The Neural Basis of Head Direction and Spatial Context in the Insect Central Complex

Home , Flesh fly, Head direction cell

THE NEURAL BASIS OF HEAD DIRECTION AND SPATIAL CONTEXT IN THE INSECT CENTRAL COMPLEX

by ADRIENN G. VARGA

Submitted in partial fulfillment of requirements For the degree of Doctor of Philosophy

Advisor: Dr. Roy E. Ritzmann

Department of Biology CASE WESTERN RESERVE UNIVERSITY May 2017

CASE WESTERN RESERVE UNIVERSITY SCHOOL OF GRADUATE STUDIES

We hereby approve the thesis/dissertation of Adrienn G. Varga Candidate for the Doctor of Philosophy degree*.

Committee Chair: Hillel J. Chiel

Committee Member: Roy E. Ritzmann

Committee Member: Mark A. Willis

Committee Member: Jessica L. Fox

Committee Member: David Friel

Date of Defense: November 18th, 2016

*We also certify that written approval has been obtained for any proprietary material contained therein.

Dedication

For my Family

Table of Contents

Thesis Summary ...... 1

Chapter 1: Introduction ...... 3

Mammalian navigation circuits ...... 6

Adaptive navigation ...... 6

Head direction cells ...... 10

Sensory cues underlying the head direction signal ...... 13

Head direction network ...... 16

Relationship with other networks in the navigation system ...... 17

Neural control of insect navigation in the central complex ...... 19

Central complex anatomy ...... 19

Cellular composition of the central complex ...... 21

Directional sensory signal processing in the central complex ...... 23

Selection and maintenance of behavior ...... 26

Visual and spatial memory ...... 29

Visual pattern recognition ...... 29

Detour paradigm and spatial working memory ...... 31

Visual place learning ...... 32

The physiological correlates of orientation in the central complex ...... 33

Local Field Potentials ...... 35

The origin and function of local field potentials ...... 35

Local field potentials in mammalian navigation circuits ...... 37

Rats rely on theta rhythm to synchronize navigational circuits ...... 38

Bat navigational network activity ...... 39

Local field potentials in the arthropod brain ...... 41

Oscillations in the insect brain ...... 41

Crayfish central complex network activity...... 42

Summary ...... 44

Chapter 2: Cellular Basis of Head Direction and Contextual Cues in the Insect Brain ...... 47

Summary ...... 48

Introduction ...... 49

Materials and Methods ...... 53

Surgical procedures ...... 53

Recording procedures ...... 54

Spike sorting and data analysis ...... 57

Results ...... 61

Central complex neurons encode head direction ...... 61

Tuning characteristics of head direction encoding neurons ...... 62

Head direction encoding CX neurons rely upon allothetic and/or idiothetic cues 67

Head direction coding persists even in the absence of visual landmarks ...... 73

Central complex units encode rotation direction history ...... 75 ii

Discussion ...... 79

Chapter 3: Modulation of Central Complex Local Field Potentials by Head Direction and Spatial Context...... 85

Summary ...... 86

Introduction ...... 87

Materials and Methods ...... 92

Experimental Procedures ...... 92

Experimental subjects ...... 92

Recording procedures ...... 93

Data analysis and statistics ...... 95

Results ...... 98

Description of spontaneous LFPs in the cockroach central complex ...... 98

Head direction modulation of central complex network activity ...... 101

Delta-band activity encodes head direction independent of the underlying sensory cues ...... 104

Sensory context does not affect relative response magnitudes in the delta-band 106

Sensory context modulates the average power of central complex network activity in the theta-, beta- and gamma-bands, but not in the delta-band ...... 111

Discussion ...... 114

Chapter 4: Conclusion ...... 120

Considerations of the neural basis of navigation ...... 121

iii

Broad discussion of major findings and their significance to the field of neurobiology ...... 123

Single neurons and LFPs encode head direction in the insect CX ...... 124

Single neurons and LFPs encode spatial context cues in the insect CX ...... 128

Conclusions and Future Directions ...... 130

Appendix ...... 131

Bibliography ...... 135

List of Tables

Chapter 2:

Table 2.1: . Landmark rotation experiment results indicate that CX units utilize five sensory strategies when encoding head direction...... 70

Chapter 3:

Table 3.1: Summary of LFP analysis results...... 114

List of Figures

Chapter 1:

Figure 1.1: Schematic illustration of the neural circuitry underlying mammalian navigation and context discrimination...... 12

Figure 1.2: Schematic illustration of the cockroach brain and the central complex (CX)...... 21

Chapter 2:

Figure 2.1: Experimental design and paradigms to test head direction coding in the cockroach...... 55

Figure 2.2: CX units encode head direction by changes in firing rate...... 64

Figure 2.3: Tuning characteristics of angle-modulated CX units...... 67

Figure 2.4: Visual landmark position determines head direction coding...... 72

Figure 2.5: CX units encode head direction in the absence of visual landmarks or any visual input...... 75

Figure 2.6: Past rotation direction affects CX unit firing rate during the stationary epochs...... 78

Chapter 3:

Figure 3.1: Example spontaneous LFPs in the cockroach CX...... 99

Figure 3.2: Average spontaneous LFP in the CX is dominated by delta-band activity...... 101

Figure 3.3: Delta-band activity reflects head direction more than any other frequency band...... 103

Figure 3.4: Sensory context and landmark exposure does not affect relative LFP response magnitudes when the animals can initially rely on a landmark...... 109

Figure 3.6: Modulation of average FFT power by sensory context is independent of head direction coding by CX LFPs...... 113

Appendix

Appendix Figure S1: Related to Figure 2.1 Outline of the experimental design showing the order and number of conditions the animals were exposed to...... 131

Appendix Figure S2: Related to Figure 2.2. Locations of recording sites within two neuropils of the cockroach central complex...... 132

Appendix Figure S3: Related to Figure 2.2 and 2.3. Central complex units possess the capacity to represent any angle relative to the animal, just like a compass...... 133

Appendix Figure S4: Related to Figures 2.2, 2.4 and 2.5. CX units encode head direction with similar precision under control conditions with a landmark and in head- covered landmark-naïve animals...... 134

vii

Acknowledgments

I would especially like to thank my advisor, Dr. Roy Ritzmann for his wise advice, continuous support and his unwavering confidence in me. I would also like to thank the members of my committee, Dr. Mark Willis, Dr. Jessica Fox and Dr. David Friel. The invaluable technical assistance of Al Pollack is greatly acknowledged. I also would like to thank Dr. Nick Kathman for the countless inspiring conversations, help with data analysis and support throughout these years. And this could not have been possible without the loving support from my family. I am most grateful for my husband, Dan, for being my endless source of motivation and inspiration, and for providing both scientific and emotional support throughout this part of our journey.

This work was supported by the National Science Foundation [IOS-1120305 and IOS- 1557228 to R.E.R.].

viii

The Neural Basis of Head Direction and Spatial Context in the Insect Central

Complex

Abstract

ADRIENN G. VARGA

A question of wide importance in neuroscience is how the brain controls behavior. How does sensory information get transformed into a spatially organized representation about our current state in the world and how is this abstract representation utilized when producing motor commands that lead to successful navigation? When navigating in a complex environment, all animals must encode information about their position and orientation in a rich sensory environment. In vertebrates this may occur by means of distributed activity across several navigation circuits located in the hippocampal formation.

Arthropods, however, lack a hippocampal formation and thus it is unclear what circuits mediate navigation.

A wide range of studies indicate that the central complex (CX), is not only involved in directional sensory information processing and the control of motor commands, but also plays a role in orientation coding in polarized light guided navigation and landmark orientation. All of these neural mechanisms point in the direction that single neurons in the

CX might be directly involved in head direction coding, as well as other aspects of adaptive navigation.

In the work described in this dissertation I used multi-channel extracellular recording techniques to uncover the neural correlates of head direction coding and spatial context cues in the cockroach CX. Specifically, I used tetrodes to record the activity of single neurons in the CX while the animal was passively rotated around on a platform surrounded by a circular arena (Chapter 2). In the same setting I also recorded local field potentials (LPFs) in the CX to uncover how navigational information modulates the network’s activity in a more global manner (Chapter 3). I found that single units, as well as LFPs in the cockroach CX encode the animal’s head direction relative to a salient visual cue. However, when landmarks are not available to the animal, both single neuron and network-level activity can rely upon idiothetic motion cues to update the animal’s relative heading in a landmark-free setting. In addition to these results, I found that a subpopulation of single neurons and some of the LFP frequency bands encoded the rotation direction history of the animal, a common spatial context cue. These results suggest that the CX navigation circuit is involved in environmental context discrimination processes that might be utilized by spatial memory circuits in the insect brain.

Taken together, these results provide a solid foundation for future studies on the neural basis of adaptive navigation in insects. By placing these results in a wider context of adaptive navigation in all animals and by comparing them to the mechanisms described in mammalian navigation circuits, these data also contribute to a broad comparative approach to understand the general principles of navigation, as well as the diversity of the neural substrates of navigation across evolutionarily distinct animals.

Chapter 1

Introduction

Portions of this material were previously published in the journal Frontiers in Behavioral Neuroscience: Varga, A. G. Kathman, N. D., Martin, J. P., Guo, P., Ritzmann, R. E. (2017)

Successful navigation is a critical component of all animal behaviors. Sea turtles migrate between foraging and nesting grounds (Sale and Luschi, 2009), homing pigeons easily find their way home even from large distances (Walcott, 1996), migrating fish, such as the Atlantic salmon, migrate between saltwater and freshwater at certain stages in their lives (Hansen et al., 1993). Monarch butterflies travel from North America to central

Mexico to mate and lay eggs and when the time comes, the next generation returns to the

North (Reppert et al., 2016). Animals rely upon navigation to find water and food sources, spot potential mates, or locate shelters. Navigation in any environment requires sophisticated brain circuits that are capable of integrating external sensory cues with the internal state of the animal and use this compressed signal to shape behaviors.

Insects perform a variety of outstanding navigational tasks, yet the neural dynamics underlying insect navigation are not well understood. Although the details may vary from insect to insect, brain regions that could control navigational movements would have to contain some common properties. At a minimum, they would need to receive appropriate forms of sensory information for the navigational task and they would have to be able to use that information to influence movement. For many tasks it would also be important to have a sense of where the insect is in its environment and how it got there.

The central complex (CX) is a group of neuropils located in the midline of the insect brain that is thought to be the center for sensorimotor integration that could potentially underlie navigation (Pfeiffer and Homberg, 2014). The CX is innervated by other brain regions carrying preprocessed sensory information of various modalities with directional components. Directional sensory cues encoded by the CX include visual information

(Kathman et al., 2014; Seelig and Jayaraman, 2013), mechanical antennal cues (Guo and

Ritzmann, 2013; Ritzmann et al., 2008), polarized light (Heinze and Homberg, 2007;

Heinze and Homberg, 2009; Heinze and Reppert, 2011; Heinze et al., 2009; Heinze et al.,

2013) and multisensory information (Homberg, 1985; Milde, 1988; Phillips-Portillo,

2012).

A wide range of studies support the CX’s role in directional motor control. Genetic manipulations in Drosophila melanogaster demonstrated the CX’s involvement in the selection and maintenance of locomotion, as well as the generation of asymmetrical limb movements (Martin et al., 1999; Strauss, 2002) . CX lesions in cockroaches lead to turning behaviors and circling in one direction (Ridgel et al., 2007), as well as turning deficits in a

U-shaped track (Harley and Ritzmann, 2010). Tetrode recordings in tethered cockroaches provided physiological evidence for the CX’s role in the directional control of locomotion

(Guo and Ritzmann, 2013). In addition to participating in sensorimotor integration processes, genetic studies in fruit flies have established the CX’s role in spatial memory and place learning (for review see, Pfeiffer and Homberg, 2014).

More recently a study by Martin and colleagues (2015) indicated that the CX supervises directed locomotion and indirectly controls local reflex circuits in the limbs of cockroaches. Another recent paper by Seelig and Jayaraman (2015) showed that a subpopulation of CX cells are involved in navigation by controlling landmark orientation in tethered fruit flies. Major questions still remain regarding the neural correlates of navigation in the CX. Seelig and Jayaraman (2015) used Ca2+ imaging to look at the dendritic responses of a single cell type in the CX. Thus, we do not know whether other populations of cells also contribute to adaptive navigation. Although, their data showed that the observed population of neurons encoded the animal’s head direction relative to a

5 landmark, it is not clear how single neurons contribute to this directional signal.

Additionally, the tethered recordings used in this particular study do not allow for the complete dissection of the underlying sensory cues and their hierarchy in establishing the head direction signal.

The work presented in this thesis stems from research I performed using multichannel extracellular recording techniques to investigate the neural correlates of head direction coding and spatial context cues in the cockroach CX. This technique allowed me to examine neural responses both at the single neuron level and local network level at the same time. My results demonstrate that single neurons, as well as local field potentials indicating network activity in the CX have the capacity to encode head direction and spatial context cues in passively rotated animals. I also extended our knowledge of the underlying sensory processes contributing to the head direction signal by elucidating the sensory hierarchy through manipulations of the animal’s environment.

In this introductory chapter I will provide detailed background information that places my results in the wider context of adaptive navigation in both insects and mammals.

Mammalian navigation circuits

Adaptive navigation

Spatial navigation is one of the most complex behaviors that animals exhibit. It requires the integration of both external and internal sensory information. External sensory cues - also called allothetic cues - are visual, olfactory, auditory and mechanosensory

6 information about the environment. Internal sensory cues – also called idiothetic cues – are derived from self-motion in the form of vestibular cues (or mechanosensory cues), optic flow, proprioceptive feedback and motor efference copy from the limbs. The integrated sensory code representing the current environment of the animal is used in motor centers to induce and shape optimal motor commands that lead to successful navigation.

Navigational tasks, or components of navigational tasks, that keep repeating in similar situations facilitate learning and memory, and the learned spatial information contributes to adaptive navigation. Most behaviors leading to adaptive navigation are goal-directed, whether that goal be finding food (foraging) or shelter (homing). Considering these criteria, the most important steps during adaptive navigation are: (1) sensory information processing (external and internal), (2) integration of sensory information with internal state-dependent information (motivational factors and goals; hormone levels; the animal’s physiological state, for instance: hunger; etc.) and retrieval of relevant memories, (3) selection and initiation of an appropriate behavioral program based on the integrated sensory context and internal state, (4) updates to memory traces and evaluation of behavioral outcomes (Mizumori et al., 2009). Each of these components has the power to impact the function of other components, and thus do not strictly follow each other in the above described order. Our current understanding of adaptive navigation in mammals is that both the basal ganglia (specifically the striatum) and the hippocampal formation process spatial information, such as the inner representation of spatial location, the animal’s head-orientation, navigational task-related rewards and movement (Mizumori et al., 2009; Penner and Mizumori, 2012).

The dorsal striatum is thought to assist the navigational system by evaluating the consequences of behaviors in the current navigational context. As a result of this analysis, planned motor actions can be fine-tuned to appropriately fit the current context. This process, as well as the motor actions approved by the striatum, have spatial components, which suggests that spatial context processing takes place within the striatum. Information about the environment and the animal’s position in this environment can be derived from preprocessed sensory information that arrives at the striatum from sensory areas, other associative areas and the limbic system (McGeorge and Faull, 1989). Lesion studies showed that impairment of the striatum leads to selective spatial deficits, especially during tasks that require learning (Mizumori et al., 2009). Extracellular recordings in freely behaving animals confirmed that some striatal neurons are sensitive to directional motor components of navigation, such as angular velocity and forward walking and navigational context cues, such as a reward’s location (Lavoie and Mizumori, 1994). The striatum, similarly to the hippocampus, contains place cells that are thought to encode the inner representation of the animal’s position in its environment through sensory integration. In addition to place cells, the striatum also contains head direction cells, which encode the head angle or orientation of the animal relative to a landmark (Figure 1.1; Mizumori et al.,

2000).

