF size ϭ 38.2°, R2 ϭ 0.80) and the horizontal bar (bottom; RF size ϭ 33.6°, R2 ϭ 0.99). F: 2-dimensional heat map (color code ϭ response amplitude in delta firing rate normalized to maximum response to horizontal and vertical bar presentation) displaying the circular receptive field of this unit when mapped on a 100° region of visual space. In all cases stimulus location is described as distance in both micrometers and degrees of visual space from the edge of the array.

J Neurophysiol • doi:10.1152/jn.00368.2015 • www.jn.org SPATIAL VISION IN ADVANCED RETINAL DEGENERATION 1325 is apparent in the relationship between response amplitude and a range of ages in this work. A comparison of receptive field position of the light bar in the two representative units pre- sizes between relatively young (3–6 mo) and older (Ͼ6 mo) sented in Fig. 2, A and D. In all 31 units, the peak response was mice indicated no very dramatic change in this parameter elicited by bars away from the edge of the MEA, allowing us (Fig. 3A). Given the fragile nature of the rd/rd cl retina, it was to approximate the receptive field by fitting a Gaussian func- not always possible to keep track of its orientation with respect tion to at least one limb (Fig. 2, C and E). Receptive field to dorso-ventral and temporo-nasal axes during dissection. diameter was estimated by applying a 90% confidence limit However, this was achieved in three preparations. The active (Ϯ1.64 SDs from the mean) as measured from a best-fit recording sites from these preparations were biased toward the Gaussian applied to the relationship between bar position and ventral retina (which should have allowed us to sample from all the normalized response amplitude. This demonstrated that ipRGC types; Hughes et al. 2013), but across this region the these receptive fields ranged in size from 8° to 50° (mean receptive field center of light-responsive units appeared fairly receptive field diameter ϭ 22.3 Ϯ 1.9°) of visual angle, evenly distributed and there was no strong relationship be- although most fell between 15° and 25° (Fig. 3A). Eighteen of tween receptive field size and retinal eccentricity (Fig. 3B). these units were mapped in two dimensions, and we found that Some ipRGCs send axon collaterals back into the retina (Joo for these cells receptive fields were roughly circular, with et al. 2013). If these were to drive spiking activity in target differences in receptive field diameter estimates in each dimen- neurons, we might record melanopsin-evoked responses in Downloaded from sion typically being no more than 8% of the mean. To confirm locations very distant from that of light exposure. In fact, we that our estimates of receptive field size were not greatly always found that the strongest light responses occurred when overestimated by our choice of 10° bar widths, three of the the light bar fell above the electrode from which the activity of single units were mapped with 5° bars. We found that these that unit was recorded. three units fell well within the range of cells mapped with a 10° We next turned our attention to the dLGN. We first bar, again showing roughly circular receptive fields with di- identified melanopsin-driven responses by presenting a full- ameters of 21.3°, 15.0°, and 29.8° of visual angle. field white light pulse to isoflurane-anesthetized rd/rd cl http://jn.physiology.org/ Although our study did not set out to systematically assess mice. In common with previous reports, we found slow but the impact of aging on receptive field structure, we did include sustained increases in firing in response to this stimulus at ϳ37% of channels (71/192 light-responsive channels). We next applied a receptive field mapping protocol analogous to that used for the retina in vitro. In brief, 10° vertical bars were presented for 60 s at 19 locations spanning 180° of the visual field with an interstimulus interval of 240 s. In pilot

