(Zebrafish and Ambon Damselfish) Emmanuel Ma

Home , Damselfish

Visual learning and its underlying neural substrate in two species of teleost fish

(Zebrafish and Ambon damselfish)

Emmanuel Marquez-Legorreta

Bachelor in Psychology

Master in Neuroscience

A thesis submitted for the degree of Doctor of Philosophy at

The University of Queensland in 2019

The Faculty of Medicine Abstract

The visual sense is one of the main sources of information for many animals. In particular, many species of fish rely on vision for survival. Whether a fish needs to distinguish between possible predators, sources of food, a possible mate or competitors for their territory, learning to discriminate between visual stimuli is fundamental part of their life.

Recently, studies have shown that multiple species of fish are able to solve visual discrimination tasks that were thought to be too complex for these organisms (Brown et al., 2011). Furthermore, neuroanatomical studies have also found that although the layout of the central nervous system of teleost fish is different, multiple structures seem to have homologues in mammalian brains (Mueller et al., 2011). Together, these findings have reinforced the role of fish as a model system to study the neuronal substrates of simple and complex behaviours (Gerlai, 2014).

A particular case is the one of zebrafish, whose development as a model for neuroscience has been exceptionally rapid (Stewart et al., 2010; Blaser and Vira, 2014; Kalueff et al., 2014; Stewart et al., 2014a; d'Amora and Giordani, 2018). In the last decade, the combination of genetic and optical technology has allowed the neuronal activity imaging of the whole brain of zebrafish larvae (Ahrens et al., 2013a; Wolf et al., 2015; Vanwalleghem et al., 2018). These powerful technologies are presenting us for the first time with the opportunity to see and analyse the processing of visual information in the whole brain as fish learn.

To accomplish the goal of analysing visual learning in the larval zebrafish’s whole brain, I chose to use a habituation paradigm. Habituation is a simple form of learning defined as a reduction of an innate response to a frequently repeated stimulus. Zebrafish larvae show habituation to visual and auditory stimuli (Best et al., 2008; Roberts et al., 2011; Wolman et al., 2011; Randlett et al., 2019), and provide an appealing platform from which to study habituation’s circuit-level mechanisms. When presented with a looming stimulus that resembles an approaching predator, larvae respond with a rapid escape behaviour. I first showed that zebrafish larvae are capable of habituation to repetitive looms, decreasing the probability of escape responses. Also, changing features of the stimulus, like speed and inter stimulus interval, modulated the habituating responses. Next, I used a selective plane illumination microscopy (SPIM) microscope and calcium imaging to visualize neuronal activity and localize the regions associated in this learning behaviour. The imaging experiments showed that different populations of neurons display a range of habituating responses across the brain, varying in their habituation rates. I then used graph theory to model the network connectivity changes during habituation, and found that some parts of the network disengage early while a smaller fragment

sustains its connectivity. A third fraction of the network, whose elements are mostly located in the optic tectum, disconnect gradually and seem to be involved in the reengagement of the network when recovery occurs. Finally, I performed habituation experiments with fmr1 mutant zebrafish larvae, a fragile X syndrome (FXS) model. The results of both behavioural experiments and network analysis of the neuronal activity suggest that fmr1 mutants habituate more slowly and present an enhanced recovery after a period of rest.

Although the zebrafish larva offers important experimental advantages, experiments in this system are restricted to basic learning paradigms. Because of this, a part of this thesis is devoted to complex visual learning in the Ambon damselfish. This fish is particularly skilled at visual discrimination tasks at it uses facial patterns to discriminate among individuals and other similar species (Siebeck, 2004; Siebeck et al., 2010). Using operant conditioning methods to train the fish for visual discrimination tasks and immunohistological techniques, I attempted to localize the possible forebrain areas involved in visual learning. Exploratory results suggest the involvement of parts of the telencephalon during these tasks. Finally, I attempted to disrupt their visual learning performance while simulating an ocean acidification scenario, as it has been shown to have sensory detrimental consequences in fish (Clements and Hunt, 2015; Nagelkerken and Munday, 2016). However, my results showed an absence of notable differences in visual discrimination and visual learning, which suggest that adult Ambon damselfish visual abilities were not affected.

The results of this thesis identified multiple teleost brain structures involved in visual learning, many of which are homologues of subcortical mammalian pathways. The relevance of these findings, including in relation to the human nervous system, is addressed in the final discussion of the thesis. Altogether, these results are important to better understand vertebrate’s visual learning and they open future directions to further investigate visual cognition in teleosts.

Declaration by author

This thesis is composed of my original work, and contains no material previously published or written by another person except where due reference has been made in the text. I have clearly stated the contribution by others to jointly-authored works that I have included in my thesis.

I have clearly stated the contribution of others to my thesis as a whole, including statistical assistance, survey design, data analysis, significant technical procedures, professional editorial advice, financial support and any other original research work used or reported in my thesis. The content of my thesis is the result of work I have carried out since the commencement of my higher degree by research candidature and does not include a substantial part of work that has been submitted to qualify for the award of any other degree or diploma in any university or other tertiary institution. I have clearly stated which parts of my thesis, if any, have been submitted to qualify for another award.

I acknowledge that an electronic copy of my thesis must be lodged with the University Library and, subject to the policy and procedures of The University of Queensland, the thesis be made available for research and study in accordance with the Copyright Act 1968 unless a period of embargo has been approved by the Dean of the Graduate School.

I acknowledge that copyright of all material contained in my thesis resides with the copyright holder(s) of that material. Where appropriate I have obtained copyright permission from the copyright holder to reproduce material in this thesis and have sought permission from co-authors for any jointly authored works included in the thesis.

III

Publications included in this thesis

– Marquez-Legorreta E, Piber M, and Scott EK. (2019) Visual Escape in Larval Zebrafish: Stimuli, Circuits, and Behavior. In: The behaviour genetics of zebrafish (Danio rerio). Elsevier.

Submitted manuscripts included in this thesis

– Marquez-Legorreta E, Constantin L, Piber M, Favre-Bulle IA, Taylor MA, Vanwalleghem GC, and Scott EK (submitted. Under review in Nature) Brain-wide visual habituation networks in wild-type and fmr1 zebrafish.

Other publications during candidature

Oral presentations

– Marquez-Legorreta E, Constantin L, Piber M, Favre-Bulle I, Taylor M, Vanwalleghem G, Scott E. Brain-wide visual habituation networks in wild type and fmr1 zebrafish. Annual Conference of the Australasian Neuroscience Society, Adelaide, 2019. – Marquez-Legorreta E, Constantin L, Piber M, Favre-Bulle I, Taylor M, Vanwalleghem G, Scott E. Brain-wide visual habituation networks in wild type and fmr1 zebrafish. Zebrafish Neural Circuit and Behaviour, Cold Spring Harbor Laboratories, 2019. – Marquez-Legorreta E, Favre-Bulle I, Taylor M, Heap, L, Vanwalleghem G, Scott E. Habituation to looming stimuli in zebrafish larvae. 38th Meeting of the Australasian Neuroscience Society, 2018. – Marquez-Legorreta E, Favre-Bulle I, Taylor M, Heap, L, Vanwalleghem G, Scott E. Habituation to looming stimuli in zebrafish larvae. 9th SBMS International Biomedical Postgraduate Symposium. The University of Queensland, 2018. – Marquez-Legorreta E, Favre-Bulle I, Taylor M, Heap, L, Vanwalleghem G, Scott E. Habituation to looming stimuli in zebrafish larvae. 19th Australia and New Zealand Zebrafish Meeting, Wollongong, Australia, 2018. – Marquez-Legorreta E, Bender D, Parker AN, Rosinski C, Siebeck UE. Visual learning in a changing climate. Heron Island Research Station Research Seminars. 9th of June 2017. – Marquez-Legorreta E, Bender D, Parker AN, Rosinski C, Siebeck UE. Fish visual discrimination and learning in a changing world. 43rd Annual Conference of the Australasian Society for the Study of Animal Behaviour. July of 2016.

Poster presentations

– Marquez-Legorreta E, Favre-Bulle I, Taylor M, Heap, L, Vanwalleghem G, Scott E. Habituation to looming stimuli in zebrafish larvae. International Congress of Neuroethology, Brisbane, Australia, 2018. – Marquez-Legorreta E, Favre-Bulle I, Taylor M, Heap, L, Vanwalleghem G, Scott E. Zebrafish larvae habituation to looming stimuli. 8th SBMS International Biomedical Postgraduate Symposium. The University of Queensland, 2017. – Marquez-Legorreta E, Bender D, Parker AN, Rosinski C, Siebeck UE. Fish visual discrimination and learning in a changing world. 7th SBMS International Biomedical Postgraduate Symposium. The University of Queensland, 2016.

Publications included in this thesis

Marquez-Legorreta E, Piber M, and Scott EK. (2019) Visual Escape in Larval Zebrafish: Stimuli, Circuits, and Behavior. In: The behaviour genetics of zebrafish (Danio rerio). Elsevier. – incorporated as Chapter 2.

Contributor Statement of contribution

Author Marquez-Legorreta E (Candidate) Conception and design (60%)

Analysis and interpretation (75%)

Drafting and production (60%)

Author Piber M Conception and design (20%)

Analysis and interpretation (25 %)

Drafting and production (20%)

Author Scott EK Conception and design (20%)

Analysis and interpretation (0%)

Drafting and production (20%)

Contributions by others to the thesis

E. Scott (Scott Laboratory, The University of Queensland), G. Vanwalleghem (Scott Laboratory, The University of Queensland) and U. Siebeck (Visual Neuroethology Lab, the University of Queensland) assisted with suggestions while writing this thesis. E. Scott (Scott Laboratory, The University of Queensland) and M. Piber (School of Medicine, Medical Sciences & Nutrition, University of Aberdeen) contributed to conceptualise, interpret and draft chapter 2. E. Scott (Scott Laboratory, The University of Queensland) and G. Vanwalleghem (Scott Laboratory, The University of Queensland) assisted designing and analysing the experiments for chapter 3 and chapter 4. G. Vanwalleghem (Scott Laboratory, The University of Queensland), assisted conceptualising and implementing the bioinformatics pre-processing and analysis of the data for chapters 3 and 4. I. Favre-Bulle (Scott Laboratory and Optical micromanipulation group, The University of Queensland) and M. Taylor (Scott Laboratory, The University of Queensland) build the light sheet microscope used for experiments in chapters 3 and 4. L. Heap (Van Swinderen Lab, The University of Queensland) assisted maintaining the Gcamp6s zebrafish line and building the set up for experiments in chapter 3 and 4. L. Constantin (Scott Laboratory, The University of Queensland) assisted establishing the Gcamp6s line used in chapters 3 and 4 and also establishing, maintaining and providing the fmr1 line and genotyping the fish for experiment 4. M. Piber (School of Medicine, Medical Sciences & Nutrition, University of Aberdeen) assisted performing and analysing the behavioural experiments in chapter 4. U. Siebeck (Visual Neuroethology Lab, the University of Queensland) assisted designing the experiments for chapter 5. H. Middleton (Sensory Neurobiology Group, the University of Queensland) assisted slicing brains from experiments detailed in chapter 5. D. Whitehead (SBMS Histology facilities, the University of Queensland) assisted designing histological protocols for the experiments in chapter 5. H. Middleton (Sensory Neurobiology Group, the University of Queensland) contributed by performing part of the visual discrimination and learning behavioural experiments, and with the logistics of the field trip to Heron Island in chapter 5. D. Bender (Coral Reef Ecosystems

Lab, The University of Queensland) designed the CO2 aquarium system and analysed the water chemistry for experiments in chapter 5. A. N. Parker (Visual Neuroethology Lab, the University of Queensland) and C. Rosinski (Biology Department, Stonehill College) assisted building the aquarium system and performing behavioural experiments for experiments in chapter 5. S. Blomberg (School of Biological Sciences, The University of Queensland), assisted designing the statistical analysis for chapter 5.

Statement of parts of the thesis submitted to qualify for the award of another degree

No works submitted towards another degree have been included in this thesis.

Research Involving Human or Animal Subjects

For zebrafish work: SBMS/378/16. AEC Committee Molecular Biosciences.

For Ambon damselfish work: SBMS/163/15/ARC. AEC Committee Native and exotic wildlife and marine animals.

VII

Acknowledgements

I would like to first thank my supervisors. Thank you Ethan for accepting me in your lab, for giving me the opportunity to work in such a fascinating field and with such a wonderful group. It was a challenging task as everything was new to me, but your effective guidance was key to make this project successful. I have learn a lot from you and I am very thankful for all your support. Uli, thank you for giving me the opportunity to come to Australia and introducing me to the marine and animal behaviour worlds, I loved them. You gave me a freedom I had never experienced before, which was daunting sometimes but that helped me to build independence. More importantly, you taught me to be brave and daring in science and to challenge the general assumptions. I really appreciate that. Gilles, thank you for your never-ending help and for all your advice, whether it was related to my project or to life as a scientist. Your enthusiasm, curiosity and humour made this journey much lighter.

I would also like to thank all the members from both labs. Some of our interactions were short but they were always very pleasant. Thank you Amira for all your help when I just arrived and for all your advice. Thank you Heather for helping me so much too. The field trips with you both were awesome and always fun. Itia and Michael, the light wizards, without you I just wouldn’t have been able to do this project. One of the reasons I wanted to do neuroscience was for its multidisciplinary nature. With you I have learn the true meaning of that, and how powerful it is. Thank you Lena for all your help, without you this lab would fall into pieces; we all owe you so much. Lucy and Andrew, thank you for placing the foundations of my project, your legacy in the lab has been very important. Thank you Marielle for enduring working with me, it was always a lot of fun. I tried to be a good guide and I learned a lot about it thanks to you. Thank you Rebecca, Sebastian, Tessa, Harry, Cory, Jaime and Neha, for all the good times.

Multiple people helped me throughout my PhD in various ways, I won’t be able to name them all but among them I would like to thank Dr Jean Giacomotto, Dr Mike Piper and his lab, Dr Simone Blomberg, Dr Darryl Whitehead, Dr Shaun Walters and all the staff from the SBMS histology and imaging facilities and the amazing staff from the Heron island and Lizard island research stations.

I would like to thank CONACYT and UQ for supporting me financially. I hope that someday I will be able to give back to Mexico what it was given to me.

I would also like to thank all my friends who helped to keep my sanity and enjoy all this years. Hugo, Nur, Barbara, Fabian, Lars, Cedric, Andrew, Matt, Laura, Carrie, the volleyball club, the soccer group, and many others. Gracias en especial a ti Marta, fue increíble, te estaré por siempre agradecido.

VIII

Thank you to all my friends in Mexico, Spain, Netherlands, Switzerland, Germany and anywhere you are, you kept entertaining me despite the exile I got into.

Finally, to my family, infinitamente gracias! Sin su apoyo y cariño incondicional jamás hubiera llegado tan lejos. Gracias por formar está loca y extraña familia tan especial, eternamente curiosa y viajera pero siempre unida a pesar de la distancia. Lamento estar siempre tan lejos… ha sido un duro precio a pagar por estas aventuras. Gracias Herma, por todo tu apoyo. Gracias papá y mamá, el ejemplo que me dieron con su valor y sabiduría, haciendo de la ciencia una forma de vida y el mundo entero un lugar por descubrir, es lo que siempre me ha mantenido en el camino. No saben cuan orgulloso estoy de ustedes, como padres, como personas y como científicos. Gracias.

Financial support

– Scholarship for a PhD abroad (Doctorate degree at The University of Queensland, Australia). Granted by the National Council for Science and Technology of Mexico (CONACyT). From 01/10/2015 to 30/09/2019.

This research was also supported by an Australian Government Research Training Program Scholarship:

– Top-Up Assistance Plan. Tuition fee award. Granted by The University of Queensland, Australia. From 01/10/2015 to 30/09/2019.

– University of Queensland Research Training Tuition Fee Offset. Granted by The University of Queensland, Australia. From 01/10/2015 to 30/09/2019

– Terry Walker Award. Australian Coral Reef Society, Australia, 2017.

– Heron Island Research Station Scholarship. The University of Queensland, Australia. 09/2016

– Moreton Bay Research Station Scholarship. The University of Queensland, Australia. 09/2016

– ANS support grant, 2019

– UQ HDR International travel scholarship, 2019.

– UQ Career Development Award (CDA), 2019.

– UQ HDR domestic travel scholarship, 2018.

Keywords visual learning, zebrafish larvae, habituation, neuroanatomy, calcium imaging, microscopy, optic tectum, pallium, ambon damselfish, visual discrimination

Australian and New Zealand Standard Research Classifications (ANZSRC)

ANZSRC code: 110999, Neurosciences not elsewhere classified, 50%

ANZSRC code: 170299 Cognitive Sciences not elsewhere classified, 25%

ANZSRC code: 060805 Animal Neurobiology, 25%

Fields of Research (FoR) Classification

FoR code: 1109, Neurosciences, 50%

FoR code: 1702, Cognitive Sciences, 20%

FoR code: 0608, Zoology, 25%

Table of Contents

List of Figures ...... XV

List of Tables ...... XVI

List of Abbreviations used in the thesis ...... XVI

1 Introduction...... 1

1.1 Teleost as models to study visual learning...... 1

1.1.1 Preamble...... 1

1.1.2 Comparing fish and mammals visual system...... 2

1.1.3 Learning and visual processing in fish brains ...... 5

1.2 Part 2. Fish in visual neuroscience ...... 10

1.2.1 Zebrafish larvae-optics and genetics ...... 10

1.2.2 Fish visual discrimination and Ambon damselfish ...... 11

1.3 Part 3. Combining models to study visual learning ...... 14

2 Visual escape in larval zebrafish: Stimuli, Circuits, and Behaviour ...... 17

2.1 Contributions ...... 17

2.2 Summary ...... 17

2.3 Introduction ...... 18

2.4 The essential properties of loom stimuli (What causes an animal to escape?) ...... 19

2.5 Execution of escape behaviours ...... 24

2.6 Core visual escape circuitry ...... 26

2.7 Modulation of the core escape circuit ...... 32

2.8 Open questions about the loom circuit ...... 35

3 Habituation to looming stimuli in zebrafish larvae ...... 39

3.1 Contributions ...... 39

3.2 Summary ...... 39

3.3 Introduction ...... 40

3.4 Preliminary results: ...... 43

XII

3.4.1 Visual habituation in larval zebrafish ...... 43

3.4.2 Locate visual habituating neuronal responses...... 46

3.5 Methods ...... 50

3.6 Results ...... 56

3.6.1 Habituation of visual escape behaviour in larval zebrafish ...... 56

3.6.2 Brain-wide characterisation of neural activity during habituation ...... 58

3.6.3 Temporal stimulus properties modulate brain-wide responses ...... 67

3.6.4 Network Modelling of Visual Loom Habituation ...... 70

3.7 Discussion ...... 74

3.7.1 Populations subtypes of visual habituating neurons ...... 76

3.7.2 A brain-wide network model of visual habituation ...... 80

3.7.3 Future directions ...... 82

3.7.4 Conclusions ...... 84

4 Visual habituation in fmr1-mutant zebrafish larvae ...... 85

4.1 Contributions ...... 85

4.2 Summary ...... 85

4.3 Introduction ...... 85

4.4 Methods ...... 88

4.5 Results ...... 91

4.5.1 fmr1 mutants present behavioural habituation differences...... 91

4.5.2 Neuronal loom responses across genotypes are similar...... 92

4.5.3 Brain distribution of responding neurons ...... 99

4.5.4 fmr1-/- mutant larvae show network-level habituation deficits ...... 102

4.6 Discussion ...... 110

5 Visual discrimination learning in Ambon damselfish: Mapping the dorsal pallium, testing pallial learning and disrupting visual and learning abilities...... 115

5.1 Contributions ...... 115

5.2 Summary ...... 115 XIII

5.3 Introduction ...... 116

5.4 Methods ...... 120

5.5 Results ...... 127

5.5.1 Neuroanatomy of the Ambon damselfish pallium ...... 127

5.5.2 Pallial visual learning in Ambon damselfish ...... 130

5.5.3 Disruption of visual discrimination and learning ...... 131

5.6 Discussion ...... 135

5.6.1 Pallium neuroanatomy and its involvement in visual processing ...... 135

5.6.2 Visual discrimination and learning are not affected by lower pH ...... 136

5.6.3 Conclusions ...... 139

6 Discussion ...... 141

6.1 General discussion ...... 141

6.2 Whole-brain visual habituation ...... 141

6.3 Visual processing of threat and learning in the teleost pallium ...... 145

6.4 Conclusions ...... 150

Bibliography ...... 151

Appendices ...... 184

Appendix 1: Ethics approvals...... 184

Appendix 2: Manuscript- Visual Escape in Larval Zebrafish: Stimuli, Circuits, and Behavior. . 188

XIV

List of Figures

Figure 1.1. The human visual system...... 4 Figure 1.2. Ascending visual pathways in rodents and fish...... 5 Figure 1.3. Visual telencephalic connections...... 7 Figure 1.4. Development of the telencephalon in rodents and cyprinids...... 9 Figure 1.5. Ambon damselfish and zebrafish as animal models...... 16 Figure 2.1. Components and computations of the loom...... 24 Figure 2.2. Loom relevant information arriving to the tectal neuropil...... 30 Figure 2.3. The core visual escape circuit...... 32 Figure 2.4. Modulatory structures of the core visual escape circuit...... 35 Figure 2.5. The confluence of retinal, thalamic, and modulatory information in the tectal neuropil.38 Figure 3.1. Preliminary behavioural habituation experiments with zebrafish larvae...... 45 Figure 3.2. Calcium imaging and behavioural responses to looms...... 48 Figure 3.3. Location of loom responsive and habituating ROIs ...... 49 Figure 3.4. Calcium traces of ROIs responding to loom presentation and sound...... 50 Figure 3.5. Modulation of habituation by stimulus features...... 57 Figure 3.6. The activity of individual ROIs and their functional clusters during habituation...... 59 Figure 3.7. Similarities among three types of fast habituating ROI...... 62 Figure 3.8. t-SNE analysis of the five functional clusters...... 63 Figure 3.9. Anatomical distributions of five habituating clusters...... 65 Figure 3.10. Prominence of motor-correlated strongly habituating ROIs in the hindbrain...... 66 Figure 3.11. Sound and multisensory responsive ROIs...... 67 Figure 3.12. Brain-wide responses during different loom stimulus trains...... 69 Figure 3.13. The visual loom network, and the changes that occur during habituation...... 74 Figure 4.1. fmr1 mutant zebrafish larvae present altered behaviour...... 92 Figure 4.2. The subpopulations of loom-habituating neurons...... 95 Figure 4.3. Similar responses per genotype...... 96 Figure 4.4. The number of loom-responding ROIs across genotypes...... 97 Figure 4.5. Distribution of loom responses...... 98 Figure 4.6. Similar distribution to S20 clusters...... 100 Figure 4.7. Brain distribution per genotype...... 101 Figure 4.8. ROI distribution in relevant brain areas per genotype...... 102 Figure 4.9. Network-wide changes in fmr1-/- larvae...... 106

Figure 4.10. Functionally sorted brain-wide networks for WT and fmr1-/- larvae...... 108 Figure 4.11. Spatially sorted brain-wide networks for WT and fmr1-/- larvae...... 110

Figure 5.1. Project design for visual testing CO2 effects...... 124 Figure 5.2. Nissl, NADPH-d and Parvalbumin staining of adult Ambon damselfish brains...... 128 Figure 5.3. HIRS behavioural experiment...... 130 Figure 5.4. c-Fos in Ambon damselfish pallium...... 130 Figure 5.5. Learning discrimination task of experiment 1...... 131

Figure 5.6. Experiment 1- effect of high CO2 on visual discrimination...... 134

Figure 5.7. Experiment 2- Learning under high CO2 conditions...... 134 Figure 6.1. Evoked responses in the telencephalon of larval zebrafish...... 147

List of Tables

Table 5.1. Water chemistry and calculated pCO2 ...... 125

List of Abbreviations used in the thesis

AFC: alternative forced-choice paradigm

ADHD: attention -deficit/hyperactivity disorder

AFs: Arborisation fields

AMPA: α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid

ANOVA: analysis of variance

ANTs: advanced normalization tools

ASD: autism spectrum disorder

CART: cocaine-and amphetamine-regulated transcript

Cb: cerebellum c-fos: Fos proto-oncogene c-Fos: protein encoded by the c-fos gene

CNMF: non-negative matrix factorisation XVI

Dc: central division of the dorsal pallium

Dcd: dorsal division of Dc

Dd: dorsal division of the dorsal pallium

DF: degrees of freedom

Dl: lateral division of the dorsal pallium

Dld: dorsal part of the Dl

Dm: medial division of the dorsal pallium dpf: days post-fertilization

DTI: diffusion tensor imaging

EmT: eminentia thalami f20: stimulus train with fast looms and 20 seconds ISI f60: stimulus train with fast looms and 60 seconds ISI fmr1/Fmr1/FMR1: fragile X mental retardation 1 fmr1-/-: fmr1 mutants with a nonsense-mediated-decay and the loss of the protein. fMRI: functional magnetic resonance imaging

FMRP: fragile X mental retardation protein

FXS: Fragile x syndrome

GABA: gamma-Aminobutyric acid

GCaMP: calcium indicator formed by GFP, calmodulin and M13 peptide

GCaMP6s: version 6 “slow” form the GCaMP genetically encoded calcium indicator

GFP: green fluorescent protein

GLMM: generalized linear mixed model

HB: hindbrain

Hb: habenula

Hets: heterozygotes XVII

HIRS: Heron Island research station

HPI: hypothalamic-pituitary-interrenal

IR: infrared

ISI: inter stimulus interval

LGN: lateral geniculate nucleus

LIRS: Lizard Island research station

LP: Lateroposterior nucleus of the thalamus

LTD: increased long-term depression

LTP: long-term potentiation

χ2: likelihood ratio chi-square test

M2: migrated posterior tubercular area

MDNs: movement detector neurons

M-cells: Mauthner cells

M-system: reticulospinal neurons that include the Mauthner cells and their segmental homologues

NADPH-d: nicotinamide adenine dinucleotide phosphate diaphorase

NMDA: N-methyl-D-aspartate

NI: nucleus isthmi nMLF: nucleus of the medial longitudinal fasciculus

OCT: optimal cutting temperature

OT: optic tectum

P1: period 1

P2: period 2

Pal: pallium

PBS: phosphate-buffered saline pCO2: partial pressure of carbon dioxide XVIII

PFC: prefrontal cortex

PG: preglomerular nucleus pHNBS: pH on the National Bureau of Standards/ National Institute of Standards and Technology scale ppm: parts per million

Pt:pretectum

Pulv: pulvinar

PV: parvalbumin

PVIN: periventricular interneurons

PVL: periventricular layer

PVN: periventricular neurons

PVPN: periventricular projection neurons

RGC: retinal ganglion cell

ROIs: regions of interest

S+: positive stimulus

S-: negative stimulus s20: stimulus train with slow looms and 20 seconds ISI s60: stimulus train with slow looms and 60 seconds ISI

SAC: stratum album centrale

SBMS: school of biomedical sciences

SC: superior colliculus

SD: standard deviation

SEM: standard error of the mean

SFGS: stratum fibrosum et griseum superficiale

SGC: stratum griseum centrale

SIN: superficial inhibitory neuron XIX

SO: stratum opticum

Sp: subpallium

SPIM: selective plane illumination microscope

Sy: sulcus ypsiloniformis

TeO: tectum opticum

Tec: tectum

Tg: tegmentum

TIFF: tagged image file format

Th: thalamus t-SNE: t-Distributed Stochastic Neighbour Embedding

V1: Primary visual cortex

V2: Secondary visual cortex

VPA: valproic acid

WT: wild type

1 Introduction

1.1 Teleost as models to study visual learning

1.1.1 Preamble

The abilities to detect and discriminate relevant visual stimuli, as well as to learn about them, are fundamental to humans and most animals. The vast information flow coming in from the environment needs to be sorted, analysed and interpreted so that an appropriate response can be determined. Furthermore, the experience of each organism is shaped through learning, increasing the likelihood that an appropriate response is taken, and modifications are implemented as needed. There is a strong ecological compulsion to get these decisions correct as errors may prove costly or even deadly (e.g. poisoning and predation). The interest in the visual system has led to a considerable amount of research in the human and animal visual nervous system, but how experience and learning modify visual perception has focused mostly in cortical studies, the results of which are difficult to generalise to other vertebrates.

Research in humans and other mammals has focussed on the role of the cortex and its local or network plasticity to explain visual learning (Grady et al., 1998; Kourtzi, 2010; Sagi, 2011; Cooke and Bear, 2014; Poort et al., 2015; Dosher and Lu, 2017; Frangou et al., 2019). This tendency is because primates (including humans) dedicate a significant proportion of their cortex to visual processing (Van Essen et al., 1992). Visual learning improves humans and other mammals abilities of visual discrimination and categorisation, processes that are considered to be performed mainly in cortical structures (Freedman and Miller, 2008; Seger and Miller, 2010; Jurjut et al., 2017; Zatka-Haas et al., 2019). However, fish are capable of complex visual discriminations but do not have a cortical telencephalic structure, and its possible homologues are still in debate (Northcutt, 2011; Salva et al., 2014). How fish learn and solve these tasks, or in particular, which structures of the brain are involved in this process, is something that is not yet understood. This information would be relevant not only for a better understanding of those animals’ nervous system but also to improve our interpretation of the subcortical pathway’s role in cognitive processes.

Although innate visual behaviours and general neuroanatomy of visual areas (especially the optic tectum and retina) have been extensively studied, not much was known about the function of brain areas related to fish visual processing until recently (Salva et al., 2014). This lack of interest was partly because, traditionally, the fish brain has been regarded as an early step in the evolution along an increasingly complex linear design, and that at this point the main structures needed for complex 1

tasks had not yet evolved. This idea has now been rejected and the thought that vertebrates share the same fundamental structures, although in simpler forms, is growing stronger (Jarvis et al., 2005; Salas et al., 2006; Liebeskind et al., 2016).

As the previous ideas were changing, fish started to be considered a good model for visual perception research because of their great taxonomic diversity, the presence of visually guided behaviours and because some of them can be trained to solve visual discrimination tasks (Siebeck et al., 2008; Van- Eyk et al., 2011; Newport et al., 2013; Champ et al., 2014; Salva et al., 2014). According to Gerlai (Gerlai, 2014), one of the main reasons to use fish as a model for behavioural neuroscience is that it allows the comparison of the findings with other animals, distant from an evolutionary point of view like birds or mammals. By studying different animal models, the essential mechanisms of biological phenomena can be better depicted, and common conclusions about the nervous system can be drawn, which increases their translational relevance. The second reason is that “older phylogenetic designs” can lead us to the core of the mechanisms. Despite the apparent simplicity of a fish’s neuroanatomy, it still possesses most of the neuroanatomical, neurophysiological and behavioural fundamental features of mammals. These fundamental features can, therefore, be investigated accessibly, as the “noise” of the complexity of mammalian brains is avoided. In this way, fish models can enrich the research of the homology and the evolution of the nervous systems which are still far from being understood (Liebeskind et al., 2016; Strausfeld and Hirth, 2016). Additionally, the use of laboratory animals such as the zebrafish, whose genome is known and can be manipulated, is key to a better understanding of how evolution shaped behaviour and the nervous system (Grant, 2016). All these features make fish an exciting model for visual processing, but in order to draw accurate conclusions and generalisations, we need to understand how their visual nervous system is comparable to the mammalian one and which parts of it could be involved in visual learning.

1.1.2 Comparing fish and mammals visual system.

The visual system of vertebrates is highly preserved across different species where most fundamental structures are shared (Sanes and Zipursky, 2010). Vertebrates that rely on vision transform the energy of light into neuronal action potentials at the retina, where the opsins of the photoreceptors produce a chain of chemical reactions that will lead to activity of retinal ganglion cells. Complex sorting of the information already happens at this stage where horizontal and amacrine neurons modulate the information that reaches the ganglion cells before they send the visual information to the brain (Sanes and Zipursky, 2010). At this level, the main difference between the mammalian and teleost retina lies

in the types of photoreceptors. For example, some fish can have up to five spectral sensitivities (Losey et al., 1999; Siebeck and Marshall, 2001; Nava et al., 2011) and all fish present double cones, a morphological receptor type absent in mammals (Marchiafava, 1985; Fang et al., 2005).

In both fish and mammals, most of the information from each eye crosses to the contralateral hemisphere (Larsson, 2015) and enters the brain, but the main target of these retinal projections for mammals and fish are different. In mammals, most of the ganglion axons will reach the thalamus (mainly targeting the lateral geniculate nucleus, LGN) which process further the visual information and serves as relay nucleus that sends projections to various parts of the isocortex. The main target of visual thalamic projections is the primary visual cortex (V1) in the occipital lobe. Then, corticocortical connexions will further process the information. This pathway is known as the geniculate or lemniscal pathway, and it is believed to process complex and detailed visual information (for example, reading or fine visual discrimination) (Figure 1.1). In contrast, a second minor pathway goes directly from the retina to the superior colliculus in the mesencephalon, where relevant processing of the visual information is done. The superior colliculus sends ascendant projections, most of which target the suprageniculate and Lateroposterior/Pulvinar (rodents and primates respectively) nucleus in the thalamus, which in turn projects to the secondary visual cortex (Huerta and Harting, 1984) (Figure 1.1 and Figure 1.2). The precise roles of both pathways are not yet fully uncovered, but it is thought that the geniculate pathway is related to more complex and detailed visual analysis. In contrast, the extra-geniculate pathway is considered to be involved in the automatic processing of moving objects, spatial localization and multisensory information that needs a quick response (Vievard et al., 1986; Rafal et al., 1991; Ro et al., 2004; Paroczy et al., 2006; Nagy et al., 2008; Rokszin et al., 2010; Zhang et al., 2015).

The lemniscal or geniculate pathway is the one considered to drive the information necessary for visual discrimination. The primary visual cortex and the inferotemporal cortices are needed to make the analysis of the stimuli, then the prefrontal cortex seems to be in charge of the categorization (Miller et al., 2002; Freedman et al., 2003; Jiang et al., 2007; Freedman and Miller, 2008; Freeman et al., 2010) (for a review of the areas related to categorization, see Seger and Miller (2010)). Other non-cortical structures such as the hippocampus and the basal ganglia seem to be related to the learning component through corticostriatal loops (Seger, 2009). The studies done with neural imaging in non-human primates show that although many cortical areas are related to visual processing (occipital, inferior temporal and middle temporal cortices), it is in the prefrontal cortex (PFC) where the categorisation takes place. PFC neurons seem to be involved in discriminating stimuli according to different learned abstract rules not bounded to specific features, and actually adapt to changing

boundaries (Freedman and Miller, 2008). The key role of the PFC in visual categorisation makes sense as the connections of the PFC with sensory, motor, emotional and motivational areas make it a singular place for this kind of processing (Freedman and Miller, 2008).

Figure 1.1. The human visual system.

In humans and other primates, most of the retinal projections reach the lateral geniculate nucleus in the thalamus, which then sends the visual information to the primary visual cortex (V1). Further cortical projections target different areas that will continue to process the visual information. A smaller portion of retinal axons targets the superior colliculus (SC) before reaching the thalamus. Image adapted from Tamietto and de Gelder (2010).

In contrast, the fish ascending visual pathways are different, as a distinct thalamic relay nucleus homologous to the lateral geniculate body has not been found. Recent results suggest that a homologue of the geniculate pathway was shared in early vertebrates, but in some species of fish, it lost relevance or even disappeared (Hagio et al., 2018). In teleost, most of the visual information coming from the retina is processed in the optic tectum, passing then to a relay nucleus (preglomerular nucleus in zebrafish) and finally reaching the telencephalon (Figure 1.2), where it is not yet clear if a similar structure to the isocortex exists. It seems, therefore, that the fish pathway follows the

mammalian extra-geniculate layout. From these differences in the anatomical structures and connections, an important question arises: How and where are visual learning and discrimination processed in fish, given their lack of brain structures thought to be responsible for these abilities in mammals?

Figure 1.2. Ascending visual pathways in rodents and fish.

Top. The geniculate or lemniscal pathway in mammals has the thalamus as a central relay station between the retina and the visual cortex (red). Bottom. Instead, the main visual pathway in fish has a first stop in the tectum opticum (TeO), similar to the mammalian extralemniscal pathway (orange). Then the visual information arrives mainly to the preglomerular nucleus, a diencephalic relay structure, to then continue to the telencephalon (purple). Other ascending pathways seem to exist but are not yet well-understood (orange dashed line). Image adapted from Mueller (2012).

1.1.3 Learning and visual processing in fish brains

As mentioned above, the main ascending visual pathway to the pallium in teleosts differs from the lemniscal pathway in mammals (Figure 1.2). The retina of teleosts projects to different diencephalic

and mesencephalic nuclei which both in turn have ascending and descending projections. However, in contrast to mammals, the main fish visual pathway to the pallium makes the first relay in the superficial layers of the optic tectum (equivalent to the superior colliculus of mammals) before reaching the diencephalon (Vanegas and Ito, 1983; Northcutt and Wullimann, 1988; Wullimann, 1997a; Northmore, 2011) and finally the telencephalon. Those tectal retinorecipient layers have ascending projections to two main structures, a dorsal thalamic structure (Braford, 1995; Butler, 2011; Mueller, 2012) and a diencephalic structure; the latter being the principal relay nucleus that projects visual information to the pallium in teleost (Vanegas and Ito, 1983; Braford, 2009; Mueller, 2012). It is likely that the information sent through this pathway is relevant for learning spatial information or recognition of objects (Northmore, 2011). The nomenclature of this diencephalic structure is variable between species, but due to their connections, it seems to be the same preserved structure that has been named in different ways. Some of those names are the prethalamic nucleus (Vanegas and Ito, 1983), nucleus rotundus (Ebbesson and Vanegas, 1976) and preglomerular complex (Demski, 2003; Northcutt, 2006, 2008; Yamamoto and Ito, 2008; Mueller, 2012; Demski, 2013). This relay station has bidirectional connections with the telencephalon, sending projections to three of the pallium subdivisions, the lateral (Dl), medial (Dm) and central (Dc) divisions of the dorsal pallium, with the highest projection density sent to the dorsal part of the Dl (Dld) (Vanegas and Ito, 1983; Demski, 2003; Northcutt, 2006; Yamamoto and Ito, 2008) (Figure 1.3). Those subdivisions interact highly with each other, especially with the Dc, which receives projections in its dorsal part (Dcd) from the visual area of the Dld (Demski, 2013). Finally, the visual information seems to make a loop as the Dc projects downwards to different nuclei, including the visual layers of the optic tectum (Luiten, 1981; Vanegas and Ito, 1983; Demski, 2003; Northcutt, 2006; Northmore, 2011; Demski, 2013). This last projection suggests a regulation from the pallium to the visual optic tectum (Figure 1.3) that could modulate visual perception based on previous experiences. Apart from hodological studies, evidence supporting that these areas process visual information is scarce, but a couple of studies found changes in neuronal activity after eye enucleation and electrophysiological responses to visual stimuli in both the Dl and Dc (Prechtl et al., 1998; Saidel et al., 2001).

Figure 1.3. Visual telencephalic connections. a. Lateral view of a green sunfish brain with a schematic representation of the visual system (adapted from previous illustrations made by Wullimann, Northcutt and Northmore (Wullimann and Northcutt, 1988; Northmore, 2011)). Green arrows represent retinal projections. Dark blue represent tectal projections. Light blue represents thalamic and pretectal efferents. Magenta represents pallial efferents. b. Schematic representations of coronal sections of the teleost pallium. The left drawing represents the afferents of different sensory modalities. The drawing in the middle represents efferent projections from the subdivisions of the dorsal pallium. The right drawing is an example of the visual circuits in the telencephalon. Adapted from Ito and Yamamoto (2009).

Learning behaviours have been related to three main structures in the fish nervous system: the cerebellum, the Dm and the Dl (Broglio et al., 2005; Hurtado-Parrado, 2010; Braithwaite, 2011; Demski, 2013). The cerebellum is implicated in classical conditioning learning. For example, experiments have conditioned eye retraction after the association of light and a mild shock in goldfish (Rodriguez et al., 2005; Gomez et al., 2010). Emotional learning, mostly studied with experimental

designs using aversive stimuli, has been related to the Dm. Studies with drug-seeking approaches, light avoidance behaviour experiments and taste aversion are some of the examples supporting emotional learning (Portavella et al., 2002; Broglio et al., 2005; Salas et al., 2006; Martin et al., 2011; von Trotha et al., 2014; Perathoner et al., 2016). Finally, lesions or drug induction studies and experiments with activity markers have shown that the Dl pallium is used by fish for similar tasks as the mammalian hippocampus, i.e. spatial learning (Rodriguez et al., 2002; Broglio et al., 2003; Gomez et al., 2006; Vargas et al., 2006; Broglio et al., 2015; Uceda et al., 2015) and individual recognition (Harvey-Girard et al., 2010). These data suggest that the subdivisions of the dorsal pallium have functional homology to those of some telencephalic mammalian structures. The Dm is now considered comparable to the amygdala and the Dl to the hippocampus (Broglio et al., 2011; Butler, 2011; Mueller et al., 2011; Mueller, 2012; Yamamoto et al., 2017). The homological functions of the Dc and the dorsal division of the dorsal pallium (Dd) are still under debate, but some data suggest that they could be similar to the mammalian isocortex (Vargas et al., 2009; Mueller et al., 2011; Northcutt, 2011; Harvey-Girard et al., 2012; Mueller, 2012). Although there is still controversy on how precisely these nuclei are aligned, it is now believed that their arrangement is the result of eversion of the telencephalon during development (Figure 1.4). The eversion leads to a different anatomical organisation compared to mammals, which has hidden their homology until recently (Yamamoto et al., 2007; Broglio et al., 2011; Butler, 2011; Mueller et al., 2011).

Figure 1.4. Development of the telencephalon in rodents and cyprinids.

Coronal sections show how, during development, an eversion in the teleost brain produces a different arrangement of the structures compared to mammals. As a result of this, the topology of telencephalic domains is inverted. The rodent hippocampus comes from the medial pallium (MP, number 6) and it is located medially in the brain. Instead, the fish MP ends up in a lateral position (Dl). On the other hand, the amygdala, originated from the ventral pallium (VP, number 4) moves to a dorsomedial region (Dm) in teleost. Finally, the mammalian isocortex develops from the dorsal pallium (DP, number 5); however, it is still under debate if an equivalent structure exists in teleost. Taken from Perathoner et al. (2016).

Taking the previous information about visual and learning structures into account, when doing functional experiments, we would expect neuronal activity in some specific regions of the fish telencephalon while performing visual related and learning tasks (Yamamoto, 2009; Demski, 2013). The flow of information would be retina-OT-PG/Th (relay structure)-Dl (pallium intercommunications)-Dc (Figure 1.3). A similar ascending pathway has been proposed for fish electrical communication and learning (Harvey-Girard et al., 2010; Giassi et al., 2012b; Giassi et al., 2012a). It is therefore probable that we could see indicators of neuronal activity in the Dld and Dcd 9

when performing a visual task, specific activity in the Dl when learning a visual related task and emotional related visual activity (seeing a predator, prey or a possible mate) could be present in the Dm.

1.2 Part 2. Fish in visual neuroscience

1.2.1 Zebrafish larvae-optics and genetics

So far, I have focused on the questions arising around fish visual perception and learning and the neuroanatomy related to these behaviours. Many of the new research tools used in neuroscience have been developed for classical laboratory animal models (e.g. rats, mice, zebrafish, fruit fly, roundworm). Among these animals, zebrafish has become one of the most used models in neuroscience. As this species has several technical, neurobiological and behavioural advantages, it has become a fantastic research tool (Blaser and Heyser, 2015). In particular, the fact that its genome is known, and it is an easy and quick species to breed, makes it a powerful tool where mutant fish can be used for highly specific research (Scott and Baier, 2009; Stewart and Kalueff, 2012; Wolman and Granato, 2012; Feierstein et al., 2015; Randlett et al., 2015). Also, being a highly visual animal, it is well suited for the study of the visual system (Bilotta and Saszik, 2001; Neuhauss, 2003). Moreover, because of the exponential growth in zebrafish research, its neuroanatomy is much better known compared to any other fish species (Wullimann et al., 1996; Arrenberg and Driever, 2013; Randlett et al., 2015; Mueller and Wullimann, 2016). Additionally, state of the art technology is being developed to exploit the advantages of zebrafish larvae as a model for neuroscience. Recently, the methodological foundations for a live view of brain activity in larval zebrafish have been developed, allowing researchers to see individual neuronal activity responding to a wide range of sensory stimulation, including visual stimuli. In this paradigm, a genetically modified zebrafish is made transparent and expresses fluorescent proteins sensitive to calcium entering the neuron (with neuronal action potentials). As these larvae are small and transparent, a microscope can be used to detect and record the neuronal activity across the whole brain. This activity can be then analysed and localised to find out where and in which fashion stimuli are processed (Muto et al., 2013; Renninger and Orger, 2013; Feierstein et al., 2015; Thompson et al., 2016).

But how capable are zebrafish of visual learning tasks? As some experts suggest, the main issue with the zebrafish model is the behavioural limitations, not because the zebrafish are not capable of a wide range of behaviours, but because the scientific community is just starting to learn about them (Sison et al., 2006; Sison and Gerlai, 2010; Kalueff et al., 2013; Gerlai, 2014). Visual research in zebrafish

has usually focused on innate behaviours (Neuhauss, 2003; Orger et al., 2004). Recently more studies have been approaching visual discrimination with adult zebrafish, using abstract or biologically relevant stimuli, showing that they can discriminate colour, shapes, animated fish, predators and conspecifics among other things (Colwill et al., 2005; Bass and Gerlai, 2008; Saverino and Gerlai, 2008; Gerlai, 2010; Ahmed et al., 2011; Ahmed et al., 2012; Luca and Gerlai, 2012; Mueller and Neuhauss, 2012; Oliveira et al., 2015). Furthermore, adult zebrafish learning behaviour has started to be a growing research field (Sison and Gerlai, 2010; Blaser and Vira, 2014; Bailey et al., 2015; Gerlai, 2016).

However, although adult zebrafish presents multiple advantages and are capable of multiple complex behaviours, they do not provide the optical or high-throughput benefits of that larvae do. Using larval zebrafish, many studies have investigated behavioural reactions to various visual stimuli and the neuronal activity of escape responses, visual acuity, movement detection or colour discrimination (Rinner et al., 2005; Haug et al., 2010; Nikolaou et al., 2012; Ahmad and Richardson, 2013; Preuss et al., 2014; Barker and Baier, 2015; Temizer et al., 2015; Bogli et al., 2016; Filosa et al., 2016; Li et al., 2016; Peeters et al., 2016). Learning in larval zebrafish has also started to be investigated, especially with habituation paradigms (Best et al., 2008; Sumbre et al., 2008; Roberts et al., 2011; Wolman et al., 2011; Roberts et al., 2013; O'Neale et al., 2014; Riley et al., 2015; Roberts et al., 2016), as associative learning experimental data has produced conflicting results (Aizenberg and Schuman, 2011; Valente et al., 2012; Hinz et al., 2013). Yet, visual habituation responses in the brain have not been explored. By following visual habituation paradigms, it would be possible to track down the activity within the fish brain during a simple form of visual learning. Finding which areas of the brain zebrafish larvae use for visual habituation would open new possibilities in basic and clinical research. It would be then possible to use mutant zebrafish larvae as genetic models of neurological diseases where learning or visual discrimination is impaired. The possible physiological reasons for these problems could be then investigated to understand how the nervous system is affected by these problems. (Guo, 2004; Homer and Rutherford, 2008; Stewart et al., 2010; Stewart and Kalueff, 2012).

1.2.2 Fish visual discrimination and Ambon damselfish

The many benefits associated with Zebrafish larvae make it an incomparable model for vertebrate nervous system research. However, its learning abilities are limited. Therefore, other models are needed to further understand more complex cognitive visual abilities, like visual discrimination. The

processes underlying visual discrimination both in terms of its performance and improvement have been of great interest to many investigators studying humans (Sowden et al., 2000; Jiang et al., 2006; Becker et al., 2011; Burgess, 2011; Rossion, 2014). Also, visual discrimination relevance as a tool for the research of complex forms of associative learning has been proven useful for studying cognitive behaviours. Because of this, visual discrimination can be used to compare healthy individuals with others suffering from neurological and psychiatric diseases (Franklin et al., 2008; Homer and Rutherford, 2008; Samson et al., 2012; Gori and Facoetti, 2014; Loffler, 2015; Silvanto, 2015; Beilharz et al., 2016). This clinical relevance has motivated its use in translational research between animals and humans (Morton et al., 2006; Keeler and Robbins, 2011; Nithianantharajah and Grant, 2013; Nithianantharajah et al., 2013; Bonaccorsi et al., 2014; Silverman et al., 2015). This translational research can highly benefit from looking at other animals, including fish, as many of them also rely on vision (Gerlai, 2014). Whether a fish needs to distinguish between possible predators, sources of food and prey, a possible mate or competitors for their territory, discrimination of visual stimuli and learning are a fundamental part of the life of many species (Siebeck, 2004; Siebeck et al., 2010).

Fish vision has been an object of scientific research for many years, leading us to a better understanding of the physiological and morphological features of the fish visual system (Douglas and Djamgoz, 1990; Kroger, 2011). Most of these studies focus on innate visual behaviours, including the optomotor response, dorsal light reaction or phototaxic behaviour (Lyon, 1904; Douglas and Djamgoz, 1990; Schuster et al., 2011a). Furthermore, visual discrimination behavioural experiments have been used to test fish visual abilities such as object recognition, colour discrimination, spatial vision, detection of movement, visual acuity, size constancy or shape discrimination (Sutherland, 1961; Mackintosh and Sutherland, 1963; Yager, 1967; Sutherland, 1969; Karplus et al., 1982; Douglas et al., 1988; Neumeyer, 1992; Siebeck et al., 2008; Siebeck et al., 2009; Schuster et al., 2011a; Van-Eyk et al., 2011; Temple et al., 2013; Champ et al., 2014). All this research has given us a much better understanding of the capabilities of the fish visual system (Salva et al., 2014), but they have also paved the way to more complex visual behavioural experiments as they show that fish can be trained using associative learning. This has recently led to some remarkable new findings. For example, that fish share some visual illusions consistent with humans (Wyzisk and Neumeyer, 2007; Sovrano and Bisazza, 2009; Gori et al., 2014; Sovrano et al., 2015; Simpson et al., 2016), and that fish are capable of discriminating human faces (Newport et al., 2016).

This new wave of visual research has helped dispel the long-held belief that fish are simple organisms not capable of “cognitive abilities” (Brown et al., 2011; Patton and Braithwaite, 2015). Learning

visual discrimination is a cognitive, complex task (Zhang et al., 2010; Horner et al., 2013; Nithianantharajah and Grant, 2013; Nithianantharajah et al., 2013) that falls within Shettleworth’s definition of animal cognition. According to Shettleworth (Shettleworth, 2001), cognition is a process that will allow the animal to interact with the environment by using the sensory systems to internalise information, process it by learning and memory and then act based on this information. After a first analysis of the sensory input has been carried out by specialised detectors (e.g. colour, shape, movement) the visual system needs to perform perceptual bindings at a higher level to be able to develop adaptive behaviours (Salva et al., 2014).

In the last five to ten years, research has shown that many different species of fish are capable of learning and solving complex visual discrimination tasks, which are considered to involve different cognitive processes (Salva et al., 2014). Apart from the frequent use of the goldfish, a wide range of fish species including mosquitofish, guppies, cichlids, bamboo sharks, archerfish and redtail splitfin among others, have also been trained with visual discrimination tasks to test complex behaviours. Some examples are numerical abilities (Agrillo et al., 2009; Bisazza et al., 2010; Agrillo et al., 2012; Agrillo et al., 2014), categorization of objects (Schluessel et al., 2012; Schluessel and Duengen, 2015), hierarchy of the stimuli (Truppa et al., 2010), distinguishing mirror images, rotation and symmetry (Gierszewski et al., 2013; Schluessel et al., 2014b; Schluessel et al., 2014a), concept learning (Newport et al., 2014, 2015) and visual search (Rischawy and Schuster, 2013). These animals are able to perform this tasks even when they are thought to be dependent on specific cortical areas in humans and other mammals (Bussey et al., 2002; Hoffman and Logothetis, 2009; Freeman et al., 2010; Petruno et al., 2013; Warburton and Brown, 2015).

A few species of reef fish have also being used for visual research (Siebeck et al., 2008; Mitchell et al., 2017; Parker et al., 2017; Phillips et al., 2017; Green et al., 2018). Fish living in reef habitats are a great model for the study of visual neuroscience as they live in visually rich and complex environments (Marshall and Cheney, 2011; Marshall et al., 2019). Many of these species have developed bigger brains with larger telencephalic and visual structures through evolution (Wullimann, 1997b; Kotrschal et al., 1998; Demski, 2013). Some of the studies using fish from those habitats have shown that these species are easily trainable and are often able to solve visual discrimination tasks within a few days after capture (Siebeck et al., 2009; Champ et al., 2014; Siebeck et al., 2014; Silveira et al., 2015). One of those species is the Ambon damselfish (Pomacentrus amboinensis), a territorial species of damselfish which will chase intruders out of their nest, including conspecific rivals. The Ambon damselfish lives in social groups with one reproductive male, 1-7 females and some juveniles. These groups remain stable for at least 3 months and the males will

protect the nest and spawn with females of their own and other groups (McCormick, 2016). This fish uses distinctive UV reflectance facial patterns to recognise individuals from not only their own but also a closely related species, tending to be more aggressive to conspecifics (Siebeck, 2004; Siebeck et al., 2010). Ambon damselfish can also discriminate colours and complex shapes (Siebeck et al., 2008; Siebeck et al., 2009; Siebeck et al., 2014). Furthermore, this species also show an advanced form of learning by presenting anticipatory behaviour when performing these visual discrimination tasks (Siebeck et al., 2009). Although not a commonly used model, the demanding visual ethology and their ability to quickly learn visual discrimination tasks make Ambon damselfish and ideal model to look for parts of the brain involved in complex visual learning in teleost.

1.3 Part 3. Combining models to study visual learning

The overall aim of this project is to study fish visual learning and its underlying mechanisms, with a focus on the brain areas involved in it. I approached this aim with the use of two fish models, the Ambon damselfish (Pomacentrus amboinensis) and the zebrafish (Danio rerio), which provided distinct relevant and complimentary features for this study (Figure 1.5). While both model fish have clear advantages for my study, neither species would suffice on its own. While complex visual behaviours can be tested in the damselfish, not much is known of the details of their neuroanatomy. Also, being a wild animal, uncommon in non-behavioural laboratory research, the available tools to explore its nervous system are limited to traditional methods. On the other hand, while larval zebrafish stands out as a powerful model in neuroscience due to the optical accessibility of its brain, their learning abilities are somewhat limited and still being explored. Hence, the project aims to harness the benefits of both models so that general conclusions about the neuroanatomy of visual learning can be drawn.

As mentioned above, the zebrafish larva is a valuable model for neuroscience as its genome is well known, they are transparent, and their brains are small. Their fast development includes their nervous system, which allows them to present avoidance behaviours just after 1dpf for tactile stimuli and to visual stimuli as early as 4dpf (Saint-Amant and Drapeau, 1998; Yao et al., 2016). Although still in development, a general layout and function of the different structures of their brain are in place, including the morphological configuration of their telencephalon (Dirian et al., 2014). This fast development has allowed researchers to generate transgenic lines that provide optimal tools for the field of neuroscience, including a wide range of optogenetic tools available for the study of its nervous system. In the first part of the thesis, I present how I used zebrafish larvae with genetic modifications

to detect neuronal activity while performing visual habituation experiments using threatening stimuli. A habituation paradigm is ideal with this animal model, as behavioural experiments have shown that this simple form of learning can be reliably studied with both auditory and visual stimuli (Best et al., 2008; Wolman et al., 2011). However, the neuronal population activity during this learning behaviour had not yet been explored. In the 2nd chapter of this thesis, I review the visual escape circuit involved in the responses to threatening stimuli. This work establishes a neuroanatomical framework on which to evaluate the functional responses, locate the habituating structures and analyse their role in the system. In the 3rd chapter, I show how combining the whole-brain analysis of neuronal activity with a habituation paradigm allowed me to investigate the zebrafish visual system and its learning dynamics, something that would not be possible to do with any other vertebrate. Then, using the same methods, in chapter 4 I tested a model for fragile X syndrome’s habituation and evaluated if this disease model has disrupted visual learning. The results of these studies will be relevant for a better understanding of the basic principles of visual learning in vertebrates, and also will provide new tools for further research in this field.

