Neuroethology of the Zebrafish

Neuroethology of the Zebrafish:

Describing the Neural Circuits that Control Innate Behaviors

A dissertation presented

Adam Raymond Kampff

The Department of Molecular and Cellular Biology

in partial fulfillment of the requirements

for the degree of

Doctor of Philosophy

in the subject of

Biology

Harvard University

Cambridge, Massachusetts

October 2008

Florian Engert Adam Raymond Kampff

Neuroethology of the Zebrafish: Describing the Neural Circuits that Control Innate Behaviors

Abstract

The complexity of the animal brain begs for simple tools that will assist scientists attempting to unravel the beautifully entangled mess. I begin by describing an experimental strategy for indentifying the components of neural circuits that control innate behaviors. My approach emphasizes the role of behavior analysis and the insights offered by neuroethology. While describing the neuroethological strategy, I will introduce the larval zebrafish and tools that I have developed to study the neural basis of innate visual behavior. The second chapter reports an implementation of the strategy that successfully revealed components of the circuit underlying the zebrafish optomotor reflex. We discovered distinct subsets of spinal projection neurons that were responsible for directing the swims and turns that constitute an important visual response to whole- field motion. The success of this study relied on the ability to record behavior-related neural activity in a restrained zebrafish, but this was found to be difficult for other visual behaviors. The third chapter addresses this problem by describing a novel technique to monitor neural activity in an unrestrained, behaving zebrafish. Employing the bioluminescent calcium reporter, Aequorin, we were able to detect the activity of

Hypocretin neurons in freely swimming zebrafish. The success of this technique will allow an extensive survey of how identified populations of neurons collectively choreograph the complex behavior patterns of a developing vertebrate.

iii

Table of Contents

Chapter 1: Introduction to the Neuroethology of Larval Zebrafish

1.0 - A neural description of behavior 1.1 - The neuroethological approach to identifying neural circuits 1.2 - The larval zebrafish as a model system for the neuroethological approach 1.3 - Implementing and augmenting the neuroethological strategy 1.4 - References

Chapter 2: Neural Correlates of the Optomotor Response

2.0 - Introduction 2.1 - Behavioral responses to stabilized whole-field motion 2.2 - Motor patterns underlying the optomotor response 2.3 - Calcium imaging of responses in spinal-projection neurons 2.4 - Projection neuron responses to whole-field motion 2.5 - Laser ablation of turning neurons 2.6 - Discussion 2.7 - Methods 2.8 - References

Chapter 3: Monitoring Neural Activity in Freely Behaving Zebrafish

3.0 - Introduction 3.1 - Potential strategies for neural recordings in unrestrained zebrafish 3.2 - Aequorin and its use as neural Ca2+-reporter 3.3 - Using GFP-Aequorin in freely-swimming larval zebrafish 3.4 - Neuroluminescent zebrafish 3.5 - Genetically-targeted recordings from hypocretin neurons during behavior 3.6 - Future Directions 3.7 - Methods 3.8 - References

Acknowledgements

I wish to thank Florian Engert; it is perhaps unfair to him and his amazing support to thank anyone else. So I won’t…at least not in the same paragraph.

I wish to thank my collaborators in the Engert Lab, but in particular, Mike Orger and Eva Naumann. I wish to thank my advisory committee. I wish to thank my parents. Most of all, I wish to thank YOU, the reader, but you are very likely one of the people mentioned already.

Oh, and the zebrafish, thanks for all the…fish.

Chapter 1

Introduction to the Neuroethology of Larval Zebrafish

1.0: A neural description of behavior

The nervous system controls an organism’s behavior, and I want to understand

how it does this. This is a daunting goal1 and rather than flounder in aimless

disappointment, I have focused on answering a simplified question: what are the

components of the neural circuit responsible for controlling an innate response to a

sensory stimulus? In the following chapter, I describe an experimental strategy for

addressing this question and a model preparation that is uniquely suited to this strategy.

However, when combined with well designed experimental techniques, I believe this

strategy can be used to identify many such circuits, underlying many different behaviors,

at work in many different brains. I am not naïvely suggesting that we would ultimately

achieve a satisfactory understanding of the brain, but I am proposing that we will have

made a significant step towards this goal. Having described some of the brain’s

functional circuits, we will be in a position to ask how they develop, how they change

during learning, and how they fail when damaged.

1.1: The neuroethological approach to identifying neural circuits

Attempts to describe the circuitry of the brain are complicated by the difficult task

of identifying neurons that participate in a dedicated circuit. The experimental

neuroscientist must consider a tremendous range of sensory stimuli, any of which might

evoke a response from the repertoire of an animal’s behavior. An uninformed search for

related stimuli and responses would be futile. Fortunately, ethologists have taken a

perspective informed by the natural ecology and evolution of animal behavior2. They

have had remarkable success in identifying very specific stimuli that evoke unique

behaviors3. Such behaviorally-relevant stimuli, sometimes termed ‘sign stimuli’ or

‘releasing mechanisms’3, provide an excellent tool for uncovering the components of a

dedicated neural circuit. If a neuroscientist is able to present these relevant stimuli, or

sufficiently accurate reproductions, to an experimental preparation from which neural

activity can be recorded, then it is informative to simply record which neurons are

activated (or inhibited). These specifically active neurons are possible participants in the

circuit controlling the related behavior. Furthermore, subtle changes to the relevant stimulus often eliminate, or significantly alter, the behavioral response3. Comparing the

neurons activated by the relevant stimulus and similar stimuli that have no effect, or

different effects, on behavior, the neuroscientist can isolate subsets of neurons that might

be specific to the behavior. Subsequent studies are needed to address the causal role of

these isolated neurons in controlling the behavior; their necessity and sufficiency can be

assayed with targeted ablation and stimulation experiments. This experimental strategy

for circuit identification, which I will call the neuroethological approach because of its

emphasis on detailed behavioral analysis, is outlined in Figure 1.1.

Figure 1.1

Figure 1.1- The Neuroethological Approach to Circuit Identification

A successful implementation of the neuroethological strategy requires an

appropriate experimental preparation. I will motivate my decision to work with larval

zebrafish by comparing the preparation to an ideal neuroethological model system.

1.2: The larval zebrafish as a model system for the neuroethological approach

The zebrafish (Danio rerio) is a freshwater teleost native to the rivers of India4.

The larvae develop externally and are almost completely translucent, contributing to their popularity as a developmental model organism5. At only one day post fertilization (1dpf),

zebrafish show behavioral responses to touch as well as spontaneous motor activity6.

Visual responses appear by day three, almost immediately after the axons of ganglion

cells leaving the eye reach their targets7, and remarkably, after just 5 days, they begin

visually hunting prey8. Such a rapid development has inspired many research groups to

address the genetic basis of zebrafish behavior9, and the research community has produced a host of genetic tools useful in the investigation of the neural circuits controlling behavior10, 11. However, the criteria for a successful neuroethological system are often distinct from those of a general model system. In the following, I will discuss how the larval zebrafish, along with the tools that I have developed, provides an almost- ideal system for implementing the three stages of the neuroethological strategy: behavior analysis, neural recordings, and causality tests.

Behavior Analysis (ideal):

In a laboratory setting, assays for identifying behaviorally-relevant stimuli require precise control of the stimulus environment while simultaneously monitoring the animal’s behavior. Difficulties in controlling the sensory environment result primarily from the modality being controlled (e.g. it is currently much easier to produce a spatially patterned visual scene than a spatially patterned olfactory space), and in some cases, certain animal preparations will allow the use of more easily controlled stimuli12.

However, it is often more challenging to accurately measure the behavior of an animal,

and this will provide a more stringent criterion for evaluating potential neuroethological systems.

For many neuroscience model systems: primates, rodents, birds and most insects, behavior consists of the intricate coordination of muscle contraction, directing the motion of multiple limbs, and resulting in complex movement through three-dimensional space.

Reducing the description of behavior to a measurable variable usually produces a

grotesque simplification of the response, e.g. describing motor coordination as the

amount of time it takes a mouse to fall off a rotating rod13. For many purposes, however,

these descriptors are adequate, but the ideal neuroethological model system would

provide a rich description of behavior obtainable with straightforward techniques.

Behavior Analysis (larval zebrafish):

Zebrafish larvae demonstrate a variety of visual behaviors after only five days of

development7. The ease of controlling visual stimuli along with this rich behavioral

repertoire guided my decision to investigate the neural circuits underlying visual

behavior. More important, however, is the relative ease with which larval zebrafish

behavior can be measured. When restricted to a shallow arena, only a few millimeters

deep, larval zebrafish behavior occurs primarily in a two-dimensional plane.

Furthermore, other than the motion of small pectoral fins and eye positioning, zebrafish

behavior is represented by the undulating pattern of muscle contraction that gives rise to

swimming and turning8. Many distinct patterns of swimming and turning have been

characterized, and all can be accurately measured with high-speed video recording. By

surrounding a swim arena with a 360° projection screen14, displaying computer generated

visual stimuli, I constructed a behavior assay (Figure 1.2) allowing the real-time control

of the visual environment while simultaneously monitoring nearly the entire behavioral

repertoire of the larval zebrafish.

Figure 1.2

Figure 1.2 – Assay for visual behavior of the larval zebrafish

A Larval zebrafish swim in a transparent cylinder surrounded by a 360° projection

screen. An infrared CCD or CMOS camera monitors zebrafish location and tail positions

at frame rates reaching 500Hz. A video projector presents computer-controlled stimuli

that can be updated in response to fish behavior in real-time. B The position of a

zebrafish larva swimming freely in the arena recorded continuously for one hour.

With the ability to accurately monitor the zebrafish while presenting arbitrary

visual scenes, we are in a position to search for behaviorally relevant stimuli. Where to

begin? Some relevant visual stimuli had previously been identified: whole-field motion

was shown to elicit both an optokinetic (OKR)15 and optomotor response (OMR)16.

However, both of these visual behaviors do not require the optic tectum17, the main

recipient of retinal ganglion cell afferents. In our attempts to identify relevant stimuli that

more fully engage the zebrafish visual system, we were guided by the ethological

perspective. Zebrafish begin hunting at only 5dpf and must successfully find food in

order to survive (much data, not shown). It was found that zebrafish use vision to track

and reorient towards moving prey prior to a capture attempt18. When presented with prey-

like stimuli (small moving spots), zebrafish reliably reoriented towards the stimulus

(Figure 1.3a); the swim kinematics of this behavior were similar to the small-angle

tracking turns seen during natural hunting19. We also discovered that for larger stimuli,

larval zebrafish reacted with a yet uncharacterized visual avoidance behavior (Figure

1.3b) that consisted of a single large-angle turn away from the offending stimulus.

Figure 1.3

Figure 1.3 – Visual attraction and avoidance behavior

Examples of the turning responses produced by larval zebrafish presented with moving

stimuli. The orientation of the fish is plotted for multiple presentations of left (+ angle,

red) and right (- angle, blue) moving bright spots. The stimulus began directly in front of

the fish (0°) and moved to +/-90° at 60°/second, where it remained stationary for 1

second before disappearing. A For small spots (1° diameter), the zebrafish reoriented

towards the stimulus with a series of small angle turns. B For larger spots (5° diameter),

the zebrafish responded with a single large turn away from the stimulus.

These two classes of stimuli, large and small moving objects differing by just a

few degrees in size, reliably evoke two radically different behaviors. In terms of the neuroethological strategy, they represented excellent probe stimuli for isolating the neurons participating in distinct circuits. That is, of course, only true if we can record from neurons while a zebrafish is producing these distinct behaviors.

Neural Recordings (ideal):

The ability to record from neurons during the presentation of relevant-stimuli and performance of related behaviors is perhaps the most important, and certainly the most difficult, requirement of a neuroethological model system. The ideal system would allow recordings from every neuron, simultaneously, in a behaving animal, during stimulus

presentation. No preparation satisfies all these requirements, so it is worth discussing the

trade-offs. In many preparations, electrophysiology provides access to any individual

neuron in the intact brain, but the goal is to identify all of the potential participants in a

neural circuit. To this end, monitoring populations of neurons with optical techniques has emerged as a popular and proven strategy20-22. Optophysiology provides tools for

recording neural activity on many different spatial scales23, but circuit identification

requires the ability to resolve individual neurons in vivo. Current techniques, such as laser

scanning microscopy, are limited in their ability to penetrate intact neural tissue and often

require invasive surgery to remove an opaque layer of overlying tissue. The ideal

neuroethological system would provide non-invasive optical access to the every neuron

of the intact animal’s brain.