experiments we found that such very long timescales were by 10.220.33.4 on March 3, 2017 required to ensure robust responses. However, this meant that each repeat of the receptive field mapping protocol lasted2h15min. Of these 71 light-responsive channels, 58 multiunit traces within the LGN responded reliably to at least one spatial location of the bar across multiple repeats. From the 58 channels showing reproducible responses to the spatial stimuli, we isolated 28 single units for detailed receptive field analysis. The response of two such units is shown in Fig. 4. The response of these cells when the bar was in the center of their receptive field had the poor temporal resolution expected for melanopsin-driven activity (n ϭ 28 cells; average time to peak firing ϭ 32.5 Ϯ 2.0 s; Fig. 4A). In all cases, cells responded to at least two spatial locations of the bar on multiple repeats (Fig. 4B), allowing us to approximate receptive fields by fitting Gaussian functions to the relationship between response amplitude and bar position (Fig. 4C). We confirmed that large receptive fields (such as that in Fig. 4C, bottom) were not an artifact of poor spike sorting by calculating the mean spike waveform evoked during bar presentation at both the center (above bar position 100°) and edges (above bar position 50° and 150°) of the receptive field and found these to be consistent throughout our analysis. Fig. 3. Receptive field properties of ipRGC population recorded from the rd/rd Across the population of responsive units, receptive field cl retina. A: population data displaying receptive field size of 31 single units sizes ranged from 18.2° to 92.6° of visual angle (Fig. 5A). recorded from 5 rd/rd cl retinas. Receptive field diameter ranged from 8.8° to Receptive field centers were distributed across a wide range of 49.9° of visual angle (n ϭ 31 units; mean Ϯ SD receptive field size ϭ 22.3 Ϯ bar positions (Fig. 5B). Across this range, there was no obvious 10.4°). B: plot of receptive field location for 18 single units projected onto map of retinal surface from 3 rd/rd cl mice. Lines depict the extent of receptive relationship between the location and size of the receptive fields for individual neurons in horizontal axis (in ␮m of retinal surface); scale fields. Interestingly, none of the units had receptive field bar and indication of retinal orientation are provided on right. centers in the most nasal 60° sampled (although given the large

J Neurophysiol • doi:10.1152/jn.00368.2015 • www.jn.org 1326 SPATIAL VISION IN ADVANCED RETINAL DEGENERATION Downloaded from http://jn.physiology.org/

Fig. 4. Receptive field properties of 2 representative melanopsin-driven neurons in the rd/rd cl dLGN. A: PSTH demonstrating the average melanopsin light response of 2 representative units with a narrow (top) or wide (bottom) receptive field in the dLGN when presented with a 60-s, 10° vertical bar in its preferred spatial location (time bin ϭ 1 s). B: trial bin counts demonstrating the reproducibility of this melanopsin response to spatial bar position over 4 repeats of the receptive field mapping protocol (time bin ϭ 5 s; color code ϭ spikes/s). C: the best-fit Gaussian of the relationship between normalized response amplitude and spatial position of the vertical bar (top, R2 ϭ 0.98; bottom, R2 ϭ 0.80) demonstrating small and large receptive field sizes of 29° and 89°, respectively. The average spike waveform evoked during the 60-s bar presentation is shown for bars at the edge of the large receptive field (bar position 50° and 150°) as well as the peak response (bar position 100°). receptive field