Despite the advantage of being able to see its neuronal activity, zebrafish larvae are not able to learn visual discrimination tasks. In order to investigate how this learning is performed, I need to use a fish capable of such visual and cognitive tasks. Ambon damselfish live in a highly complex visual environment and use patterns to distinguish between members of their own and other species (Siebeck et al., 2010). It has almost ten times better visual acuity than the zebrafish larvae (Haug et al., 2010; Parker et al., 2017) and previous studies have demonstrated their capabilities for learning a range of ecologically relevant as well as abstract visual discrimination tasks. Therefore, in 5th chapter, I will show how I used behavioural experiments and neuroanatomical methods to study the brain areas involved in visual discrimination learning in Ambon damselfish. Based on the reviewed literature, I focused my research on the possible role of the pallium and its subdivisions in this type of learning. Histochemical methods and early gene expression immunohistochemistry were used on the Ambon damselfish brains to locate visual neuronal activity in learning conditions. Finally, intending to test how the activity of these areas plays a role in visual learning, an ecologically relevant tool (decrease in pH) was explored to disrupt normal visual learning. This chapter contains interesting exploratory results; however, time constraints prevented this part of the project to be completed. During part of my PhD, I worked with both projects trying to use the advantages of both animal models to study visual learning. Yet, although both projects suggested attractive preliminary results, it became clear that I was not going to be able to address both of them to the desired extent for the remaining duration of my PhD. As the zebrafish part of the project was advancing faster and logistic and practical reasons

made it easier to progress further, I eventually choose to focus on the visual habituation project. I decided, nonetheless, to include in this thesis the preliminary findings with my work with Ambon damselfish as I consider they are appealing and relevant.

Figure 1.5. Ambon damselfish and zebrafish as animal models.

Summary of pros and cons of the animal models that were used in to study visual learning and its neuroanatomical substrate. Ulrike Siebeck took the Ambon damselfish images on the left. The image from the right is from the 2009 Photomicrography Competition and belongs to Dominik Paquet.

2 Visual escape in larval zebrafish: Stimuli, Circuits, and Behaviour

2.1 Contributions

E. Marquez-Legorreta and M. Piber conceptualised and performed the literature review. E. Marquez- Legorreta wrote the “Introduction”, “Core visual escape circuitry”, “Modulation of the core escape circuit” and “Open questions about the loom circuit” sections and elaborated all figures except Figure 1. M. Piber wrote “The essential properties of loom stimuli” and “Execution of escape behaviours” sections and elaborated Figure 2.1. E. K. Scott revised and corrected the manuscript and provided conceptual feedback during the writing process.

2.2 Summary

Visual escape behaviour is important for survival, and elements of this behaviour are conserved from insects to humans. Because it needs to be robust and rapid, but also open to modulation, it is an excellent system in which to study visual processing and sensorimotor gating. Recent studies, especially in the transparent larvae of the zebrafish model system, have begun to shed light on the intricacies of visual escape circuitry, and in this chapter, we will review this progress. First, we will explore the essential properties of loom stimuli, including their movement, edges, and luminance changes, and will discuss how these stimulus properties, alone or in combination, can contribute to eliciting startle behaviour. Next, we will describe the escape behaviour itself, including the sequence of kinematic events that carries the animal away from the perceived threat and the various forms that this behaviour can take depending on the stimulus and context. We will then provide an in-depth review of the core circuitry that lies between the stimulus and response, beginning with the retinal and thalamic projections that carry loom relevant information to the tectum. We will also discuss how this information is likely to be processed in the tectum and the visuomotor projections to premotor cells in the hindbrain, including the well-known Mauthner neurons. Finally, we will describe ways in which context, such as alertness or hunger, can alter an animal’s responses to threatening visual stimuli and the ways in which specific brain regions may detect these conditions and impinge on the core escape circuit to modulate behaviour. We will conclude with perspectives on the important outstanding questions about visual escape circuits and specific experiments that might help in addressing them.

2.3 Introduction

The ability to detect an incoming predator and avoid its attack is a quintessential example of an evolutionarily selected behaviour, as the consequence of failure is death. Because of this, most animals perform behavioural responses shaped through evolution to increase the likelihood of escaping an imminent threat. Animals use a range of sensory modalities to detect and respond to threats, but vision provides a particularly rich set of information on a predator’s size, shape, position, and movement. It is unsurprising, therefore, that visual escape behaviour is widespread across the animal kingdom (Cooper and Blumstein, 2015; Sillar et al., 2016). This has inspired studies in systems as disparate as insects, crustaceans, fish, amphibians, reptiles, birds, and mammals (including humans), into the relationships that exist between visual stimuli and escape responses, and the neural circuits that perceive, analyse, and trigger responses to the stimuli (Schiff et al., 1962; Hayes and Saiff, 1967; Ewert, 1970; Sun and Frost, 1998; Carlile et al., 2006; Preuss et al., 2006; de Vries and Clandinin, 2012; Oliva and Tomsic, 2012; Landwehr et al., 2013; Yilmaz and Meister, 2013; De Franceschi et al., 2016; Fink et al., 2019).

Often an animal will respond to a visual threat with an explosive movement in the opposite direction, but responses also need to be adaptable to different situations and conditions. As a result, although the escape response is a strong innate behaviour, it can be modulated to suit a particular animal’s past experience or circumstances. This modulation can take the form of favouring a particular evasive response, tuning the kinematic speed and strength of that response, or indeed deciding not to respond at all. In order to understand both the core circuitry necessary for visual escape and the broader networks that modulate responses, we have to study the relationships that exist among various threatening visual stimuli, a range of possible output behaviours, and the brain activity that lies between.

In recent years, larval zebrafish have emerged as a useful model in which to study these questions. Zebrafish embryos show evasive behaviours as early as 1 day post-fertilization (dpf) in response to touch stimuli (Eaton and Farley, 1973; Saint-Amant and Drapeau, 1998; Downes and Granato, 2006), and larvae execute ballistic visual escapes from looming stimuli as early as 4dpf (Yao et al., 2016). Zebrafish are externally fertilised, meaning that animals are readily observable at these early developmental stages. Importantly, they are also optically transparent as larvae, meaning that internal structures, including the brain, can be observed through microscopes without the need to dissect, cannulate, or otherwise interfere with the larva’s development or neural activity (Simmich et al., 2012).

These biological attributes provide a nearly ideal platform for recent advances in genetically encoded fluorescent indicators of neural activity (Broussard et al., 2014; Lin and Schnitzer, 2016), in 2-photon and selective plane illumination microscopy (SPIM) (Ahrens et al., 2013a; Wolf et al., 2015), and in quantitative approaches for detecting and modelling brain-wide patterns of neural activity from large imaging datasets (Chen et al., 2018b). Combined, these approaches have made it possible to image activity throughout the larval zebrafish brain at cellular resolution, and to make sense of the collective activity of the tens of thousands of neurons composing it (Vanwalleghem et al., 2018). These capabilities have recently been extended to free-swimming larval zebrafish (Cong et al., 2017; Kim et al., 2017) and can be also integrated with powerful tools for functionally testing the relevant circuitry (using optogenetics or ablations) and describing the morphology of the neurons involved (using single-cell labelling or immunofluorescence) (Leung et al., 2013; Feierstein et al., 2015; Dunn et al., 2016b; dal Maschio et al., 2017; Lovett-Barron et al., 2017; Helmbrecht et al., 2018; Tabor et al., 2018).

In summary, these new approaches, combined with their biological attributes, make zebrafish larvae uniquely suited to the study of the circuitry that underlies visual escape. In this chapter, we will focus on zebrafish larvae as we describe behavioural responses to threatening visual stimuli and the neural circuits that drive these behaviours. This will include descriptions of the types of stimuli and stimulus features that can trigger visual escape, along with the various escape manoeuvres that can result. We will focus mostly, however, on recent discoveries about the brain regions, local networks, and cellular microcircuits that larval zebrafish use to detect and respond to visual threats. Finally, we will suggest some of the future research needed to better understand the circuitry and the mechanisms of the visual escape response in zebrafish.

2.4 The essential properties of loom stimuli (What causes an animal to escape?)

Zebrafish, like most animals, have a powerful instinct for avoiding oncoming predators, and these responses have been explored extensively in experiments using live predators, animated predators presented on screens, and robotic simulated predators (Bass and Gerlai, 2008; Saverino and Gerlai, 2008; Gerlai et al., 2009; Colwill and Creton, 2011; Ahmed et al., 2012; Luca and Gerlai, 2012; Ladu et al., 2015). In general, zebrafish show stronger evasive responses to sympatric predators (those that the zebrafish might encounter in their natural habitat), than to allopatric predators or nonpredatory fish, suggesting some innate selection for relevant visual threats that allows adaptive behaviour at the first encounter (Bass and Gerlai, 2008; Ahmed et al., 2012). This makes sense from an evolutionary

standpoint, but does little to reveal the specific visual properties contained in these stimuli that drive escape behaviour.

Addressing this requires a reductionist approach in which stimulus properties can be presented in isolation or controlled combinations to gauge their efficacy in driving behaviour. For example, some simple and unnatural stimuli, such as a moving rectangle, elicit strong responses (Ahmed et al., 2012). Other studies exposing the zebrafish to foreign objects that have a set of eyes drawn onto them show that various other simple shapes can induce evasive responses (Blaser and Gerlai, 2006). The presentation of silhouettes of predators, including birds (Dill, 1974), powerfully elicits responses in zebrafish. Interestingly, responses to these predatory silhouettes were stronger than those to actual predators, suggesting that they not only conserve, but may accentuate the properties that zebrafish find salient (Luca and Gerlai, 2012). These properties may be quite simple, as further experiments reducing the complexity of the stimuli to simple growing dots proved that these, too, elicited escape behaviour (Luca and Gerlai, 2012).

This simple stimulus, a dark expanding circle on a lighter background (hereafter referred to as a “loom”), has become the stimulus of choice for studying startle circuitry and behaviour in model systems such as larval zebrafish. Even this simple stimulus, however, contains multiple possible cues for a larva to process when deciding whether to execute an escape. One is a drop in luminance that occurs as the black dot occludes an increasing portion of the visual field as it “approaches” the zebrafish. It also contains edge information, represented as the moving boundary line between the growing dark circle and its lighter background (see Figure 2.1A-C). The specific contribution that each cue (dimming and moving edges) makes has been the topic of considerable study, but their exact roles in driving escape behaviour are still not fully understood. Studies in mice, for instance, show that light looms on dark backgrounds do not elicit startles, nor do dimming stimuli that lack moving edges (Yilmaz and Meister, 2013), suggesting that both cues must be present in order for startles to occur in mice. Convergent results have been found in larval zebrafish in a study using both receding and approaching stimuli in both light and dark configurations, along with dimming stimuli (Temizer et al., 2015). In these experiments, dark looms elicited escapes in more than 90% of trials while the light looms on dark background triggered startles only 20-30% of the time. Receding stimuli had an even weaker effect below 10%. To test whether the change in luminance alone was sufficient to cause escape, the full-sized disk was dimmed with similar temporal dynamics to the dimming that occurs during loom. The dimming stimulus failed to cause startles, suggesting that luminance information alone is insufficient for driving escape behaviour. The unique contributions of moving edges can be explored using an expanding checkerboard stimulus on an isoluminant grey background. This delivers

movement similar to that in the loom stimulus, but without a dimming component. Using this approach, Dunn et al (Dunn et al., 2016a) found that checkerboard stimuli drove behaviour at a similar rate to looms, while Heap et al (Heap et al., 2018b) found a much lower rate of escape to checkerboard stimuli than to looms. Differences in the details of the checkerboard stimuli likely account for these different results, as discussed in Heap et al (Heap et al., 2018b). The interpretations of the combined results from these studies are that moving edges appear to be necessary for eliciting escape behaviour, but that the direction of the luminance change (brightening or dimming) is important, and, indeed, dimming may make a major contribution to the likelihood of escape behaviour when moving edges are present.

The temporal and spatial properties of looming stimuli also have implications for the probability and timing of the ensuing escape response. Researchers across model systems use a common language to describe the mathematical properties of these stimuli. The expanding loom projects a dark disk on the zebrafish’s retina at a distance (d), with an angular size (θ) at the retina that increases with time (Figure 2.1D). Trigonometrically, (θ) is twice the inverse tangent of half the object size (l). The relationship between (l) and the object’s approach velocity (v), expressed as the (l/v) ratio, dictates the rate of change for the retinal size of the object. Varying this (l/v) ratio allows investigators to manipulate the temporal dynamics of the looming stimulus to study which of its varying aspects are used by an animal to guide its behaviour.

For a simulated predator approaching at a constant rate, the image on the retina grows at an ever- increasing speed (Figure 2.1E). When detected monocularly, the growth of (θ), and the associated changes in its rate of expansion (ρ)(Figure 2.1F), characterise the approach, including the “time to collision” (τ)(Figure 2.1G) (Hatsopoulos et al., 1995). This, in theory, presents the animal with three key optical parameters, the object’s size, velocity, and rate of expansion, with which it could detect the threat. Naturally, this has led to a search of the neurons encoding these parameters in the brain.

Indeed, loom responsive and escape executing neurons have been found in locusts that rely on information about both θ and ρ (Fotowat et al., 2011). These neurons’ firing increases with the growth of the looming object, peaking at a given angular threshold that also correlates with the initiation of an escape response, and then dropping off. In other words, the neuron responds when the approaching object reaches a certain angular size. Systematic tests with a range of (l/v) ratios demonstrated that it is the value of θ that determines the timing of this response. This information is then combined with information on other properties of the loom stimulus to allow the locust to make an appropriately timed escape maneuver (Gabbiani et al., 1999).

Interestingly, in the pigeon, all three optical variables of image expansion were found to correspond to separate categories of loom sensitive neurons (Sun and Frost, 1998). Firstly, τ cells, whose spiking activity starts at a particular time-to-collision, regardless of object size and velocity, initiates the behavioural escape response. Secondly, ρ cells respond once a threshold for angular velocity is surpassed. They signal the rate of change of visual angle θ’ and the absolute rate of expansion. Finally, η cells’ firing rate peaks at a critical stimulus size threshold without regard to the velocity of the approaching object. A recent set of studies in larval zebrafish has hinted at the existence of η neurons in the optic tectum (Temizer et al., 2015; Dunn et al., 2016a; Bhattacharyya et al., 2017). Yet, These studies empirically calculated the size threshold angle based on the reaction times to stimuli with varying sizes and growth rates, finding rather divergent thresholds of 20° (Temizer et al., 2015), 35° (Bhattacharyya et al., 2017), and 72° (Dunn et al., 2016a) as the angle at which the response occurred. This variability can be explained by differences in the experimental approaches, where Temizer et al (2015) presented looms horizontally to head-embedded larvae, Bhattacharyya et al (2017) presented a range of more natural stimuli horizontally to free-swimming larvae, and Dunn et al (2016a) presented loom stimuli from below to free-swimming larvae. The studies converge on the interpretation that there is an angular size threshold for triggering escape behaviour in larval zebrafish, but their differences serve as a reminder that numerous factors related to the stimulus (size, shape, growth rate, and direction) and the animal’s circumstances (head embedded versus free-swimming, in this case) impact on the animal’s behavioural decisions. This reinforces the fundamental properties of visual escape: it is a powerful and innate behaviour that nonetheless remains subject to variability from an unpredictable world, and to modulation from the animal’s internal state.

Figure 2.1. Components and computations of the loom.

A-C: representations of the loom (A), checkerboard (B), and dim (C) stimuli, from the onset of the stimulus (left) to the final frame (right). D: Schematic representation of the eye of an observer at the visual angle θ, subtended by an object with a half size l. The distance d will decrease as a function of time with approach of the object at the velocity v. If the approach velocity is constant, the angle θ(t) subtended by the object and its rate of increase [ρ(t) = θ'(t)] both grow non-linearly as a function of time. E-G: Evolution of the kinematic parameters of the looming stimulus during approach. θ (t) is the loom angular size, ρ (t) is the loom angular velocity and τ (t) is the time-to-collision. Adapted from Peek & Card (2016).

2.5 Execution of escape behaviours

A threatening visual stimulus can lead to any of several behaviours. In the case of the fully developed adult zebrafish, these may include startle, freezing, and a range of avoidance behaviours (thrashing, zig-zagging, leaping and diving (Luca and Gerlai, 2012; Bishop et al., 2016)) that are, presumably, selected based on the nature of the predator and the animal’s surroundings. For example, thrashing is usually performed in shallow water and is followed by freezing, a sequence that is thought to stir up debris and camouflage the fish from the potential predator (Bass and Gerlai, 2008; Luca and Gerlai, 2012). Similarly, the selection of a particular behaviour may be influenced by the animal’s available cover (such as shadows or vegetation)(Hein et al., 2018). This is an interesting and nuanced space, but is only peripherally related to the startle behaviour seen in larval zebrafish, so we will not focus on it here.

The most common escape behaviour across fish species, including in larval zebrafish, is a C-start that can be triggered by touch (Fetcho, 1991), vibrations including auditory cues (Burgess and Granato, 2007), water flow (Stewart et al., 2013), and vision (Eaton et al., 1977). The escape latency is very short, with responses beginning 5-10 msec after stimulus onset in adult zebrafish (Eaton et al., 1977). It involves a fast, powerful activation of head and body muscles, unilaterally contracting the axial muscles in the trunk and tail, thus bending the body into a C shape (Eaton et al., 1977). This manoeuvre can be sectioned into three stages, whereby the initial 15-20 msec are stereotyped (Eaton et al., 1977). The extensive contraction described above (Stage 1) usually occurs after 5-8 msec, and is followed by a further contraction on the opposite side of the body (Stage 2), during which the tail straightens. These two stages lead to rapid propulsion of the fish away from the predator, and are generally complete after approximately 20 msec. Following this, the fish may show any of a number

of behaviours including, gliding, braking, or continued swimming (Eaton et al., 1977). The response in larvae follows the same motion sequence as in the adult, but there are some quantitative differences in terms of the speed, kinematics, and time course of the behaviour (Kimmel et al., 1974; Nair et al., 2015). Larvae have a slightly slower response latency of under 15 msec, and reach full contraction more slowly than adults, yet head displacement speed relative to body length is three times greater (a function of their having a deeper C-bend) (Eaton et al., 1977).

The circuitry involved in the detection and processing of visual threats will be discussed in a later section, but we will briefly discuss the motor circuits relevant to startle here. Short latency C-starts are mediated by the Mauthner cells (M-cells), a pair of bilateral reticulospinal neurons that innervate motor neurons throughout the spine. A single action potential in one M-cell initiates a C-bend on the opposite side of the body (Zottoli, 1977; Eaton et al., 1982). However, escape behaviours also occur if, instead of activating the M-cells, M-cell homologues are activated (Kimmel et al., 1980; Liu and Fetcho, 1999; Kohashi and Oda, 2008). These responses, the long-latency C-starts, take longer to initiate, with an average escape latency of 28.3 msec in larvae (Liu and Fetcho, 1999; Budick and O'Malley, 2000; Burgess and Granato, 2007). They tend to result from slowly approaching loom stimuli, and unlike short-latency C starts, involve the usage of the pectoral fins, which may permit the fine-tuning of the escape movement (McClenahan et al., 2012; Bhattacharyya et al., 2017). By contrast, faster loom approach rates typically evoke short latencies C-starts (Bhattacharyya et al., 2017). The short- and long-latency C-starts, therefore, provide different behavioural options depending on the urgency of the threat. M-cell driven short latency C-starts provide a rapid and dramatic response to an immediate threat, sacrificing adjustment and control of speed and power. Long latency C-starts are slower, variable, and kinematically more complex, likely permitting control over the speed or direction of the escape.

These different attributes of short- and long-latency C-starts, and their links to the nature of the threat facing the animal, occupy a small part of the fish’s broader behavioural repertoire and environmental context. Feldman Barrett and Finlay (2018) suggest that defensive behaviours are context dependent, purposeful, flexible actions. Exclusive use of short latency C-bends would be energetically inefficient, and would make the escape behaviour predictable for the predator. Thus, the ability to appropriately estimate the level of threat, and to have a range of possible responses, enhances the probability of survival (Nair et al., 2017). At longer durations, complex details of their environments, such as hiding places, can be integrated into their decisions, adding another level of complexity to the decision-making (Hein et al., 2018). Aspects of an animal’s internal state, such as motivation to feed, mate, or defend territory, could also influence its appetite for risk, and therefore, its responses to

threats. A core escape circuit that provides a reliable, fast, and strong response to urgent threats, but that is modulated by other circuits responsible for encoding past experience, environmental context, and internal state, could provide a means for delivering the correct adaptive behaviour in response to a range of threats and circumstances. In the next two sections, we will discuss the core escape circuit in larval zebrafish, and will then present evidence for the modulation of this circuit by a number of other brain regions.

2.6 Core visual escape circuitry

So far, we have reviewed the key features of loom stimuli and the characteristics of the visually evoked escape responses that result. Of course, these stimulus-response relationships rely on the neural circuitry that perceives and integrates sensory information, performs calculations about the nature of the threat, and recruits the appropriate motor circuits to elicit an adaptive response. Elements of these circuits have been identified for decades in a range of model systems, but the entire core circuit is just now coming into focus, due in a large part to new optophysiological and optogenetic methods described in the Introduction. This newfound ability to image activity across large populations of neurons has made zebrafish larvae a particularly valuable model for studying visual startle circuits, among others, and has permitted the basic loom circuit to be observed in its entirety across the brain.

The first neurons to respond to visual stimuli are in the retina, and visual information is then sent through the Retinal Ganglion Cell (RGC) axons to multiple arborisation fields (AFs) in the brain. A majority of RGCs extend to AF10, which is the neuropil of the tectum (Burrill and Easter, 1994; Robles et al., 2014) (red projections to the tectum in Figure 2.2 and Figure 2.3). These terminals are ordered in a precise topographic map (across the rostro-caudal and medial-lateral axes) that is duplicated in the various dorsal (shallow) to ventral (deep) layers of the tectal neuropil in a way that is highly conserved in most species of fish (Vanegas and Ito, 1983; Vonbartheld and Meyer, 1987; Baier et al., 1996; Robles et al., 2013; Kita et al., 2015). Different subtypes of RGC, relaying different types of visual information from the retina, target specific dorso-ventral layers of the tectal neuropil and other AFs (Robles et al., 2013; Robles et al., 2014). The nature of the visual information emerging from the retina is important, since this provides the working material with which circuits in the tectum and elsewhere will decode and interpret visual stimuli such as looms. In recent years, RGC subtypes have been shown to deliver specific orientation, direction, and size relevant information to the tectum (Gabriel et al., 2012; Nikolaou et al., 2012; Lowe et al., 2013; Preuss et al., 2014; Semmelhack et al.,

2014; Barker and Baier, 2015), including the possibility that tectum receives input from loom-specific RGCs in the retina (Temizer et al., 2015).

Figure 2.2 summarizes the loom related information that RGCs deliver to the neuropil, including details of the neuropil laminae and sublaminae that may have specific roles in the subsequent processing. RGCs carrying information on motion and changes in luminance are targeted mostly to the stratum fibrosum et griseum superficiale (SFGS) and also to the stratum griseum centrale (SGC)(Temizer et al., 2015). Dimming stimuli mostly activate axons innervating the deep SFGS6 layer, while SFGS2- SFGS5 inputs are more active during looms than dimming. That RGCs innervating the SFGS2-5 layers respond to an approaching object more than to a dim (change in luminance) suggests that they are processing information related to the moving edges. Interestingly, this last group of RGC axons selectively innervate the tectum, and not other AFs (Robles et al., 2014). The prominence of loom information arriving in the tectum suggests an important role for this structure in the detection of visual threats, a notion that is supported by experiments in which lesions of the tectal neuropil markedly reduced escape responses in zebrafish larvae (Temizer et al., 2015).

Other studies of the tectal neuropil’s laminar structure suggest a mechanism by which the tectum could distinguish looms from other types of visual stimuli. Apart from luminance changes and moving edge information, the tectum also processes details related to the size and directionality of moving objects. It seems that the deeper layers of the SFGS are preferentially activated by bigger objects, while its superficial layers and the stratum opticum (SO) are most active in response to smaller objects (Figure 2.2) (Preuss et al., 2014; Semmelhack et al., 2014). This filtering is mediated by the tectal Superficial Inhibitory Neurons (SINs), which are located in the SO, and encode size and direction information (Del Bene et al., 2010; Preuss et al., 2014; Barker and Baier, 2015; Abbas et al., 2017; Yin et al., 2018). SINs tuned to specific stimulus parameters modulate the activity of the periventricular interneurons (PVIN) in their area, which in turn activate the periventricular projection neurons (PVPN) that send their axons to premotor areas (Del Bene et al., 2010; Helmbrecht et al., 2018). In this manner, SINs may be gating the information that flows from the RGCs to the PVNs, making stimuli with specific characteristics (such as looms) more salient. Dunn and colleagues (Dunn et al., 2016a), in modelling this interaction, found that activity in SINs and RGC axons could explain the PVN’s responses to loom stimuli. Providing functional support for this idea, Barker and Baier (Barker and Baier, 2015) ablated SINs, altering avoidance responses to large objects, although this effect was not observed in another similar ablation study (Yin et al., 2018).

the ways in which other visual regions of the brain could contribute to escape behaviour. Notably, a recent study has revealed a specific and important role for the thalamus in loom processing (Heap et al., 2018b). In this study, Heap and colleagues characterised visual responses in the thalamus, finding it to be responsive to looms. They then identified projections from the thalamus to the tectal neuropil (mainly localised in deeper layers, orange in Figure 2.2 and Figure 2.3), and found that these projections were active specifically during looms. To assess the functional contributions of these projections, they ablated the thalamo-tectal tract and found a selective reduction in tectal responses to looms, but not to other visual stimuli, meaning that the thalamus was contributing loom-relevant information to visual processing in the tectum. Ablation of the thalamo-tectal tract also had a surprising impact on escape behaviour. These ablations not only reduced the probability of loom- induced escape, but reversed their directionality, causing the larvae to startle toward, rather than away from, the looming stimulus. To gauge the stimulus property responsible for thalamic loom responses, they presented checkerboard and dimming stimuli, and found that the drop in luminance (dimming) was exclusively responsible for the thalamus’ loom responsiveness. This luminance information is likely delivered to the thalamus by RGC axons terminating in AF8 (Temizer et al., 2015). In further behavioural experiments with these stimuli, Heap and colleagues found that checkerboard looms had a decreased startle rate compared to dark looms, but that dim stimuli failed to elicit responses, consistent with prior studies (Temizer et al., 2015; Dunn et al., 2016a). This suggests that thalamic luminance information cannot drive escapes on its own, but increases the probability of a response, and controls its direction. To confirm these proposed roles for luminance and the thalamus in visual escape, Heap and colleagues presented a split stimulus in which a dim was presented to one eye while and expanding checkerboard was presented to the other. This stimulus led to a higher probability of escape than dimming or looming checkerboards individually, and produced more responses toward the looming checkerboard stimulus than away from it. The results of this study support a model in which information about moving edges, carried directly from RGCs into the tectum, are necessary for the escape responses, but that luminance information brought to the tectum through the thalamus is necessary to direct the escape away from the predator (the side receiving the earlier or more dramatic drop in luminance). Although this challenges the basic retino-tectal model of visual escape, it still places the tectum at the centre of loom processing. As the point of confluence for moving edge and luminance information, it is well positioned to perform the key calculations necessary for visual predator detection.

The details of the tectal microcircuits combining these streams of information remain largely unknown, but the results of these calculations are delivered to the dendrites of PVPNs in the deep

layers of the neuropil (Scott and Baier, 2009; Robles et al., 2011) (Figure 2.2), and the PVPNs then relay these outputs to premotor neurons to shape the motor response (Figure 2.3)(Zottoli et al., 1987; Sato et al., 2007; Yao et al., 2016). Recent findings suggest that two subpopulations of PVPNs are involved. One projects to contralateral reticulospinal neurons and the nucleus of the medial longitudinal fasciculus (nMLF), while the other innervates ipsilateral reticulospinal neurons (Helmbrecht et al., 2018) (dark and light blue neurons, respectively, in Figure 2.3). These subtypes of PVPNs presumably permit the activation of the contralateral M-system (driving a C-bend) to be coordinated with other postural adjustments and the inhibition of ipsilateral spinal motor circuits (Fetcho and Faber, 1988; Fetcho, 1991; Song et al., 2015) (for reviews, see (Eaton et al., 1991; Korn and Faber, 1996; Eaton et al., 2001; Korn and Faber, 2005; Medan and Preuss, 2014; Hale et al., 2016)). Interestingly, tectal projections to the M-cells fire equally during flashes and looms, but the M-cell responses are more specific to loom stimuli (Yao et al., 2016). These results suggest that the M-cells may require further loom-specific information in order to reach threshold, or that inhibitory mechanisms could keep the M-cells in check during other stimuli. As noted above, escape responses still occur after M-cell ablation, showing that these responses can be generated and modulated by multiple categories of reticulospinal neurons (Eaton et al., 1982; Liu and Fetcho, 1999; Dunn et al., 2016a; Naumann et al., 2016), with the M-cells mediating short-latency C-bends and other reticulospinal neurons driving long-latency C-bends (Kohashi and Oda, 2008; Bhattacharyya et al., 2017). This difference in motor circuitry is important, because it underlies the previously described choice between a rapid, invariant escape response (typically used for rapid looms and imminent collisions) and a slower, adjustable escape (used for less urgent threats) (Bhattacharyya et al., 2017).

After the initial reorientation produced by the C-bend, the escape response involves a swim burst, propelling the fish forward and away from the threat. Considering the results of previous studies, it is likely that this motor response is produced by the nMLF (light green nucleus in Figure 2.3) as it has been involved with escape responses and forward swimming behaviours, and has visually responsive neurons (Gahtan et al., 2002; Orger et al., 2008; Sankrithi and O'Malley, 2010; Severi et al., 2014; Thiele et al., 2014; Wang and McLean, 2014).

Figure 2.2. Loom relevant information arriving to the tectal neuropil.

Left: RGC axons (in red) carry movement information to layers 2-5 of the SFGS and dimming information to layer 6 of the SFGS and the deep portion of the SGC (Temizer et al., 2015). Thalamic axons (in orange) carrying dimming information target the deep region of the SGC and also deep parts of the stratum album centrale (SAC) (Heap et al., 2018b). Right: Smaller objects are preferentially processed in the SO and superficial layers of the SFGS while larger objects are mostly processed at deeper layers of the SFGS (Preuss et al., 2014; Semmelhack et al., 2014).

Figure 2.3. The core visual escape circuit.

Schematic summary of the brain regions involved in detecting and responding to visual looming stimuli.

2.7 Modulation of the core escape circuit

C-bend escapes are disruptive and energetically expensive, so while they are critical to survival in some contexts, it is also important that the animal can modulate or prevent these behaviours when they are not needed. In this section, we will highlight the ways in which internal states, recent experience, and other types of context impact visual escape behaviour. We will also review the brain regions and circuits that may be responsible for encoding these factors, and the ways in which this information reaches and influences the activity of the core escape circuit to produce behaviour appropriate to the circumstances facing the animal. These pathways are summarized in Figure 2.4.

An example of an internal state that can impact visual escape has recently been provided by Lovett- Barron and colleagues (Lovett-Barron et al., 2017), who combined calcium imaging with visual escape assays, seeking to find neurons whose activity before the stimulus correlated with quick reactions (which they used as a proxy for alertness). By performing immunohistochemistry and registering this information against their calcium imaging results, they identified specific functional properties of the neurons that were active prior to rapid escapes. This approach identified noradrenergic neurons in the locus coeruleus; cocaine-and amphetamine-regulated transcript (CART) and cholinergic neurons in the tegmentum; and dopaminergic, serotoninergic, and neuropeptide-Y neurons of the hypothalamus (purple structure in Figure 2.4). Similar activity was found in these brain structures after caffeine was applied, which also decreased the reaction time. On the other hand, activity in somatostatin neurons in the hypothalamus was correlated with longer reaction times. These results led Lovett-Barron and colleagues to conclude that the neuromodulatory structures active before short reaction times are involved in alertness and that this internal state of the animal reduces the reaction time to the threatening stimulus, although the mechanisms by which this influences the core visual escape circuit remain unclear.

(black neuron in Figure 2.4). When a looming stimulus is presented, these dopaminergic neurons do not activate the inhibitory interneurons, allowing free communication between the tectum and reticulospinal neurons and permitting the escape response. This circuit provides a “safety” that prevents unnecessary C-bends, even when a visual stimulus contains some properties (motion or edges) in common with threatening stimuli.

A third study by Filosa and colleagues involved the relationships among visual stimuli, approach and escape behaviour, and the hypothalamic-pituitary-interrenal (HPI) axis (Filosa et al., 2016). Under pressure to find food, starved larvae become less risk-averse, showing a decreased tendency to startle from visual stimuli and increased approach behaviour associated with feeding. This state was associated with less cortisol and less activity in the HPI axis. They then showed that starved larvae also had increased activity in raphe serotoninergic neurons that project to the tectum (pink in Figure 2.4 and Figure 2.5), and that by manipulating the activity of these neurons they could modulate the behaviour of both fed and starved larvae. Finally, they found that the activity of the HPI axis and serotoninergic neurons modifies tectal responses to visual circles of various sizes (where large stimuli generally elicit startles and small stimuli cause approach behaviour) in a way that matched the behavioural results. Their findings suggest that the activity of these serotonergic neurons is tuning the sensitivity of tectal circuits, thus tipping the balance in downstream premotor regions toward avoidance or approach, depending on the animal’s immediate survival pressures.

Involvement of the hypothalamus is a common thread among these studies, and it appears to have distinct effects on escape behaviour and to impinge on different parts of the core escape circuit under different conditions. Since the hypothalamus also sends projections directly to deep layers of the tectal neuropil (purple axons in Figure 2.5)(Heap et al., 2018a), it may influence escape behaviour in ways that have not yet been studied. This makes the hypothalamus, and its circuit-level interactions with loom-sensitive brain regions, intriguing topics for further studies of how visual startle is modulated.

Figure 2.4. Modulatory structures of the core visual escape circuit.

Schematic representation of the brain regions that are known to modulate (solid lines) or are likely to modulate (dashed lines) the core visual escape circuit.

2.8 Open questions about the loom circuit

Above, we have outlined both the core visual escape circuit and other brain regions that are involved in modulating it. The picture of loom processing that this paints, however, is incomplete in two important ways. First, it is currently understood at the level of brain regions, with little knowledge of the microcircuits within those regions responsible for the processing. While some of the studies that we have described resolve the individual neurons within these regions during calcium imaging, it is generally without information on those neurons’ morphologies or functional characteristics. This shortcoming is illustrated by our current knowledge of the tectum’s contributions to visual escape. The tectum receives loom-relevant input directly from the retina, along with luminance information from the thalamus, and likely modulatory information from the hypothalamus and elsewhere. The structures and connectivity of the tectal neurons that receive this information, and the functional architecture of the networks through which this varied information flows and interacts, remain mostly unknown. In the tectum, as well as other structures involved in visual startle, the mapping of these microcircuits (through anatomical and histochemical studies) and functional testing of their relationships with one another (using optogenetics) will be necessary to register specific types of neuron to the types of information that they process and the roles that they play in the broader network.

The goal will be to elaborate the region-based model to one that describes the microcircuits and neuronal networks mediating visual escape, and a few promising studies have begun to employ this approach in other visual pathways. These studies work at the nexus of population-scale functional imaging (with calcium or voltage indicators), anatomical labelling (Forster et al., 2017a; Forster et al., 2018), and light sculpting to produce custom holograms in the brain (Favre-Bulle et al., 2015; Accanto et al., 2019). This combination of techniques has permitted neurons with particular response profiles to stimuli (judged by calcium imaging) to be labelled with photoactivatable GFP (Patterson and Lippincott-Schwartz, 2002) through the use of a targeted hologram on the relevant neuron (dal Maschio et al., 2017; Kramer et al., 2019), linking the neurons’ morphologies to their activity. A similar approach could be used for the selective activation or silencing of particular functional categories of neurons, using sculpted light and optogenetic actuators (dal Maschio et al., 2017). In

terms of the neurons’ functional characteristics, careful registration of in vivo calcium imaging and post mortem immunohistochemistry has revealed the neurochemical properties of neurons with interesting patterns of activity during sensory processing (Lovett-Barron et al., 2017). In a conceptually similar approach, transgenic marker lines have been combined with calcium imaging to provide further information about the active neurons in real time (Dunn et al., 2016a; Yao et al., 2016; Heap et al., 2018b). A broad library of transgenic marker lines (Davison et al., 2007; Scott et al., 2007; Scott and Baier, 2009; Satou et al., 2013; Kimura et al., 2014; Kawakami et al., 2016; Forster et al., 2017b; Tabor et al., 2019) should make this a flexible and powerful approach in the future.

The second way in which the model is incomplete is that it almost certainly still omits whole brain regions contributing to the decision making and execution of visual escape. Promising brain regions include additional loom-responsive areas like the pretectum, midbrain tegmentum, and forebrain (Dunn et al., 2016a; Chen et al., 2018b), which have yet to be thoroughly explored functionally. Among these structures, the pretectum (yellow in Figure 2.4) receives direct inputs from the retina (Burrill and Easter, 1994; Yanez et al., 2018), including luminance information into AF6 (Temizer et al., 2015). In addition, it has reciprocal connections with the tectum (Yanez et al., 2018) and has been linked to other visually evoked behaviours in zebrafish (Kubo et al., 2014; Portugues et al., 2014; Semmelhack et al., 2014; Naumann et al., 2016). The tegmentum, another loom-responsive area, includes a nucleus that is also likely to modulate escape responses. Studies in goldfish and sunfish have shown that the nucleus isthmi (NI) responds to looming stimuli, is more sensitive to novel visual stimuli, and has reciprocal connections with the deeper layers of the tectum (green area in Figure 2.4) (Vanegas and Ito, 1983; Striedter and Northcutt, 1989; King and Schmidt, 1993; Northmore and Gallagher, 2003; Gallagher and Northmore, 2006). This raises the possibility that an NI-tectal loop is involved in the processing information about approaching objects (Northmore and Graham, 2005; Graham and Northmore, 2007). This visual feedback circuit is preserved across vertebrates (Gruberg et al., 2006) and its role has been studied in birds. Several studies of owls, chickens, and pigeons strongly suggest that the NI modulates tectal responses to salient visual stimulus and is involved with visual attention. (Marin et al., 2007; Asadollahi et al., 2010; Mysore and Knudsen, 2013; Goddard et al., 2014; Mysore and Knudsen, 2014; Basso and May, 2017).

is likely to receive most of its visual information from the preglomerular nucleus (brown relay structure in Figure 2.4). According to tracing experiments in weakly electric fish, rainbow trout, carp, and goldfish, this diencephalic structure receives projections from the tectum and establishes connections with the dorsal pallium, which in turn closes the loop by sending axons to the tectum (brown in Figure 2.5) (Echteler and Saidel, 1981; Luiten, 1981; Demski, 2003; Folgueira et al., 2004; Yamamoto and Ito, 2005; Northcutt, 2006; Yamamoto et al., 2007; Yamamoto and Ito, 2008; Giassi et al., 2012b; Giassi et al., 2012a; Demski, 2013). These connections are reminiscent of the visual pathway involved in the visual escape responses triggered by looms in mice (Mueller, 2012; Zhao et al., 2014b; Carr, 2015; Wei et al., 2015; Yuan and Sus, 2015; Perathoner et al., 2016; Pereira and Moita, 2016; Shang et al., 2018; Zhou et al., 2019) and that may be important in the processing of threatening visual stimuli in humans (McFadyen et al., 2017). Given the similarities to the amygdala’s connectivity in mammals, it is tempting to think that the Dm may play a similar role in fish.

The circuitry involved in the visual escape response can be approached in its minimal form with a retina-thalamus-tectum-hindbrain circuit (Figure 2.3) or as a more complex interaction with multiple modulating structures (summarised in Figure 2.4). Viewed either way, the tectum is the structure where most of the visual information from the retina, thalamus, NI, telencephalon, and other structures converges (Figure 2.5), and is where the separate features of loom stimuli are integrated. As such, the tectum appears to serve as a visuomotor pivot where visual information is gathered and interpreted, and where premotor signals for an escape response are generated.

A framework for these microcircuits is illustrated in Figure 2.5, were we summarise the laminar structure of the tectal neuropil as it relates to various visual and modulatory inputs. This highlights the fact that retinal inputs dominate the SFGS, where information about movement, edges, and topography are delivered, providing the raw material for visual processing. In deeper layers, inputs from other regions, with relevance to luminance, context, and internal state are more prominent, providing further detail that may be necessary for the direction, kinematics, or indeed, the decision to startle. Registering this against the anatomical structure of the tectum (with superficial retinorecipient SINs, PVINs with dendrites in the superficial neuropil, and PVPNs with dendrites in the deep neuropil) provides a structural scaffold through which processing and modulation could occur as information progresses from the superficial to the deep neuropil, and then on to downstream structures. The elucidation of the individual cell types and tectal microcircuits carrying out this processing are of particular interest for future work, as they will both provide a fine-grained functional map of this important transformation, and will reveal the circuit-level mechanisms by which modulatory information is incorporated.

Figure 2.5. The confluence of retinal, thalamic, and modulatory information in the tectal neuropil.

Left: The two core visual afferents to the tectal neuropil are the RGCs (red) and projections from the thalamus (orange) (Temizer et al., 2015; Heap et al., 2018b). Right: Structures with known afferents to the tectal neuropil, and that are likely to modulate loom processing, include the hypothalamus (purple), raphe (pink), nucleus isthmi (green) and the telencephalon (brown) (Luiten, 1981; King and Schmidt, 1993; Folgueira et al., 2004; Filosa et al., 2016; Heap et al., 2018a). Middle: the projections from each of these structures terminate in stereotyped laminae and sublaminae in the neuropil, with a trend toward visual (especially retinal) input to superficial layers and modulatory input to deeper layers. Retinal information is received and processed by the SINs and PVINs. The PVINs then innervate the PVPNs which, in turn, send axons to premotor structures (Del Bene et al., 2010; Nevin et al., 2010). Sketches of PVNs and SINs are based on those in Nevin et al, 2010.

3 Habituation to looming stimuli in zebrafish larvae

3.1 Contributions

G. Vanwalleghem, I. A. Favre-Bulle and M. A. Taylor contributed to the work of this chapter. G. Vanwalleghem assisted with the data analysis of the calcium imaging. I. A. Favre-Bulle assisted in building the behavioural rig. I. A. Favre-Bulle and M. A. Taylor built and maintained the SPIM microscope. E. K. Scott assisted conceptualising the experiments and editing the associated manuscript.

3.2 Summary

Habituation, a simple form of learning defined as a diminishment of an innate response to a frequently repeated stimulus, is critical to our everyday focus and attention, and the neural mechanisms underlying habituation remain unclear. Zebrafish larvae show habituation to visual and auditory stimuli and provide an appealing platform from which to study habituation’s circuit-level mechanisms. When presented with a looming stimulus that resembles an approaching predator, larvae respond with a rapid escape behaviour. In this study, I explored whether zebrafish larvae are capable of habituation to repetitive looms, and if features of the stimulus modulate the habituation dynamics. I also used a SPIM microscope and calcium imaging to visualise neuronal activity and localise the regions associated with this learning behaviour. In the behavioural experiments, I used a screen presenting looms below an arena with freely swimming larval zebrafish. Our results confirm that they habituate to visual threats, as they gradually decrease the probability of startle responses during a train of looming stimuli. Neuronal functional imaging experiments with GCaMP6s zebrafish larvae helped me uncover the neural circuits and patterns of activity responsible for this habituation. These experiments involved the simultaneous observation of activity across thousands of neurons spanning the entire brain early and late in the habituation process, and the results showed populations with different habituation profiles across the brain. Graph theory analysis helped me observe the network dynamics during habituation and how stimulus features affect these networks. The overall results point to a disengagement of part of the network as habituation occurs, mediated by integrating elements in the optic tectum, while a core visual circuit only habituates weakly.

3.3 Introduction

Habituation is a simple form of non-associative learning that is present in a wide range of animals (Harris, 1943; Christoffersen, 1997). It consists of the decrease of an innate response to a stimulus after multiple presentations. Habituation has been studied extensively both behaviourally and at a synaptic level. This has led to a consensus on the main characteristics of habituation that distinguish it from other forms of learning and sensory adaptation or fatigue (Thompson and Spencer, 1966; Rankin et al., 2009; Thompson, 2014). One of these key characteristics is a spontaneous recovery after a period of rest when the stimulus is absent (Thompson and Spencer, 1966). Another, referred to as dishabituation, is the recovery of the response after a different stimulus is presented (Rodgers et al., 1963; Rankin et al., 1990; Giles and Rankin, 2009; Faria et al., 2019). Finally, another way to differentiate habituation from other phenomena is by being able to modulate its rate with changes in the features of the stimulus (e.g. intensity of frequency) (Russell, 1967; Askew, 1970; Rooney and Laming, 1986; Laming and Mckinney, 1990; Rankin and Broster, 1992; Post and von der Emde, 1999; Gray, 2005; Giles and Rankin, 2009; Fischer et al., 2011).

Although habituation has been extensively characterised behaviourally, many of its neural mechanisms remain unclear. Prior work has focussed on the synaptic and molecular changes that underlie habituation and the reciprocal phenomenon, sensitisation. Studies in Aplysia and C. elegans have linked habituation to a decrease in the neurotransmitter released by the presynaptic neuron (Castellucci et al., 1970; Castellucci and Kandel, 1974; Gover and Abrams, 2009), resulting from silencing at the release sites (Gover et al., 2002), probably mediated by potassium channels (Engel and Wu, 1998; Typlt et al., 2013; Zaman et al., 2017). This presynaptic depression is a general effect of habituation and provides an explanation specifically for short-term habituation. Long-term habituation, which needs protein synthesis and can last multiple hours, is modulated by AMPA and NMDA receptors (Rose et al., 2002; Ezzeddine and Glanzman, 2003; Glanzman, 2009). So far, despite the significant progress, studies in Aplysia have led to a better understanding of sensitisation than habituation (Glanzman, 2009, 2010). However, exploiting the simple nervous systems of invertebrates, studies in C. elegans allowed further progress in the understanding of habituation. Its well-known neuroanatomy and genetics make it an ideal model for the research of the fundamental mechanism underlying habituation. After a first characterisation of this learning behaviour and its modulation (Rankin et al., 1990; Rankin and Broster, 1992), subsequent research was able to study the circuits implicated in habituation (Wicks and Rankin, 1997) and also showed that it is mediated by multiple mechanisms (Ardiel and Rankin, 2008; Giles and Rankin, 2009; Bozorgmehr et al., 2013). The continuous technological progress combined with the genetic editing possible in C. elegans keeps

revealing molecular factors involved, like the relevance of dopamine and peptides (Ardiel et al., 2016; Ardiel et al., 2017; McDiarmid et al., 2018b; McDiarmid et al., 2018a).

Due to their small size, external development, and optical transparency, the zebrafish has emerged as a useful vertebrate model system for studying sensory processing and sensorimotor gating (Vanwalleghem et al., 2018). Zebrafish habituate to auditory and visual stimuli, the mechanisms of which have implicated NDMA receptors and protein synthesis (Best et al., 2008; Roberts et al., 2011; Wolman et al., 2011; Roberts et al., 2016). Furthermore, studies in zebrafish have also shown the role of individual variability, circuitry, genetic factors and molecular modulation in habituation (Wolman et al., 2014; Marsden and Granato, 2015; Wolman et al., 2015; Pantoja et al., 2016; Jain et al., 2018; Jordi et al., 2018; López-Schier, 2019; Randlett et al., 2019).

Collectively, these studies have thoroughly mapped the relationships between stimuli and behaviour during the process of habituation, and have provided important insights into some of the synaptic and cellular processes that are affected. Other work has addressed activity across the whole brain during habituation using fMRI (Kleinhans et al., 2016). Little is known, however, about the intervening scales of local microcircuits and regional networks through which sensorimotor transformation changes as habituation proceeds. These networks, and the ways in which they change during habituation, can only be addressed by observing activity in whole populations of neurons at single- cell resolution. Among vertebrate model systems, zebrafish larvae are uniquely suited to this approach, and recent studies have succeeded in recording activity in individual neurons, using fluorescent calcium indicators, across the entire brain (Ahrens et al., 2013a; Wolf et al., 2015; Chen et al., 2018b; Vanwalleghem et al., 2018). This offers the prospect of studying habituation in cellular microcircuits and the networks that they form.

To work with this approach in zebrafish that is just a few days old requires a strong innate behaviour that is still subject to habituation. Visual escape behaviour has previously been shown to habituate in different types of fish (Laming, 1987; Rooney and Laming, 1988). As such, a looming visual stimulus, which simulates an approaching predator and drives escape behaviour in larval zebrafish, provides an appealing option for studying habituation. The behavioural responses to looms, the features of the stimulus driving escape behaviour, and the key brain structures involved in processing and producing the escape responses in zebrafish have recently been described (Temizer et al., 2015; Dunn et al., 2016a; Yao et al., 2016; Bhattacharyya et al., 2017; Heap et al., 2018b). The visual escape circuit was described in detail in the previous chapter. Briefly, previous studies suggest that visual information from the retina travels to the tectum, directly or through the thalamus, and that the tectum is a likely centre for integrating the multiple features of visual information and calculating that a 41

threat is imminent (Temizer et al., 2015; Dunn et al., 2016a; Heap et al., 2018b). Tectal projection neurons then send this information to the Mauthner cells and their segmental homologs (M-system) to produce a C-bend startle response (Sato et al., 2007; Dunn et al., 2016a; Yao et al., 2016; Bhattacharyya et al., 2017). However, additional structures including the telencephalon, habenulae, pretectum, midbrain tegmentum, and hindbrain also respond to looms (Temizer et al., 2015; Dunn et al., 2016a; Chen et al., 2018b). Others, including the hypothalamus, modulate the visual escape responses in contexts other than habituation (Filosa et al., 2016; Yao et al., 2016; Lovett-Barron et al., 2017). These results suggest a broader brain-wide network for visual startle and its modulation than the core circuit comprising the retina, thalamus, and tectum, and leaves many possibilities for how and where key network changes take place during habituation.

In order to investigate the network of implicated brain areas and its changes while habituation learning takes place, I decided to use a graph theory analysis. This approach has become more popular in the last decade due to the fMRI development and has been used to study brain networks and its characteristics in human cognition and neurological disorders (Farahani et al., 2019). This approach establishes a network based on the similarity of the activity patterns (usually measured with a cross- correlation), which can then be analysed on its multiple graph measures (Zalesky et al., 2012; Sporns, 2018; Farahani et al., 2019). Special tools have been developed to facilitate that analysis (Rubinov and Sporns, 2010), which has allowed the investigation of complex zebrafish whole-brain resting and evoked activity (Betzel, 2018). These analyses are especially interesting for the analysis of dynamic networks, like in the case of development or learning. Graph theory has been used to research the zebrafish larvae optic tectum’s modularity, and its changes as the brain develops (Avitan et al., 2017; Triplett et al., 2018). A graph theory analysis at different time points of a network is a promising approach to investigate learning dynamics. Previously, an analysis of the cortical network changes while learning a motor skill suggested that the changes in the network modularity are indicative of learning dynamics. In particular, the flexibility of nodes from one session helps predict the learning progress of the next session (Bassett et al., 2011). Using this approach to see the differences in the brain networks as habituation occurs could help reveal the important elements and changes responsible for habituation.

In this study, I established a behavioural paradigm for habituation to visual loom stimuli and explored the temporal details of the stimulus that influence the rate and persistence of habituation. I then performed whole-brain calcium imaging at cellular resolution to identify the brain regions involved and how individual cellular responses change during the habituation process. These data feed into a

network model of the brain-wide visual habituation, including the changes in information flow that are proposed to underlie decreased behavioural responses to the repeated stimuli.

3.4 Preliminary results:

In order to design experiments suitable for testing loom habituation and the influence of the stimulus features on habituation rates, I explored and tested different paradigms both in the free-swimming behavioural set up and in calcium imaging experiments. This exploration allowed me to adjust stimulus parameters to what I finally considered an optimal experiment that could be performed in both set ups. In this section of the chapter I will present the results of this experiments as they are also informative of the habituation learning behaviour in zebrafish larvae.

3.4.1 Visual habituation in larval zebrafish

To explore the habituation response, I designed pilot behavioural experiments for habituation using the protocol from Wolman et al, 2011 as a guide for long term habituation in zebrafish larvae. This consisted of blocks of stimuli separated by resting intervals; in this case, 4 blocks of 10 min each (52 looms per block) with 5 min rests between them. I manually scored the first pilot studies, and their results suggested the presence of habituation but also showed that the lack of an escape response was sometimes due to freezing responses. I noticed that the larvae often froze for long periods of time and that these freezing responses presented some stereotyped features that included more frequent eye movements and sinking (Kalueff et al., 2013). This long freezing events were easier to distinguish from shorter ones, which could look like the intermittent swim bouts. Therefore, I started to categorise the responses as either escape response, a freezing response (not moving on the 10s before, during and 10s after a loom presentation) or non-response when the fish would not present either of the previous fear-related responses. In the second set of experiments, I also included a different stressor (poking with a micropipette) before a 5th block of 52 looms to test for a possible dishabituation of the fear responses (Figure 3.1a). To explore the possibility of better habituation responses at older stages, I tested the habituation in both 6dpf and 12dpf (Figure 3.1b). In both groups, the probability of the fish not responding with a fear response increased gradually. This supported the use of 6dpf larvae, which are standard for calcium imaging as a good trade-off between complex behaviours and the ability to remain healthy while immobilized for long periods in agarose. We can observe a small recovery of the fear responses after each period of rest (spontaneous recovery), and the results also suggest dishabituation after the introduction of a new stressor (tactile stimulation, Figure 3.1a, b). However, a caveat of these experiments is the need for manual scoring. This could make the experiments more sensitive to subjective errors and limit the quantification of the behaviours.

Therefore, to do an unsupervised categorisation of the escape responses and being able to analyse the behaviour in greater depth (like changes in distance and speed), I implemented the use of tracking software. This also allowed collecting and analysing multiple larvae while measuring relevant parameters. I adapted the behavioural rig with an infrared filter to filter the light from the screen below the arena. This allowed the tracking of the fish with computer software like Ctrax (Branson et al., 2009) (Figure 3.1c) and Zebralab (Viewpoint Life Sciences, France). I used these tools on a behavioural experiment to test if the interstimulus interval would affect the habituation rate based on a short habituation protocol from Marsden and Granato (Marsden and Granato, 2015). This experiment was done in 8 larval zebrafish and consisted of a first resting period of 5 min followed by 5 looms with an interstimulus interval (ISI) of 120 seconds and then 4min of looms with 12s ISI. This train was presented 3 times to also test for inter-session habituation. Once again the results suggest a decrease in escape responses (Figure 3.1d) and also indicate that more frequent looms produce faster habituation, as has been described in the literature (Rankin et al., 2009).

Finally, I designed stimulus trains that could also be tested in calcium imaging experiments, these needed to be shorter but still elicit habituation (e.g. dishabituation, spontaneous recovery). I presented 30 looms either at fix rate of 30s or with variable intervals with an average ISI of 15s. Both experiments included 3 repeated 300Hz sound stimuli that were played before the 21st loom to produce dishabituation. The variable presentation of looms at 15s shows that the fish habituates rapidly to the looms, and that the lack of periods of rest diminishes the probability of recovery (Figure 3.1e). On the other hand, presenting 30 looms at a regular 30s ISI produces escape responses that gradually habituate, and that present recoveries after the periods of rest (Figure 3.1f). However, dishabituation by the sound stimuli was not clear.

Figure 3.1. Preliminary behavioural habituation experiments with zebrafish larvae. a. Probability of an escape, freezing or not responding to a looming stimulus by 6dpf larvae. Note the decrease in the probability of no response after another stressor is introduced (block 21). b. Same experiment as in a, but with both 6dpf and 12dpf larvae. c. Larvae being tracked with Ctrax software. The larvae swim in individual agar arenas above a screen that projects individual looms. d. Probability of fish presenting fast responses (≥30mm/s) at 120s (in blue) and 12s ISI (n=8). e. Probability of fish presenting fast responses to 15s ISI looms. f. Probability of fish presenting fast responses to 30s ISI looms. Light grey bars represent periods of rest of 5 min for all graphs except for f, where the rest was for 2 min. 45

3.4.2 Locate visual habituating neuronal responses.

Using a diffusive digitally scanned light-sheet microscope (Taylor et al., 2018), I was able to collect visual responses to the looms while imaging the whole brain of 6dpf larvae. I imaged 14 fish, presented with a movie of 10 looms separated by 30s intervals during which the screen progressively faded back to white to avoid responses to the increase in luminance. In those fish, the tail was also freed to record behavioural responses.