Because both place cell and head direction cell responses significantly change in rearranged or novel environments, the spatial code in the striatum is thought to be highly context-dependent. This supports the hypothesis that the striatum evaluates behavioral, or in this case navigational consequences and selects the motor actions that can potentially lead to those consequences in a context-dependent manner.

The hippocampal formation is thought to be a spatial context discriminator. Its role in adaptive navigation is to compare the current spatial context to the expected spatial context, which would be achieved through the motor actions approved by the striatum in the basal ganglia. The discrimination process requires access to spatial memory and the ability to detect and encode information about novelty. Novelty in the environment induces exploration or goal-directed navigation, because the current spatial context does not match the expected spatial context, or the animal has not reached the navigational goal yet. This process not only induces navigation, but also facilitates learning and memory (Paulsen and

Moser, 1998). Thus, the hippocampal formation needs to continuously integrate sensory information about the environment, sensory information derived from movement and the current motivational state of the animal as a function of space and time, which translates to the current spatial context of the animal. A hippocampal place cell’s preferred location, the place field, is hypothesized to be the result of the specific spatially and temporally relevant organization of the above listed information. In other words, a place cell’s place field contains an abstract description of the animal’s current spatial context and internal state (Mizumori et al., 2009; Penner and Mizumori, 2012). Thus, the comparison between the current and expected spatial context may be achieved with the help of place cells in the hippocampus (Figure 1.1).

Based on the previous sections, it is clear that navigational processes in all brain regions require at least two types of spatial information. The brain needs information about the animal’s location in an environment (encoded by place cells), and to initiate directed movements it needs information about the current heading of the animal (encoded by head direction cells). In the following section I will discuss the role of head direction cells in

9 adaptive navigation and the head direction network’s connections with other spatial networks, such as hippocampal place cells and entorhinal grid cells (Figure 1.1).

Head direction cells

Head direction cells in the mammalian brain are neurons that are responsible for encoding the animal’s spatial orientation in its proximal environment independent of the animal’s location. They were originally discovered by James Ranck and extensively characterized by Jeffrey Taube (Ranck, 1985; Taube et al., 1990a, 1990b; Taube, 2007).

Head direction cells are located in several brain areas (see Figure 1.1), including, but not limited to the entorhinal cortex, postsubiculum, anterodorsal thalamic nucleus, the dorsal striatum of the basal ganglia and the hippocampal CA1 area (Taube, 2007). Head direction cells found in these brain regions encode spatial orientation in a similar manner, and the reason behind such widespread replication of the orientation signal remains to be elucidated.

Each head direction cell is tuned to a single preferred head orientation, and together the network covers the entire 360° environment like a compass. A head direction cell’s firing rate reaches the maximum when the animal faces the cell’s preferred direction, and as the animal turns away from that angle, the firing rate drops down to near zero almost linearly. The preferred angle range, where the firing rate is above the minimum, or zero, is usually between 60 - 150° and averages around 90°.

The peak firing rate at the preferred angle ranges between 5-120 spikes/s. It is still not known whether the cells exhibiting various firing rates are in any way different from

10 each other. It has also been shown that anterodorsal thalamic head direction cells may exhibit irregular firing patterns even in the same constant environment (Taube, 2010). Head direction cell firing rate is also independent of the head’s pitch or roll within ~90° of the horizontal plane, as well as any kind of ongoing behavior, which usually includes walking.

However, head direction cells in some brain areas also encode angular velocity, which results in increased firing rates at the preferred angle when the animal quickly turns its head through this angle, and slightly decreased activity when the turn is slow or the animal is stationary (Taube and Burton, 1995).

Figure 1.1: Schematic illustration of the neural circuitry underlying mammalian navigation and context discrimination. A: Schematic of a rat brain. Sagittal section where Bregma represents 0 mm (marked by black vertical line). Gray lines indicate the location of sections illustrated in the right side of the panel relative to Bregma. Navigation centers are color coded (all rat brain diagrams were created based upon (Paxinos and Watson, 1997). B: Arrows indicate the direction of communication between brain regions. Brain regions are color coded based on the types of spatial cells that can be found in those locations. The exact roles of the above illustrated structures and the connections within the navigation circuit are described in more detail in the text. Based on (Jankowski et al., 2013; Mizumori et al., 2009; Taube, 2007; Whitlock et al., 2008).

Sensory cues underlying the head direction signal

The sensory cues underlying and forming the head direction signal have been studied extensively in rodents (more information can be found in this review: Taube, 2007).

In the most common experimental design to test head direction cells, a rat is placed in a

~1m diameter circular arena with uniform walls and floors. The arena is separated from the rest of the room with curtains and the only navigational reference point available to the animal is usually a single, prominent visual cue card. Manipulations to the cue card’s position produce changes in head direction cell preferred angles, suggesting that head direction cells establish their preferred orientations primarily based on visual information.

Shifting the cue card’s position usually leads to corresponding shifts in head direction cell preferred directions, indicating that the head direction signal may be anchored to visual landmarks (Taube, 1995; Taube et al., 1990b). However, this is only the case when the animal is removed from the arena while the cue card’s position is being changed.

Otherwise, when a well-established visual cue is moved in front of the animal, the shift in the preferred angle is not always that accurate (Taube et al., 1990b).

A prominent visual landmark can become well-established and anchor head direction cell preferred angles within minutes of exposure (Goodridge et al., 1998).

Interestingly, the removal of visual landmarks from the environment, or turning off the lights, does not abolish head direction cell firing even when no other allothetic cues

(olfactory or tactile) are present, although the preferred angles might drift (Goodridge et al., 1998; Taube et al., 1990b).

Optic flow, although visual in nature, is considered to be an important idiothetic cue during navigation, because it is usually a result of head or body movements. Since such

13 movements lead to the angular displacement of the head, which is closely monitored by the vestibular system of mammals, optic flow influences the head direction system predominantly as a vestibular cue that is processed by the vestibular nuclei and forwarded to other brain structures (Arleo et al., 2013; Waespe and Henn, 1977). To test whether optic flow alone, with no corresponding vestibular effects, can influence head direction cell firing, Arleo and colleagues projected and rotated a visual pattern on a curtain that was surrounding a circular platform where the animal was standing (Arleo et al., 2013). When they slowly rotated the pattern around, they found that the recorded head direction cells’ preferred directions shifted around in the same direction. Thus, the rotation of the pattern simulated a shift in the arena environment and resulted in a corresponding change in preferred directions. These findings suggest that the head direction signal can be maintained based on visual information alone or in combination with other sensory sources.

Several earlier findings indicated that unlike visual cues, vestibular cues alone may not be sufficient to maintain head direction cell activity and these cues may be integrated with proprioceptive and motor feedback at an early processing stage (Taube, 2007). For instance, the strength of the directional signal does not change in motionless animals compared to when they are walking, even though, in this case, there are no vestibular cues contributing to the neural representation of head direction. Also, normally when animals move from a familiar environment into a novel environment that lacks a familiar landmark, head direction cells are still able to encode the earlier established preferred direction.

However, when the animal is passively transported from the old room into the new one, the preferred directions randomly shift by ~70°, even though vestibular cues were available to the system (Stackman et al., 2003). Thus, the general understanding in the early stages

14 of head direction studies were that vestibular cues without proprioceptive feedback and motor efference copy might not be sufficient in situations where the animal is passively traveling.

The motor system’s involvement in head direction coding was further supported by experiments where the effects of passive rotation through a preferred angle were compared to active walking (Knierim et al., 1995; Taube, 1995; Taube et al., 1990b). These studies showed that the firing rate at the preferred angle is reduced in passively displaced animals.

The subjects of these experiments were usually tightly wrapped in a towel to restrain them while they were being moved. Even though extensive acclimation and pre-training reduced stress that might be caused by the tight restraint, it still resulted in decreased head direction cell activity. These results were similar to the findings of carefully executed experiments, where Zugaro and colleagues trained unrestrained rats to remain immobile in the middle of a platform that they rotated around in a lit room with stable background visual cues

(Zugaro et al., 2001). They found that when compared to the freely walking trials, peak firing rates of head direction cells were reduced by 27% on average during passive rotations in the anterodorsal thalamus. These results further confirmed that being restrained was not the source of the reduction in firing rate, but the passive movement itself caused those changes.

In contrast to the previously discussed studies, more recent behavioral, physiological and lesion studies suggest that out of all motion-derived information, vestibular cues are in fact the most critical for generating the head direction signal and they indeed have the capacity to do so without the involvement of other idiothetic cues.

Neurotoxic lesions of the vestibular labyrinth (Stackman and Taube, 1997) and inactivation

15 of the vestibular hair cells (Stackman et al., 2002) completely abolishes the head direction signal for up to several months after the lesion. Similar studies in chinchillas and mice showed that occlusion of all three semicircular canals in both ears also leads to the disruption of head direction coding (Muir et al., 2009; Valerio and Taube, 2016). The animals involved in the experiments described above displayed a variety of behavioral deficits, such as periodic ataxia, thigmotaxis (specifically walking close to the arena walls) and spinning around in circles. A newer set of passive rotation experiments from the Taube laboratory showed that passive rotations do not lead to any changes in head direction cell activity if in addition to restraining the rat’s body, the head is also fixed (Shinder and

Taube, 2011). This indicates that the earlier observed reductions in head direction cell firing rate were likely due to unrestricted head and neck movements. They also found that the directional signal can be maintained even when passive rotations take place in complete darkness, supporting not only that vestibular cues are more important than any other idiothetic input, but also that the head direction system can rely on vestibular inputs when visual landmarks are not available.

Head direction network

A major advantage of extracellular recordings is that a single electrode can record the activity of multiple cells simultaneously. This technique has been traditionally used in navigation studies and provided researchers with the opportunity to look at the relationship between several head direction cells simultaneously. Results from the Knierim lab provided evidence suggesting that head direction cells, at least within one brain region, might function together as a network (Yoganarasimha et al., 2006). They found that

16 sensory manipulations to the environment, such as landmark removal, result in approximately equal shifts in preferred directions of all recorded cells. However, the amount of the shift is unpredictable, and the neural processes leading to the shift are still not known. Nevertheless, because each cell responded similarly and with equal shifts, we can assume that the specific inputs driving this change similarly affect all head direction cells in that particular brain region. Thus, head direction cells resemble a coherent neural network where the preferred directions are always a fixed angle apart from each other and perturbations to the environment lead to changes in every individual cell’s firing patterns.

Relationship with other networks in the navigation system

The head direction network is a fundamental component in the vertebrate navigation system. Since the two critical pieces of information necessary for navigation are location and orientation, accurate navigation is not possible without head direction cells.

Positional information is encoded by the aforementioned place cells in the form of a compressed integrated signal, comprised of internal and external sensory cues, motivational and internal state information, and if available, relevant memories

(McNaughton et al., 2006; O’Keefe and Dostrovsky, 1971). Because positional information is independent of orientation, there might not be a direct link between place cells and head direction cells, however to my knowledge, this hypothesis has not been supported or rejected to date. Nevertheless, the hippocampus has direct connections with many brain areas that contain head direction cells (Figure 1.1).

Another major component of the vertebrate navigation circuits is the grid system

(Hafting et al., 2005; Rowland et al., 2016). Grid cells are principal cells in the medial entorhinal cortex that similarly to place cells, fire when the animal crosses specific locations within an environment (Figure 1.1). While place cells only have a single place field where they fire, grid cell firing fields are hexagonally arranged and repeat at regular intervals over the entire environment creating a grid-like structure of place fields. This grid- like firing pattern contains two types of spatial information, position in the environment and a regular metric of distance. Because grid cells in different layers of the medial entorhinal cortex span multiple scales and orientations (larger/smaller distances in the grid pattern and different orientations based on external cues), combinations of grid cells can provide information about distance and position in any environment (McNaughton et al.,

2006; Rowland et al., 2016). The exact source of positional information and thus the relationship between place cells and grid cells is still unknown, however there is physiological evidence supporting interactions between the two populations of spatial cells

(Figure 1.1; Rowland et al., 2016).

A direct connection between grid cells and head direction cells has been recently established. An elegant study from Winter and colleagues showed that grid cells rely upon head direction cells to encode orientation (Figure 1.1; Winter et al., 2015a, Winter et al.,

2015b). They lesioned the head direction system located in the anterior thalamic nuclei with a reversible lidocaine injection and found that the inactivation of this orientation signal source disrupts grid cell firing. The animals recovered from the lidocaine injections within

~1.5 hours and so did the recorded grid patterns. Saline injections did not have any effect on grid cell firing. To test whether any long-term compensatory mechanisms exist that

18 could provide a backup system for grid pattern generation, they permanently lesioned the source of the head direction signal. They found that recovery after the permanent damage activated the hypothesized secondary mechanism and spared the grid pattern. The neural basis of the compensatory mechanism, and what other regions the orientation information may come from, remains to be found. Nevertheless, these data provided the first piece of evidence showing that grid cells receive orientation cues directly from the anterior thalamic head direction network and that the representation of distance, and to some degree position, is highly dependent on the orientation input from the head direction system.

Neural control of insect navigation in the central complex

The central complex (CX) is located in the center of the insect brain. This central location provides an ideal opportunity to receive and integrate various types of sensory inputs from any brain region. Indeed, a wide range of anatomical, behavioral, physiological and genetic studies support its role in sensory information processing, motor control, as well as navigation and spatial memory processes (Pfeiffer and Homberg, 2014).

Central complex anatomy

The CX consists of the protocerebral bridge (PB), the fan-shaped body (FB) and the ellipsoid body (EB), as illustrated in Figure 1.2. In pterygote (winged) insects, such as the cockroach and locust, the most ventral components of the CX are the paired noduli

(Pfeiffer and Homberg, 2014). In the cockroach, the PB is shaped like an elongated handlebar and contains 16 columns. Located between the mushroom body calyces can be

19 found the kidney- or fan-shaped body FB with 16 columns and the EB with 8 columns

(Pfeiffer and Homberg, 2014; Strausfeld, 1999). Both the FB and EB are also organized into horizontal layers.

Due to its central location, the CX has direct and indirect connections with a wide range of brain areas. In locusts and, presumably cockroaches, the CX is only indirectly linked to primary sensory areas, nevertheless, it receives a great wealth of preprocessed sensory information through connections with other associative structures (Homberg et al.,

1991; Pfeiffer and Homberg, 2014).

One of the most important direct projections goes to the lateral accessory lobes

(LAL), and transfers ascending and descending information between the thoracic ganglia and the CX (Homberg, 1994; Okada et al., 2003). Although they have not been experimentally demonstrated yet, ascending projections likely carry sensory feedback associated with motor behaviors to the CX. Meanwhile, descending projections relay higher order motor commands from the CX towards the nerve cord.

There are also three main indirect visual pathways connecting the eyes and the CX

(Pfeiffer and Homberg, 2014). Two of these pathways relay visual information through the

LAL to the EB and FB respectively. The third, indirect visual pathway enters the PB through the posterior optic tubercle.