size that is not to say that we failed to record of the mouse dLGN (data not shown). In one mouse, we also by 10.220.33.4 on March 3, 2017 responses for bars at these locations). One possible explanation found that these two tracers preferentially labeled distinct for this is that our electrode placements targeting the central locations of the medial and lateral regions of V1 and V2M, dLGN had failed to adequately sample very rostral portions of indicating that retinotopic order remains at least partially intact the nucleus receiving strongest input from the temporal retina. after retinal degeneration (Fig. 6, E and F). The implication would be that the remaining visual input to the The retention of spatial order in the visual information rd/rd cl dLGN retains some retinotopic order. To explicitly reaching the dLGN of rd/rd cl mice raises the possibility of investigate this possibility, we used histological techniques to melanopsin-driven light responses in this animal supporting estimate the location of the electrodes from which these re- spatial discrimination. However, two aspects of the responses had been recorded (Fig. 6A). To maximize the sample sponses we record would substantially reduce the utility of size for this analysis, we turned to multiunit activity, which this information under natural viewing conditions. First, we allowed inclusion of responses from units that could not be find that the magnitude and timing of melanopsin-driven unambiguously spike sorted. Correlating the receptive field responses vary substantially between presentations. This is center with the anatomical location of the light-responsive apparent in the activity of a representative unit across channel revealed a smooth gradient from nasal to temporal multiple repeats of a 60-s full-field pulse over half an hour visual field across a ventro-medial to dorso-lateral axis of the (Fig. 7A). A second, most substantial problem is their very dLGN (Fig. 6B). This relationship exists both within individual poor temporal fidelity. Increases in firing rate build up and rd/rd cl mice and when combining the single or multiunit decay very slowly and therefore do not track light incre- receptive fields of all six animals and is consistent with ments and decrements very accurately (Brown et al. 2010; previous reports for visually intact animals. We continued to Fig. 4A and Fig. 7, A and B). Moreover, to record the data demonstrate the retention of retinotopic order by injecting presented here we have had to include very long interstimu- cholera toxin ␤-subunit conjugated to either Alexa 488 or lus intervals, indicating very gradual recovery from prior Alexa 555 into the medial and lateral portions of the dLGN, light exposure. To further explore the implication of these respectively (Fig. 6C). Here, from two injected mice, we found characteristics for the ability of melanopsin to track dy- that the retrograde transport of these tracers preferentially namic changes in light intensity in these retinally degenerate labels distinct portions of the temporal and nasal retina as animals, we finally recorded the response to full-field sinu- predicted given the retinotopic organization of the mouse soidal modulations (92% Michelson contrast) over a wide dLGN (Fig. 6D). We also found Alexa 488 labeling in the most range of temporal frequencies, 10 Hz, 1 Hz, 0.1 Hz, and 0.01 temporal regions of the ipsilateral retina, which is in agreement Hz. Across 22 light response channels (or 16 single units with the injection site of this tracer spanning the binocular zone isolated from them), we did not find a single example of a