To explore the spatial feature of the responses, I tested different analysis paradigms. I first performed a pixel-based analysis, where I used a maximal intensity projection of the acquired optical slices of each fish, and then used linear regression on each pixel with a regressor simulating GCaMP responses to each of the looms. This analysis was useful to find general areas that responded to looms (Figure 3.2). Manually selecting ROIs outlined by this pixel selection showed different patterns of responses. However, this method is limited as it is not segmenting the neurons, and the 3D information is lost.

To better analyse the data from the calcium imaging experiments, I processed the images with computational methods developed for similar data sets (Pnevmatikakis et al., 2016b). The constrained non-negative matrix factorisation (CNMF) allow us to identify regions of interest (ROIs) and denoise the calcium traces. In the next step, I used a linear regression analysis with simulated calcium spike responses to the loom stimuli to select all the ROIs with a coefficient of determination (r2) higher than 0.1. With these selected ROIs, I then applied the K-means clustering method to group similar response profiles. Then, I selected the clusters of interest (the ones with responses to the visual stimuli), and used their mean responses as regressors for another linear regression with all the ROIs. To avoid the inclusion of noise, I increased the selection criteria for the ROIs to an r2>0.5. I clustered these ROIs again to divide them into different response profiles and map them back into the brains of the individual fish.

The data analysed in this manner was more comprehensive and suggested different response patterns (Figure 3.3). The blue cluster represents areas that show little to no habituation. The pink group present a gradual intensity decrease, but a response to every stimulus and finally clusters green and red seem to stop firing after a few presentations. These results support the idea that both habituated and non-habituated responses could be found in the zebrafish larval brain. In the fish presented as an example, we can observe many interesting results (Figure 3.3). One of them is that many of fast habituating ROIs (Red) are located in both hemispheres while the others seem to be predominately contralateral to the loom presentation. It is also interesting that the majority of gradually habituating responses (pink) are mostly located in the tectum. Furthermore, it is also relevant that the pretectum

seems to have a population of non-habituating responses (blue) and that some of these neurons can also be found in the most dorsal part of the pallium. This suggest that different parts of the visual pathway continues to response to the visual stimuli independently of the habituating responses in other areas. Finally, another interesting result is the presence of habituating neurons in what is possibly the medial division of the dorsal pallium (Dm). This suggests that the Dm responds and habituates to threatening stimuli and supports the idea that the Dm plays a similar function than the mammalian amygdala.

To test for characteristic features of habituation, like the recovery after a period of rest and dishabituation, I performed calcium imaging experiments with variable loom presentations (average 15s), no periods of rest and that included a sound presentation to produce dishabituation. The analysis of this experiment showed at least two types of habituating responses and a sound responding group. (Figure 3.4a). An analysis with regressors that included responses to both the visual and auditory stimuli also found multisensory responses. However, the dishabituation to the visual stimuli was not clear, which could be due to a strong habituation caused by a high frequency and no periods of rest. Parallel to these experiments, I also tested periods of 5 min rest after a block of ten looms (presented in the Results section below). The results of both series were important to confirm the relevance of periods of rest to produce the recovery of the response.

The previous behavioural and calcium imaging experiments were useful to design a stimulus train that could be used in both free-swimming and whole-brain imaging experiments. With these first experiments I confirmed that a relatively small number of looms was enough to produce habituation, that periods of rest produced spontaneous recoveries, and that different ISI influenced the habituation rate (Figure 3.1-3.4). I also confirmed that I could locate different profiles of responses across the brain using segmentation and clustering techniques. With this information, I then designed and performed experiments manipulating the stimulus features to explore the habituation dynamics in the brain. These are presented in the rest of the chapter.

Figure 3.2. Calcium imaging and behavioural responses to looms. a. 2D reconstruction of a calcium imaging stack of an example fish. A maximum projection of the 3D stack with a mask of pixels which had an r2> 0.3 to loom responses. The mask is colour-coded by the number of times a response to a loom has a beta coefficient higher than 2SD. 1, 2 and OT are pointing at ROIs in which calcium trace is shown in B. b. Example of a tail movement. c. Calcium traces of responsive ROIs in the OT and the Dm. Dm2 presents calcium spikes during tail movements (asterisk). Y-axis are the z-scored fluorescent traces and the x-axis is time.

Figure 3.3. Location of loom responsive and habituating ROIs a. Dorsal view of a 3D reconstruction with ROIs that responded to loom presentations in an example fish. 1. A population of pink ROIs in the optic tectum. 2. Blue ROIs in the pretectum. 3. Green and red ROIs in the Dm. b. On the left is the mean of the calcium traces of the ROIs that formed each of the 4 clusters. The y-axis represents the fluorescence normalised as a Z score. On the right are the raster plots of each of the clusters, which are heat maps of each individual ROI calcium traces. These ROIs belong to 7 fish. The x-axis for both plots is time (1=0.5s).

Figure 3.4. Calcium traces of ROIs responding to loom presentation and sound. a. Left. Mean of the traces of ROIs responding to looms (blue and red) and to sound (green). Right. Raster plots of each of the clusters which have a heat map of the individual ROIs calcium traces. b. Left. Mean of the calcium traces of the ROIs that responded to both visual stimuli and sound. Right. Raster plots of each of the clusters. The y-axis represents the fluorescence normalised as a Z score. The x-axis for both plots is time (1=0.5s).

3.5 Methods

Animals. All zebrafish (Danio rerio) work was performed in accordance with The University of Queensland Animal Welfare Unit (approval SBMS/378/16). Adults were reared and maintained in a Tecniplast zebrafish housing system under standard conditions using the rotifer polyculture method for early feeding 5 to 9 days post-fertilisation. For the visual habituation experiments with different stimulus trains, I used nacre zebrafish embryos of the TL strain expressing the transgene, elavl3:H2B‐ GCaMP6s (Chen et al., 2013). All fish were produced by natural spawning and reared in Petri dishes with embryo medium (1.37 mM NaCl, 53.65 µM KCl, 2.54 µM Na2HPO4, 4.41 µM KH2PO4, 0.13

mM CaCl2, 0.16 mM MgSO4, and 0.43 mM NaHCO3 at pH 7.2) at 28.5 °C on a 14-hour light: 10- hour dark cycle.

Stimulus train for behavioural experiments. The stimulus train consisted of three blocks of 10 looms with 5min of rest (with a white screen) between each block. The loom was initiated with a dot that started expanding after 1s. The minimum angle of the loom was ~11° of the larva field of view, and the maximum angle of the loom was ~90°. The fast looms reached their maximum angle in 2s, and the slow looms in 4s. This was followed by 2 seconds of a black screen and a 9s slow fade back to white, designed to avoid any neural OFF responses to changes in luminance. The screen remained white until the next loom initiation for a variable duration, depending on the desired inter stimulus intervals (ISI). This was of 18, 20, or 22s for the f20 and s20 paradigms, and 54, 60, or 66s for f60 and s60. A sound stimulus of 300Hz at ~85 dB was played 3 times for 1s with 1s ISI. The first presentation of the sound was 25s before the 21st loom. The video and sound were displayed by a monitor (10.1 1366x768 Display IPS + Speakers - HDMI/VGA/NTSC/PAL, Little Bird, Australia).

Behavioural experiments. Individual 6dpf larvae were placed in each of the 12 wells arena (circular plugs of agar were removed to produce the wells). The wells were filled with embryo medium and placed at 1cm above a screen inside a dark chamber, and all larvae received the same stimulus train. The chamber was kept in the dark but was illuminated with infrared LEDs, which are invisible to zebrafish larvae. A Basler acA1920 camera recorded the movements from above, a lens (40mm Thorlabs) and a 665nm longpass filter (FGL665 - Ø25 mm RG665 Colored Glass Filter, Thorlabs) delivered infrared light to the camera with a weak signal from the screen that confirmed the timing of the looming stimuli. Movements were tracked in bins of 1s using the zebrafish tracking Viewpoint software (ZebraLab, ViewPoint Life Sciences, France), tracking three speed categories: <0.5mm/s, 0.5-30mm/s, and >30mm/s. The output of the tracking was then analysed using a Matlab script. Escape responses were defined as one or more movements above 30mm/s during a loom presentation. Further statistical analysis and graphs were made in GraphPad Prism v7.04 and R 3.5.1 (R core team, 2018). The sound failed to produce a clear dishabituation, so this effect was not further analysed. The fitted curves were done in GraphPad Prism v7.04 with the exponential one-phase decay curve from the 1st to the 10th loom of each block, using a Least Squares regression and plateau to 0.

Larvae preparation for calcium imaging. Imaging was performed on 6dpf larvae that were embedded upright in 2% low melting point agarose (Sigma, A9045) and transferred to a 3D printed imaging chamber (Favre-Bulle et al., 2018). Imaging 51

chambers were filled with embryo medium once the agarose had set and the tail was freed (Thompson et al., 2016) so that escape responses could be monitored. The imaging chamber was composed of a 3D-printed base (24 × 24 mm) with four posts (3 × 3 × 20 mm) raised along the four corners of the platform. The four outward faces of the chamber were fixed with a glass coverslip (20 × 20 mm, 0.13- 0.16 mm thick). A glass window on the bottom of the chamber allowed the filming of tail movements (Favre-Bulle et al., 2018).

Loom stimulus train for calcium imaging. Looms were presented on a 75 × 55 mm LCD generic PnP monitor (1024 × 768 pixels, 85 Hz, 32-bit true colour) with an NVIDIA GeForce GTX 970 graphics card. The monitor was positioned 30mm to the right of the larvae and was covered by a coloured-glass alternative filter (Newport, 65CGA- 550) with a cut-on wavelength of 550 nm. The minimum angle of the loom was ~10°, and the maximum angle of the larvae field of view the loom covered was ~82°. The auditory stimulation (a 100Hz sound at 100dB before the 21st loom) was presented with two audio speakers (Logitech Z213) placed at ~20cm from the fish. The background noise level was 40dB.

Microscopy. elavl3:H2B‐GCaMP6s zebrafish larvae, individually mounted in the imaging chamber, were imaged on a custom-built SPIM microscope (Favre-Bulle et al., 2018; Taylor et al., 2018). To avoid stimulating the eyes with the light sheet, the side laser path of the SPIM was blocked, and the front SPIM plane was restricted to a space between the eyes using a vertical slit. Captured images were binned 4 times to a final resolution of 640 × 540 pixels at 16-bit in a tagged image file (TIFF) format. Fifty transverse sections at 5µm increments were captured and imaged at 2 Hz. Recording of the brain activity started 30s before the first stimulus onset and stopped after the return to white from the last loom of each block, resulting in three separated acquisitions. To image the larva and record its tail movements, a 4x 0.1NA Olympus microscope objective (PLN 4X) was placed below the sample chamber (Favre-Bulle et al., 2017), coupled with a tube lens projecting the image onto a Basler acA1920 camera, recording at 30 fps.

At the end of each experiment, a single high-definition scan of non-binned images was recorded with 100ms exposure time and 2µm increments to be used for the registration of the brain of each fish (see below).

Analysis of calcium imaging data. Calcium imaging data from the three acquisitions were concatenated in ImageJ v1.52c as a combined time series and then separated into individual slices (50 planes per fish). Motion correction was

performed using Non-Rigid Motion Correction (NoRMCorre) algorithm (Pnevmatikakis and Giovannucci, 2017), and fluorescence traces were extracted and demixed from the time series using the CaImAn package (version 0.9)(Pnevmatikakis et al., 2016a; Giovannucci et al., 2019) (http://github.com/flatironinstitute/CaImAn). I used 4000 components per slice to ensure that we would not miss any ROIs during the initialisation step of CaImAn. The risk of over-segmentation was mitigated by a merge step using a threshold of 0.8 to merge overlapping ROIs. The order of the autoregressive model was set at 1 to account for the decay of the fluorescence, our acquisition speed being too slow to account for the rise time. The gSig (half-size of neurons) was set at 2, based on estimates of the sizes of the nuclei in our images. I did not use any temporal or spatial down-sampling and the initialisation method was ‘greedy_roi’. The components were updated before and after the merge steps, empty components were discarded, and the components were ranked for fitness as previously (Pnevmatikakis et al., 2016a).

Analysis of whole-brain activity data. For the experiment with four stimulus trains, the resulting ROIs and fluorescent traces from the CaImAn package were pooled from larvae of each stimulus train (n of the 4 datasets: f20=11, f60=8, s20=10, s60=10), and then z-scored per dataset. A k-means clustering by cityblock distance with 50 components and 5 replicates was done for each dataset. Clusters were manually selected based on their profile responses to the looms or sound and their presence across datasets and individual fish. This produced 7 clusters selected from the f20 and s20 datasets and which were used as regressors for subsequent analysis of the four datasets: three strongly habituating, a moderately habituating, a weakly habituating, an inhibited and a sound responsive clusters. All ROIs from each of the 4 datasets were modelled by linear regression to each of these regressors. As the 60 sec ISI time series were longer, these time series were trimmed around the 30 looms to perform the linear regression. ROIs with an r2 value higher than 0.3 were then selected for further analysis. The selected ROIs were categorised by correlation to each of the 7 selected regressors. I confirmed that the clusters could be found in most or all larvae, but 3 fish (1 from f20 and 2 from f60) were discarded because their ROIs contribution to one of the habituating clusters was above 50% of the total number of ROIs for that cluster, so they were deemed as outliers in terms of responsiveness. To find the motor evoked calcium responses, I first used ImageJ to detect the tail movements from the behavioural imaging. I used a polygon ROI covering half of the fish’s tail to extract the mean grey values of the time series. Substantial tail movements produced large peaks and were flagged as movement events. I then build regressors for individual larvae inserting a stereotypical GCamp6s trace to the movement timing for

each larva. Finally, I used a linear regression with the motor regressor of each larva as for the habituating clusters and selected ROIs with an r2 value higher than 0.2.

For the t-SNE (van der Maaten and Hinton, 2008) (Figure 3.7g and Figure 3.8) I used the Matlab function with a correlation-based distance and the following parameters: Perplexity=100, Exaggeration=20. For further analysis, I pooled together the three strongly habituating clusters and I excluded the sound response cluster, resulting in four main clusters.

To calculate the proportions of ROIs for a given cluster that appear in each brain region (Figure 3.12b), we divided the number of ROIs of each cluster in each brain region by the total number of ROIs of that cluster in the whole brain. I did this for each larva, created a mean for each dataset, and then averaged these values across all four datasets. For the analysis of the normalized responses in the tectum (Figure 3.12c), a mean of the tectal ROIs’ response for each cluster was calculated for each fish, then the maximum response per loom was calculated based on the maximum z-score value in the window of the loom presentation adjusted by the baseline before each loom. These values were normalised to the first loom response, and a mean of the normalised maximum response was calculated for each dataset. To compare these tectal responses with the matching behavioural results we used the Pearson correlations coefficients.

To locate the subset of strongly habituating neurons that are involved in motor behaviours (Figure 3.10), I calculated the Spearman correlation coefficient between each strongly habituating ROI from the f20 dataset and the motor regressor of its respective fish. I then selected the ROIs above a correlation coefficient of 0.3066 (the mean=0.1522, plus one SD=0.1544). Finally, we calculated their proportion compared to the strongly habituating ROIs of each of the brain regions previously analysed.

The analysis was done using Matlab R2018b and GraphPad Prism v7.04.

Correlation matrices and graphs. For graph theory, I simplified the system from the 144,709 responsive ROIs while preserving the functional identity and anatomical location of the responses. To do so, I performed a k-means clustering on the 3-dimensional spatial coordinates of the ROIs (Betzel, 2018) of each functional cluster, in each brain region, with k number of clusters. k was defined based on the number of ROIs. For regions with fewer than 200 ROIs, no node was placed; between 200-500, 1 node; between 500- 1000, 2 nodes, between 1000-3000, 3 nodes; and >3000, 4 nodes. This was intended to strike a balance between including relatively sparse populations that may, nonetheless, make functional contributions, and weighting our analysis to some degree toward more abundant response types. This

produced 102 nodes, but we discarded three nodes that had three or fewer fish contributing to them. For the remaining 99 nodes, I cross-correlated the mean loom response of their ROIs and generated individual matrices for each larva and each loom presentation. I then averaged the matrices of each dataset across larvae. To identify the network most similar to the 11th trial of the f20 and f60 datasets, I performed a correlation between the matrices of the first 10 looms and the 11th loom of the relevant dataset and identified the loom with the highest Pearson correlation coefficient.

I used the Brain Connectivity Toolbox (Rubinov and Sporns, 2010) to perform the graph analysis. I first generated weighted connectivity matrices and filtered out edges with an absolute correlation value below 0.75. I then subtracted each of the f20 loom matrices from the f60 matrices. The width and colour of the edges are indicative of the subtraction weight. The participation coefficient was calculated between the four functional clusters identified previously (strongly habituating, moderately habituating, weakly habituating, and inhibited).

Registration to a reference brain. I used Advanced Normalization Tools (ANTs, https://github.com/ANTsX/ANTs) to register our results on the H2B-RFP reference of Zbrain (Avants et al., 2008; Avants et al., 2011; Randlett et al., 2015). The high definition stacks were used to build a common template, before registering this template to the Zbrain atlas (Favre-Bulle et al., 2018). The resulting warps were sequentially applied to the centroids of extracted ROIs to map them all in the same frame of reference. The Warped ROI coordinates were then placed in each of the 294 brain regions defined in the Zbrain atlas (Randlett et al., 2015).

Data visualisation. I used Unity to represent each ROI centroid as a sphere. Their diameter was adjusted based on the number of ROIs to be able to visualise the different clusters (Strongly habituating=2; Moderately habituating=3, Weakly habituating=4; Inhibited=6). An isosurface mesh of the zebrafish brain was generated from the Zbrain masks for the diencephalon, mesencephalon, rhombencephalon, telencephalon and eyes using ImageVis3D (Institute). The mesh was imported in Unity and overlaid to the ROIs.

The colormaps were generated using two Matlab® functions: The cbrewer function, https://au.mathworks.com/matlabcentral/fileexchange/34087-cbrewer-colorbrewer-schemes-for- matlab (Accessed on May 2019) which includes specifications and designs developed by Cynthia Brewer (http://colorbrewer.org/), and the MatPlotLib 2.0 default colormaps ported to Matlab,

https://au.mathworks.com/matlabcentral/fileexchange/62729-matplotlib-2-0-colormaps- perceptually-uniform-and-beautiful (Accessed on May 2019).

Figures were produced using Matlab R2018b and GraphPad Prism v7.04, and assembled in Adobe Illustrator CS6.

3.6 Results

3.6.1 Habituation of visual escape behaviour in larval zebrafish

To characterise the escape behaviour of larval zebrafish exposed to looming stimuli, I designed a 12- well apparatus in which each well contained a larva receiving its own loom stimulus from below (Figure 3.5a). I presented looms in blocks of 10, with five minutes between blocks and an auditory tone at the end of the second rest period (for dishabituation before the 21st loom stimulus). To explore the relationships between stimulus properties and behavioural habituation, I used looming stimuli of two expansion speeds (a fast stimulus that filled the bottom of the well in 2 sec and a slow stimulus taking 4 sec) and two inter-stimulus intervals (ISIs) of 20 or 60 sec between looms. This resulted in four stimulus trains: f20, f60, s20, and s60 (Figure 3.5b).

Each led to habituation of loom-elicited startle responses (Figure 3.5c), and two patterns arose across the four stimulus trains. First, the slow-growing stimuli led to stronger habituation than the fast stimuli did, especially in the first block of 10 looms. Second, the stimulus trains with 20 sec ISIs produced faster habituation within blocks, but the habituation produced by trains with 60 sec ISIs showed less recovery after the 5-minute rest periods. A multiple linear regression of the first block indicated a significant effect of the loom presentation number (beta= -0.052918, p= 1.83e-11) on response probability, confirming habituation. The loom speed also affected response probability strongly (beta= -0.245827, p= 3.36e-09), as well as a weaker but significant impact from the ISI (beta= - 0.075312, p= 0.0228). Together, the speed, ISI, and presentation number explain almost 80% of the variance (Adjusted R2=0.7991, F(3,36)=52.7, p= 2.994e-13). These effects are consistent with past studies in zebrafish and other diverse model systems (Rankin and Broster, 1992; Best et al., 2008; Fischer et al., 2011; Wolman et al., 2011), suggesting a relationship between stimuli features and habituation behaviour that is broadly conserved. Explaining this relationship requires an exploration of the underlying circuitry and the ways in which it changes during habituation.

Figure 3.5. Modulation of habituation by stimulus features. a. Schematic representation of our setup for measuring visual habituation behaviour. A 12-well chamber with one larva in each well (top right) was filmed on a horizontal screen (left) on which the looms were presented. Automated tracking recorded periods of swim bouts (green) and burst swim (red) for each larva (bottom right) b. Stimulus train properties across the 4 experimental groups. c. Probability of response across the 4 groups during three blocks of ten loom presentations. d. Smoothed curves of the response probability for each group.

3.6.2 Brain-wide characterisation of neural activity during habituation

I performed whole-brain imaging of the elavl3:H2B‐GCaMP6s line (Chen et al., 2013) using selective plane illumination microscopy (SPIM) as previously described (See Methods). For each larva, this produced 50 dorso-ventral planes, at 5μm intervals, spanning the rostro-caudal and medio-lateral extents of the brain, with a volumetric acquisition rate of 2Hz. I performed morphological segmentation of these images to identify regions of interest (ROIs) generally corresponding to individual neurons, and extracted fluorescent traces from these ROIs, as described before (See Methods).

Snapshots of responses across the brain during this repetitive stimulation (shown for f20 in Figure 3.6a-c) show a sharp decrease in responsive ROIs between the first and second stimuli, and a further drop in responses by the 10th stimulus. Figure 3.6d and 6e show the response of each ROI in the second and 10th trial as a proportion of its response in the first. Habituation is conspicuous across all loom-responsive brain regions, including the tectum, thalamus, medial hindbrain, tegmentum, and telencephalon, suggesting involvement for these regions in the habituation process.

To address these possible mechanisms, I used k-means clustering to identify seven categories (clusters) of loom-responsive neurons with distinct functional properties (and one auditory cluster). Based on their highly similar response properties, I merged three clusters of ROIs showing strong and rapid habituation (Figure 3.7) into a single strongly habituating cluster (Figure 3.6f, g). I characterised the remaining three clusters as moderately habituating, weakly habituating, inhibited, and I also located a motor-associated group of ROIs using regressors customised to each animal’s movements (Figure 3.6f, g). A t-SNE analysis (van der Maaten and Hinton, 2008) (Figure 3.8) shows functional segregation among these clusters, suggesting that they are distinct categories of loom- responsive neurons. A detailed distribution of the five functional clusters is shown in Figure 3.9.

Figure 3.6. The activity of individual ROIs and their functional clusters during habituation. a. Responses of ROIs across the brain to a looming stimulus, colour coded for the intensity of their response. b, c. The same ROIs’ responses to the second and tenth looms. d, e. The degree of habituation in each of these ROIs in the second and 10th trials, calculated as the ratio of response to the first loom. This analysis was restricted to ROIs showing clear responses for the first loom stimulus (with a coefficient of determination (r2 value) >0.5 for the linear regression between their response and a regressor simulating a calcium signal). Raster plots (f) and mean responses (g) of the ROIs composing each of five functional clusters. h. Anatomical locations for the ROIs belonging to each functional cluster. Since different animals startled in different trials, I identified the motor cluster using a different regressor for each animal. The mean responses are shown for a single animal in this

cluster in g, with yellow lines indicating the relevant neurons from that animal in f. Distribution of these clusters is detailed in virtual sections in Figure 3.9. Data show the pooled responses of 11 larvae to the f20 stimulus train. Relevant anatomical brain regions are indicated in the bottom right corner of (h), each shown for only one side of the brain. Pallium, Pal; subpallium, Sp; thalamus, Th; habenula, Hb; pretectum, Pt; tectum, Tec; tegmentum, Tg; cerebellum, Cb; and hindbrain, HB. R, rostral; C, caudal.

Figure 3.7. Similarities among three types of fast habituating ROI. a. The average responses of the ROIs composing three of the clusters produced by k-means. All are powerfully responsive the first loom stimulus, with strongly attenuated responses in the second and third trials (using the f20 stimulus train). Responses are essentially absent from trials 4-30 across three blocks of ten stimuli each. b. The response strengths and temporal dynamics are similar across these three groups during the first three trials, although fast habituating cluster #1 shows the sharpest response profile. c. The anatomical locations of these ROIs are indicated, showing a high degree of overlap in their distributions (with clusters shown individually in d-f). Finally, a t-SNE analysis (g) fails to reveal clear functional distinctions among these groups, with extensive overlap and intermingling across these three clusters, especially for the Fast Habituating 2 cluster (green). These functional and anatomical analyses form the basis for pooling these three groups into a single “fast habituating” cluster in our subsequent analyses.

Figure 3.8. t-SNE analysis of the five functional clusters.

The calculated distributions of ROIs belonging to five functional clusters are represented following a t-SNE analysis. The motor associated cluster was not included because the ROIs in this cluster show different patterns of activity in different fish.

Figure 3.9. Anatomical distributions of five habituating clusters.

For each functional cluster shown in Figure 3.6, a dorsal view, lateral view, and four virtual coronal sections are shown. The ranges included in the coronal sections, scale bar and axis are indicated in the strongly habituating images at the top.

Strongly habituating ROIs are spread across several brain regions (Figure 3.6h and Figure 3.9), most prominently in the tectum, thalamus, medial hindbrain, pallium, and tegmentum. In the hindbrain, these ROIs are concentrated in a longitudinal rostro-caudal strip along the pathway of the tectobulbar projections, meaning that they likely include reticulospinal premotor neurons (Sato et al., 2007; Helmbrecht et al., 2018).

Moderately habituating ROIs are tightly concentrated in the central region of the tectal periventricular layer (PVL) of the left tectum (Figure 3.6h and Figure 3.9). This laterality is unsurprising, since the stimulus was presented to the right eye, and since all retinal projections are contralateral in zebrafish larvae. This position is consistent with a role for the associated neurons in the spatially registered processing of visual information, and their decreased responses may represent an important element of the overall circuit’s reduced responsiveness during habituation.

Weakly habituating ROIs are prominent in the tectum, habenulae, pretectum, and pallium (Figure 3.6h and Figure 3.9). There is moderate laterality toward the contralateral side to the stimulus in most of these regions. In the pallium, responses are concentrated around the dorsal edge of the pallium in what will likely become the lateral division of the dorsal pallium (Dl), although they also concentrated the medial division (Dm, Figure 3.9).

Inhibited ROIs are rare and mostly localised to the contralateral tectum and rostral thalamus (Figure 3.6h and Figure 3.9). Motor-associated ROIs are concentrated in the cerebellum. However, some can be found in the anterior and lateral hindbrain, and small numbers occur in the thalamus and pallium (Figure 3.6h and Figure 3.9). These ROIs are presumably involved in the coordination and delivery of the escape responses.

To analyse the possible relationship of the strongly habituating neurons and motor responses, I searched within the ROIs of this cluster for the ones that highly correlated with the movements of its respective fish (Figure 3.10). Results show that the highest proportion of strongly habituating ROIs involved in movements is located in the hindbrain, followed by other motor-related structures (cerebellum and pretectum).

Finally, an exploration of the sound stimulation shows responses in expected areas, like the medial octavolateral nucleus, the torus semicircularis and thalamus (Vanwalleghem et al., 2017) (Figure 3.11a,b). Interestingly, there are also sound responding ROIs in the telencephalon. When searching for specific multisensory ROIs using a regressor with responses to looms and sound, we found responses in the medial hindbrain, tegmentum, tectum, thalamus and Dm (Figure 3.11a,c).

Figure 3.10. Prominence of motor-correlated strongly habituating ROIs in the hindbrain. a. The distribution of all strongly habituating neurons across the brain for the f20 stimulus train. b. The subset of the ROIs from (a) that show >1s.d. correlation with motor responses during loom stimuli on a trial-by-trial basis. c. The proportion of strongly habituating ROIs that shows this motor correlation, by brain region.

Figure 3.11. Sound and multisensory responsive ROIs. a. Raster plot showing the traces of ROIs that responded to the sound presentation or to both the sound and loom presentations. The colour-code represents the strength of the response normalised (Z score). The time scale bar at the bottom applies to both raster plots. b. Top view of the localisation of sound responding ROIs. c. Top view of the localisation of multisensory ROIs.

3.6.3 Temporal stimulus properties modulate brain-wide responses

By looking at the changes in these distributions across our four stimulus trains (Figure 3.5), I next aimed to characterise brain-wide activity under conditions that lead to different rates and persistence of habituation in free-swimming larvae. The fundamental brain-wide habituation network was conserved across these treatments, but specific functional differences emerged (Figure 3.12a). One was a greater number of strongly habituating ROIs in the hindbrain strip for the s20 and s60 experiments (Figure 3.5). Another came in experiments with 60 sec ISIs, where I observed a greater number of weakly habituating ROIs in the dorsal hindbrain on the side contralateral to the stimulus (Figure 3.12). This could reflect the stronger preservation of habituation across breaks in the 60 sec ISI experiments.

Since a large proportion of loom-responsive ROIs is in the tectum, especially for the moderately habituating cluster (Figure 3.12b), I next looked at the relationship between the stimulus train’s properties and the responses of each functional cluster in the tectum (Figure 3.12c). This revealed only subtle difference across the stimulus trains for the response profiles of fast habituating neurons. 67

For moderately habituating neurons, differences arose with intriguing parallels to the behavioural outputs. In experiments with 60sec ISIs, habitation is slower, and recovery is less dramatic than for 20sec ISI experiments. In experiments that used slow stimuli, habituation occurs faster than in the corresponding experiments with fast loom stimuli. For weakly habituating neurons, experiments with 60sec ISIs lead to less habituation throughout the experiment, while other correlates of behaviour are less clear. Overall, moderately habituating ROIs repeatedly had the strongest correlation to escape probability in free-swimming experiments (Pearson correlation values: f20= 0.6896; f60= 0.6896; s20= 0.6368; s60= 0.7724), suggesting that among our functional clusters, it is the moderately habituating ROIs in the tectum whose dynamics most closely reflect behaviour.

Figure 3.12. Brain-wide responses during different loom stimulus trains. a. Brain-wide distributions of the functional clusters from Figure 3.6 for each of four loom habituation paradigms. b. The proportions of ROIs from each functional cluster located in the indicated brain regions. c. The average response profiles for fast habituating (left) moderately habituating (centre), and weakly habituating (right) tectal ROIs in each of the habituation paradigms.

3.6.4 Network Modelling of Visual Loom Habituation

As an approach to modelling visual loom processing and the network changes that produce habituation, I applied graph theory to the brain-wide activity data. To generate a tractable dataset and allow comparison across fish, I downsampled the 144,709 responsive ROIs into 99 nodes that represent the ROIs’ functional clusters and anatomical locations. From these 99 time series, I produced matrices representing the correlations in activity across each of these nodes at different times during the experiments (Figure 3.13a, see Methods). I then compared these matrices in larvae exposed to the f20 and f60 habituation paradigms to identify the network-level correlates of behavioural habituation. As expected, both paradigms produced high correlation values in response to the first loom, and the matrices for the two paradigms were highly similar. As habituation proceeded, network correlations remained somewhat higher in the f60 paradigm, reflecting differences in the behavioural responses during the f20 and f60 experiments (Figure 3.5c, d). By the 10th loom, most of these correlations had dropped dramatically for both paradigms, with high values mostly restricted to correlations between weakly habituating (red) nodes. The f20 paradigm shows a stronger recovery across the network in the 11th trial, reflecting the stronger behavioural recovery that takes place in this paradigm.

As an approach to judge both the rate at which these correlations were lost during the first block of stimuli and the degree to which they recovered in the 11th trial, I used a Pearson correlation to match the matrix of the 11th trial to the most closely related matrix from the first block of stimuli. The highest Pearson correlation coefficients were for the 4th trial for f20 and the 6th trial for f60, indicating both that the correlations are lost more rapidly in f20 (the paradigm in which habituation is strongest), and that the recovery is weaker in f60 (the paradigm that produces more indelible behavioural habituation). Notably, the patterns of correlations across the matrices during mid-habituation trials (4th for f20, Figure 3.13a, and 6th for f60, now shown) strikingly resemble those in the 11th trials, suggesting that the network is returning to a partially habituated state.

These results show that the loss of correlations across nodes in the network reflects behavioural outputs. To describe the networks containing these nodes, I represented them spatially and mapped the relative correlation strengths between nodes in the f20 and f60 paradigms (Figure 3.13b). Each edge (node-to-node relationship) in the graph is represented by its correlation value in the f20 paradigm minus its value in f60 paradigm. As expected, because the first trial is identical, both paradigms show robust correlations across numerous edges in the first trial, with most edges near a zero value and no net weighting of the graph toward positive or negative. By the 10th trial, the graph has lost most edges, and the remaining activity is biased toward stronger correlations in f60 (shown 70

in red), reflecting the slower habituation. The f20 paradigm shows stronger recovery, however, and this is captured in a shift toward positive values (blue) in the 11th trial.

I then quantified the participation of each node in the graph, where participation is defined as the proportion of a node’s highly correlated edges that are shared with nodes from a different functional cluster (as defined in Figure 3.6). Participation dropped over the course of 10 stimuli (Figure 3.13c), but this drop was slower in f60, suggesting that habituation is driven not only by a drop in correlation across nodes, but specifically by a loss of communication between different response types. This is reinforced by the higher participation in the 11th trial of the f20 paradigm, where strong behavioural recovery is echoed by a recovery in participation. Raster plots of participation by each node across the first 11 trials (Figure 3.13d) show this trend, further suggesting that it is weakly habituating (red nodes) that maintain much of their participation as habituation proceeds, and that recovery is accompanied by a resumption of participation by various strongly (green) and moderately (blue) habituating nodes.

To address which brain regions are involved in this process, I mapped the correlation strengths of edges between nodes across five regions containing a majority of the nodes (the pallium, thalamus, tectum, tegmentum, and hindbrain, Figure 3.13e). The values for each edge, represented by a dot, show the correlation in the 10th trial minus the correlation in the 11th trial, thus giving negative values to edges that became stronger during recovery. Violin plots show the total distributions of edges between different functional clusters. The results confirm that certain types of edges, especially those between two weakly habituating (red) nodes, play a relatively small role in recovery, owing to their strong unhabituated responses in the 10th trial. Other types of edges, especially those including a strongly and moderately habituating nodes, tend to have highly negative values, indicating that they contribute to the part of the network that is lost during habituation and regained during recovery.

Collectively, these results converge to produce a model of the brain-wide network that produces visual escape and the mechanisms by which these responses are suppressed during learning. The initial process of habituation appears to rest on the loss of correlation (and presumed communication) among neurons of different functional clusters. This is manifested as a dramatic drop in correlation values for edges between different clusters (Figure 3.13a), the restriction of the active network principally to edges between nodes of the same type (especially weakly habituating nodes, Figure 3.13b, e), and the loss of participation during the course of habituation (Figure 3.13c). The striking similarity between mid-habituation matrices and those of partially recovered networks (Figure 3.13a) indicates that the matrix changes that underlie habituation are likely the same as those that are reversed during recovery. This suggests that the onset of habituation works through the same circuit- 71

level changes that recovery does, and that there are not separable network-level mechanisms for the acquisition and retention of behavioural habituation.

Figure 3.13. The visual loom network, and the changes that occur during habituation. a. Correlation matrices for activity across 99 nodes representing ROIs across the whole brain. The functional clusters to which each node belongs are indicated on the axes, using the colour code from Figure 3.6. Darker blue shades represent stronger positive correlations for any given pairing, and red indicates negative correlations (see colour scale, a). b. A graphic representation of correlations across the 99 nodes, whose functional clusters are indicated by their colours and anatomical positions represented spatially. The colours and width of the lines indicate the relative correlation across the f20 and f60 experiments (f20 correlation minus f60 correlation), where red indicates stronger correlations in f60 and blue indicates stronger correlations in f20 (see colour scale). Only edges with correlations above 0.75 in either the f20 or the f60 matrices are displayed. c. A heat map of the participation for each of the 99 nodes during the 1st, 2nd, 3rd, 10th, and 11th loom stimuli of the f20 and f60 experiments. d. Raster plots showing the participation of each node across the first 11 stimuli for f20 and f60, and the relative participation (f20 value minus f60 value) where blue indicates stronger f20 participation and red indicates stronger f60 participation. The functional clusters for each node are indicated, using the colour code from Figure 3.6. e. Changes in correlation strength for all inter- node edges from the 10th to the 11th looms of f20, indicating the impacts of the recovery from habituation. Values shown are calculated for each edge as its correlation in the 10th loom minus its value in the 11th loom, with more negative values indicating edges that showed more pronounced recovery between the 10th and 11th looms (top). The functional clusters for each edge’s two nodes are colour-coded and the brain regions that the edges span are indicated on the left. Violin plots (bottom) show the cumulative distributions of edges connecting different types of functional clusters, as indicated on the left.

3.7 Discussion

The behavioural results show that zebrafish larvae habituate to loom presentations and that this habituation is tuned by the ISI and speed as predicted by the characteristics of habituation (Thompson and Spencer, 1966; Dragoi, 2002; Rankin et al., 2009; Thompson, 2014)). These results corroborate what has been found in the literature in terms of the relevance of the intensity and the stimulus interval to modulate the rate of habituation and the strength of the recovery (Russell, 1967; Askew, 1970; Rooney and Laming, 1986; Laming and Mckinney, 1990; Rankin and Broster, 1992; Post and von der Emde, 1999; Gray, 2005; Giles and Rankin, 2009; Fischer et al., 2011). In the experiments

presented in this chapter, the intensity (in the form of speed) is an important factor at modulating habituation to the looms, as it is suggested by the results and the statistical analysis. Faster looms take longer to habituate than slower looms. Also, and similarly to what has been previously found, a shorter ISI presentation had a stronger recovery than a long ISI. However, a limitation in this study was the fact that the sound stimulus did not seem to produce dishabituation on the recovery of the response at the 3rd block. A plausible explanation is that the stimulus was not strong or salient enough to produce a clear dishabituation effect. For most of the fish, the sound did not produce an escape response. Similar difficulties have been reported in other larval zebrafish habituation studies, where a dark flash did not manage to produce dishabituation to acoustic startles (Wolman et al., 2011). The reason suggested was that the flash did not elicit an escape response, meaning that it was not a salient enough stimulus to induce dishabituation. Another limitation of my study is that agarose embedded zebrafish larvae did not execute similar numbers of escape responses as free-swimming larvae. This problem seems to be a common phenomenon of similar setups (Lovett-Barron et al, 2017 figure S1H). Recent findings suggest that in similar preparations, the lack of visual feedback to the attempted movements could decrease the likelihood of the larvae trying to move (Mu et al., 2019). Nevertheless, the tail movements of head embedded fish could be used to localise the neurons that were active during motor responses. These results are similar to the movements produced by an aversive stimulus (Haesemeyer et al., 2018). Similarly, as in the behavioural experiments, our sound stimulus did not produce a clear dishabituation effect in the calcium imaging experiments.

A very interesting result is that, in the calcium imaging experiments, all groups of responding profiles already presented substantial habituation at the second presentation (independent of the stimulus interval), which shows that habituation can happen very quickly. Although single-trial habituation has been recently reported for tactile stimuli in zebrafish larvae (Roberts et al., 2019), this result contrasts with the more gradual habituation of the behavioural experiments, particularly for the fast loom presentations. Fast habituation of neuronal responses in the optic tectum to visual stimuli, including to a single trial, have been reported in goldfish before (Obenar, 1976). These visual habituating neurons seem to be sensitive to changes in the stimulus features like the position in the visual field, which suggest that presenting our stimulus from the same origin could facilitate faster habituation. Such results have been previously observed in other animals (Horn and Rowell, 1968; Gray, 2005; Raderschall et al., 2011). Therefore, the difference of habituation in the fixed head larvae and the free-swimming fish could be explained by different visual field stimulation. This would be due to the fact that in the latter setup, the larvae could change their body position. Therefore, visual stimulation would likely come from different parts of the visual field. Also, I considered that during

the behavioural experiments the loom stimuli are more salient as it is the only sensory stimulation to the fish in the behavioural rigs. However, during a calcium imaging experiment the set up produces some unavoidable stimuli (e.g. noises from the microscope, camera and computer, laser light, etc…). These other stimuli probably reduce the salience of the loom even when it is detected, which could facilitate its faster habituation.

3.7.1 Populations subtypes of visual habituating neurons

A key finding in this study is that visual habituation is happening at different rates across different neuronal populations. Previous studies suggested that neurons from the tectum could habituate at different rates in fish (Obenar, 1976; Guthrie and Banks, 1978; Kawasaki and Aoki, 1983), and across brain regions (Laming, 1980). These results and the ones presented in this study indicate that the nervous system has subpopulations of habituating neurons. In a recent study, a behavioural correlation of this phenomenon was found as different behavioural outputs of visual habituation were shown to be happening in parallel (Randlett et al., 2019). Although wide variability of responses can be found across the brain, the approach I used allowed me to group thousands of neurons into a few clusters that shared key characteristics. These functional subtypes can serve as a general model of variable and parallel habituation to the same stimulus train. It is probable that the differences in habituating profiles we are seeing are the combination of multiple factors. These responses might be showing different neuronal types that tend to respond in different ways. However, we probably also see some variation in the responses due to individual fish responses (some fish could habituate faster than others). Finally, changes in the calcium activity of these neurons could also be a result of the modulation by the different stimulus strains. This is maybe why we see more distinctive subtypes in the f60 dataset than in the s20 or s60 datasets (Figure 3.12). This could also explain the presence of weak habituating neurons in the cerebellum of fish with 60ISI and not in the fish with 20ISI. Nonetheless, besides these different factors playing a role, the categorisation of habituating neurons presented here provides a first and very valuable framework to study whole-brain habituation.

The loom responding neurons were found in the same areas as previous experiments (Dunn et al., 2016a; Chen et al., 2018b; Heap et al., 2018b; Henriques et al., 2019). The group of fast habituating neurons are very widespread across the brain. The analysis found three subtypes that I decided to merge due to their similarity. However, the fast habituating subtype 1 and 3 seem to present some subtle differences. These subtypes were a common result from the k-means clustering across datasets and were present in most fish. These facts suggest that there are different firing patterns presented

from different neurons. Some differences in distribution further suggest that they could belong to neuronal subtypes. It seems that more rostral areas (particularly pallium and thalamus) are more likely to contain the fast habituating subtype 1. On the other hand, the majority of broad fast habituating neurons seem to be located in the caudal part of the tectal PVN, in a similar location as some of the GABAergic neurons described previously (see (Gabriel et al., 2012) Fig 7B). Moreover, although is the group less clearly defined, most of the strongly habituating responses in the hindbrain belong to the 2nd subtype. Finally, a recent study also described different profiles of loom responding neurons, some more sustained than the others (Henriques et al., 2019). However, as they didn’t analyse their responses to multiple presentations, it is impossible to do an accurate comparison with the results I found.

It is also interesting how the thalamus is mostly populated by fast habituating neurons, especially considering that it plays a key role in the processing of loom and dim information involved in escape responses. This previous study did not show habituation of the thalamic axon activity located in the tectum of multiple fish (see figure 2E of (Heap et al., 2018b). Therefore, it is possible that the diming information provided by the thalamus, and relevant for the adequate escape response, rises from the subpopulation of weak habituating neurons located in this region.

Furthermore, the fast habituating neurons can also be found in the tegmentum and the medial part of the hindbrain. The responses in both of these areas probably come from projections of the tectum described by Helmbrecht and colleagues (Helmbrecht et al., 2018). The neurons located in the medial hindbrain likely include premotor neurons that are in the pathway of the tectobulbar projections (Sato et al., 2007; Helmbrecht et al., 2018). Unfortunately, our registration didn’t allow us to resolve if they were part of the M-System. Interestingly, most of these ROIs are located in the side contralateral to the loom presentation, while an escape response away from the stimulus should be generated in the ipsilateral side (see the previous chapter). Furthermore, these strong responses are present despite the lack of escape response in some of the fish. This suggests that these ROIs are implicated in a modulation of the escape response, rather than producing it.

The responses in both hemispheres of the tegmentum likely involve the nucleus isthmi, which receives projections from the tectum (Helmbrecht et al., 2018). This structure has been related to responses to looming stimuli, to novel visual stimulations and to habituate very quickly to stimuli that originate from the same part of the visual field (Northmore and Gallagher, 2003; Gallagher and Northmore, 2006). It has also been suggested that a NI-tectal loop is involved in the avoidance of approaching objects (Northmore and Graham, 2005; Graham and Northmore, 2007). Very recent studies have described this nucleus responses to looming stimuli in larval zebrafish. These researches 77

found that the nucleus isthmi is involved in the discriminating the most salient loom, that its optogenetic stimulatin modulates visual responses in the tectum, and that its ablation alters contrast sensitivity to looms (Fernandes et al., 2019; Henriques et al., 2019). The interconnections between the nucleus isthmi and the optic tectum are highly conserved across vertebrates (Kunzle and Schnyder, 1984; Gruberg et al., 2006; Dudkin and Gruberg, 2009), and are considered to be involved in detecting salient stimuli and in maintaining visual attention (Basso and May, 2017). Due to all these facts, and as mentioned in Chapter 2, it would be very interesting to investigate the role of this nucleus in escape responses, and in its possible role in habituation to looming stimuli. It is possible that it participates in changing the saliency of a repeated stimulus and facilitates its habituation.

Finally, a very interesting group of strongly habituating neurons is located in the medial area of the Pallium, which is most likely the Dm. Previously, responses to approaching objects in this area were reported in goldfish (Laming, 1980). This is an exciting result as this area has been suggested to be the amygdala-like structure of teleosts (Wullimann and Mueller, 2004; Yamamoto et al., 2007; Mueller et al., 2011; Mueller and Neuhauss, 2012; von Trotha et al., 2014; Carr, 2015; Perathoner et al., 2016; Lal et al., 2018), which suggests that a 6dpf larvae could be processing the looming stimulus with fear-like responses. This is a relevant finding as the amygdala is involved in the processing of many important emotional responses, presents habituation, and disruptions of its normal activity is present in different pathologies (Plichta et al., 2014; Tam et al., 2017). It would be a very promising structure to study in a model animal that also allows you to use all the technological advantages as a zebrafish larva does.

The weak habituating responses in the tectum, pretectum, habenula and telencephalon suggest that elements of the visual pathway are fairly resistant to habituation. This pathway could involve relays in the tectum or other retinorecipient structures that would then project to higher structures (Lau et al., 2011). The latter idea is corroborated by the fact that RGC axons don’t present the habitation seen in tectal cells (Obenar, 1976), and could, therefore, produce constant stimulation in other retinorecipient areas. An example of one of these pathways seems to be the retinal-EmT-Hb pathway (Zhang et al., 2017), which could explain the weak habituating activity that we see in the habenula. It is worth mentioning that we also observe a higher population of visual responses in the left habenula, which was previously described as sensitive to luminance (Dreosti et al., 2014).

The connections to the weak-habituating neurons of the pallium could be relayed by visual projections from the migrated posterior tubercular area (M2-future preglomerular nucleus (PG)) or from the dorsal thalamus, which receive projections from the tectum (Northcutt, 2006). The role of the thalamus proper of fish as a relay station is still in debate (Mueller and Neuhauss, 2012), and most 78

studies suggest that the main relay structure seems to be the PG. Unfortunately, the M2’s location makes it difficult to image in our setup, and I was not able to confirm this pathway in this study. This structure, or the group of cells that will form it, are localised directly behind the eye, making it impossible to image in our experiments. Interestingly, recent research suggests that this structure is not yet fully developed at 6dpf and that its projections to the telencephalon are not yet present (Bloch, 2019). On the other hand, I could find multiple kinds of visual responses in the thalamus proper, and some of its subdivisions are thought to have connections to the telencephalon. Particularly, the central posterior region of the dorsal thalamus seems to project mostly to the Dm, but also to the Dl (Echteler and Saidel, 1981; Murakami et al., 1983; Folgueira et al., 2004; Yamamoto and Ito, 2005; Northcutt, 2006). Electrophysiological recordings in this area found that it responds to multiple sensory inputs, including visual stimuli (Kirsch et al., 2002). Further and specific research would be needed to tackle this question. How does the visual information reach the telencephalon in 6dpf? Do these circuits change when reaching adulthood? These studies would be very interesting to better understand the ascending visual pathway in zebrafish and other teleosts.

Whether the visual ascending pathway to the pallium is from the M2 or the thalamus, it is likely that the pallial weak habituating neurons are located in an area that will become the lateral division of the dorsal pallium (Dl). This area has been described as the main recipient of visual input in the telencephalon (Saidel et al., 2001; Demski, 2003; Northcutt, 2006; Yamamoto and Ito, 2008; Ito and Yamamoto, 2009; Demski, 2013). Furthermore, the fact that there is a population of fast habituating and weak habituating neurons in the telencephalon suggest that even at this early stage the brain has the wiring and the processing of visual information in higher structures. This is very interesting as in many species of fish, ablation of, or discrete lesions in the telencephalon of adults produces a decrease in habituation (Rooney and Laming, 1986; Laming, 1987; Rooney and Laming, 1988; Laming and Mckinney, 1990), indicating that it could modulate how visual responses are perceived. Top-down modulation in habituation has been suggested in other animals (Quick and Laming, 1990; Krasne and Teshiba, 1995). Therefore, the telencephalic role of sensory processing and habituation in zebrafish larvae, including looming stimulus, is an area to explore in the future.

Although small in number, the location of the inhibited neurons in the rostral thalamus across datasets is consistent. This area is close to where inhibited neurons to auditory and vestibular stimuli were also found (Favre-Bulle et al., 2018). This suggests that multisensory information is modulated in this part of the thalamus. Moreover, some of the inhibited neurons were also found in the tectum and tectal neuropil. A similar result was found in electrophysiological recordings of the goldfish tectum,

where neurons with large receptive fields were inhibited to light presentations and didn’t show habituation in their response (Obenar, 1976).

Although some variations were found, the different subtypes of habituating responses were distributed in a very similar way across datasets. Multiple structures displayed very similar recruitment of responding neurons, the tectum being the main structure in terms of loom responding neurons. This result was expected from the key role of the tectum holds in visual processing (Northmore, 2011). A closer look at the responses in this structure suggests that the stimulus features modulate the activity of the different neuronal subtypes (Figure 3.12). The variations in the habituating responses in the tectum, despite the much faster habituation, still show a similar influence of the ISI and the speed of the stimulus as in behavioural experiments. These effects are more clearly displayed on the moderately habituating neurons, which are mostly located in the tectum. All of this highly suggests that the tectum is a convergence point for visual information that results in the habituating response we see in behavioural experiments. However, functional experiments are needed to confirm this idea.

3.7.2 A brain-wide network model of visual habituation

Using graph theory, I was able to model the networks involved in the visual habituation. From an anatomical perspective, the core loom perception circuit can be inferred from the edges that remain active through habituation. These include edges within and among the tectum, thalamus, and pallium (Figure 3.13b). The absence of habituation in these edges suggests that they are involved in perceiving a looming stimulus and that they are upstream of the sensorimotor transformation that controls behavioural outputs. The regions most affected during habituation (especially the hindbrain, but also including a subset of ROIs in the pallium) are likely downstream of this transformation. This places the tectum at an intriguing pivot point in the overall network. It has a confirmed role as an important recipient of loom information (Temizer et al., 2015; Dunn et al., 2016a; Filosa et al., 2016; Yao et al., 2016; Heap et al., 2018b; Helmbrecht et al., 2018), and communicates in different ways with different brain regions. This includes non-habituating correlations with the pallium and likely outputs to the hindbrain that habituate strongly (Figure 3.5b). Combined with the high density of moderately habituating ROIs, whose activity most closely mirrors behavioural habituation (Figure 3.12c) in the tectum (Figure 3.6h), this raises the possibility that circuits within the tectum are responsible for the key changes in the sensorimotor transformation that produce habituation. This idea is reinforced by the drops in correlation between moderately habituating ROIs and weakly habituating ROIs (blue-

red edges) and between moderately habituating and strongly habituating ROIs (blue-green edges) during habituation. This provides a mechanism by which moderately habituating neurons in the tectum could uncouple the core visual circuit of weakly habituating (red) neurons from downstream circuits as habituation proceeds. These uncoupled circuits, principally comprising strongly habituating ROIs (green), show interesting diversity, reflective of distinct impacts that novelty and saliency play in different brain regions. The hindbrain’s strongly habituating ROIs are more likely to correlate with the animal’s actual escape responses (Figure 3.10), as are those in other motor- associated regions including the cerebellum, pretectum, thalamus, and tegmentum. This suggests an interaction with these regions’ premotor and motor circuits (Sato et al., 2007; Helmbrecht et al., 2018) and an acute role in escape. Other strongly habituating ROIs that uncouple from the tectum occupy the pallium, including the Dm, a fear processing area (Mueller et al., 2011; Lal et al., 2018), and these are less likely to correlate to behaviour on a trial-by-trial basis (Figure 3.10), suggesting reduced higher-order representations of threat during habituation that are independent of trial-by-trial escape. The overall interpretation is that habituation involves the uncoupling of various downstream elements from visual perception circuitry, and implicates the tectum as the likely switch for this sensorimotor transformation.

It is worth to mention that the clustering methods and selection of ROIs belonging to different habituation subtypes would influence the the main decorrelation dynamics among nodes. This is due to the fact that the analysis evaluates the correlation between nodes that don’t habituate, nodes that habituate gradually and nodes that habituate quickly. In this sense, finding a decrease in edges as habituation occurs is to be expected. However, the observations using graph theory and the approaches used in this work allow for a deeper analysis of this phenomenon. The graph analysis allowed to quantify how the decorrelation during habituation and recovery involves connectivity changes across functional clusters (Figure 3.13c). If the decrease of edges due to decorrelation would have been only between nodes of the same cluster, we would see a gradual decrease in connectivity but the participation index would remain stable. However, the results suggest that that the loss of edges during habituation affects connection across functional clusters. It is important to note also how these participation changes differ between nodes of the same functional type (Figure 3.13d), which is related to a second argument that supports the meaningfulness of this approach. This is that the division of the clusters in nodes across the different anatomical regions allowed this study to take into consideration the variance across the different brain regions and even with in brain regions. Only by doing this, a comparison of how the connectivity recovered at different rates across the brain was possible (Figure 3.13e).

An alternative approach, tested when performing the analysis of the data, is to populate the nodes with neighboring ROIs before they are filtered by the linear regression to the stimulus train. This approach has the advantage that all the nodes would have the same ROIs sample from all the fish. However, the average time series of the categorized nodes would be then “diluted” with nonresponsive or noisy ROIs that would have been discarded otherwise. This extra noise could affect our ability to detect differences across groups or between genotypes (for chapter four). Therefore, this approach was discarded. Finally, the value and strength of the graph analysis can be exploited by combining the temporal dimension in the network to my graph theory approach. For example, further analysis could be performed using a dynamic community detection analysis (Mucha et al., 2010; Garcia et al., 2018; Vaiana and Muldoon, 2018), which is likely to produce interesting findings.

3.7.3 Future directions

This study is the first to describe whole-brain habituation dynamics at cellular resolution, and its results are very relevant for a better understanding of habituation and visual processing in the zebrafish larvae’s brain. However, these advances are opening interesting questions to be addressed in the future. I will briefly outline some of them.

An interesting but intriguing result in this study is the fast habituating response in the calcium imaging experiments, where we see a remarkable decrease in response already at the second loom presentation (Figure 3.6). As mentioned above, this could be due to the fact that the looms were presented every time in the same visual field, which seems to produce a fast decrease of responses in tectal and nucleus isthmi neurons (Obenar, 1976; Northmore and Gallagher, 2003; Gallagher and Northmore, 2006). Therefore, whole-brain imaging experiments with loom presentations from different locations would be very valuable to see how this new stimulus train habituates. This would be interesting in multiple ways; for example, it could help understand if local circuits in the tectum, sensitive to specific parts of the visual field, are key to the habituation. On the other hand, it would also be interesting to see how upstream structures like the thalamus or telencephalon respond and habituate by integrating the information from different eyes. Currently, I am supervising an undergraduate student that is starting to explore some of these questions.