Even if neurons can be imaged, it is often impossible to stably record from them

during behavior. Neuroscience has made much progress recording sensory responses in

the neurons of anesthetized animals, however, the logic of the neuroethological strategy

requires recording neural activity while an animal reacts to behaviorally-relevant stimuli; this is not possible for paralyzed or anesthetized animals. Scientists have taken two approaches to overcoming this limitation: study ‘fictive’ behaviors than can be monitored in restrained animals or develop techniques to record neural activity during unrestrained behavior. The ideal neuroethological model system would be amenable to both of these approaches.

Neural Recordings (larval zebrafish):

The translucence of the larval zebrafish, particularly of pigmentation mutants that

preserve eye pigment while eliminating body melanophores24, makes it an obvious

candidate for optical techniques for recording neural activity during visual stimulation25.

Although electrophysiology is possible26, 27, and might ultimately be necessary to dissect the function of identified circuits, I will focus on optical techniques that allow recording neural activity from populations of neurons throughout the entire fish brain.

Intact neural tissue, even when translucent, is prone to optical scattering that

limits the depth penetration of conventional imaging strategies, e.g. confocal microscopy.

However, with the development of two-photon microscopy28, 29, diffraction limited

optical sectioning was possible to depths exceeding 500µm30, the maximum thickness of

the larval zebrafish brain. In addition, two-photon microscopy uses an infrared excitation

source that is invisible to zebrafish (and human) photoreceptors, and is thus ideally suited

to neural recordings during the presentation of visual stimuli. I designed and constructed

a custom two-photon laser scanning microscope (Figure 1.4) paired with a setup for presenting visual stimuli to restrained zebrafish.

Figure 1.4

Figure 1.4 – Custom two-photon microscope for zebrafish visual neuroethology

Schematic of the principal optomechanics and instrumentation integrated into a unique

system for neuroethology, allowing recording neural and behavioral responses to visual

stimuli in a restrained zebrafish.

Two-photon imaging provided access to every neuron of the zebrafish brain, with single-cell resolution (Figure 1.5), the only vertebrate model system for which this is possible. A number of optical probes of neural activity have been used with zebrafish and

when combined with two-photon microscopy, it will be possible to record responses to

behaviorally-relevant stimuli in every potential component of the controlling circuit21, 31.

Figure 1.5

Figure 1.5 – Two-photon imaging in the larval zebrafish brain

A transgenic larval zebrafish (left), expressing a fluorescent reporter pan-neuronally31

was imaged with a two-photon microscope. A single dorsal section through the optic-

tectum, cerebellum, and hindbrain (middle) was acquired at high spatial resolution,

allowing every individual neuron to be resolved (zoom-in, right).

The more stringent requirement for monitoring neural recordings in a

neuroethological system is the ability to record during behavior. Two-photon microscopy requires the larval zebrafish to be restrained in agar. Fortunately, restrained zebrafish will continue to react to some relevant-stimuli, albeit in an obviously restricted manner. Agar- embedded larvae will move their eyes in response to rotating gratings (OKR), and if their tail is removed from the agar and free to move, whole-field motion stimuli induce head-

restrained zebrafish to produce swimming and turning tail patterns consistent with the

freely-swimming OMR25. This ‘fictive’ OMR was vital to the success of the experiments

reported in Chapter 2. However, when we looked at the responses of restrained zebrafish to other behaviorally-relevant stimuli, such as moving spots (Figure 1.3), the behavior was severely attenuated (Figure 1.6).

Figure 1.6

Figure 1.6 – Whole-field motion versus moving-spot response of restrained zebrafish

Motion response of fully-embedded zebrafish presented with rotational gratings (A) and

moving-spots (B) (Motion Frames/500ms Bin). The stimuli started at 5 seconds (gray

bar) and alternated directions on (Figure 1.6 continued) subsequent presentations; the

responses to each direction were averaged (n = 12 larvae, 6-9dpf, 500ms bins). The

motion responses to small and large spots were combined in B, and in contrast to freely-

swimming fish, responses these same spot stimuli were absent.

The effect of restraint on zebrafish behavior may not be surprising, but it is still

rather disappointing; it represents a serious limitation of the larval zebrafish as a

neuroethological preparation. In order to study behavior for which no ‘fictive’ version

can be characterized in restrained fish, such as visual approach and avoidance (Figure

1.3), a technique is required to record neural activity in unrestrained larvae. Chapter 3

will describe just such a technique.

Causality Tests (ideal):

Tools for manipulating the circuits of the brain have been traditionally rather

crude. However, tissue ablations (intended or otherwise)32 and electrical stimulation have

provided some of the most fascinating evidence about the functional organization of

nervous systems. In order to complete a neuroethological identification of neural circuits,

tools for specifically ablating and stimulating the correlates identified from activity

recordings will be necessary. Recently, laser techniques have facilitated targeted

ablations10 (see Chapter 2), but the most promising new tools have emerged from

genetics. Cell-type specific expression of neurotoxins has been used to test for loss of

function33 and, more recently, genetically-encoded, optically-gated channels have been

used to specifically excite34 and inhibit35 neural populations. These tools will continue to

develop, and the ideal neuroethological model system should be genetically tractable and

thus able to fully exploit these impressive technical advances.

Causality Tests (larval zebrafish):

As explained above, the popularity of zebrafish as a vertebrate model has resulted

in a community of researchers developing genetic tools for all types of study, including

the ablation and stimulation of neurons. Again, the translucence of the zebrafish makes it particularly well-suited to optical strategies (including “optogenetics”). Targeted laser ablations have been used to determine the causal role of entire brain regions and single neurons10, and recent reports of successful use of channelrhodopsin-227 in zebrafish opens

the possibility of addressing the role of genetically-specified neural populations. In my experience, the ‘causality test’ stage of the neuroethological strategy has been the most difficult. It is encouraging to observe the wide range of new tools being developed and exciting to imagine the new experiments they will allow.

1.3: Implementing and augmenting the neuroethological strategy

The following chapters will report: an implementation of the neuroethological strategy that successfully revealed components of the circuit underlying the zebrafish optomotor reflex (chapter 2) and a novel technique to monitor neural activity in a freely behaving zebrafish (chapter 3).

1.4: References

1. Neuroscientist. Every Neuroscience Paper Ever Published.

2. Lorenz, K. On aggression (Harcourt, New York,, 1966).

3. Tinbergen, N. The study of instinct (Clarendon Press, Oxford [Eng.], 1951).

4. Spence, R., Gerlach, G., Lawrence, C. & Smith, C. The behaviour and ecology of the zebrafish, Danio rerio. Biol.Rev.Camb.Philos.Soc. 83, 13-34 (2008).

5. Haffter, P. & Nusslein-Volhard, C. Large scale genetics in a small vertebrate, the zebrafish. Int J Dev Biol 40, 221-7 (1996).

6. Saint-Amant, L. & Drapeau, P. Time course of the development of motor behaviors in the zebrafish embryo. Journal of Neurobiology 37, 622-632 (1998).

7. Fleisch, V.C. & Neuhauss, S.C. Visual behavior in zebrafish. Zebrafish 3, 191- 201 (2006).

8. Budick, S.A. & O'Malley, D.M. Locomotor repertoire of the larval zebrafish: swimming, turning and prey capture. Journal of Experimental Biology 203 Pt 17, 2565-2579 (2000).

9. Baier, H. Zebrafish on the move: towards a behavior-genetic analysis of vertebrate vision [In Process Citation]. Current Opinion in Neurobiology 10, 451- 455 (2000).

10. Gahtan, E. & Baier, H. Of lasers, mutants, and see-through brains: functional neuroanatomy in zebrafish. Journal of Neurobiology 59, 147-161 (2004).

11. McLean, D.L. & Fetcho, J.R. Using imaging and genetics in zebrafish to study developing spinal circuits in vivo. Dev Neurobiol 68, 817-34 (2008).

12. Heiligenberg, W. Neural nets in electric fish (MIT Press, Cambridge, Mass., 1991).

13. Crawley, J.N. Behavioral phenotyping of transgenic and knockout mice: experimental design and evaluation of general health, sensory functions, motor abilities, and specific behavioral tests. Brain Res 835, 18-26 (1999).

14. Orger, M.B. et al. Behavioral screening assays in zebrafish. Methods Cell Biol 77, 53-68 (2004).

15. Easter, S.S., Jr. & Nicola, G.N. The development of eye movements in the zebrafish (Danio rerio). Dev Psychobiol 31, 267-76 (1997).

16. Orger, M.B., Smear, M.C., Anstis, S.M. & Baier, H. Perception of fourier and non-fourier motion by larval zebrafish. Nat.Neurosci. 3, 1128-1133 (2000).

17. Roeser, T. & Baier, H. Visuomotor Behaviors in Larval Zebrafish after GFP- Guided Laser Ablation of the Optic Tectum. Journal of Neuroscience 23, 3726- 3734 (2003).

18. McElligott, M.B. & O'Malley D, M. Prey tracking by larval zebrafish: axial kinematics and visual control. Brain Behav Evol 66, 177-96 (2005).

19. Borla, M.A., Palecek, B., Budick, S. & O'Malley, D.M. Prey capture by larval zebrafish: evidence for fine axial motor control. Brain Behav.Evol. 60, 207-229 (2002).

20. Ohki, K., Chung, S., Ch'ng, Y.H., Kara, P. & Reid, R.C. Functional imaging with cellular resolution reveals precise micro-architecture in visual cortex. Nature 433 (2005).

21. Niell, C.M. & Smith, S.J. Functional imaging reveals rapid development of visual response properties in the zebrafish tectum. Neuron 45, 941-951 (2005).

22. Ramdya, P. & Engert, F. Emergence of binocular functional properties in a monocular neural circuit. Nat Neurosci (2008).

23. Yuste, R., Lanni, F. & Konnerth, A. Imaging neurons (Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y., 1999).

24. Lister, J.A., Robertson, C.P., Lepage, T., Johnson, S.L. & Raible, D.W. nacre encodes a zebrafish microphthalmia-related protein that regulates neural-crest- derived pigment cell fate. Development 126, 3757-3767 (1999).

25. Orger, M.B., Kampff, A.R., Severi, K.E., Bollmann, J.H. & Engert, F. Control of visually guided behavior by distinct populations of spinal projection neurons. Nat Neurosci 11, 327-33 (2008).

26. Ramdya, P., Reiter, B. & Engert, F. Reverse correlation of rapid calcium signals in the zebrafish optic tectum in vivo. J Neurosci Methods 157, 230-7 (2006).

27. Douglass, A.D., Kraves, S., Deisseroth, K., Schier, A.F. & Engert, F. Escape behavior elicited by single, channelrhodopsin-2-evoked spikes in zebrafish somatosensory neurons. Curr Biol 18, 1133-7 (2008).

28. Denk, W., Strickler, J.H. & Webb, W.W. Two-photon laser scanning fluorescence microscopy. Science 248, 73-76 (1990).

29. Svoboda, K. & Yasuda, R. Principles of two-photon excitation microscopy and its applications to neuroscience. Neuron 50, 823-839 (2006).

30. Helmchen, F. & Denk, W. Deep tissue two-photon microscopy. Nat Methods 2, 932-40 (2005).

31. Higashijima, S., Masino, M.A., Mandel, G. & Fetcho, J.R. Imaging neuronal activity during zebrafish behavior with a genetically encoded calcium indicator. J Neurophysiol 90, 3986-97 (2003).

32. Ratiu, P. & Talos, I.F. Images in clinical medicine. The tale of Phineas Gage, digitally remastered. N Engl J Med 351, e21 (2004).

33. Shah, E.M. & Jay, D.G. Methods for ablating neurons. Curr Opin Neurobiol 3, 738-42 (1993).

34. Boyden, E.S., Zhang, F., Bamberg, E., Nagel, G. & Deisseroth, K. Millisecond- timescale, genetically targeted optical control of neural activity. Nat.Neurosci. 8, 1263-1268 (2005).

35. Zhang, F. et al. Multimodal fast optical interrogation of neural circuitry. Nature 446, 633-639 (2007).

Chapter 2

[This chapter reports published1 work done in collaboration with Dr. Michael Orger.