J Neurophysiol • doi:10.1152/jn.00368.2015 • www.jn.org SPATIAL VISION IN ADVANCED RETINAL DEGENERATION 1327

non-M1 types. These values are in agreement with the anatomical estimates of dendritic field diameters of ipRGCs (149–477 ␮m) in wild-type retinas (Berson et al. 2010; Ecker et al. 2010; Estevez et al. 2012; Schmidt and Kofuji 2009). Of interest is that the non-M1 ipRGCs, which are believed to target the dLGN, were found to possess center-surround receptive fields while M1 ipRGCs, which predominantly target NIF centers, do not (Zhao et al. 2014). Given that surround antagonism originates in the circuitry linking rods/cones to ganglion cells, it is not unsurprising that we did not observe center-surround elements of receptive fields in either the retina or dLGN of rd/rd cl mice. This would, though, impact the quality of spatial information provided by ipRGCs in advanced retinal degeneration. In other respects, our data are broadly consistent with data

from wild-type mice. In our experiments in the rd/rd cl retina Downloaded from we found that receptive field diameter was commonly between 15° and 25° of visual angle, corresponding to between 450 and 750 ␮m in diameter on the retina (mean Ϯ SE diameter ϭ 668 Ϯ 56 ␮m). This is consistent with estimates of ipRGC dendritic field size and morphological evidence that outer retinal degeneration does not greatly affect the integrity of the ipRGC population (Semo et al. 2003; Vugler et al. 2008). http://jn.physiology.org/ However, a small proportion of RCGs (6/31) possessed receptive fields that were more than double that predicted for a single ipRGC with even the largest dendritic field. What could result in these larger-diameter receptive fields? One possibility is that a melanopsin-driven light response propagates through gap junctions, creating a localized spatial network of excitabil- Fig. 5. Receptive field properties of melanopsin-driven responses in the rd/rd ity in the retina. ipRGCs make gap junction connections with cl dLGN. A: distribution of receptive field widths in the horizontal axis for 28 displaced amacrine cells in the ganglion cell layer (Bramley et by 10.220.33.4 on March 3, 2017 single units recorded from the dLGN of 6 rd/rd cl mice. Receptive field al. 2011; Schmidt et al. 2008; Sekaran et al. 2003), which can diameter ranged from 18.2° to 92.6° of visual angle (mean Ϯ SD receptive field size ϭ 56.6 Ϯ 22.0°). B: relationship between the diameter (in horizontal relay this information to a heterologous population of RGCs axis) and location in visual space of the receptive fields from 28 single units. (Muller et al. 2010). This is supported by strongly correlated Single units recorded from each of the 6 different rd/rd cl mice are presented firing of RGCs via gap junctions that has been reported in the in a single color. The location of receptive field centers is provided in degrees rd1 retina (Borowska et al. 2011; Toychiev et al. 2013). If that of visual angle, using a reference of 0° for directly in front of the mouse. Note that few receptive fields are centered in the nasal part of the visual field and were the case, then we might expect factors that modulate gap that there is no clear correlation between receptive field position and receptive junction coupling in the inner retina to further improve the field size. quality of spatial information provided by ipRGCs in advanced retinal degeneration. convincing modulation in firing to any of these sinusoids (P We found that the receptive fields of melanopsin-driven Ͼ 0.001; Fig. 7, C and D). LGN neurons are substantially larger than those recorded from ipRGCs in the retina. In the rd/rd cl retina, an LGN neuron’s receptive field size would be defined by the number of ipRGCs DISCUSSION from which it receives input (directly or indirectly) and the Using in vitro and in vivo extracellular electrophysiological dendritic field of each of these ipRGCs. The latter parameter recordings, we mapped spatial receptive fields in the retina and would vary between subtypes of ipRGC. Nevertheless, a sim- dLGN of rd/rd cl mice. In each method, we found that ple comparison between the average receptive field diameter of receptive fields are large but clearly defined and cover the light-responsive units in the rd/rd cl retina (diameter ϭ 22°, mouse visual field. Furthermore, these responses adopt a func- area ϭ 380°2 and dLGN diameter ϭ 56°, area ϭ 2,463°2) tional retinotopic organization in the dLGN. This argues that suggests that each light-responsive dLGN unit is influenced on low-resolution spatial discrimination could be possible from average by at least six light-responsive RGCs with nonover- melanopsin-driven light responses in the rd/rd cl mouse. How- lapping receptive fields. However, there must be great variation ever, the very poor temporal fidelity of melanopsin-driven light in this value on the individual unit level. Thus if the dLGN unit response in this model represents a substantial barrier to with the largest receptive field (diameter ϭ 92°, area ϭ 6,500°2 realizing that potential. if symmetrical) received its input from those retinal units with In the intact retina, spatial receptive fields have been de- the smallest receptive fields (diameter ϭ 9°, area ϭ 64°2) then scribed for individual ipRGCs in the rat and mouse (Estevez et it must be downstream of at least 100 ganglion cells (even al. 2012; Wong et al. 2007; Zhao et al. 2014). Receptive field assuming that the ganglion cells influencing it tile the retina diameter was found to vary between ipRGC subtypes, being perfectly). Conversely, we find dLGN units whose receptive ϳ500 ␮m for M1 cells and ranging from 200 to 500 ␮m for field diameters are within the range of those found in the retina,

J Neurophysiol • doi:10.1152/jn.00368.2015 • www.jn.org 1328 SPATIAL VISION IN ADVANCED RETINAL DEGENERATION

Fig. 6. Retinotopic organization in the rd/rd cl visual system. A: reconstructed positions of the recording sites superimposed on schematic representatives of the dLGN extent in coronal sections through the mouse brain atlas at Ϫ1.82 mm and Ϫ2.18 mm relative to bregma. Reconstruction was based on DiI labeling of the electrode shank. Colored dots (orange) depict recording sites from which visually responsive multiunit traces were recorded. Brain schematics based on Paxinos and Franklin mouse atlas (Paxinos and