Similarly, it is possible that the fish in the calcium imaging preparation habituates so quickly and fail to present behaviours because the lack of a sensory feedback to its motor movements (Mu et al., 2019). Previous studies have used closed loop system to study visuomotor behaviours (Portugues and Engert, 2011; Ahrens et al., 2013b; Trivedi and Bollmann, 2013; Jouary et al., 2016; Kawashima et

al., 2016; Henriques et al., 2019). It would be very interesting to see if a close loop system, where tail movements from escape responses could be tracked and interrupt the looming stimulus, would change the habituation rates.

This comprehensive analysis of neuronal responses to the looms also adds to the knowledge of how the visually evoked escape responses are processed and which parts of the brain are implicated (Temizer et al., 2015; Dunn et al., 2016a; Yao et al., 2016; Chen et al., 2018b; Heap et al., 2018b). However, this analysis does not differentiate between neurons that are processing different visual features of the loom (i.e. diming and moving edges). It would be of great interest to further elucidate what is the role of these different visual components, and the circuits that process them. A previous study from our group demonstrated that luminance sensitive neurons of the thalamus are crucial for processing the directionality of the threat (Heap et al., 2018b). On the other hand, previous studies have shown that while a checkerboard loom (moving edges without a drop in luminance) is still able to produce escape responses, a dim is not sufficient (Temizer et al., 2015; Dunn et al., 2016a; Heap et al., 2018b). Furthermore, how these features could modulate habituation is also unknown. At the end of my PhD, I helped with the supervision of the project of Marielle Piber, which consisted in presenting isoluminant looms (checkerboards), dims and looms in habituation paradigms and whole- brain imaging. Her results are very promising, and after further analysis of the data, they will likely become an interesting publication in the future.

The advantages of using zebrafish as a model to study the circuits involved in habituation demonstrated to be useful to investigate how this could be happening in different brain areas, but could also help to find out which type of neurons are involved in these circuits. This is particularly relevant to test the negative-image model or inhibitory potentiation theory proposed by Dr Ramaswami (Ramaswami, 2014). In this interesting theory, an inhibitory neuron that targets a sensory pathway enhances its synaptic activity to a particular stimulus. This creates a “negative-image” of the stimulus which competes with the excitatory activity, modulating the response of the excitatory neurons involved in the processing of that stimulus. The increment on inhibitory activity in such circuit is Dr Ramaswami’s proposed mechanism of habituation. Whole-brain habituation analysis, with the support of specific lines or methods to detect neuronal subpopulations, could locate such inhibitory circuitries. In our analysis, using nuclear targeted Gcamp6s, we did not find populations of neurons that incremented their activity (plausible GABAergic neurons). However, this phenomenon could maybe be detected at a synaptic level. For instance, experiments using a zebrafish line with a calcium indicator in the synaptic activity of tectal GABAergic neurons could test if such circuits are involved in loom habituation. This method could also be used to test if higher order structures could

also be generating such inhibitory potentiation in the tectum or other brain structures. Future work in this topic is likely to find interesting results.

3.7.4 Conclusions

Overall, I have shown for the first time that neurons with distinct response profiles to repetitive visual stimuli are simultaneously present throughout the brain and that these responses can be modulated by the saliency and temporal details of the stimuli. These activity traces reflect the rates of behavioural habituation to repeated looms, providing a framework for understanding the brain-wide network changes that mediate habituation. Using graph theory, I have shown that behavioural habituation tracks with a functional disconnection of a principally visual circuit in the fore- and midbrain and a response circuit that includes known premotor regions located in the hindbrain and higher-order representations of threats in the forebrain. The central location of the tectum (homologous to the mammalian superior colliculus) in this functional network, and the prominence of moderately habituating tectal neurons, whose activity reflects behavioural habituation rates, suggests that this region is involved in visual learning. Given these properties, it could serve as a pivot point for the sensorimotor transformation, a role that may be conserved in birds and primates (Dutta and Gutfreund, 2014). Furthermore, as habituation is affected in multiple mental diseases (McDiarmid et al., 2017), the study of psychiatric disorders models could also benefit from visual habituation protocols both behaviourally and with whole-brain imaging to try to find the specific population of neurons that are affected during this behaviour.

4 Visual habituation in fmr1-mutant zebrafish larvae

4.1 Contributions

M. Piber, G. Vanwalleghem, and L. Constantin all contributed to the work in this chapter. M. Piber performed the behavioural experiments and I assisted her in analysing the data. G. Vanwalleghem assisted with the data analysis of the calcium imaging. L. Constantin provided the zebrafish larvae and performed the preparation of the tissue for genotyping. E. K. Scott assisted in the conceptualisation of the experiments.

4.2 Summary

To test the validity and explore the utility of the network proposed in the previous chapter, we next looked for phenotypes across the habituation network in a zebrafish (Danio rerio) model of Fragile X syndrome (FXS), an inherited disorder characterised by intellectual disability, social deficits, and sensory phenotypes. We used a nonsense mutation in the highly conserved fmr1 gene (den Broeder et al., 2009), the perturbation of which causes FXS in humans. Given the learning deficits in humans with FXS, including slow habituation (Van der Molen et al., 2012; Schneider et al., 2013; Ethridge et al., 2016), we explored whether and how behaviour and brain-wide habituation networks are altered in fmr1-mutant zebrafish. Using the s20 habituation paradigm in our free-swimming preparation, we found that in fmr1-/- mutants, habituation is slower and recovery after a break is more dramatic. I then analysed the brain-wide loom habituation responses and once more found the main subtypes of habituating responses described in Chapter 3. Finally, I used graph theory to explore how the network of fmr1-/- fish could differ from wild types (WTs). While fmr1-/- animals undergo habituation through similar overall mechanisms to those seen in WT animals, their network loses these correlations more slowly and recovers them more dramatically after a rest, consistent with these animals’ behavioural phenotype.

4.3 Introduction

The previous chapter shows that zebrafish larvae habituate to looms by reducing the probability of escape from a looming stimulus after multiple presentations. Moreover, whole-brain calcium imaging results of multiple loom presentations uncovered that different populations of neurons habituate at different rates and are localised in particular brain regions. Furthermore, using graph theory, I was able to find that the connections between these different functional types of neurons reflect 85

habituation dynamics. These results suggest a disengagement of the circuit’s strongly habituating neurons as habituation occurs, probably mediated by moderately habituating neurons, while visual- specific neurons remain active and only habituate weakly. The previous findings are relevant for a better understanding of habituation as a phenomenon that involves multiple brain structures and different functional networks. However, these findings are also relevant because they show that habituation dynamics can be studied at a population level in larval zebrafish, and this could be useful for biomedical research. It is now possible to use this paradigm to study how habituation could be affected in animal models of psychiatric diseases. In particular, whole-brain visual habituation could be explored in animals modelling diseases where habituation is affected, such as schizophrenia, attention -deficit/hyperactivity disorder (ADHD), Fragile X syndrome, and autism spectrum disorder (ASD), among others (McDiarmid et al., 2017).

Research in ASD and related pathologies has mostly considered social and cognitive impairments. However, multiple studies suggest that the sensory system of autistic subjects is altered and that this could affect cognitive and affective processes (Rogers and Ozonoff, 2005; Robertson and Baron- Cohen, 2017; South and Rodgers, 2017). Among those altered sensory symptoms, a deficit in habituation among ASD patients to tactile, visual, and auditory stimuli has been reported (Webb et al., 2010; Green et al., 2015; Hudac et al., 2018; Vivanti et al., 2018). In particular, various imaging studies that tested habituation to emotional faces in ASD and healthy subjects suggest that the amygdala, a structure involved in emotional reactions, presents slower habituation in subjects with autism (Wiggins et al., 2014; Kleinhans et al., 2016; Tam et al., 2017). Furthermore, this slower habituation could involve a modulating circuit from the prefrontal cortex (Swartz et al., 2013; Green et al., 2015). Slower habituation observed in these studies seems to correlate with social impairments; however, the mechanisms underlying this habituation impairment are not yet understood.

Closely related to autism, the fragile X syndrome (FXS), caused by the mutation of the fragile X mental retardation 1 gene (FMR1), also presents habituation deficits to multiple sensory modalities (Miller et al., 1999). This mutation silences the production of the fragile X mental retardation protein (FMRP), involved in synaptic plasticity, and affects normal brain development that, among other symptoms, can cause features of autism (Saldarriaga et al., 2014; Hagerman et al., 2017; Mila et al., 2018). The habituation impairments in this disorder include affected short and long-term cortical habituation to auditory stimuli (Castren et al., 2003; Van der Molen et al., 2012; Ethridge et al., 2016), which can be recovered by minocycline treatment (Schneider et al., 2013). Furthermore, as in the ASD studies mentioned above, FXS also shows impaired habituation to human faces (Bruno et al.,

2014; Rigoulot et al., 2017), which could be due to altered GABAergic signalling (Olmos-Serrano et al., 2010).

The sensory alterations of FXS are easier to observe compared to more complex behaviours. Easier to track, these alterations are ideal to study in animal models, which enable experimental approaches. This allows better control of the variables and the possibility of establishing causal relationships of the involved elements, which is fundamental for understanding how sensory processing and habituation are affected (Sinclair et al., 2017; Rais et al., 2018). The FMRP is highly conserved across taxa (van 't Padje et al., 2005), which has allowed for the generation of multiple animal models. Previous studies using rodent and fly models with mutations in the Fmr1 gene have proven to be useful to investigate how the silencing of this gene affects the cellular and molecular pathways that are involved in FXS (Santos et al., 2014). Among these studies, habituation deficits were found in rodent Fmr1 mutants (Lovelace et al., 2016).

Zebrafish have become an important animal model for biomedical research, and in recent years, multiple groups have generated zebrafish models of psychiatric diseases. This is because they allow for the use of powerful genetic tools and present optical advantages and high-throughput benefits that can be exploited to understand the mechanisms underlying these conditions (Stewart et al., 2014a) (Stewart et al., 2015; Vaz et al., 2019). Among them, multiple zebrafish models of ASD have been developed (Stewart et al., 2014b; Lee et al., 2018; Meshalkina et al., 2018), including a zebrafish with mutations in the highly conserved fmr1 gene that are meant to model FXS (den Broeder et al., 2009).

Research in zebrafish with a nonsense mutation in the fmr1 gene has found behavioural phenotypes like an increased activity at different life stages in open field and novel environments (Ng et al., 2013; Kim et al., 2014; Shamay-Ramot et al., 2015; Wu et al., 2017). Avoidance learning impairment in adults, as well as reduced long-term potentiation (LTP) and increased long-term depression (LTD) in the telencephalon, have also been found (Ng et al., 2013). Additionally, fmr1 mutants present earlier shoaling behaviour, likely promoted by an increase in anxiety (Wu et al., 2017). These behaviours may result from structural changes in the axons and synapses of this animal model. Mutant larvae present increased axonal branching and synaptic density of at least glutamatergic and cholinergic neurons. These morphological effects could be caused by abnormal activity of Adar enzymes, which are normally regulated by FMRP and are involved in RNA editing (Shamay-Ramot et al., 2015). Despite previous efforts towards the characterisation of the fmr1 mutant zebrafish, it is still unclear if they also present altered habituation or visual processing defects.

A characterisation of visual habituation in fmr1 mutant zebrafish larvae, using the methods used in the previous chapter, would also allow for the investigation of how specific subpopulations of neurons across the brain and their network dynamics could be altered during habituation learning in this animal model. This would be relevant to understanding how this mutation alters brain-wide dynamics and which networks could be more affected, revealing potential target areas for therapy.

4.4 Methods

Animals. All zebrafish (Danio rerio) work was performed in accordance with The University of Queensland Animal Welfare Unit (approval SBMS/378/16). Adults were reared and maintained in a Tecniplast zebrafish housing system under standard conditions using the rotifer polyculture method for early feeding 5 to 9 days post-fertilisation. fmr1 mutant zebrafish embryos were bred by incrossing the fourth generation of zebrafish heterozygous for fmr1hu2787 (den Broeder et al., 2009) and elavl3:H2B‐ GCaMP6s to produce clutches with a 1:2:1 Mendelian ratio (wild type: heterozygous: homozygous) for fmr1hu2787. The fmr1hu2787 mutants have a change (C to T) in the fmr1 coding region, leading to a nonsense-mediated-decay and the loss of the protein (den Broeder et al., 2009). All fish were produced by natural spawning and reared in Petri dishes with embryo medium (1.37 mM NaCl, 53.65

µM KCl, 2.54 µM Na2HPO4, 4.41 µM KH2PO4, 0.13 mM CaCl2, 0.16 mM MgSO4, and 0.43 mM

NaHCO3 at pH 7.2) at 28.5 °C in a 14-hour light: 10-hour dark cycle. After the fmr1 experiments, larvae were genotyped as previously described (Ng et al., 2013).

Stimulus train for behavioural experiments. The stimulus train was a shorter version of the s20 described in the previous chapter. This stimulus train elicits faster habituation, which could help to find differences among phenotypes that would be masked by the ceiling effect of slower habituating stimulus trains. It consisted of two blocks of 10 looms with 5 min of rest (with a white screen) between the blocks. The loom was initiated with a dot that started expanding after 1s. The minimum angle of the loom was ~11°, and the maximum angle of the loom was ~90°. The looms reached their maximum angle in 4s. This was followed by 2 seconds of black screen and a 9s slow fade back to white, designed to avoid any neural OFF responses. The screen remained white for a variable duration of 18, 20, or 22s until the next loom initiation. The video was displayed by a monitor (10.1 1366x768 Display IPS + Speakers - HDMI/VGA/NTSC/PAL, Little Bird, Australia).

Behavioural experiments. Individual 6dpf larvae were placed in each well of the 12-well arena (circular plugs of agar were removed to produce the wells). The wells were filled with embryo medium and were placed 1 cm above a screen inside a dark chamber. All larvae received the same stimulus train. Experiments were videorecorded, and the movements tracked in bins of 1s using Viewpoint zebrafish tracking software (ZebraLab, ViewPoint Life Sciences, France). The output of the tracking was then analysed using a Matlab script. As before, fast movements (>30mm/s) were considered escapes. Further statistical analysis and graphs were made in GraphPad Prism v7.04 and R 3.5.1 (R core team, 2018). When the experiment ended, larvae were processed for genotyping. The quantification of the data was performed blind to the genotype for the fmr1 experiments. The binomial test was performed one- sided with the escape responses of the fmr1 or Het larvae in each loom versus the probability of response of the WT for that same loom.

Sample preparation, stimulus train, and calcium imaging. Imaging was performed on 6dpf larvae as described before. Briefly, larvae were embedded in 2% low melting point agarose (Sigma, A9045), transferred to a 3D printed imaging chamber, and filled with embryo medium. At the end of the experiments, larvae were processed for genotyping. Looms were presented in the same manner and with the same screen described in the previous chapter. The minimum angle of the loom was ~10°, and the maximum angle the loom covered was ~82°. As for the behavioural experiment, the stimulus was a shorter version of the s20 stimulus train. Zebrafish larvae were imaged for elavl3:H2B‐GCaMP6s on the custom-built SPIM microscope. To avoid stimulating the eyes with the light sheet, the side laser path of the SPIM was blocked and the front SPIM plane was restricted to a space between the eyes using a vertical aperture. The images were captured as described in the previous chapter. At the end of each experiment, a single high-definition scan of non-binned images was recorded with 100ms exposure time and 2µm increments, to be used for the registration of the brain of each fish (see below).

Analysis of calcium imaging data. Calcium imaging data were processed as described in the previous chapter. Briefly, the two acquisitions were concatenated, separated into individual slices, motion-corrected (NoRMCorre) (Pnevmatikakis and Giovannucci, 2017), and fluorescence traces were extracted and demixed from the time series using the CaImAn package (v0.9; http://github.com/flatironinstitute/CaImAn) (Pnevmatikakis et al., 2016a; Giovannucci et al., 2019).

Analysis of whole-brain activity data. I performed a K-means cluster analysis with 50 components with the traces of all of the fish and manually selected 8 clusters based on their likely loom responses. As in the previous chapter, I performed a linear regression using the chosen clusters as regressors and selected the regions of interests (ROIs) with an r2 value above 0.3. To compare the responses of this experiment with the functional clusters previously found (Chapter 3), I classified the ROIs into the functional clusters from our original s20 dataset using a correlation. All data were quantified blind to genotype.

For the representation of the ROIs per genotype, I chose a random sample (n=11) of heterozygotes (Hets) to match WTs (n=10) and fmr1-/- (n=11). For this chapter, no fish was excluded based on the strength of the responses or distribution among clusters as these results could be due to an effect of the phenotype.

The analysis was done using Matlab® R2018b and GraphPad Prism v7.04.

Correlation matrices and graphs. For the graph theory analysis, this fmr1 dataset was treated similarly to the wild type study described above, using the spatial nodes from that previous dataset. ROIs were assigned to each of the wild type nodes based on the smallest Euclidian distance and functional cluster identity. By again discarding nodes with less than 3 larvae, I ended up with 90 nodes for this analysis. As before, I performed cross-correlations and generated individual fish matrices for each loom presentation. I then averaged the matrices of each dataset across larvae. The connectivity matrices were analysed as above using the Brain Connectivity Toolbox (Rubinov and Sporns, 2010). I first generated weighted connectivity matrices and filtered out edges with an absolute correlation value below 0.75. The graphs were created with the subtraction of the fmr1-/- mutants from the WTs. The width and colour of the edges are indicative of the subtraction weight. To confirm the effects in the networks observed across genotypes (fmr1 vs WTs), I cross-validated the dataset with a leave-one-out approach. This consists in systematically excluding one fish from the average matrix and then generating multiple matrices for each loom presentation for the WT and fmr1 datasets. This allows to see the weight of individual fish and if the results are constant. The participation coefficient was calculated between the four functional clusters identified previously (strongly habituating, moderately habituating, weakly habituating, and inhibited).

Registration to a reference brain. Advanced Normalization Tools (ANTs, https://github.com/ANTsX/ANTs) was used to register results on the H2B-RFP reference of Zbrain (Avants et al., 2008; Avants et al., 2011; Randlett et al., 2015), as described in Chapter 3.

Data visualisation. I used Unity to represent each ROI centroid as a sphere with the parameters described in Chapter 3. The colourmaps were generated using two Matlab functions: the cbrewer function (https://au.mathworks.com/matlabcentral/fileexchange/34087-cbrewer-colorbrewer-schemes-for- matlab; accessed in May 2019), which includes specifications and designs developed by Cynthia Brewer (http://colorbrewer.org/), and the MatPlotLib 2.0 default colourmaps ported to Matlab (https://au.mathworks.com/matlabcentral/fileexchange/62729-matplotlib-2-0-colormaps- perceptually-uniform-and-beautiful; accessed in May 2019).

The circular graphs (Figure 4.10 and Figure 4.11) were made with a modified version of the code from Matlab’s circularGraph toolbox (https://www.mathworks.com/matlabcentral/fileexchange/48576-circulargraph/; accessed in May 2019).

Figures were produced using Matlab R2018b and GraphPad Prism v7.04 and assembled in Adobe Illustrator CS6.

4.5 Results

4.5.1 fmr1 mutants present behavioural habituation differences.

To characterise the escape behaviour of fmr1 larval zebrafish exposed to looming stimuli, I assisted M. Piber in using the 12-well apparatus I designed for Chapter 3, in which each well contained a larva receiving its own loom stimulus from below (see Methods and Figure 3.5a in Chapter 3). We presented looms in blocks of 10, with five minutes between blocks. The results suggest that compared to wild type (WT) siblings, fmr1 larvae display a slower habituation rate from the 2nd to the 4th loom as well as a more pronounced recovery after the 5 min break, while Hets show an intermediate phenotype. Significant differences are found between fmr1 mutants and Hets when compared to WT at the 2nd loom (Figure 4.1).

Figure 4.1. fmr1 mutant zebrafish larvae present altered behaviour.

Probability of escape response after multiple stimulus presentations in free-swimming larvae. Over the course of two blocks of 10 stimuli, fmr1-/- larvae show slower habituation and stronger recovery than wild type (WT) siblings, and heterozygotes (Hets) show an intermediate phenotype. Binomial test: fmr1-/- versus WT: 2nd Loom (p=3.056e-5); 3rd Loom (p=0.034); 11th Loom (p=0.055), and 16th Loom (p=0.039). Hets versus WT: 2nd Loom (p=0.001) and 9th Loom (p=0.039). All other comparisons (p>0.1). Significance cut-off using a Bonferroni correction for multiple comparisons is p<0.00125 (0.05/40).

4.5.2 Neuronal loom responses across genotypes are similar.

To explore and characterise the loom-habituating responses and possible differences between genotypes, I used a similar approach (K-means clustering) as for the characterisation of habituating neurons in Chapter 3. This approach sorts response profiles into a selected number of clusters (k=50) based on their similarities, which would allow for new patterns of responses possibly dependent on genotype effects to be detected in an unsupervised manner. Among the 50 clusters generated, I selected 8 clusters with clear loom responses and manually categorised them based on their similarity to the clusters described in Chapter 3. I again found strongly habituating neurons in which the response was dramatically reduced already at the 2nd loom presentation and that rarely responded or 92

had very weak responses to other loom presentations (Figure 4.2, first 3 rows). As in Chapter 3, I also found subtypes of these strongly habituating neurons, in which response decay lasts longer but is not as strong (Figure 4.2, first and second row). On the other hand, the 3rd row shows a profile of strongly habituating neurons whose response decays faster but that reach higher calcium activity. I also found moderately habituating neurons, but in this case, two subtypes emerged (Figure 4.2, rows 4–5). Similarly, I again found weakly habituating neurons, also represented in 2 clusters (rows 6–7), and inhibited neurons (Figure 4.2, last row). Comparing the means of the responses of each cluster per genotype shows a great similarity overall (a closer view of the first 3 and the 11th loom are shown in Figure 4.3). Averaged calcium responses to looms present the same rate of habituation and identical dynamics (e.g. raising and decay). The response to the first loom seems to be equally strong across genotypes in all clusters, which suggests that the fmr1 mutants are neither hypo- nor hypersensitive to a looming stimulus. Further responses are strikingly similar, with occasional higher responses for fmr1 larvae in the first 3 clusters (see arrows in Figure 4.3). Comparing the average responses between genotypes with a Pearson correlation corroborates this observation as the average correlation for each cluster is very high (R>0.97, see Figure 4.3).

In order to determine whether the fmr1 mutants present differences in the number of ROIs responding to visual stimuli, I compared the number of ROIs that passed the selection threshold (r2 > 0.3 to any of the cluster regressors) across genotypes. fmr1 mutants and Hets present slightly fewer ROIs than WTs, but the differences are not significant (Figure 4.4a). This was not due to a difference in the total number of ROIs segmented per genotype, as there were no significant differences in the number of ROIs per brain (Figure 4.4b). It is possible that the selection methods used were applying a bias that buffered these differences, however, as I was selecting for ROIs whose responses were similar to the regressors used. Therefore, to further investigate the possibility of a decreased response in fmr1 larvae, I compared the number of ROIs presenting a change in fluorescence that confidently indicated activity (2 SD above the mean) regardless of the trace profile. Again, fmr1 mutants and Hets presented less responsive ROIs; however, the differences were not significant (Figure 4.4c). Finally, to elucidate if the different clusters were preferentially populated by a specific genotype, I compared their ROI distribution (see Figure 4.4d). In general, WTs presented more ROIs in each cluster, but these differences seemed more pronounced in the strongly habituating and the first moderately habituating cluster.

To investigate whether the responses to the looms varied among genotypes, I analysed the normalised distribution of the maximum responses to the first 3 loom presentations and at the 11th loom (Figure 4.5). The distribution curve of responses across genotypes for all clusters is very similar for the first

loom. Interestingly, for the second loom, there is a subtle tendency of the fmr1 distribution towards higher values for some of the clusters. Specifically, the 3 strongly habituating and the first moderately habituating clusters present this shift. Responses to the 3rd and 11th looms are more homogenous again, except that values are slightly for the 11th loom for both of the moderately habituating clusters. These differences are subtle, but they could indicate different habituating responses from a small group of neurons in the fmr1 mutants or that it is more likely that some of these animals are showing slower habituation.

Together, these findings suggest that although there are no pronounced differences in the strength and profile of the responses across genotypes, fmr1 mutants could be recruiting fewer visual responsive neurons, particularly of the strongly habituating subtype. Furthermore, fmr1 mutants could be more likely to present strongly habituating neurons that take longer to habituate and moderately habituating neurons that present a stronger recovery. These last results are especially interesting as they could be related to the behavioural phenotype previously observed (Figure 4.1Figure 4.1).

Figure 4.2. The subpopulations of loom-habituating neurons. a. Raster plot of the ROIs forming each of the subpopulation clusters. b. Mean of the calcium activity of the visually responsive neurons for fish exposed to 20 looms. The insets of the strongly habituating clusters represent the responses to the 1st and 2nd looms. The mark on the y-axis indicates 2 SD of the normalised calcium traces and that on the x-axis indicates 60s. The same scales apply for the vertical and horizontal scale bars at the bottom of the Inhibited cluster. This scale bar applies to all clusters. 95

Figure 4.3. Similar responses per genotype.

Zoomed-in view of each genotype’s average calcium activity for the first 3 and the 11th loom presentations of each cluster shown in Figure 4.2. The traces are very similar, as indicated by the correlation between the genotypes’ traces. The correlation index for each cluster is at the top corner of each plot. The dashed line indicates the cropping of the trace between the first 3 loom presentations and the 11th presentation. The y-axis indicates the Z score. The horizontal bar in the top left plot indicates the time scale. This scale applies to all plots. 96

Figure 4.4. The number of loom-responding ROIs across genotypes. a. Number of ROIs per genotype that passed the r2 > 0.3 threshold for any of the regressors. b. Number of ROIs in the whole brain detected with our segmentation method across the 3 genotypes. c. Number of ROIs that responded to the first loom with a fluorescence change greater than 2 SD. d. Number of ROIs per genotype for each of the clusters shown in Figure 4.2. None of the comparisons analysed were statistically significant. Error bars are the SEM. WT n=10; Hets n=20; fmr1-/- n=11.

Figure 4.5. Distribution of loom responses.

Distribution of the maximum responses of the ROIs to the 1st, 2nd, 3rd, and 11th looms. Each row displays the ROIs belonging to the clusters in the first column. The insets of the strongly habituating clusters represent the responses to the 1st and 2nd looms. The mark on the y-axis indicates 2 SD of the normalised calcium traces and that on the the x-axis indicates 60s. The second column shows the average ROIs of each cluster found in each genotype. The error bars represent the SEM. WT n=10; Hets n=20; fmr1-

/- n=11. From the 3rd to the 6th column, the normalised distribution of responses to the 1st, 2nd, 3rd, and

11th looms are shown. WT larvae are represented in blue, Hets in grey, and fmr1-/- mutants in red. The x-axis indicates the maximum value of the z-scored response. The y-axis indicates the normalised frequency. The asterisk indicates a small shift to stronger responses in the distribution of the fmr1-/- larvae compared to WTs in response to the 2nd loom of the first 4 clusters and the 3rd and 4th clusters for the 11th presentation.

4.5.3 Brain distribution of responding neurons

To further characterise the loom-responsive clusters and the possible differences across genotypes, I analysed their distribution across the brain. The distribution of the ROIs of each cluster was very similar to the distribution of the ROIs of the matching cluster in the S20 dataset (Figure 4.6). Due to the comparable distribution and response profiles, and to facilitate generalisations and comparisons, I classified the ROIs from the 8 new clusters into the 4 main categories used for the previous dataset (Chapter 3). To do so, I assigned their identity based on the highest correlation to the averaged S20 responses for the strongly habituating, moderately habituating, weakly habituating, and inhibited clusters.

A view of the distribution of the ROIs across the brain shows that while all genotypes had fundamentally similar distributions, there was a reduction in strongly habituating ROIs in the hindbrain (Figure 4.7) in fmr1-/- larvae. There were also fewer weakly habituating ROIs in the pallium and tectum of fmr1-/- larvae, but on the other hand, there was a trend toward more numerous weakly habituating ROIs in the cerebellum. Despite their magnitude, none of these differences were significant due to the individual variability within each genotype.

Figure 4.6. Similar distribution to S20 clusters. a. Top view of the ROI distribution in the brain for each of the clusters in Figure 4.2. The ROIs of each cluster are colour-coded with their putative functional identities based on their traces: green for strongly habituating ROIs, blue for moderately habituating ROIs, red for weakly habituating ROIs, and magenta for inhibited ROIs. The ROIs of all genotypes are plotted together. b. Top view of the ROI distribution in the brain for the clusters from the stimulus train (S20) in Chapter 3. Very similar patterns of distribution that match the functional cluster identity can be found in both datasets. 100

Figure 4.7. Brain distribution per genotype.

Top view of the brain-wide distributions of ROIs for the three genotypes, colour-coded for the functional clusters as in Figure 4.6. A noticeable decrease in ROIs can be seen between WTs and fmr1-/- mutants, especially in strongly habituating ROIs (green) in the hindbrain.

101

Figure 4.8. ROI distribution in relevant brain areas per genotype.

The number of functional cluster ROIs found in different brain regions. Fewer strongly habituating ROIs can be found in the hindbrains of fmr1-/- larvae compared to controls. In addition, fewer weakly habituating ROIs can be found in the pallium and tectum. None of the comparisons were significant. Error bars are SEM.

4.5.4 fmr1-/- mutant larvae show network-level habituation deficits

To investigate whether network activity is affected in fmr1 mutant larvae, I applied graph theory to these results, based on the analysis performed in Chapter 3. I analysed the cross-correlations among 90 nodes (having eliminated nine of the original 99 nodes with the requirement that all nodes need to be represented in at least three larvae), applied a 0.75 correlation threshold, and subtracted the connectivity matrix of the mutants from that of the WTs (Figure 4.9a). Generally, correlations across

102

the network were stronger in WTs than in fmr1-/- mutants in the first trial (resulting in positive values shown in blue). This is reversed in the 2nd and 3rd trials, where the WT network habituates more quickly, leaving negative (red) values that indicate persistent fmr1-/- network activity. This results hold even after cross-validation analysis (Figure 4.9b). Consistent with behavioural data, the overall correlations across the WT and fmr1-/- networks are equivalent by the 10th trial, but WT networks are stronger across the core perceptual pathway described above (tectum, thalamus, and pallium), while fmr1-/- correlations are stronger across edges that habituate quickly in WTs. Again echoing free- swimming behaviour, the fmr1-/- animals show a dramatically broader and stronger network correlation in the 11th trial, following a break in the stimulus. A closer look reveals that one of the main differences between genotype networks is that the hemispheres seem to be more connected in fmr1-/- larvae as habituation proceeds (note the number of red lines crossing the midline in Figure 4.9a). In particular, the strongly habituating nodes from the tegmentum maintain strong connectivity at the 2nd and 3rd loom presentations (Figure 4.9 and Figure 4.11). Another population that establishes multiple connections with the contralateral side is that represented by the weakly habituating nodes of the tectum and pretectum (Figure 4.9 and Figure 4.11), this is observable even at the 10th loom presentation. In contrast, the weakly habituating nodes from the habenula and pallium from that hemisphere remain more connected in WT larvae.

On the contralateral side of the stimulus presentation, clear differences can be found in the strongly habituating nodes of the hindbrain at the 2nd loom, but most particularly at the 11th loom. Stronger connections to the hindbrain could represent visually evoked premotor activity or its modulation (Sato et al., 2007; Helmbrecht et al., 2018). The tectum also presents higher connectivity of its green nodes at the 2nd and 11th loom presentations. This is a very interesting result as these nodes probably represent tectal neurons that are not directly exited by retinal projections (hence their strong habituation) but that could be relevant for the habituation process or that could also be projecting to other structures to produce, for instance, escape responses (Sato et al., 2007; Dunn et al., 2016a; Helmbrecht et al., 2018). As for the side ipsilateral to the stimulus, the contralateral green nodes of the tegmentum also show maintained connectivity and recovery. Among the structures of the tegmentum, it is likely that some of these nodes represent connections to the nucleus isthmi. This structure is reciprocally connected to the tectum, is sensitive to looms, and is related to the detection of relevant visual stimuli (Vanegas and Ito, 1983; Striedter and Northcutt, 1989; King and Schmidt, 1993; Northmore and Gallagher, 2003; Gallagher and Northmore, 2006; Fernandes et al., 2019; Henriques et al., 2019). The failure of its connections to decrease, as in the WTs, could be important to understand the behavioural effects seen in free-swimming larvae and the rest of the circuit. Higher

103

on the visual pathway, the thalamus also presents green nodes with stronger connections at the 2nd and 11th looms than in WTs. The thalamus has traditionally been seen as a relay structure to higher centres, although this role is not as clear in teleosts (Mueller, 2012). However, it has also been found to be involved in the visual processing that modulates escape responses evoked by looming stimuli (Heap et al., 2018b). Finally, as in the previously described structures, the pallium also presents green nodes with altered habituation, which prevails even after 10 looms. These nodes seem to be aggregated in the medial part of the pallium and likely represent neurons from the medial division of the dorsal pallium (Dm). All of these correlation-based observations carry through to participation across the networks (Figure 4.9c,d), including in the same nodes whose edges showed differences in correlation (Figure 4.9a).

By assessing correlation strengths across the network in a way that represents the functional and anatomical properties of nodes, I outlined the overall functional architecture of the habituating fmr1- /- brain versus that of WTs. First, I organised brain-wide node-to-node relationships by functional cluster (Figure 4.10), allowing the level of correlation within and across clusters to be assessed. This shows that by the 2nd stimulus, there are still strong functional connections among red–red edges and along blue–blue edges, and a few nodes from these groups establish connections with green nodes in WT animals. The connections are then largely restricted to red–red edges by the 3rd trial, with only a handful of red–blue connections remaining. By the 10th trial, strong correlations only exist in red–red edges (and a few to inhibited (purple) nodes). A subset of red–blue, blue–blue, and blue–green nodes reconnect in the 11th trial, reflecting recovery. In all regards, these effects resemble the habituating network dynamics shown for the f20 paradigm in Chapter 3, where behavioural habituation tracks with a loss of communication were seen between weakly habituating (red) nodes and strongly habituating (green) nodes, connected through moderately habituating (blue) nodes. By comparison, fmr1-/- animals show more strong correlations between nodes in the 2nd and 3rd trials, as well as following recovery in the 11th trial (Figure 4.10). Correlated edges are similar between the genotypes in the first and 10th trials (Figure 4.10), showing that the networks are closer to equivalent in the naïve state and following habituation. Consistent with the analysis of the behavioural experiments, this suggests that uncoupling across functional clusters occurs more slowly and recovers more completely in fmr1-/- animals, providing a mechanism by which the sensorimotor transformation is slanted towards downstream network activity and behavioural responsiveness in these animals.

To explore the spatial properties of this phenotype, I represented these data organised by brain region (Figure 4.11). In WT animals, this analysis shows an extensive correlation between nodes across all brain regions in the first trial that is progressively winnowed to the core perceptual circuit described 104

above (mainly connections among the tectum, thalamus, and pallium on the side contralateral to the stimulus) as habituation proceeds. In the 2nd, 3rd, and 11th trials (and to a lesser degree, the 10th trial), this network is more extensive in fmr1-/- animals, showing stronger functional relationships between the tectum and other regions, and has a greater number of highly correlated edges from the hindbrain to other regions. This, in turn, echoes observations from the previous chapter, which suggests that an uncoupling of spatially distinct perceptual and downstream networks drives habituation.

105

Figure 4.9. Network-wide changes in fmr1-/- larvae. a. Node-based graphs showing relative correlations (WT correlation minus fmr1-/- correlation), where blue indicates correlations that are stronger in WT animals and red indicates correlations that are

106

stronger in fmr1-/-animals. b. Results of cross-validation analysis after a leave-one-out approach. A comparison of the total number of edges >0.75 at different loom presentations corroborate the results presented in a. WT present more edges at the 1st loom and fmr1 fish at the following loom presentations, including the recovery loom (11th). (WT vs fmr1 >0.75 edges Predicted mean diff. and adjusted p values of Sidak’s multiple comparisons test: 1st Loom=546.9, p= <0.0001; 2nd Loom=- 192.3, p=<0.0001; 3rd Loom=-121.7, p=<0.0001; 4th Loom=-94.98, p=0.0001; 5th Loom=-165.6, p=<0.0001; 10th Loom=-81.07, p=0.0039; 11th Loom=-311.1, p=0.0002). c. Heat maps of participation for all nodes across habituation and recovery. d. A raster plot of relative participation (WT participation minus fmr1-/- participation) for each node through the first 11 trials.

107

Figure 4.10. Functionally sorted brain-wide networks for WT and fmr1-/- larvae.

Edges with correlations above 0.75 are shown between all combinations of nodes, and nodes are arranged by their functional clusters (colours of nodes). Networks are shown for trials 1, 2, 3, 10

(top), and 11 (bottom). WT networks are shown in blue, and fmr1-/- are shown in red. The brain region to which each node belongs is indicated. Pallium, Pal; subpallium, Sp; thalamus, Th; habenula, Hb; pretectum, Pt; tectum, Tec; tegmentum, Tg; cerebellum, Cb; and hindbrain, HB.

108

109

Figure 4.11. Spatially sorted brain-wide networks for WT and fmr1-/- larvae.

Edges with correlations above 0.75 are shown between all combinations of nodes for trials 1, 2, 3, 10, and 11, and nodes are arranged by brain region. Functional clusters of nodes are identified by colour. Empty nodes (black) are added to match the right side (ipsilateral to the visual stimulus) to the left side spatially, despite having fewer nodes. Correlations are enriched on the left side of each brain, consistent with stimulus presentation to the right eye. Abbreviations are the same as in Figure 4.10.

4.6 Discussion

The behavioural experiment in this study found a moderately altered habituation phenotype in fmr1 mutants (Figure 4.1). Mutants presented slower habituation to the first train and a stronger recovery at the 11th trial after the rest period. This phenotype is in line with the habituation deficits found in FXS patients (Miller et al., 1999; Castren et al., 2003; Van der Molen et al., 2012; Ethridge et al., 2016) and encourages further characterisation in future studies. For instance, a recent impressive study using C. elegans was able to characterize multiple phenotypes presented in animals with mutations of autism-associated genes (McDiarmid et al., 2020). Many of these genes affect probability of escape responses as habituation occurs. However, these genes often varied on how they affected other traits of habituation, as the duration of the response or distance travelled by the animal. Parallel networks in visual habituation had been already suggested in zebrafish (Randlett et al., 2019), so we attempted to do such analysis observing possible changes in other features like the total number of escape responses for a loom, distance travelled during escapes, or maximum velocity of escapes. However, our methods were not sensitive enough to detect significant difference in these traits. Further experiments with better time and spatial resolution and more sensitive software would allow a more detailed analysis of the changes in other responses and could show other features of the behaviour on which fmr1 mutant fish are affected. Finally, recent studies also shown that habituation at different ISIs is mediated by different molecular mechanisms (Ardiel et al., 2018; McDiarmid et al., 2019). Therefore, it would be interesting to test how fmr1 mutants’ habituation is affected at different ISI and see if the fmr1 gene could affect a specific pathway. If this could be related to other genes with similar phenotypes we would be able to learn more about the molecular mechanisms behind fmr1 mutation effects.

Interestingly, the behavioural deficits in habituation contrast with our findings of calcium activity. Overall, our results show that the averaged neuronal activity was very similar across the three

110

genotypes (Figure 4.3). We were not able to find clear differences in the strength of the responses to the loom presentations. However, subtle differences in the distribution of the responses to the 2nd and 11th stimulus (Figure 4.5) in strongly habituating and moderately habituating clusters could be indicative of effects of the fmr1 mutation. As suggested by the previous chapter, these subtypes could be responsible for the sensorimotor transformation. The fact that we find a shift towards higher responses in these clusters could be indicative of alterations in the circuits involved in visual habituation.

Another difference found between genotypes is the number of ROIs responding to the visual stimuli (Figure 4.4, 4.7 and 4.8). However, this comparison was not conclusive due to the variability of the sample. This is a result of being inclusive regarding the fish used in the study in order to avoid biasing our results and ensure all phenotypes were detected. Thus, the criteria to include a fish for further analysis was simply the detection of some responses in the tectum, indicating that the fish was not blind, and the fish being healthy at the end of the experiment (when it was freed from the agar for genotyping). These criteria allowed for a wider diversity of responses, including very weak ones, which limited the quantitative analysis. However, there was a trend of fewer responsive ROIs in the fmr1 mutants, especially of the strongly habituating subtype and predominantly in the hindbrain. There were also fewer weakly habituating ROIs in the pallium and the tectum of the fmr1 mutant fish compared to controls. Further experiments with either a larger sample or a more restrictive criteria for inclusion would be needed to clarify whether these trends are just a result of individual variability or if they could indicate a phenotype. This would be interesting as it seems that the total number of ROIs detected in the brain doesn’t differ between genotypes, meaning that fmr1 mutants could be recruiting fewer neurons when responding to visual stimuli.

Although the outcome was different, a study with an ASD mouse model also found fewer reactive neurons to looming stimuli and evidence of altered connections in the loom-evoked escape response pathway (Hu et al., 2017). In that study, valproic acid (VPA)-treated mice show decreased defensive responses to looming stimuli and less neuronal activity in the ascending visual pathway (superior colliculus, thalamus, and amygdala) measured with c-Fos localisation. Viral injections and retrograde labelling confirmed that this pathway was affected in VPA-treated mice. Finally, behavioural and diffusion tensor imaging (DTI) scans of children with ASD also show fewer responses and less activity in this subcortical pathway. It is possible that the sensory alterations presenting in ASD and FXS patients in the form of hypo- or hypersensitivity are due to circuit alterations (Port et al., 2014; Contractor et al., 2015; Subramanian et al., 2015; Zerbi et al., 2018). Zebrafish embryos treated with VPA have also previously been used as ASD animal models (Zimmermann et al., 2015; Chen et al.,

111

2018a; Lee et al., 2018; Dwivedi et al., 2019). The use of VPA-treated zebrafish in combination with the methods used in this chapter would be ideal for investigating these circuit alterations and further study the developmental and molecular mechanisms of these changes. It would be also relevant to study how the VPA effects on zebrafish brain circuits are different from the fmr1 mutation’s (Cowden et al., 2012; Shamay-Ramot et al., 2015; Baronio et al., 2018) and how both changes can still trigger autism-like behaviours (Hazlett et al., 2009; Hoeft et al., 2011; Hazlett et al., 2012; Subramanian et al., 2015; Zerbi et al., 2018).

A clear effect found in this study is the maintenance of strong connections in fmr1 mutants as habituation occurs and their enhanced reestablishment at the moment of recovery (Figure 4.9–11). These results suggest that the fmr1-/- larvae present habituation deficits not in terms of the strength of their responses but in terms of the connectivity of circuits. While fmr1-/- animals seem to undergo habituation through similar overall mechanisms to those seen in WT animals, their network loses these correlations more slowly and recovers them more dramatically after a rest, consistent with these animals’ behavioural phenotype (Figure 4.1). An interesting feature found among the differences between fmr1 mutants and controls is that connectivity between both hemispheres is more prevalent in fmr1 larvae (Figure 4.9 and Figure 4.11). This is very conspicuous for the strongly habituating nodes of the tegmentum and the weakly habituating nodes of the tectum and pretectum (Figure 4.9 and Figure 4.11). Interestingly, these connections remain until the 10th loom presentation, which is surprising because most of the retinal projections target the contralateral side of the stimulated retina. However, it is striking that the weakly habituating nodes from the habenula and pallium located on the ipsilateral side of the stimulus presentation remain more connected in WT larvae. This could indicate that they play a role in the modulation of the network that is lost in fmr1 mutants. Together, these results suggest that the core visual pathway remains strongly connected in a different way for WTs and fmr1 mutants, which could point towards a failure of the visual circuits to modulate the activity of parts of this pathway in the mutant larvae.

In the contralateral side of the stimulus presentation, where as expected we found most of the neuronal activity, the most conspicuous differences in connectivity seem to involve strongly habituating nodes from the hindbrain, tegmentum, tectum, thalamus, and telencephalon (Figure 4.9–11). The nodes from this last structure are likely located in the Dm, which is considered to be the homologue of the mammalian amygdala and could be involved in emotional responses including fear like responses (Wullimann and Mueller, 2004; Broglio et al., 2005; Mueller et al., 2011; von Trotha et al., 2014; Perathoner et al., 2016; Lal et al., 2018). Finding impaired habituation in the connectivity of this structure is particularly relevant as many of the human visual habituation studies of ASD and FXS

112

patients find altered habituation in this structure (Kleinhans et al., 2009; Wiggins et al., 2014; Kleinhans et al., 2016; Tam et al., 2017).

The experiments outlined in this chapter confirm that the main functional neuronal subtypes discovered in Chapter 3 are an accurate description of the subtypes of loom-habituating responses. Albeit with slight differences, I found the same main patterns of activity in this new dataset. Additionally, the methods used again identified subtypes of the strongly habituating cluster, and in particular a strong and sharp group and a less strong and wider response group, which suggest that these activity patterns could belong to different neuronal populations. However, in this analysis I also found moderate and weakly habituating clusters subgroups. Some of the differences found between the former and the present study could be due to differences in the variables and the methods used to localise the activity patterns. Although I used the same clustering methods and parameters (e.g. K- means, cityblock distance, 50 clusters), the present dataset was shorter (20 vs 30 looms), and there was no sound presentation. These features influence the number of variables taken into account to divide clusters in our method, leading to more variability in the former dataset. As the final number of clusters was the same (k=50), having fewer possible profiles could have influenced the analysis of the present dataset to split groups that would have been merged otherwise.

However, apart from the similarity with the averaged traces from the previous dataset, another confirmation comes from the very similar distribution of these clusters in the brain (Figure 4.6). The clear localisation in particular areas of the brain of some of these patterns of activity in both datasets is a strong confirmation that they correspond to specific neuronal populations. However, a more detailed analysis is needed, taking a careful and systematic approach based on their location in the brain, to draw accurate conclusions. Future studies could combine neuroanatomical techniques such as immunolocalisation (Lovett-Barron et al., 2017), or targeted photoactivatable GFP (dal Maschio et al., 2017) with functional activity to learn more about these neuronal subtypes. This would be of special interest in animal models of psychiatric diseases such as the fmr1-/- zebrafish used in this study as they could point to specific populations of neurons affected by these disorders. Such information would be of great relevance for further development of targeted drug therapies.

The analysis of the relevant networks was carried out using graph theory. This approach has been extensively used over the last 10 years for the study of human brain activity and structural correlation networks of healthy and psychiatric disease individuals (Sporns, 2018; Yamashita et al., 2018; Farahani et al., 2019; Hallquist and Hillary, 2019), including ASD and FXS patients (Saggar et al., 2015; McKinnon et al., 2019). Interestingly, these methods have described how the amygdala and putamen are less interactive in the network of FXS individuals and that they develop different 113

modularity than controls (Bruno et al., 2017). Very recently, these methods have also started to be used in calcium imaging datasets of zebrafish larvae models of psychiatric diseases (Light and Jontes, 2019; Liu and Baraban, 2019). These studies and the work presented here show the relevance of network analysis in the study of brain connectivity in humans and animal models of psychiatric diseases.

The study presented in this chapter used state of the art whole-brain calcium imaging technology and a combination of advanced analytical methods, including graph theory, to better understand how the fmr1 mutation could be altering the nervous system. Another piece of progress from this study is the incorporation of a whole-brain habituation paradigm at cellular resolution in the study of psychiatric diseases. Habituation paradigms have recently been used successfully in zebrafish studies trying to decode the genetic and molecular basis for habituation using high-throughput methods (Wolman et al., 2011; Wolman et al., 2014; Marsden and Granato, 2015; Wolman et al., 2015; Jain et al., 2018). The combination of the simplicity of the habituation paradigms and the advantages of zebrafish for studying the nervous system is likely to be very relevant in future studies investigating this learning process and how it is affected in psychiatric diseases.

In summary, I have shown that the loom habituation responses and overall network found in wild type animals in the previous chapter are present in fmr1-/- animals but that the dynamics are shifted towards higher network correlations, greater transmission through the tectum, and ultimately slower behavioural habituation. This reveals a brain-wide mechanism for slower sensorimotor learning that reflects previously observed behavioural phenomena in animal models and humans with FXS. Importantly, it provides a departure point for targeted explorations of the circuit-level causes of learning and sensorimotor deficits in FXS and other psychiatric disorders.

114

5 Visual discrimination learning in Ambon damselfish: Mapping the dorsal pallium, testing pallial learning and disrupting visual and learning abilities.

5.1 Contributions

D. Bender, A. N. Parker, C. Rosinski, H. Middleton and U. Siebeck contributed to the work of this chapter. H. Middleton collected and trained fish in the Heron Island field trip. She also sliced some of the brains for the neuroanatomical analysis. D. Bender designed and trained me on the use of the

CO2 setup, and she performed the chemical analysis of the water samples. D. Bender, A. N. Parker,

C. Rosinski and I build the CO2 setup for the visual disruption experiment on Lizard Island. A. N. Parker and U. Siebeck collected fish near Lizard Island and trained H. Middleton and me to catch fish. U. Siebeck and A. N. Parker trained C. Rosinski, H. Middleton and me to conduct behavioural experiments. A. N. Parker, C. Rosinski and I trained fish for the visual disruption experiment. U. Siebeck designed the visual discrimination task for visual disruption experiment. U. Siebeck and I designed the visual disruption experiment timeline.

5.2 Summary

In this chapter, I focus on Ambon damselfish (Pomacentrus amboinensis) as model to study more complex visual abilities in teleosts, i.e. visual discrimination learning. Based on histochemical staining, I show that the pallium organisation of Ambon damselfish is similar to that of other teleosts. Also, my results suggest a likely involvement of the pallium in learning visual discrimination tasks. Furthermore, I describe an experiment which was designed to disrupt visual discrimination and visual learning by decreasing the pH using increasing CO2 concentrations in the aquarium water. This method has been previously reported to produce disruptive effects across various sensory systems, including vision (Chung et al., 2014). However, my results show that the visual abilities of adult P. amboinensis required for discrimination and learning are not affected by this treatment. Finally, I discuss the implications of the results obtained with the neuroanatomical mapping, localisation of visual activity in the pallium and visual learning disruption experiments in Ambon damselfish.

115

5.3 Introduction

Many species of fish rely on vision as one of their main sources of information from the outside world (Kotrschal et al., 1998; Gerlai, 2014). These animals use this information and their experiences to adapt their behaviour depending on the circumstances to increase their chances of survival. An example of this is Neolamprologus pulcher, a territorial cichlid. This fish uses visual cues to learn to individually recognise its neighbours and detects when they change location, or a new neighbour is introduced displaying more aggressiveness to new fish to protect their territory (Saeki et al., 2018). Furthermore, recent results suggest that the triggerfish Rhinecanthus aculeatus could be using colour and pattern edge cues to learn and generalise which food is appropriate to eat (Newport et al., 2017; Green et al., 2018). Finally, visual information paired with social or olfactory cues can speed up learning to recognise predators (Csanyi and Doka, 1993; Kelley and Magurran, 2003). Dascyllus marginatus is a species of damselfish which, as adults, show a selective avoidance behaviour towards predators but not to non-predator fish. This selective response seems to be learned, as juveniles show a generalised response (Karplus et al., 2006). Interestingly, when paired with experienced adults, the juveniles adjust their behaviour, suggesting that they learn from seeing the responses of their conspecifics (Karplus et al., 2006). Another damselfish showed learning after a visual presentation of a possible predator. In this study, researchers shown that Spiny Chromis modulate their escape responses based on previous visual and olfactory exposure to a predator (Ramasamy et al., 2015). Olfactory pairing and social facilitation of visual learning experiences have also been tested in zebrafish with non-natural stimuli. After pairing an olfactory alarm cue with a red light, the single presentation of the light produced avoidance behaviour (Hall and Suboski, 1995). When these trained fish were placed together in a shoal with naïve fish, the latter would also display avoidance behaviours to the red light, even after being separated from the experienced fish (Hall and Suboski, 1995). Together, these studies show examples of how fish are able to process different visual features, like colour, size or shape, to distinguish and learn from these experiences and display appropriate responses in the future.

To investigate how fish process visual information, researchers have developed different creative ways of testing fish visual abilities (Douglas and Hawryshyn, 1990; Kelber et al., 2003; Schuster et al., 2011b). One of these methods for the study of fish vision, adapted from human psychophysics, is the alternative forced-choice paradigm (AFC) (Sutherland, 1961; Douglas and Hawryshyn, 1990). Using this paradigm, researchers have been able to investigate, among other topics, colour vision (Neumeyer, 1992; Siebeck et al., 2008; Pignatelli et al., 2010; Van-Eyk et al., 2011), visual acuity (Temple et al., 2013; Champ et al., 2014), size constancy (Douglas et al., 1988) and shape

116

discrimination (Siebeck et al., 2009). In this paradigm, fish are trained to select one specific stimulus in order to receive a food reward. Following training, different visual stimuli (objects or 2D images) are presented, and the fish has to identify the trained or positive stimulus. Only by approaching the positive stimulus (S+), the fish is rewarded. This operant conditioning technique allows training and testing fish to see if they can distinguish and discriminate among the different visual stimuli.

Using this technique, researchers have been testing the visual discrimination abilities of fish for more than half a century. For example, in some early studies, researchers tested goldfish visual discrimination of different shapes (Mackintosh and Sutherland, 1963; Sutherland, 1969). Since then multiple visual discrimination tasks have been used to investigate fish visual processing, but in the last decades more complex stimuli and tasks have been tested, not only to learn more about fish visual capabilities but also to investigate their perception and cognitive skills. Some fish, for example, can perceive visual illusions like the Kanizsa figures or the Ebbinghaus illusion (Wyzisk and Neumeyer, 2007; Sovrano and Bisazza, 2009; Sovrano et al., 2015). Moreover, an interesting study showed that cichlids are able to categorise objects and after being trained with sample images, they were able to accurately select the stimuli belonging to the same category, even when the image was of a different object (Schluessel et al., 2012). These cichlids are also able to recognise 3D rotated objects (Schluessel et al., 2014a). Similarly, a recent study showed that archerfish can recognise rotated human faces (Newport et al., 2018). This is striking, as it is a completely unfamiliar stimulus class. In general, these studies show that some species of fish can be used to study complex visual discrimination tasks despite their relatively simple brains and the fact that no homological cortex structure has been found in their nervous system (Northcutt, 2011).

The ability to learn and solve many visual discrimination tasks is believed to require the involvement of the cortex in mammals. Image rotation, object categorisation and face or object recognition are a few examples (Hoffman and Logothetis, 2009; Petruno et al., 2013; Warburton and Brown, 2015). Recently, researchers using mice expressing a calcium indicator and optogenetic stimulation in the cortex could follow the activity related to a visual discrimination task and also showed that inhibition of the visual cortex affects discrimination performance (Burgess et al., 2017). However, not only the cortex is involved in visual learning and discrimination. Recently, some evidence has come to light showing that the mammalian hippocampus is also involved in visual learning and discrimination (Knutson et al., 2012; Hales et al., 2015; Liu et al., 2017; Kim et al., 2018). The Dl of the teleost pallium has been described as a possible homologue of the mammalian hippocampus (Mueller et al., 2011), and it receives visual information (Vanegas and Ito, 1983; Demski, 2003; Northcutt, 2006; Yamamoto and Ito, 2008). These facts make this structure an appealing target for visual learning

117

research in teleost, yet the role of the Dl in visual learning has not yet been explored. An analysis of the potential involvement of this part of the fish brain in solving complex visual discrimination tasks will help our understanding of the homology of brain structures and visual discrimination abilities between fish and mammals. Furthermore, these results will lead to a better understanding of the possible role of the mammalian subcortical pathways in visual processing, and the phylogenetic origins of visual cognitive behaviour in general.

Ambon damselfish have incredible visual discrimination abilities, and they have been shown to quickly learn novel tasks (Siebeck et al., 2009; Siebeck et al., 2014). These features make them an ideal model to investigate visual learning. Therefore, I decided to use Ambon damselfish to investigate if its pallium, and particularly the Dl, is involved in learning and solving visual discrimination tasks. To be able to accurately locate the Dl in the Ambon damselfish telencephalon, it is necessary to be able to differentiate it from the other dorsal pallial structures. The authors of many important brain atlases including rat’s, mouse’s and human’s (Paxinos and Franklin, 2001; Paxinos and Watson, 2007; Mai et al., 2015), pointed out how useful staining with multiple markers is in anatomical research (Watson et al., 2010). Some Immunolocalisation and histochemical techniques commonly used to identify neuronal tissues include the Nissl method, the localisation of the calcium-binding proteins like Parvalbumin (PV) and reaction for Nicotinamide adenine dinucleotide phosphate-diaphorase (NADPH-d) (Baimbridge et al., 1992; Elphick, 1997). These staining methods can be used in Ambon damselfish tissue to map its pallium for the first time. Furthermore, the results obtained using these common markers can then be compared with those of other fish for which similar neuroanatomical analyses already exist (Crespo et al., 1999; Diaz- Regueira and Anadon, 2000; Castro et al., 2006; Giraldez-Perez et al., 2008; Maruska and Tricas, 2009; Mueller et al., 2011; Puschina and Varaksin, 2011; Grana et al., 2012; Harvey-Girard et al., 2013; Dewan and Tricas, 2014; Deshpande and Jadhao, 2015; Palande et al., 2015).