Even though physiological studies indicate that the CX receives and processes mechanosensory inputs coming from the antennae and other body parts, the underlying circuitry has not been resolved yet (Homberg, 1994; Homberg, 1985; Ritzmann et al., 2012,

2008). Olfactory information is transferred from the antennal lobes to the mushroom bodies

(Strausfeld et al., 2009). Considering that the CX is located in between the mushroom bodies, one might expect that antennal inputs could be further transferred to the CX from there. In spite of this ideal position, only a small number of direct connections between the

CX and the mushroom bodies have been traced to date (Heinze et al., 2013; Phillips-

Portillo, 2012; Strausfeld, 1976).

Figure 1.2: Schematic illustration of the cockroach brain and the central complex (CX). Highlighted by the rectangle are three CX subunits, the protocerebral bridge (PB; orange), fan-shaped body (FB; yellow) and the ellipsoid body (EB; red). The structures highlighted in green are the mushroom bodies, and the structures highlighted in blue are the lateral accessory lobes (LAL).

Cellular composition of the central complex

The majority of our knowledge regarding the CX’s neural composition comes from locusts (Heinze and Homberg, 2008; Strausfeld, 1999). There are four major groups of neuronal cell types in the CX. The first group contains tangential neurons, which were named after tangentially innervating single layers of the PB (TB neurons), the FB (TU neurons), the EB (TL neurons) and the noduli (TN neurons). These neurons represent the

21 most diverse input source to the CX, and thus have extensive ramifications outside of the

CX. The different types of tangential neurons connect specific brain areas with specific subunits of the CX. Most of these connections carry sensory inputs to the CX from different parts of the protocerebrum, and a small subset of them carry ascending feedback from the thoracic ganglia (Strausfeld, 1999).

The second most prominent group of CX cells are the columnar neurons. Columnar neurons are the main output elements of the CX. Columnar neurons’ cell bodies are located outside of the CX, in the pars intercerebralis, and have extensive spiny ramifications mostly in the PB. These neurons connect single columns of the PB and or the FB/EB with the LAL in a very precise, regulated manner. For instance, most (but not all) columnar neurons’ axons cross the midline before entering the contralateral hemisphere of the FB. Together these projections form a symmetrical interhemispheric network of columnar connections that have the capacity to converge sensory information perceived on the left side with information coming from the right side of the animal. Both tangential and columnar cells play a role in navigation by processing polarized light visual cues in a topographic manner

(Heinze and Homberg, 2007).

The other two cell types found in the CX are only located in the FB. Amacrine neurons are anaxonal, and their somata are located in the pars intercerebralis. Pontine neurons are intrinsic to the CX, and their somata are also located in the pars intercerebralis.

Pontine neurons connect pairs of columns in the FB in a precise manner, sending ramifications from one or two slices of one hemisphere to one or two slices in the contralateral hemisphere (Heinze and Homberg, 2008).

Directional sensory signal processing in the central complex

As mentioned before, the CX’s ideal location and diverse connections with other brain regions make it a multisensory hub, where preprocessed sensory information of different modalities (visual, mechanosensory, olfactory) is integrated into a more compressed multimodal neural code. This abstract representation of the sensory environment then presumably gets integrated with information regarding the internal state of the animal, to provide the CX’s downstream targets with behaviorally relevant and spatially organized descending commands.

Although, the majority of CX studies focused on visual processes, there is evidence supporting the CX’s role in mechanosensory information processing. Ritzmann and colleagues performed extracellular multichannel recordings from the CX of restrained cockroaches while mechanically stimulating the animal’s antennae (Ritzmann et al., 2008).

During the experiment, one antenna or both antennae were deflected in a range of directions at varying velocities. The results indicate that most CX units are sensitive to mechanical stimulation of either antenna with no directional preference. Meanwhile, other units responded to deflections in a specific direction or preferred deflection of one of the antennae. They also found that many of these antenna-movement sensitive neurons also respond to changes in ambient light intensity, suggesting that the CX is involved in multisensory convergence. This same group later reported that CX neurons respond to both external stimulations of the antennae and also self-generated contact with an object located by the antennae(Guo and Ritzmann, 2013). Further evidence for both mechanosensory processing and multisensory integration comes from intracellular studies exploring the responses of CX neurons to combinations of flashing light, air puffs and olfactory stimuli

(Homberg, 1985; Milde, 1988; Phillips-Portillo, 2012). For instance, in 1985, Homberg reported that single FB cells might increase their firing rates in response to a variety of sensory stimuli, including a light stimulus, an air stream over the animal’s head and an odor puff. Taken together, these data indicate that the CX is a multisensory structure

(Homberg, 1985).

The CX’s role in directional sensory signal processing is supported by a wide range of studies. Since the majority of these studies focused on polarized light vision, the neural computations and circuitry underlying visual processes have been described in great detail, especially in the context of polarized light guided long-distance migration and spatial orientation. Nevertheless, the circuitry described below is likely similar to the circuitry governing general visually-guided navigation in most insects.

The sky is a rich source of visual information that can aid animals in spatial navigation. Many insects exploit the polarization pattern of the sky to successfully orient themselves. Polarized light is detected by photoreceptors in the dorsal rim area of the compound eyes and then travels through the anterior polarization pathway. This pathway carries visual information through the anterior lobe of the lobula, through the medial and lateral bulbs located in the LALs, to TL2 and TL3 tangential neurons in the EB (Homberg et al. 2003; Labhart and Meyer, 1999; Pfeiffer et al., 2005; Träger et al., 2008).

Polarized light-evoked information passes through three stages of processing within the CX, the above described anterior polarization pathway is the initial, input stage of the network. Following this step, at the intermediate stage, polarized light information is transferred from the EB to the contralateral hemisphere of the PB through CL1 columnar neurons (Heinze and Homberg, 2009). The output stage involves two pathways. In the first

24 pathway of the output stage, columnar neurons forward the polarized light signal directly back to the LAL. In the second pathway TB1 and CL1 neurons located in the PB integrate the signal from several CL1 cells across both hemispheres of the PB. As a result of this process, a topographically arranged compass-like representation of preferred E-vector orientations (the polarization angles of the sunlight) is generated in the output areas of TB1 tangential neurons. This internal visual compass covers approximately 2 X 180° along the

16 PB columns (Heinze and Homberg, 2007). The topographical arrangement of the visual information is preserved through the next step, where it is forwarded onto different types of columnar cells in the PB (CP1, CP2, and CPU1). Finally, as the last step of the output stage, these neurons project to several regions of the LAL, where they may influence descending commands (Heinze and Homberg, 2007; Homberg et al., 2011).

In addition to the sky’s polarization patterns, other landmarks, such as the position of the sun, or for nocturnal insects the moon and the Milky Way, may also visually guide navigation (el Jundi et al., 2015). The sun being the brightest spot in the sky, on its own provides a good reference point for orientation. Additionally, when sunlight is scattered in the sky, it not only produces a polarization pattern, but a chromatic and intensity gradient of long wavelength radiation as well. Results of intracellular recordings indicate that neurons in the primary visual areas, as well as the CX, are sensitive to such unpolarized skylight signals. El Jundi and colleagues recorded from CL1, TB1 and CPU1 neurons in the locust brain and reported that all of these neurons were sensitive to simulated sun chromatic cues (el Jundi et al., 2014a). When combined with directional cues derived from the sky’s polarization patterns, chromatic cues can make navigation led by a sky compass more reliable in ambiguous or conflicting situations. It is likely that the sun’s position and

25 the resulting changes in chromatic gradient may also signal the time of day to the animal, thus this mechanism could also provide the animal with an estimate of the time spent with traveling.

Information about time is especially important in long-distance migratory animals, such as monarch butterflies. The sun’s position constantly shifts throughout the day, and so does the polarization pattern in the sky, thus to stay on track, migratory animals need to compensate for this shift in their heading. It has been proposed that, at least in monarchs, directional visual information gets time-compensated in the CX to account for the celestial shifts (Reppert et al., 2016). The proposed pathway involves timing-related information communicated to the CX from the peripheral antennal clocks and the brain circadian clocks which are located in the pars lateralis. Although, the exact mechanisms supporting these hypotheses remain to be determined, time-processing pathways will be crucial components of the navigation circuits in every insect’s brain.

Selection and maintenance of behavior

The CX’s connections with upstream sensory areas and downstream motor centers set this region up for higher level, associative processes, such as action selection and maintenance of behaviors. The arriving preprocessed sensory information is further processed and integrated with information about the animal’s internal state. Such multisensory information can then be used by the CX, or its downstream targets, to guide behavior and fine-tune motor patterns.

The first piece of evidence for the CX’s role as a premotor area came from Huber in 1960. Huber showed that electrical stimulation of the CX promotes locomotor activity, while CX ablation leads to a decreased walking duration in crickets (Huber, 1960). Later, extracellular recordings in cockroaches revealed that the activity of some neurons in the

CX correlates with, and sometimes predicts, stepping frequency (Bender et al., 2010).

These results are highly similar to speed cells found in the rat entorhinal cortex, which is also an associative brain area (Figure 1.1; Kropff et al., 2015). Bender and his colleagues were also able to manipulate the animal’s walking speed by stimulating presumably the same cells through the recording electrodes. These results were complemented by genetic studies on fruit flies, where structural mutations and CX neuronal knockouts showed changes in walking speed and behavioral activity levels (Martin et al., 1999; Poeck et al.,

2008; Strauss et al., 1992; Strauss and Heisenberg, 1993).

In addition to controlling the duration and speed of walking, the CX also influences the directional components of locomotion. Both mechanical and electrical lesions in the

CX lead to inappropriate turning behaviors. Mechanical lesions resulted in cockroaches turning or circling around preferentially in one direction and causing them to run into obstacles (Ridgel et al., 2007). More specific electrolytic lesions that affected smaller areas of the CX (approximately one to two columns) had similar effects on turning behaviors

(Harley and Ritzmann, 2010). Lateral lesions in the FB produced wrong turns in a U- shaped track. Lesions in the midline failed to affect turning behaviors, but impaired the animal’s climbing abilities. Tetrode recordings in tethered cockroaches walking on an air- suspended ball provided physiological evidence for the CX’s role in the directional control of locomotion (Guo and Ritzmann, 2013). Changes in the activity of some lateral CX

27 neurons preceded changes in angular velocity in the ipsilateral direction. Additionally, when the same regions of the CX were electrically stimulated, the elicited locomotion’s direction was also biased.

These findings were extended by Martin and colleagues, who recorded the extracellular activity of CX neurons while cockroaches were freely exploring an arena or climbing over a barrier (Martin et al., 2015). They found that even in freely behaving animals, some neurons selectively predicted movement in the left, right or forward directions, others predicted the speed of walking, while a group of neurons predicted both the speed and direction of movements. Stimulation at these regions of the CX through one of the recording electrodes induced consistent trajectories of forward walking or turning.

Further tests on the animals where stimulation induced turning revealed a direct link between CX activity and changes in motor neuron activity and an inter-joint reflex in the legs that resembled alterations that occur during turning. On the other hand, climbing over a barrier resulted in altered firing rates that corresponded to the changes in navigational context.

Genetic studies on two Drosophila mutants with perturbed PB structure, ocelliless1 and tay bridge1 were tested in a gap-climbing paradigm where the animal has to cross a gap raised between two platforms (Triphan et al., 2010). Both mutants exhibited normal motor function, but failed to align their bodies in the correct direction, and thus fell into the gap.

These results suggest that impairments in the PB cause a deficit in directional visual targeting, rather than inappropriate motor control. Other genetic tests revealed that all three of the CX subunits (PB, EB and FB) are important for controlling walking direction and speed. For instance, the no-bridge mutation in fruit flies causes structural defects in the PB

28 of the CX, and these flies display reduced walking speeds and disturbed leg-coordination during turning, supporting the PB’s role in motor control (Strauss et al., 1992).

As a premotor center responsible for the selection and maintenance of behavioral activity, the CX also controls motor functions that are not related to navigation. For instance, in crickets, grasshoppers, and fruit flies the CX is involved in the motor control of courtship behaviors and acoustic communication. Electrical and chemical stimulation of the CX can lead to consistent sound production, however the songs are atypical with unnatural temporal dynamics (for review see: Pfeiffer and Homberg, 2014).

Visual and spatial memory

The CX’s role in learning and memory has been predominantly supported by genetic studies in the fruit fly, Drosophila melanogaster (e.g. Zars et al., 2000; Liu et al.,

2006; Neuser et al., 2008; Li et al., 2009; Pan et al., 2009; Ofstad et al., 2011). The results of these studies indicate that most CX structural mutants perform poorly in sensory learning paradigms with spatial components. Similarly to studies investigating basic sensory processes, most of the evidence for the CX’s role in learning and memory comes from studies employing visual paradigms.

Visual pattern recognition

Visual pattern recognition has been heavily studied in fruit flies (Dill and

Heisenberg, 1995; Dill et al., 1993; Dill et al., 1995; Ernst and Heisenberg, 1999; Tang et al., 2004). When fixed in a flight simulator, facing an artificial panorama display with visual patterns, flies can remember the color, position/elevation, vertical compactness and contour orientation of the pattern with the help of short-term memory pools. Liu et al.

(2006) tested ten mutants with structural deficits in the CX and found that all ten failed the pattern recognition test. The flies did not show any motor impairments, and three out of the ten mutants were able to discriminate between patterns, suggesting that the mutation affected visual learning performance rather than pattern recognition itself. They found that visual memory traces were stored within distinct CX structures depending on the visual components of the patterns. All visual pattern memory processes were linked to the expression of the rutabaga gene (rut), a gene known to be involved in many forms of learning and memory in fruit flies (Zars et al., 2000). Short-term memory traces for pattern elevation and contour orientation were linked to F5 neurons (dorsal FB neurons) and F1 neurons (ventral FB neurons) respectively (Liu et al., 2006; Pan et al., 2009). Similarly, expression of rut in R2/R4m neurons (ring neurons) in the EB of flies was also required for storing most features of a visual pattern (Pan et al., 2009).

Another important gene, the foraging gene (for) also plays a role in visual learning.

The foraging gene in fruit flies encodes a cGMP-mediated protein kinase (PKG) (Kaun et al., 2007). Under natural circumstances, individual differences in for gene produces two behavioral types in the larval stage called “sitters” and “rovers”. Elevated for expression, among other things, increases PKG production that results in increased foraging activities, hence the name “rovers”(Kent et al., 2009). forS mutant flies that have reduced PKG

30 activity cannot store memory traces of visual patterns efficiently. However, their learning ability is at least partially rescued after ectopic expression of for+ in R2/R4m neurons in the EB or in the F1 or F5 neurons of the FB of these mutants (Li et al., 2009).