Franklin 2004) with permission; solid lines Downloaded from depict boundaries of dorsal and ventral LGN and dashed line the intergeniculate leaﬂet. B: 2-dimensional heat map correlating anatomical position of electrodes in the dLGN with position of receptive ﬁelds in the visual scene for multiunit activity recorded at that site. A gradient from ventro-medial to dorso-

lateral dLGN matches the progression from http://jn.physiology.org/ nasal to ventral visual fields (n ϭ 58 multiunit traces within the LGN from 6 rd/rd cl mice). C, top: cholera toxin ␤-subunit injections into the medial (Alexa 488; green) and lateral (Alexa 555; red) regions of the dLGN. Bottom: overlay of injection pattern onto mouse atlas to illustrate the extent of cholera toxin labeling in the dorsal LGN. Adapted from Paxinos and Franklin (2004) with permission. D, top: wholemount retina contralateral to the injected dLGN shown in by 10.220.33.4 on March 3, 2017 C demonstrating distinct regions of labeling with cholera toxin ␤-subunit conjugated to Alexa 488 in the nasal retina and Alexa 555 in the temporal retina. Bottom: magnified image of cholera toxin-labeled retinal ganglion cell bodies from nasal (green highlighted box, top) and temporal (red highlighted box, top) retina. E: regions of V1 and V2ML (3.5 mm posterior to bregma) in the ipsilateral hemisphere to the dLGN injection site demonstrate distinct regions of labeling with both cholera toxin ␤-subunit conjugates. F: magnified image highlighting labeled cell bodies in lateral and medial V2M by cholera toxin ␤-subunit conjugated to Alexa 488 and 555, respectively.

indicating that they could be downstream from a single light- routed via other neurons in the thalamus or elsewhere in the responsive RGC. Current literature reports that between one brain. and four RGCs synapse with a single neuron in the wild-type An important determinant of the potential contribution of LGN (Chen and Regehr 2000; Hooks and Chen 2008). Our ipRGCs to spatial vision is whether their projections match the data do not directly address the question of whether this retinotopic order of other ganglion cell terminals in the dLGN. arrangement is altered in retinal degeneration. Although many Estevez et al. (2012) reported that a tracer injected into the of our light-responsive dLGN units clearly lie downstream medial dLGN in wild-type mice labeled ipRGCs in the dorsal from more than four RGCs, this need not reflect direct synaptic retina, indicating that they do. Our functional characterization input, as within the timescales of response there is ample of visual responses in the dLGN is consistent with this view. opportunity for the visual information reaching them to be Thus we find that the distribution of receptive fields in visual

J Neurophysiol • doi:10.1152/jn.00368.2015 • www.jn.org SPATIAL VISION IN ADVANCED RETINAL DEGENERATION 1329

Fig. 7. Reproducibility and temporal ﬁdelity of the melanopsin-driven light-responses in the rd/rd cl dLGN. A and B: trial bin count (A) and corresponding PSTH (B) of a representative melanopsin-driven light response in the rd/rd cl dLGN (time bin ϭ 5 s; color code ϭ spikes/s) reveals poor temporal resolution and great variation in response latency, peak response amplitude, and recovery time over 10 repeats of a 60-s light pulse (interstimulus interval ϭ 120 s; time bin ϭ 5 s). C: mean Ϯ SE ﬁring rate for a represen- Downloaded from tative light-responsive unit across sinusoidal modulations in irradiance (1.24 ϫ 1015 photons·cmϪ2·sϪ1) at 4 different temporal frequencies—10 Hz, 1 Hz, 0.1 Hz, and 0.01 Hz [100 repeats (20 repeats for 0.01 Hz); time bin ϭ 100 data points per period for each temporal frequency; ␹2 periodogram