Immunohistochemistry methods can also be used to detect the expression of immediate early genes like c-Fos, the localization of which has been related to neuronal activity and learning (Curran and Morgan, 1987; Mccabe and Horn, 1994; Terleph and Tremere, 2006; Kawashima et al., 2014; Arias et al., 2015; Minatohara et al., 2016). Studies exist where c-Fos or Egr1 were used to find neuronal activity in general and learning related activity in multiple fish (Burmeister and Fernald, 2005; Burmeister, 2007; Harvey-Girard et al., 2010; Rajan et al., 2011; Wood et al., 2011; Barbosa et al., 2012; O'Connell et al., 2012; O'Connell et al., 2013; Yaeger et al., 2014). The key advantage of this method is that neuronal activity can be analysed following the completion of a task. The combination

118

of visual learning tasks and the neuroanatomical techniques described here provide a powerful tool to test the hypothesis that the dorsal pallium is involved in visual learning.

Finally, I decided to disrupt the Ambon damselfish nervous system, as I intended to explore if visual or learning alterations would affect the neuronal activity of the pallium. To do this, I used a new ecologically relevant paradigm which is based on the reported impact of ocean acidification on animal sensory systems and behaviour. As CO2 levels are rising and predicted to continue to increasingly do so, a new research field developed over the last decade which is dedicated to the effects of this acidification on marine animals. Typically, behavioural experiments testing the effect of CO2 levels predicted to occur by 2100 have been carried out investigating reproduction, escape response, predator avoidance, settlement and lateralization, among other behaviours (Briffa et al., 2012; Heuer and Grosell, 2014; Clements and Hunt, 2015; Nagelkerken and Munday, 2016). Some evidence was found which suggested that low pH treatment could be altering vision and learning related behaviours in fish (Forsgren et al., 2013; Jutfelt et al., 2013; Chung et al., 2014). Such impairments in learning and visual behaviours have also been suggested for juvenile Ambon damselfish, which seems to be one of the most sensitive species to this treatment (Ferrari et al., 2011b). Pre-settlement Ambon damselfish exposed for a few days to a lower than normal pH failed to learn the olfactory or visual cues required to avoid predators and generally showed reduced antipredator responses (Ferrari et al., 2012a; Ferrari et al., 2012b; Chivers et al., 2014). The proposed mechanisms for these effects are neurophysiological disturbances, in particular, a different ion flow across the GABA-A receptors (Nilsson et al., 2012; Heuer and Grosell, 2014). GABA-A receptors are ubiquitous in the brain and can also be found in the retina and tectum, two main structures of the visual systems of fish (Lin and Yazulla, 1994; Anzelius et al., 1995b; Anzelius et al., 1995a; Pirone et al., 2006; Ruan et al., 2014). Finally, GABA-A receptors have been related to visual processing and learning (Wan et al., 2004; Mora-Ferrer et al., 2005; Mora-Ferrer and Neumeyer, 2009). Together, the previous findings and the proposed GABA-A theory suggest that a low pH treatment could lead to a disruption of the abilities of Ambon damselfish to learn and solve a visual discrimination tasks, but a systematic and controlled approach has not been attempted yet.

In this chapter, I will show my exploratory results of labelling and mapping the Ambon damselfish pallium and for the localisation of c-Fos expression in this area after visual learning. I will then present the experiments designed to disrupt Ambon damselfish visual learning with a low pH treatment.

119

5.4 Methods

Animals. I used 45 adult Ambon damselfish, 6 for pallium mapping only, 20 for evaluation of pallial learning and 19 for visual disruption studies. For the neuroanatomical studies, animals were caught with hand nets while on SCUBA on reefs near Lizard Island, Great Barrier Reef, Australia (14°41′S, 145°27′E) (fisheries permit: 162472; GBRMPA permit: G15/37305.1). They were then brought to aquarium facilities at the University of Queensland in Brisbane (in individual plastic bags half-filled with fresh seawater and topped up with oxygen) using a commercial aquarium trader (Cairns Marine Aquarium, Stratford, QLD, Australia). Due to their territorial nature, fish were held in individual aquariums with a PVC tube for shelter and water was supplied with a sump system with 5-7 fish per system. Their tanks were cleaned and a water change was performed regularly. Fish were fed a mixture of water and fish flakes 5 times per week. For the pallial learning experiments, 20 were caught from reefs near Heron Island Great Barrier Reef, Australia (23° 26'S, 151° 54'E) and held in individual aquaria at the Heron Island Research Station (fisheries permit: 189852; GBRMPA permit: G15/37305.1). Water was supplied with a flow-through system available at the station. They were fed daily a mixture of water and fish flakes as part of the experiments. At the end of the experiment, 18 fish were euthanised for immunohistochemistry analysis (see below), and 2 fish were released to their collection point. For the visual disruption studies, animals were caught with hand nets while on SCUBA on reefs near Lizard Island, Great Barrier Reef, Australia (14°41′S, 145°27′E) (fisheries permit: 162472; GBRMPA permit: G15/37305.1). 19 adult Pomacentrus amboinensis, of an average standard length of 8.36 cm (SD=0.78) were then evenly distributed by body size into control (n=9) and experimental (n=10) groups. All fish were held individually in aquaria for the duration of the experiment and were supplied with seawater via the flow-through system available in the aquarium rooms of the Lizard Island Research Station. All experiments were conducted according to the animal welfare act Australia and approved by the ethics committee of the University of Queensland (ethics permit SBMS/163/15/ARC).

Histochemistry and immunohistochemistry. Fish were anesthetized with clove oil prior to being euthanised. Their brains were dissected, fixed in 4% paraformaldehyde overnight and stored in PBS at 4°C. Then they were cryoprotected by gradually increasing the concentration of sucrose until the brains sank at 30% sucrose. At that point, they were immersed in OCT compound (Scigen). I then collected coronal 30μm sections separated in 6 series using a cryostat. I performed Nissl staining using the protocol of the SBMS Histology facilities, based on previous methods (Clark and Sperry, 1945). Briefly, sections were washed with water and immersed

120

in Thionin solution (0.05% LiCO3) for ~6min. NADPH-d histochemistry was performed following the methods used by Lopez et al. 2017. The sections were incubated in a medium made of 1 mM b-NADPH (Sigma), 0.8 mM nitroblue tetrazolium (Sigma) and 0.06% Triton X-100 in PB, at 37°C in darkness for 1–2 hours. The reaction was stopped by successive rinses in cold PB. For immunohistochemistry of Parvalbumin, and adaptation of the protocol that I developed previously was followed (Marquez- Legorreta et al., 2016). Briefly, sections were incubated overnight in the primary antiserum (1:1000 monoclonal mouse anti-PV, P3088, Sigma or MAB1572, Chemicon) diluted in PBS with 0.5% TX and 5% bovine serum, followed by secondary biotinylated anti-serum (1:200 goat anti-mouse; 115-065-166, Jackson ImmunoResearch). Then sections were then processed by the avidin-biotin peroxidase complex procedure (ABC; Vectastain, Vector) following the manufacturer’s specifications, and then by standard histochemistry for peroxidase, with or without heavy metal intensification. c-Fos immunolocalisation was performed overnight incubation with 1:500 rabbit anti-c-Fos primary antibody (Santa Cruz Biotechnology, sc-253) diluted in PBS with 0.5% TX and 5% bovine serum, followed by 1:200 goat anti- rabbit (BA-1000, Vector Laboratories) and then processed for the histochemical reaction as for Parvalbumin. Finally, after any of the staining methods, sections were gradually dehydrated with a battery of Ethanol at increasing concentrations, immersed in Xylene and covered using DPX (Ajax Finechem).

Image processing. Photos were taken with a Nikon DS-Fi1 camera on a Nikon ECLIPSE 50i microscope using the NIS- Elements BR 4.0.06 software. Images were cropped, and their brightness and contrast were adjusted uniformly with Adobe Photoshop software. Illustrations were arranged into plates using Adobe Illustrator software.

Visual and learning areas of the pallium-Behavioural Experiment. We performed this behavioural experiment at the Heron Island Research Station. To test the fish with a visual discrimination task, we used a pre-training operant conditioning approach followed by a training and testing period.

Pre-Training: Two different observers trained fish during two sessions every day following procedures established in our lab (Siebeck et al., 2009; Siebeck et al., 2014). Briefly, we first habituated them to feed off a feeding tube, which contained a mixture of fish flakes and water. Using operant conditioning, we gradually trained them to “tap” (push with their mouth) a laminated printed pattern attached to a plastic board. We separated the food delivery in time and location from the stimulus to avoid olfactory cues. Only after tapping the stimulus, the fish received their food reward. Stimuli were 3cm x3cm laminated paper print-outs, displaying either a blue circle or a yellow cross

121

on a white background. We trained half of the fish on the circle and the other half on the cross. This was done to control for a possible bias towards a particular stimulus.

Training: When fish had an association with the positive stimulus (S+), we introduced the negative stimulus (S-) and continued training them using a 2 alternative forced-choice paradigm (2AFC). For the fish to reach the learning criterion, they had two complete two consecutive sessions with 75% or higher accuracy. Stimulus position was randomised to prevent positional learning, but with the following constraints: S+ could not be in the same position in the first two trials and could not be on the same side more than twice in a row to avoid the development of side bias. The board displaying the stimuli was removed from the aquarium following a correct completion of the task, which was rewarded with food. The stimuli were also removed if a fish took more than 2 min to complete the task, and the next fish was tested. Fish completed 12 trials twice a day.

The experiment consisted of training Ambon damselfish to discriminate images and euthanising them at different points during learning to locate the areas involved in visual learning with the use of c-Fos immunolocalisation. 20 fish were trained to discriminate a blue circle and a yellow cross. 18 successfully trained fish (reached ≥75% in two consecutive sessions) were separated into 5 groups: Experienced (n=4) fish, which continued training after reaching criterion for another 10 sessions. Learners (n=4), which were euthanised after reaching criterion. 2ndPattern (n=4), which were trained to the second pair of images after they reached criterion for the first one and were euthanised when they reached criterion on the second pair. Sham (n=4), which were trained and reached criterion, then continued to be tested but in their 2 last sessions (1 day) we introduced the board and were periodically fed, but no stimuli were presented. The previous 4 groups were euthanised 1h after their last session. The Baseline group (n=2), was trained and reached criterion, then continued to be tested but were euthanised>4hrs after their last session so c-Fos activity was not related to the visual task. The hypothesis was that Learners would present the most activity in the Dc and Dl, followed by the 2ndPattern group, the Experience group, sham group and finally the baseline group in decreasing order. I expected that 2ndPattern group would show only activity related to learning to discriminate visual stimuli, while the Learner group would also show the activity produced by the training. When the behavioural part of the experiment was completed, brains were dissected and fixed following the methods of previous studies for localising c-Fos in fish (O'Connell et al., 2012; Yaeger et al., 2014).

122

Visual disruption-behavioural experiment. Two behavioural experiments (Figure 5.1) were planned for a month-long field trip which was considered enough time for the different learning, testing steps and visual disruption based on previous studies (Siebeck et al., 2009; Munday et al., 2010; Ferrari et al., 2012b; Chung et al., 2014; Siebeck et al., 2014).

Pre-Training, training and testing: Three different observers trained an identical number of control and experimental fish (4-4, 3-3 and 3-3). Fish were trained during two sessions every day following the procedures described above. Stimuli were 3cm x3cm laminated paper print-outs, displaying either the outline of a blue triangle or a circle on a white background. Half of the fish were trained on the triangle and the other half on the circle. Training and testing phases were performed as described above but with 10 trials twice a day. We deemed the fish to have learned the task if they reached the learning criterion of two consecutive sessions at an accuracy of at least 70%.

Experiment 1- Low pH visual discrimination: CO2 was added to the sump of the treatment fish, lowering the pH gradually by 0.1 units every 24h until we reached a pH of 7.8. pH 7.8 was selected because it represents the worst-case scenario predicted to occur by the year 2100, (>1000 pCO2, -0.3 pH RCP8.5 scenario [2014 SPM IPCC] (Pachauri et al., 2014)) and as such it has been commonly used in previous studies. Overall, during experiment 1, the treatment fish were exposed to elevated

CO2 levels for a total time of 184 hours and at pH 7.8 for 136 hours. During this time, both groups of fish continued to be tested twice daily to assess any changes in their visual discrimination accuracy.

Experiment 2 – Low pH visual learning: Following experiment 1, we tested whether the ability of

CO2 exposed fish to learn a new discrimination task (Odd One Out task) differed from that of the control fish. The stimuli were 3cm squares that displayed four symbols each. The four symbols were identical for S- and one of the four symbols was different from the other three for S+. To receive a food reward, the fish had to identify the S+ stimulus containing the odd symbol. Stimuli were never repeated so that the fish saw a new pair of stimuli every single trial.

123

Figure 5.1. Project design for visual testing CO2 effects. a. Pair of stimuli used for Experiment 1 and examples of pairs used for Experiment 2. b. Timeline of Experiment 1 and 2. Experiment 1 consisted of the training and testing of the discrimination of the first pair of stimuli in a prior and after acute and maintained exposure to increased CO2. For experiment 1, I compared the discrimination accuracy of fish at period 1 (P1) against period 2 (P2). The acidification of the water for the Treatment group started at the end of the 12th day and reached a pH of 7.8 at the end of the 14th day. Experiment 2 consisted of the training to the odd one out task.

Visual disruption-CO2 procedure We pumped ocean seawater into two 300L sumps and then into the individual aquaria (26L, ten per condition) at a continuous flow of 880-1040 mL/min, which guaranteed a complete water turn over at least every 30 min (Eheim 1262 Universal Water Pump; Output GPH: 910). The use of sumps allowed us to control the pH and pCO2 and the exposure in all aquaria of each treatment group simultaneously. For both groups, we used identical sumps which were fed by the same water supply and exposed to identical light and temperature conditions.

For the treatment group, a pH probe (Mettler Toledo, InPro4501VP), which recorded the pH was placed into the sump. It was connected to a pH interface (Aquatronica, ACQ210N-PH) and then to a controller unit (Aquatronica ACQ 110) which opened a solenoid to allow the flow of food-grade CO2 124

into the sump when the pH was above the desired value. It also logged the pH every 30 min. The average pH of the seawater (April 2016) was 8.1. When the treatment started, the pH was lowered gradually (0.1 every 24h) until pH 7.8 was reached. The pH in the control group was not altered and had natural fluctuations, the sump pH was logged every 30 min, and the value was around 8.1 (SD= 0.036). We calibrated pH probes daily to avoid any drifting. Daily, pH of alternated halves of the Treatment groups’ individual aquaria was recorded to confirm that the pH in the sump and treatment aquaria matched. Water samples were taken from both sumps as specified below to calculate the pCO2 that was on the sumps during the experiments. A temperature probe logged the temperature every 30 min and was alternated between sumps. Both sumps had an average temperature of 28.82 °C with natural daily fluctuations (± 1°C).

Visual disruption-Water chemistry analysis We collected samples for water carbonate chemistry analyses from both sumps at midday and midnight 4 times during the study. On the first day that the treatment group reached pH 7.8, and then at the 4th, the 8th and the 11th days after. The samples were taken in glass containers, pH and salinity were logged. Total alkalinity was measured using Gran titration method (Kline et al., 2012) at the

University of Queensland (Table 5.1). Total alkalinity and pH data were used to calculate pCO2 using the programme CO2sys (http://www.cdiac.ornl.gov/oceans/co2rprt.html) using pH Scale: NBS scale

(mol/kg-H2O), CO2 Constants: K1, K2 from Mehrbach et al., 1973 refit by Dickson and Millero, 1987, KSO4 Source: Dickson, Total Boron Source: Uppstrom, 1974. Salinity (35.2) was recorded from the sensor-floats deployed at Lizard Island (Integrated Marine Observing System (IMOS), 2017).

Table 5.1. Water chemistry and calculated pCO2

Group Temperature Salinity pHNBS TA (µmol/kg pCO2 (ppm) pCO2 at 25° (C°) (ppm) SW) (ppm) Control 28.56± 35.2 8.07±0.02 2282.27± 560.13± 483.53± 0.52 8.73 37.87 29.22 Treatment 28.56± 35.2 7.8±0.01 2281.84±6.85 1151.91±53.9 1001.4±36.95 0.52

mean±SD of the Temperature, Salinity, pHNBS, Total Alkalinity and pCO2. The pCO2 was calculated using the programme CO2sys (see Methods).

125

Visual disruption-Data analysis. During each session, we scored the number of correct choices (taps on target) as correct or incorrect. For Experiment 1, two periods of 40 trials were taken as a sample. P1: the 40 trials before the start of acidification and P2: the last 40 trials before Experiment 2. With the exception of fish 8 whose P1 sample was of 27 trials, at it reached 2 min without responding in some sessions. Only fish that had reached criterion on the 2AFC discrimination task were taken into account for the analysis of Experiment 1. The other fish were still trained or kept doing the 2AFC task and were later included in Experiment 2.

The raw data were exported from Excel sheets and analysed using R 3.3.0 (R Development Core Team, 2018). Most of the results were analysed by generalised linear mixed-effects models to assess the effect of the Treatment group and the Period. For these data sets, mixed-effect models offer a better fit of the data than the repeated measures ANOVA. This method was also chosen as it allowed us to control for random effects (e.g. influence of trainers or individual fish). We used the lme4 package to perform our model analysis, with a binominal distribution (logit). The final models were obtained by a backwards stepwise term deletion routine, where significance of the elements of the starting model where assessed and eliminated when pertinent. The final model to evaluate the effects of the treatment group and period was:

Y = Group + Period + (1|Trainer) + (1|Fish)

To analyse the differences between groups to reach criterion the model was:

Y = Group + (1|Trainer) + (1|Fish)

For analysis, the contribution of the parameters Group and Period were obtained using the maximum likelihood method (Laplace Approximation). This technique creates a likelihood ratio, known to approximate a chi-square distribution that was used to assess if a first model is significantly different from a second model. For the analysis between groups before the start of the treatment, I used similar procedures as above. Briefly, I used GLMM with the glmer or glmmadmb R functions of the lme4 package with binomial, Poisson or “truncpoiss” distributions depending on the response variable analysed.

To calculate if the group performance was above chance, I used the Intercept results of the Z test of the model. For the individual fish accuracy analysis, I used the binomial test since the outcome of all

126

the experiments were either a correct or incorrect choice. This analysis was done in Excel, and the cumulative probability of the correct choices [P(X≥x)] from each period (40 trials) was calculated.

5.5 Results

5.5.1 Neuroanatomy of the Ambon damselfish pallium

The sections presented heterogeneity of tissue fixation, meaning that I still needed to optimise the protocols I followed in order to get better results. However, some labelling patterns were nevertheless noticeable. With the results of the Nissl staining (Figure 5.2a), I was able to identify 3 of the main divisions of the dorsal pallium (Figure 5.2b). The lateral division was easily identifiable by its location at the lateral edge of the pallium and by the high density of small and round Nissl stained neurons. It extends from the sulcus ypsiloniformis (Sy) until the sulcus externus that separates it from the ventral pallium (not shown). Dorsally, the Sy makes a clear boundary with the Dm3, which has a lower density of neurons. Medially to the Dm3, a very densely populated area known as Dm4, continues as a thin structure through the medial edge and reaches subpallial structures. These three structures surround the Dc and are separated from it by a circular boundary. Furthermore, the Dc is the least populated area, with more segregated and larger neurons than the surrounding structures (Figure 5.2b and e). The areas were defined using the work of Dewan and Tricas on butterflyfish as a reference (Dewan and Tricas, 2014). NADPH-d staining also presented clear patterns across these dorsal structures (Figure 5.2c), with the exception of a group of large neurons around a fibre bundle in the hypothalamus (Figure 5.2f). Most of the labelling was found in the neuropil of the dorsal pallial structures with different patterns of intensity. The most stained area was the Dm4, followed by the Dl and in a similar intensity the Dc. Finally, the Dm3 region is the one that presented the weakest staining. Finally, PV immunohistochemistry also reflects a distinctive pattern. Both the Dl and the Dc present darker staining due to labelling of the neuropil of these areas. Very small, segregated PV positive neurons can be found in the outer bound of the Dm and the Dl (Figure 5.2d).

127

Figure 5.2. Nissl, NADPH-d and Parvalbumin staining of adult Ambon damselfish brains. a. Ambon damselfish sample coronal sections ordered from caudal to rostral. Pallium stained sections for Nissl (b), NADPH-d (c) and PV (d). The Dl presents both small round parvalbumin-positive neurons and parvalbumin-positive fibres, while the Dm presents only positive neurons and the Dc only positive fibres. e. Zoom into a Nissl stained pallium section to see cytoarchitectural details. f. Example of NADPH-d stained neurons in the hypothalamus.

128

129

Figure 5.3. HIRS behavioural experiment. a. Ambon damselfish being trained with a blue circle as positive stimulus. b. Ambon damselfish in a testing session. c. Same Ambon damselfish as in b being rewarded after the correct choice. d. Accuracy of the individual fish during testing sessions across the different groups.

Figure 5.4. c-Fos in Ambon damselfish pallium. c-Fos labelling as dark dots in the Dl. a. Ambon damselfish of the 2ndPattern group with positive c- Fos neurons in the Dl. b. Zoom of rectangle depicted in a. Some of the positive c-Fos neurons are indicated by arrows. c. Ambon damselfish of the Experienced group with positive c-Fos neurons in the Dl.

5.5.2 Pallial visual learning in Ambon damselfish

We successfully trained fish following our operant conditioning training (example in Figure 5.3a-c) and the ones that were tested after reaching criterion maintained a high performance (Figure 5.3d). I processed a subset of the samples from the learning experiment with some promising but inconclusive results. As mentioned above, the tissue fixation was not optimal and labelling was very heterogeneous. Nonetheless, I found c-Fos labelling in telencephalic areas in at least 5 fish, which suggest that interesting results would come after further optimisation of the methods. Positive c-Fos labelled neurons could be found in the Dl (an example is shown in Figure 5.4), but also in some cases in the Dc and in medially located nuclei from the subpallium (not shown). Interestingly, samples from negative control groups (Baseline and Sham) did not present c-Fos positive neurons in the telencephalon. However, their absence could be to the state of the tissue fixation.

130

5.5.3 Disruption of visual discrimination and learning

Experiment 1: Learning the discrimination task 15 fish (6 Control and 9 Treatment) were tapping their reward stimulus within 3-6 days after being captured, and on the 8th day 15 fish (6 Controls and 9 Treatment group) had completed pre-training and were doing the 2AFC task. By the time of the start of the acidification treatment, 18 fish were on the 2AFC task, and 12 had reached criterion.

From the model analysis, the variance explained by random effects was higher for individual fish (0.1154, SD= 0.3397) than for trainers (0.07531; SD= 0.2744). The other random variables controlled (time of the day and period) showed a variance close to 0. There was no difference between groups 2 in the time needed to start the 2AFC task (χ 1 = 1.0707; p= 0.3008) or criterion (Deviance=1.2094; DF=1; p= 0.2715). Twelve fish reached criterion before the treatment started (4 control and 8 treatment fish) and neither was there a difference in the number of fish that reached criterion per 2 group (χ 1 = 3.4165; p= 0.06455).

Although there was individual variability in the time it took each fish to associate the S+ with a food reward, and in the number of sessions needed to reach criterion (Figure 5.5a), both groups showed similar improvement rates (Figure 5.5b). Overall, no significant differences were observed between the groups before the manipulation of CO2 levels.

Figure 5.5. Learning discrimination task of experiment 1. a. Example of the performance of four individual fish (coloured lines) from the start of the 2AFC task until criterion. The black dashed line represents the 50% accuracy = chance and the red dashed line represents 70% accuracy. Fish had to achieve two consecutive sessions of ≥70% accuracy to reach criterion. b. The percentage of fish with an accuracy of ≥70% is shown for both groups and all sessions prior to the start of the CO2 treatment. 131

Experiment 1: Differences in discrimination performance between Periods P1 and P2

Average group accuracy for the treatment group was 73.125% (SD ± 10.807) at P1 (before CO2 treatment) and 68.125% (SD ± 11.973) at P2 (following CO2 treatment) (Figure 5.6a). For the control group, the average accuracy was 80.833% (SD ± 9.519) at P1 and 75% (SD ± 11.319) at P2. Both groups had an accuracy above chance at P1 and at P2 (Controls-P1: Accuracy= 80.8% -Z value= 0.2776; p= 4.6e-07-; Controls-P2: Accuracy= 75% -Z value= 0.2731; p= 3.31e-05-; Treatment-P1: Accuracy= 73.1% -Z value= 0.2074; p= 5.21e-07-; Treatment-P2: Accuracy= 68.1%, Z value= 0.2044; p= 1.52e-04-).

No interaction between time (P1/P2) and treatment (control/CO2) was found for the level of accuracy 2 2 of the fish (χ 1= 0.0334; p= 0.855). Also, there was no difference in accuracy between groups (χ 1= 2 1.4609; p= 0.2268), and between periods (χ 1= 2.4504; p= 0.1175).

Experiment 1: Individual changes Analysis of individual performance shows that no individual fish had a significant effect on the group 2 drop at P2 (χ 2 = 2.853; p= 0.2401, fish analysed as a random variable). There were also no significant 2 changes in individual accuracy between P1 and P2 (χ 11 = 19.098; p= 0.05936). This was still true for the fish with the biggest drop (Figure 5.6b; fish 13: Accuracy at P1= 0.75%; Accuracy at P2= 0.525 %; p= 0.07205).

I also could not find a relation between fish size and any possible effect of the acidification treatment (Figure 5.6c).

132

133

Figure 5.6. Experiment 1- effect of high CO2 on visual discrimination. a. Performance accuracy per group for the two different periods analysed (before and following exposure to CO2). The black dashed line at 50% represents chance accuracy. b. The individual average performance of each fish at P1 and P2 ordered by size, the smallest at the left. Bars in red represent treatment group fish, and bars in blue represent control fish group. The black dashed line at 50% represents chance accuracy, and the red dashed line at 63% represents the threshold were p<0.05 in a binominal distribution. For fish 8 the threshold for a p<0.05 at P1 would be at 68% approximatively. c. Individual difference of percentage accuracy between P1 and P2 in relation to the total length and a fitted line for Treatment and Control groups.

Experiment 2: Learning the odd one out task 17 Fish reached criterion in the new visual discrimination task (9 from the control group and 8 from the treatment group). 11 fish had learned the task within 2 days (4 from control group and 7 from treatment group) (Figure 5.7a). The control fish took on average 3.44 (SD ± 1.922) sessions and the treatment group 2.25 (SD ± 0.707) sessions. However, this difference was not significant. (Deviance=2.6958; DF=1; p= 0.1006).

There was also no significant difference between the group’s accuracy, in the 2 sessions to reach 2 criterion (χ 1 = 0.5544; p= 0.4565). The treatment group achieved 80% (SD ± 9.258) accuracy and the control 76.666 % (SD ± 5.59) (Figure 5.7b).

Figure 5.7. Experiment 2- Learning under high CO2 conditions.

134

a. Cumulative percentage of fish that reached criterion in consecutive sessions. b. Experiment 2- Performance. The average accuracy of both groups with the 2 sessions prior to criterion. The black dashed line represents the 50% chance accuracy.

5.6 Discussion

5.6.1 Pallium neuroanatomy and its involvement in visual processing

The Nissl results suggest a very similar pallial cytoarchitecture to what has been described for butterflyfish (Dewan and Tricas, 2014), adult zebrafish (Wullimann et al., 1996) and goldfish (Northcutt, 2006). In general, the Dm and Dl, separated by the Sy, have a more dense population of neurons and surround the Dc, which has a less dense distribution and larger neurons. The other labelling methods present some variability across species and studies, although they remain useful for distinguishing the pallium subdivisions. For example, researchers that used NADPH-d labelling in goldfish found that all the dorsal pallial areas contain labelled cells (Giraldez-Perez et al., 2008; Giraldez-Perez et al., 2013), however, others only found them in the Dc and Dl in the same species (Villani and Guarnieri, 1995). The latter pattern is also described for adult zebrafish (Mueller et al., 2011) and except for the labelling in the Dm4 is also the pattern suggested based on my results. However, although the Dc and Dl of Ambon damselfish brains presented stronger blue staining than the Dm3, I could not find clear labelling of neurons in this area. Furthermore, PV immunolocalisation in the Ambon damselfish dorsal pallium is also similar to that found in adult zebrafish (Mueller et al., 2011; von Trotha et al., 2014). In zebrafish, PV stains predominantly the Dc’s and Dl’s neuropil but not the Dm’s. This pattern only differs from the one found in Ambon damselfish from the Dm positive PV neurons found in the present study. Positive cells in this area have been described before in the Tinca tinca (Crespo et al., 1999), which suggest that although there is a general pattern in teleosts, some differences of PV expression can be found across species. Altogether, these results suggest that the pallium of Ambon damselfish has a similar conformation as those of other teleosts. As a result, it is possible to generalise the results found for Ambon damselfish and compare them with results found for other teleosts.

The presence of c-Fos positive cells in the Dl of fish after performing visual discrimination tasks s that these cells were active during this behaviour. However, weather this activity was due to a learning process or to other non-related activity is elusive. The localization of this immediate early gene has been related to cell structural changes associated to a strong activity (Curran and Morgan, 1987; Minatohara et al., 2016), which strongly suggests the involvement of this structure in visual

135

processing and possibly learning. This idea is supported by the fact that no c-Fos positive neurons were found in the negative control groups. Nonetheless, it is not possible to confidently discard the possibility that their absence is not just due to the fixation problems, and that some non-cognitive or baseline activity could explain the results found. An optimisation of the protocols and further more detailed studies are needed to confirm and expand these observations.

These preliminary results suggest that the abilities of the Ambon damselfish, and likely also those of other highly visual fish species, to perform visual discrimination tasks could be used to investigate which structures in the brain are involved in processing the information in the absence of cortical structures. This will help us understand the overlooked role of subcortical mammalian pathways in visual learning and other cognitive tasks (Zackon et al., 1997; Parvizi, 2009). So far, studies with these species have focused on behavioural experiments, but with the existing labelling and electrophysiological methods, functional experiments could also be implemented. In particular, these studies could help us to understand the role of the optic tectum (superior colliculus in mammals) and the Dl (hippocampus) in visual learning. A better understanding of this pathway would be relevant for developing therapies for patients with cortical damage (Garrido, 2012; Bertini et al., 2016; Hadid and Lepore, 2017).

5.6.2 Visual discrimination and learning are not affected by lower pH

The results of this study indicate that exposure to lower pH levels did not affect the visual discrimination performance of adult Ambon damselfish. There was a small non-significant drop at P2, which was also present in the control group. This could be due to a loss of motivation in some of the fish, maybe due to the regular feeding. Furthermore, looking at the individual fish difference between P1 and P2, most of them kept performing significantly above chance, which means that they were able to accurately discriminate the stimuli. To test for possible effects in the learning process we trained the fish to a new discrimination task. The odd one out task was chosen as it is a more complex exercise, which we thought would make it more sensitive to alterations caused by the treatment. In this kind of task, a general rule needs to be learned as the stimuli presented are always new, and there is no single feature (size, shape or colour) that will serve as a specific cue. Therefore, it is considered a task where concept learning is involved, as fish need to learn that the oddity of a stimulus is the cue for the reward. Our results show that fish from both groups were able to do it within a few days. From the treatment group, only 2 fish did not learn the task in the time we gave them. It is possible that they would have learned it eventually.

136

It is important to consider that interpreting non-significant results most be with caution (Makin and Orban de Xivry, 2019). Indeed, a lack of significant effect produced by a decrease in pH in this study could be due to a small sample size, which only allows to detect large effects. A bigger sample would be needed to allow a higher statistical power and a more accurate interpretation. However, even a larger sample size could find a mild but significant effect of increased CO2 in visual learning, the results from this study suggest that this effect would unlikely be enough to alter the fish performance in a meaningful way.

The previous studies that analysed visual and learning effects on juvenile Ambon damselfish did so trying to address ecologically relevant stimuli and conditions, especially with responses to predators (Ferrari et al., 2012a; Ferrari et al., 2012b; Chivers et al., 2014). The problem with this approach is that, although it is done to be more realistic, not all variables can be properly controlled and behavioural results could be explained by other effects, like an increase in boldness for example (Munday et al., 2014). Likewise, although there are some studies suggesting learning impairment in juveniles Ambon damselfish (Ferrari et al., 2012a; Chivers et al., 2014) more detailed research is needed to see if the learning process was actually impaired, or if it was indirectly affected by an alteration of the sensory system tested (the olfactory system). Finally, in a previous study with Acanthochromis polyacanthus damselfish, researchers found changes in the retinal activity, which could affect temporal resolution sensitivity (Chung et al., 2014). It is possible that the changes in the retinal activity produced by lower pH are not enough to produce a visual discrimination impairment of static objects but could affect the processing of moving objects. This would require further investigation.

Our water analysis indicates that we successfully matched the ambient pH and pCO2 levels and were equal or more severe than in previous experiments (Table 5.1) (Ferrari et al., 2012a; Ferrari et al., 2012b; Chivers et al., 2014). This discards a failure in the set up as explanation for the lack of effect in our study. Likewise, previous studies claim that 4 days are enough to produce an effect on juvenile

Ambon damselfish exposed to 700ppm of CO2, and at higher levels (850ppm) even 2 days are enough (Munday et al., 2010). Besides, Exp2 was performed after 2 more days under a lower pH, and it also involved visual discrimination. This means that in my experiments fish were exposed for a long enough time to produce an effect according to the published literature.

An important difference between our study and previous ones is that they used Ambon damselfish juveniles (Ferrari et al., 2011b; Ferrari et al., 2011a; Devine et al., 2012b; Ferrari et al., 2012a; Ferrari et al., 2012b; Allan et al., 2013; Couturier et al., 2013; Chivers et al., 2014; Munday et al., 2016). We needed to use adult Ambon damselfish to be able to train them to the 2AFC task, but it is possible 137

that adult fish do not present sensitivity to this treatment. However, studies in which a similar CO2 protocol was used for adult fish of comparable or bigger size than adult Ambon damselfish still found behavioural effects (Cripps et al., 2011; Devine et al., 2012a; Jutfelt et al., 2013; Munday et al., 2013; Dixson et al., 2015; Ferrari et al., 2015; Lai et al., 2015; Heuer et al., 2016). Furthermore, researchers that evaluated the responses to chemical alarm cues in both, juvenile and adult Acanthochromis polyacanthus, have found olfactory effects at both life stages (Welch et al., 2014; Heuer et al., 2016; Schunter et al., 2016). These last studies suggest that adults of a species that shows effects when young, are likely to present also behavioural changes as adults.

Following my results and a critical review of the literature, I consider that the GABA-A theory (Nilsson et al., 2012; Heuer and Grosell, 2014) needs to be reviewed. The ubiquity of GABA-A in the brain would predict a much wider array of effects; however mostly olfactory impairments have been found, and findings are sometimes contradictory even when the same species is used (Welch et al., 2014; Sundin et al., 2017). Furthermore, gene expression analysis of tissue under high CO2 conditions has only found small differences in GABA related genes (Schunter et al., 2016; Lai et al., 2017). Finally, although multiple studies have shown different behavioural effects, there are also numerous reports where fish behaviour remains unaltered after low pH and CO2 exposure (Nowicki et al., 2012; Jutfelt and Hedgarde, 2013; Maneja et al., 2013; Jutfelt and Hedgarde, 2015; Maneja et al., 2015) (Duteil et al., 2016; Heinrich et al., 2016; Kwan et al., 2017). This variability of results strongly suggests that a careful and replicable approach is needed to really isolate the possible effects of exposure to low pH. Interestingly, a very recent and thorough publication presents results of research that tested and tried to replicate multiple behavioural effects of high CO2 conditions in various species of damselfish, across different life stages and geographical regions (Clark et al., 2020). This study was unable to find consistent effects after treatment exposure, leading to the conclusion that end of the century levels of CO2 do not alter coral reef fish behaviour in a meaningful way.

The GABA-A theory was postulated considering some known basic facts about fish physiology in hypercapnia and acid base regulation, together with the GABA excitatory effect of developing nervous systems. Briefly, to reduce acidity, fish accumulate HCO3- which then leads to a reduction of Cl- to compensate (Heuer and Grosell, 2014). Under normal conditions, GABA-A receptors would allow the inflow of Cl- and HCO3- which produces the hyperpolarization of the neuron. The GABA- A theory proposes that the ion concentration changes produced by the acid based regulation of the fish causes an inverse flow of ions when the GABA-A receptors are open, leading to a depolarization (Nilsson et al., 2012). A similar phenomenon that involves ion transporters has been described in

138

developing neurons and could be related to some psychiatric diseases (Knoflach et al., 2016). The main support for the GABA-A theory was obtained by a few experiments where researches used the GABA-A receptor antagonist gabazine and were able to recover normal behaviours (Nilsson et al., 2012; Chivers et al., 2014; Chung et al., 2014; Lai et al., 2015; Munday et al., 2016). However, a recent and thorough review of the GABA-A receptor theory pointed out at the complexity of the physiological mechanisms that would be involved and how little they have been approached by previous research so far (Tresguerres and Hamilton, 2017). Among their comments, the authors highlight the need to test other pharmacological and molecular ways to test the GABA-A theory apart of the use of gabazine. They draw attention to the fact that many studies have found effects at much lower pCO2 concentrations than the one needed to actually alter the ion concentration enough to produce depolarization (Heuer and Grosell, 2014). This could be due to the fact that the models have incomplete information of what is actually happening in a living organisms. Therefore, they encourage making accurate measurements of the cerebrospinal fluid and neuron cytoplasm, which has not yet been done. Finally, they also remind the community that the nervous system of fish most have regulatory mechanisms and ion transporters to maintain a functional physiological homeostasis, and that this would interact with the alterations produced by the CO2 acidification. Agreeing with their recommendations, I consider that a comprehensive electrophysiological analysis and exploration of other possible reasons for the observed effects are necessary. For example, if acidosis in the extracellular fluid does occur, local pH changes could affect the acid-sensitive channels (Wemmie et al., 2008; Soto et al., 2018), directly affect GABA-A receptors (Wilkins et al., 2005; Ruusuvuori and Kaila, 2014), interactions between these two channels (Zhao et al., 2014a) (Storozhuk et al., 2016) or other channels (Liu and Chen, 2014).

5.6.3 Conclusions

My results suggest that the pallium of the Ambon damselfish has an organisation and histochemical labelling similar to other teleost and that common labelling techniques would help in more detailed studies of Ambon damselfish neuroanatomy. Furthermore, finding positive c-Fos neurons in the Dl of the Ambon damselfish after learning and performing visual discrimination tasks suggest that this area could be involved in cognitive visual processing. This means that this animal model would be suitable for exploring how the subcortical visual pathway to the hippocampus is implicated in visual tasks, which would help us understand how this pathway might also be relevant in mammals. However, optimisation of the fixation and histochemical methods are needed to confirm these findings and continue with this investigation. Finally, adult Ambon damselfish visual discrimination 139

and learning are not affected by exposure to the lower pH predicted to occur in the near future. The nervous system of this animals can cope with these environmental changes and still allow the fish to behave normally in the tested task. This means that the method was not useful to disrupt visual abilities, probably because it might not affect the nervous system by the mechanisms proposed so far. On the other hand it means that at least the Ambon damselfish visual capabilities are not in danger, at least for currently predicted ocean acidification scenarios.

140

6 Discussion

6.1 General discussion

The studies presented in this thesis have explored different aspects of visual learning using two species of teleost. First, using whole-brain calcium imaging to study visual habituation in larval zebrafish, I was able to explore for the first time how different populations of neurons that participate in the visual responses to a threatening stimulus present different activity profiles and disengage at different rates as habituation occurs. Using theoretical network analysis, I was able to find results that suggest that moderately habituating neurons in the tectum are likely key modulating these responses by integrating information from other types of responses and inputs from different parts of the brain.

Using the same approach, I found that fmr1 mutants present slower habituation and more pronounced recovery of the loom response. Analysing the networks involved, it seems that compared to wild types, the fmr1 mutants maintain and reengage more connections with a functional type of neurons that habituates quickly and that is likely responsible for motor responses and higher processing of the visual stimulus. Finally, and although more work is needed, I tried to describe the relationship of the

Ambon damselfish Dl in visual learning. The results suggest that this structure is involved during visual discrimination and possibly during visual learning. Overall, the work presented in these chapters show that these two species of teleost are valuable to study visual learning, and that the results of the research conducted in this thesis add new knowledge to multiple fields, including habituation, subcortical visual processing, teleost visual neuroanatomy and how these can help understand mammalian nervous system and psychiatric disorders.

6.2 Whole-brain visual habituation

A key finding from chapter 3 is that visual habituation responses show variable rates across different neuronal populations. Simultaneous different profiles of habituation were also found before in the

Aplysia (Falk et al., 1993) and crab (Tomsic et al., 2003). Particularly, it is very interesting that a 141

simpler nervous system model, like the abdominal ganglion of the Aplysia, presents clusters of habituating neurons that are very similar to the ones described in this study. Using voltage-sensitive dye recordings, Falk and colleagues described a group of neurons that gradually habituate; however, they continue responding at every stimulus presentation, like the moderately habituating neurons found in my dataset. In this study researchers also show a group of neurons that habituate minimally and maintain a similar response for the rest of the presentations, which seems analogous to the weakly habituating neurons. Furthermore, they describe also a group of neurons that habituate after a few presentations and remains virtually inactive, similar to the strongly habituating cluster. Finally, although less clearly, they present also a group of neurons that seem to be inhibited by the stimulus.

The similarity in the findings of the study in the Aplysia and the present study is impressive. It suggests that a common dynamic phenomenon could emerge from networks habituating to a stimulus.

Moreover, these results indicate that the nervous systems of very different animals produce subpopulations of habituating neurons, suggesting that this is a conserved mechanism of habituation.

A relevant implication of this finding is also that it reinforces the idea that habituation is not happening only at the level of small circuits (Marsden and Granato, 2015), but that it involves a substantial number of neurons and the interaction of their responses. Recently, multiple studies have shown that a wide range of molecular mechanisms can be co-affecting habituation, or that the same mechanisms can affect habituation in different circumstances (McDiarmid et al., 2019). This has led to the idea that habituation is actually a very complex phenomenon where no single mechanism seems to control it, and rather there are multiple factors that influence the outcomes of the process. To this new complex view of habituation, we have now to add the idea that a compound network composed of elements with different activity is interacting and modulating habituation. It is the combination of the multiple molecular mechanisms and the interaction of these networks that finally produce the behavioural output.

142

However, despite the complexity of the elements that form habituation, they seem to converge in similar functional properties. I mention the similar clusters found in the abdominal ganglion of the

Aplysia and their different profiles of habituating responses to tactile stimuli. Similarly, Tomsic and colleagues describe visual neurons in the crab that respond to moving edges, and that habituate gradually (Tomsic et al., 2003). Interestingly, other visual neurons don’t seem to habituate or habituate minimally. However, even more strikingly, the responses of the movement detector neurons

(MDNs) that do habituate seem to match the behavioural escape responses of the crab. This result is similar to the one I found with the moderately habituating neurons, mostly located in the optic tectum.

Unfortunately, the behavioural output in the calcium imaging experiments was not the same as that in the free-swimming setup. However, the optic tectum responses were the ones that had more similar dynamics compared to the behaviour (Chapter 3, Figure 3.12). Furthermore, the network analysis suggested that connections that involved moderately habituating (blue) nodes, mostly localized in the tectum, where the ones more sensitive to recovery (Chapter 3, Figure 3.13e). This result suggests that these nodes are more likely to present this key characteristic of habituation, as in the behavioural results. The habituation dynamics of these neurons are likely the result of the integration of other populations with different habituating profiles. The study from Tomsic and the one presented in chapter 3 suggest that although the visual responses are processed by multiple elements, this information converges in a population of neurons that process it and that it is likely key in the behavioural transformation.

In the case of zebrafish and most teleost, the optic tectum is the most likely visual structure in charge of integrating the information that would lead to the behavioural habituation. Its connectivity strongly supports this idea as it communicates with higher centres and with premotor areas. Its connections with the retina, nucleus isthmi, pretectum, thalamus, telencephalon and reticulospinal neurons, allow the optic tectum to be a converging point of information. Interestingly, very similar habituating responses have been found in the equivalent structures of other vertebrates. In electrophysiological

143

recordings in the superior colliculus of monkeys, Boehnke and colleagues found visual habituating neurons with different responding profiles, but showed a strong similarity to the responses from the moderately habituating neurons presented here (Boehnke et al., 2011)(See Boehnke et al, 2011,

Figure 1B and 2A). This group of neurons were also sensitive to the changes in the ISI, and a subset of them also responded when performing eye movements. Finally, their responses were also modulated by the relative brightness of the stimulus. These results suggest that the superior colliculus/OT is involved in the detection of salient visual stimuli and orienting movements towards them, which seems to be a preserved role found in mammals, birds and fish (Dutta and Gutfreund,

2014; Zhaoping, 2016; Northmore, 2017) (White et al., 2017).

Despite the apparent key role of the tectum in the processing of visual inputs and its likely integration of multiple features that lead to the habituation response, the responses across multiple brain regions suggest that there is not a single group of neurons that drive habituation visual escape responses on their own. Different brain structures would play a role and modulate habituation in intricate ways.

Similarly, the acoustic startle response of rats can be elicited by a simple pathway in the brain stem, but that can also be directly or indirectly modulated by multiple structures including the cortex, the basal ganglia, hippocampus and amygdala (Koch and Schnitzler, 1997). As outlined in chapter 2, the circuits involved in the visual escape responses in zebrafish include multiple nuclei apart from the core visuomotor circuits and could modulate loom habituating responses. For example, both the pretectum and the thalamus seem to modulate habituation to looming stimulus in toads (Ewert, 1970).

This opens many questions that would be very interesting to follow and that the technological advances compatible with zebrafish larvae would be able to address.

An interesting question that arises from this study is the identity and characteristics of the neurons that have different habituating profiles across the brain. Immunofluorescent methods could be used in convergence with calcium imaging to identify the neurochemical identity of these neurons (Lovett-

Barron et al., 2017). Furthermore, targeted labelling of individual loom habituating neurons with

144

photoactivatable GFP would reveal their morphology and connectivity (Patterson and Lippincott-

Schwartz, 2002; dal Maschio et al., 2017; Helmbrecht et al., 2018; Kramer et al., 2019; Kunst et al.,

2019). Combining these techniques could comprehensively describe the circuits implicated in loom processing and the role of the neurons that form it. This could further reveal how habituation is generated from this system. Experiments implementing substances known to affect habituation in zebrafish like some NMDA blockers (Best et al., 2008; Roberts et al., 2011; Wolman et al., 2011) or the conspecific alarm substance (Lima et al., 2016) could also be used to observe changes in specific brain structures and habituating subtypes. Moreover, more sophisticated tools could disrupt habituation in zebrafish larvae, taking full advantage of this model. One option would be to use optogenetics to silence or activate selective neuronal groups that would interfere with the normal habituation activity (Carter and de Lecea, 2011; Yizhar et al., 2011). Another option is to use a laser to damage axons or neurons in specific areas that could be involved in learning and visual responses

(Roeser and Baier, 2003; Gahtan and Baier, 2004; Aizenberg and Schuman, 2011). By finding the molecular characteristics, specific connections or manipulating the activity of the different habituating neurons, we will be more likely to understand their role in network and how their combined activity produces behavioural outputs. Knowing how the single elements of the circuits contribute to generating habituation will be key to uncover mechanisms of this phenomenon and its complex interactions. Such an approach is needed to investigate a system that includes multiple parts of the brain and neuronal populations, as the results in this thesis suggest.

6.3 Visual processing of threat and learning in the teleost pallium

Among the responses found in multiple parts of the brain after loom presentations, some of the most interesting ones are found in the telencephalon of zebrafish larvae (Figure 6.1). Among those, the responses in the Dm which suggest that the 6 dpf larvae are responding to a threatening visual stimulus with the structure thought to be responsible for aversive and fear-like processing, as it is

145

considered homologous to the amygdala. The relevance of the amygdala as a part of the brain involved in emotional and learning related behaviours (Phelps and LeDoux, 2005; LeDoux, 2007;

Janak and Tye, 2015), makes the Dm one of the appealing structures to further research with the technologies mentioned above.

Previous studies have found that this part of the telencephalon can modulate habituating responses to moving stimuli (Rooney and Laming, 1986; Laming, 1987), that it relevant for social behaviour

(Geng and Peterson, 2019), and is involved with aversive stimulus processing in adult zebrafish (Lau et al., 2011; von Trotha et al., 2014; Ruhl et al., 2017; Lal et al., 2018). However, until now, no descriptions of its involvement and activity to sensory stimuli had been done at the larval stage. This is a great opportunity to investigate the role and development of this amygdala-like structure in ways that were not possible before. Targeted approaches, for example using optogenetics, could further research the role of this structure when processing threatening stimuli. For example, it would be interesting to see if previous stimulation of the Dm would facilitate or increase the responses to a loom. Similarly, optical stimulation of this area could be tested to produce dishabituation. To understand how the Dm could be modulating visual responses, we would also need to understand its connectivity. How does the visual information reaches the Dm is still unclear, as the main ascending pathway through the preglomerular nucleus might not be developed (Lau et al., 2011; Bloch, 2019).

Finally, single-cell targeted labelling with photoactivatable-GFP could be used to find different morphological types of neurons in this area and the target of its projections. This would help reveal its role in the visual processing of threatening stimuli.

146

Figure 6.1. Evoked responses in the telencephalon of larval zebrafish.

Coronal plane equivalent to section 1 of Figure 3.9. Strongly habituating ROIs are shown in green, weakly habituating neurons in red, blue represents multisensory neurons and magenta motor related responses. Most strongly habituating neurons can be found in the Dm, which also has a notable population of multisensory neurons. Most weakly habituating neurons are located in what likely is the Dl.

There is a strong incentive to study the role of this structure and its connectivity, is because it is part of a pathway that processes visual stimuli and that is conserved in mammals (Day-Brown et al., 2010;

Zhao et al., 2014b; Carr, 2015; Wei et al., 2015; Pereira and Moita, 2016; Shang et al., 2018; Zhou et al., 2019). More importantly, recent research has found that this route is present in primates and humans (Tamietto and de Gelder, 2010; Tamietto et al., 2012; Rafal et al., 2015; Elorette et al., 2018).

Interestingly, this pathway seems to be important to process visual threats. Particularly, it has been involved in the faster processing of emotionally related visual stimuli, like human faces or dangerous animals, that don’t need a conscious processing to elicit a response (Morris et al., 1999; Tamietto et

147

al., 2010; Cecere et al., 2013; Silverstein and Ingvar, 2015; McFadyen et al., 2017; Koller et al., 2019;

McFadyen et al., 2019). Although the exact contribution of this pathway in visual processing is still in debate, there is an agreement that it is involved in emotional related processing (Pessoa and

Adolphs, 2010; de Gelder et al., 2011; Pessoa and Adolphs, 2011). This is relevant for clinical research, because as mentioned before, the amygdala’s responses to human faces in autism patients are altered (Kleinhans et al., 2009; Wiggins et al., 2014; Kleinhans et al., 2016; Tam et al., 2017), and recent research suggests that this subcortical pathway could be implicated in these responses

(Hadjikhani et al., 2017). Altogether, these studies make this pathway of special interest for future research, as it could be altered in multiple psychiatric disorders like autism, Parkinson and anxiety, and could be affecting the social and emotional wellbeing of these patients (Soares et al., 2017; Aincy et al., 2018; McFadyen, 2019). Working with zebrafish larvae could be very beneficial to investigate how these diseases particularly affect this subcortical circuit (Geng and Peterson, 2019). As presented in this thesis, using functional analysis of the responses and analysing the circuits from the tectum to the Dm, we could reveal the mechanisms behind these alterations and ways to approach future therapies.

The visual subcortical pathway to the Dm is very promising but is not the only ascending circuit that deserves future attention. The visual projections to the Dl have been described extensively using traditional methods (Northcutt, 2006; Yamamoto and Ito, 2008), and multiple experiments have shown the role of the Dl in similar tasks as the mammalian hippocampus (Rodriguez et al., 2002;

Broglio et al., 2015). In this thesis, promising preliminary results after using Ambon damselfish suggest that this part of the pallium is involved in learning and performing visual discrimination tasks.

Fish that can perform this kind of task would be very useful for understanding how this structure can play a role in learning and perceiving visual cues. Experiments with more detailed analysis using multiple immunofluorescence labels could identify the types of neurons implicated in these processes.

Furthermore, the use of drugs would allow searching for the molecular mechanisms involved in this

148

task. For example, MK-801, a non-competitive NMDA receptor antagonist, has been shown to affect visual discrimination learning and performance in rats and humans (Murray et al., 1995; Meuwese et al., 2013). It would be interesting to test this and other drugs at different stages of the visual learning protocol that I used in chapter 5, to see how they affect visual discrimination and learning in teleost.

Although the studies in this thesis were different in many ways, it is interesting to note that in both species I found neuronal activity in the Dl. This supports the idea that the visual ascending pathway that reaches this area is well conserved among teleost. Also relevant, as mentioned for the Dm, it is striking that the telencephalon of a 6dpf zebrafish larvae is processing visual information. It is unknown at this stage if this activity could be modulating the habituating responses indirectly via Dc projections to the tectum (Luiten, 1981; Giassi et al., 2012b)(Luiten, 1981; Giassi et al, 2012). Those connections haven not been reported in fish of this age yet. Further research on these neurons would be necessary to see which role they play in visual learning. A recent study showed that 3-week old zebrafish are able to do spatial and visual learning in an avoidance task (Yashina et al., 2019). It would be interesting to study whether the Dl is involved in this behaviour, as it has been shown to participate in spatial learning (Rodriguez et al., 2002; Broglio et al., 2015; Uceda et al., 2015).

The visual pathway that reaches the Dl could be relevant for clinical research, as it could be similar to subcortical circuits involved in visual learning where cortical processing is impaired, a phenomenon called blindsight (Hadid and Lepore, 2017). Understanding this pathway could help establish rehabilitation programs in patients with cortical damage of visual areas (Huxlin et al., 2009;

Das et al., 2014). This research is now more appealing as in the last decade multiple studies have shown that the hippocampus is involved in visual processing and visual learning of complex images

(Lee et al., 2012)Kim et al, 2011; (Kim et al., 2011; Liu et al., 2017; Dalton et al., 2018; Kim et al.,

2018). Furthermore, it has been suggested that the role of the hippocampus in visual learning and discrimination could arise from the need to maintain the visual features in the working memory so as to allow discrimination (Knutson et al., 2012; Hales et al., 2015). All these features could be studied

149

in teleost to understand how visual cognitive processes evolved in vertebrates (Yamamoto, 2009;

Oliveira, 2013; Bshary and Brown, 2014; Murray et al., 2018).

6.4 Conclusions

Using two different species of teleost and different approaches, I was able to investigate visual learning and the parts of the brain that are likely to be involved in this behaviour. First, I was able to show that habituation, a universal form of learning, involves the interaction of multiple subpopulations of neurons across the brain of vertebrates. This result, combined with previous habituation research, suggests that this phenomenon is more complex than previously thought. It also opens interesting questions on the specific roles and mechanisms that dictate the interaction between these populations and the brain areas they belong. Among all the structures observed, my results and an analysis of the literature suggest that the information converges in the tectum, which is the likely modulator of the disengagement of other parts of the system as habituation occurs. Furthermore, my results suggest that the pallium of 6dpf zebrafish larvae and Ambon damselfish are involved in visual processing and learning. Particularly, I found that the Dm and Dl of zebrafish larvae respond to visual stimuli and that the Dl of Ambon damselfish is likely involved in visual discrimination learning.

Altogether, the findings presented in this thesis suggest that multiple structures in the midbrain and forebrain process visual information, and are sufficient to produce simple and complex learning behaviours. However, these structures are part of the subcortical circuits in humans, and their role in visual learning has been mostly overlooked. Future research in these structures looks promising and full of interesting questions that could contribute to a more complete understanding of the nervous system of all vertebrates.