Detour paradigm and spatial working memory

The fruit fly’s ability to perform well in visual learning tasks comes from its natural attraction to visual features. When presented with two identical vertical lines on the opposite sides of a circular arena, flies tend to walk back and forth between these two lines

(Bülthoff et al., 1982). In a detour paradigm, individual flies are placed in the middle of a similar arena, with similar visual pattern displays on the two opposite sides of the arena, which are removed after the fly crosses the midline (Neuser et al., 2008). Then a distractor target is displayed at a 90° angle compared to the flies heading. Wild type flies tend to turn towards this new visual target if it was present for at least 500ms. When the fly is facing the distractor target, it disappears within 1s. When wild type flies were left in the arena with no visual targets, they remembered their original, pre-distractor heading and started walking in that direction again. Thus, these flies were able to store and recall the position of a former target even though it was not present in the environment anymore. Flies with silenced EB ring neurons performed poorly in the detour paradigm, for instance, they did not remember their pre-distractor heading, suggesting that these neurons are important components of a spatial working memory circuit. To elucidate which molecular pathways may play a role in this process, Neuser and colleagues first tested mutants for the rutabaga gene (Neuser et al., 2008). As mentioned previously, rut mutants are unable to learn visual patterns, thus as expected, they also failed the detour paradigm. Next, they investigated the 31 role of a gene called ignorant (ign). ign58/1 flies patrolled the two target stripes like wild types, however, they were unable to recall the positions of the original targets after the distractor disappeared. Rescue experiments showed that expression of ign is especially important in R3 and/or R4d neurons for working orientation memory.

Visual place learning

The Morris water maze is a behavioral test commonly used with rodents to study place learning (Morris, 1981; Morris et al., 1982). In this task, the subject is placed in a circular pool of water and it has to find a visible or hidden platform that allows it to escape from the water. Under natural circumstances, the amount of time spent with searching for the platform decreases gradually as the animal learns its location. Since most insects are not great swimmers, the insect version of the maze is a circular arena with heated floor tiles and a single cold tile which serves as the rescue platform (Mizunami et al., 1998; Ofstad et al., 2011). When tested in this paradigm, wild type fruit flies quickly learn (one trial, 5 min) to locate the cold tile, by relying on the visual patterns displayed on the arena walls. When the pattern is rotated, it takes the flies approximately 10 trials to improve their time required to locate the cold tile. Thus, they need ~10 trials to learn to associate the new position of the visual features with the cold tile’s location. In a series of behavioral tests, individuals with silenced R1 neurons in the EB displayed normal locomotor, optomotor, visual pattern discrimination and olfactory learning performances, yet they failed the spatial learning paradigm. These results indicate that R1 neurons are specifically responsible for some aspect of visually-guided place learning that is independent of basic sensory and locomotor functions or olfactory learning function (Ofstad et al., 2011). 32

Together these studies indicate that the CX is responsible for various important functions which rely upon the spatial context by integrating specific components of the sensory environment. Such processed, spatial information can be encoded, retained and recalled within the CX and from there it can be forwarded to motor areas to shape behaviors.

The physiological correlates of orientation in the central complex

In the previous sections I discussed how the CX processes directional sensory information and how it controls directional behaviors. Until recently, we had no information on what happens between the sensory and motor processing stages, and how

CX neurons turn directional sensory cues into appropriately directed movements. For the first time, Seelig and Jayaraman provided physiological evidence for the CX’s direct involvement in encoding an animal’s orientation in a navigational context (Seelig and

Jayaraman, 2015). The authors used two-photon Ca2+ imaging to monitor the dendritic responses of a set of 16 columnar neurons (called ring neurons or wedge neurons) that send projections to 16 columns of the EB in the Drosophila CX. Unlike some other insects, the fruit fly’s EB is ring shaped, or elliptical, so the columns divide it into 16 radial wedges.

They picked this particular type of neuron, because it has previously been shown to process visual information and play a role in feature detection (Seelig and Jayaraman, 2013).

During the experiments fruit flies were head-fixed, but they were free to walk on an air- suspended polyurethane foam ball in an LED arena. The arena was part of a closed-loop system, where the fly’s movements controlled the position of the projected image on the

LED panels. They found that at certain headings relative to the displayed pattern’s position,

33 active cells formed a so-called ‘activity bump’, wherein projections going to approximately

5-6 wedges would show increased activity. Whenever the fly changed its heading, the activity bump rotated as well. Importantly, any kind of visual scenery evoked this specific response, ranging from a single vertical stripe to complex visual features. This indicates that the neurons were not encoding the visual information itself, rather the animal’s orientation relative to the visual landmark(s).

By varying the closed-loop gain that matched the ball’s rotational movements to the visual landmark’s movements, Seelig and Jayaraman (2015) observed that CX activity integrated visual cues more heavily, than self-motion cues. Experiments conducted in darkness revealed similar results. The flies were able to maintain the EB wedge neuron activity bumps with no visual cues, but only for a limited period of time. This indicates that the fruit fly’s navigation system accumulates error over time when the only updates on the fly’s relative orientation come from walking. This was the first study to provide evidence for the CX’s role as a navigation center with a compass-like function that integrates sensory information about the animal’s orientation and through unknown downstream targets, guides movements accordingly.

The results from Seelig and Jayaraman (2015) combined with the results of the study conducted by Martin and colleagues (2015) on directional motor control provide the first piece of evidence for the CX’s direct involvement in orientation coding and the utilization of this orientation signal during behavior. These two studies were also the first to bridge the gap between sensory information processing and motor control in the CX and opened up new questions about the CX’s role as a higher order, associative brain center.

The second chapter of this thesis describes experiments that use methods developed in rat

34 head direction studies to test for similar orientation-sensitive cell responses in cockroach

CX.

Local Field Potentials

The third chapter of this thesis describes an additional analysis of the experiments discussed in Chapter 2, but this time examining local field potentials (LFPs). In this section,

I will describe what LFPs are and how they can extend our understanding of complex neural systems.

The origin and function of local field potentials

The functional organization of neural networks that underlie complex behaviors is multifold. There is a high level of integration across and within the different levels of processing, from the cellular level to local networks and global neural systems. Single neurons are connected to each other with synapses in a highly regulated manner and populations of coordinated cells form functional units, that are often referred to as circuits

(Watson and Buzsáki, 2015).

Functional local circuits in the brain are spatially segregated from each other into brain regions, such as the central complex neuropils in insects, or the hippocampus in mammals. The temporal organization of neural activity, also referred to as a neuron’s temporal dynamics, is important for the proper function of a brain region. Temporally organized activity of neurons which share similar intrinsic dynamics (ionic content, types

35 and distribution of ion channels, calcium flux dynamics, etc.) leads to synchronized fluctuations in electrical potentials, also called oscillations or LFPs.

The fluctuation in electric potentials are not purely the result of intrinsic cellular properties (Buzsáki et al., 2012). Among other events, synaptic currents, subthreshold intrinsic currents, gap junctions and neuron-glia interactions all contribute to extracellular fields. Thus, the LFP is the superposition of all ionic processes in the brain rather than a cellular population code or the simple summation of multiunit activity. Local oscillations can spread to multiple distant brain areas and carry information about regional effects

(Trongnetrpunya et al., 2015). On the other hand, global changes that affect multiple brain areas can also result in oscillations and these global oscillations can interact with and modify local oscillations (Buzsáki and Draguhn, 2004). Oscillations can synchronize neural activity to multiple oscillatory rhythms. By synchronization via oscillations the brain is able to divide complex information into transmittable segments, such as synchronized spike trains in a population of neurons, which are considered less taxing to decode than a random signal (Buzsáki and Freeman, 2015). As framed by Buzsáki and

Freeman, just like words in a sentence, if the segments are appropriately organized in space and time, together they can convey meaningful messages from one brain region to another.

Due to this rhythmic arrangement of information flow, oscillations are alternating combinations of synchronous and asynchronous activity (Hasselmo, 2005). Synchronous network activity is associated with increased synaptic plasticity and strongly coupled neuronal ensembles. Contrary to this, if the segments are not properly coordinated, it can lead to asynchronous activity and the weakening of neural signals. Indeed, asynchronous field oscillations are commonly observed in neurological disorders associated with

36 impaired cognitive, sensory, and motor function (Buzsáki and Watson, 2012; Palop and

Mucke, 2010). Therefore, the organized segments of information need to be transmitted within the synchronizing periods of an oscillation for proper functioning (Cardin et al.,

2009).

Due to the multiple levels of information processing, neuronal networks operate at several oscillatory frequencies in the range of ~0.05-500 Hz (Buzsáki and Draguhn, 2004).

Oscillation frequency is mainly determined by the limited speed of neuronal conduction and the size of the network. Thus, higher frequency oscillations reflect activity at a small, local neuronal space, while slow oscillations reflect global effects and the activity of large networks (Steriade, 2001). Oscillatory frequencies can be grouped into bands based on experimental observations. Several frequency bands can co-occur within a neural network and they can interact with each other. Also, the same frequency bands can temporally coexist in distant networks which contributes to global coherence even between otherwise unrelated brain regions (Buzsáki and Draguhn, 2004). The most important frequency bands that contribute to adaptive navigation and spatial memory are theta-band activity

(depending on the brain region and species, ~ 4 - 10 Hz) and delta-band activity (also varies with brain region and species, ~0.1 – 4 Hz ) (Buzsáki and Draguhn, 2004; Buzsáki et al.,

2013; Ulanovsky and Moss, 2007).

Local field potentials in mammalian navigation circuits

Rats rely on theta rhythm to synchronize navigational circuits

Local oscillations coordinate neural ensembles in the navigation circuit to reliably integrate spatial information into a neural code that downstream target areas can further process and utilize to produce behavior. In rats, synchronized theta rhythm has been observed within and between brain regions participating in navigation, such as the hippocampus, entorhinal cortex, and striatum (Buzsáki, 2002; Igarashi et al., 2014;

Malhotra et al., 2012; Mizuseki et al., 2009; Patel et al., 2012; Penner and Mizumori, 2012;

Tort et al., 2008). Theta-band activity in the rat hippocampus can modulate neuronal firing patterns and synchronize place cell firing (Buzsáki, 2002; Buzsáki and Draguhn, 2004;

O’Keefe and Recce, 1993). Theta rhythm may coordinate striatal and entorhinal networks with the hippocampus (Igarashi et al., 2014; Malhotra et al., 2012; Mizuseki et al., 2009;

Penner and Mizumori, 2012; Tort et al., 2008). Striatal oscillations can become entrained to hippocampal theta during specific spatial tasks (DeCoteau et al., 2007; Tort et al., 2008).

For instance, in goal-directed navigational tasks when the animal is engaged in decision making, hippocampal and striatal theta-band network activity becomes increasingly coherent. Grid cells in the entorhinal cortex interact with place cells in the hippocampus and both networks are synchronized by theta oscillations that originate from the hippocampus during navigation and (spatial) memory processes (Buzsáki and Moser,

2013). As a result of network coherence, changes in the environment, such as repositioning a landmark, lead to equal shifts in firing fields and remapping in both place cells and grid cells. On the other hand, interference with or disruption of theta oscillations leads to impaired task performance and failure to recall learned spatial behaviors (Buzsáki and

Draguhn, 2004; Penner and Mizumori, 2012).

In addition to the grid system’s synchronization, limited amount of evidence showed that the activity of the entorhinal head direction network is also coordinated by theta rhythm (Brandon et al., 2013; Tsanov et al., 2011). Brandon and colleagues examined the relationship between head direction cell spiking patterns and entorhinal theta cycles.

They found that the activity of two subpopulations of head direction cells was entrained to alternate theta cycles, wherein one population skipped a cycle compared to the other, and both populations showed a reliably fixed phase or anti-phase relationship with theta

(Brandon et al., 2013). The observed firing pattern segregation correlated with tighter tuning curves and the population of neurons entrained to theta were highly synchronized.

In a separate study, pharmacological lesioning of an upstream brain region reduced theta- band activity and abolished grid cell firing in the entorhinal cortex, which in turn disrupted theta rhythm synchrony in all head direction cells, while maintaining head direction specificity and tuning (Brandon et al., 2011). This experiment was performed on freely behaving rats that explored a large arena with a single cue card during the recording sessions. Despite the changes in network synchrony, the rats spent similar amounts of time in the arena and their walking speeds did not differ during the control and lesion trials.

Considering that the rats were not expected to perform a navigational task that required spatial memory, it is possible that the behavioral effects of the lesion were masked by the random exploratory behavior of the animals.

Bat navigational network activity

The first experiments concerned with bat navigation brought exciting results to the vertebrate navigation literature. Bats have been shown to have place cells, grid cells, head

39 direction cells and even border cells in the same brain structures as rats (Ulanovsky and

Moss, 2007). The place fields of these cells are 3 dimensional, which is not surprising considering that bats navigate by flying, thus performing volumetric navigation. On the other hand, some significant differences occurred in the network dynamics that accompanied these cells (Ulanovsky and Moss, 2007; Yartsev et al., 2011). For instance, even though LFPs in the theta-band were present, these oscillations had a smaller spectral peak and co-occurred with a large delta-band component, unlike in rats. Also, in rats theta is continuous during locomotion, but in bats it only appeared intermittently, in short 1-2 second bouts and exclusively during periods of environmental exploration through echolocation. Moreover, unlike in rats, place cell and grid cell firing patterns did not show any theta rhythmicity, suggesting that theta oscillations may not play an important role in bat navigation. These findings in bats corroborate the results of primate and human studies where theta rhythmicity also only appears in short bouts (Ekstrom et al., 2005; Stewart and

Fox, 1991).

These interspecies similarities on the single neuron level, and the major discrepancies on the network level highlight the importance of employing a wider range of scientific approaches in order to fully understand brain function. Studying single neuron activity, as well as oscillation patterns, can lead to a more complete understanding of the functional organization of a circuit and how it might interact with other circuits. A population code retrieved from single neuron data alone likely would not provide us with the same results.

Local field potentials in the arthropod brain

Oscillations in the insect brain

Although, the majority of electrophysiological studies in arthropods, specifically in insects, have been focusing on single cell analysis, measurements of LFPs are becoming widely used. As a result of this, correlations between specific behaviors and oscillatory patterns– similar and sometimes analogous to those apparent in mammals - have been observed in several insect species (Kay, 2015; Kirschfeld, 1992; Kirszenblat and van

Swinderen, 2015; Laurent and Davidowitz, 1994; Nitz et al., 2002; Paulk et al., 2013; Paulk et al., 2014; Paulk et al., 2015; Stopfer et al., 1997; van Swinderen and Brembs, 2010; van

Swinderen and Greenspan, 2003; van Swinderen et al., 2009). For instance, synchronization by oscillations has a similar role in the locust olfactory system as in mammalian brains. In the presence of an olfactory stimulus, LFPs in the 20-30 Hz frequency band likely originate in the antennal lobes and spread to the mushroom bodies.

Although, these evoked oscillations indicate the presence of an odor, they do not contain any information about the chemical identity of the stimulus, rather they reflect network level synchronized activity that leads to odor perception (Laurent, 1996; Laurent, 1999;

Laurent and Davidowitz, 1994; Laurent and Naraghi, 1994; Wehr and Laurent, 1996).

Similarly, beta oscillations (15-35 Hz) are prevalent in rat olfactory circuits during odor sampling and show high coherence with activity in interconnected olfactory structures

(Kay, 2014; Kay, 2015; Kay and Beshel, 2010).

Oscillatory activity in the high frequency beta-band (~20-30 Hz) in the central protocerebrum of fruit flies has been suggested to play a role in the selection and

41 suppression of novel and salient stimuli and to participate in attention-like processes (de

Bivort and van Swinderen, 2016; Frye and Dickinson, 2003; Tang and Juusola, 2010; van

Swinderen, 2007; van Swinderen and Greenspan, 2003; van Swinderen et al., 2009). It has also been indicated that specific changes in LFPs recorded from the central brain of

Drosophila can be linked to different behavioral states, such as different stages of arousal and sleep (Kirszenblat and van Swinderen, 2015; Nitz et al., 2002; van Swinderen et al.,

2004). The function of these specific changes in oscillatory rhythms remains unclear, but based on the extensive literature on sleep in mammals, it is likely that they support the neural mechanisms underlying learning and memory (Alphen et al., 2013; Kirszenblat and van Swinderen, 2015; Nitz et al., 2002).