Although Dr. Orger oversaw analysis of the behavior and Ca2+ imaging and I oversaw the

development of the technical assays for behavior analysis, two-photon imaging and laser

ablations, every aspect of the project reflects our mutual contribution and fruitful

collaboration.]

Neural Correlates of the Optomotor Response

2.0: Introduction

In a behaving animal, the brain communicates its intentions to muscles via the

pattern of activity in descending projection neurons2. In vertebrates, these cells respond to

the detection and processing of sensory stimuli and transmit their motor command to the

local networks of the spinal cord, which in turn initiate and coordinate muscle

contraction3. A fundamental question in neuroscience is how the commands that initiate behaviors are encoded in the activation of these projection neurons4-6.

The spinal projection system of fish provides an excellent model for studying this

code7. Six day old zebrafish show a diverse range of swimming behaviors8-10 mediated

by a descending projection to the spinal cord consisting of fewer than 300 neurons11, 12.

These neurons are easily labeled with fluorescent indicators13, are optically accessible

with modern imaging techniques, and are arranged in a stereotyped pattern such that the

same cells or groups of cells can easily be identified from one fish to the next14. These

neurons are morphologically diverse with distinct dendritic fields, and axonal projection

patterns14, 15, suggesting that they serve different behavioral functions. Nevertheless,

determining how differing patterns of activity in these spinal projection neurons produce

different motor outputs has proved difficult.

An important step towards achieving this goal is to find ways to reliably and

specifically activate distinct types of movement. One approach has been to use tactile

stimuli to the head and tail of zebrafish, which evoke different escape swims and

differentially activate the segmental homologs of the Mauthner cell13. Activity in the

Mauthner cell is sufficient to initiate escape sequences in goldfish16. Ablation of these

neurons in zebrafish can eliminate short latency, high performance turns in response to

tactile and acoustic stimuli, but the fish still show slower, long latency responses17, 18, and

tactile stimuli were found to activate virtually all reticulospinal neurons19. Similar

widespread activation of motor control neurons has been observed in both invertebrate20,

21 and vertebrate model systems22, 23. This suggests that, although small numbers of neurons may be sufficient to activate different behaviors24, 25, the actual command is

encoded in distributed activity throughout the population of control neurons3.

The swimming behaviors of zebrafish, including the complex sequence of turns and swims that make up the escape response, can be broken down into basic kinematic elements9. It is possible that the motor commands for these basic behaviors originate

from distinct populations of neurons, which when combined, would appear as a

‘distributed command’. We found that whole-field visual motion, with the direction

dynamically locked to the fish’s body axis, is able to selectively evoke some of these

basic swim patterns. Calcium imaging of stimulus-evoked responses in the complete

population of neurons projecting to the spinal cord revealed that these control neurons are

indeed organized into distinct functional groups. A causal link between this observed

organization and control of the fishes’ behavior was established via targeted single-cell

ablations.

2.1: Behavioral responses to stabilized whole-field motion

Larval zebrafish show a reliable optomotor response, turning and swimming to

follow large moving gratings presented from below26. This requires that the fish respond

to different stimulus directions with different swimming trajectories. To analyze the

components of this complex behavior we needed to eliminate the effect of the fishes’

swimming on the stimulus orientation and motion. We achieved this by tracking the

fishes’ position and heading using a high speed camera, and translating and rotating the

stimulus to cancel the fishes’ own motion (Figure 2.0a). Thus, if a fish viewing a right

to left grating turns to the left, the stimulus will rotate with him so that he continues to

see leftward motion.

We found that the fishes’ swimming behavior is strongly tuned for stimulus

direction (Figure 2.0b). Stabilized gratings moving in 24 different directions relative to

the fishes’ body axis trigger 24 different swimming trajectories. Gratings moving from

tail to head cause vigorous forward swimming; gratings that drift towards the fishes’ left

or right cause turning in the corresponding direction. The basic swim parameters, forward

velocity and angular velocity, both show strong tuning for stimulus direction (Figure

2.0c–d).

Figure 2.0 – Behavioral responses to drifting gratings

a Schematic of the behavioral set up. b Average trajectories of 6 fish responding to

gratings drifting at 24 equally spaced directions relative to their body axis. Trajectories

are color coded according to stimulus direction as shown in the key, with up representing

forward, or tail to head motion (green). The fish, superimposed for scale, is shown at the

start position. Gratings have a spatial period of 1 cm, 5 mm below the fish, and drift with

a temporal frequency of 1 Hz for 6 seconds. c, d Tuning of swimming parameters to

grating direction, for the same 6 fish. e High speed kinematics of the swim bouts. Images

of a fish taken at 5 ms intervals during 3 modes of swimming are presented from left to

right. Colored arrows show the stimulus direction. Red dots indicate the starting position of the fish; the blue dot in the last frame indicates its final position and the overall trajectory is described to the right. f Distributions of swim bout parameters, forward motion and change in angle of the fishes’ body axis, in response to 8 directions of motion

(10 fish, 22540 swim events). The fish viewed gratings of 50% contrast with the same spatiotemporal parameters as b. The gratings were stationary relative to the fish for 5

seconds, to measure a baseline swimming distribution, and then drifted for 5 seconds.

Each swim bout is characterized in terms of distance moved along the fishes’ initial

heading, and the angle through which the body axis turns. Distributions are color coded

by stimulus direction, with the peak of the distribution set to the maximum intensity. The

8 panels representing the 4 cardinal and 4 oblique directions are arranged in a square,

with tail to head motion at the top. The center panel shows the distributions for tail to

head (green), back left (blue) and back right (red) stimulus motion superimposed. g Polar

plot of swim-type frequency (forward swim, 0-5mm forward, < 10º turned and turns, >

(Figure 2.0 continued)

30º) versus stimulus direction h Cumulative turning of one fish in response to multiple 10

second presentations of back left (+135º, blue) and back right (-135º, red) motion. Bold

line indicates the mean response.

Figure 2.0

2.2: Motor patterns underlying the optomotor response

Larval zebrafish swimming behavior occurs in discrete bouts alternating with

brief periods of inactivity. These swimming bouts can be grouped into a small number of

distinct categories, which have been characterized in detail9, 10, 27. We analyzed the

visually-evoked responses at the level of single swim bouts. The high-speed kinematics

of three typical swims evoked by different grating directions, a left turn, a right turn and a

forward swim, are compared in Figure 2.0e. The left and right turns consist of a sharp tail flick in one direction in which the tip of the tail bends around up to 180 degrees. This

may correspond to the visually evoked large angle turns seen in a recent kinematic

study18. The forward swim consists of ~30 Hz tail oscillations and corresponds to the

‘slow swim’9.

We compare the distribution of swimming bouts, characterized by two

parameters, distance swum forward and angle turned, in response to moving gratings in 8

different directions (Figure 2.0f). In the center we have superimposed, for comparison,

the distributions for forward (tail to head), back left and back right motion. Two features

of the distributions are clear: the responses to each direction are highly clustered,

indicating that they are stereotyped across trials and across fish. 84 percent of all swim

bouts recorded during forward motion propel the fish forward 0.5-4.0 mm with less than

9º change in body angle. In contrast, 80 percent of bouts during back left motion are left

turns larger than 45º. Secondly, there is less than three percent overlap between the

distribution of swimming bouts evoked by forward motion, and the turning responses

shown superimposed in the center panel. From this, we conclude that whole-field drifting

gratings can be used to isolate distinct components of zebrafish swimming behavior.

Clustering of turns and forward swims is clearly seen when responses to all 8 directions

are combined (Figure 2.1). The stimulus tuning for each distinct behavior (Figure 2.0g)

can now be compared with the stimulus tuning of neural responses. A complete mapping

of neural activity requires that the responses persist over multiple presentations of the

stimuli. This is the case for the whole-field motion responses, which show little or no

reduction over time (Figure 2.0h).

Figure 2.1

Figure 2.1 – Clustering of behavioral responses to whole-field motion

a Combined distribution of 22540 swim bouts shown in 8 outer panels of Figure 2.0f.

Three clusters representing forward swims and left and right turns are evident. b

Histogram of angle turned for the data in a.

2.3: Calcium imaging of responses in spinal-projection neurons

All motor commands from the fishes’ brain to its spinal cord must pass through

the array of descending projection neurons. Calcium imaging was used to monitor the

activation of all such neurons during presentation of the behaviorally-relevant stimuli

described above. Two-photon excitation was essential, in part because the infrared

excitation light was invisible to the fish, and also because it provides access to the

deepest cells of the reticulospinal system. Responses were typically stable for several

hours, allowing imaging at many focal planes and subsequent reconstruction of the

response properties for the whole population.

We illustrate the analysis of such an experiment for a group of cells in the right

rhombomere 1 (Figure 2.2a).

Figure 2.2

Figure 2.2 – Measuring calcium responses in spinal projection neurons

a Z-projection of data acquired during calcium imaging at different z planes. The image is shown with inverted luminance here and in subsequent figures for visual clarity; all images are presented rostral side up. Two cells are indicated and analysis of their calcium signals is shown to the right. b Mean fluorescence traces from the 3D regions of interest that define the two cells in a. The stimulus consisted of sinusoidal gratings drifting in 8 different directions presented for 10 seconds each (shaded regions). The directions are indicated by arrows. c Polar plot showing time averaged fluorescence changes for the 8

(Figure 2.2 continued)

directions of motion for the same two cells. The directionality vectors are shown to the

right. d The projected outlines of all 16 cells shown in a are color coded according to

their directionality vector using the inset look-up table (maximum intensity corresponds

to 0.5 DPI, direction preference index). e Analysis as in a-d performed on 151 labeled

neurons in one fish. f Schematic of our cell classification system, where thin lines

represent single neurons and bold lines denote groups, based on a z-projection of cells

retrogradely labeled using Texas Red dextran (Invitrogen, USA).

Two example cells are indicated and their calcium responses, averaged over their

3D volume, are plotted (Figure 2.2b). The stimulus was a sinusoidal grating, which

drifted at 1 Hz for 10 seconds in 8 directions (shaded regions), and was stationary for 17

seconds in between. The responses of these two cells are clearly selective for different

directions. Notably, all the projection neurons with stimulus-evoked responses (> 0.1 Δf/f

to at least one direction) had strong direction selectivity (Figure 2.3). We plot the cells’

responses, in polar form, as the average percentage increase in fluorescence for each

stimulus direction and also as a ‘directionality vector’28 (Figure 2.2c). The results of this

analysis can be summarized in a color-coded image (Figure 2.2d). Projections of the 3D

regions of interest used for data analysis are outlined in white. We fill each region with a

color according to its directionality vector, with hue representing direction and intensity representing magnitude (see inset look up table). Here we can see that cell 1 is colored

blue indicating a preference for forward motion and cell 2 is red indicating a rightward

preference.

Figure 2.3

Figure 2.3 – Motion responsive spinal projection neurons are direction selective

a Distribution of DPI for all stimulus responsive cells (> 0.1 Δf/f to at least one direction).

b Histogram of direction preference for three groups of neurons, defined as the direction

of the vector sum of their responses (stimulus directions x response magnitudes), with 16

equally spaced bins. Only direction-selective cells (vector sum magnitude > 0.05 Δf/f)

are included. Positive angles indicate motion towards the cell’s side of the brain. c

Histogram of V cell direction preferences using 32 bins.

The end result of such an analysis for one fish with scattered labeling throughout

the spinal projection system is shown in Figure 2.2e. Most cells either did not respond

(black) or preferred forward motion (blue). A few cells prefer gratings moving in

directions that elicit turning (red/green). The organization of forward motion responsive

cells seen in this fish is repeated in all individuals that we have examined. We divided

the cells into 20 morphological groups for analysis, and show here a schematic of our

classification system (Figure 2.2f).

2.4: Projection neuron responses to whole-field motion

Identifying the neural correlates of the zebrafish whole-field motion response

requires a complete mapping of the activity of projection neurons in response to drifting

gratings. We analyzed responses from 1465 neurons in 50 fish. We show here a

summary of the responses of all cells (Figure 2.4). For each category, we show the

directionality vector of every single responsive neuron (left brain in green, right brain in

red) and the average tuning curve for that group in polar form. The cells fall into distinct

functional groups. The stimulus tuning of each group can be compared with the stimulus

tuning of the behavioral patterns (Figure 2.0g), allowing us to hypothesize which of these neurons might participate in each behavior.