P ϭ 1.00]. Data are double plotted, with the http://jn.physiology.org/ associated change in irradiance shown with the solid line above. No unit showed a significant modulation in firing at any frequency (␹2 periodogram P Ͼ 0.001). D: this is also apparent in the mean (ϮSE) firing rate across all 22 multiunit traces from the LGN of 2 rd/rd cl mice exposed to sinusoidal modulations at these temporal frequencies (P Ͼ 0.001). by 10.220.33.4 on March 3, 2017

space across the dLGN of rd/rd cl mice matches that recently fidelity of the melanopsin-driven response in this genotype described for conventional RGC input in wild types (Piscopo et provides a straightforward explanation for that finding. The al. 2013). This is supported by our cholera toxin tracer study, very long latencies for changes in firing following both appear- which demonstrates that at least some order is retained in a ance and disappearance of light bars here are consistent with heterologous population of RGCs projecting from the retina to many previous reports of the light response of deafferented the dLGN in this model compared with the known retinotopic ipRGCs (Berson et al. 2002; Do et al. 2009). To explore how organization of the wild-type dLGN (Piscopo et al. 2013) and that might impact their ability to track dynamic modulations in visual cortex (Schuett et al. 2002). To our knowledge, ours is local luminance, we presented full-field sinusoidal oscillations the first report of functional retinotopy in the dLGN after in light intensity over a wide range of temporal frequencies. retinal degeneration. The fact that it is broadly retained is We were unable to find any units in the rd/rd cl dLGN that encouraging for the potential of therapies currently under convincingly tracked sinusoidal oscillations over any temporal development aimed at restoring vision in patients with ad- frequency. One interesting question is whether this is a funda- vanced retinal degeneration. These are largely focused on mental limit to melanopsin’s capacity or, in part, a conse- providing visual information to the surviving RGCs and will quence of retinal degeneration. The kinetics of the melanopsin rely upon those targeted neurons making appropriate connec- single-photon response are poor (Do et al. 2009), but neither tions within the brain. onset nor decay latencies approach those observed here for Although the receptive field sizes of neurons in the rd/rd cl dLGN responses in rd/rd cl mice. The reasons for this discrep- dLGN are large, the fact that they exist and adopt a retinotopic ancy are unclear. However, kinetics of the melanopsin-driven order implies that they could support low-acuity spatial vision. responses in animals with an intact visual system more closely In fact, although there is evidence that melanopsin can support match those of the single-photon response, suggesting that the visual discrimination in advanced retinal degeneration (Zaidi et very sluggish activity observed here may be a consequence of al. 2007), rd/rd (and rd/rd cl) mice perform extremely poorly rod/cone loss. In this context, our data revealing discrete in even the most rudimentary spatial navigation tasks (Brown receptive fields and retinotopic order to melanopsin-driven et al. 2012; Thyagarajan et al. 2010). The very poor temporal responses in the dLGN have a couple of interesting implica-

J Neurophysiol • doi:10.1152/jn.00368.2015 • www.jn.org 1330 SPATIAL VISION IN ADVANCED RETINAL DEGENERATION tions. First, the fundamental capacity to provide spatial infor- Hooks BM, Chen C. Vision triggers an experience-dependent sensitive period mation these data reveal might be translated into a genuine at the retinogeniculate synapse. J Neurosci 28: 4807–4817, 2008. contribution to pattern vision in intact animals and at earlier Hughes S, Watson TS, Foster RG, Peirson SN, Hankins MW. Nonuniform distribution and spectral tuning of photosensitive retinal ganglion cells of the stages of degeneration. Second, any method of addressing the mouse retina. Curr Biol 23: 1696–1701, 2013. poor temporal kinetics of the melanopsin-driven response in Joo HR, Peterson BB, Dacey DM, Hattar S, Chen SK. Recurrent axon retinal degeneration might represent a viable strategy for re- collaterals of intrinsically photosensitive retinal ganglion cells. Vis Neurosci storing crude spatial discrimination. 30: 175–182, 2013. Lucas RJ, Freedman MS, Lupi D, Munoz M, David-Gray ZK, Foster RG. Identifying the photoreceptive inputs to the mammalian circadian system ACKNOWLEDGMENTS using transgenic and retinally degenerate mice. Behav Brain Res 125: The authors thank J. Wynne for technical assistance. 97–102, 2001. Lucas RJ, Freedman MS, Munoz M, Garcia-Fernandez JM, Foster RG. Regulation of the mammalian pineal by non-rod, non-cone, ocular photo- GRANTS receptors. Science 284: 505–507, 1999. Lucas RJ, Hattar S, Takao M, Berson DM, Foster RG, Yau KW. Dimin- This research was funded by the European Research Council via award ished pupillary light reﬂex at high irradiances in melanopsin-knockout mice. 268970 to R. J. Lucas. Science 299: 245–247, 2003.