Crayfish central complex network activity

The majority of our knowledge specifically on CX oscillations comes from crayfish sleep studies. Ramon and colleagues recorded LFPs simultaneously from several major brain areas, including the CX, in both freely moving and tethered crayfish for extended periods of time (Mendoza-Angeles et al., 2007; Mendoza-Angeles et al., 2010; Ramón et al., 2004). They found that in active behavioral states oscillations in the ~15-20 Hz frequency band are only present in the CX. However, as the crayfish becomes relaxed and shifts to a sleeping state, this activity spreads to the rest of the brain. This initial observation suggested that the CX might serve the role of a main oscillator in the brain. In an elegant control experiment they showed that cooling the CX eliminates the 15-20 Hz activity, which confirmed the CX’s role in generating and spreading rhythmic activity during sleep

(Mendoza-Angeles et al., 2010). Whether and why the CX serves as a central oscillator in insect brains will need to be addressed by future studies.

Based on these examples it becomes clear that information gained from comparative studies can lead to a better understanding on how different circuits in different nervous systems might serve analogous roles in similar ways or might have become specialized during evolution. Hence, it is very important to extend the analysis of insect

CX navigation studies to the realm of LFPs as is provided in Chapter 3.

Summary

As we examine how the brain controls behavior we must identify the types of information that are necessary for each behavior. What are the computations and processes that transform sensory information about our current state in the world into appropriate motor commands to change it to our desired state in the world? When navigating in a complex environment, all animals must be aware of their position and orientation in a rich sensory scenery. Knowledge of position and head direction are the necessary starting points for any navigational task. During goal-directed navigation an animal’s current spatial context is compared to an expected spatial context. If the current state and the expected goal do not match, short-term or long-term memory traces of previous encounters with similar situations need to be retrieved. Such learned information can then be combined with the internal representation of the sensory environment to select and initiate motor actions that will lead to the navigational goal.

The available literature on the CX, including but not limited to the studies discussed in the Introduction, provided information about some of the neural components and circuits underlying the above described navigational process. All of these neural mechanisms and functions pointed in the direction that the CX might be supervising adaptive navigation, by participating in a range of processes from sensory information acquisition to directed motor control, but until recently there was no direct evidence supporting this hypothesis.

In this dissertation, I further investigated the CX’s role in adaptive navigation. I used multi-channel extracellular recording techniques to uncover the neural correlates of head direction coding and spatial context cues in the cockroach CX. Specifically, in the experiments described in Chapter 2, I used tetrodes to record the activity of single neurons

44 in the CX while the animal was passively rotated around on a platform surrounded by a circular arena. I found that single units in the cockroach CX encode the animal’s head direction relative to a salient visual cue. In addition to these results, I found single neurons that encoded the rotation direction history of the animal, which is a common spatial context cue. These results suggest that the CX navigation circuit is involved in environmental context discrimination processes that might be utilized by spatial memory circuits in the insect brain.

Tetrode recordings not only provide us with single unit data, but can also be utilized to simultaneously examine network-level LFP activity in the brain. Thus, when conducting the experiments described in Chapter 2, I also recorded the changes in CX network activity represented by LFPs. As described in the Introduction, LFPs are considered to be important mediators of nervous system function, especially in the context of mammalian navigation.

I provide the first description of spontaneous LFPs in the insect CX. The LFP recordings described in Chapter 3 also provide us with the first evidence that oscillations in the insect brain also have the capacity to encode spatially-relevant information and thus contribute to adaptive navigation. I found that LFP activity in all frequency-bands encoded the animal’s head direction during the stationary periods of the passive rotation experiments. The recorded LFPs also encoded spatial context cues, such as the presence or absence of a landmark, or the past rotation direction of the animal through changes in relative response magnitude and average LFP power. I predict that the sensory processes supported by LFP oscillations might have an important role in adaptive navigation, by serving a network- level foundation for sensory context discrimination. The results of Chapters 2 and 3 together provide a solid foundation for future studies on the neural basis of adaptive

45 navigation in insects and will contribute to a broad comparative approach to understand the general principles as well as the diversity of all navigation circuits.

Chapter 2

Cellular Basis of Head Direction and Contextual Cues in the Insect Brain

This material was previously published in the journal Current Biology: Varga, A. G. and Ritzmann, R. E. (2016)

Summary

Animals rely upon integrated sensory information for spatial navigation. A question of wide importance in navigation is how sensory cues get transformed into neural codes which represent the animal’s orientation within its proximal environment. Here, we investigated the possibility of head-direction coding in the central complex of the cockroach, Blaberus discoidalis. We used extracellular recordings in restrained animals that were rotated on a platform relative to a fixed landmark. The passive rotations allowed us to test for head direction coding in the absence of self-generated motion cues. Our results indicate that individual cells in the central complex encode the animal’s heading relative to a landmark’s position in several ways. In some cells, directional tuning was established even in the absence of visual cues, suggesting that the directional code can be maintained solely based on the internal motion cues derived from the passive rotations. Additionally, some cells in the central complex encoded rotation direction history, a navigational context cue, by increasing or decreasing the firing rate during the stationary periods following clockwise or counterclockwise rotations. Together, these results unveil head direction cell like activity in the insect central complex, which highly resemble similarly functioning cells in the mammalian brain that encode head direction. We predict, that the observed head-orientation coding and directionally sensitive cells are essential components of the brain circuitry mediating insect navigation.

Introduction

Successful navigation is fundamental for animal survival. The most basic aspect of navigation is establishing one’s spatial orientation in the environment. In all mammals studied to date, spatial orientation coincides with the continuous updating of the animal’s azimuthal head-angle among so-called ‘head direction cells’. This process is based upon the animal’s directional heading within its environment, independent of location (Taube,

1995; Taube, 2007; Taube et al., 1990a; Taube et al., 1990b). While insects are known to perform remarkable navigational tasks, it is unknown whether individual ‘head direction cells’ with properties similar to those observed in mammals exist in insects.

In mammals, head direction cells rely upon a variety of sensory inputs to update the highly dynamic directional signal animals receive during navigation. Each head direction cell encodes a single preferred head-orientation, which can be established in reference to environmental sensory cues (viz., allothetic cues, including visual and olfactory information) that serve the role of a spatial landmark (Taube et al., 1990b).

Alternatively, the preferred angle can be established based on continuous updates derived from internal sensory cues that integrate self-movement information (idiothetic cues), such as vestibular flow, optic flow, proprioceptive feedback, and motor efference copy (Taube,

2007; Valerio and Taube, 2012; Yoder et al., 2011) . Head direction cells primarily rely upon allothetic cues to encode orientation, although, in the absence of landmarks, idiothetic cues can update the head direction signal (Taube, 2007; Yoder et al., 2011). A properly updated directional signal without information about the navigational context, however, is not behaviorally meaningful (Griffin et al., 2007; McNaughton et al., 2006; Mizumori et al., 2009). The navigation system needs to be informed about contextual cues, such as a

49 navigational goal, a phase of a behavioral task, or the relative direction of the movements

(left vs. right) that lead to a new head direction, but are not directly encoded in the head direction signal. Such contextual information may also contribute to spatial memory, and thus to adaptive navigation, by providing the animal with a cellular-based reference of past orientation, whereby to compare ongoing and/or possibly future headings (Griffin et al.,

2007; Hasselmo and Eichenbaum, 2005; Mizumori et al., 2009). Whether cells in the insect brain function in this manner to encode spatial information is currently unknown.

The central complex (CX) of the insect brain represents directional components of sensory information processing and navigation in a variety of species (Pfeiffer and

Homberg, 2014) including locusts (Bech et al., 2014; Homberg et al., 2011), crickets

(Sakura et al., 2008), monarch butterflies (Heinze and Reppert, 2011; Reppert et al., 2016), dung beetles (el Jundi et al., 2015), fruit flies (Seelig and Jayaraman, 2015) and cockroaches (Guo and Ritzmann, 2013; Martin et al., 2015). Neurons in the CX represent celestial information in some insects and adjust the animal’s heading towards a defined goal (Bech et al., 2014; el Jundi et al., 2015; Heinze and Homberg, 2007; Heinze and

Reppert, 2011; Homberg et al., 2011; Reppert et al., 2004). Our understanding of the CX’s role in head direction coding, however, is still far from complete (Heinze, 2015; Heinze and Homberg, 2007; Seelig and Jayaraman, 2015). Recently, Seelig and Jayaraman monitored the activity of a population of neurons in one neuropil of the CX, the ellipsoid body (EB), with Ca2+ imaging and found that this population represented the tethered fly’s orientation similarly to mammalian head direction cells (Seelig and Jayaraman, 2015). This is a highly significant advance for the field of insect navigation. Major questions still remain, however, regarding the function of the CX in orientation. Particularly, because the

50 imaging method utilized by Seelig and Jayaraman focused upon quantification of population dynamics, rather than analysis of the activity of individual neurons, it is difficult to draw direct comparisons between their data and the single cells in the mammalian brain known to participate in head direction coding. Thus, we do not know the sensory mechanisms whereby individual neurons in the CX encode head direction. Moreover, do these cells have the capacity to encode other types of spatial information, such as navigational context cues? These are significant questions which, if answered, will complement the population code data (Seelig and Jayaraman, 2015), thereby, extending models of insect navigation strategies.

We considered the cockroach to be an ideal model to understand the fundamental coding schemes underlying navigation. First, cockroaches, versus Drosophila, are nocturnal insects which, similarly to rats, forage and live in dark, maze-like environments

(Bell et al., 2007). Second, cockroaches, versus mammals, must accomplish feats of navigation with a limited central nervous system, void of the executive and affective influences known to modulate mammalian behaviors. These features, combined with the tractability of physiological approaches, make the cockroach a strong contender to unlock mysteries of navigation. Therefore, here we adopted the methods and controls of well- established head direction cell studies in rodents (Shinder and Taube, 2011; Taube et al.,

1990a; Taube et al., 1990b), to allow us to draw comparisons between the observed response characteristics of CX units and those elegantly described in mammalian navigation. We utilized extracellular recordings in the cockroach, Blaberus discoidalis, to test the overall hypothesis that single neurons in the CX encode head direction. We found that individual neurons in the CX of the cockroach encode head direction, as well as

51 information regarding the navigational context, similar to those reported extensively in mammals (Finkelstein et al., 2014; Muir et al., 2009; Robertson et al., 1999; Rubin et al.,

2014; Taube, 2007; Valerio and Taube, 2016) including humans (Chadwick et al., 2015;

Ekstrom et al., 2003; Jacobs et al., 2010; Vass and Epstein, 2013), and thereby provide a compelling example of a highly conserved or convergent navigational system. We propose that further work in the cockroach model will yield additional fundamental computations used by nervous systems to resolve navigation in more complex models (i.e., rodents to humans).

Materials and Methods

Surgical procedures

Cockroaches (Blaberus discoidalis) were housed together in 5 gallon plastic bins in a room with a maintained temperature of 27C and a 12:12-h dark-light cycle. Subjects were 27 healthy adult males with intact limbs and appendages. Each animal was moved from the plastic bin into a smaller colony (<5 subjects at once) housed in a plastic container within the same room, approximately 2 weeks prior to the experiments. The animals were then water deprived (had access to water for one hour 1x a week) to limit hemolymph production during the surgery. The animals had access to food ad libitum.

In the morning of the experiment, subjects were cold anesthetized and implanted with a tetrode (procedures described in detail elsewhere (Guo et al., 2014)). Briefly, anesthetized animals were gently mounted onto a cork with pins and the head was secured in place with a plastic collar and wax. The animals were next placed into a plastic surgical tube (45mm in height, 40mm i.d.) that was surrounded by ice to keep them immobile during the procedure. To expose the brain, a small opening was cut in the cuticle between the ocelli, and the connective tissue and trachea were gently tucked under the remaining cuticle areas, or if necessary, some portions were removed.

Next, a small section of the sheath surrounding the brain was removed in the central areas using fine forceps and a small insect pin. A second opening was created in the cuticle between the compound eyes with a pin. A reference electrode bundle (7/46 SPSN Litz wire,

MWS Wire Industries, CA, USA) was lowered into this second opening. A tetrode composed of four copper-plated 12µm NiCr wires (Kanthal RO-800, Sandvik,

Hallstahammar, Sweden) was slowly lowered into the central complex using a

53 micromanipulator. The tetrode was fixed in the brain at the location that had the best signal:noise ratio and where the units responded to sensory stimuli (i.e., light on or off and/or antennal contact with a small stick (Kathman et al., 2014; Ritzmann et al., 2008)).

The reference electrode and the tetrode were fixed in the head at the same time by injecting light curable clear glue (Loctite 3555 transparent Light Cure Adhesive) into the cavities.

Both the recording and reference electrodes were also fixed to the neck supporting plastic collar with wax, to create a strain relief point for a stable recording site. During head- covered trials, we carefully covered the entire head of the animal with foil obstructing all visual cues (both the ocelli and compound eyes were covered) and most antenna movements. The foil was fixed to the neck collar and the strain relief point without causing any disturbance or quality changes in the recording.

Recording procedures

After tetrode implantation, the subjects were transferred into a custom 3D-printed polylactic acid (PLA) restraint apparatus that was marked with green and red tape for body axis tracking (Figure 2.1A). The restraint apparatus consisted of a 27mm i.d. PLA tube measuring 75mm in length. Because the animal’s head and neck were fixed with a plastic collar, body axis values are referred to as head angle values. The subjects were also slightly immobilized within the restraint apparatus with a 5mm layer of sponge to minimize leg movements. After recovery, the restraint apparatus was placed onto an Arduino-controlled

(Arduino Uno by Arduino Italy/USA) gear platform in a way that the animal’s head was exactly in the midpoint of the rotating platform. The restraint was stabilized on the platform with a thin piece of magnet that matched the size of the restraint (75mm x 75 mm), which

54 also served as a control for excluding possible magnetic cues. The platform was located in the middle of the recording arena, which is a uniformly painted matte black cylinder

(d=40cm; height=30cm) with a single white paper landmark placed on the wall

(width=21.5cm; height=28cm; angular extent=60°; Figure 2.1A). The contrast ratio of the background and landmark was 22.1 (background luminance: 0.07 candela/m2; landmark luminance: 1.55 candela/m2). The arena was covered and surrounded by a white curtain to block out conflicting visual cues. The Arduino board controlled the platform rotations through a DC motor instructed by a custom Arduino motor stepper script.

Figure 2.1: Experimental design and paradigms to test head direction coding in the cockroach. (A) Illustration of the experimental preparation which consisted of a raised rotating platform, cockroach restraint tube, and a single visual landmark which contrasted with the otherwise landmark-void black recording arena. During recordings, physiological data were digitized through a headstage and head-angle was captured by a camera. (B) A 360° rotation (viz., 1 trial) consisted of 12x30° rotations with 10 sec immobile periods between each one of them. Data were only analyzed during these immobile periods (* activity during rotation was not analyzed). See also Figure S1.

All physiological recordings were performed with a Neuralynx Cheetah system at

30 kHz (Bozeman, MT, USA) along with synchronized video-tracking of the head angle 55

(30Hz; recorded with a JAI CV-S3200 camera). Each recording session started with a 5 min period of spontaneous activity while the animal was facing the dark wall of the arena.