Figure 2.4 – Summary of direction tuning of all spinal projection neurons

In the top left is a schematic summary of the directional preference of spinal projection neurons. Cells that prefer tail to head motion are colored blue, whereas cells with a left or right preference are colored green or red respectively, and mixed populations are striped. Along the right and bottom, two plots are shown for each cell group. The left plot of each pair shows the directionality vectors for every responsive cell within a group and the right plots give the average tuning curves for the population, shown in polar form as in Figure 2.1c. Data from cells on the right side of the brain is shown in red; data from left cells is in blue; up represents the tail to head direction. The label for each plot gives the category name followed by the number of cells recorded on the left and the right side of the brain respectively. The colored boxes group functionally similar classes of neurons according to their stimulus preference (forward = blue, turn (right/left) = red/green, no response = black).

Figure 2.4

Cells involved in initiating slow swims should be selectively activated by forward

motion and show bilaterally symmetric responses. Both of these properties were seen for

cells in the nucleus of the medial longitudinal fasciculus (nucMLF) including the large

MeL and MeM cells, RoL1, RoR1 and RoM1c (blue box). Cells of the MiR and MiM

types also occasionally show weak forward responses (light blue box), but this was not

consistent across fish. Three examples of well labeled nucMLF cells, color-coded

according to their directional preference are shown (Figure2.5a). In every fish forward preferring neurons are scattered throughout the nucleus.

We expected neurons involved in turning to show two characteristics. They should be activated by stimuli that evoke turns (Figure 2.0g), and their responses should

be lateralized: a left preferring neuron should have a contralateral partner with the

opposite preference. In only two places are these conditions met: in RoM1r (red/green

box) and in the ventromedial (V) cells of rhombomeres 3-5 (red/green/blue box). Cells in the RoM1r group are small and not always well labeled. However, they are clearly identifiable by shape and location and showed strikingly lateralized response properties selective for ‘turn’ stimuli. The spatial organization of this direction preference in rhombomere 1 is consistent from fish to fish (Figure 2.5b).

Figure 2.5 – Spatial distribution of visually responsive cells

Examples of the response distribution in different fish. a Bilaterally labeled nucMLF

cells in three different fish with a z-projection on the left and color-coding by

directionality vectors on the right, as presented in Figure 2.2. b Three examples of rhombomere 1 cell groups from left and right brain. Left and right motion preference is restricted to small rostral RoM cells and is consistent across fish. c Responses in the ventromedial portion of rhombomeres 3-5 in a fish with good bilateral labeling. Each group contains a mixture of forward and turn-stimuli preferences.

The average tuning curve of the V cells includes both bilateral forward motion responses and lateralized responses to left and right stimuli. However, the response vectors of these cells span a wide range of directions. This implies that these cells, although grouped together based on anatomical features, are in fact functionally heterogeneous. This is best demonstrated by looking at all V cells individually in a single fish (Figure 2.5c). These clusters contain mixtures of neurons with different response

properties. Many cells show a strong forward motion preference, but some show a

lateralized preference for turning stimuli. Therefore, this group could be involved in

generating both slow swims and turns. We plotted the distribution of preferred directions

for neurons in responsive groups (Figure 2.3b–c). The remaining groups of cells show no

consistent responses (black box), although most of these have been shown to have

calcium responses to tactile stimuli19 and are likely to participate in other behaviors.

2.5: Laser ablation of turning neurons

Calcium imaging revealed that projection neurons are organized into distinct

functional groups related to distinct behaviors, which suggests that removing one of these groups should result in a loss of its associated behavioral response. To test for the expected causal relationship, we chose to remove neurons responsive to ‘turn’ stimuli,

RoM1r and V cells, since there are relatively few and the two laterally symmetrical populations provide an internal control. To remove the selected cells, we adapted a technique29, based on non-linear absorption, for the rapid ablation of single identified

neurons in the zebrafish. A mode-locked laser is scanned in a spiral pattern over a

selected cell until a highly localized plasma is formed30, destroying the target but leaving

adjacent cells intact (Figure 2.6a–c). To test the efficacy of this technique, we ablated the

Mauthner cell and its segmental homologs reproducing the side specific increase in

escape response latency that was found in previous studies17 (data not shown).

Figure 2.6 – Laser ablation of turning selective neurons

a Z-projection of spinal projection neurons labeled with Texas Red dextran. b, c A right

RoM2l neuron before and after laser ablation. The cell and its processes are destroyed

and cell debris is visible. The adjacent RoM2m cell (arrow) and its processes are unaffected. d–f Ablation of all the RoM cells unilaterally in rhombomeres 2 and 3 causes no phenotype. d shows which neurons were selected on a schematic image. The fish were tested before and after ablation using gratings drifting in two directions: back left and back right. e and f shows the results of one such ablation. The cumulative turning responses to many repetitions of the stimulus are plotted, with those to back left in green, back right in red, and the mean turning is shown in bold (as in Figure 2.0h). There is no

obvious difference between the pre and post ablation behavior. g–l Ablation of 8-10 V

cells in the left RoV3-MiV2 region can completely eliminate stimulus-evoked turns

towards the ablated side. m Average ablation phenotypes. Change in turning following

ablation is expressed as a percentage change from pre-ablation behavior (-100 means fish

no longer turns; +100 means fish turns twice as much). Unilateral ablation of RoM2 and

3 cells, or rostral RoM1 cells has no significant effect on turning. However ablation of V

cells unilaterally, alone, or in conjunction with a RoM1r ablation, causes a severe deficit

in turning with respect to control fish (p < 0.001; n=14,9, two-tailed Student’s T-test,

error bars show s.e.m.).

Figure 2.6

We measured the behavioral responses of fish, before and after ablation, to stimuli

which reliably evoked turning, backward left and backward right motion. In order to test

the behavior around the threshold for responding, the stimuli were ramped up in contrast

from 0 to 20 percent over 20 seconds. As an additional control, we unilaterally ablated all

the large RoM cells in rhombomeres 2-3, since they show no grating responses. Fish were always able to turn in both directions after this control ablation, as seen in raw data from an example fish (Figure 2.6d–f). For the V cells, we unilaterally ablated 2-4 cells from each cluster, since we couldn’t predict by morphology which would show the turning responses or guarantee that all cells were labeled. Nevertheless, in several cases, this resulted in a complete elimination of turning towards the ablated side (Figure 2.6g– l). In cases where turning towards the ablated side was eliminated, forward swims and turns away from the ablated side were still performed normally (Figure 2.7). On average, with this stochastic method, we saw a large, significant, decrease in turning whenever we ablated cells in these clusters (Figure 2.6m). 7 out of 23 ablations resulted in a greater than 90% reduction in turning towards the ablated side. 3 fish had less than a

10% reduction, and the rest had intermediate, graded phenotypes. Surprisingly, no significant change in turning was observed when only the RoM1r cells were ablated. We

therefore argue that this small subpopulation of functionally lateralized ventromedial

neurons is a necessary component of the circuit underlying turning in the optomotor

response.

Figure 2.7

Figure 2.7 – V-Cell ablation alters swim bout distribution

Pre and post ablation distributions of swim bout parameters for the six V-cell ablated fish

with the largest deficit. Positive turn angles indicate turns towards the ablated side.

Colored lines from dark blue (low) to red (high) are contours of log frequency. Following

the classification of swim bouts in Figure 1g, the distribution of forward swims and turns away from and towards the ablated side is: pre-ablation 4566:1595:1624; post-ablation

4176:1624:75.

2.6: Discussion

This study identifies groups of spinal projection neurons selective for visual

stimuli that evoke basic swim patterns in zebrafish, and reveals, with cellular detail, the

spatial organization of a vertebrate motor control system. These functional groups are

distinct and are composed of an unexpectedly small number of neurons. The discrete

organization we see, which differs from previously observed distributed activity5, 19, 31, is made apparent by the ability to evoke simple motor patterns with controlled visual stimulation. The fact that more complex behaviors are associated with distributed activation of many neurons is consistent with a model in which subsets of neurons initiate distinct components of the behavior32, 33.

Additional evidence for a distributed motor command has come from a number of

studies finding that ablations do not abolish specific behaviors, although they can alter their latency and kinematics17, 18, 34-36. Here, an ablation guided by an observed functional

organization was able to completely abolish a specific behavior: large angle turns in one direction can be specifically eliminated by ablation of the small number of cells responsive to turn-evoking stimuli. Why do we see such a strong phenotype? Two aspects of the approach contributed to this: dynamically stabilized visual stimuli were used to elicit very specific motor patterns and responses in all projection neurons were assessed, thereby identifying all possible participants in the behavior.

This characterization of a functionally organized motor control system presents new opportunities to address the development and function of systems underlying vertebrate behavior. Zebrafish are a well-established model system for vertebrate development. Hundreds of thousands of neurons and their synaptic connections are formed in the first few days of the zebrafish’s life. In an intriguing study37, Mendelson

grouped reticulospinal neurons according to the timing of their axonal growth.

Remarkably, there is a one to one correspondence between Mendelson’s classification

and ours: all neurons in the developmental categories ‘nucMLF’ and ‘second wave’ show responses to forward motion, while all others do not. Therefore, the projection neurons

active during forward slow swims constitute a distinct developmental as well as

functional group. Interestingly, the spinal interneurons involved in slow swimming behavior also appear a few hours later than those involved in escapes38.

The spinal projection neurons link sensory processing in the brain to motor output

in the spinal cord, and therefore provide an excellent starting point for studying the

sensorimotor transformations underlying behavior. Having identified which neurons

control particular motor patterns, we can now ask how their activity is decoded within the

spinal cord to produce the associated behavior. Tremendous progress has been made

recently in understanding the organization and function of specific cell-types and circuits

in zebrafish spinal cord39-41, encouraging the hope that a connection between descending

motor commands and the resulting motor patterns is within reach. More obscure, however, is the upstream circuitry that leads to the selective activation of these descending control neurons. Recently developed optical techniques, which are capable of revealing activity and connectivity in large numbers of neurons42-44, can now be focused

on determining the inputs to neurons with known response selectivities and identified

roles in behavior. Together, this opens up the possibility of studying complete circuits

underlying complex behaviors in a vertebrate with a small, transparent brain.

2.7: Methods

Zebrafish

Six to seven day old zebrafish from AB and WIK strains were used in all

experiments. nacre -/- fish45 were used for imaging and ablation studies since they lack

pigment in the skin, but retain normal eye pigmentation. Their behavior is

indistinguishable from wild-type siblings in our assay. All experiments were approved by

Harvard University’s standing committee on the use of animals in research and training.

Behavioral setup

Zebrafish larvae swam freely in a 5 cm diameter acrylic arena. Fish were

illuminated by an array of infrared light-emitting diodes and their swimming behavior

was recorded at 200 Hz using an infrared sensitive high speed CMOS camera (Mikrotron,

Germany). Stimuli were projected directly onto a 10 cm screen 5 mm below the fish

using a DLP projector (Optoma, USA). Custom image processing software (Visual C++,

Microsoft, USA) extracted the position and orientation of the fish at the acquisition frame

rate. This information was used to update a stimulus rendered in real-time using

DirectX3D (Microsoft, USA).

Calcium Imaging

Spinal projection neurons were filled as previously described13. Briefly, a 50%

w/v solution of dextran-conjugated calcium green (Invitrogen, USA) was injected into the

spinal cord of tricaine-anesthetized fish 24 hours before imaging. Fish were embedded in

1.5% low melting temperature agarose, paralyzed using an injection of a-bungarotoxin and imaged with a custom two-photon microscope46, with a pulsed Ti-sapphire laser

tuned to 920 nm (Spectra Physics, USA). The stimulus was projected from below with a

DLP projector [Optoma, USA] and passed through a red long-pass filter, which allowed

simultaneous visual stimulation and fluorescence detection. In each experiment frames

were acquired at 3.6 Hz. After one or more repetitions of each stimulus, the focus was

moved to a different z plane. The resulting time series were combined to yield a 4D

picture of the scanned volume of the brain. This image was passed through image

segmentation filters in Matlab (Mathworks, USA) to extract the volume occupied by each

cell. Cell identities were assigned by hand based on morphological criteria14. Cells with

an evoked response greater than 0.1 Δf/f to at least one stimulus direction were assigned

directionality vectors and a directional preference index (DPI) calculated as in Euler et

al.28

Two-photon laser ablation

Targeted cells were identified and a central sub-region was selected from a full frame scan. The power of a mode-locked laser (850 nm) was linearly increased while the beam was scanned in a spiral pattern throughout the targeted region. Scanning was immediately terminated upon the detection of brief flashes of saturating intensity, which are presumed to result from the creation of a highly localized plasma via multi-photon absorption by water molecules29, 30. This procedure always results in destruction of the

cell despite immediately adjacent cells appearing unaffected, as verified using both a pan-

neuronally expressing transgenic line (HuC:YC2.144) and retrogradely labeled spinal projection cells.