Marc RE, Jones BW, Anderson JR, Kinard K, Marshak DW, Wilson JH, Downloaded from DISCLOSURES Wensel T, Lucas RJ. Neural reprogramming in retinal degeneration. Invest Ophthalmol Vis Sci 48: 3364–3371, 2007. No conflicts of interest, financial or otherwise, are declared by the author(s). Muller LP, Do MT, Yau KW, He S, Baldridge WH. Tracer coupling of intrinsically photosensitive retinal ganglion cells to amacrine cells in the AUTHOR CONTRIBUTIONS mouse retina. J Comp Neurol 518: 4813–4824, 2010. Paxinos G, Franklin K. The Mouse Brain in Stereotaxic Coordinates (2nd Author contributions: C.A.P., T.M.B., and R.J.L. conception and design of ed.). San Diego, CA: Gulf Professional, 2004. research; C.A.P., A.E.A., N.M., and R.J.L. interpreted results of experiments; Peirce JW. Generating stimuli for neuroscience using PsychoPy. Front Neu- http://jn.physiology.org/ C.A.P. and R.J.L. drafted manuscript; C.A.P., C.G.E., R.S., A.E.A., N.M., roinform 2: 10, 2008. T.M.B., and R.J.L. edited and revised manuscript; C.A.P., R.S., T.M.B., and Piscopo DM, El-Danaf RN, Huberman AD, Niell CM. Diverse visual R.J.L. approved final version of manuscript; C.G.E. and R.J.L. performed features encoded in mouse lateral geniculate nucleus. J Neurosci 33: experiments; R.S., A.E.A., and R.J.L. analyzed data; R.S. and R.J.L. prepared 4642–4656, 2013. figures. Schmidt TM, Alam NM, Chen S, Kofuji P, Li W, Prusky GT, Hattar S. A role for melanopsin in alpha retinal ganglion cells and contrast detection. REFERENCES Neuron 82: 781–788, 2014. Schmidt TM, Kofuji P. Functional and morphological differences among Berson DM, Castrucci AM, Provencio I. Morphology and mosaics of intrinsically photosensitive retinal ganglion cells. J Neurosci 29: 476–482, melanopsin-expressing retinal ganglion cell types in mice. J Comp Neurol 2009. 518: 2405–2422, 2010. Schmidt TM, Taniguchi K, Kofuji P. Intrinsic and extrinsic light responses by 10.220.33.4 on March 3, 2017 Berson DM, Dunn FA, Takao M. Phototransduction by retinal ganglion cells in melanopsin-expressing ganglion cells during mouse development. J that set the circadian clock. Science 295: 1070–1073, 2002. Neurophysiol 100: 371–384, 2008. Borowska J, Trenholm S, Awatramani GB. An intrinsic neural oscillator in Schuett S, Bonhoeffer T, Hubener M. Mapping retinotopic structure in the degenerating mouse retina. J Neurosci 31: 5000–5012, 2011. mouse visual cortex with optical imaging. J Neurosci 22: 6549–6559, 2002. Bramley JR, Wiles EM, Sollars PJ, Pickard GE. Carbenoxolone blocks the Sekaran S, Foster RG, Lucas RJ, Hankins MW. Calcium imaging reveals a light-evoked rise in intracellular calcium in isolated melanopsin ganglion network of intrinsically light-sensitive inner-retinal neurons. Curr Biol 13: cell photoreceptors. PloS One 6: e22721, 2011. 