Following this period, the subject was exposed to 4 or 6 (depending on the group) experimental paradigms (Figure S1). In each paradigm the cockroach was rotated around the arena 4-6x (4-6 trials/paradigm). Each trial consisted of 12x30° rotations (2 sec; rotation speed 15°/sec) followed by a 10 sec immobile period to allow the animal to sample its environment and for us to be able to look at purely orientation responses (Figure 2.1B).

The maximum duration of a recording was ~1.5 hours. At the end of the experiment the brain was lesioned and copper was deposited from the electrode tips by running DC current through the wires. Tetrode locations were later identified by Timm’s histological staining of the copper deposits (see Figure S2; (Guo et al., 2014)).

A total of 27 cockroaches were tested. The test conditions are demonstrated in

Figure S1. In more detail, subjects in Cohort 1 (8 animals) were exposed to 4 conditions –

‘landmark in control position + CW rotations’; ‘landmark in control position + CCW rotations’, ‘landmark position shifted + CW rotations’; ‘landmark position shifted + CCW rotations’. Subjects in Cohort 2 (13 animals) were exposed to 6 conditions, where we added two more conditions to the already mentioned control conditions: ‘no landmark + CW rotations’ and ‘no landmark + CCW rotations’. Subjects in Cohort 3 (6 animals) were exposed to 4 conditions – ‘head covered + CW rotations’; ‘head covered + CCW rotations’;

‘landmark in control position + CW rotations’; and ‘landmark in control position + CCW rotations’ respectively. Rotation direction (CW vs. CCW) was randomized.

Spike sorting and data analysis

Single unit analysis was performed off-line in Spike2 v7.15 (CED, Cambridge UK).

To sort multi-unit activity into single unit activity, we used user-supervised semi- automated tetrode template matching and K-means assisted principle component analysis.

Any clusters with >3% of all the spike events falling within the 2ms inter-spike-interval criteria were excluded from analysis. Only single-unit activity with a stable amplitude was used for later analyses.

Spiking activity during 8 sec of the immobile periods were analyzed – the first and last 1 sec were not analyzed in order to exclude all possible movement-related activity (that possibly occurring during the start or stop of the platform rotation). The corresponding head angles were identified, binned into 30°angle bins, and exported into Microsoft’s Excel using custom written Spike2 scripts. Spike counts were then organized for each condition/unit and the average spike count/angle bin was calculated in Excel.

All circular statistics were performed on the average spike count/angle bin for each experimental condition separately in MATLAB (MathWorks, Natick, MA) using the

Circstat toolbox (Berens, 2009). We used the Rayleigh test to determine the degree of angle modulation for unimodal units (significance = p<0.05). The same script provided us with the mean vector length (R-vector) values that were used to indicate the spread of the tuning curves, as well as mean vector positions (in angles) that indicated the averaged peak of the tuning curve. We also calculated the half width of the tuning curves in degrees to be able to link the tuning curves described by the R-values to a more intuitive measure (data not shown). To establish such a link, we tested the correlation between R-values and half widths, which showed a strong negative correlation between these variables (p<0.001).

Thus, a large R-value and a small half width of the tuning curve both indicate narrow tuning to an angle. On the other hand a small R-value and a large half width of the tuning curve characterizes a broadly-tuned unit. To better illustrate the tuning characteristics of all angle-modulated units, we normalized the firing rates of each unit (with the peaks aligned) and plotted the individual normalized tuning curves in a two-dimensional histogram

(Figure 2.3). To quantify tuning across the entire population of CX neurons, using the R- values, we sorted all angle-modulated units into quartiles (1: 0< R≤0.25; 2: 0.25

0.5

To test whether there is a correlation between high ‘background’ activity and broad tuning schemes, and low ‘background’ activity (~ minimum firing rate) and narrow tuning schemes, we calculated the signal to noise ratio of all of the unimodal angle-modulated units. We defined signal to noise ratios as the average minimum spike magnitude divided by the average maximum spike magnitude for each unit. We calculated the signal to noise ratios for every unit across all control trials separately for CW and CCW rotations. The signal to noise ratios were then compared to the R-values of the units to uncover a possible correlation between the two variables.

Peak firing rates indicated by gray boxes in all figures represent the R-vector position results of a significant Rayleigh test. For the landmark rotation (LR), ‘no landmark’ and head covered conditions, mean vector positions were used to calculate the amount of shift compared to the expected peak vector position (original mean vector position + amount of landmark shift [90° or 180°]) or the original peak vector position. We considered the peak vector positions to be similar to the expected (or original) angles, if

58 they were no more than 60° away (2x30° bins; based on average errors reported in (Taube et al., 1990a)). Units that responded to two preferred angles were considered bimodal and were tested for deviations from uniformity with the Hodges-Ajne test within the same

MATLAB script (Berens, 2009). These two angles were usually 180° from each other, except when the units developed a bimodal response following 90° landmark position shifts. In these cases the two peaks were 90° apart, which corresponded with the original and new landmark positions. If the distribution of a single unit was significantly different from a uniform bimodal distribution (significance at p<0.05) mean vector positions were obtained through angle doubling (Batschelet, 1981; Berens, 2009) and compared to the expected and/or original mean vector positions. All mean vector errors are reported as average ± standard deviation (SD). Mean spike counts were converted to mean firing rate

(spikes/second) for all figures. Error bars in every figure represent the standard deviation

(± SD) unless otherwise stated. Some figures demonstrate the instantaneous firing rate of an example unit, with a red line overlaid the histogram. The red line indicates the convolved firing rate of that same unit. Convolution was done by applying a Gaussian kernel (Ϭ=10s).

Unimodal units with significant Rayleigh test results were also tested for changes in periodicity (Figure S4). In other words, we looked at how much error accumulated in the signal, or how much the preferred angle shifted over the course of many trials. To test this, we examined the autocorrelation of the spiking frequencies for all of the units passing the above mentioned two criteria. We assigned a directional stability score to each of these units and then cross-correlated the 12 directional firing bins across trials (3-6trials) and averaged the resultant correlation coefficients (for 3 trials: directional stability = (trial1: trial2 + trial1: trial3 + trial2: trial3) / 3; significance at r > 0.576; (Winter et al., 2015b)).

To measure the relative amount of error accumulation in each experiment, we averaged the directional stability scores during the control, landmark rotated, no landmark and head- covered trials. We tested for group differences with a Kruskal-Wallis nonparametric test

(significance at p<0.05).

Results

Central complex neurons encode head direction

We recorded from the CX of head-fixed cockroaches that were rotated in a circular direction (Figure 2.1A). Each cockroach experienced 360° rotations, in 30° increments, 4-

6 times in both clockwise (CW) and counterclockwise (CCW) directions. To test the role of a visual landmark in head direction coding, our recording arena consisted of a uniform black environment containing a large (60° angular extent), solid white landmark which was removable. Our recording paradigm was designed to afford the analysis of CX responses when the animal was immobile, following each 30° rotation step (Figure 2.1B). It is important to emphasize that, even when discussing the effects of rotation direction history, we restricted all of our analyses to epochs wherein the animal is stationary. Further, our paradigm allowed us to selectively test for head direction coding in the context of passive rotations, wherein internal motion cues are the only idiothetic input that can be used by the animal to update a potential navigational system, allowing us to further narrow down the sources of sensory inputs that contribute to the directional signal (Figure S1).

We recorded a total of 173 single units from two CX neuropils, the ellipsoid body

(EB; 99/173 units) and the fan-shaped body (FB; 74/173 units) from 27 cockroaches

(EB:15; FB:12; see recording sites in Figure S2). As illustrated in Figure 2.2A, we established that some units displayed altered firing rates depending upon the animal’s head angle. The single unit shown in Figure 2.2A, for instance, displayed fairly regular firing rates while facing one angle (Figure 2.2A1, ‘non-preferred angle’) which then increased in frequency while the animal was facing a different angle (Figure 2.2A2, ‘preferred angle’).

Indeed, across the entire recording, which included 5 rotations, this unit repeatedly displayed consistent phasic increases and decreases in its firing rate (Figure 2.2A3). This unit’s firing rate, averaged across all rotation angles, displayed a statistically significant mean vector position (i.e. peak) at ~160º, which we will term the unit’s ‘preferred angle’

(p<0.001 Rayleigh test). Looking across the entire population of CX units (n = 173), across all conditions, 37.5% (65/173; 19 animals) of CX units were significantly modulated by head direction during at least one condition (p<0.05 Rayleigh test; 3-7 trials/unit, see also

Figure S4A and E). Among all units that significantly encoded an angle during the control trials (those with the landmark in the original location), the entire 360° environment was represented uniformly (p=0.11; Rayleigh test of unimodal units’ peaks; Figure S3). Thus, similarly to reports of rat head direction cells (Taube, 2007), these units possess the capacity to represent any angle relative to the animal, just like a compass.

Tuning characteristics of head direction encoding neurons

Head direction cells in rats are narrowly-tuned to a preferred angle and as the head turns away from this angle, the firing rate drops near zero (Taube, 2007; Taube et al.,

1990a). Thus, we asked whether directionally sensitive CX cells are narrowly or broadly- tuned to their preferred angles. To test this, we categorized the 65 angle-modulated units based upon their mean resultant vector lengths (R-value). The R-value defines the circular spread, or tuning curve, of a given unit (Berens, 2009) and it is strongly negatively correlated with the half width of a tuning curve (r (df CW = 29) = - 0.81; p<0.001; r (df CCW

=33) = - 0.89; p<0.001). Therefore, we defined units as being narrowly-tuned if their R-

62 values were closer to 1 - the majority of the spikes occurred closely around the preferred angle, and cells as being broadly-tuned if their R-values were closer to 0 – spiking activity was less localized around the preferred angle.

We found that CX units encoded head direction with a range of narrow to broad coding schemes (see examples in Figure 2.2B and C). To quantify tuning across the entire population of CX neurons, using the R-values, we sorted all angle-modulated units into quartiles (1: 0< R≤0.25; 2: 0.25

Figure 2.3A and B show that the majority of head direction coding units are broadly-tuned to their preferred angles, with 83% of units falling into quartiles 1 and 2. The remainder of the modulated CX units are narrowly-tuned, and more precisely encode angle. Next we investigated the relationship between the tuning schemes and ‘background’ firing rates of angle-modulated units. For both rotation directions, the R-values and minimum firing rate showed a strong negative correlation (r (df CW = 29) = -0.59; p<0.001 and r (df CCW = 33)

= -0.54; p<0.001, Pearson’s correlation). Thus, narrowly-tuned units have lower

‘background’ firing rates, while broadly-tuned units have higher ‘background’ firing rates.

We also found a strong negative correlation between signal:noise (minimum firing rate divided by maximum firing rate) and the R-value of angle-modulated units (r (df CW = 29)

= -0.86; p<0.001 and r (df CCW = 33) = -0.85; p<0.001, Pearson’s correlation) suggesting that the width of the tuning curve increases as the signal:noise decreases.

Figure 2.2: CX units encode head direction by changes in firing rate. (A) Example unit response (A1) Single- (SUA) and multi-unit activity (MUA) of tetrode channel (ch) 3 during 8sec of facing a non-preferred angle. Also shown are waveforms for channels 1-4 and the arrangement of the tetrode channels. (A2) Same as in (A1), but from when the animal is facing a preferred angle. (A3) Average firing rate/angle bin (8 sec of stationary period each) of the single unit in A1 and A2 for the first 3 trials. The landmark (horizontal black bar) was maintained in control position, CW rotations. # and ^ indicate the 8sec periods shown as the example traces above. (A4) Average firing rate ± SD across 5 trials following CW rotations. The unit was significantly modulated by head direction. p<0.001 Rayleigh test. (B) Example of a narrowly-tuned unit and (C) Example of a broadly-tuned unit. (B1) and (C1) Circular representation of firing following clockwise rotations. Purple to green indicates the increase in firing rate as the animal rotates from non-preferred to preferred angles. Arch represents landmark position (65°-125°). The units were significantly modulated by head direction, p<0.001 Rayleigh test. Red line represents R-vector (length and direction). (B2) and (C2) Mean firing rate for each angle bin in 5 trials ± SD. (B3) and (C3) show the unit’s instantaneous firing rate in 1 sec bins for each trial of the experiment. Red line = Gaussian filtered (Ϭ = 10s) unit activity. See also Figures S2, S3 and S4. 64

Does the location of the recording within the cockroach CX (EB vs. FB) affect the unit’s tuning characteristics? During control CW and CCW trials, 37.4% (37/99; 11 animals) of units recorded in the EB significantly encoded an angle (unimodal units) and

5% (5/99; 4 animals) encoded two angles (bimodal units). Only 10.8% of FB units (8/74 units; 5 animals) were unimodal and 4% (3/74 units; 3 animals) were bimodal. Thus, the

EB contained significantly more angle-modulated units than the FB (χ2 (1) =7.471; p=0.006, two-tailed with Yates correction). When categorizing the R-values of unimodal

EB units, we found that the majority of them were broadly-tuned (CW: 83%; CCW: 77% in quartiles 1 and 2) and the remaining portion of them were narrowly-tuned (CW: 17% in quartiles 3 and 4; CCW: 23% in quartile 3). All categorized FB units were unimodal and broadly-tuned (CW:100% in category 1; CCW: 60% in category 1 and 40% in category 2).

These results suggest that a subpopulation within the EB may participate more in narrowly- tuned head direction coding than the FB (χ2 (1) =0.575; p=0.448, two-tailed with Yates correction). Importantly, broadly and narrowly-tuned units were found within the same recordings in the EB on several occasions (largest range of R-values in one site: 0.05 to

0.93), suggesting a lack of anatomical segregation for the different coding schemes at least within the EB. Altogether, the above analyses are consistent with the hypothesis that a subset of individual neurons in two subunits of the CX encode head direction and do so along a variety of coding schemes.

We next asked whether the rotation direction in which the animals were moved influenced the encoding of head direction in CX neurons. We compared tuning curves across all angle-modulated units in the control CW and CCW conditions. The tuning curves

65 of the entire ensemble of angle-modulated units were statistically similar (p=0.822; paired two-tailed t-test on R-values). We found that only 40% of these units encoded an angle during both CW and CCW control conditions with an average of 39.9°±44.9° difference in peak positions (ranging from 2.7°- 197.2°), thus we categorized the data separately based on rotation direction. All categorized R-values were divided into two groups for significant angle response following CW (n=31 units; Figure 2.3C1-C2) and CCW (n=35 units; Figure

2.3D1-D2) control rotations. Using these criteria, we found that the majority of units across the two groups were broadly-tuned (CW: 85%; CCW: 80%) and the remaining portion of them were narrowly-tuned (CW:15%; CCW: 20%). These data show that rotation direction history affects the position of the encoded angles, but does not affect the breadth of tuning.

Figure 2.3: Tuning characteristics of angle-modulated CX units. (A) 2 dimensional histogram illustrating the normalized firing rates of angle-modulated single units for every 30° bin. The normalized tuning curves were aligned at the peak. We utilized R-values to indicate the spread or tuning of a cell. An R of 0-0.5 indicates that the cell is broadly-tuned, whereas an R of 0.5-1 means the cell is narrowly-tuned to an angle. Based on this criteria the normalized tuning curves were organized into four quartiles (Q1-Q4). The average normalized tuning curves ± SD are depicted in line graphs to the right. (B) Pie chart of R-value distribution showing that the majority of units were broadly-tuned (R=0-0.5) during all control trials. (C1) and (C2) R- value distribution separately for CW rotations and (D1) and (D2) for CCW rotations only. Rotation direction does not significantly impact angle tuning characteristics (p=0.822; paired two-tailed t-test). See also Figure S3.