2.8: References

1. Orger, M.B., Kampff, A.R., Severi, K.E., Bollmann, J.H. & Engert, F. Control of visually guided behavior by distinct populations of spinal projection neurons. Nat Neurosci 11, 327-33 (2008).

2. Orlovsky, G.N., Deliagina, T.G. & Grillner, S. Neuronal control of locomotion (Oxford University Press, 1999).

3. Rossignol, S., Dubuc, R.J. & Gossard, J.P. Dynamic sensorimotor interactions in locomotion. Physiological Reviews 86, 89-154 (2006).

4. Zelenin, P.V. Activity of individual reticulospinal neurons during different forms of locomotion in the lamprey. European Journal of Neuroscience 22, 2271-2282 (2005).

5. Deliagina, T.G., Zelenin, P.V. & Orlovsky, G.N. Encoding and decoding of reticulospinal commands. Brain Research 40, 166-177 (2002).

6. Korn, H. & Faber, D.S. The Mauthner cell half a century later: a neurobiological model for decision-making? Neuron 47, 13-28 (2005).

7. Gahtan, E. & Baier, H. Of lasers, mutants, and see-through brains: functional neuroanatomy in zebrafish. Journal of Neurobiology 59, 147-161 (2004).

8. Saint-Amant, L. & Drapeau, P. Time course of the development of motor behaviors in the zebrafish embryo. Journal of Neurobiology 37, 622-632 (1998).

9. Budick, S.A. & O'Malley, D.M. Locomotor repertoire of the larval zebrafish: swimming, turning and prey capture. Journal of Experimental Biology 203 Pt 17, 2565-2579 (2000).

10. Burgess, H.A. & Granato, M. Modulation of locomotor activity in larval zebrafish during light adaptation. Journal of Experimental Biology 210, 2526-2539 (2007).

11. Lee, R.K. & Eaton, R.C. Identifiable reticulospinal neurons of the adult zebrafish, Brachydanio rerio. J.Comp Neurol. 304, 34-52 (1991).

12. Kimmel, C.B., Powell, S.L. & Metcalfe, W.K. Brain neurons which project to the spinal cord in young larvae of the zebrafish. J.Comp Neurol. 205, 112-127 (1982).

13. O'Malley, D.M., Kao, Y.H. & Fetcho, J.R. Imaging the functional organization of zebrafish hindbrain segments during escape behaviors. Neuron 17, 1145-1155 (1996).

14. Metcalfe, W.K., Mendelson, B. & Kimmel, C.B. Segmental homologies among reticulospinal neurons in the hindbrain of the zebrafish larva. J.Comp Neurol. 251, 147-159 (1986).

15. Gahtan, E. & O'Malley, D.M. Visually guided injection of identified reticulospinal neurons in zebrafish: a survey of spinal arborization patterns. J.Comp Neurol. 459, 186-200 (2003).

16. Nissanov, J., Eaton, R.C. & DiDomenico, R. The motor output of the Mauthner cell, a reticulospinal command neuron. Brain Research 517, 88-98 (1990).

17. Liu, K.S. & Fetcho, J.R. Laser ablations reveal functional relationships of segmental hindbrain neurons in zebrafish. Neuron 23, 325-335 (1999).

18. Burgess, H.A. & Granato, M. Sensorimotor gating in larval zebrafish. Journal of Neuroscience 27, 4984-4994 (2007).

19. Gahtan, E., Sankrithi, N., Campos, J.B. & O'Malley, D.M. Evidence for a widespread brain stem escape network in larval zebrafish. J.Neurophysiol. 87, 608-614 (2002).

20. Briggman, K.L., Abarbanel, H.D. & Kristan, W.B., Jr. Optical imaging of neuronal populations during decision-making. Science 307, 896-901 (2005).

21. Wu, J.Y., Cohen, L.B. & Falk, C.X. Neuronal activity during different behaviors in Aplysia: a distributed organization? Science 263, 820-823 (1994).

22. Bosch, T.J., Maslam, S. & Roberts, B.L. Fos-like immunohistochemical identification of neurons active during the startle response of the rainbow trout. J.Comp Neurol.2001.Oct.22.;439.(3):306.-14. 439, 306-314 (2001).

23. Zelenin, P.V., Orlovsky, G.N. & Deliagina, T.G. Sensory-motor transformation by individual command neurons. Journal of Neuroscience 27, 1024-1032 (2007).

24. Wiersma, C.A. & Ikeda, K. Interneurons commanding swimmeret movements in the crayfish, Procambarus clarkii (Girard). Comp Biochem.Physiol 12, 509-525 (1964).

25. Pearson, K.G. Common principles of motor control in vertebrates and invertebrates. Annual Review of Neuroscience 16, 265-297 (1993).

26. Orger, M.B., Smear, M.C., Anstis, S.M. & Baier, H. Perception of fourier and non-fourier motion by larval zebrafish. Nat.Neurosci. 3, 1128-1133 (2000).

27. Borla, M.A., Palecek, B., Budick, S. & O'Malley, D.M. Prey capture by larval zebrafish: evidence for fine axial motor control. Brain Behav.Evol. 60, 207-229 (2002).

28. Euler, T., Detwiler, P.B. & Denk, W. Directionally selective calcium signals in dendrites of starburst amacrine cells. Nature 418, 845-852 (2002).

29. Chung, S.H., Clark, D.A., Gabel, C.V., Mazur, E. & Samuel, A.D. The role of the AFD neuron in C. elegans thermotaxis analyzed using femtosecond laser ablation. BMC.Neurosci. 7, 30 (2006).

30. Vogel, A. & Venugopalan, V. Mechanisms of pulsed laser ablation of biological tissues. Chem.Rev. 103, 577-644 (2003).

31. Leonardo, A. & Fee, M.S. Ensemble coding of vocal control in birdsong. Journal of Neuroscience 25, 652-661 (2005).

32. d'Avella, A., Saltiel, P. & Bizzi, E. Combinations of muscle synergies in the construction of a natural motor behavior. Nat.Neurosci. 6, 300-308 (2003).

33. Bizzi, E., d'Avella, A., Saltiel, P. & Tresch, M. Modular organization of spinal motor systems. Neuroscientist. 8, 437-442 (2002).

34. Gahtan, E. & O'Malley, D.M. Rapid lesioning of large numbers of identified vertebrate neurons: applications in zebrafish. Journal of Neuroscience Methods 108, 97-110 (2001).

35. Zottoli, S.J., Newman, B.C., Rieff, H.I. & Winters, D.C. Decrease in occurrence of fast startle responses after selective Mauthner cell ablation in goldfish (Carassius auratus). J.Comp Physiol [A] 184, 207-218 (1999).

36. Gahtan, E., Tanger, P. & Baier, H. Visual prey capture in larval zebrafish is controlled by identified reticulospinal neurons downstream of the tectum. Journal of Neuroscience 25, 9294-9303 (2005).

37. Mendelson, B. Development of reticulospinal neurons of the zebrafish. II. Early axonal outgrowth and cell body position. J.Comp Neurol. 251, 172-184 (1986).

38. Kimura, Y., Okamura, Y. & Higashijima, S. alx, a zebrafish homolog of Chx10, marks ipsilateral descending excitatory interneurons that participate in the regulation of spinal locomotor circuits. Journal of Neuroscience 26, 5684-5697 (2006).

39. Bhatt, D.H., McLean, D.L., Hale, M.E. & Fetcho, J.R. Grading movement strength by changes in firing intensity versus recruitment of spinal interneurons. Neuron 53, 91-102 (2007).

40. McLean, D.L., Fan, J., Higashijima, S., Hale, M.E. & Fetcho, J.R. A topographic map of recruitment in spinal cord. Nature 446, 71-75 (2007).

41. Chong, M. & Drapeau, P. Interaction between hindbrain and spinal networks during the development of locomotion in zebrafish. Dev.Neurobiol. 67, 933-947 (2007).

42. Niell, C.M. & Smith, S.J. Functional imaging reveals rapid development of visual response properties in the zebrafish tectum. Neuron 45, 941-951 (2005).

43. Sato, T., Hamaoka, T., Aizawa, H., Hosoya, T. & Okamoto, H. 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 (2007).

44. Higashijima, S.I., Masino, M.A., Mandel, G. & Fetcho, J.R. Imaging neuronal activity during zebrafish behavior with a genetically encoded calcium indicator. J.Neurophysiol. 90, 3986-3997 (2003).

45. Lister, J.A., Robertson, C.P., Lepage, T., Johnson, S.L. & Raible, D.W. nacre encodes a zebrafish microphthalmia-related protein that regulates neural-crest- derived pigment cell fate. Development 126, 3757-3767 (1999).

46. Denk, W., Strickler, J.H. & Webb, W.W. Two-photon laser scanning fluorescence microscopy. Science 248, 73-76 (1990).

Chapter 3

[This chapter reports a technique developed in collaboration with Eva Naumann. Eva and

I worked together in developing the bioluminescence assay, experimental protocols, and

data analysis. Eva produced all pan-neuronal transgenic and transient Aequorin

expressing zebrafish. Dr. David Prober generated the transgenic Aequorin-Hypocretin

zebrafish used in the study.]

Monitoring Neural Activity in Freely Behaving Zebrafish

3.0: Introduction

The definitional goal of neuroethology, to understand the relationship between the

activity of neural circuits and the behavior of an organism, demands techniques for

monitoring brain activity during behavior. Unfortunately, few such techniques exist, all

possess serious limitations, and none are currently applicable to the neuroethology-suited

larval zebrafish1-4. Current tools for measuring neural activity in zebrafish require larvae to be restrained and/or paralyzed5, 6. This requirement restricts current techniques to

studying the few behaviors for which a ‘fictive’ version can be evoked and monitored in

the stabilized animal (e.g. the optomotor response, see Chapter 27). Even if a ‘fictive’

behavior can be found, inferring the relationship between neural responses in this

artificial setting and those occurring during natural behavior presents an awkward

challenge. With the goal of directly measuring the neural activity corresponding to

natural behaviors, we developed a technique to monitor activity-dependent Ca2+ signals in unrestrained, freely swimming zebrafish larvae.

3.1: Potential strategies for neural recordings in unrestrained zebrafish

The small (length: 3-4mm) larval zebrafish is unable to support the infrastructure required for the transmission of electrical data3, and thus electrophysiology in freely-

swimming zebrafish larvae is not feasible. However, the translucence of the larval

zebrafish suggests an optical approach to detecting neural signals during unrestrained

behavior. I will first discuss the virtues and limitations of existing optical tools in order to

motivate our decision to develop a novel approach.

Optophysiology, the optical recording of brain activity, has found most of its

success measuring the Ca2+-signals indirectly related to neural activity. The commonly

used optical Ca2+ indicators relate a change in fluorescence intensity to a change in intracellular Ca2+levels. Transient increases in intracellular Ca2+ resulting from action potentials and/or synaptic activation, detected by either synthetic or genetically-encoded

fluorescent probes, have been used to infer neural activity in restrained larval zebrafish7-

10. However, given the larvae’s small size and translucence, it would be possible to

illuminate an entire behavior arena with excitation light and monitor Ca2+-dependent

changes in fluorescence intensity as the fish swam freely. Such a technique would

provide only temporal information about the activity of a population of cells containing

the Ca2+-reporter, but genetically-encoded sensors could be targeted to express in only

the specific neurons of interest. When combined with genetic-targeting, a fluorescence-

based strategy for recording from unrestrained zebrafish would provide a powerful tool

for monitoring the activity of distinct neural populations during natural behavior11.