1290–1298, 2003. Brown TM, Gias C, Hatori M, Keding SR, Semo M, Coffey PJ, Gigg J, Semo M, Lupi D, Peirson SN, Butler JN, Foster RG. Light-induced c-fos in Piggins HD, Panda S, Lucas RJ. Melanopsin contributions to irradiance melanopsin retinal ganglion cells of young and aged rodless/coneless (rd/rd coding in the thalamo-cortical visual system. PloS Biol 8: e1000558, 2010. cl) mice. Eur J Neurosci 18: 3007–3017, 2003. Brown TM, Tsujimura SI, Allen AE, Wynne J, Bedford R, Vickery G, Strettoi E, Pignatelli V. Modifications of retinal neurons in a mouse model of Vugler A, Lucas RJ. Melanopsin-based brightness discrimination in mice retinitis pigmentosa. Proc Natl Acad Sci USA 97: 11020–11025, 2000. and humans. Curr Biol 22: 1134–1141, 2012. Thyagarajan S, van Wyk M, Lehmann K, Lowel S, Feng GP, Wassle H. Chen CF, Regehr WG. Developmental remodeling of the retinogeniculate Visual function in mice with photoreceptor degeneration and transgenic synapse. Neuron 28: 955–966, 2000. expression of Channelrhodopsin 2 in ganglion cells. J Neurosci 30: 8745– Do MT, Kang SH, Xue T, Zhong H, Liao HW, Bergles DE, Yau KW. 8758, 2010. Photon capture and signalling by melanopsin retinal ganglion cells. Nature Toychiev AH, Ivanova E, Yee CW, Sagdullaev BT. Block of gap junctions 457: 222–223, 2009. eliminates aberrant activity and restores light responses during retinal Ecker JL, Dumitrescu ON, Wong KY, Alam NM, Chen SK, LeGates T, degeneration. J Neurosci 33: 13972–13977, 2013. Renna JM, Prusky GT, Berson DM, Hattar S. Melanopsin-expressing Vugler AA, Semo M, Joseph A, Jeffery G. Survival and remodeling of retinal ganglion-cell photoreceptors: cellular diversity and role in pattern melanopsin cells during retinal dystrophy. Vis Neurosci 25: 125–138, 2008. vision. Neuron 67: 222–223, 2010. Wong KY, Dunn FA, Graham DM, Berson DM. Synaptic influences on rat Estevez ME, Fogerson PM, Ilardi MC, Borghuis BG, Chan E, Weng S, ganglion-cell photoreceptors. J Physiol 582: 279–296, 2007. Auferkorte ON, Demb JB, Berson DM. Form and function of the M4 cell, Zaidi FH, Hull JT, Peirson SN, Wulff K, Aeschbach D, Gooley JJ, an intrinsically photosensitive retinal ganglion cell type contributing to Brainard GC, Gregory-Evans K, Rizzo JF, Czeisler CA 3rd, Foster RG, geniculocortical vision. J Neurosci 32: 222–223, 2012. Moseley MJ, Lockley SW. Short-wavelength light sensitivity of circadian, Freedman MS, Lucas RJ, Soni B, von Schantz M, Munoz M, David-Gray pupillary, and visual awareness in humans lacking an outer retina. Curr Biol Z, Foster R. Regulation of mammalian circadian behavior by non-rod, 17: 2122–2128, 2007. non-cone, ocular photoreceptors. Science 284: 222–223, 1999. Zhao X, Stafford BK, Godin AL, King WM, Wong KY. Photoresponse Govardovskii VI, Fyhrquist N, Reuter T, Kuzmin DG, Donner K. In search diversity among the five types of intrinsically photosensitive retinal ganglion of the visual pigment template. Vis Neurosci 17: 509–528, 2000. cells. J Physiol 592: 1619–1636, 2014.

J Neurophysiol • doi:10.1152/jn.00368.2015 • www.jn.org