150

Bibliography

Abbas F, Triplett MA, Goodhill GJ, Meyer MP (2017) A Three-Layer Network Model of Direction Selective Circuits in the Optic Tectum. Frontiers in neural circuits 11. Accanto N, Chen IW, Ronzitti E, Molinier C, Tourain C, Papagiakoumou E, Emiliani V (2019) Multiplexed temporally focused light shaping through a gradient index lens for precise in- depth optogenetic photostimulation. Scientific Reports 9. Agrillo C, Miletto Petrazzini M, Bisazza A (2014) Numerical acuity of fish is improved in the presence of moving targets, but only in the subitizing range. Animal cognition 17:307-316. Agrillo C, Dadda M, Serena G, Bisazza A (2009) Use of Number by Fish. PloS one 4. Agrillo C, Petrazzini MEM, Piffer L, Dadda M, Bisazza A (2012) A new training procedure for studying discrimination learning in fish. Behavioural brain research 230:343-348. Ahmad F, Richardson MK (2013) Exploratory behaviour in the open field test adapted for larval zebrafish: Impact of environmental complexity. Behav Process 92:88-98. Ahmed O, Seguin D, Gerlai R (2011) An automated predator avoidance task in zebrafish. Behavioural brain research 216:166-171. Ahmed TS, Fernandes Y, Gerlai R (2012) Effects of animated images of sympatric predators and abstract shapes on fear responses in zebrafish. Behaviour 149:1125-1153. Ahrens MB, Orger MB, Robson DN, Li JM, Keller PJ (2013a) Whole-brain functional imaging at cellular resolution using light-sheet microscopy. Nat Methods 10:413-+. Ahrens MB, Huang KH, Narayan S, Mensh BD, Engert F (2013b) Two-photon calcium imaging during fictive navigation in virtual environments. Frontiers in neural circuits 7. Aincy M, Meziane H, Herault Y, Humeau Y (2018) Synaptic dysfunction in amygdala in intellectual disorder models. Prog Neuro-Psychoph 84:392-397. Aizenberg M, Schuman EM (2011) Cerebellar-dependent learning in larval zebrafish. The Journal of neuroscience : the official journal of the Society for Neuroscience 31:8708-8712. Allan BJM, Domenici P, McCormick MI, Watson S-A, Munday PL (2013) Elevated CO₂ Affects Predator-Prey Interactions through Altered Performance. PloS one 8:e58520. Anzelius M, Ekstrom P, Mohler H, Richards JG (1995a) Immunocytochemical Localization of the Gaba(a) Benzodiazepine Receptor Beta(2)/Beta(3) Subunits in the Optic Tectum of the Salmon. J Recept Signal Tr R 15:413-425. Anzelius M, Ekström P, Möhler H, Grayson Richards J (1995b) Immunocytochemical localization of GABAA receptor β2β3-subunits in the brain of Atlantic salmon (Salmo salar L). Journal of chemical neuroanatomy 8:207-221. Ardiel E, Rankin C (2008) Behavioral Plasticity in the C. elegans Mechanosensory Circuit. J Neurogenet 22:239-255. Ardiel EL, Yu AJ, Giles AC, Rankin CH (2017) Habituation as an adaptive shift in response strategy mediated by neuropeptides. NPJ science of learning 2:9. Ardiel EL, Giles AC, Yu AJ, Lindsay TH, Lockery SR, Rankin CH (2016) Dopamine receptor DOP- 4 modulates habituation to repetitive photoactivation of a C. elegans polymodal nociceptor. Learn Memory 23:495-503. Ardiel EL, McDiarmid TA, Timbers TA, Lee KCY, Safaei J, Pelech SL, Rankin CH (2018) Insights into the roles of CMK-1 and OGT-1 in interstimulus interval-dependent habituation in Caenorhabditis elegans. P Roy Soc B-Biol Sci 285. Arias N, Mendez M, Arias JL (2015) The recognition of a novel-object in a novel context leads to hippocampal and parahippocampal c-Fos involvement. Behavioural brain research 292:44- 49. 151

Arrenberg AB, Driever W (2013) Integrating anatomy and function for zebrafish circuit analysis. Frontiers in neural circuits 7. Asadollahi A, Mysore SP, Knudsen EI (2010) Stimulus-driven competition in a cholinergic midbrain nucleus. Nature neuroscience 13:889-U138. Askew HR (1970) Effects of Stimulus Intensity and Intertrial Interval on Habituation of Head-Shake Response in Rat. J Comp Physiol Psych 72:492-&. Avants BB, Epstein CL, Grossman M, Gee JC (2008) Symmetric diffeomorphic image registration with cross-correlation: Evaluating automated labeling of elderly and neurodegenerative brain. Med Image Anal 12:26-41. Avants BB, Tustison NJ, Song G, Cook PA, Klein A, Gee JC (2011) A reproducible evaluation of ANTs similarity metric performance in brain image registration. NeuroImage 54:2033-2044. Avitan L, Pujic Z, Mölter J, Van De Poll M, Sun B, Teng H, Amor R, Scott EK, Goodhill GJ (2017) Spontaneous Activity in the Zebrafish Tectum Reorganizes over Development and Is Influenced by Visual Experience. Current Biology 27:2407-2419.e2404. Baier H, Klostermann S, Trowe T, Karlstrom RO, NussleinVolhard C, Bonhoeffer F (1996) Genetic dissection of the retinotectal projection. Development 123:415-425. Bailey JM, Oliveri AN, Levin ED (2015) Pharmacological analyses of learning and memory in zebrafish (Danio rerio). Pharmacol Biochem Be 139:103-111. Baimbridge KG, Celio MR, Rogers JH (1992) Calcium-binding proteins in the nervous system. Trends Neurosci 15:303-308. Barbosa A, Hoffmann A, Alves FL, Wolkers CPB, Hoshiko FM, Ide LM, Silva LECM, Santos P (2012) Western Blotting Assay for Egr-1 Immediate Early Gene in Brain Tissue of Zebrafish Applied to Neuroethological Study. In: Zebrafish Protocols for Neurobehavioral Research (Kalueff VA, Stewart MA, eds), pp 331-342. Totowa, NJ: Humana Press. Barker AJ, Baier H (2015) Sensorimotor Decision Making in the Zebrafish Tectum. Current Biology 25:2804-2814. Baronio D, Puttonen HAJ, Sundvik M, Semenova S, Lehtonen E, Panula P (2018) Embryonic exposure to valproic acid affects the histaminergic system and the social behaviour of adult zebrafish (Danio rerio). Brit J Pharmacol 175:797-809. Bass SLS, Gerlai R (2008) Zebrafish (Danio rerio) responds differentially to stimulus fish: The effects of sympatric and allopatric predators and harmless fish. Behavioural brain research 186:107- 117. Bassett DS, Wymbs NF, Porter MA, Mucha PJ, Carlson JM, Grafton ST (2011) Dynamic reconfiguration of human brain networks during learning. Proceedings of the National Academy of Sciences of the United States of America 108:7641-7646. Basso MA, May PJ (2017) Circuits for Action and Cognition: A View from the Superior Colliculus. Annu Rev Vis Sci 3:197-226. Becker DV, Anderson US, Mortensen CR, Neufeld SL, Neel R (2011) The face in the crowd effect unconfounded: happy faces, not angry faces, are more efficiently detected in single- and multiple-target visual search tasks. Journal of experimental psychology General 140:637-659. Beilharz FL, Atkins KJ, Duncum AJF, Mundy ME (2016) Altering Visual Perception Abnormalities: A Marker for Body Image Concern. PloS one 11. Bertini C, Grasso PA, Ladavas E (2016) The role of the retino-colliculo-extrastriate pathway in visual awareness and visual field recovery. Neuropsychologia 90:72-79. Best JD, Berghmans S, Hunt JJ, Clarke SC, Fleming A, Goldsmith P, Roach AG (2008) Non- associative learning in larval zebrafish. Neuropsychopharmacol 33:1206-1215. Betzel RF (2018) Organizing principles of whole-brain functional connectivity in zebrafish larvae. bioRxiv:496414. Bhattacharyya K, McLean DL, Maciver MA (2017) Visual Threat Assessment and Reticulospinal Encoding of Calibrated Responses in Larval Zebrafish. Current Biology 27:2751-+. Bilotta J, Saszik S (2001) The zebrafish as a model visual system. Int J Dev Neurosci 19:621-629. 152

Bisazza A, Piffer L, Serena G, Agrillo C (2010) Ontogeny of Numerical Abilities in Fish. PloS one 5. Bishop BH, Spence-Chorman N, Gahtan E (2016) Three-dimensional motion tracking reveals a diving component to visual and auditory escape swims in zebrafish larvae. Journal of Experimental Biology 219:3981-3987. Blaser R, Gerlai R (2006) Behavioral phenotyping in zebrafish: Comparison of three behavioral quantification methods. Behavior Research Methods 38:456-469. Blaser R, Heyser C (2015) Spontaneous object recognition: a promising approach to the comparative study of memory. Frontiers in behavioral neuroscience 9. Blaser RE, Vira DG (2014) Experiments on learning in zebrafish (Danio rerio): A promising model of neurocognitive function. Neuroscience and biobehavioral reviews 42:224-231. Bloch S (2019) Divergent Evolution of Brain Structures and Convergence of Cognitive Functions in Vertebrates : the Example of the Teleost Zebrafish. In: Université Paris-Saclay. Boehnke SE, Berg DJ, Marino RA, Baldi PF, Itti L, Munoz DP (2011) Visual adaptation and novelty responses in the superior colliculus. European Journal of Neuroscience 34:766-779. Bogli SY, Afthinos M, Bertolini G, Straumann D, Huang MYY (2016) Unravelling Stimulus Direction Dependency of Visual Acuity in Larval Zebrafish by Consistent Eye Displacements Upon Optokinetic Stimulation. Invest Ophth Vis Sci 57:1721-1727. Bonaccorsi J, Berardi N, Sale A (2014) Treatment of annblyopia in the adult: insights from a new rodent model of visual perceptual learning. Frontiers in neural circuits 8. Bozorgmehr T, Ardiel EL, McEwan AH, Rankin CH (2013) Mechanisms of plasticity in a Caenorhabditis elegans mechanosensory circuit. Front Physiol 4. Braford MR, Jr. (1995) Comparative aspects of forebrain organization in the ray-finned fishes: touchstones or not? Brain, behavior and evolution 46:259-274. Braford MR, Jr. (2009) Stalking the everted telencephalon: comparisons of forebrain organization in basal ray-finned fishes and teleosts. Brain, behavior and evolution 74:56-76. Braithwaite VA (2011) SENSORY SYSTEMS, PERCEPTION, AND LEARNING | Fish Learning and Memory. In: Encyclopedia of Fish Physiology (Farrell AP, ed), pp 707-712. San Diego: Academic Press. Branson K, Robie AA, Bender J, Perona P, Dickinson MH (2009) High-throughput ethomics in large groups of Drosophila. Nat Methods 6:451-U477. Briffa M, de la Haye K, Munday PL (2012) High CO(2) and marine animal behaviour: potential mechanisms and ecological consequences. Marine pollution bulletin 64:1519-1528. Broglio C, Rodriguez F, Salas C (2003) Spatial cognition and its neural basis in teleost fishes. Fish and Fisheries 4:247-255. Broglio C, Gómez A, Durán E, Salas C, Rodríguez F (2011) Brain and Cognition in Teleost Fish. In: Fish Cognition and Behavior, pp 325-358: Wiley-Blackwell. Broglio C, Gomez A, Duran E, Ocana FM, Jimenez-Moya F, Rodriguez F, Salas C (2005) Hallmarks of a common forebrain vertebrate plan: specialized pallial areas for spatial, temporal and emotional memory in actinopterygian fish. Brain research bulletin 66:277-281. Broglio C, Martín-Monzón I, Ocaña F, Gómez A, Durán E, Salas C, Rodríguez F (2015) Hippocampal pallium and map-like memories through vertebrate evolution. Journal of Behavioral and Brain Science 5:109. Broussard GJ, Liang RQ, Tian L (2014) Monitoring activity in neural circuits with genetically encoded indicators. Front Mol Neurosci 7. Brown C, Laland K, Krause J (2011) Fish Cognition and Behaviour. In: Fish Cognition and Behavior, pp 1-9: Wiley-Blackwell. Bruno JL, Garrett AS, Quintin EM, Mazaika PK, Reiss AL (2014) Aberrant Face and Gaze Habituation in Fragile X Syndrome. Am J Psychiat 171:1099-1106.

153

Bruno JL, Hosseini SMH, Saggar M, Quintin EM, Raman MM, Reiss AL (2017) Altered Brain Network Segregation in Fragile X Syndrome Revealed by Structural Connectomics. Cerebral cortex 27:2249-2259. Bshary R, Brown C (2014) Fish cognition. Current Biology 24:R947-R950. Budick SA, O'Malley DM (2000) Locomotor repertoire of the larval zebrafish: Swimming, turning and prey capture. Journal of Experimental Biology 203:2565-2579. Burgess AE (2011) Visual perception studies and observer models in medical imaging. Seminars in nuclear medicine 41:419-436. Burgess CP, Lak A, Steinmetz NA, Zatka-Haas P, Reddy CB, Jacobs EAK, Linden JF, Paton JJ, Ranson A, Schroder S, Soares S, Wells MJ, Wool LE, Harris KD, Carandini M (2017) High- Yield Methods for Accurate Two-Alternative Visual Psychophysics in Head-Fixed Mice. Cell Reports 20:2513-2524. Burgess HA, Granato M (2007) Sensorimotor gating in larval zebrafish. Journal of Neuroscience 27:4984-4994. Burmeister SS (2007) Genomic responses to behavioral interactions in an African cichlid fish: Mechanisms and evolutionary implications. Brain Behav Evolut 70:247-256. Burmeister SS, Fernald RD (2005) Evolutionary conservation of the Egr-1 immediate-early gene response in a teleost. Journal of Comparative Neurology 481:220-232. Burrill JD, Easter SS (1994) Development of the Retinofugal Projections in the Embryonic and Larval Zebrafish (Brachydanio-Rerio). Journal of Comparative Neurology 346:583-600. Bussey TJ, Saksida LM, Murray EA (2002) Perirhinal cortex resolves feature ambiguity in complex visual discriminations. European Journal of Neuroscience 15:365-374. Butler AB (2011) BRAIN AND NERVOUS SYSTEM | Functional Morphology of the Brains of Ray- Finned Fishes. In: Encyclopedia of Fish Physiology (Farrell AP, ed), pp 37-45. San Diego: Academic Press. Carlile PA, Peters RA, Evans CS (2006) Detection of a looming stimulus by the Jacky dragon: selective sensitivity to characteristics of an aerial predator. Animal Behaviour 72:553-562. Carr JA (2015) I'll take the low road: the evolutionary underpinnings of visually triggered fear. Frontiers in Neuroscience 9. Carter ME, de Lecea L (2011) Optogenetic investigation of neural circuits in vivo. Trends in molecular medicine 17:197-206. Castellucci V, Pinsker H, Kupfermann I, Kandel ER (1970) Neuronal Mechanisms of Habituation and Dishabituation of Gill-Withdrawal Reflex in Aplysia. Science 167:1745-+. Castellucci VF, Kandel ER (1974) Quantal Analysis of Synaptic Depression Underlying Habituation of Gill-Withdrawal Reflex in Aplysia. Proceedings of the National Academy of Sciences of the United States of America 71:5004-5008. Castren M, Paakkonen A, Tarkka IM, Ryynanen M, Partanen J (2003) Augmentation of auditory N1 in children with fragile X syndrome. Brain Topogr 15:165-171. Castro A, Becerra M, Manso MJ, Anadon R (2006) Calretinin immunoreactivity in the brain of the zebrafish, Danio rerio: Distribution and comparison with some neuropeptides and neurotransmitter-synthesizing enzymes. I. Olfactory organ and forebrain. Journal of Comparative Neurology 494:435-459. Cecere R, Bertini C, Ladavas E (2013) Differential Contribution of Cortical and Subcortical Visual Pathways to the Implicit Processing of Emotional Faces: A tDCS Study. Journal of Neuroscience 33:6469-6475. Champ C, Wallis G, Vorobyev M, Siebeck U, Marshall J (2014) Visual acuity in a species of coral reef fish: Rhinecanthus aculeatus. Brain, behavior and evolution 83:31-42. Chen JF, Lei L, Tian LJ, Hou F, Roper C, Ge XQ, Zhao YX, Chen YH, Dong QX, Tanguay RL, Huang CJ (2018a) Developmental and behavioral alterations in zebrafish embryonically exposed to valproic acid (VPA): An aquatic model for autism. Neurotoxicol Teratol 66:8-16.

154

Chen TW, Wardill TJ, Sun Y, Pulver SR, Renninger SL, Baohan A, Schreiter ER, Kerr RA, Orger MB, Jayaraman V, Looger LL, Svoboda K, Kim DS (2013) Ultrasensitive fluorescent proteins for imaging neuronal activity. Nature 499:295-300. Chen XY, Mu Y, Hu Y, Kuan AT, Nikitchenko M, Randlett O, Chen AB, Gavornik JP, Sompolinsky H, Engert F, Ahrens MB (2018b) Brain-wide Organization of Neuronal Activity and Convergent Sensorimotor Transformations in Larval Zebrafish. Neuron 100:876-+. Chivers DP, McCormick MI, Nilsson GE, Munday PL, Watson SA, Meekan MG, Mitchell MD, Corkill KC, Ferrari MC (2014) Impaired learning of predators and lower prey survival under elevated CO2 : a consequence of neurotransmitter interference. Global change biology 20:515-522. Christoffersen GRJ (1997) Habituation: Events in the history of its characterization and linkage to synaptic depression. A new proposed kinetic criterion for its identification. Prog Neurobiol 53:45-66. Chung WS, Marshall NJ, Watson SA, Munday PL, Nilsson GE (2014) Ocean acidification slows retinal function in a damselfish through interference with GABAA receptors. The Journal of experimental biology 217:323-326. Clark G, Sperry M (1945) A Simplified Nissl Stain with Thionin. Stain Technol 20:23-24. Clark TD, Raby GD, Roche DG, Binning SA, Speers-Roesch B, Jutfelt F, Sundin J (2020) Ocean acidification does not impair the behaviour of coral reef fishes. Nature 577:370-375. Clements JC, Hunt HL (2015) Marine animal behaviour in a high CO2 ocean. Marine Ecology Progress Series 536:259-279. Colwill RM, Creton R (2011) Imaging escape and avoidance behavior in zebrafish larvae. Rev Neuroscience 22:63-73. Colwill RM, Raymond MP, Ferreira L, Escudero H (2005) Visual discrimination learning in zebrafish (Danio rerio). Behav Process 70:19-31. Cong L, Wang ZG, Chai YM, Hang W, Shang CF, Yang WB, Bai L, Du JL, Wang K, Wen Q (2017) Rapid whole brain imaging of neural activity in freely behaving larval zebrafish (Danio rerio). Elife 6. Contractor A, Klyachko VA, Portera-Cailliau C (2015) Altered Neuronal and Circuit Excitability in Fragile X Syndrome. Neuron 87:699-715. Cooke SF, Bear MF (2014) How the mechanisms of long-term synaptic potentiation and depression serve experience-dependent plasticity in primary visual cortex. Philosophical transactions of the Royal Society of London Series B, Biological sciences 369:20130284. Cooper WE, Blumstein DT (2015) Escaping from predators : an integrative view of escape decisions. Couturier CS, Stecyk JAW, Rummer JL, Munday PL, Nilsson GE (2013) Species-specific effects of near-future CO2 on the respiratory performance of two tropical prey fish and their predator. Comp Biochem Phys A 166:482-489. Cowden J, Padnos B, Hunter D, MacPhail R, Jensen K, Padilla S (2012) Developmental exposure to valproate and ethanol alters locomotor activity and retino-tectal projection area in zebrafish embryos. Reprod Toxicol 33:165-173. Crespo C, Porteros A, Arevalo R, Brinon JG, Aijon J, Alonso JR (1999) Distribution of parvalbumin immunoreactivity in the brain of the tench (Tinca tinca L., 1758). The Journal of comparative neurology 413:549-571. Cripps IL, Munday PL, McCormick MI (2011) Ocean acidification affects prey detection by a predatory reef fish. PloS one 6:e22736. Csanyi V, Doka A (1993) Learning Interactions between Prey and Predator Fish. Mar Behav Physiol 23:63-78. Curran T, Morgan JI (1987) Memories of Fos. Bioessays 7:255-258. d'Amora M, Giordani S (2018) The Utility of Zebrafish as a Model for Screening Developmental Neurotoxicity. Frontiers in Neuroscience 12.

155

dal Maschio M, Donovan JC, Helmbrecht TO, Baier H (2017) Linking Neurons to Network Function and Behavior by Two-Photon Holographic Optogenetics and Volumetric Imaging. Neuron 94:774-+. Dalton MA, Zeidman P, McCormick C, Maguire EA (2018) Differentiable Processing of Objects, Associations, and Scenes within the Hippocampus. Journal of Neuroscience 38:8146-8159. Das A, Tadin D, Huxlin KR (2014) Beyond blindsight: properties of visual relearning in cortically blind fields. The Journal of neuroscience : the official journal of the Society for Neuroscience 34:11652-11664. Davison JM, Akitake CM, Goll MG, Rhee JM, Gosse N, Baier H, Halpern ME, Leach SD, Parsons MJ (2007) Transactivation from Gal4-VP16 transgenic insertions for tissue-specific cell labeling and ablation in zebrafish. Developmental biology 304:811-824. Day-Brown JD, Wei HY, Chomsung RD, Petry HM, Bickford ME (2010) Pulvinar projections to the striatum and amygdala in the tree shrew. Frontiers in neuroanatomy 4. De Franceschi G, Vivattanasam T, Saleem AB, Solomon SG (2016) Vision Guides Selection of Freeze or Flight Defense Strategies in Mice. Current Biology 26:2150-2154. de Gelder B, van Honk J, Tamietto M (2011) Emotion in the brain: of low roads, high roads and roads less travelled. Nat Rev Neurosci 12. de Vries SEJ, Clandinin TR (2012) Loom-Sensitive Neurons Link Computation to Action in the Drosophila Visual System. Current Biology 22:353-362. Del Bene F, Wyart C, Robles E, Tran A, Looger L, Scott EK, Isacoff EY, Baier H (2010) Filtering of Visual Information in the Tectum by an Identified Neural Circuit. Science 330:669-673. Demski L (2003) In a Fish’s Mind’s Eye: The Visual Pallium of Teleosts. In: Sensory Processing in Aquatic Environments (Collin S, Marshall NJ, eds), pp 404-419: Springer New York. Demski LS (2013) The pallium and mind/behavior relationships in teleost fishes. Brain, behavior and evolution 82:31-44. den Broeder MJ, van der Linde H, Brouwer JR, Oostra BA, Willemsen R, Ketting RF (2009) Generation and Characterization of Fmr1 Knockout Zebrafish. PloS one 4. Deshpande KV, Jadhao AG (2015) Calcium binding protein calretinin (29 kD) localization in the forebrain of the cichlid fish: An immunohistochemical study. Gen Comp Endocr 220:93-97. Devine BM, Munday PL, Jones GP (2012a) Homing ability of adult cardinalfish is affected by elevated carbon dioxide. Oecologia 168:269-276. Devine BM, Munday PL, Jones GP (2012b) Rising CO2 concentrations affect settlement behaviour of larval damselfishes. Coral Reefs 31:229-238. Dewan AK, Tricas TC (2014) Cytoarchitecture of the telencephalon in the coral reef multiband Butterflyfish ( Chaetodon multicinctus : Perciformes). Brain, behavior and evolution 84:31- 50. Diaz-Regueira S, Anadon R (2000) Calretinin expression in specific neuronal systems in the brain of an advanced teleost, the grey mullet (Chelon labrosus). The Journal of comparative neurology 426:81-105. Dill LM (1974) The escape response of the zebra danio (Brachydanio rerio) I. The stimulus for escape. Animal Behaviour 22:711-722. Dirian L, Galant S, Coolen M, Chen WB, Bedu S, Houart C, Bally-Cuif L, Foucher I (2014) Spatial Regionalization and Heterochrony in the Formation of Adult Pallial Neural Stem Cells. Dev Cell 30:123-136. Dixson DL, Jennings AR, Atema J, Munday PL (2015) Odor tracking in sharks is reduced under future ocean acidification conditions. Global change biology 21:1454-1462. do Carmo Silva RX, Lima-Maximino MG, Maximino C (2018) The aversive brain system of teleosts: Implications for neuroscience and biological psychiatry. Neuroscience & Biobehavioral Reviews 95:123-135. Dosher B, Lu ZL (2017) Visual Perceptual Learning and Models. Annu Rev Vis Sci 3:343-363.

156

Douglas R, Djamgoz MBA (1990) The Visual system of fish, 1st Edition. London ; New York: Chapman and Hall. Douglas RH, Hawryshyn CW (1990) Behavioural studies of fish vision: an analysis of visual capabilities. In: The visual system of fish, pp 373-418: Springer. Douglas RH, Eva J, Guttridge N (1988) Size Constancy in Goldfish (Carassius-Auratus). Behavioural brain research 30:37-42. Downes GB, Granato M (2006) Supraspinal input is dispensable to generate glycine-mediated locomotive behaviors in the zebrafish embryo. J Neurobiol 66:437-451. Dragoi V (2002) A feedforward model of suppressive and facilitatory habituation effects. Biol Cybern 86:419-426. Dreosti E, Llopis NV, Carl M, Yaksi E, Wilson SW (2014) Left-Right Asymmetry Is Required for the Habenulae to Respond to Both Visual and Olfactory Stimuli. Current Biology 24:440-445. Dudkin E, Gruberg E (2009) Evolution of Nucleus Isthmi. In: Encyclopedia of Neuroscience (Binder MD, Hirokawa N, Windhorst U, eds), pp 1258-1262. Berlin, Heidelberg: Springer Berlin Heidelberg. Dunn TW, Gebhardt C, Naumann EA, Riegler C, Ahrens MB, Engert F, Del Bene F (2016a) Neural Circuits Underlying Visually Evoked Escapes in Larval Zebrafish. Neuron 89:613-628. Dunn TW, Mu Y, Narayan S, Randlett O, Naumann EA, Yang CT, Schier AF, Freeman J, Engert F, Ahrens MB (2016b) Brain-wide mapping of neural activity controlling zebrafish exploratory locomotion. Elife 5. Duteil M, Pope EC, Perez-Escudero A, de Polavieja GG, Furtbauer I, Brown MR, King AJ (2016) European sea bass show behavioural resilience to near-future ocean acidification. Royal Society open science 3. Dutta A, Gutfreund Y (2014) Saliency mapping in the optic tectum and its relationship to habituation. Front Integr Neurosc 8. Dwivedi S, Medishetti R, Rani R, Sevilimedu A, Kulkarni P, Yogeeswari P (2019) Larval zebrafish model for studying the effects of valproic acid on neurodevelopment: An approach towards modeling autism. Journal of Pharmacological and Toxicological Methods 95:56-65. Eaton RC, Farley RD (1973) Development of Mauthner Neurons in Embryos and Larvae of Zebra Fish, Brachydanio-Rerio. Copeia:673-682. Eaton RC, Bombardieri RA, Meyer DL (1977) The Mauthner-initiated startle response in teleost fish. Journal of Experimental Biology 66. Eaton RC, Lavender WA, Wieland CM (1982) Alternative Neural Pathways Initiate Fast-Start Responses Following Lesions of the Mauthner Neuron in Goldfish. Journal of Comparative Physiology 145:485-496. Eaton RC, Didomenico R, Nissanov J (1991) Role of the Mauthner Cell in Sensorimotor Integration by the Brain-Stem Escape Network. Brain Behav Evolut 37:272-285. Eaton RC, Lee RKK, Foreman MB (2001) The Mauthner cell and other identified neurons of the brainstem escape network of fish. Prog Neurobiol 63:467-485. Ebbesson SO, Vanegas H (1976) Projections of the optic tectum in two teleost species. The Journal of comparative neurology 165:161-180. Echteler SM, Saidel WM (1981) Forebrain Connections in the Goldfish Support Telencephalic Homologies with Land Vertebrates. Science 212:683-685. Elorette C, Forcelli PA, Saunders RC, Malkova L (2018) Colocalization of Tectal Inputs With Amygdala-Projecting Neurons in the Macaque Pulvinar. Frontiers in neural circuits 12. Elphick MR (1997) Localization of nitric oxide synthase using NADPH diaphorase histochemistry. Methods in molecular biology 72:153-158. Engel JE, Wu CF (1998) Genetic dissection of functional contributions of specific potassium channel subunits in habituation of an escape circuit in Drosophila. Journal of Neuroscience 18:2254- 2267.

157

Ethridge LE, White SP, Mosconi MW, Wang J, Byerly MJ, Sweeney JA (2016) Reduced habituation of auditory evoked potentials indicate cortical hyper-excitability in Fragile X Syndrome. Transl Psychiat 6. Ewert JP (1970) Neural Mechanisms of Prey-Catching and Avoidance Behavior in Toad (Bufo-Bufo L). Brain Behav Evolut 3:36-+. Ezzeddine Y, Glanzman DL (2003) Prolonged habituation of the gill-withdrawal reflex in Aplysia depends on protein synthesis, protein phosphatase activity, and postsynaptic glutamate receptors. Journal of Neuroscience 23:9585-9594. Falk CX, Wu JY, Cohen LB, Tang AC (1993) Nonuniform expression of habituation in the activity of distinct classes of neurons in the Aplysia abdominal ganglion. The Journal of neuroscience : the official journal of the Society for Neuroscience 13:4072-4081. Fang MR, Li JC, Wai SM, Kwong WH, Kung LS, Yew DT (2005) Retinal twin cones or retinal double cones in fish: Misnomer or different morphological forms? Int J Neurosci 115:981- 987. Farahani FV, Karwowski W, Lighthall NR (2019) Application of Graph Theory for Identifying Connectivity Patterns in Human Brain Networks: A Systematic Review. Front Neurosci 13:585. Faria M, Prats E, Novoa-Luna KA, Bedrossiantz J, Gomez-Canela C, Gomez-Olivan LM, Raldua D (2019) Development of a vibrational startle response assay for screening environmental pollutants and drugs impairing predator avoidance. Sci Total Environ 650:87-96. Favre-Bulle IA, Stilgoe AB, Rubinsztein-Dunlop H, Scott EK (2017) Optical trapping of otoliths drives vestibular behaviours in larval zebrafish. Nat Commun 8. Favre-Bulle IA, Vanwalleghem G, Taylor MA, Rubinsztein-Dunlop H, Scott EK (2018) Cellular- Resolution Imaging of Vestibular Processing across the Larval Zebrafish Brain. Current Biology 28:3711-+. Favre-Bulle IA, Preece D, Nieminen TA, Heap LA, Scott EK, Rubinsztein-Dunlop H (2015) Scattering of Sculpted Light in Intact Brain Tissue, with implications for Optogenetics. Scientific Reports 5. Feierstein CE, Portugues R, Orger MB (2015) Seeing the Whole Picture: A Comprehensive Imaging Approach to Functional Mapping of Circuits in Behaving Zebrafish. Neuroscience 296:26- 38. Feldman Barrett L, Finlay BL (2018) Concepts, goals and the control of survival-related behaviors. Current Opinion in Behavioral Sciences 24:172-179. Fernandes AM, Larsch J, Donovan JC, Helmbrecht TO, Mearns D, Kölsch Y, Dal Maschio M, Baier H (2019) Neuronal circuitry for stimulus selection in the visual system. bioRxiv:598383. Ferrari MC, Manassa RP, Dixson DL, Munday PL, McCormick MI, Meekan MG, Sih A, Chivers DP (2012a) Effects of ocean acidification on learning in coral reef fishes. PloS one 7:e31478. Ferrari MC, Munday PL, Rummer JL, McCormick MI, Corkill K, Watson SA, Allan BJ, Meekan MG, Chivers DP (2015) Interactive effects of ocean acidification and rising sea temperatures alter predation rate and predator selectivity in reef fish communities. Global change biology 21:1848-1855. Ferrari MCO, McCormick MI, Munday PL, Meekan MG, Dixson DL, Lonnstedt Ö, Chivers DP (2011a) Putting prey and predator into the CO2 equation – qualitative and quantitative effects of ocean acidification on predator–prey interactions. Ecology letters 14:1143-1148. Ferrari MCO, Dixson DL, Munday PL, McCormick MI, Meekan MG, Sih A, Chivers DP (2011b) Intrageneric variation in antipredator responses of coral reef fishes affected by ocean acidification: implications for climate change projections on marine communities. Global change biology 17:2980-2986. Ferrari MCO, McCormick MI, Munday PL, Meekan MG, Dixson DL, Lönnstedt O, Chivers DP (2012b) Effects of ocean acidification on visual risk assessment in coral reef fishes. Functional Ecology 26:553-558. 158

Fetcho JR (1991) Spinal Network of the Mauthner Cell. Brain Behav Evolut 37:298-316. Fetcho JR, Faber DS (1988) Identification of Motoneurons and Interneurons in the Spinal Network for Escapes Initiated by the Mauthner Cell in Goldfish. Journal of Neuroscience 8:4192-4213. Filosa A, Barker AJ, Dal Maschio M, Baier H (2016) Feeding State Modulates Behavioral Choice and Processing of Prey Stimuli in the Zebrafish Tectum. Neuron 90:596-608. Fink AJP, Axel R, Schoonover CE (2019) A virtual burrow assay for head-fixed mice measures habituation, discrimination, exploration and avoidance without training. Elife 8. Fischer TM, Jacobson DA, Counsell AN, Pelot MA, Demorest K (2011) Regulation of low-threshold afferent activity may contribute to short-term habituation in Aplysia californica. Neurobiology of learning and memory 95:248-259. Folgueira M, Anadon R, Yanez J (2004) Experimental study of the connections of the telencephalon in the rainbow trout (Oncorhynchus mykiss). II: Dorsal area and preoptic region. Journal of Comparative Neurology 480:204-233. Forsgren E, Dupont S, Jutfelt F, Amundsen T (2013) Elevated CO2 affects embryonic development and larval phototaxis in a temperate marine fish. Ecology and evolution 3:3637-3646. Forster D, Dal Maschio M, Laurell E, Baier H (2017a) An optogenetic toolbox for unbiased discovery of functionally connected cells in neural circuits. Nat Commun 8. Forster D, Kramer A, Baier H, Kubo F (2018) Optogenetic precision toolkit to reveal form, function and connectivity of single neurons. Methods 150:42-48. Forster D, Arnold-Ammer I, Laurell E, Barker AJ, Fernandes AM, Finger-Baier K, Filosa A, Helmbrecht TO, Kolsch Y, Kuhn E, Robles E, Slanchev K, Thiele TR, Baier H, Kubo F (2017b) Genetic targeting and anatomical registration of neuronal populations in the zebrafish brain with a new set of BAC transgenic tools. Scientific Reports 7. Fotowat H, Harrison RR, Gabbiani F (2011) Multiplexing of Motor Information in the Discharge of a Collision Detecting Neuron during Escape Behaviors. Neuron 69:147-158. Frangou P, Emir UE, Karlaftis VM, Nettekoven C, Hinson EL, Larcombe S, Bridge H, Stagg CJ, Kourtzi Z (2019) Learning to optimize perceptual decisions through suppressive interactions in the human brain. Nat Commun 10:474. Franklin A, Sowden P, Burley R, Notman L, Alder E (2008) Color perception in children with autism. Journal of autism and developmental disorders 38:1837-1847. Freedman DJ, Miller EK (2008) Neural mechanisms of visual categorization: insights from neurophysiology. Neuroscience and biobehavioral reviews 32:311-329. Freedman DJ, Riesenhuber M, Poggio T, Miller EK (2003) A comparison of primate prefrontal and inferior temporal cortices during visual categorization. Journal of Neuroscience 23:5235- 5246. Freeman JB, Rule NO, Adams RB, Jr., Ambady N (2010) The neural basis of categorical face perception: graded representations of face gender in fusiform and orbitofrontal cortices. Cerebral cortex 20:1314-1322. Gabbiani F, Laurent G, Hatsopoulos N, Krapp HG (1999) The many ways of building collision- sensitive neurons. Trends in Neurosciences 22:437-438. Gabriel JP, Trivedi CA, Maurer CM, Ryu S, Bollmann JH (2012) Layer-Specific Targeting of Direction-Selective Neurons in the Zebrafish Optic Tectum. Neuron 76:1147-1160. Gahtan E, Baier H (2004) Of lasers, mutants, and see-through brains: functional neuroanatomy in zebrafish. J Neurobiol 59:147-161. Gahtan E, Sankrithi N, Campos JB, O'Malley DM (2002) Evidence for a widespread brain stem escape network in larval zebrafish. Journal of neurophysiology 87:608-614. Gallagher SP, Northmore DPM (2006) Responses of the teleostean nucleus isthmi to looming objects and other moving stimuli. Visual neuroscience 23:209-219. Garcia JO, Ashourvan A, Muldoon SF, Vettel JM, Bassett DS (2018) Applications of community detection techniques to brain graphs: Algorithmic considerations and implications for neural function. Proc IEEE Inst Electr Electron Eng 106:846-867. 159

Garrido MI (2012) Brain Connectivity: The Feel of Blindsight. Current Biology 22:R599-R600. Geng Y, Peterson RT (2019) The zebrafish subcortical social brain as a model for studying social behavior disorders. Disease models & mechanisms 12. Gerlai R (2010) Zebrafish antipredatory responses: A future for translational research? Behavioural brain research 207:223-231. Gerlai R (2014) Fish in behavior research: Unique tools with a great promise! Journal of neuroscience methods 234:54-58. Gerlai R (2016) Learning and memory in zebrafish (Danio rerio). Zebrafish: Cellular and Developmental Biology, Pt B: Developmental Biology 134:551-586. Gerlai R, Fernandes Y, Pereira T (2009) Zebrafish (Danio rerio) responds to the animated image of a predator: Towards the development of an automated aversive task. Behavioural brain research 201:318-324. Giassi AC, Ellis W, Maler L (2012a) Organization of the gymnotiform fish pallium in relation to learning and memory: III. Intrinsic connections. The Journal of comparative neurology 520:3369-3394. Giassi AC, Duarte TT, Ellis W, Maler L (2012b) Organization of the gymnotiform fish pallium in relation to learning and memory: II. Extrinsic connections. The Journal of comparative neurology 520:3338-3368. Gierszewski S, Bleckmann H, Schluessel V (2013) Cognitive abilities in Malawi cichlids (Pseudotropheus sp.): matching-to-sample and image/mirror-image discriminations. PloS one 8:e57363. Giles AC, Rankin CH (2009) Behavioral and genetic characterization of habituation using Caenorhabditis elegans. Neurobiology of learning and memory 92:139-146. Giovannucci A, Friedrich J, Gunn P, Kalfon J, Brown BL, Koay SA, Taxidis J, Najafi F, Gauthier JL, Zhou PC, Khakh BS, Tank DW, Chklovskii DB, Pnevmatikakis EA (2019) CaImAn an open source tool for scalable calcium imaging data analysis. Elife 8. Giraldez-Perez RM, Gaytan SP, Pasaro R (2013) Cholinergic and nitrergic neuronal networks in the goldfish telencephalon. Acta neurobiologiae experimentalis 73:338-353. Giraldez-Perez RM, Gaytan SP, Ruano D, Torres B, Pasaro R (2008) Distribution of NADPH- diaphorase and nitric oxide synthase reactivity in the central nervous system of the goldfish (Carassius auratus). Journal of chemical neuroanatomy 35:12-32. Glanzman DL (2009) Habituation in Aplysia: The Cheshire Cat of neurobiology. Neurobiology of learning and memory 92:147-154. Glanzman DL (2010) Common Mechanisms of Synaptic Plasticity Minireview in Vertebrates and Invertebrates. Current Biology 20:R31-R36. Goddard CA, Mysore SP, Bryant AS, Huguenard JR, Knudsen EI (2014) Spatially Reciprocal Inhibition of Inhibition within a Stimulus Selection Network in the Avian Midbrain. PloS one 9. Gomez A, Duran E, Salas C, Rodriguez F (2010) Cerebellum Lesion Impairs Eyeblink-Like Classical Conditioning in Goldfish. Neuroscience 166:49-60. Gomez Y, Vargas JP, Portavella M, Lopez JC (2006) Spatial learning and goldfish telencephalon NMDA receptors. Neurobiology of learning and memory 85:252-262. Gori S, Facoetti A (2014) Perceptual learning as a possible new approach for remediation and prevention of developmental dyslexia. Vision research 99:78-87. Gori S, Agrillo C, Dadda M, Bisazza A (2014) Do Fish Perceive Illusory Motion? Scientific Reports 4. Gover TD, Abrams TW (2009) Insights into a molecular switch that gates sensory neuron synapses during habituation in Aplysia. Neurobiology of learning and memory 92:155-165. Gover TD, Jiang XY, Abrams TW (2002) Persistent, exocytosis-independent silencing of release sites underlies homosynaptic depression at sensory synapses in aplysia (vol 22, pg 1942, 2002). Journal of Neuroscience 22:3302-3302. 160

Grady CL, McIntosh AR, Rajah MN, Craik FI (1998) Neural correlates of the episodic encoding of pictures and words. Proceedings of the National Academy of Sciences of the United States of America 95:2703-2708. Graham BJ, Northmore DPM (2007) A spiking neural network model of midbrain visuomotor mechanisms that avoids objects by estimating size and distance monocularly. Neurocomputing 70:1983-1987. Grana P, Huesa G, Anadon R, Yanez J (2012) Immunohistochemical study of the distribution of calcium binding proteins in the brain of a chondrostean (Acipenser baeri). The Journal of comparative neurology 520:2086-2122. Grant SGN (2016) The molecular evolution of the vertebrate behavioural repertoire. Philos T R Soc B 371. Gray JR (2005) Habituated visual neurons in locusts remain sensitive to novel looming objects. Journal of Experimental Biology 208:2515-2532. Green NF, Urquhart HH, van den Berg CP, Marshall NJ, Cheney KL (2018) Pattern edges improve predator learning of aposematic signals. Behavioral Ecology 29:1481-1486. Green SA, Hernandez L, Tottenham N, Krasileva K, Bookheimer SY, Dapretto M (2015) Neurobiology of Sensory Overresponsivity in Youth With Autism Spectrum Disorders. Jama Psychiat 72:778-786. Gruberg E, Dudkin E, Wang Y, Marin G, Salas C, Sentis E, Letelier J, Mpodozis J, Malpeli J, Cui H, Ma R, Northmore D, Udin S (2006) Influencing and interpreting visual input: The role of a visual feedback system. Journal of Neuroscience 26:10368-10371. Guo S (2004) Linking genes to brain, behavior and neurological diseases: what can we learn from zebrafish? Genes Brain Behav 3:63-74. Guthrie DM, Banks JR (1978) Receptive-Field Structure of Visual Cells from Optic Tectum of Freshwater Perch (Perca-Fluviatilis). Brain research 141:211-225. Hadid V, Lepore F (2017) From Cortical Blindness to Conscious Visual Perception: Theories on Neuronal Networks and Visual Training Strategies. Front Syst Neurosci 11. Hadjikhani N, Johnels JA, Zurcher NR, Lassalle A, Guillon Q, Hippolyte L, Billstedt E, Ward N, Lemonnier E, Gillberg C (2017) Look me in the eyes: constraining gaze in the eye-region provokes abnormally high subcortical activation in autism. Scientific Reports 7. Haesemeyer M, Robson DN, Li JM, Schier AF, Engert F (2018) A Brain-wide Circuit Model of Heat- Evoked Swimming Behavior in Larval Zebrafish. Neuron 98:817-+. Hagerman RJ, Berry-Kravis E, Hazlett HC, Bailey DB, Moine H, Kooy RF, Tassone F, Gantois I, Sonenberg N, Mandel JL, Hagerman PJ (2017) Fragile X syndrome. Nat Rev Dis Primers 3. Hagio H, Sato M, Yamamoto N (2018) An ascending visual pathway to the dorsal telencephalon through the optic tectum and nucleus prethalamicus in the yellowfin goby Acanthogobius flavimanus (Temminck & Schlegel, 1845). The Journal of comparative neurology 526:1733- 1746. Hale ME, Katz HR, Peek MY, Fremont RT (2016) Neural circuits that drive startle behavior, with a focus on the Mauthner cells and spiral fiber neurons of fishes. J Neurogenet 30:89-100. Hales JB, Broadbent NJ, Velu PD, Squire LR, Clark RE (2015) Hippocampus, perirhinal cortex, and complex visual discriminations in rats and humans. Learn Memory 22:83-91. Hall D, Suboski MD (1995) Visual and Olfactory Stimuli in Learned Release of Alarm Reactions by Zebra Danio Fish (Brachydanio-Rerio). Neurobiology of learning and memory 63:229-240. Hallquist MN, Hillary FG (2019) Graph theory approaches to functional network organization in brain disorders: A critique for a brave new small-world. Netw Neurosci 3:1-26. Harris JD (1943) Habituatory Response Decrement in the Intact Organism. Psychol Bull 40:385-422. Harvey-Girard E, Giassi AC, Ellis W, Maler L (2012) Organization of the gymnotiform fish pallium in relation to learning and memory: IV. Expression of conserved transcription factors and implications for the evolution of dorsal telencephalon. The Journal of comparative neurology 520:3395-3413. 161

Harvey-Girard E, Giassi ACC, Ellis W, Maler L (2013) Expression of the cannabinoid CB1 receptor in the gymnotiform fish brain and its implications for the organization of the teleost pallium. Journal of Comparative Neurology 521:949-975. Harvey-Girard E, Tweedle J, Ironstone J, Cuddy M, Ellis W, Maler L (2010) Long-term recognition memory of individual conspecifics is associated with telencephalic expression of Egr-1 in the electric fish Apteronotus leptorhynchus. The Journal of comparative neurology 518:2666- 2692. Hatsopoulos N, Gabbiani F, Laurent G (1995) Elementary computation of object approach by wide- field visual neuron. Science (New York, NY) 270:1000-1003. Haug MF, Biehlmaier O, Mueller KP, Neuhauss SCF (2010) Visual acuity in larval zebrafish: behavior and histology. Front Zool 7. Hayes WN, Saiff EI (1967) Visual Alarm Reactions in Turtles. Animal Behaviour 15:102-&. Hazlett HC, Poe MD, Lightbody AA, Styner M, MacFall JR, Reiss AL, Piven J (2012) Trajectories of Early Brain Volume Development in Fragile X Syndrome and Autism. J Am Acad Child Psy 51:921-933. Hazlett HC, Poe MD, Lightbody AA, Gerig G, MacFall JR, Ross AK, Provenzale J, Martin A, Reiss AL, Piven J (2009) Teasing apart the heterogeneity of autism: Same behavior, different brains in toddlers with fragile X syndrome and autism. J Neurodev Disord 1:81-90. Heap LA, Vanwalleghem GC, Thompson AW, Favre-Bulle I, Rubinsztein-Dunlop H, Scott EK (2018a) Hypothalamic Projections to the Optic Tectum in Larval Zebrafish. Frontiers in neuroanatomy 11. Heap LAL, Vanwalleghem G, Thompson AW, Favre-Bulle IA, Scott EK (2018b) Luminance Changes Drive Directional Startle through a Thalamic Pathway. Neuron 99:293-+. Hein AM, Gil MA, Twomey CR, Couzin ID, Levin SA (2018) Conserved behavioral circuits govern high-speed decision-making in wild fish shoals. Proceedings of the National Academy of Sciences of the United States of America 115:12224-12228. Heinrich DDU, Watson SA, Rummer JL, Brandl SJ, Simpfendorfer CA, Heupel MR, Munday PL (2016) Foraging behaviour of the epaulette shark Hemiscyllium ocellatum is not affected by elevated CO2. Ices J Mar Sci 73:633-640. Helmbrecht TO, dal Maschio M, Donovan JC, Koutsouli S, Baier H (2018) Topography of a Visuomotor Transformation. Neuron 100:1429-+. Henriques PM, Rahman N, Jackson SE, Bianco IH (2019) Nucleus Isthmi Is Required to Sustain Target Pursuit during Visually Guided Prey-Catching. Current Biology 29:1771-+. Heuer RM, Grosell M (2014) Physiological impacts of elevated carbon dioxide and ocean acidification on fish. American journal of physiology Regulatory, integrative and comparative physiology 307:R1061-1084. Heuer RM, Welch MJ, Rummer JL, Munday PL, Grosell M (2016) Altered brain ion gradients following compensation for elevated CO2 are linked to behavioural alterations in a coral reef fish. Scientific Reports 6. Hinz FI, Aizenberg M, Tushev G, Schuman EM (2013) Protein synthesis-dependent associative long- term memory in larval zebrafish. The Journal of neuroscience : the official journal of the Society for Neuroscience 33:15382-15387. Hoeft F, Walter E, Lightbody AA, Hazlett HC, Chang C, Piven J, Reiss AL (2011) Neuroanatomical Differences in Toddler Boys With Fragile X Syndrome and Idiopathic Autism. Arch Gen Psychiat 68:295-305. Hoffman KL, Logothetis NK (2009) Cortical mechanisms of sensory learning and object recognition. Philos T R Soc B 364:321-329. Homer M, Rutherford MD (2008) Individuals with autism can categorize facial expressions. Child Neuropsychol 14:419-437. Horn G, Rowell CHF (1968) Medium and Long-Term Changes in Behaviour of Visual Neurones in Tritocerebrum of Locusts. Journal of Experimental Biology 49:143-&. 162

Horner AE, Heath CJ, Hvoslef-Eide M, Kent BA, Kim CH, Nilsson SRO, Alsio J, Oomen CA, Holmes A, Saksida LM, Bussey TJ (2013) The touchscreen operant platform for testing learning and memory in rats and mice. Nat Protoc 8:1961-1984. Hu Y, Chen ZM, Huang L, Xi Y, Li BX, Wang H, Yan JJ, Lee TMC, Tao Q, So KF, Ren CR (2017) A translational study on looming-evoked defensive response and the underlying subcortical pathway in autism. Scientific Reports 7. Hudac CM, DesChamps TD, Arnett AB, Cairney BE, Ma RQ, Webb SJ, Bernier RA (2018) Early enhanced processing and delayed habituation to deviance sounds in autism spectrum disorder. Brain Cognition 123:110-119. Huerta MF, Harting JK (1984) Connectional organization of the superior colliculus. Trends in Neurosciences 7:286-289. Hurtado-Parrado C (2010) Mecanismos neuronales del aprendizaje en los peces teleósteos. Universitas Psychologica 9:16. Huxlin KR, Martin T, Kelly K, Riley M, Friedman DI, Burgin WS, Hayhoe M (2009) Perceptual relearning of complex visual motion after V1 damage in humans. The Journal of neuroscience : the official journal of the Society for Neuroscience 29:3981-3991. Institute SCaI "ImageVis3D" An interactive visualization software system for large-scale volume data. Scientific Computing and Imaging Institute (SCI). In. Integrated Marine Observing System (IMOS) (2017) Lizard Island Deployments of the Facility for the Automated Intelligent Monitoring of Marine Systems - FAIMMS. In: https://apps.aims.gov.au/metadata/view/efc69c33-528f-4853-99aa-74d73e0daffa, accessed 11-Sep-2019. Ito H, Yamamoto N (2009) Non-laminar cerebral cortex in teleost fishes? Biology letters 5:117-121. Jain RA, Wolman MA, Marsden KC, Nelson JC, Shoenhard H, Echeverry FA, Szi C, Bell H, Skinner J, Cobbs EN, Sawada K, Zamora AD, Pereda AE, Granato M (2018) A Forward Genetic Screen in Zebrafish Identifies the G-Protein-Coupled Receptor CaSR as a Modulator of Sensorimotor Decision Making. Current biology : CB 28:1357-1369 e1355. Janak PH, Tye KM (2015) From circuits to behaviour in the amygdala. Nature 517:284-292. Jarvis ED et al. (2005) Avian brains and a new understanding of vertebrate brain evolution. Nat Rev Neurosci 6:151-159. Jiang X, Rosen E, Zeffiro T, Vanmeter J, Blanz V, Riesenhuber M (2006) Evaluation of a shape- based model of human face discrimination using FMRI and behavioral techniques. Neuron 50:159-172. Jiang X, Bradley E, Rini RA, Zeffiro T, VanMeter J, Riesenhuber M (2007) Categorization training results in shape- and category-selective human neural plasticity. Neuron 53:891-903. Jordi J, Guggiana-Nilo D, Bolton AD, Prabha S, Ballotti K, Herrera K, Rennekamp AJ, Peterson RT, Lutz TA, Engert F (2018) High-throughput screening for selective appetite modulators: A multibehavioral and translational drug discovery strategy. Science advances 4:eaav1966. Jouary A, Haudrechy M, Candelier R, Sumbre G (2016) A 2D virtual reality system for visual goal- driven navigation in zebrafish larvae. Scientific Reports 6. Jurjut O, Georgieva P, Busse L, Katzner S (2017) Learning Enhances Sensory Processing in Mouse V1 before Improving Behavior. The Journal of neuroscience : the official journal of the Society for Neuroscience 37:6460-6474. Jutfelt F, Hedgarde M (2013) Atlantic cod actively avoid CO2 and predator odour, even after long- term CO2 exposure. Front Zool 10. Jutfelt F, Hedgarde M (2015) Juvenile Atlantic cod behavior appears robust to near-future CO2 levels. Front Zool 12. Jutfelt F, Bresolin de Souza K, Vuylsteke A, Sturve J (2013) Behavioural disturbances in a temperate fish exposed to sustained high-CO2 levels. PloS one 8:e65825. Kalueff AV, Stewart AM, Gerlai R (2014) Zebrafish as an emerging model for studying complex brain disorders. Trends in Pharmacological Sciences 35:63-75. 163

Kalueff AV et al. (2013) Towards a Comprehensive Catalog of Zebrafish Behavior 1.0 and Beyond. Zebrafish 10:70-86. Karplus I, Goren M, Algom D (1982) A Preliminary Experimental-Analysis of Predator Face Recognition by Chromis-Caeruleus (Pisces, Pomacentridae). Z Tierpsychol 58:53-65. Karplus I, Katzenstein R, Goren M (2006) Predator recognition and social facilitation of predator avoidance in coral reef fish Dascyllus marginatus juveniles. Marine Ecology Progress Series 319:215-223. Kawakami K, Asakawa K, Hibi M, Itoh M, Muto A, Wada H (2016) Gal4 Driver Transgenic Zebrafish: Powerful Tools to Study Developmental Biology, Organogenesis, and Neuroscience. Advances in genetics 95:65-87. Kawasaki M, Aoki K (1983) Visual Responses Recorded from the Optic Tectum of Japanese Dace, Tribolodon-Hakonensis. J Comp Physiol 152:147-153. Kawashima T, Okuno H, Bito H (2014) A new era for functional labeling of neurons: activity- dependent promoters have come of age. Frontiers in neural circuits 8. Kawashima T, Zwart MF, Yang CT, Mensh BD, Ahrens MB (2016) The Serotonergic System Tracks the Outcomes of Actions to Mediate Short-Term Motor Learning. Cell 167:933-+. Keeler JF, Robbins TW (2011) Translating cognition from animals to humans. Biochem Pharmacol 81:1356-1366. Kelber A, Vorobyev M, Osorio D (2003) Animal colour vision - behavioural tests and physiological concepts. Biol Rev 78:81-118. Kelley JL, Magurran AE (2003) Learned predator recognition and antipredator responses in fishes. Fish and Fisheries 4:216-226. Kim DH, Kim J, Marques JC, Grama A, Hildebrand DGC, Gu WC, Li JM, Robson DN (2017) Pan- neuronal calcium imaging with cellular resolution in freely swimming zebrafish. Nat Methods 14:1107-+. Kim J, Castro L, Wasserman EA, Freeman JH (2018) Dorsal hippocampus is necessary for visual categorization in rats. Hippocampus 28:392-405. Kim L, He L, Maaswinkel H, Zhu LQ, Sirotkin H, Weng W (2014) Anxiety, hyperactivity and stereotypy in a zebrafish model of fragile X syndrome and autism spectrum disorder. Prog Neuro-Psychoph 55:40-49. Kim S, Jeneson A, van der Horst AS, Frascino JC, Hopkins RO, Squire LR (2011) Memory, visual discrimination performance, and the human hippocampus. The Journal of neuroscience : the official journal of the Society for Neuroscience 31:2624-2629. Kimmel CB, Patterson J, Kimmel RO (1974) The development and behavioral characteristics of the startle response in the zebra fish. Developmental Psychobiology 7:47-60. Kimmel CB, Eaton RC, Powell SL (1980) Decreased fast-start performance of zebrafish larvae lacking mauthner neurons. Journal of Comparative Physiology ? A 140:343-350. Kimura Y, Hisano Y, Kawahara A, Higashijima S (2014) Efficient generation of knock-in transgenic zebrafish carrying reporter/driver genes by CRISPR/Cas9-mediated genome engineering. Scientific Reports 4. King WM, Schmidt JT (1993) Nucleus Isthmi in Goldfish - Invitro Recordings and Fiber-Connections Revealed by Hrp Injections. Visual neuroscience 10:419-437. Kirsch JA, Hofmann MH, Mogdans J, Bleckmann H (2002) Responses of diencephalic neurons to sensory stimulation in the goldfish, Carassius auratus. Brain research bulletin 57:419-421. Kita EM, Scott EK, Goodhill GJ (2015) Topographic wiring of the retinotectal connection in zebrafish. Developmental neurobiology 75:542-556. Kleinhans NM, Richards T, Greenson J, Dawson G, Aylward E (2016) Altered Dynamics of the fMRI Response to Faces in Individuals with Autism. Journal of autism and developmental disorders 46:232-241.