Two features of fluorescence seriously hinder the strategy described above: most

fluorescent reporters have moderate baseline emission intensity (when unbound to Ca2+)

and, regardless of this basal fluorescence, auto-fluorescence in biological tissue will

provide a significant background signal. With these sources of background emission, any

changes in excitation or detection efficiency caused by fish motion will produce

substantial changes in detected fluorescence, vastly exceeding the small intensity changes

dependent on neural Ca2+. Even with a restrained preparation, motion artifacts, such as

those resulting from a tactile stimulus, induce so much noise that fluorescence recordings

during these periods must be discarded12. In addition to noise in the background signal,

the use of fluorescent reporters requires an intense, visible excitation light to

homogenously fill the behavior arena; this will disrupt any assay of visual behavior.

Given these above constraints, we can list the characteristics of an ideal optical Ca2+

reporter for use with freely swimming zebrafish: zero background emission at basal

neural Ca2+ levels, excitation independence, genetically-encoded, non-toxic to neurons,

stable at room temperature, fast Ca2+-sensing kinetics, and sensitive to the large dynamic

range in Ca2+-concentrations achieved by neurons. Conveniently (perhaps amazingly?),

each of these criteria is met by the fascinating protein Aequorin, a Ca2+-dependent

bioluminescent enzyme produced in nature by the jellyfish Aequorea Victoria13.

3.2: Aequorin and its use as neural Ca2+-reporter

Aequorin, upon binding calcium, catalyzes the oxidation of a cofactor that results

in the release of a blue photon14. Soon after its discovery, purified Aequorin was employed as an optical indicator of intracellular Ca2+in many cell-types including

neurons15, 16. The cloning of the Aequorin gene17, 18 provided researchers with the ability

to target its expression and the molecule received much use as a genetically-encoded

Ca2+-sensor19-21. However, the advent of synthetic fluorescent reporters, used with

conventional imaging setups rather than expensive low-light detectors, contributed to an

otherwise inexplicable, yet conspicuous decrease in Aequorin’s popularity as an optical

Ca2+-reporter.

In jellyfish, Aequorin naturally occurs as a complex with green-fluorescent

protein (GFP), and via a process termed chemiluminescence resonance energy transfer,

the energy from the cofactor oxidation is transferred to GFP and results in the emission of

green light22. Interestingly, the efficiency of Ca2+-dependent photoemission from

Aequorin is enhanced when associated with GFP (from 10% to 90%) which inspired the production of a GFP-Aequorin fusion (GA) 23. GA retains the fast kinetics of Aequorin

(6-30ms rise time24) and sensitivity to Ca2+-concentrations ranging from 100nM-10uM25, on par with the best synthetic Ca2+ sensors. Additionally, the associated GFP provides a

fluorescent tag that can be imaged with conventional methods to characterize Aequorin

expression. The development of GA, with its improved signal efficiency, fostered new

interest in Aequorin as a Ca2+-reporter in neurons26, 27.

3.3: Using GFP-Aequorin in freely-swimming larval zebrafish

Our goal of detecting neural activity in a freely swimming zebrafish required that

the reporter produce no background signal, and at intracellular resting Ca2+ levels

Aequorin mediated luminescence is effectively zero23. Aequorin’s excitation-

independence, zero-background, high-sensitivity to Ca2+, and fast kinetics, have allowed

the design of an Aequorin-based system for measuring neural activity in freely behaving fish. However, some caveats remain:

Unlike fluorescent reporters, an Aequorin molecule will emit only a single photon with each Ca2+-binding event and the bioluminescent signal from an individual neuron is

thus expected to be small. However, photomultiplier tubes are capable of detecting

individual photons and Aequorin assays have employed this technology to monitor Ca2+- dependent events28.

Although independent of excitation light, all bioluminescent enzymes

(luciferases) require a cofactor (luciferin) that is responsible for light-emission22. In the case of Aequorin, Ca2+-dependent bioluminescence is only possible upon reconstitution

with the luciferin coelenterazine. Coelenterazine, a small molecule for which the

biosynthetic pathway is unknown22, must be delivered to the intracellular Aequorin.

Fortunately, it readily permeates lipid membranes29 and under the correct conditions,

described in section 3.4, will be absorbed directly from the surrounding water. In fact,

many coelenterates are not thought to synthesize their own coelenterazine, but instead

acquire it via their diet30.

3.4: Neuroluminescent zebrafish

Ca2+-dependent bioluminescence resulting from neural activity,

neuroluminescence, requires expression of Aequorin-GFP (GA) specifically in neurons

and reconstitution of Aequorin with coelenterazine; our strategy is outlined in Figure 3.1

and further described in the methods (section 3.7).

Figure 3.1

Figure 3.1: - Creating a neuroluminescent zebrafish

The neuron-specific expression of GFP-Aequorin (GA) is achieved by injecting single-

cell embryos with a plasmid encoding GA downstream of a neuronal promoter. Germ-

line integration of the GA gene results in a transgenic founder fish with stable expression

(Figure 3.1 continued) of GA driven by the cell-type specific promoter. Apoaequorin is

then reconstituted with coelenterazine to form the luminescent Aequorin by soaking the

entire larva in normal fish water containing 10-40µM of coelenterazine dissolved in

cyclodextrin. After 24hours of reconstitution, neural Ca2+-dependent light emission,

neuroluminescence, can be detected during natural, unrestrained behavior.

The neuron-specific promoter neural-β-tubulin was used to drive GA expression

specifically in neurons. When injected as plasmid-DNA at the single-cell stage, mosaic

neuron specific expression was apparent (Figure 3.2c-d). Stable germ-line integration of

the plasmid resulted in transgenic zebrafish with pan-neuronal expression of GA (Figure

3.2a-b).

Figure 3.2 – Neural specific expression of GFP-Aequorin (GA)

A Bright-field and fluorescence micrographs of a transgenic larval zebrafish (4dpf)

expressing GA pan-neuronally (scale bar: 0.5mm). B A two-photon optical section

through the dorsal midbrain and hindbrain; both somatic and neuropil regions contain GA

(scale bar: 100µm). C-D Two-photon maximum intensity z-projections of the left optic

tectum in two zebrafish larvae that transiently expressed GA following plasmid injection.

The promoter neural- -tubulin specifically targeted GA expression to neurons, which

appear healthy (scale bar: 20µm).

Figure 3.2

Reconstitution of Apoaequorin was achieved by soaking young larvae in

coelenterazine dissolved in 2-hydroxypropyl-b-cyclodextrin (CDX), which is required to

increase coelenterazine’s water solubility31. After 12-36 hours in coelenterazine-CDX,

the fish were removed from the soaking solution and placed into the neuroluminescence

assay; the fish swam in a small chamber (12.5mm) while a large-area (25mm) photon-

counting PMT detected green photons from above the chamber. Meanwhile, an infrared

CCD camera, illuminated with a ring of IR LEDs, monitored fish behavior (Figure 3.3).

Fascinatingly, as the fish swims freely in the dark, its brain emits green light!

Figure 3.3

Figure 3.3 – Monitoring the neural activity of freely behaving zebrafish

A photon-counting PMT is situated directly above a small, transparent behavior chamber.

The highly-sensitive PMT is protected by an IR-blocking filter such that a ring of IR

light-emitting diodes can be used illuminate an infrared-sensitive CCD camera positioned

below the chamber. The spectral separation allows the simultaneous recording of

neuroluminescence signals (200Hz) and the behavior of freely swimming zebrafish larvae

(60Hz).

The neuroluminescence arising from pan-neuronal GA-expressing fish required

both Aequorin and coelenterazine (Figure 3.4), suggesting that fish do not naturally emit green light and that the photons were generated by reconstituted Aequorin.

Figure 3.4

Figure 3.4 – Photo-emission requires reconstituted Aequorin

Groups of 5 pan-neuronal GA-expressing zebrafish larvae (5-6dpf) were tested together

for luminescence emission following 24hours of soaking in normal fish water (green) and

water containing 40µm coelenterazine-CDX (blue). Groups of 5 wild-type zebrafish

larvae (5-6dpf) were similarly tested after soaking in normal fish water (gray) and water

(Figure 3.4 continued) containing 40µm coelenterazine-CDX (orange). Each trace was

offset by 25 counts for clarity (temporal bin size: 50ms). Both Aequorin and coelenterazine were required for luminescence, suggesting that detected photons were generated by reconstituted Aequorin; the events detected for the other conditions arise from dark counts generated at the PMT photocathode.

The pan-neuronal GA transgenic lines were then used to optimize experimental conditions for subsequent neuroluminescence assays targeting a small population of identified neurons. We initially set out to determine whether coelenterazine dissolved with cyclodextrin (CDX) was able to reconstitute Aequorin as well as ethanol or methanol, which previous studies have used as a solubility agent. We found that CDX achieved equivalent, possibly improved results without the potential side-effects of long-

term exposure to an alcohol solution32 (Figure 3.5a). Using CDX, we also found that replenishing coelenterazine every two days would allow the detection of neuroluminescence until at least 11dpf (Figure 3.5b).

Figure 3.5

Figure 3.5 – Reconstitution of Aequorin with coelenterazine dissolved in cyclodextrin

A The average number of photons within a neuroluminescent event comparing pan-

neuronal GA fish soaked in ethanol vs. cyclodextrin. Fish were tested in groups of 5 (n =

30). We are unable to determine whether the increased average event size in the CDX

condition reflects improved reconstitution with Aequorin or decreased health of the

zebrafish exposed to ethanol. B CDX-dissolved coelenterazine was able to reconstitute

Aequorin and sustain neuroluminescence until at least 11dpf (the last day tested). The

coelenterazine bath solution was replaced every 2 days.

Neuroluminescent signals detected from pan-neuronal GA larval fish were

reliable (Figure 3.6a), fast (Figure 3.6b), and, in stark contrast to bleaching prone fluorescence-based methods, could be recorded continuously for days (Figure 3.6c).

Figure 3.6

Figure 3.6 – Neuroluminescence signals are consistent, fast, and stable for long periods

A Neuroluminescent signals evoked by an electromagnetic tap device striking the

behavior chamber. Responses from an individual pan-neuronal GA larva (4dpf) with low

(Figure 3.6 continued) spontaneous activity demonstrate a consistent signal amplitude

evoked by the repeated vibration stimulus. B The neuroluminescent events seen in A are

aligned relative to the signal onset. The signal peak is reached within 15-20ms and then

decay with a 30ms time constant. C Portions of a neuroluminescence recording acquired

at 9:00PM and 9:00AM on the following day. A 6dpf larva, despite being removed from

the coelenterazine bath for 12 hours during neuroluminescence recording, continued to

produce large spontaneous signals.

In summary, we are able to detect neuroluminescence from freely swimming

zebrafish larva expressing GFP-Aequorin pan-neuronally. The signals require

coelenterazine, have fast kinetics, and can be recorded for days. However, the most

powerful application of this technique will require expressing GA specifically in

identified neural populations and thus allow directly relating the activity of different

groups of neurons to the spectrum of zebrafish behaviors.

3.5: Genetically-targeted recordings from hypocretin neurons during behavior

The hypocretin/orexin (HCRT) system has been implicated in the control of

arousal in mammals33. Disruption of HCRT signaling, as happens for dogs and mice in

which the HCRT receptor is mutated34, 35, produces symptoms similar to those of human

narcolepsy, a disorder characterized by the sudden, undesired onset of sleep. Electrical

recording from HCRT+ neurons in rodents have found that these cells are active during

periods of wakefulness as well as transiently during phasic REM sleep36, 37, and specific

stimulation of HCRT neurons in mice via the optically-gated channelrhodopsin-2, led to

an increased probability of awakening from slow-wave sleep38. In non-mammalian

vertebrates, the limited evidence has suggested a similar role for HCRT in the regulation

of arousal state. For example, zebrafish for which HCRT over-expression was under the

control of a heat-shock promoter maintained much higher rates of activity following

induction39. However, it still remains to be seen whether the activity of HCRT+ neurons

in zebrafish is associated with periods of arousal, as has been observed in mammals.

We set out to test whether neuroluminescence could report the activity of HCRT+ neurons by specifically expressing GA in these cells (Figure 3.7). A promoter element identified in the lab of Alex Schier39 was used to generate stable transgenic zebrafish with

GA expression specifically in HCRT+ neurons.