Head direction encoding CX neurons rely upon allothetic and/or idiothetic cues

Directional responses in rodents may encode the inner representation of, and tuning to, allothetic cues, or they could be a product of self-motion (idiothetic) cues (McNaughton

67 et al., 2006; Taube, 2007). To test whether CX cells are tuned to the artificial landmark in the animal’s proximal environment (allothetic cue), we rotated the landmark’s position by

180° or 90° following control trials (in 15 and 6 animals respectively). 38.3% of the 128 units recorded (a subset of those presented above in Figures 2.2-2.3) were angle-modulated during at least one of the above mentioned trials (49 units/16 animals; 69% EB and 31%

FB units; Figure S4B and E). We analyzed landmark positional effects separately for CW and CCW trials, to avoid confounds which may arise from also incorporating rotation direction. We calculated the mean vector position (i.e., peak) for each unit during control and landmark-rotated (LR) trials. Then, we compared the peaks across LR trials to the expected peaks (control peak + the amount of shift in landmark position [180° or 90°]). To account for the possibility that binning the data might have led to slightly shifted peaks, and taking into consideration that even rat head direction cells have an average error of

~20° (ranging up to 48° (Taube et al., 1990b)), we set a maximum of 60°of acceptable error when comparing any peak positions.

We found five categories of effects among the 49 angle-modulated units (see Table

2.1 for details). In 20.4% of CX neurons (10/49 units; 8 animals) the shift in landmark position coincided with a near-equal shift of the cells’ peaks (example Figure 2.4A; Table

2.1, 1.Allothetic). These data support the hypothesis that the CX participates in allocentric navigational processes, wherein a unit’s directional tuning is determined by external reference points in the environment.

The second category of angle-modulated units (14 units; 28.6%; 7 animals) had persisting peak positions that were locked to the initial landmark position even after it had been shifted (example Figure 2.4B and details in Table 2.1, 2.Idiothetic). These units were

68 either modulated by an unknown, but stable, landmark, or they used idiothetic cues to maintain their directional tuning. Alternatively, these units might have relied upon short- term memory processes to encode the original preferred angle established in reference with the control landmark position throughout the entire experiment regardless of subsequent changes in landmark position.

The third category of units (12 units; 24.5%; 7 animals) were bimodal and encoded two preferred angles (Figure 2.4C; detailed results in Table 2.1, 3.Two peaks, 180°). A subpopulation of these units (5/12 units; 2 animals) significantly encoded one angle during the control trials and developed the additional second peak during the LR trials. In all of these cases one peak encoded the original single preferred angle and the second peak followed the landmark shift, and encoded the expected preferred angle. 4/12 of the units (3 animals) had two peaks during the control trials and only one peak during the LR trials.

Except for one unit, all of these examples (3/4) encoded one of the original peaks during the LR trials, which suggests that this consistent peak was established using idiothetic cues.

The remaining 3/12 units from the third category experienced a 90° rotation in landmark position (2 animals; data in Table 2.1, 3.Two peaks, 90°). All 3 of these units were bimodal during the control trials and 1/3 unit responded to the rotated landmark with a single peak, that persisted to encode one of the original peaks. The other 2/3 units (from the same animal) persisted to display two peaks during the rotated trials as well. In both cases the two peaks from the control CW trials remained approximately in their original positions during CW LR trials, but shifted following the landmark shift during CCW LR trials compared to the control CCW trials.

Table 2.1: . Landmark rotation experiment results indicate that CX units utilize five sensory strategies when encoding head direction. AVG AVG AVG error AVG error Shift error if error if N of Ctrl vs. if if in modulated modulated Category units % original/ modulated modulated % in land- after after of sensory (% of in expected after CW after CW FB mark CCW ctrl CCW ctrl strategy modu- EB peak(s) ctrl or CW ctrl and posi- or CCW and lated) p-value LR CW LR tion LR CCW LR (± SD) (± SD) (± SD) (± SD) 90° 1. 10 26.6± 8.6°± 31.2°± 50% 50% or 0.416 17.8° Allothetic (20.4%) 17.8° 5.1° 14.5° 180° 90° 2. 14 31.9°± 26.0°± 29.4°± 26.8°± 93% 7% or 0.764 Idiothetic (28.6%) 16.2° 14.9° 16.6° 12.5° 180° Two peaks 10.3°± 10.3°± 8.1°± 8.1°± 180° 0.598 during 10.3° 10.3° 2.9° 2.9° LR Two 3. 12 peaks 34.2°± 28.8°± Two 83% 17% 180° 0.681 N/A NA (24.5%) during 34.3° 32.5° peaks ctrl Two peaks 14.2°± 14.2°± 18.1°± 18.1°± 90° 0.625 during 12.7° 12.7° 9.9° 9.9° ctrl CW LR at expected 0.760 27.7°± 31.8°± 24.5°± and 12.3° 19.3° 19.9° 1.9° CCW LR at 90° 4. 10 original 50% 50% or Mixed (20.4%) 180° CW LR at original 32.7°± 30.9°± and 0.495 38.7° 41.4° 21.2° 10.5° CCW LR at expected AVG error AVG error when when 5. compared 3 compared Random 33% 67% 180° 0.06 to (6.1%) to original: shift expected: 102.5°± 77.5°± 19.2° 19.2° Total number of units = 128. Number of unmodulated units = 49 (38.3%). EB, ellipsoid body; FB, fan-shaped body; LR, landmark-rotation trials. AVG error, average error ± SD in peak shift compared to the expected and/or original peak. NA, not applicable. p-values indicate the results of paired two-tailed t-tests.

The fourth category of angle-modulated units (10/49; 7 animals) were ‘mixed’ in their responses (Table 2.1, 4.Mixed). These units were unimodal, but displayed different response patterns following CW and CCW rotations. Specifically, the peaks were locked to the control landmark position after one rotation direction and shifted with the landmark when it was rotated around the other way. These examples support the hypothesis that CX neurons not only encode head direction, but may also respond to contextual changes, such as rotation direction history, by selecting a different spatial coding strategy (allothetic vs. idiothetic).

The remaining category of 3/49 angle-modulated units unpredictably shifted their peaks which could indicate a general remapping in the directional code due to the landmark positional changes (3 animals; data in Table 2.1, 5.Random shift).

Taken together, these categories of CX units reflect that head direction coding by

CX units may occur by means of encoding the inner representation of, and tuning to, allothetic cues and/or they could be a product of self-motion (idothetic) cues (McNaughton et al., 2006; Taube, 2007).

Figure 2.4: Visual landmark position determines head direction coding. (A) Head direction coding tuned to allothetic cues, wherein angle-modulated units follow the shift in landmark position by shifting their peaks. Example unit’s mean firing rate ± SD over 6 trials. (B) Head direction coding tuned to idiothetic cues, wherein angle-modulated units persist to encode the original peak. Example unit’s mean firing rate ± SD over 4 trials. (C) Bimodal responses during landmark rotation trials. These units developed a second peak in response to the new landmark position, while the original peak persisted to encode the peak from the control trials. Example unit’s mean firing rate ± SD over 6 trials. Black line: current landmark position. Gray line: previous landmark position. All examples were modulated by head direction, p<0.05 Rayleigh test. Gray boxes show peaks (i.e. original, shifted and expected mean vector positions). See also Figure S4.

Head direction coding persists even in the absence of visual landmarks

The preceding analyses suggest that visual cues are sufficient to subserve head direction coding in CX neurons. To test whether visual cues are necessary for head direction coding, we removed the landmark from the recording arena. In these experiments, which followed a subset of the previous experiments (81 units, 13 animals), we first rotated the animals around with the landmark present, then shifted the landmark position (LR trials, same data as in Figure 2.4 and Table 2.1), and lastly removed the landmark from the arena

(‘no landmark’ trials; see Figure S4C and E). 27.1% of the recorded units (9 animals; 22/81 units) established a directional tuning during the control trials of the experiment. A subpopulation of angle-modulated units continued to encode head direction even in the absence of the landmark (36.4%, 8/22 units; 6 animals; Figure 2.5A).

Units that encoded the same head direction during both control and LR trials (5/8;

62.5%) persisted to encode the same angle during the ‘no landmark’ trials (average error during CW trials: 23.7°±19.1°; during CCW trials: 18.2°±21.9°; control vs. ‘no landmark’ peaks: p=0.757; paired two-tailed t-test). These data strongly support the hypothesis that idiothetic cues are sufficient to inform head direction coding in the CX. While one unit

(1/8; 12.5%) unpredictably shifted its peak, the remaining units in this subpopulation (2/8;

25%) encoded the same angle during LR trials and ‘no landmark’ trials (average error during CW trials: 18.6°±7.8°; during CCW trials: 33.0°±5.8°; LR vs. ‘no landmark’ peaks; p=0.944; paired two-tailed t-test).

These examples support the hypothesis that some units might rely upon short-term memory or idiothetic cues to update the directional signal when no visual landmarks are available.

We confirmed this finding by recording from a cohort of 6 additional animals (45 units) none of which had experience with the recording arena or the visual landmark. These

‘landmark naïve’ animals were placed in the arena with no landmark, then we completely covered their heads with aluminum foil to block all possible visual cues. The head-covered animals were then rotated in CW and CCW directions with absolutely no stationary visual information available to them. Following these trials, the foil was removed and the animals were rotated around CW and CCW, experiencing the control conditions with a single visual landmark. Strikingly, even in the absence of visual cues, only relying upon the motion cues derived from the rotations, some units significantly encoded head direction (8/45; 17.8%;

Figure 2.5B and Figure S4D and E). During the second, control part of the experiment, 9 additional CX units (17/45; 37.8%) established a significant angle response. We analyzed all of the units that were angle-modulated during the head-covered and/or following control trials. Most units (14/17 units for CW (χ2 (1) = 3.712; p=0.054) and 11/17 units for CCW

(χ2 (1) = 0.485; p=0.4761) both two-tailed with Yates correction) unpredictably shifted the peak position after the foil was removed suggesting a general remapping in the directional signal caused by the newly available visual cues (CW average shift: 118.9°±41.3°; CCW average shift: 130.3°±38.5°; head-covered vs. control condition peaks; p=0.152; paired two-tailed t-test). Meanwhile, other units (2/17 for CW and 5/17 for CCW) did not significantly change their peak positions after the landmark became visible (CW average shift: 34.76°±33.2°; CCW average shift: 25.2°±17.5°; head-covered vs. control condition peaks; p=0.860; paired two-tailed t-test). These findings not only support the hypothesis that internally available references derived from passive movement are sufficient for

74 encoding head-orientations, but further confirm that some units rely solely upon idiothetic cues even when visual cues become available.

Figure 2.5: CX units encode head direction in the absence of visual landmarks or any visual input. (A) CX units are able to maintain already established directional activity in the absence of visual landmarks; p<0.05 Rayleigh test. (B) Example unit that encoded head direction in a head-covered landmark naïve animal; p<0.05 Rayleigh test. The second graph shows the average firing rate of the same unit over 4 trials in 1 sec bins. Red line = Gaussian filtered (Ϭ = 10s) unit activity. See also Figure S4.

Central complex units encode rotation direction history

Some neurons in the hippocampal formation of mammals are influenced by, and encode navigational context cues, such as the history of rotation direction (Jacobs et al.,

2010). When only considering the control CW and CCW conditions, we found that 68.8%

(119/173) of CX units did not encode head direction. We tested whether these units encode rotation direction history by comparing their firing rates during the stationary periods following CW vs. CCW rotations. Despite the fact that on the population level rotation direction history did not affect the firing rate (p=0.863; paired two-tailed t-test on mean spike number per unit during all CW vs. CCW trials), we found that 69.7% (83/119; 25 animals) of individual units displayed significantly greater firing rates at each angle during the stationary periods after the animal was rotated in a particular direction (p<0.05; paired two-tailed t-tests per unit comparing mean spike number across every angle bin for all CW vs. CCW trials; Figure 2.6A, B, D and E). 29/83 units (35%; 14 animals) encoded the direction it was first exposed to with a higher firing rate, while significantly more units

(54/83; 65%; 19 animals) had increased firing rates in response to the new rotational direction (χ2 (1) = 4.481; p=0.034; two-tailed with Yates correction). These results show that a subpopulation of units in the CX may play a different role in navigation by storing information about navigational context, such as directional history.

Upon finding that some neurons encode directional history, we considered that an optimal navigational strategy would be for some neurons to represent two discrete features of spatial information (head direction and the memory of rotational direction), in a manner sometimes termed ‘multiplexing’. We tested this possibility across all units showing angle- modulation during the control trials (54/173 units; 31.2%). In these units, we tested whether the firing rates during exclusively immobile periods depended upon the recent rotation direction. Although, we found that on the population level rotational history did not affect the average firing rate of angle-modulated units (p=0.419; paired two-tailed t-test on mean spike number per unit during all CW vs. CCW trials), a subpopulation of these units (37/54

76 units; 67.3%; 19 animals) did encode rotational history as well, by significantly changing spiking activity (p<0.05; paired two-tailed t-tests per unit comparing mean spike number across every angle bin for all CW vs. CCW trials; Figure 2.6C-E). A smaller portion of these multiplexing units (43.2%, 16/37; 9 animals) responded to being rotated in the original direction with a higher firing rate, while 56.8% of units (21/37; 10 animals) responded with an increase when the platform was being turned in a new direction (χ2 (1)

= 0.221; p=0.638; two-tailed with Yates correction). Together, these results show that some individual neurons in the CX multiplex two major types of orientation information, head direction and rotational history, in a manner possibly serving adaptive navigation.

Figure 2.6: Past rotation direction affects CX unit firing rate during the stationary epochs. (A) Example unit not modulated by angle increased its firing rate following CW rotations. (B) Example unit that increased its firing rate following CCW rotations. The two units are from the same recording. p<0.05, two-tailed paired t-test. (C) A representative example of a CX unit that significantly encoded a preferred head direction and increased its relative firing rate during the stationary epochs following CW rotations. p<0.05 for both Rayleigh test and two-tailed paired t-test. (D) Mean firing rate across all trials and angle bins during stationary epochs following rotations in the preferred direction and in the non-preferred direction separately for units that did not respond to head direction and ones that significantly encoded head direction. The population mean ± SD is marked by the horizontal and vertical black lines. *** p<0.001, paired two-tailed t-test. (E) Distribution of p-values indicating the level of significance when firing rate following preferred rotation directions was compared to the firing rate following the non-preferred rotations.

Discussion

The neural basis of navigation is a fundamental problem to solve in order to understand the mechanisms underlying critical animal behaviors. Here, we report that individual cells in the CX of the cockroach encode head direction, as well as the history of rotation direction in manners strikingly similar to those reported in rodents (Muir et al.,

2009; Taube et al., 1990a; Yoder and Taube, 2009), bats (Finkelstein et al., 2014; Rubin et al., 2014), non-human primates (Robertson et al., 1999), and humans (Chadwick et al.,

2015; Ekstrom et al., 2003; Jacobs et al., 2010; Vass and Epstein, 2013) and do so through modulation of the firing rates of individual cells. We predict these angle-coding neurons are essential components of the brain circuitry mediating insect navigation.

It was recently discovered that population codes within the Drosophila EB reflect head-orientation during walking (Seelig and Jayaraman, 2015). Here, we add to this growing body of literature on CX navigational coding schemes by exploring sensory orientation of individual neurons within the populations that make up the cockroach CX.