164

Kleinhans NM, Johnson LC, Richards T, Mahurin R, Greenson J, Dawson G, Aylward E (2009) Reduced Neural Habituation in the Amygdala and Social Impairments in Autism Spectrum Disorders. Am J Psychiat 166:467-475. Kline DI et al. (2012) A short-term in situ CO2 enrichment experiment on Heron Island (GBR). Scientific Reports 2. Knoflach F, Hernandez MC, Bertrand D (2016) GABA(A) receptor-mediated neurotransmission: Not so simple after all. Biochem Pharmacol 115:10-17. Knutson AR, Hopkins RO, Squire LR (2012) Visual discrimination performance, memory, and medial temporal lobe function. Proceedings of the National Academy of Sciences of the United States of America 109:13106-13111. Koch M, Schnitzler HU (1997) The acoustic startle response in rats - circuits mediating evocation, inhibition and potentiation. Behavioural brain research 89:35-49. Kohashi T, Oda Y (2008) Initiation of Mauthner- or Non-Mauthner-Mediated Fast Escape Evoked by Different Modes of Sensory Input. Journal of Neuroscience 28:10641-10653. Koller K, Rafal RD, Platt A, Mitchell ND (2019) Orienting toward threat: Contributions of a subcortical pathway transmitting retinal afferents to the amygdala via the superior colliculus and pulvinar. Neuropsychologia 128:78-86. Korn H, Faber DS (1996) Escape behavior - Brainstem and spinal cord circuitry and function. Current opinion in neurobiology 6:826-832. Korn H, Faber DS (2005) The Mauthner cell half a century later: A neurobiological model for decision-making? Neuron 47:13-28. Kotrschal K, Van Staaden MJ, Huber R (1998) Fish brains: evolution and environmental relationships. Rev Fish Biol Fisher 8:373-408. Kourtzi Z (2010) Visual learning for perceptual and categorical decisions in the human brain. Vision research 50:433-440. Kramer A, Wu Y, Baier H, Kubo F (2019) Neuronal Architecture of a Visual Center that Processes Optic Flow. Neuron. Krasne FB, Teshiba TM (1995) Habituation of an Invertebrate Escape Reflex Due to Modulation by Higher Centers Rather Than Local Events. Proceedings of the National Academy of Sciences of the United States of America 92:3362-3366. Kroger RHH (2011) Physiological Optics in Fishes. Encyclopedia of Fish Physiology: From Genome to Environment, Vols 1-3:102-109. Kubo F, Hablitzel B, Dal Maschio M, Driever W, Baier H, Arrenberg AB (2014) Functional Architecture of an Optic Flow-Responsive Area that Drives Horizontal Eye Movements in Zebrafish. Neuron 81:1344-1359. Kunst M, Laurell E, Mokayes N, Kramer A, Kubo F, Fernandes AM, Forster D, Dal Maschio M, Baier H (2019) A Cellular-Resolution Atlas of the Larval Zebrafish Brain. Neuron 103:21-+. Kunzle H, Schnyder H (1984) The Isthmus-Tegmentum Complex in the Turtle and Rat - a Comparative-Analysis of Its Interconnections with the Optic Tectum. Experimental Brain Research 56:509-522. Kwan GT, Hamilton TJ, Tresguerres M (2017) CO2-induced ocean acidification does not affect individual or group behaviour in a temperate damselfish. Royal Society open science 4. Ladu F, Bartolini T, Panitz SG, Chiarotti F, Butail S, Macrì S, Porfiri M (2015) Live Predators, Robots, and Computer-Animated Images Elicit Differential Avoidance Responses in Zebrafish. Zebrafish 12:205-214. Lai F, Jutfelt F, Nilsson GE (2015) Altered neurotransmitter function in CO2-exposed stickleback (Gasterosteus aculeatus): a temperate model species for ocean acidification research. Conserv Physiol 3. Lai F, Fagernes CE, Jutfelt F, Nilsson GE (2017) Expression of genes involved in brain GABAergic neurotransmission in three-spined stickleback exposed to near-future CO2 (vol 4, pg 1, 2016). Conserv Physiol 5. 165

Lal P, Tanabe H, Suster ML, Ailani D, Kotani Y, Muto A, Itoh M, Iwasaki M, Wada H, Yaksi E, Kawakami K (2018) Identification of a neuronal population in the telencephalon essential for fear conditioning in zebrafish. Bmc Biol 16. Laming PR (1980) Electroencephalographic Studies on Arousal in the Goldfish (Carassius-Auratus). J Comp Physiol Psych 94:238-254. Laming PR (1987) Behavioral Arousal and Its Habituation in the Squirrel Fish, Holocentrus-Rufus - the Role of the Telecephalon. Behav Neural Biol 47:80-104. Laming PR, Mckinney SJ (1990) Habituation in Goldfish (Carassius-Auratus) Is Impaired by Increased Interstimulus-Interval, Interval Variability, and Telencephalic Ablation. Behavioral neuroscience 104:869-875. Landwehr K, Brendel E, Hecht H (2013) Luminance and contrast in visual perception of time to collision. Vision research 89:18-23. Larsson M (2015) Binocular vision, the optic chiasm, and their associations with vertebrate motor behavior. Frontiers in Ecology and Evolution 3. Lau BYB, Mathur P, Gould GG, Guo S (2011) Identification of a brain center whose activity discriminates a choice behavior in zebrafish. Proceedings of the National Academy of Sciences of the United States of America 108:2581-2586. LeDoux J (2007) The amygdala. Current Biology 17:R868-R874. Lee AC, Yeung LK, Barense MD (2012) The hippocampus and visual perception. Frontiers in human neuroscience 6:91. Lee S, Chun HS, Lee J, Park HJ, Kim KT, Kim CH, Yoon S, Kim WK (2018) Plausibility of the zebrafish embryos/larvae as an alternative animal model for autism: A comparison study of transcriptome changes. PloS one 13. Leung LC, Wang GX, Mourrain P (2013) Imaging zebrafish neural circuitry from whole brain to synapse. Frontiers in neural circuits 7. Li X, Li X, Chen D, Guo JL, Feng DF, Sun MZ, Lu Y, Chen DY, Zhao X, Feng XZ (2016) Evaluating the biological impact of polyhydroxyalkanoates (PHAs) on developmental and exploratory profile of zebrafish larvae. Rsc Adv 6:37018-37030. Liebeskind BJ, Hillis DM, Zakon HH, Hofmann HA (2016) Complex Homology and the Evolution of Nervous Systems. Trends in ecology & evolution 31:127-135. Light SEW, Jontes JD (2019) Multiplane Calcium Imaging Reveals Disrupted Development of Network Topology in Zebrafish pcdh19 Mutants. eNeuro 6. Lima MG, Silva RXD, Silva SDD, Rodrigues LDD, Oliveira KRHM, Batista EDO, Maximino C, Herculano AM (2016) Time-dependent sensitization of stress responses in zebrafish: A putative model for post-traumatic stress disorder. Behav Process 128:70-82. Lin MZ, Schnitzer MJ (2016) Genetically encoded indicators of neuronal activity. Nature neuroscience 19:1142-1153. Lin ZS, Yazulla S (1994) Heterogeneity of Gaba(a) Receptor in Goldfish Retina. Journal of Comparative Neurology 345:429-439. Liu J, Baraban SC (2019) Network Properties Revealed during Multi-Scale Calcium Imaging of Seizure Activity in Zebrafish. eNeuro 6. Liu KS, Fetcho JR (1999) Laser ablations reveal functional relationships of segmental hindbrain neurons in zebrafish. Neuron 23:325-335. Liu Y, Chen LM (2014) Towards bridging the gap between acid-base transporters and neuronal excitability modulation. International journal of physiology, pathophysiology and pharmacology 6:221-225. Liu ZX, Shen K, Olsen RK, Ryan JD (2017) Visual Sampling Predicts Hippocampal Activity. Journal of Neuroscience 37:599-609. Loffler G (2015) Probing intermediate stages of shape processing. J Vision 15. López-Schier H (2019) Neuroplasticity in the acoustic startle reflex in larval zebrafish. Current opinion in neurobiology 54:134-139. 166

Losey GS, Cronin TW, Goldsmith TH, Hyde D, Marshall NJ, McFarland WN (1999) The UV visual world of fishes: a review. J Fish Biol 54:921-943. Lovelace JW, Wen TH, Reinhard S, Hsu MS, Sidhu H, Ethell IM, Binder DK, Razak KA (2016) Matrix metalloproteinase-9 deletion rescues auditory evoked potential habituation deficit in a mouse model of Fragile X Syndrome. Neurobiol Dis 89:126-135. Lovett-Barron M, Andalman AS, Allen WE, Vesuna S, Kauvar I, Burns VM, Deisseroth K (2017) Ancestral Circuits for the Coordinated Modulation of Brain State. Cell 171:1411-+. Lowe AS, Nikolaou N, Hunter PR, Thompson ID, Meyer MP (2013) A Systems-Based Dissection of Retinal Inputs to the Zebrafish Tectum Reveals Different Rules for Different Functional Classes during Development. Journal of Neuroscience 33:13946-13956. Luca RM, Gerlai R (2012) In search of optimal fear inducing stimuli: Differential behavioral responses to computer animated images in zebrafish. Behavioural brain research 226:66-76. Luiten PG (1981) Afferent and efferent connections of the optic tectum in the carp (Cyprinus carpio L.). Brain research 220:51-65. Lyon EP (1904) On rheotropism. I. Rheotropism in fishes. Am J Physiol 12:149-161. Mackintosh J, Sutherland N (1963) Visual discrimination by the goldfish: The orientation of rectangles. Animal Behaviour 11:135--141. Mai JK, Majtanik M, Paxinos G (2015) Atlas of the human brain: Academic Press. Makin TR, Orban de Xivry JJ (2019) Ten common statistical mistakes to watch out for when writing or reviewing a manuscript. Elife 8. Maneja RH, Frommel AY, Browman HI, Clemmesen C, Geffen AJ, Folkvord A, Piatkowski U, Durif CMF, Bjelland R, Skiftesvik AB (2013) The swimming kinematics of larval Atlantic cod, Gadus morhua L., are resilient to elevated seawater pCO(2). Mar Biol 160:1963-1972. Maneja RH, Frommel AY, Browman HI, Geffen AJ, Folkvord A, Piatkowski U, Durif CMF, Bjelland R, Skiftesvik AB, Clemmesen C (2015) The swimming kinematics and foraging behavior of larval Atlantic herring (Clupea harengus L.) are unaffected by elevated pCO(2). J Exp Mar Biol Ecol 466:42-48. Marchiafava PL (1985) Cell Coupling in Double Cones of the Fish Retina. Proc R Soc Ser B-Bio 226:211-215. Marin G, Salas C, Sentis E, Rojas X, Letelier JC, Mpodozis J (2007) A cholinergic gating mechanism controlled by competitive interactions in the optic tectum of the pigeon. Journal of Neuroscience 27:8112-8121. Marquez-Legorreta E, Horta JDC, Berrebi AS, Saldana E (2016) Organization of the Zone of Transition between the Pretectum and the Thalamus, with Emphasis on the Pretectothalamic Lamina. Frontiers in neuroanatomy 10. Marsden KC, Granato M (2015) In vivo Ca 2+ imaging reveals that decreased dendritic excitability drives startle habituation. Cell reports 13:1733-1740. Marshall NJ, Cheney K (2011) VISION | Color Vision and Color Communication in Reef Fish. In: Encyclopedia of Fish Physiology (Farrell AP, ed), pp 150-158. San Diego: Academic Press. Marshall NJ, Cortesi F, de Busserolles F, Siebeck UE, Cheney KL (2019) Colours and colour vision in reef fishes: Past, present and future research directions. J Fish Biol 95:5-38. Martin I, Gomez A, Salas C, Puerto A, Rodriguez F (2011) Dorsomedial pallium lesions impair taste aversion learning in goldfish. Neurobiology of learning and memory 96:297-305. Maruska KP, Tricas TC (2009) Central projections of octavolateralis nerves in the brain of a soniferous damselfish (Abudefduf abdominalis). The Journal of comparative neurology 512:628-650. Mccabe BJ, Horn G (1994) Learning-Related Changes in Fos-Like Immunoreactivity in the Chick Forebrain after Imprinting. Proceedings of the National Academy of Sciences of the United States of America 91:11417-11421. McClenahan P, Troup M, Scott EK (2012) Fin-Tail Coordination during Escape and Predatory Behavior in Larval Zebrafish. PloS one 7:e32295. 167

McCormick MI (2016) Protogyny in a tropical damselfish: females queue for future benefit. Peerj 4. McDiarmid TA, Bemardos AC, Rankin CH (2017) Habituation is altered in neuropsychiatric disorders-A comprehensive review with recommendations for experimental design and analysis. Neuroscience and biobehavioral reviews 80:286-305. McDiarmid TA, Yu AJ, Rankin CH (2018a) Beyond the response-High throughput behavioral analyses to link genome to phenome in Caenorhabditis elegans. Genes Brain Behav 17:e12437. McDiarmid TA, Yu A, Rankin CH (2019) Habituation Is More Than Learning to Ignore: Multiple Mechanisms Serve to Facilitate Shifts in Behavioral Strategy. Bioessays 41. McDiarmid TA, Au V, Loewen AD, Liang J, Mizumoto K, Moerman DG, Rankin CH (2018b) CRISPR-Cas9 human gene replacement and phenomic characterization in Caenorhabditis elegans to understand the functional conservation of human genes and decipher variants of uncertain significance. Disease models & mechanisms 11. McDiarmid TA, Belmadani M, Liang J, Meili F, Mathews EA, Mullen GP, Hendi A, Wong WR, Rand JB, Mizumoto K, Haas K, Pavlidis P, Rankin CH (2020) Systematic phenomics analysis of autism-associated genes reveals parallel networks underlying reversible impairments in habituation. Proceedings of the National Academy of Sciences of the United States of America 117:656-667. McFadyen J (2019) Investigating the Subcortical Route to the Amygdala Across Species and in Disordered Fear Responses. J Exp Neurosci 13. McFadyen J, Mattingley JB, Garrido MI (2019) An afferent white matter pathway from the pulvinar to the amygdala facilitates fear recognition. Elife 8. McFadyen J, Mermillod M, Mattingley JB, Halasz V, Garrido MI (2017) A Rapid Subcortical Amygdala Route for Faces Irrespective of Spatial Frequency and Emotion. Journal of Neuroscience 37:3864-3874. McKinnon CJ et al. (2019) Restricted and Repetitive Behavior and Brain Functional Connectivity in Infants at Risk for Developing Autism Spectrum Disorder. Biological psychiatry Cognitive neuroscience and neuroimaging 4:50-61. Medan V, Preuss T (2014) The Mauthner-cell circuit of fish as a model system for startle plasticity. J Physiol-Paris 108:129-140. Meshalkina DA, Kizlyk MN, Kysil EV, Collier AD, Echevarria DJ, Abreu MS, Barcellos LJG, Song C, Warnick JE, Kyzar EJ, Kalueff AV (2018) Zebrafish models of autism spectrum disorder. Experimental neurology 299:207-216. Meuwese JDI, van Loon AM, Scholte HS, Lirk PB, Vulink NCC, Hollmann MW, Lamme VAF (2013) NMDA Receptor Antagonist Ketamine Impairs Feature Integration in Visual Perception. PloS one 8. Mila M, Alvarez-Mora MI, Madrigal I, Rodriguez-Revenga L (2018) Fragile X syndrome: An overview and update of the FMR1 gene. Clin Genet 93:197-205. Miller EK, Freedman DJ, Wallis JD (2002) The prefrontal cortex: categories, concepts and cognition. Philos T Roy Soc B 357:1123-1136. Miller LJ, McIntosh DN, McGrath J, Shyu V, Lampe M, Taylor AK, Tassone F, Neitzel K, Stackhouse T, Hagerman RJ (1999) Electrodermal responses to sensory stimuli in individuals with fragile X syndrome: A preliminary report. Am J Med Genet 83:268-279. Minatohara K, Akiyoshi M, Okuno H (2016) Role of Immediate-Early Genes in Synaptic Plasticity and Neuronal Ensembles Underlying the Memory Trace. Front Mol Neurosci 8. Mitchell L, Cheney KL, Cortesi F, Marshall NJ, Vorobyev M (2017) Triggerfish uses chromaticity and lightness for object segregation. Royal Society open science 4:171440. Mora-Ferrer C, Neumeyer C (2009) Neuropharmacology of vision in goldfish: a review. Vision research 49:960-969.

168

Mora-Ferrer C, Hausselt S, Hoffmann RS, Ebisch B, Schick S, Wollenberg K, Schneider C, Teege P, Jurgens K (2005) Pharmacological properties of motion vision in goldfish measured with the optomotor response. Brain research 1058:17-29. Morris JS, Ohman A, Dolan RJ (1999) A subcortical pathway to the right amygdala mediating "unseen" fear. Proceedings of the National Academy of Sciences of the United States of America 96:1680-1685. Morton AJ, Skillings E, Bussey TJ, Saksida LM (2006) Measuring cognitive deficits in disabled mice using an automated interactive touchscreen system. Nat Methods 3:767-767. Mu Y, Bennett DV, Rubinov M, Narayan S, Yang CT, Tanimoto M, Mensh BD, Looger LL, Ahrens MB (2019) Glia Accumulate Evidence that Actions Are Futile and Suppress Unsuccessful Behavior. Cell 178:27-+. Mucha PJ, Richardson T, Macon K, Porter MA, Onnela JP (2010) Community structure in time- dependent, multiscale, and multiplex networks. Science 328:876-878. Mueller KP, Neuhauss SCF (2012) Automated visual choice discrimination learning in zebrafish (Danio rerio). J Integr Neurosci 11:73-85. Mueller T (2012) What is the Thalamus in Zebrafish? Frontiers in Neuroscience 6. Mueller T, Wullimann MF (2016) Chapter 2 - Atlas of Cellular Markers in Zebrafish Neurogenesis: Atlas. In: Atlas of Early Zebrafish Brain Development (Second Edition) (Mueller T, Wullimann MF, eds), pp 27-157. San Diego: Elsevier. Mueller T, Dong Z, Berberoglu MA, Guo S (2011) The dorsal pallium in zebrafish, Danio rerio (Cyprinidae, Teleostei). Brain research 1381:95-105. Munday PL, Warner RR, Monro K, Pandolfi JM, Marshall DJ (2013) Predicting evolutionary responses to climate change in the sea. Ecology letters 16:1488-1500. Munday PL, Cheal AJ, Dixson DL, Rummer JL, Fabricius KE (2014) Behavioural impairment in reef fishes caused by ocean acidification at CO2 seeps. Nature Clim Change 4:487-492. Munday PL, Dixson DL, McCormick MI, Meekan M, Ferrari MC, Chivers DP (2010) Replenishment of fish populations is threatened by ocean acidification. Proceedings of the National Academy of Sciences of the United States of America 107:12930-12934. Munday PL, Welch MJ, Allan BJM, Watson SA, McMahon SJ, McCormick MI (2016) Effects of elevated CO2 on predator avoidance behaviour by reef fishes is not altered by experimental test water. Peerj 4. Murakami T, Morita Y, Ito H (1983) Extrinsic and Intrinsic Fiber-Connections of the Telencephalon in a Teleost, Sebastiscus-Marmoratus. Journal of Comparative Neurology 216:115-131. Murray EA, Wise SP, Graham KS (2018) Representational specializations of the hippocampus in phylogenetic perspective. Neuroscience letters 680:4-12. Murray TK, Ridley RM, Snape MF, Cross AJ (1995) The Effect of Dizocilpine (Mk-801) on Spatial and Visual-Discrimination Tasks in the Rat. Behav Pharmacol 6:540-549. Muto A, Ohkura M, Abe G, Nakai J, Kawakami K (2013) Real-time visualization of neuronal activity during perception. Current biology : CB 23:307-311. Mysore SP, Knudsen EI (2013) A shared inhibitory circuit for both exogenous and endogenous control of stimulus selection. Nature neuroscience 16:473-U143. Mysore SP, Knudsen EI (2014) Descending Control of Neural Bias and Selectivity in a Spatial Attention Network: Rules and Mechanisms. Neuron 84:214-226. Nagelkerken I, Munday PL (2016) Animal behaviour shapes the ecological effects of ocean acidification and warming: moving from individual to community-level responses. Global change biology 22:974-989. Nagy A, Paroczy Z, Markus Z, Berenyi A, Wypych M, Waleszczyk WJ, Benedek G (2008) Drifting grating stimulation reveals particular activation properties of visual neurons in the caudate nucleus. European Journal of Neuroscience 27:1801-1808. Nair A, Azatian G, McHenry MJ (2015) The kinematics of directional control in the fast start of zebrafish larvae. The Journal of experimental biology 218:3996-4004. 169

Nair A, Nguyen C, McHenry MJ (2017) A faster escape does not enhance survival in zebrafish larvae. P Roy Soc B-Biol Sci 284. Naumann EA, Fitzgerald JE, Dunn TW, Rihel J, Sompolinsky H, Engert F (2016) From Whole-Brain Data to Functional Circuit Models: The Zebrafish Optomotor Response. Cell 167:947-+. Nava SS, An S, Hamil T (2011) Visual detection of UV cues by adult zebrafish (Danio rerio). J Vision 11. Neuhauss SCF (2003) Behavioral genetic approaches to visual system development and function in zebrafish. J Neurobiol 54:148-160. Neumeyer C (1992) Tetrachromatic Color-Vision in Goldfish - Evidence from Color Mixture Experiments. J Comp Physiol A 171:639-649. Nevin LM, Robles E, Baier H, Scott EK (2010) Focusing on optic tectum circuitry through the lens of genetics. Bmc Biol 8. Newport C, Wallis G, Siebeck UE (2014) Concept learning and the use of three common psychophysical paradigms in the archerfish (Toxotes chatareus). Frontiers in neural circuits 8:39. Newport C, Wallis G, Siebeck UE (2015) Same/Different Abstract Concept Learning by Archerfish (Toxotes chatareus). PloS one 10. Newport C, Wallis G, Siebeck UE (2018) Object recognition in fish: accurate discrimination across novel views of an unfamiliar object category (human faces). Animal Behaviour 145:39-49. Newport C, Wallis G, Temple SE, Siebeck UE (2013) Complex, context-dependent decision strategies of archerfish, Toxotes chatareus. Animal Behaviour 86:1265-1274. Newport C, Wallis G, Reshitnyk Y, Siebeck UE (2016) Discrimination of human faces by archerfish (Toxotes chatareus). Scientific Reports 6. Newport C, Green NF, McClure EC, Osorio DC, Vorobyev M, Marshall NJ, Cheney KL (2017) Fish use colour to learn compound visual signals. Animal Behaviour 125:93-100. Ng MC, Yang YL, Lu KT (2013) Behavioral and Synaptic Circuit Features in a Zebrafish Model of Fragile X Syndrome. PloS one 8. Nikolaou N, Lowe AS, Walker AS, Abbas F, Hunter PR, Thompson ID, Meyer MP (2012) Parametric Functional Maps of Visual Inputs to the Tectum. Neuron 76:317-324. Nilsson GE, Dixson DL, Domenici P, McCormick MI, Sorensen C, Watson S-A, Munday PL (2012) Near-future carbon dioxide levels alter fish behaviour by interfering with neurotransmitter function. Nature Clim Change 2:201-204. Nithianantharajah J, Grant SGN (2013) Cognitive components in mice and humans: Combining genetics and touchscreens for medical translation. Neurobiology of learning and memory 105:13-19. Nithianantharajah J, Komiyama NH, McKechanie A, Johnstone M, Blackwood DH, St Clair D, Emes RD, van de Lagemaat LN, Saksida LM, Bussey TJ, Grant SGN (2013) Synaptic scaffold evolution generated components of vertebrate cognitive complexity. Nature neuroscience 16:16-U37. Northcutt RG (2006) Connections of the lateral and medial divisions of the goldfish telencephalic pallium. The Journal of comparative neurology 494:903-943. Northcutt RG (2008) Forebrain evolution in bony fishes. Brain research bulletin 75:191-205. Northcutt RG (2011) Do Teleost Fishes Possess a Homolog of Mammalian Isocortex? Brain, behavior and evolution 78:136-138. Northcutt RG, Wullimann MF (1988) The Visual System in Teleost Fishes: Morphological Patterns and Trends. In, pp 515-552. New York, NY: Springer New York. Northmore D (2011) VISION | Optic Tectum. In: Encyclopedia of Fish Physiology (Farrell AP, ed), pp 131-142. San Diego: Academic Press. Northmore DPM (2017) Holding visual attention for 400 million years: A model of tectum and torus longitudinalis in teleost fishes. Vision research 131:44-56.

170

Northmore DPM, Gallagher SP (2003) Functional relationship between nucleus isthmi and tectum in teleosts: Synchrony but no topography. Visual neuroscience 20:335-348. Northmore DPM, Graham BJ (2005) Avoidance behavior controlled by a model of vertebrate midbrain mechanisms. Lect Notes Comput Sc 3561:338-345. Nowicki JP, Miller GM, Munday PL (2012) Interactive effects of elevated temperature and CO2 on foraging behavior of juvenile coral reef fish. J Exp Mar Biol Ecol 412:46-51. O'Connell LA, Matthews BJ, Hofmann HA (2012) Isotocin regulates paternal care in a monogamous cichlid fish. Horm Behav 61:725-733. O'Connell LA, Rigney MM, Dykstra DW, Hofmann HA (2013) Neuroendocrine Mechanisms Underlying Sensory Integration of Social Signals. J Neuroendocrinol 25:644-654. O'Neale A, Ellis J, Creton R, Colwill RM (2014) Single stimulus learning in zebrafish larvae. Neurobiology of learning and memory 108:145-154. Obenar JD (1976) Electrophysiology of Neural Units in Goldfish Optic Tectum. Brain research bulletin 1:529-541. Oliva D, Tomsic D (2012) Visuo-motor transformations involved in the escape response to looming stimuli in the crab Neohelice (=Chasmagnathus) granulata. Journal of Experimental Biology 215:3488-3500. Oliveira J, Silveira M, Chacon D, Luchiari A (2015) The Zebrafish World of Colors and Shapes: Preference and Discrimination. Zebrafish 12:166-173. Oliveira RF (2013) Mind the fish: zebrafish as a model in cognitive social neuroscience. Frontiers in neural circuits 7. Olmos-Serrano JL, Paluszkiewicz SM, Martin BS, Kaufmann WE, Corbin JG, Huntsman MM (2010) Defective GABAergic Neurotransmission and Pharmacological Rescue of Neuronal Hyperexcitability in the Amygdala in a Mouse Model of Fragile X Syndrome. Journal of Neuroscience 30:9929-9938. Orger MB, Kampff AR, Severi KE, Bollmann JH, Engert F (2008) Control of visually guided behavior by distinct populations of spinal projection neurons. Nature neuroscience 11:327- 333. Orger MB, Gahtan E, Muto A, Page-McCaw P, Smear MC, Baier H (2004) Behavioral screening assays in zebrafish. Method Cell Biol 77:53-68. Pachauri RK, Allen MR, Barros V, Broome J, Cramer W, Christ R, Church J, Clarke L, Dahe Q, Dasgupta P (2014) Climate change 2014: synthesis Report. Contribution of working groups I, II and III to the fifth assessment report of the intergovernmental panel on climate change: IPCC. Palande NV, Bhoyar RC, Biswas SP, Jadhao AG (2015) Short-term exposure to L-type calcium channel blocker, verapamil, alters the expression pattern of calcium-binding proteins in the brain of goldfish, Carassius auratus. Comp Biochem Phys C 176:31-43. Pantoja C, Hoagland A, Carroll EC, Karalis V, Conner A, Isacoff EY (2016) Neuromodulatory Regulation of Behavioral Individuality in Zebrafish. Neuron 91:587-601. Parker AN, Fritsches KA, Newport C, Wallis G, Siebeck UE (2017) Comparison of functional and anatomical estimations of visual acuity in two species of coral reef fish. The Journal of experimental biology 220:2387-2396. Paroczy Z, Nagy A, Markus Z, Waleszczyk WJ, Wypych M, Benedek G (2006) Spatial and temporal visual properties of single neurons in the suprageniculate nucleus of the thalamus. Neuroscience 137:1397-1404. Parvizi J (2009) Corticocentric myopia: old bias in new cognitive sciences. Trends in cognitive sciences 13:354-359. Patterson GH, Lippincott-Schwartz J (2002) A photoactivatable GFP for selective photolabeling of proteins and cells. Science 297:1873-1877. Patton BW, Braithwaite VA (2015) Changing tides: ecological and historical perspectives on fish cognition. Wiley interdisciplinary reviews Cognitive science 6:159-176. 171

Paxinos G, Franklin KB (2001) The mouse brain in stereotaxic coordinates. 2. San Diego: Academic. Paxinos G, Watson C (2007) The rat brain in stereotaxic coordinates, 6th Edition. Amsterdam ; Boston ;: Academic Press/Elsevier. Peek MY, Card GM (2016) Comparative approaches to escape. Current opinion in neurobiology 41:167-173. Peeters BWMM, Moeskops M, Veenvliet ARJ (2016) Color Preference in Danio rerio: Effects of Age and Anxiolytic Treatments. Zebrafish 13:330-334. Perathoner S, Cordero-Maldonado ML, Crawford AD (2016) Potential of zebrafish as a model for exploring the role of the amygdala in emotional memory and motivational behavior. J Neurosci Res 94:445-462. Pereira AG, Moita MA (2016) Is there anybody out there? Neural circuits of threat detection in vertebrates. Current opinion in neurobiology 41:179-187. Pessoa L, Adolphs R (2010) Emotion processing and the amygdala: from a 'low road' to 'many roads' of evaluating biological significance. Nat Rev Neurosci 11:773-782. Pessoa L, Adolphs R (2011) Emotion and the brain: multiple roads are better than one. Nat Rev Neurosci 12:425-U472. Petruno SK, Clark RE, Reinagel P (2013) Evidence That Primary Visual Cortex Is Required for Image, Orientation, and Motion Discrimination by Rats. PloS one 8. Phelps EA, LeDoux JE (2005) Contributions of the amygdala to emotion processing: From animal models to human behavior. Neuron 48:175-187. Phillips GAC, How MJ, Lange JE, Marshall NJ, Cheney KL (2017) Disruptive colouration in reef fish: does matching the background reduce predation risk? The Journal of experimental biology 220:1962-1974. Pignatelli V, Champ C, Marshall J, Vorobyev M (2010) Double cones are used for colour discrimination in the reef fish, Rhinecanthus aculeatus. Biology letters 6:537-539. Pirone A, Giannaccini G, Betti L, Lucacchini A, Mascia G, Fabbrini L, Italiani P, Uccelli A, Lenzi C, Fabiani O (2006) Autoradiographic localization and binding study of benzodiazepines receptor sites in carp brain (Cyprinus carpio L.). Journal of chemical neuroanatomy 31:139- 145. Plichta MM, Grimm O, Morgen K, Mier D, Sauer C, Haddad L, Tost H, Esslinger C, Kirsch P, Schwarz AJ, Meyer-Lindenberg A (2014) Amygdala habituation: A reliable fMRI phenotype. NeuroImage 103:383-390. Pnevmatikakis EA, Giovannucci A (2017) NoRMCorre: An online algorithm for piecewise rigid motion correction of calcium imaging data. Journal of neuroscience methods 291:83-94. Pnevmatikakis EA, Soudry D, Gao YJ, Machado TA, Merel J, Pfau D, Reardon T, Mu Y, Lacefield C, Yang WJ, Ahrens M, Bruno R, Jessell TM, Peterka DS, Yuste R, Paninski L (2016a) Simultaneous Denoising, Deconvolution, and Demixing of Calcium Imaging Data. Neuron 89:285-299. Pnevmatikakis EA, Soudry D, Gao Y, Machado TA, Merel J, Pfau D, Reardon T, Mu Y, Lacefield C, Yang W, Ahrens M, Bruno R, Jessell TM, Peterka DS, Yuste R, Paninski L (2016b) Simultaneous Denoising, Deconvolution, and Demixing of Calcium Imaging Data. Neuron 89:285-299. Poort J, Khan AG, Pachitariu M, Nemri A, Orsolic I, Krupic J, Bauza M, Sahani M, Keller GB, Mrsic- Flogel TD, Hofer SB (2015) Learning Enhances Sensory and Multiple Non-sensory Representations in Primary Visual Cortex. Neuron 86:1478-1490. Port RG, Gandal MJ, Roberts TPL, Siegel SJ, Carlson GC (2014) Convergence of circuit dysfunction in ASD: a common bridge between diverse genetic and environmental risk factors and common clinical electrophysiology. Front Cell Neurosci 8. Portavella M, Vargas JP, Torres B, Salas C (2002) The effects of telencephalic pallial lesions on spatial, temporal, and emotional learning in goldfish. Brain research bulletin 57:397-399.

172

Portugues R, Engert F (2011) Adaptive locomotor behavior in larval zebrafish. Front Syst Neurosci 5. Portugues R, Feierstein CE, Engert F, Orger MB (2014) Whole-Brain Activity Maps Reveal Stereotyped, Distributed Networks for Visuomotor Behavior. Neuron 81:1328-1343. Post N, von der Emde G (1999) The "novelty response" in an electric fish: response properties and habituation. Physiology & behavior 68:115-128. Prechtl JC, von der Emde G, Wolfart J, Karamursel S, Akoev GN, Andrianov YN, Bullock TH (1998) Sensory processing in the pallium of a mormyrid fish. The Journal of neuroscience : the official journal of the Society for Neuroscience 18:7381-7393. Preuss SJ, Trivedi CA, vom Berg-Maurer CM, Ryu S, Bollmann JH (2014) Classification of Object Size in Retinotectal Microcircuits. Current Biology 24:2376-2385. Preuss T, Osei-Bonsu PE, Weiss SA, Wang C, Faber DS (2006) Neural representation of object approach in a decision-making motor circuit. Journal of Neuroscience 26:3454-3464. Puschina EV, Varaksin AA (2011) Hydrogen Sulfide-, Parvalbumin-, and GABA-Producing Systems in the Masu Salmon Brain. Neurophysiology+ 43:90-102. Quick IA, Laming PR (1990) Relationship between Ecg, Eeg and Sps Responses during Arousal in the Goldfish (Carassius-Auratus). Comp Biochem Phys A 95:459-471. R Development Core Team (2018) R: A language and environment for statistical computing. In, 3.5.2 Edition. Vienna, Austria: R Foundation for Statistical Computing. Raderschall CA, Magrath RD, Hemmi JM (2011) Habituation under natural conditions: model predators are distinguished by approach direction. Journal of Experimental Biology 214:4209- 4216. Rafal R, Henik A, Smith J (1991) Extrageniculate Contributions to Reflex Visual Orienting in Normal Humans - a Temporal Hemifield Advantage. Journal of Cognitive Neuroscience 3:322-328. Rafal RD, Koller K, Bultitude JH, Mullins P, Ward R, Mitchell AS, Bell AH (2015) Connectivity between the superior colliculus and the amygdala in humans and macaque monkeys: virtual dissection with probabilistic DTI tractography. Journal of neurophysiology 114:1947-1962. Rais M, Binder DK, Razak KA, Ethell IM (2018) Sensory Processing Phenotypes in Fragile X Syndrome. Asn Neuro 10. Rajan KE, Ganesh A, Dharaneedharan S, Radhakrishnan K (2011) Spatial learning-induced egr-1 expression in telencephalon of gold fish Carassius auratus. Fish Physiol Biochem 37:153-159. Ramasamy RA, Allan BJM, McCormick MI (2015) Plasticity of Escape Responses: Prior Predator Experience Enhances Escape Performance in a Coral Reef Fish. PloS one 10. Ramaswami M (2014) Network plasticity in adaptive filtering and behavioral habituation. Neuron 82:1216-1229. Randlett O, Haesemeyer M, Forkin G, Shoenhard H, Schier AF, Engert F, Granato M (2019) Distributed Plasticity Drives Visual Habituation Learning in Larval Zebrafish. Current Biology 29:1337-+. Randlett O, Wee CL, Naumann EA, Nnaemeka O, Schoppik D, Fitzgerald JE, Portugues R, Lacoste AMB, Riegler C, Engert F, Schier AF (2015) Whole-brain activity mapping onto a zebrafish brain atlas. Nat Methods 12:1039-1046. Rankin CH, Broster BS (1992) Factors Affecting Habituation and Recovery from Habituation in the Nematode Caenorhabditis-Elegans. Behavioral neuroscience 106:239-249. Rankin CH, Beck CDO, Chiba CM (1990) Caenorhabditis-Elegans - a New Model System for the Study of Learning and Memory. Behavioural brain research 37:89-92. Rankin CH, Abrams T, Barry RJ, Bhatnagar S, Clayton DF, Colombo J, Coppola G, Geyer MA, Glanzman DL, Marsland S, McSweeney FK, Wilson DA, Wu CF, Thompson RF (2009) Habituation revisited: An updated and revised description of the behavioral characteristics of habituation. Neurobiology of learning and memory 92:135-138. 173

Renninger SL, Orger MB (2013) Two-photon imaging of neural population activity in zebrafish. Methods 62:255-267. Rigoulot S, Knoth IS, Lafontaine MP, Vannasing P, Major P, Jacquemont S, Michaud JL, Jerbi K, Lippe S (2017) Altered visual repetition suppression in Fragile X Syndrome: New evidence from ERPs and oscillatory activity. Int J Dev Neurosci 59:52-59. Riley E, Kopotiyenko K, Zhdanova I (2015) Prenatal and acute cocaine exposure affects neural responses and habituation to visual stimuli. Frontiers in neural circuits 9:41. Rinner O, Rick JM, Neuhauss SCF (2005) Contrast sensitivity, spatial and temporal tuning of the larval zebrafish optokinetic response. Invest Ophth Vis Sci 46:137-142. Rischawy I, Schuster S (2013) Visual search in hunting archerfish shares all hallmarks of human performance. Journal of Experimental Biology 216:3096-3103. Ro T, Shelton D, Lee OL, Chang E (2004) Extrageniculate mediation of unconscious vision in transcranial magnetic stimulation-induced blindsight. Proceedings of the National Academy of Sciences of the United States of America 101:9933-9935. Roberts AC, Bill BR, Glanzman DL (2013) Learning and memory in zebrafish larvae. Frontiers in neural circuits 7. Roberts AC, Pearce KC, Choe RC, Alzagatiti JB, Yeung AK, Bill BR, Glanzman DL (2016) Long- term habituation of the C-start escape response in zebrafish larvae. Neurobiology of learning and memory 134 Pt B:360-368. Roberts AC, Reichl J, Song MY, Dearinger AD, Moridzadeh N, Lu ED, Pearce K, Esdin J, Glanzman DL (2011) Habituation of the C-Start Response in Larval Zebrafish Exhibits Several Distinct Phases and Sensitivity to NMDA Receptor Blockade. PloS one 6. Roberts AC, Chornak J, Alzagatiti JB, Ly DT, Bill BR, Trinkeller J, Pearce KC, Choe RC, Campbell CS, Wong D, Deutsch E, Hernandez S, Glanzman DL (2019) Rapid habituation of a touch- induced escape response in Zebrafish (Danio rerio) Larvae. PloS one 14. Robertson CE, Baron-Cohen S (2017) Sensory perception in autism. Nat Rev Neurosci 18:671-684. Robles E, Smith SJ, Baier H (2011) Characterization of genetically targeted neuron types in the zebrafish optic tectum. Frontiers in neural circuits 5. Robles E, Filosa A, Baier H (2013) Precise Lamination of Retinal Axons Generates Multiple Parallel Input Pathways in the Tectum. Journal of Neuroscience 33:5027-5039. Robles E, Laurell E, Baier H (2014) The Retinal Projectome Reveals Brain-Area-Specific Visual Representations Generated by Ganglion Cell Diversity. Current Biology 24:2085-2096. Rodgers WL, Segal JR, Melzack R (1963) Tail Flip Response in Goldfish. J Comp Physiol Psych 56:917-&. Rodriguez F, Lopez JC, Vargas JP, Broglio C, Gomez Y, Salas C (2002) Spatial memory and hippocampal pallium through vertebrate evolution: Insights from reptiles and teleost fish. Brain research bulletin 57:499-503. Rodriguez F, Duran E, Gomez A, Ocana FM, Alvarez E, Jimenez-Moya F, Broglio C, Salas C (2005) Cognitive and emotional functions of the teleost fish cerebellum. Brain research bulletin 66:365-370. Roeser T, Baier H (2003) Visuomotor behaviors in larval zebrafish after GFP-guided laser ablation of the optic tectum. The Journal of neuroscience : the official journal of the Society for Neuroscience 23:3726-3734. Rogers SJ, Ozonoff S (2005) Annotation: What do we know about sensory dysfunction in autism? A critical review of the empirical evidence. J Child Psychol Psyc 46:1255-1268. Rokszin A, Markus Z, Braunitzer G, Berenyi A, Benedek G, Nagy A (2010) Visual Pathways Serving Motion Detection in the Mammalian Brain. Sensors-Basel 10:3218-3242. Rooney DJ, Laming PR (1986) Cardiac and Ventilatory Arousal Responses and Their Habituation in Goldfish - Effects of the Intensity of the Eliciting Stimulus. Physiology & behavior 37:11-14.

174

Rooney DJ, Laming PR (1988) Effects of Telencephalic Ablation on Habituation of Arousal Responses, within and between Daily Training Sessions in Goldfish. Behav Neural Biol 49:83-96. Rose JK, Kaun KR, Rankin CH (2002) A new group-training procedure for habituation demonstrates that presynaptic glutamate release contributes to long-term memory in Caenorhabditis elegans. Learn Memory 9:130-137. Rossion B (2014) Understanding individual face discrimination by means of fast periodic visual stimulation. Experimental brain research Experimentelle Hirnforschung Experimentation cerebrale 232:1599-1621. Ruan JM, Hu K, Zhang HX, Wang Y, Zhou AL, Zhao YN, Yang XL (2014) Distribution and quantitative detection of GABA(A) receptor in Carassius auratus gibelio. Fish Physiol Biochem 40:1301-1311. Rubinov M, Sporns O (2010) Complex network measures of brain connectivity: Uses and interpretations. NeuroImage 52:1059-1069. Ruhl T, Zeymer M, von der Emde G (2017) Cannabinoid modulation of zebrafish fear learning and its functional analysis investigated by c-Fos expression. Pharmacol Biochem Be 153:18-31. Russell EM (1967) Changes in Behaviour of Lebistes Reticulatus Upon a Repeated Shadow Stimulus. Animal Behaviour 15:574-&. Ruusuvuori E, Kaila K (2014) Carbonic Anhydrases and Brain pH in the Control of Neuronal Excitability. In: Carbonic Anhydrase: Mechanism, Regulation, Links to Disease, and Industrial Applications (Frost SC, McKenna R, eds), pp 271-290. Dordrecht: Springer Netherlands. Saeki T, Sogawa S, Hotta T, Kohda M (2018) Territorial fish distinguish familiar neighbours individually. Behaviour 155:279-293. Saggar M, Hosseini SM, Bruno JL, Quintin EM, Raman MM, Kesler SR, Reiss AL (2015) Estimating individual contribution from group-based structural correlation networks. NeuroImage 120:274-284. Sagi D (2011) Perceptual learning in Vision Research. Vision research 51:1552-1566. Saidel WM, Marquez-Houston K, Butler AB (2001) Identification of visual pallial telencephalon in the goldfish, Carassius auratus: a combined cytochrome oxidase and electrophysiological study. Brain research 919:82-93. Saint-Amant L, Drapeau P (1998) Time course of the development of motor behaviors in the zebrafish embryo. J Neurobiol 37:622-632. Salas C, Broglio C, Duran E, Gomez A, Ocana FM, Jimenez-Moya F, Rodriguez F (2006) Neuropsychology of Learning and Memory in Teleost Fish. Zebrafish 3:157-171. Saldarriaga W, Tassone F, Gonzalez-Teshima LY, Forero-Forero JV, Ayala-Zapata S, Hagerman R (2014) Fragile X Syndrome. Colomb Medica 45:190-198. Salva OR, Sovrano VA, Vallortigara G (2014) What can fish brains tell us about visual perception? Frontiers in neural circuits 8. Samson F, Mottron L, Soulieres I, Zeffiro TA (2012) Enhanced visual functioning in autism: An ALE meta-analysis. Hum Brain Mapp 33:1553-1581. Sanes JR, Zipursky SL (2010) Design principles of insect and vertebrate visual systems. Neuron 66:15-36. Sankrithi NS, O'Malley DM (2010) Activation of a Multisensory, Multifunctional Nucleus in the Zebrafish Midbrain during Diverse Locomotor Behaviors. Neuroscience 166:970-993. Santos AR, Kanellopoulos AK, Bagni C (2014) Learning and behavioral deficits associated with the absence of the fragile X mental retardation protein: what a fly and mouse model can teach us. Learn Memory 21:543-555. Sato T, Hamaoka T, Aizawa H, Hosoya T, Okamoto H (2007) Genetic single-cell mosaic analysis implicates ephrinB2 reverse signaling in projections from the posterior tectum to the hindbrain in zebrafish. Journal of Neuroscience 27:5271-5279. 175

Satou C, Kimura Y, Hirata H, Suster ML, Kawakami K, Higashijima S (2013) Transgenic tools to characterize neuronal properties of discrete populations of zebrafish neurons. Development 140:3927-3931. Saverino C, Gerlai R (2008) The social zebrafish: Behavioral responses to conspecific, heterospecific, and computer animated fish. Behavioural brain research 191:77-87. Schiff W, Caviness JA, Gibson JJ (1962) Persistent Fear Responses in Rhesus Monkeys to Optical Stimulus of Looming. Science 136:982-&. Schluessel V, Duengen D (2015) Irrespective of size, scales, color or body shape, all fish are just fish: object categorization in the gray bamboo shark Chiloscyllium griseum. Animal cognition 18:497-507. Schluessel V, Fricke G, Bleckmann H (2012) Visual discrimination and object categorization in the cichlid Pseudotropheus sp. Animal cognition 15:525-537. Schluessel V, Kraniotakes H, Bleckmann H (2014a) Visual discrimination of rotated 3D objects in Malawi cichlids (Pseudotropheus sp.): a first indication for form constancy in fishes. Animal cognition 17:359-371. Schluessel V, Beil O, Weber T, Bleckmann H (2014b) Symmetry perception in bamboo sharks (Chiloscyllium griseum) and Malawi cichlids (Pseudotropheus sp.). Animal cognition 17:1187-1205. Schneider A, Leigh MJ, Adams P, Nanakul R, Chechi T, Olichney J, Hagerman R, Hessl D (2013) Electrocortical changes associated with minocycline treatment in fragile X syndrome. J Psychopharmacol 27:956-963. Schunter C, Welch MJ, Ryu TW, Zhang HM, Berumen ML, Nilsson GE, Munday PL, Ravasi T (2016) Molecular signatures of transgenerational response to ocean acidification in a species of reef fish. Nat Clim Change 6:1014-1018. Schuster S, Machnik P, Schulze W (2011a) VISION | Behavioral Assessment of the Visual Capabilities of Fish. In: Encyclopedia of Fish Physiology (Farrell AP, ed), pp 143-149. San Diego: Academic Press. Schuster S, Machnik P, Schulze W (2011b) Behavioral Assessment of the Visual Capabilities of Fish. Encyclopedia of Fish Physiology: From Genome to Environment, Vols 1-3:143-149. Scott EK, Baier H (2009) The cellular architecture of the larval zebrafish tectum, as revealed by Gal4 enhancer trap lines. Frontiers in neural circuits 3. Scott EK, Mason L, Arrenberg AB, Ziv L, Gosse NJ, Xiao T, Chi NC, Asakawa K, Kawakami K, Baier H (2007) Targeting neural circuitry in zebrafish using GAL4 enhancer trapping. Nat Methods 4:323-326. Seger CA (2009) The Involvement of Corticostriatal Loops in Learning Across Tasks, Species, and Methodologies. Adv Behav Biol 58:25-39. Seger CA, Miller EK (2010) Category learning in the brain. Annual review of neuroscience 33:203- 219. Semmelhack JL, Donovan JC, Thiele TR, Kuehn E, Laurell E, Baier H (2014) A dedicated visual pathway for prey detection in larval zebrafish. Elife 3. Severi KE, Portugues R, Marques JC, O'Malley DM, Orger MB, Engert F (2014) Neural Control and Modulation of Swimming Speed in the Larval Zebrafish. Neuron 83:692-707. Shamay-Ramot A, Khermesh K, Porath HT, Barak M, Pinto Y, Wachtel C, Zilberberg A, Lerer- Goldshtein T, Efroni S, Levanon EY, Appelbaum L (2015) Fmrp Interacts with Adar and Regulates RNA Editing, Synaptic Density and Locomotor Activity in Zebrafish. Plos Genet 11. Shang CP, Chen ZJ, Liu AX, Li Y, Zhang JJ, Qu BL, Yan F, Zhang YN, Liu WX, Liu ZH, Guo XF, Li DP, Wang Y, Cao P (2018) Divergent midbrain circuits orchestrate escape and freezing responses to looming stimuli in mice. Nat Commun 9. Shettleworth SJ (2001) Animal cognition and animal behaviour. Animal Behaviour 61:277-286.

176

Siebeck UE (2004) Communication in coral reef fish: the role of ultraviolet colour patterns in damselfish territorial behaviour. Animal Behaviour 68:273-282. Siebeck UE, Marshall NJ (2001) Ocular media transmission of coral reef fish - can coral reef fish see ultraviolet light? Vision research 41:133-149. Siebeck UE, Wallis GM, Litherland L (2008) Colour vision in coral reef fish. The Journal of experimental biology 211:354-360. Siebeck UE, Litherland L, Wallis GM (2009) Shape learning and discrimination in reef fish. The Journal of experimental biology 212:2113-2119. Siebeck UE, Parker AN, Sprenger D, Mathger LM, Wallis G (2010) A species of reef fish that uses ultraviolet patterns for covert face recognition. Current biology : CB 20:407-410. Siebeck UE, Wallis GM, Litherland L, Ganeshina O, Vorobyev M (2014) Spectral and spatial selectivity of luminance vision in reef fish. Frontiers in neural circuits 8:118. Sillar KT, Picton L, Heitler WJ (2016) The neuroethology of predation and escape. Silvanto J (2015) Why is "blindsight" blind? A new perspective on primary visual cortex, recurrent activity and visual awareness. Conscious Cogn 32:15-32. Silveira MM, Oliveira JJ, Luchiari AC (2015) Dusky damselfish Stegastes fuscus relational learning: evidences from associative and spatial tasks. J Fish Biol 86:1109-1120. Silverman JL, Gastrell PT, Karras MN, Solomon M, Crawley JN (2015) Cognitive Abilities on Transitive Inference Using a Novel Touchscreen Technology for Mice. Cerebral cortex 25:1133-1142. Silverstein DN, Ingvar M (2015) A multi-pathway hypothesis for human visual fear signaling. Front Syst Neurosci 9. Simmich J, Staykov E, Scott E (2012) Zebrafish as an appealing model for optogenetic studies. Optogenetics: Tools for Controlling and Monitoring Neuronal Activity 196:145-162. Simpson EE, Marshall NJ, Cheney KL (2016) Coral reef fish perceive lightness illusions. Sci Rep 6:35335. Sinclair D, Oranje B, Razak KA, Siegel SJ, Schmid S (2017) Sensory processing in autism spectrum disorders and Fragile X syndrome-From the clinic to animal models. Neuroscience and biobehavioral reviews 76:235-253. Sison M, Gerlai R (2010) Associative learning in zebrafish (Danio rerio) in the plus maze. Behavioural brain research 207:99-104. Sison M, Cawker J, Buske C, Gerlai R (2006) Fishing for genes influencing vertebrate behavior: zebrafish making headway. Lab Animal 35:33-39. Soares SC, Maior RS, Isbell LA, Tomaz C, Nishijo H (2017) Fast Detector/First Responder: Interactions between the Superior Colliculus-Pulvinar Pathway and Stimuli Relevant to Primates. Frontiers in Neuroscience 11. Song JR, Ampatzis K, Ausborn J, El Manira A (2015) A Hardwired Circuit Supplemented with Endocannabinoids Encodes Behavioral Choice in Zebrafish. Current Biology 25:2610-2620. Soto E, Ortega-Ramirez A, Vega R (2018) Protons as Messengers of Intercellular Communication in the Nervous System. Front Cell Neurosci 12. South M, Rodgers J (2017) Sensory, Emotional and Cognitive Contributions to Anxiety in Autism Spectrum Disorders. Frontiers in human neuroscience 11. Sovrano VA, Bisazza A (2009) Perception of subjective contours in fish. Perception 38:579-590. Sovrano VA, Albertazzi L, Salva OR (2015) The Ebbinghaus illusion in a fish (Xenotoca eiseni). Animal cognition 18:533-542. Sowden PT, Davies IR, Roling P (2000) Perceptual learning of the detection of features in X-ray images: a functional role for improvements in adults' visual sensitivity? Journal of experimental psychology Human perception and performance 26:379-390. Sporns O (2018) Graph theory methods: applications in brain networks. Dialogues in clinical neuroscience 20:111-120.

177

Stewart A, Kadri F, DiLeo J, Min Chung K, Cachat J, Goodspeed J, Suciu C, Roy S, Gaikwad S, Wong K (2010) The developing utility of zebrafish in modeling neurobehavioral disorders. International Journal of Comparative Psychology 23. Stewart AM, Kalueff AV (2012) The Developing Utility of Zebrafish Models for Cognitive Enhancers Research. Curr Neuropharmacol 10:263-271. Stewart AM, Braubach O, Spitsbergen J, Gerlai R, Kalueffl AV (2014a) Zebrafish models for translational neuroscience research: from tank to bedside. Trends in Neurosciences 37:264- 278. Stewart AM, Nguyen M, Wong K, Poudel MK, Kalueff AV (2014b) Developing zebrafish models of autism spectrum disorder (ASD). Prog Neuro-Psychoph 50:27-36. Stewart AM, Ullmann JFP, Norton WHJ, Parker MO, Brennan CH, Gerlai R, Kalueff AV (2015) Molecular psychiatry of zebrafish. Mol Psychiatr 20:2-17. Stewart WJ, Cardenas GS, McHenry MJ (2013) Zebrafish larvae evade predators by sensing water flow. Journal of Experimental Biology 216:388-398. Storozhuk M, Kondratskaya E, Nikolaenko L, Krishtal O (2016) A modulatory role of ASICs on GABAergic synapses in rat hippocampal cell cultures. Mol Brain 9. Strausfeld NJ, Hirth F (2016) Introduction to 'Homology and convergence in nervous system evolution'. Philos T R Soc B 371. Striedter GF, Northcutt RG (1989) Two distinct visual pathways through the superficial pretectum in a percomorph teleost. The Journal of comparative neurology 283:342-354. Subramanian M, Timmerman CK, Schwartz JL, Pham DL, Meffert MK (2015) Characterizing autism spectrum disorders by key biochemical pathways. Frontiers in Neuroscience 9. Sumbre G, Muto A, Baier H, Poo MM (2008) Entrained rhythmic activities of neuronal ensembles as perceptual memory of time interval. Nature 456:102-U106. Sun HJ, Frost BJ (1998) Computation of different optical variables of looming objects in pigeon nucleus rotundus neurons. Nature neuroscience 1:296-303. Sundin J, Amcoff M, Mateos-Gonzalez F, Raby GD, Jutfelt F, Clark TD (2017) Long-term exposure to elevated carbon dioxide does not alter activity levels of a coral reef fish in response to predator chemical cues. Behav Ecol Sociobiol 71. Sutherland NS (1961) The methods and findings of experiments on the visual discrimination of shape by animals. Cambridge, Eng.,: Heffer. Sutherland ns (1969) Shape Discrimination in Rat Octopus and Goldfish - a Comparative Study. J Comp Physiol Psych 67:160-&. Swartz JR, Wiggins JL, Carrasco M, Lord C, Monk CS (2013) Amygdala Habituation and Prefrontal Functional Connectivity in Youth With Autism Spectrum Disorders. J Am Acad Child Psy 52:84-93. Tabor KM, Smith TS, Brown M, Bergeron SA, Briggman KL, Burgess HA (2018) Presynaptic Inhibition Selectively Gates Auditory Transmission to the Brainstem Startle Circuit. Current Biology 28:2527-+. Tabor KM, Marquart GD, Hurt C, Smith TS, Geoca AK, Bhandiwad AA, Subedi A, Sinclair JL, Rose HM, Polys NF, Burgess HA (2019) Brain-wide cellular resolution imaging of Cre transgenic zebrafish lines for functional circuit-mapping. Elife 8. Tam FI, King JA, Geisler D, Korb FM, Sareng J, Ritschel F, Steding J, Albertowski KU, Roessner V, Ehrlich S (2017) Altered behavioral and amygdala habituation in high-functioning adults with autism spectrum disorder: an fMRI study. Scientific Reports 7. Tamietto M, de Gelder B (2010) Neural bases of the non-conscious perception of emotional signals. Nat Rev Neurosci 11:697-709. Tamietto M, Pullens P, de Gelder B, Weiskrantz L, Goebel R (2012) Subcortical Connections to Human Amygdala and Changes following Destruction of the Visual Cortex. Current Biology 22:1449-1455.