Figure 3.7

Figure 3.7: - GA Expression in HCRT+ Neurons

A Specific expression of GA in the HCRT+ neurons of a transgenic, 4dpf zebrafish larva.

The maximum intensity projection of two-photon image sections highlights the long,

dorsal-caudal projecting axons (inverted brightness, scale bar: 25µm). B The same

HCRT+ neurons imaged with a wide-field epi-fluorescence microscope clarifies their

position within the posterior diencephalon.

The HCRT+ neurons represented a significant challenge for the

neuroluminescence assay as there are only ~16 such neurons in the posterior

diencephalon of a 4-6dpf larva and they are deep below the dorsal surface (>300µm).

Therefore, as a means of increasing signal size, we reconstituted Aequorin with an analog of native coelenterazine, coelenterazine-h, shown to confer greater Ca2+-sensitivity to the

luminescent enzyme24. Zebrafish larvae expressing GA in HCRT+ neurons were soaked from 3dpf and neuroluminescence was measured starting in the evening on subsequent

days.

After exposure to a light-dark cycle, zebrafish will maintain a circadian

fluctuation in their rate of spontaneous swimming under constant lighting conditions40. In the constant dark conditions of the neuroluminescence assay, larvae were found to increase their rate of swim bouts each morning, shortly after the time of lights-ON in the fish rearing facility. During these morning-active periods, as well as other periods of increased activity during the night, we observed an increase in neuroluminescent signals from HCRT+ neurons (Figure 3.8).

Figure 3.8

Figure 3.8 – Overnight behavior and neuroluminescence recording of HCRT+ neurons

A zebrafish larva (4dpf) swimming freely in the neuroluminescence assay demonstrates

periods of higher spontaneous swimming rates (top panel: the red line shows the measured motion (gray trace, 1Hz) smoothed with a 10 minute sliding window). The longest active period occurs soon after the Lights-ON time (arrow) of the rearing facility.

Neuroluminescence events primarily occur during these periods of higher swimming activity (lower panel).

With so few cells specifically expressing GFP-Aequorin, we were concerned that these transient events could arise from the collective activation of many other neurons in which a small amount of Aequorin was expressed do to leaky expression from our promoter. Therefore, before we will address any correspondence between the observed

neuroluminescence and behavior of the zebrafish, we sought to determine whether the

observed signals were actually originating in the HCRT+ neurons. The existing assay

provides no spatial information about the origin of detected photons, but camera-based detectors exist that are able to detect the arrival of individual photons. Using an

intensified-CCD camera, we imaged restrained zebrafish and compared the origins of

neuroluminescence to the location of Aequorin expression reported by the fluorescence of the tethered GFP (Figure 3.9a-c). The quantum efficiency of the collection optics and photon-counting camera were low (~2%), and we used pentylenetetrazole (PTZ) to

induce epileptic events and drive strong activity waves throughout the zebrafish brain41.

This pharmacological stimulation produced transient events (Figure 3.9b), during which

~90% of the emitted photons came from an ROI surrounding the HCRT+ neurons

(Figure 3.9d). Photons appearing to originate from elsewhere within the fish were not significantly more frequent than the background rate generated by dark-counts in the detector. However, there was a small increase that could be due to photons generated in the processes of the HCRT+ neurons, which extend caudally towards the spinal cord

(Figure 3.7).

Figure 3.9

Figure 3.9 – Photons are emitted by the GA-Expressing HCRT+ neurons

A Fluorescence image acquired of GA-HCRT+ zebrafish restrained beneath an

intensified CCD camera (inverted intensity scale). B When epileptic-like neural activity

is induced by the application of pentylenetetrazole (PTZ), transient increases in the total

number photons were detected. To determine where the photons originated during these

transient events, we plot the position of detected photons arriving during these events

(Figure 3.9 continued) (C). The majority of photons arrive from the region of the

HCRT+ neurons; the spread is likely caused by scattering in the biological tissue while

the homogenous background signal results from dark counts at the detector. D The

number of counts arriving from within three ROIs (see inset): the HCRT cells (red), the

rest of the fish, excluding the HCRT cells (blue), and the remaining background (cyan).

The slight increase in photons arriving from the rest of the fish above the background

dark counts, when normalized for ROI size, may represent photons originating from the

processes of the HCRT neurons. However, after adjusting for the dark count signal, we

observe that >90% of photons arrive from HCRT region.

Having shown that neuroluminescent signals detected from GA-HCRT+ zebrafish

likely originate from the targeted cells, we then asked how a Ca2+ event in these neurons

is related to the rate of spontaneous swimming of a larval zebrafish. We monitored

individual zebrafish between the ages of 4-7dpf, beginning in the evening (7-9pm) and

continuing for at least 16 hours. In an effort to reduce the amount of background light

bleeding into our photon-counter though the IR blocking filter, we strobed the IR

illumination at 1Hz for 20ms, and thus removed 98% of this bleed-through noise. With a

1Hz acquisition rate, we classified our behavior recording into active seconds and

inactive seconds based upon whether the fish had moved since the last frame acquisition.

For each neuroluminescence event, we extracted the 30 minutes of swimming activity

surrounding the peak of the event; the average fish activity surrounding a

neuroluminescence signal in HCRT+ neurons (1617 events, 12 fish) is shown in Figure

3.10.

Figure 3.10

Figure 3.10 –Average swimming activity surrounding a HCRT+ neuroluminescent event

A The average swimming activity in the 30 minutes preceding and following

neuroluminescence events in the HCRT+ neurons (black trace, 1617 events, 12 fish). The

average activity around a randomly selected time point (red trace) provides an estimate of

the baseline spontaneous activity. The large central peak suggests that HCRT signals

coincide with a swim event and because swim events tend to occur during periods of

increased swimming (Figure 3.8)., the ~10 minutes of heightened activity surrounding

the HCRT event would reflect mostly the autocorrelation within zebrafish behavior.

However, the subtle persistent increase in activity following a HCRT event was

intriguing. B To quantify the increase in swimming activity following a HCRT event, we

(Figure 3.10 continued) pooled the activity in the minute preceding and minutes

following the neuroluminescence, excluding the 10seconds immediately before and after

the signal. A small but significant increase (*: p < 0.05) was found in the 1 minute

following (+1) with respect to the 1 minute preceding (-1) a neuroluminescent event in the population of HCRT+ neurons. The dashed blue-line denotes the average activity per minute for the 60 minutes surrounding a HCRT event.

A few features of the average zebrafish activity surrounding a HCRT event are interesting. First, a sharp peak in activity at the time of the HCRT event indicates that the neuroluminescent signal may coincide with a swim bout. High speed analysis of behavior is necessary to address this correlation in more detail, and the results from fish that we were able to monitor at 60Hz are discussed below. Second, HCRT events occur during periods of increased swimming activity relative to the baseline rate. This result can be explained by the fact that zebrafish exhibit long periods of increased activity (Figure

3.8), and a similar shape is seen in the autocorrelation of the zebrafish’s behavior recording. More interesting, however, is the small, but significant increase in activity- level following a HCRT event. Although consistent with the hypothesized role of HCRT in mammals and the effects of over-expressing the HCRT peptide in zebrafish, we cannot claim that HCRT+ neurons are causally-related to this increase in activity-level without additional experiments perturbing (stimulating or ablating) these neurons. However, the correspondence with previous results is reassuring for future studies of specific neural populations whose association to behavior is entirely unknown.

Another compelling feature of the neuroluminescence signals from HCRT+

neurons is presented in Figure 3.11a, b. It is clear that for the individual shown, HCRT

events can be easily separated into two distinct amplitude classes. We found that HCRT

signals could be similarly classified for most larvae (n = 8 of 12). Although there was no

strong difference between each amplitude class in their relationship to spontaneous

swimming activity in the minutes surrounding the signal, an interesting difference

between the large and small HCRT events was detected when zebrafish were monitored

at 60Hz under increased IR illumination.

Figure 3.11 – HCRT events of distinct amplitudes relate to different swim latencies

A Neuroluminescence recording demonstrating the two distinct amplitude classes of

HCRT signals, similar results were seen for 8 of 12 individuals. Manually determined

thresholds (400 photons/50ms in A) were used to classify the large (blue) and small (red)

amplitude signals. The classified signals from an entire recording, for individual shown in

A, are aligned in B; the thick lines indicate the mean signal within each class. C

Individual swim events following different amplitude HCRT signals (red – small event,

blue – large event). The small HCRT signals are followed by short-latency swim bouts

whereas the large signals precede bouts with a significantly longer latency (+25ms) and

often higher-velocity. This is summarized in D where the response latency is plotted for

all the small and large HCRT-events from 4 individuals (133 large events, 404 small

events).

Figure 3.11

Following a small HCRT event, larvae initiated a swim event that occurred

reliably within about 30-40ms. However, the swim event following a large HCRT event

was usually of higher peak velocity and had a delay of about 65ms (Figure 3.11c). Given

that 60Hz acquisition is not fast enough to accurately characterize the detailed kinematics

of the responses, we focused primarily on the latency of the behavior relative to the onset

of the neuroluminescence signal rather than the velocity. When measured for the 537

neuroluminescence events detected from the zebrafish monitored at high-speed, we found

a conspicuous difference (~25ms) in the response latency for swims following a small

versus a large HCRT signal (Figure 3.11d). The significance of this difference is unclear,

but suggests two distinct modes of activating the population of HCRT+ neurons. Whether

the small signals result from activating a subset of the labeled neurons or more weakly

activating the entire population is unknown, however, we are encouraged by the

quantitative properties of neuroluminescence demonstrated by this result. Not only are we

able to isolate distinct response amplitudes in a freely swimming zebrafish, but the two

response classes were reliable predictors of distinct behaviors. As we extend this

technique to different populations of neurons, exploiting the well-develop genetics of the

zebrafish to target GA expression, we can now expect to resolve not only when the

population is active during natural behavior but possibly the magnitude of these

activations; another valuable source of information to assist in decoding how the neurons

of the brain control the behavior of a vertebrate.

3.6: Future Directions

Building upon the success of the neuroluminescence technique, we will pursue

targeting GA expression to other populations of genetically-identified neurons. Using an available promoter for the serotonergic dorsal Raphe Nuclei42 as well as the

dopaminergic system43, I hope to investigate the role of these neurotransmitters in

regulating the feeding behavior of larval zebrafish. Specifically, we are interested in the

neural response resulting from a successful prey capture event that might serve as a

reward signal instructing plasticity in the zebrafish hunting circuit. However, these tools

should extend easily to the complete set of known neuromodulators and I expect that the neuroluminescence approach, coupled with genetic strategies for stimulating and ablating

these same neurons, will provide an important tool for determining how the brain

choreographs and directs the complex behavioral patterns of a developing zebrafish.

Further technical developments will also improve and advance the applicability of

this technique. Currently, our experiments have required studying behaviors that occur

naturally in darkness. The highly visual zebrafish, however, begs for assays of neural

activity during unrestrained visual behavior such as prey tracking44 and the recognition of

con-specifics45.The sensitivity of photon-counting detectors and the imperfect nature of

spectral filters prevents the use of visual stimuli during low-light level detection. We are pursuing instead a time-gated strategy that will allow us to illuminate with visible light while the detector is gated-OFF and then remove the illumination for period during which the detector is gated-ON and able to detect neuroluminescence. It is possible to rapidly alternate between illumination and detection (easily faster than 100Hz) providing the fish, and our slow human eyes, with the impression of normal visible illumination while

reserving 90% of our sample time for neuroluminescence detection. Needless to say, we are excited about these possibilities, but unfortunately, I must leave you here and get to

work on them. Cheers.

3.7: Methods

Zebrafish

Zebrafish (Danio rerio), of the mitfa-/- (nacre)46 strain were optically transparent

despite intact ocular pigmentation. Zebrafish were maintained on a 14 hr light /10 hr dark

cycle and fertilized eggs were collected and raised at 28 °C. Embryos were kept in E3

solution (5 mM NaCl, 0.17 KCl, 0.33mM CaCl2, 0.33 mM MgSO4). For reconstitution

with coelenterazine, 5-10 fish were transferred into 2ml of E3 solution containing a final

concentration of 40 µM native Coelenterazine (Invitrogen, USA or Biotium,USA) or

Coelenterazine-h (Biotium,USA) (all stock solutions at 10 mM dissolved in 45% 2- hydroxypropyl-b-cyclodextrin (Invitrogen) were kept at -80 °C to avoid auto-oxidation).

All experiments were approved by Harvard University’s Standing Committee on the Use

of Animals in Research and Training.

Vector construction and transgenic lines

GAv1 coding sequence was subcloned via PCR into Neurobetatubulin expression

vector into AgeI and NotI site (kind gift of P. Krieg) resulting in tol2:Nbt:GA5v1:tol2.

UAS:GA5v1 was constructed by sub-cloning the coding sequence behind the UAS:Eb1

sequence in UAS: Dsred Express -1 expression vector by replacing DsRed by blunt end

insertion at the AgeI and Not I site. For specific expression of GA5v1in Hypocretin

positive neurons the 2.4 kb fragment of Fugu rubripes genomic DNA containing 2 kb of

upstream sequence, the putative hcrt first exon, intron, and the beginning of the second

exon was placed upstream of GA5v1 to yield tol2:HCRT: GA5v1:tol2.39 Plasmid DNA

(20ng/µl) in 100 mM KCl was then injected at the single cell stage into the zygote for

transient expression. For the generation of transgenic zebrafish, tol2:Nbt:GA5v1:tol2 and

tol2:HCRT: GA5v1:tol2 was co-injected with tol2 transposase mRNA . Injected

embryos were grown in E3solution and screened for expression at 2-5dpf and positive

individuals were grown to adulthood and out-crossed to wild type zebrafish. Progeny

(F2) of this cross were again screened for expression at 2-5dpf, and transgenic founder

fish with best expression profile were identified. Most experiments were performed with

progeny (F3) of crosses of stable F2 transgenic and wild-type zebrafish.

Bioluminescence detection

In a custom built light tight box, a zebrafish swam freely in 1ml of E3 solution

contained in a custom recording chamber enclosed from top and bottom with cover glass.

To avoid bubble formation in long term or overnight recordings, silicone sealant was

used to keep the upper coverslip in position. The chamber was mounted directly under a

3885 PMT photon-counter (Hamamatsu, Japan), to maximize the angle of light detected

by the PMT. An infrared LED ring light (Advanced Illumination, USA) was placed

above the recording chamber such that the zebrafish could be illuminated with only

minimally introducing light into the PMT. Depending on the experiment, single fish or

groups up to 10 were included in the chamber. In some experiments, PTZ was added to

induce epileptic events. To evoke escape behaviors and the corresponding neural

response, zebrafish were stimulated with a gentle tap to the recording chamber with a

computer controlled, electromagnetic tapping device – think telegraph.

In-vivo two-photon imaging

Prior to imaging, selected zebrafish were anaesthetized using 0.02% Tricaine in

E3 and embedded in low melting point Agarose LMA (1.2w/v%). Tricaine was removed

and using a pulled glass pipette, bungarotoxin was injected in to the ventral region to

prevent movement artifacts. Expression profile was imaged at high resolution with a

custom built two-photon microscope using a Ti:Sapphire laser (920 nm) and a 20x water

immersion objective with a 0.95 numerical aperture (Olympus). All data acquisition was

done with custom Labview (National Instruments, USA) software.

Intensified CCD Camera Imaging

In a modified light tight box (Xenogen, USA), a custom designed microscope was

built to allow magnification of the zebrafish brain onto an image intensifier, which

amplifies light via an electron multiplication stage and output onto a phosphor screen that

is imaged by a CCD camera. A USB frame-grabber (Sensoray, USA) was used to acquire

raw images from the CCD, and custom Matlab (Mathworks, USA) software was used to

detect single photon positions. Zebrafish were prepared as for two-photon imaging

(described above).

3.8: References

1. Dombeck, D.A., Khabbaz, A.N., Collman, F., Adelman, T.L. & Tank, D.W. Imaging large-scale neural activity with cellular resolution in awake, mobile mice. Neuron 56, 43-57 (2007).

2. Helmchen, F., Fee, M.S., Tank, D.W. & Denk, W. A miniature head-mounted two-photon microscope. high-resolution brain imaging in freely moving animals. Neuron 31, 903-912 (2001).

3. Fee, M.S. Active stabilization of electrodes for intracellular recording in awake behaving animals. Neuron 27, 461-468 (2000).

4. Lee, A.K., Manns, I.D., Sakmann, B. & Brecht, M. Whole-cell recordings in freely moving rats. Neuron 51, 399-407 (2006).

5. Fetcho, J.R. & O'Malley, D.M. Visualization of active neural circuitry in the spinal cord of intact zebrafish. J.Neurophysiol. 73, 399-406 (1995).

6. Fetcho, J.R. & O'Malley, D.M. Imaging neuronal networks in behaving animals. Current Opinion in Neurobiology 7, 832-838 (1997).

7. Orger, M.B., Kampff, A.R., Severi, K.E., Bollmann, J.H. & Engert, F. Control of visually guided behavior by distinct populations of spinal projection neurons. Nat Neurosci 11, 327-33 (2008).

8. Niell, C.M. & Smith, S.J. Functional imaging reveals rapid development of visual response properties in the zebrafish tectum. Neuron 45, 941-951 (2005).

9. Higashijima, S.I., Masino, M.A., Mandel, G. & Fetcho, J.R. Imaging neuronal activity during zebrafish behavior with a genetically encoded calcium indicator. J.Neurophysiol. 90, 3986-3997 (2003).

10. Ramdya, P., Reiter, B. & Engert, F. Reverse correlation of rapid calcium signals in the zebrafish optic tectum in vivo. J Neurosci Methods 157, 230-7 (2006).

11. Luo, L., Callaway, E.M. & Svoboda, K. Genetic dissection of neural circuits. Neuron 57, 634-60 (2008).

12. O'Malley, D.M., Kao, Y.H. & Fetcho, J.R. Imaging the functional organization of zebrafish hindbrain segments during escape behaviors. Neuron 17, 1145-1155 (1996).

13. Shimomura, O. A short story of aequorin. Biol Bull 189, 1-5 (1995).

14. Daunert, S. & Deo, S.K. Photoproteins in bioanalysis (Wiley-VCH, Weinheim, 2006).

15. Ross, W.N. Changes in intracellular calcium during neuron activity. Annu Rev Physiol 51, 491-506 (1989).

16. Smith, S.J. & Zucker, R.S. Aequorin response facilitation and intracellular calcium accumulation in molluscan neurones. J Physiol 300, 167-96 (1980).

17. Inouye, S. et al. Cloning and sequence analysis of cDNA for the luminescent protein aequorin. Proc Natl Acad Sci U S A 82, 3154-8 (1985).

18. Prasher, D., McCann, R.O. & Cormier, M.J. Cloning and expression of the cDNA coding for aequorin, a bioluminescent calcium-binding protein. Biochem Biophys Res Commun 126, 1259-68 (1985).

19. Chiesa, A. et al. Recombinant aequorin and green fluorescent protein as valuable tools in the study of cell signalling. Biochem J 355, 1-12 (2001).

20. Nakai, J. & Ohkura, M. Probing calcium ions with biosensors. Biotechnol Genet Eng Rev 20, 3-21 (2003).

21. Knight, H. & Knight, M.R. Recombinant aequorin methods for intracellular calcium measurement in plants. Methods Cell Biol 49, 201-16 (1995).

22. Hastings, J.W. & Johnson, C.H. Bioluminescence and chemiluminescence. Methods Enzymol 360, 75-104 (2003).

23. Baubet, V. et al. Chimeric green fluorescent protein-aequorin as bioluminescent Ca2+ reporters at the single-cell level. Proc.Natl.Acad.Sci.U.S.A 97, 7260-7265 (2000).

24. Shimomura, O., Musicki, B., Kishi, Y. & Inouye, S. Light-emitting properties of recombinant semi-synthetic aequorins and recombinant fluorescein-conjugated aequorin for measuring cellular calcium. Cell Calcium 14, 373-8 (1993).

25. Curie, T., Rogers, K.L., Colasante, C. & Brulet, P. Red-shifted aequorin-based bioluminescent reporters for in vivo imaging of Ca2 signaling. Mol Imaging 6, 30-42 (2007).

26. Rogers, K.L. et al. Visualization of local Ca2+ dynamics with genetically encoded bioluminescent reporters. European Journal of Neuroscience 21, 597-610 (2005).

27. Martin, J.R., Rogers, K.L., Chagneau, C. & Brulet, P. In vivo bioluminescence imaging of Ca signalling in the brain of Drosophila. PLoS ONE 2, e275 (2007).

28. Robinson, K.R., Keating, T.J. & Cork, R.J. Inexpensive techniques for measuring [Ca2+]i changes using a photomultiplier tube. Methods Cell Biol 40, 287-303 (1994).

29. Shimomura, O. Membrane permeability of coelenterazine analogues measured with fish eggs. Biochem J 326 ( Pt 2), 297-8 (1997).

30. Haddock, S.H., Rivers, T.J. & Robison, B.H. Can coelenterates make coelenterazine? Dietary requirement for luciferin in cnidarian bioluminescence. Proc Natl Acad Sci U S A 98, 11148-51 (2001).

31. Teranishi, K. & Shimomura, O. Solubilizing Coelenterazine in Water with Hydroxypropyl-β-cyclodextrin. Biosci Biotechnol Biochem 61, 1219-1220 (1997).

32. Matsui, J.I., Egana, A.L., Sponholtz, T.R., Adolph, A.R. & Dowling, J.E. Effects of ethanol on photoreceptors and visual function in developing zebrafish. Invest Ophthalmol Vis Sci 47, 4589-97 (2006).

33. Sakurai, T. The neural circuit of orexin (hypocretin): maintaining sleep and wakefulness. Nat Rev Neurosci 8, 171-81 (2007).

34. Lin, L. et al. The sleep disorder canine narcolepsy is caused by a mutation in the hypocretin (orexin) receptor 2 gene. Cell 98, 365-76 (1999).

35. Chemelli, R.M. et al. Narcolepsy in orexin knockout mice: molecular genetics of sleep regulation. Cell 98, 437-51 (1999).

36. Mileykovskiy, B.Y., Kiyashchenko, L.I. & Siegel, J.M. Behavioral correlates of activity in identified hypocretin/orexin neurons. Neuron 46, 787-98 (2005).

37. Takahashi, K., Lin, J.S. & Sakai, K. Neuronal activity of orexin and non-orexin waking-active neurons during wake-sleep states in the mouse. Neuroscience 153, 860-70 (2008).

38. Adamantidis, A.R., Zhang, F., Aravanis, A.M., Deisseroth, K. & de Lecea, L. Neural substrates of awakening probed with optogenetic control of hypocretin neurons. Nature 450, 420-4 (2007).

39. Prober, D.A., Rihel, J., Onah, A.A., Sung, R.J. & Schier, A.F. Hypocretin/orexin overexpression induces an insomnia-like phenotype in zebrafish. J Neurosci 26, 13400-10 (2006).

40. Cahill, G.M., Hurd, M.W. & Batchelor, M.M. Circadian rhythmicity in the locomotor activity of larval zebrafish. NeuroReport 9, 3445-9 (1998).

41. Baraban, S.C., Taylor, M.R., Castro, P.A. & Baier, H. Pentylenetetrazole induced changes in zebrafish behavior, neural activity and c-fos expression. Neuroscience 131, 759-768 (2005).

42. Lillesaar, C., Tannhauser, B., Stigloher, C., Kremmer, E. & Bally-Cuif, L. The serotonergic phenotype is acquired by converging genetic mechanisms within the zebrafish central nervous system. Dev Dyn 236, 1072-84 (2007).

43. Wen, L. et al. Visualization of monoaminergic neurons and neurotoxicity of MPTP in live transgenic zebrafish. Dev Biol 314, 84-92 (2008).

44. McElligott, M.B. & O'Malley D, M. Prey tracking by larval zebrafish: axial kinematics and visual control. Brain Behav Evol 66, 177-96 (2005).

45. Sovrano, V.A. & Andrew, R.J. Eye use during viewing a reflection: behavioural lateralisation in zebrafish larvae. Behav Brain Res 167, 226-31 (2006).

46. Lister, J.A., Robertson, C.P., Lepage, T., Johnson, S.L. & Raible, D.W. nacre encodes a zebrafish microphthalmia-related protein that regulates neural-crest- derived pigment cell fate. Development 126, 3757-3767 (1999).