178

Tamietto M, Cauda F, Corazzini LL, Savazzi S, Marzi CA, Goebel R, Weiskrantz L, de Gelder B (2010) Collicular Vision Guides Nonconscious Behavior. Journal of Cognitive Neuroscience 22:888-902. Taylor MA, Vanwalleghem GC, Favre-Bulle IA, Scott EK (2018) Diffuse light-sheet microscopy for stripe-free calcium imaging of neural populations. J Biophotonics 11. Temizer I, Donovan JC, Baier H, Semmelhack JL (2015) A Visual Pathway for Looming-Evoked Escape in Larval Zebrafish. Current Biology 25:1823-1834. Temple SE, Manietta D, Collin SP (2013) A comparison of behavioural (Landolt C) and anatomical estimates of visual acuity in archerfish (Toxotes chatareus). Vision research 83:1-8. Terleph TA, Tremere LA (2006) The Use of Immediate Early Genes as Mapping Tools for Neuronal Activation: Concepts and Methods. In: Immediate Early Genes in Sensory Processing, Cognitive Performance and Neurological Disorders (Pinaud R, Tremere LA, eds), pp 1-10. Boston, MA: Springer US. Thiele TR, Donovan JC, Baier H (2014) Descending Control of Swim Posture by a Midbrain Nucleus in Zebrafish. Neuron 83:679-691. Thompson AW, Vanwalleghem GC, Heap LA, Scott EK (2016) Functional Profiles of Visual-, Auditory-, and Water Flow-Responsive Neurons in the Zebrafish Tectum. Current Biology 26:743-754. Thompson RF (2014) Habituation A History. Wiley Blackwell Handbook of Operant and Classical Conditioning:79-94. Thompson RF, Spencer WA (1966) Habituation - a Model Phenomenon for Study of Neuronal Substrates of Behavior. Psychol Rev 73:16-&. Tomsic D, Beron de Astrada M, Sztarker J (2003) Identification of individual neurons reflecting short- and long-term visual memory in an arthropodo. The Journal of neuroscience : the official journal of the Society for Neuroscience 23:8539-8546. Tresguerres M, Hamilton TJ (2017) Acid base physiology, neurobiology and behaviour in relation to CO2-induced ocean acidification. Journal of Experimental Biology 220:2136-2148. Triplett MA, Avitan L, Goodhill GJ (2018) Emergence of spontaneous assembly activity in developing neural networks without afferent input. Plos Comput Biol 14. Trivedi CA, Bollmann JH (2013) Visually driven chaining of elementary swim patterns into a goal- directed motor sequence: a virtual reality study of zebrafish prey capture. Frontiers in neural circuits 7. Truppa V, Sovrano VA, Spinozzi G, Bisazza A (2010) Processing of visual hierarchical stimuli by fish (Xenotoca eiseni). Behavioural brain research 207:51-60. Typlt M, Mirkowski M, Azzopardi E, Ruth P, Pilz PKD, Schmid S (2013) Habituation of reflexive and motivated behavior in mice with deficient BK channel function. Front Integr Neurosc 7. Uceda S, Ocana FM, Martin-Monzon I, Rodriguez-Exposito B, Duran E, Rodriguez F (2015) Spatial learning-related changes in metabolic brain activity contribute to the delimitation of the hippocampal pallium in goldfish. Behavioural brain research 292:403-408. Vaiana M, Muldoon SF (2018) Multilayer Brain Networks. Journal of Nonlinear Science. Valente A, Huang KH, Portugues R, Engert F (2012) Ontogeny of classical and operant learning behaviors in zebrafish. Learn Mem 19:170-177. Van-Eyk SM, Siebeck UE, Champ CM, Marshall J, Hart NS (2011) Behavioural evidence for colour vision in an elasmobranch. The Journal of experimental biology 214:4186-4192. van 't Padje S, Engels B, Blonden L, Severijnen LA, Verheijen F, Oostra B, Willemsen R (2005) Characterisation of Fmrp in zebrafish: evolutionary dynamics of the fmr1 gene. Dev Genes Evol 215:198-206. van der Maaten L, Hinton G (2008) Visualizing Data using t-SNE. J Mach Learn Res 9:2579-2605. Van der Molen MJW, Van der Molen MW, Ridderinkhof KR, Hamel BCJ, Curfs LMG, Ramakers GJA (2012) Auditory change detection in fragile X syndrome males: A brain potential study. Clin Neurophysiol 123:1309-1318. 179

Van Essen DC, Anderson CH, Felleman DJ (1992) Information processing in the primate visual system: an integrated systems perspective. Science 255:419-423. Vanegas H, Ito H (1983) Morphological aspects of the teleostean visual system: a review. Brain research 287:117-137. Vanwalleghem G, Heap LA, Scott EK (2017) A profile of auditory-responsive neurons in the larval zebrafish brain. Journal of Comparative Neurology 525:3031-3043. Vanwalleghem GC, Ahrens MB, Scott EK (2018) Integrative whole-brain neuroscience in larval zebrafish. Current opinion in neurobiology 50:136-145. Vargas JP, López JC, Portavella M (2009) What are the functions of fish brain pallium? Brain research bulletin 79:436-440. Vargas JP, Bingman VP, Portavella M, Lopez JC (2006) Telencephalon and geometric space in goldfish. European Journal of Neuroscience 24:2870-2878. Vaz R, Hofmeister W, Lindstrand A (2019) Zebrafish Models of Neurodevelopmental Disorders: Limitations and Benefits of Current Tools and Techniques. International journal of molecular sciences 20. Vievard A, Fabrethorpe M, Buser P (1986) Role of the Extra-Geniculate Pathway in Visual Guidance .1. Effects of Lesioning the Superior Colliculus in the Cat. Experimental Brain Research 62:587-595. Villani L, Guarnieri T (1995) Localization of Nadph-Diaphorase in the Goldfish Brain. Brain research 679:261-266. Vivanti G, Hocking DR, Fanning PAJ, Uljarevic M, Postorino V, Mazzone L, Dissanayake C (2018) Attention to novelty versus repetition: Contrasting habituation profiles in Autism and Williams syndrome. Dev Cogn Neuros-Neth 29:54-60. von Trotha JW, Vernier P, Bally-Cuif L (2014) Emotions and motivated behavior converge on an amygdala-like structure in the zebrafish. European Journal of Neuroscience 40:3302-3315. Vonbartheld CS, Meyer DL (1987) Comparative Neurology of the Optic Tectum in Ray-Finned Fishes - Patterns of Lamination Formed by Retinotectal Projections. Brain research 420:277- 288. Wan H, Warburton EC, Zhu XO, Koder TJ, Park Y, Aggleton JP, Cho K, Bashir ZI, Brown MW (2004) Benzodiazepine impairment of perirhinal cortical plasticity and recognition memory. European Journal of Neuroscience 20:2214-2224. Wang WC, McLean DL (2014) Selective Responses to Tonic Descending Commands by Temporal Summation in a Spinal Motor Pool. Neuron 83:708-721. Warburton EC, Brown MW (2015) Neural circuitry for rat recognition memory. Behavioural brain research 285:131-139. Watson CR, Paxinos G, Tokuno H (2010) Using a panel of immunomarkers to define homologies in Mammalian brains. Frontiers in human neuroscience 4:13. Webb SJ, Jones EJH, Merkle K, Namkung J, Toth K, Greenson J, Murias M, Dawson G (2010) Toddlers with Elevated Autism Symptoms Show Slowed Habituation to Faces. Child Neuropsychol 16:255-278. Wei PF, Liu N, Zhang ZJ, Liu XM, Tang YQ, He XB, Wu BF, Zhou Z, Liu YH, Li J, Zhang Y, Zhou XY, Xu L, Chen L, Bi GQ, Hu XT, Xu FQ, Wang LP (2015) Processing of visually evoked innate fear by a non-canonical thalamic pathway (vol 6, 6756, 2015). Nat Commun 6. Welch MJ, Watson SA, Welsh JQ, McCormick MI, Munday PL (2014) Effects of elevated CO2 on fish behaviour undiminished by transgenerational acclimation. Nat Clim Change 4:1086- 1089. Wemmie JA, Zha X-m, Welsh MJ (2008) Acid-Sensing Ion Channels (ASICs) and pH in Synapse Physiology. In: Structural And Functional Organization Of The Synapse (Hell JW, Ehlers MD, eds), pp 661-681. Boston, MA: Springer US.

180

White BJ, Kan JY, Levy R, Itti L, Munoz DP (2017) Superior colliculus encodes visual saliency before the primary visual cortex. Proceedings of the National Academy of Sciences of the United States of America 114:9451-9456. Wicks SR, Rankin CH (1997) Effects of tap withdrawal response habituation on other withdrawal behaviors: The localization of habituation in the nematode Caenorhabditis elegans. Behavioral neuroscience 111:342-353. Wiggins JL, Swartz JR, Martin DM, Lord C, Monk CS (2014) Serotonin transporter genotype impacts amygdala habituation in youth with autism spectrum disorders. Soc Cogn Affect Neur 9:832- 838. Wilkins ME, Hosie AM, Smart TG (2005) Proton modulation of recombinant GABA(A) receptors: influence of GABA concentration and the beta subunit TM2-TM3 domain. J Physiol-London 567:365-377. Wolf S, Supatto W, Debregeas G, Mahou P, Kruglik SG, Sintes JM, Beaurepaire E, Candelier R (2015) Whole-brain functional imaging with two-photon light-sheet microscopy. Nat Methods 12:379-380. Wolman M, Granato M (2012) Behavioral genetics in larval zebrafish: Learning from the young. Developmental neurobiology 72:366-372. Wolman MA, Jain RA, Liss L, Granato M (2011) Chemical modulation of memory formation in larval zebrafish. Proceedings of the National Academy of Sciences of the United States of America 108:15468-15473. Wolman MA, de Groh ED, McBride SM, Jongens TA, Granato M, Epstein JA (2014) Modulation of cAMP and Ras Signaling Pathways Improves Distinct Behavioral Deficits in a Zebrafish Model of Neurofibromatosis Type 1. Cell Reports 8:1265-1270. Wolman MA, Jain RA, Marsden KC, Bell H, Skinner J, Hayer KE, Hogenesch JB, Granato M (2015) A Genome-wide Screen Identifies PAPP-AA-Mediated IGFR Signaling as a Novel Regulator of Habituation Learning (vol 85, pg 1200, 2015). Neuron 87:906-907. Wood LS, Desjardins JK, Fernald RD (2011) Effects of stress and motivation on performing a spatial task. Neurobiology of learning and memory 95:277-285. Wu YJ, Hsu MT, Ng MC, Amstislavskaya TG, Tikhonova MA, Yang YL, Lu KT (2017) Fragile X Mental Retardation-1 Knockout Zebrafish Shows Precocious Development in Social Behavior. Zebrafish 14:438-443. Wullimann MF (1997a) The central nervous system. The physiology of fishes:245-282. Wullimann MF (1997b) Major patterns of visual brain organization in teleosts and their relation to prehistoric events and the paleontological record. Paleobiology 23:101-114. Wullimann MF, Northcutt RG (1988) Connections of the corpus cerebelli in the green sunfish and the common goldfish: a comparison of perciform and cypriniform teleosts. Brain, behavior and evolution 32:293-316. Wullimann MF, Mueller T (2004) Teleostean and mammalian forebrains contrasted: Evidence from genes to behavior (vol 475, pg 143, 2004). Journal of Comparative Neurology 478:427-428. Wullimann MF, Rupp B, Reichert H (1996) The brain of the zebrafish Danio rerio: a neuroanatomical atlas. In: Neuroanatomy of the Zebrafish Brain: A Topological Atlas, pp 19-87. Basel: Birkhäuser Basel. Wyzisk K, Neumeyer C (2007) Perception of illusory surfaces and contours in goldfish. Visual neuroscience 24:291-298. Yaeger C, Ros AM, Cross V, Deangelis RS, Stobaugh DJ, Rhodes JS (2014) BLOCKADE OF ARGININE VASOTOCIN SIGNALING REDUCES AGGRESSIVE BEHAVIOR AND c- FOS EXPRESSION IN THE PREOPTIC AREA AND PERIVENTRICULAR NUCLEUS OF THE POSTERIOR TUBERCULUM IN MALE AMPHIPRION OCELLARIS. Neuroscience 267:205-218. Yager D (1967) Behavioral Measures and Theoretical Analysis of Spectral Sensitivity and Spectral Saturation in Goldfish Carassius Auratus. Vision research 7:707-&. 181

Yamamoto K, Bloch S, Vernier P (2017) New perspective on the regionalization of the anterior forebrain in Osteichthyes. Development, growth & differentiation 59:175-187. Yamamoto N (2009) Studies on the teleost brain morphology in search of the origin of cognition. Japanese Psychological Research 51:154-167. Yamamoto N, Ito H (2005) Fiber connections of the anterior preglomerular nucleus in cyprinids with notes on telencephalic connections of the preglomerular complex. Journal of Comparative Neurology 491:212-233. Yamamoto N, Ito H (2008) Visual, lateral line, and auditory ascending pathways to the dorsal telencephalic area through the rostrolateral region of the lateral preglomerular nucleus in cyprinids. The Journal of comparative neurology 508:615-647. Yamamoto N, Ishikawa Y, Yoshimoto M, Xue HG, Bahaxar N, Sawai N, Yang CY, Ozawa H, Ito H (2007) A new interpretation on the homology of the teleostean telencephalon based on hodology and a new eversion model. Brain, behavior and evolution 69:96-104. Yamashita M, Yoshihara Y, Hashimoto R, Yahata N, Ichikawa N, Sakai Y, Yamada T, Matsukawa N, Okada G, Tanaka SC, Kasai K, Kato N, Okamoto Y, Seymour B, Takahashi H, Kawato M, Imamizu H (2018) A prediction model of working memory across health and psychiatric disease using whole-brain functional connectivity. Elife 7. Yanez J, Suarez T, Quelle A, Folgueira M, Anadon R (2018) Neural connections of the pretectum in zebrafish (Danio rerio). Journal of Comparative Neurology 526:1017-1040. Yao YY, Li XQ, Zhang BB, Yin C, Liu YF, Chen WY, Zeng SQ, Du JL (2016) Visual Cue- Discriminative Dopaminergic Control of Visuomotor Transformation and Behavior Selection. Neuron 89:598-612. Yashina K, Tejero-Cantero A, Herz A, Baier H (2019) Zebrafish Exploit Visual Cues and Geometric Relationships to Form a Spatial Memory. iScience 19:119-134. Yilmaz M, Meister M (2013) Rapid Innate Defensive Responses of Mice to Looming Visual Stimuli. Current Biology 23:2011-2015. Yin C, Li X, Du J (2018) Optic tectal superficial interneurons detect motion in larval zebrafish. Protein & cell. Yizhar O, Fenno LE, Davidson TJ, Mogri M, Deisseroth K (2011) Optogenetics in neural systems. Neuron 71:9-34. Yuan TF, Sus HX (2015) Fear learning through the two visual systems, a commentary on: "A parvalbumin-positive excitatory visual pathway to trigger fear responses in mice". Frontiers in neural circuits 9. Zackon DH, Casson EJ, Stelmach L, Faubert J, Racette L (1997) Distinguishing subcortical and cortical influences in visual attention - Subcortical attentional processing. Invest Ophth Vis Sci 38:364-371. Zalesky A, Fornito A, Bullmore ET (2012) On the use of correlation as a measure of network connectivity. NeuroImage 60:2096-2106. Zaman T, De Oliveira C, Smoka M, Narla C, Poulter MO, Schmid S (2017) BK Channels Mediate Synaptic Plasticity Underlying Habituation in Rats. Journal of Neuroscience 37:4540-4551. Zatka-Haas P, Steinmetz NA, Carandini M, Harris KD (2019) Distinct contributions of mouse cortical areas to visual discrimination. bioRxiv:501627. Zerbi V, Ielacqua GD, Markicevic M, Haberl MG, Ellisman MH, A-Bhaskaran A, Frick A, Rudin M, Wenderoth N (2018) Dysfunctional Autism Risk Genes Cause Circuit-Specific Connectivity Deficits With Distinct Developmental Trajectories. Cerebral cortex 28:2495-2506. Zhang B-b, Yao Y-y, Zhang H-f, Kawakami K, Du J-l (2017) Left Habenula Mediates Light- Preference Behavior in Zebrafish via an Asymmetrical Visual Pathway. Neuron 93:914- 928.e914. Zhang GR, Cao HY, Kong LX, O'Brien J, Baughns A, Jan M, Zhao H, Wang XD, Lu XG, Cook RG, Geller AI (2010) Identified circuit in rat postrhinal cortex encodes essential information for

182

performing specific visual shape discriminations. Proceedings of the National Academy of Sciences of the United States of America 107:14478-14483. Zhang P, Zhou H, Wen W, He S (2015) Layer-specific response properties of the human lateral geniculate nucleus and superior colliculus. NeuroImage 111:159-166. Zhao DB, Ning NN, Lei Z, Sun H, Wei CF, Chen DW, Li JX (2014a) Identification of a Novel Protein Complex Containing ASIC1a and GABA(A) Receptors and Their Interregulation. PloS one 9. Zhao XY, Liu MN, Cang JH (2014b) Visual Cortex Modulates the Magnitude but Not the Selectivity of Looming-Evoked Responses in the Superior Colliculus of Awake Mice. Neuron 84:202- 213. Zhaoping L (2016) From the optic tectum to the primary visual cortex: migration through evolution of the saliency map for exogenous attentional guidance. Current opinion in neurobiology 40:94-102. Zhou Z, Liu X, Chen S, Zhang Z, Liu Y, Montardy Q, Tang Y, Wei P, Liu N, Li L, Song R, Lai J, He X, Chen C, Bi G, Feng G, Xu F, Wang L (2019) A VTA GABAergic Neural Circuit Mediates Visually Evoked Innate Defensive Responses. Neuron. Zimmermann FF, Gaspary KV, Leite CE, Cognato GD, Bonan CD (2015) Embryological exposure to valproic acid induces social interaction deficits in zebrafish (Danio rerio): A developmental behavior analysis. Neurotoxicol Teratol 52:36-41. Zottoli SJ (1977) Correlation of the startle reflex and Mauthner cell auditory responses in unrestrained goldfish. The Journal of experimental biology 66:243-254. Zottoli SJ, Hordes AR, Faber DS (1987) Localization of Optic Tectal Input to the Ventral Dendrite of the Goldfish Mauthner Cell. Brain research 401:113-121.

183

Appendices

Appendix 1: Ethics approvals.

184

185

186

187

Appendix 2: Manuscript- Visual Escape in Larval Zebrafish: Stimuli, Circuits, and Behavior.

Visual Escape in Larval Zebrafish: Stimuli, Circuits, and Behavior

Emmanuel Marquez-Legorreta1, Marielle Piber1,2, and Ethan K. Scott1,3

1 School of Biomedical Sciences

The University of Queensland

Brisbane, QLD 4072

Australia

2 School of Medicine, Medical Sciences & Nutrition

University of Aberdeen

Aberdeen, AB25 2ZD

United Kingdom

3 The Queensland Brain Institute

The University of Queensland

Brisbane, QLD 4072

Australia

Keywords: zebrafish, circuits, sensory processing, vision, startle, loom, escape behavior, GCaMP, optogenetics

188

Abstract

Visual escape behavior is important for survival, and elements of this behavior are conserved from insects to humans. Because it needs to be robust and rapid, but also open to modulation, it is an excellent system in which to study visual processing and sensorimotor gating. Recent studies, especially in the transparent larvae of the zebrafish model system, have begun to shed light on the intricacies of visual escape circuitry, and in this chapter, we will review this progress. First, we will explore the essential properties of loom stimuli, including their movement, edges, and luminance changes, and will discuss how these stimulus properties, alone or in combination, can contribute to eliciting startle behavior. Next, we will describe the escape behavior itself, including the sequence of kinematic events that carries the animal away from the perceived threat and the various forms that this behavior can take depending on the stimulus and context. We will then provide an in-depth review of the core circuitry that lies between the stimulus and response, beginning with the retinal and thalamic projections that carry loom relevant information to the tectum. We will also discuss how this information is likely to be processed in the tectum and the visuomotor projections to premotor cells in the hindbrain, including the well-known Mauthner neurons. Finally, we will describe ways in which context, such as alertness or hunger, can alter an animal’s responses to threatening visual stimuli and the ways in which specific brain regions may detect these conditions and impinge on the core escape circuit to modulate behavior. We will conclude with perspectives on the important outstanding questions about visual escape circuits and specific experiments that might help in addressing them.

Introduction

The ability to detect an incoming predator and avoid its attack is a quintessential example of an evolutionarily selected behavior, as the consequence of failure is death. Because of this, most animals perform behavioral responses shaped though evolution to increase the likelihood of escaping an imminent threat. Animals use a range of sensory modalities to detect and respond to threats, but vision provides a particularly rich set of information on a predator’s size, shape, position, and movement. It is unsurprising, therefore, that visual escape behavior is widespread across the animal kingdom (Cooper and Blumstein, 2015; Sillar et al., 2016). This has inspired studies in systems as disparate as insects, crustaceans, fish, amphibians, reptiles, birds, and mammals (including humans), into the relationships that exist between visual stimuli and escape responses, and the neural circuits that

189

perceive, analyze, and trigger responses to the stimuli (Schiff et al., 1962; Hayes and Saiff, 1967; Ewert, 1970; Sun and Frost, 1998; Carlile et al., 2006; Preuss et al., 2006; de Vries and Clandinin, 2012; Oliva and Tomsic, 2012; Landwehr et al., 2013; Yilmaz and Meister, 2013; De Franceschi et al., 2016; Fink et al., 2019).

Often an animal will respond to a visual threat with an explosive movement in the opposite direction, but responses also need to be adaptable to different situations and conditions. As a result, although the escape response is a strong innate behavior, it can be modulated to suit a particular animal’s past experience or circumstances. This modulation can take the form of favoring a particular evasive response, tuning the kinematic speed and strength of that response, or indeed deciding not to respond at all. In order to understand both the core circuitry necessary for visual escape and the broader networks that modulate responses, we have to study the relationships that exist among various threatening visual stimuli, a range of possible output behaviors, and the brain activity that lies between.

In recent years, larval zebrafish have emerged as a useful model in which to study these questions. Zebrafish embryos show evasive behaviors as early as 1 day postfertilization (dpf) in response to touch stimuli (Eaton and Farley, 1973; Saint-Amant and Drapeau, 1998; Downes and Granato, 2006), and larvae execute ballistic visual escapes from looming stimuli as early as 4dpf (Yao et al., 2016). Zebrafish are externally fertilized, meaning that animals are readily observable at these early developmental stages. Importantly, they are also optically transparent as larvae, meaning that internal structures, including the brain, can be observed though microscopes without the need to dissect, cannulate, or otherwise interfere with the larva’s development or neural activity (Simmich et al., 2012).

These biological attributes provide a nearly ideal platform for recent advances in genetically encoded fluorescent indicators of neural activity (Broussard et al., 2014; Lin and Schnitzer, 2016), in 2-photon and selective plane illumination microscopy (SPIM) (Ahrens et al., 2013; Wolf et al., 2015), and in quantitative approaches for detecting and modelling brain-wide patterns of neural activity from large imaging datasets (Chen et al., 2018). Combined, these approaches have made it possible to image activity throughout the larval zebrafish brain at cellular resolution, and to make sense of the collective activity of the tens of thousands of neurons composing it (Vanwalleghem et al., 2018). These capabilities have recently been extended to free swimming larval zebrafish (Cong et al., 2017; Kim et al., 2017) and can be also integrated with powerful tools for functionally testing the relevant circuitry (using optogenetics or ablations) and describing the morphology of the neurons involved (using single cell labelling or immunofluorescence) (Leung et al., 2013; Feierstein et al., 2015; Dunn 190

et al., 2016b; dal Maschio et al., 2017; Lovett-Barron et al., 2017; Helmbrecht et al., 2018; Tabor et al., 2018).

In summary, these new approaches, combined with their biological attributes, make zebrafish larvae uniquely suited to the study of the circuitry that underlies visual escape. In this chapter, we will focus on zebrafish larvae as we describe behavioral responses to threatening visual stimuli and the neural circuits that drive these behaviors. This will include descriptions of the types of stimuli and stimulus features that can trigger visual escape, along with the various escape maneuvers that can result. We will focus mostly, however, on recent discoveries about the brain regions, local networks, and cellular microcircuits that larval zebrafish use to detect and respond to visual threats. Finally, we will suggest some of the future research needed to better understand the circuitry and the mechanisms of the visual escape response in zebrafish.

The essential properties of loom stimuli (What causes an animal to escape?)

Zebrafish, like most animals, have a powerful instinct for avoiding oncoming predators, and these responses have been explored extensively in experiments using live predators, animated predators presented on screens, and robotic simulated predators (Bass and Gerlai, 2008; Saverino and Gerlai, 2008; Gerlai et al., 2009; Colwill and Creton, 2011; Ahmed et al., 2012; Luca and Gerlai, 2012; Ladu et al., 2015). In general, zebrafish show stronger evasive responses to sympatric predators (those that the zebrafish might encounter in their natural habitat), than to allopatric predators or nonpredatory fish, suggesting some innate selection for relevant visual threats that allows adaptive behavior at the first encounter (Bass and Gerlai, 2008; Ahmed et al., 2012). This makes sense from an evolutionary standpoint, but does little to reveal the specific visual properties contained in these stimuli that drive escape behavior.

Addressing this requires a reductionist approach in which stimulus properties can be presented in isolation or controlled combinations to gauge their efficacy in driving behavior. For example, some simple and unnatural stimuli, such as a moving rectangle, elicit strong responses (Ahmed et al., 2012). Other studies exposing the zebrafish to foreign objects that have a set of eyes drawn onto them show that various other simple shapes can induce evasive responses (Blaser and Gerlai, 2006). The presentation of silhouettes of predators, including birds (Dill, 1974), powerfully elicits responses in zebrafish. Interestingly, responses to these predatory silhouettes were stronger than those to actual predators, suggesting that they not only conserve, but may accentuate the properties that zebrafish find salient (Luca and Gerlai, 2012). These properties may be quite simple, as further experiments

191

reducing the complexity of the stimuli to simple growing dots proved that these, too, elicited escape behavior (Luca and Gerlai, 2012).

This simple stimulus, a dark expanding circle on a lighter background (hereafter referred to as a “loom”), has become the stimulus of choice for studying startle circuitry and behavior in model systems such as larval zebrafish. Even this simple stimulus, however, contains multiple possible cues for a larva to process when deciding whether to execute an escape. One is a drop in luminance that occurs as the black dot occludes an increasing portion of the visual field as it “approaches” the zebrafish. It also contains edge information, represented as the moving boundary line between the growing dark circle and its lighter background (see Figure 1A-C). The specific contribution that each cue (dimming and moving edges) makes has been the topic of considerable study, but their exact roles in driving escape behavior are still not fully understood. Studies in mice, for instance, show that light looms on dark backgrounds do not elicit startles, nor do dimming stimuli that lack moving edges (Yilmaz and Meister, 2013), suggesting that both cues must be present in order for startles to occur in mice. Convergent results have been found in larval zebrafish in a study using both receding and approaching stimuli in both light and dark configurations, along with dimming stimuli (Temizer et al., 2015). In these experiments, dark looms elicited escapes in more than 90% of trials while the light looms on dark background triggered startles only 20-30% of the time. Receding stimuli had an even weaker effect below 10%. To test whether the change in luminance alone was sufficient to cause escape, the full-sized disk was dimmed with similar temporal dynamics to the dimming that occurs during loom. The dimming stimulus failed to cause startles, suggesting that luminance information alone is insufficient for driving escape behavior. The unique contributions of moving edges can be explored using an expanding checkerboard stimulus on an isoluminant grey background. This delivers movement similar to that in the loom stimulus, but without a dimming component. Using this approach, Dunn et al (Dunn et al., 2016a) found that checkerboard stimuli drove behavior at a similar rate to looms, while Heap et al (Heap et al., 2018b) found a much lower rate of escape to checkerboard stimuli than to looms. Differences in the details of the checkerboard stimuli likely account for these different results, as discussed in Heap et al (Heap et al., 2018b). The interpretations of the combined results from these studies are that moving edges appear to be necessary for eliciting escape behavior, but that the direction of the luminance change (brightening or dimming) is important, and, indeed, dimming may make a major contribution to the likelihood of escape behavior when moving edges are present.

The temporal and spatial properties of looming stimuli also have implications for the probability and timing of the ensuing escape response. Researchers across model systems use a common language to

192

describe the mathematical properties of these stimuli. The expanding loom projects a dark disk on the zebrafish’s retina at a distance (d), with an angular size (θ) at the retina that increases with time (Figure 1D). Trigonometrically, (θ) is twice the inverse tangent of half the object size (l). The relationship between (l) and the object’s approach velocity (v), expressed as the (l/v) ratio, dictates the rate of change for the retinal size of the object. Varying this (l/v) ratio allows investigators to manipulate the temporal dynamics of the looming stimulus to study which of its varying aspects are used by an animal to guide its behavior.

For a simulated predator approaching at a constant rate, the image on the retina grows at an ever- increasing speed (Figure 1E). When detected monocularly, the growth of (θ), and the associated changes in its rate of expansion (ρ)(Figure 1F), characterize the approach, including the “time to collision” (τ)(Figure 1G) (Hatsopoulos et al., 1995). This, in theory, presents the animal with three key optical parameters, the object’s size, velocity, and rate of expansion, with which it could detect the threat. Naturally, this has led to a search of the neurons encoding these parameters in the brain.

(2015) presented looms horizontally to head-embedded larvae, Bhattacharyya et al (2017) presented a range of more natural stimuli horizontally to free swimming larvae, and Dunn et al (2016a) presented loom stimuli from below to free-swimming larvae. The studies converge on the interpretation that there is an angular size threshold for triggering escape behavior in larval zebrafish, but their differences serve as a reminder that numerous factors related to the stimulus (size, shape, growth rate, and direction) and the animal’s circumstances (head embedded versus free-swimming, in this case) impact on the animal’s behavioral decisions. This reinforces the fundamental properties of visual escape: it is a powerful and innate behavior that nonetheless remains subject to variability from an unpredictable world, and to modulation from the animal’s internal state.

[Insert Figure 1 here]

Figure 1. Components and computations of the loom. A-C: representations of the loom (A), checkerboard (B), and dim (C) stimuli, from the onset of the stimulus (left) to the final frame (right). D: Schematic representation of the eye of an observer at the visual angle θ, subtended by an object with a half size l. The distance d will decrease as a function of time with approach of the object at the velocity v. If the approach velocity is constant, the angle θ(t) subtended by the object and its rate of increase [ρ(t) = θ'(t)] both grow non-linearly as a function of time. E-G: Evolution of the kinematic parameters of the looming stimulus during approach. θ (t) is the loom angular size, ρ (t) is the loom angular velocity and τ (t) is the time-to-collision. Adapted from Peek & Card (2016)

Execution of escape behaviors

A threatening visual stimulus can lead to any of several behaviors. In the case of the fully developed adult zebrafish, these may include startle, freezing, and a range of avoidance behaviors (thrashing, zig-zagging, leaping and diving (Luca and Gerlai, 2012; Bishop et al., 2016)) that are, presumably, selected based on the nature of the predator and the animal’s surroundings. For example, thrashing is usually performed in shallow water and is followed by freezing, a sequence that is thought to stir up debris and camouflage the fish from the potential predator (Bass and Gerlai, 2008; Luca and Gerlai, 2012). Similarly, the selection of a particular behavior may be influenced by the animal’s available cover (such as shadows or vegetation)(Hein et al., 2018). This is an interesting and nuanced space, but is only peripherally related to the startle behavior seen in larval zebrafish, so we will not focus on it here.

The most common escape behavior across fish species, including in larval zebrafish, is a C-start that can be triggered by touch (Fetcho, 1991), vibrations including auditory cues (Burgess and Granato, 194

2007), water flow (Stewart et al., 2013), and vision (Eaton et al., 1977). The escape latency is very short, with responses beginning 5-10 msec after stimulus onset in adult zebrafish (Eaton et al., 1977). It involves a fast, powerful activation of head and body muscles, unilaterally contracting the axial muscles in the trunk and tail, thus bending the body into a C shape (Eaton et al., 1977). This maneuver can be sectioned into three stages, whereby the initial 15-20 msec are stereotyped (Eaton et al., 1977). The extensive contraction described above (Stage 1) usually occurs after 5-8 msec, and is followed by a further contraction on the opposite side of the body (Stage 2), during which the tail straightens. These two stages lead to rapid propulsion of the fish away from the predator, and are generally complete after approximately 20 msec. Following this, the fish may show any of a number of behaviors including, gliding, braking, or continued swimming (Eaton et al., 1977). The response in larvae follows the same motion sequence as in the adult, but there are some quantitative differences in terms of the speed, kinematics, and time course of the behavior (Kimmel et al., 1974; Nair et al., 2015). Larvae have a slightly slower response latency of under 15 msec, and reach full contraction more slowly than adults, yet head displacement speed relative to body length is three times greater (a function of their having a deeper C-bend) (Eaton et al., 1977).

The circuitry involved in the detection and processing of visual threats will be discussed in a later section, but we will briefly discuss the motor circuits relevant to startle here. Short latency C-starts are mediated by the Mauthner cells (M-cells), a pair of bilateral reticulospinal neurons that innervate motor neurons throughout the spine. A single action potential in one M-cell initiates a C-bend on the opposite side of the body (Zottoli, 1977; Eaton et al., 1982). However, escape behaviors also occur if, instead of activating the M-cells, M-cell homologues are activated (Kimmel et al., 1980; Liu and Fetcho, 1999; Kohashi and Oda, 2008). These responses, the long-latency C-starts, take longer to initiate, with an average escape latency of 28.3 msec in larvae (Liu and Fetcho, 1999; Budick and O'Malley, 2000; Burgess and Granato, 2007). They tend to result from slowly approaching loom stimuli, and unlike short-latency C starts, involve the usage of the pectoral fins, which may permit the fine-tuning of the escape movement (McClenahan et al., 2012; Bhattacharyya et al., 2017). By contrast, faster loom approach rates typically evoke short latencies C-starts (Bhattacharyya et al., 2017). The short- and long-latency C-starts, therefore, provide different behavioral options depending on the urgency of the threat. M-cell driven short latency C-starts provide a rapid and dramatic response to an immediate threat, sacrificing adjustment and control of speed and power. Long latency C-starts are slower, variable, and kinematically more complex, likely permitting control over the speed or direction of the escape.

195

These different attributes of short- and long-latency C-starts, and their links to the nature of the threat facing the animal, occupy a small part of the fish’s broader behavioral repertoire and environmental context. Feldman Barrett and Finlay (2018) suggest that defensive behaviors are context dependent, purposeful, flexible actions. Exclusive use of short latency C-bends would be energetically inefficient, and would make the escape behavior predictable for the predator. Thus, the ability to appropriately estimate the level of threat, and to have a range of possible responses, enhances the probability of survival (Nair et al., 2017). At longer durations, complex details of their environments, such as hiding places, can be integrated into their decisions, adding another level of complexity to the decision-making (Hein et al., 2018). Aspects of an animal’s internal state, such as motivation to feed, mate, or defend territory, could also influence its appetite for risk, and therefore, its responses to threats. A core escape circuit that provides a reliable, fast, and strong response to urgent threats, but that is modulated by other circuits responsible for encoding past experience, environmental context, and internal state, could provide a means for delivering the correct adaptive behavior in response to a range of threats and circumstances. In the next two sections, we will discuss the core escape circuit in larval zebrafish, and will then present evidence for the modulation of this circuit by a number of other brain regions.

Core visual escape circuitry

The first neurons to respond to visual stimuli are in the retina, and visual information is then sent through the Retinal Ganglion Cell (RGC) axons to multiple arborization fields (AFs) in the brain. A majority of RGCs extend to AF10, which is the neuropil of the tectum (Burrill and Easter, 1994; Robles et al., 2014) (red projections to the tectum in Figures 2 and 3). These terminals are ordered in a precise topographic map (across the rostro-caudal and medial-lateral axes) that is duplicated in the various dorsal (shallow) to ventral (deep) layers of the tectal neuropil in a way that is highly conserved 196

in most species of fish (Vanegas and Ito, 1983; Vonbartheld and Meyer, 1987; Baier et al., 1996; Robles et al., 2013; Kita et al., 2015). Different subtypes of RGC, relaying different types of visual information from the retina, target specific dorso-ventral layers of the tectal neuropil and other AFs (Robles et al., 2013; Robles et al., 2014). The nature of the visual information emerging from the retina is important, since this provides the working material with which circuits in the tectum and elsewhere will decode and interpret visual stimuli such as looms. In recent years, RGC subtypes have been shown to deliver specific orientation, direction, and size relevant information to the tectum (Gabriel et al., 2012; Nikolaou et al., 2012; Lowe et al., 2013; Preuss et al., 2014; Semmelhack et al., 2014; Barker and Baier, 2015), including the possibility that tectum receives input from loom-specific RGCs in the retina (Temizer et al., 2015).

Figure 2 summarizes the loom related information that RGCs deliver to the neuropil, including details of the neuropil laminae and sublaminae that may have specific roles in the subsequent processing. RGCs carrying information on motion and changes in luminance are targeted mostly to the stratum fibrosum et griseum superficiale (SFGS) and also to the stratum griseum centrale (SGC)(Temizer et al., 2015). Dimming stimuli mostly activate axons innervating the deep SFGS6 layer, while SFGS2- SFGS5 inputs are more active during looms than dimming. That RGCs innervating the SFGS2-5 layers respond to an approaching object more than to a dim (change in luminance) suggests that they are processing information related to the moving edges. Interestingly, this last group of RGC axons selectively innervate the tectum, and not other AFs (Robles et al, 2014). The prominence of loom information arriving in the tectum suggests an important role for this structure in the detection of visual threats, a notion that is supported by experiments in which lesions of the tectal neuropil markedly reduced escape responses in zebrafish larvae (Temizer et al., 2015).

197

2018). In this manner, SINs may be gating the information that flows from the RGCs to the PVNs, making stimuli with specific characteristics (such as looms) more salient. Dunn and colleagues (Dunn et al., 2016a), in modelling this interaction, found that activity in SINs and RGC axons could explain the PVN’s responses to loom stimuli. Providing functional support for this idea, Barker and Baier (Barker and Baier, 2015) ablated SINs, altering avoidance responses to large objects, although this effect was not observed in another similar ablation study (Yin et al., 2018).

This provides a plausible mechanism by which the tectum could identify looms based on retinal input, but it does not take into account the retinal information that is delivered to the other nine AFs, and the ways in which other visual regions of the brain could contribute to escape behavior. Notably, a recent study has revealed a specific and important role for the thalamus in loom processing (Heap et al., 2018b). In this study, Heap and colleagues characterized visual responses in the thalamus, finding it to be responsive to looms. They then identified projections from the thalamus to the tectal neuropil (mainly localized in deeper layers, orange in Figures 2 and 3), and found that these projections were active specifically during looms. To assess the functional contributions of these projections, they ablated the thalamo-tectal tract and found a selective reduction in tectal responses to looms, but not to other visual stimuli, meaning that the thalamus was contributing loom-relevant information to visual processing in the tectum. Ablation of the thalamo-tectal tract also had a surprising impact on escape behavior. These ablations not only reduced the probability of loom-induced escape, but reversed their directionality, causing the larvae to startle toward, rather than away from, the looming stimulus. To gauge the stimulus property responsible for thalamic loom responses, they presented checkerboard and dimming stimuli, and found that the drop in luminance (dimming) was exclusively responsible for the thalamus’ loom responsiveness. This luminance information is likely delivered to the thalamus by RGC axons terminating in AF8 (Temizer et al., 2015). In further behavioral experiments with these stimuli, Heap and colleagues found that checkerboard looms had a decreased startle rate compared to dark looms, but that dim stimuli failed to elicit responses, consistent with prior studies (Temizer et al., 2015; Dunn et al., 2016a). This suggests that thalamic luminance information cannot drive escapes on its own, but increases the probability of a response, and controls its direction. To confirm these proposed roles for luminance and the thalamus in visual escape, Heap and colleagues presented a split stimulus in which a dim was presented to one eye while and expanding checkerboard was presented to the other. This stimulus led to a higher probability of escape than dimming or looming checkerboards individually, and produced more responses toward the looming checkerboard stimulus than away from it. The results of this study support a model in which information about moving edges, carried directly from RGCs into the tectum, are necessary for the

198

escape responses, but that luminance information brought to the tectum through the thalamus is necessary to direct the escape away from the predator (the side receiving the earlier or more dramatic drop in luminance). Although this challenges the basic retino-tectal model of visual escape, it still places the tectum at the center of loom processing. As the point of confluence for moving edge and luminance information, it is well positioned to perform the key calculations necessary for visual predator detection.

The details of the tectal microcircuits combining these streams of information remain largely unknown, but the results of these calculations are delivered to the dendrites of PVPNs in the deep layers of the neuropil (Scott and Baier, 2009; Robles et al., 2011) (Figure 2), and the PVPNs then relay these outputs to premotor neurons to shape the motor response (Figure 3)(Zottoli et al., 1987; Sato et al., 2007; Yao et al., 2016). Recent findings suggest that two subpopulations of PVPNs are involved. One projects to contralateral reticulospinal neurons and the nucleus of the medial longitudinal fasciculus (nMLF), while the other innervates ipsilateral reticulospinal neurons (Helmbrecht et al., 2018) (dark and light blue neurons, respectively, in Figure 3). These subtypes of PVPNs presumably permit the activation of the contralateral M-system (driving a C-bend) to be coordinated with other postural adjustments and the inhibition of ipsilateral spinal motor circuits (Fetcho and Faber, 1988; Fetcho, 1991; Song et al., 2015) (for reviews, see (Eaton et al., 1991; Korn and Faber, 1996; Eaton et al., 2001; Korn and Faber, 2005; Medan and Preuss, 2014; Hale et al., 2016)). Interestingly, tectal projections to the M-cells fire equally during flashes and looms, but the M-cell responses are more specific to loom stimuli (Yao et al., 2016). These results suggest that the M-cells may require further loom-specific information in order to reach threshold, or that inhibitory mechanisms could keep the M-cells in check during other stimuli. As noted above, escape responses still occur after M-cell ablation, showing that these responses can be generated and modulated by multiple categories of reticulospinal neurons (Eaton et al., 1982; Liu and Fetcho, 1999; Dunn et al., 2016a; Naumann et al., 2016), with the M-cells mediating short-latency C-bends and other reticulospinal neurons driving long-latency C-bends (Kohashi and Oda, 2008; Bhattacharyya et al., 2017). This difference in motor circuitry is important, because it underlies the previously described choice between a rapid, invariant escape response (typically used for rapid looms and imminent collisions) and a slower, adjustable escape (used for less urgent threats) (Bhattacharyya et al., 2017).

After the initial reorientation produced by the C-bend, the escape response involves a swim burst, propelling the fish forward and away from the threat. Considering the results of previous studies, it is likely that this motor response is produced by the nMLF (light green nucleus in Figure 3) as it has been involved with escape responses and forward swimming behaviors, and has visually responsive

199

neurons (Gahtan et al., 2002; Orger et al., 2008; Sankrithi and O'Malley, 2010; Severi et al., 2014; Thiele et al., 2014; Wang and McLean, 2014).

[Insert Figure 2 here]

Figure 2. Loom relevant information arriving to the tectal neuropil. Left: RGC axons (in red) carry movement information to layers 2-5 of the SFGS and dimming information to layer 6 of the SFGS and the deep portion of the SGC (Temizer et al, 2015). Thalamic axons (in orange) carrying dimming information target the deep region of the SGC and also deep parts of the SAC (Heap et al, 2018). Right: Smaller objects are preferentially processed in the SO and superficial layers of the SFGS while larger objects are mostly processed at deeper layers of the SFGS (Preuss et al, 2014; Semmelhack et al, 2014).

[Insert Figure 3 here]

Figure 3. The core visual escape circuit. Schematic summary of the brain regions involved in detecting and responding to visual looming stimuli.

Modulation of the core escape circuit

C-bend escapes are disruptive and energetically expensive, so while they are critical to survival in some contexts, it is also important that the animal can modulate or prevent these behaviors when they are not needed. In this section, we will highlight the ways in which internal states, recent experience, and other types of context impact visual escape behavior. We will also review the brain regions and circuits that may be responsible for encoding these factors, and the ways in which this information reaches and influences the activity of the core escape circuit to produce behavior appropriate to the circumstances facing the animal. These pathways are summarized in Figure 4.

200

noradrenergic neurons in the locus coeruleus; cocaine-and amphetamine-regulated transcript (CART) and cholinergic neurons in the tegmentum; and dopaminergic, serotoninergic, and neuropeptide-Y neurons of the hypothalamus (purple structure in Figure 4). Similar activity was found in these brain structures after caffeine was applied, which also decreased the reaction time. On the other hand, activity in somatostatin neurons in the hypothalamus was correlated with longer reaction times. These results led Lovett-Barron and colleagues to conclude that the neuromodulatory structures active before short reaction times are involved in alertness and that this internal state of the animal reduces the reaction time to the threatening stimulus, although the mechanisms by which this influences the core visual escape circuit remain unclear.

An important example of sensorimotor gating in visual escape has recently been revealed by Yao and colleagues (Yao et al., 2016). In their study, they show how hypothalamic dopaminergic neurons, which receive information from the tectum, are more active when non-threatening visual stimuli (flashes, receding circles or bright loom) are presented. This activity suppresses transmission from the tectum to the Mauthner neurons through inhibitory glycinergic interneurons of the hindbrain (black neuron in Figure 4). When a looming stimulus is presented, these dopaminergic neurons do not activate the inhibitory interneurons, allowing free communication between the tectum and reticulospinal neurons and permitting the escape response. This circuit provides a “safety“ that prevents unnecessary C-bends, even when a visual stimulus contains some properties (motion or edges) in common with threatening stimuli.

A third study by Filosa and colleagues involved the relationships among visual stimuli, approach and escape behavior, and the hypothalamic-pituitary-interrenal (HPI) axis (Filosa et al., 2016). Under pressure to find food, starved larvae become less risk-averse, showing a decreased tendency to startle from visual stimuli and increased approach behavior associated with feeding. This state was associated with less cortisol and less activity in the HPI axis. They then showed that starved larvae also had increased activity in raphe serotoninergic neurons that project to the tectum (pink in Figure 4 and 5), and that by manipulating the activity of these neurons they could modulate the behavior of both fed and starved larvae. Finally, they found that the activity of the HPI axis and serotoninergic neurons modifies tectal responses to visual circles of various sizes (where large stimuli generally elicit startles and small stimuli cause approach behavior) in a way that matched the behavioral results. Their findings suggest that the activity of these serotonergic neurons is tuning the sensitivity of tectal circuits, thus tipping the balance in downstream premotor regions toward avoidance or approach, depending on the animal’s immediate survival pressures.

201

Involvement of the hypothalamus is a common thread among these studies, and it appears to have distinct effects on escape behavior and to impinge on different parts of the core escape circuit under different conditions. Since the hypothalamus also sends projections directly to deep layers of the tectal neuropil (purple axons in Figure 5)(Heap et al., 2018a), it may influence escape behavior in ways that have not yet been studied. This makes the hypothalamus, and its circuit-level interactions with loom-sensitive brain regions, intriguing topics for further studies of how visual startle is modulated.

[Insert Figure 4 here]

Figure 4. Modulatory structures of the core of visual escape circuit. Schematic representation of the brain regions that are known to modulate (solid lines) or are likely to modulate (dashed lines) the core visual escape circuit.

Open questions about the loom circuit

imaging (with calcium or voltage indicators), anatomical labelling (Forster et al., 2017a; Forster et al., 2018), and light sculpting to produce custom holograms in the brain (Favre-Bulle et al., 2015; Accanto et al., 2019). This combination of techniques has permitted neurons with particular response profiles to stimuli (judged by calcium imaging) to be labelled with photoactivatable GFP (Patterson and Lippincott-Schwartz, 2002) through the use of a targeted hologram on the relevant neuron (dal Maschio et al., 2017; Kramer et al., 2019), linking the neurons’ morphologies to their activity. A similar approach could be used for the selective activation or silencing of particular functional categories of neurons, using sculpted light and optogenetic actuators (dal Maschio et al., 2017). In terms of the neurons’ functional characteristics, careful registration of in vivo calcium imaging and post mortem immunohistochemistry has revealed the neurochemical properties of neurons with interesting patterns of activity during sensory processing (Lovett-Barron et al., 2017). In a conceptually similar approach, transgenic marker lines have been combined with calcium imaging to provide further information about the active neurons in real time (Dunn et al., 2016a; Yao et al., 2016; Heap et al., 2018b). A broad library of transgenic marker lines (Davison et al., 2007; Scott et al., 2007; Scott and Baier, 2009; Satou et al., 2013; Kimura et al., 2014; Kawakami et al., 2016; Forster et al., 2017b; Tabor et al., 2019) should make this a flexible and powerful approach in the future.

The second way in which the model is incomplete is that it almost certainly still omits whole brain regions contributing to the decision making and execution of visual escape. Promising brain regions include additional loom-responsive areas like the pretectum, midbrain tegmentum, and forebrain (Dunn et al., 2016a; Chen et al., 2018), which have yet to be thoroughly explored functionally. Among these structures, the pretectum (yellow in Figure 4) receives direct inputs from the retina (Burrill and Easter, 1994; Yanez et al., 2018), including luminance information into AF6 (Temizer et al., 2015). In addition, it has reciprocal connections with the tectum (Yanez et al., 2018) and has been linked to other visually evoked behaviors in zebrafish (Kubo et al., 2014; Portugues et al., 2014; Semmelhack et al., 2014; Naumann et al., 2016). The tegmentum, another loom-responsive area, includes a nucleus that is also likely to modulate escape responses. Studies in goldfish and sunfish have shown that the nucleus isthmi (NI) responds to looming stimuli, is more sensitive to novel visual stimuli, and has reciprocal connections with the deeper layers of the tectum (green area in Figures 4) (Vanegas and Ito, 1983; Striedter and Northcutt, 1989; King and Schmidt, 1993; Northmore and Gallagher, 2003; Gallagher and Northmore, 2006). This raises the possibility that an NI-tectal loop is involved in the processing information about approaching objects (Northmore and Graham, 2005; Graham and Northmore, 2007). This visual feedback circuit is preserved across vertebrates (Gruberg et al., 2006) and its role has been studied in birds. Several studies of owls, chickens, and pigeons strongly suggest

203

that the NI modulates tectal responses to salient visual stimulus and is involved with visual attention. (Marin et al., 2007; Asadollahi et al., 2010; Mysore and Knudsen, 2013; Goddard et al., 2014; Mysore and Knudsen, 2014; Basso and May, 2017).

Another structure likely involved in modulating the responses to approaching predators is the medial division of the dorsal pallium (Dm, brown structure in the telencephalon in figure 4). This forebrain structure is viewed as the homolog of the mammalian amygdala (Wullimann and Mueller, 2004; Yamamoto et al., 2007; Mueller et al., 2011) and could therefore participate in responses to aversive stimuli in zebrafish (von Trotha et al., 2014; do Carmo Silva et al., 2018; Lal et al., 2018). The Dm is likely to receive most of its visual information from the preglomerular nucleus (brown relay structure in Figure 4). According to tracing experiments in weakly electric fish, rainbow trout, carp, and goldfish, this diencephalic structure receives projections from the tectum and establishes connections with the dorsal pallium, which in turn closes the loop by sending axons to the tectum (brown in Figure 5) (Echteler and Saidel, 1981; Luiten, 1981; Demski, 2003; Folgueira et al., 2004; Yamamoto and Ito, 2005; Northcutt, 2006; Yamamoto et al., 2007; Yamamoto and Ito, 2008; Giassi et al., 2012b; Giassi et al., 2012a; Demski, 2013). These connections are reminiscent of the visual pathway involved in the visual escape responses triggered by looms in mice (Mueller, 2012; Zhao et al., 2014; Carr, 2015; Wei et al., 2015; Yuan and Sus, 2015; Perathoner et al., 2016; Pereira and Moita, 2016; Shang et al., 2018; Zhou et al., 2019) and that may be important in the processing of threatening visual stimuli in humans (McFadyen et al., 2017). Given the similarities to the amygdala’s connectivity in mammals, it is tempting to think that the Dm may play a similar role in fish.

The circuitry involved in the visual escape response can be approached in its minimal form with a retina-thalamus-tectum-hindbrain circuit (Figure 3) or as a more complex interaction with multiple modulating structures (summarized in Figure 4). Viewed either way, the tectum is the structure where most of the visual information from the retina, thalamus, NI, telencephalon, and other structures converges (Figure 5), and is where the separate features of loom stimuli are integrated. As such, the tectum appears to serve as a visuomotor pivot where visual information is gathered and interpreted, and where premotor signals for an escape response are generated.

A framework for these microcircuits is illustrated in Figure 5, were we summarize the laminar structure of the tectal neuropil as it relates to various visual and modulatory inputs. This highlights the fact that retinal inputs dominate the SFGS, where information about movement, edges, and topography are delivered, providing the raw material for visual processing. In deeper layers, inputs from other regions, with relevance to luminance, context, and internal state are more prominent, providing further detail that may be necessary for the direction, kinematics, or indeed, the decision to 204

startle. Registering this against the anatomical structure of the tectum (with superficial retinorecipient SINs, PVINs with dendrites in the superficial neuropil, and PVPNs with dendrites in the deep neuropil) provides a structural scaffold through which processing and modulation could occur as information progresses from the superficial to the deep neuropil, and then on to downstream structures. The elucidation of the individual cell types and tectal microcircuits carrying out this processing are of particular interest for future work, as they will both provide a fine-grained functional map of this important transformation, and will reveal the circuit-level mechanisms by which modulatory information is incorporated.

[Insert Figure 5 here]

Figure 5. The confluence of retinal, thalamic, and modulatory information in the tectal neuropil. Left: The two core visual afferents to the tectal neuropil are the RGCs (red) and projections from the thalamus (orange) (Temizer et al, 2015; Heap et al, 2018). Right: Structures with known afferents to the tectal neuropil, and that are likely to modulate loom processing, include the hypothalamus (purple), raphe (pink), nucleus isthmi (green) and the telencephalon (brown) (Heap et al, 2018; Filosa et al, 2016; King and Schmidt, 1993; Luiten, 1981; Folgueira et al, 2004). Middle: the projections from each of these structures terminate in stereotyped laminae and sublaminae in the neuropil, with a trend toward visual (especially retinal) input to superficial layers and modulatory input to deeper layers. Retinal information is received and processed by the SINs and PVINs. The PVINs then innervate the PVPNs which, in turn, send axons to premotor structures (Del Bene et al, 2010; Nevin et al, 2010). Sketches of PVNs and SINs are based on those in (Nevin et al, 2010).

205