Since it was established in the paper by Seelig and Jayaraman, that the monitored population’s response was similar in complex and single-landmark environments (Seelig and Jayaraman, 2015), we tested the question of head direction coding by utilizing one landmark. While unlikely, it is possible that some of the responses observed during the control and LR trials are due to the units’ tuning to the landmark’s visual features rather than its position. However, units that followed the idiothetic, bimodal and mixed responses argue against this possibility (Figure 2.4 and Table 2.1). Additionally, because of our relatively narrow landmark, most of the peaks in these experiments did not line up with any features of the landmark. Rather, the peaks represented a heading relative to the

79 reference (which in many cases also happened to be outside of the animal’s visual field).

Together, our landmark removal and blind-folded experiments demonstrate that CX neurons encode head direction.

A major goal of the present study was to determine whether CX neurons rely on the same hierarchy of sensory information as angle-coding cells reported in mammals

(Finkelstein et al., 2014; Taube, 2007; Yoder et al., 2011). For instance, in rats, head direction cells primarily rely upon allothetic cues, but can update the directional signal based on idiothetic cues if needed (Taube, 2007). What idiothetic cues do animals need to establish head direction? Some answers to this question in rodents come from studies comparing angle coding in freely-walking animals to that of passively-transported ones.

Freely-walking animals have access to all idiothetic cues, including proprioception and motor feedback, as well as motion cues (vestibular cues in mammals). In contrast, passively-transported animals only have access to motion cues. Several rodent studies have confirmed that passive displacement does not abolish head direction cell activity, indicating that motion cues derived from vestibular inputs alone can maintain the signal

(Blair and Sharp, 1996; Shinder and Taube, 2011). Thus, the utilization of sensory cues by head direction cells in rodents is highly adaptive to ensure optimum navigation even in dynamic sensory contexts.

Our results support the hypothesis that head direction coding in the cockroach relies upon similar sensory pools as in rats (Blair and Sharp, 1996; Knierim et al., 1998; Taube,

2007; Taube et al., 1990b). Specifically, we found that neurons in the CX encode head direction through three main strategies (allothetic [Figure 2.4A]; allothetic and idiothetic.

[Figure 2.4C]; idiothetic [Figure 2.4B]). CX neurons following the idiothetic strategy

80 continued to rigidly encode the originally established preferred-angle even after the landmark was repositioned, possibly indicating that this angle may have been derived from internal cues. Alternatively, it is possible that idiothetic cues can override head direction cell responses when the shift in the landmark’s position is close to 180°, as reported by

Knierim et al. (1998) (Knierim et al., 1998). Thus, these neurons might have originally encoded an angle based on an allothetic cue and when that cue became unreliable (shifted), they switched to idiothetic cues to update the same angle.

In support of the hypothesis that idiothetic cues alone may be sufficient to update the head direction signal, we also found that CX neurons persisted to encode head direction following landmark removal (Figure 2.5A) and even in ‘blind-folded’ animals (Figure

2.5B). These results not only indicate that orientation coding can be based on idiothetic cues, but also because of our passive-rotation experimental design, they additionally support the hypothesis that, similarly to rodents (Blair and Sharp, 1996; Shinder and Taube,

2011), motion cues alone can update the head direction system. Importantly, such motion cues have to integrate continuous information about speed and position in order to continuously update the dynamic head direction signal. Previous work from our lab described neurons in the CX that encode the angular velocity of tethered animals, which have the potential to update the described head direction system about changes in speed

(Guo and Ritzmann, 2013). The angular velocity signal might travel through neural pathways reminiscent of the vestibular pathways in the brains of mammals, which originate from the Johnston’s organs in the insect antennae (Kamikouchi et al., 2009; Matsuo et al.,

2014; Yorozu et al., 2009). These pathways might be responsible for providing the insect navigational circuit with continuous motion cues. Future studies are needed to define the

81 brain region(s) needed to encode the position of the animal. The finding of directionally tuned cells in insects provides an essential starting point to these future directions, given the reliance of many spatial coding schemes upon head direction cells in rodents (Winter et al., 2015a).

In addition to directional and positional information, successful navigation requires updates and memories upon context-dependent information. Contextual information, such as navigational goal, or movement direction are encoded in navigation centers of rodent and human brains (Chadwick et al., 2015; Ekstrom et al., 2003; Jacobs et al., 2010;

McNaughton et al., 1983; Mizumori et al., 2009). For instance, McNaughton et al.

(McNaughton et al., 1983) showed that ‘complex-spike cells’ in the hippocampus consistently increased their firing rates when the animal was moving in a particular direction within a maze. Jacobs and colleagues (Jacobs et al., 2010) described ‘path cells’ in the human entorhinal cortex, which varied their firing rate depending on whether the patients were driving in a CW or CCW direction in a virtual-navigation game. Here, we found that a subpopulation of CX neurons encoded past rotation direction information during stationary periods (Figure 2.6A, B, D, E). The directional effect on firing rate was consistent during the long stationary periods (data not shown), indicating that the increase in firing rate is not caused by movement-related rebound effects (van der Meer et al., 2007).

Further, we resolved that some CX neurons integrate or ‘multiplex’ both the current head direction and past rotation direction through changes in firing rate (Figure 2.6C-E).

Additionally, we found examples of rotation direction impacting the applied head direction coding strategy within a single cell. This finding is highly novel, and, to our knowledge, is the first example of context-dependent spatial information coding in an insect neuron. We

82 predict that these cells may underlie complex behaviors by providing context-dependent information for action selection (Strausfeld and Hirth, 2013) and place learning (Liu et al.,

1999; Ofstad et al., 2011; Strausfeld and Hirth, 2013), which contribute to adaptive navigation (Mizumori et al., 2009).

In comparison to what is known in rodents, little is known regarding the cellular features of adaptive navigation in insects [27]. Heinze and Homberg (2007) showed that the CX in locusts contains a sun compass which can encode the animal’s head direction relative to celestial cues (Heinze and Homberg, 2007). The neural signal containing directional skylight information is altered as it travels through several stages of processing from the input areas to the output areas of the CX. Specifically, early stage neurons are more narrowly-tuned with lower background firing rates, while later stage neurons are broadly-tuned displaying higher background firing rates (Heinze et al., 2009). Our results are in accordance with these data, wherein narrowly-tuned neurons were only found within the EB and the FB only contained broadly-tuned neurons.

Elegant work by Seelig and Jayaraman (2015) established that a certain population of CX neurons encode the fly’s orientation predominantly relying upon visual landmarks

(Seelig and Jayaraman, 2015). Their results indicate that integrating idiothetic cues only may lead to error accumulation in the fly’s directional code. Our results however, showed that the amount of error accumulation is similar during control and head-covered trials, thus motion cues alone are sufficient for cockroaches to maintain a reliable directional signal (Figures 2.5 and S4). This difference could stem from the ecology of the two model animals. Drosophila are diurnal flying insects which, under natural circumstances, normally have access to external (visual) cues to update their navigation system.

Cockroaches, however, are more similar to rats, which have limited sources of visual landmarks and need a navigation system that can precisely encode orientation using idiothetic cues. Together the results of these studies provide fundamental insights into general coding strategies relying upon specific sensory cues whereby the CX may critically mediate insect navigation.

In summary, our results indicate that the navigational systems of evolutionarily distinct animals (insects to humans) might rely upon the same general coding principles, while adjusting the mechanisms to adapt to specific ecological needs. This similarity opens up a whole new level of questions regarding the conservation of navigation and spatial memory.

Chapter 3

Modulation of Central Complex Local Field Potentials by Head Direction and Spatial Context

Summary

Local field potentials (LFPs) are considered to be important mediators of navigation in mammalian spatial circuits. In this paper we examined LFPs in the CX of restrained cockroaches and found that spontaneous CX LFPs consist of coexisting slow and fast rhythms. More importantly, we found that some of the LFP spectral bands are modulated by head direction as well as changes in the environmental context in a passive rotation paradigm. Head direction coding by delta-band activity was the most prevalent even when a landmark was not available to the animal. Changes in the sensory context of the experiment led to a significant decrease in delta-band modulation by head angle. On the other hand, although a less frequent effect, theta-band activity encoded a preferred angle similarly in all experimental conditions. These results suggest that the head direction encoded by theta-band modulation is independent of the sensory context or might rely upon internally available sensory cues to encode head direction. Meanwhile, changes in relative response magnitude, as well as average power in the theta-, beta-, and gamma-bands indicated that CX LFPs may encode information about the spatial sensory context independent of head direction. These sensory processes supported by oscillations might have an important role in adaptive navigation, by serving a foundation for sensory context discrimination. We predict that analysis of LFPs underlying insect navigation will contribute to a broad comparative approach to understand the general principles as well as the diversity of navigation circuits.

Introduction

Neural activity can be observed in several forms, including action potentials and network-level local field potentials (LFPs). As it becomes clear that neurons exert control within relatively large ensembles, the timing mechanisms of elements within those ensembles becomes a critical factor (Fazelpour and Thompson, 2015; Fell and Axmacher,

2011; Lakatos et al., 2008; Rodriguez et al., 1999; Ward, 2003; Womelsdorf et al., 2007).

At least one way in which neural activity can be synchronized involves LFPs which arise from a range of neural properties including action potentials, synaptic events and channel kinetics (Buzsáki et al., 2012). LFPs synchronize neural ensembles to influence and interact with downstream targets in the brain (Kopell et al., 2010; Tiesinga and Sejnowski, 2010).

Both action potentials and LFPs are considered important in encoding sensory information, including navigationally-relevant information (Wehr and Laurent, 1996; Hasselmo, 2005).

While mammalian studies have provided us with information regarding the relationship between LFPs and spatial navigation, the exact role of these oscillations in head direction coding remains unclear (Brandon et al., 2013; Hartley et al., 2014; Huxter et al., 2008).

LFPs can be functionally segregated into frequency bands based on experimental observations (Buzsáki and Draguhn, 2004). Based on our current understanding, the most important frequency bands that contribute to adaptive navigation and spatial memory in mammals are theta-band activity (~ 4 - 10 Hz) and delta-band activity (~0.1 - 4 Hz )

(Buzsáki and Draguhn, 2004; Buzsáki et al., 2013; Ulanovsky and Moss, 2007).

Head direction cells are important components of the mammalian brain’s navigation system that directly contribute to proper grid cell function (Winter et al., 2015a).

Yet the mechanisms controlling the temporal organization of head direction cell activity

87 are currently unknown. Several studies indicate that LFPs may be responsible for synchronization within the head direction network. For instance, some populations of rat head direction cells exhibit a high degree of rhythmicity with LFPs within the theta band range (Boccara et al., 2010; Tsanov et al., 2011). These cells encode head direction predominantly through spike trains discharged at theta frequency. Similarly, Brandon and colleagues found that some head direction cells fire action potentials during alternating theta cycles (Brandon et al., 2013). Such theta-driven rhythmicity is thought to facilitate the temporal segregation of neurons that have overlapping preferred directions. Additional observations suggest that the relationship between head direction cells and the segregation by oscillatory cycles is stable, which indicates that oscillations can possibly contribute to the organization of the head direction signal (Tsanov et al., 2011; Tsodyks et al., 1996).

These data suggest that the mammalian head direction system is a component of, or at least partially affected by, the hippocampal theta network and therefore may be organized by theta oscillations (Jeffery, Donnett and O’Keefe, 1995; Deshmukh et al., 2010). If true, this suggestion would mean that a complete understanding of the role of head direction cells in navigation must include analysis of LFP oscillations.

The specific role of theta-band oscillations in navigation is not widely accepted, however. Indeed, in contrast to rodent systems, described above, navigation networks in humans, primates and bats rarely fire in theta-rhythmic manner, and even when they do, oscillations in the theta-band are intermittent and only last for a short period of time

(Ekstrom et al., 2005; Stewart and Fox, 1991; Ulanovsky and Moss, 2007; Yartsev et al.,

2011). For instance, in bats, theta rhythmicity is only present during periods of environmental exploration via echolocation and is accompanied by large delta-band

88 components (Ulanovsky and Moss, 2007; Yartsev et al., 2011). Nevertheless, these animals are still able to perform extraordinary navigational behaviors, which questions the exact role of the synchronizing theta rhythm, at least within non-rodent navigation circuits.

In every animal, navigation requires two main sources of information: the inner representation of the animal’s current position and orientation, and memories of previous positions and movements that led to them (Buzsáki and Moser, 2013). In rats, it has been hypothesized that a navigational path and the corresponding positions and orientations of the animal may be encoded by mechanisms similar to episodic memory, wherein sequences of information are temporally organized by theta oscillations (Buzsáki and Moser, 2013;

Huxter et al., 2003). However, the results of bat, primate, and human studies force us to rethink the role of theta in navigation. Is it possible that theta synchronization is more important for memory than navigation? Could non-rodent navigation systems be organized by different, non-theta, oscillatory patterns? Are there other alternative solutions to organizing navigational network activity?

Another form of synchronization is inferred by computational models. In this case head direction cells that encode similar preferred angles appear to synchronize their activity independent of oscillations and work in a so-called ‘ring-attractor network’ (Knierim and

Zhang, 2012; McNaughton et al., 2006; Skaggs et al., 1995). In a ring-attractor network, synchronized groups of head direction cells form an activity bump that moves on a virtual ring as the animal moves its head or walks in a different direction. The connections between the participating cells are assumed to be internally organized, thus the system works even when external cues are not available to update the cells. This model was supported by

89 recordings in rats where the activity bumps of several head direction circuits remained temporally organized even during sleep, when the cells do not receive any external sensory inputs, nor vestibular inputs (Peyrache et al., 2015). Also, rat pups have normally organized head direction cells even before they open their eyes, which further supports the idea of internal organization that may be independent of oscillations (Bjerknes et al., 2015). Of course, attractor dynamics and LFP influence are not necessarily mutually exclusive.

Insects provide us with a tractable system to study the neural dynamics of navigation, since their brains are several folds smaller, yet their ability to navigate is comparable to and in some cases exceeds that of mammals. Separate studies from Seelig and Jayaraman, as well as from our laboratory have provided initial evidence for head direction coding in the insect central complex (CX) (Seelig and Jayaraman, 2015; Varga and Ritzmann, 2016). Our results showed that CX neurons in the cockroach fan-shaped body (FB) and ellipsoid body (EB) encode head direction in a manner highly similar to mammalian head direction cells (Varga and Ritzmann, 2016). We reported that the head direction signal depends on the same range and hierarchy of sensory inputs as in rats. The recorded cells established their preferred angles based on a landmark’s position and rotated in response to landmark shifts. The head direction signal was present even in situations where motion cues were providing the only updates to the system. Seelig and Jayaraman

(2015) used two-photon Ca2+ imaging to monitor the responses of columnar neurons in the ellipsoid body (EB) in the Drosophila CX in a closed-loop orientation experiment. At certain headings relative to the displayed landmark’s position, active cells formed an activity bump wherein projections going to several neighboring areas showed increased activity. Whenever the fly changed its heading, the activity bump rotated. These

90 observations support the idea of an internally organized ring attractor underlying the head direction signal in the insect CX.

LFP oscillations are considered important mediators of insect nervous system function (e.g. Laurent and Davidowitz, 1994; Wehr and Laurent, 1996; Nitz et al., 2002; van Swinderen and Greenspan, 2003; Broome, Jayaraman and Laurent, 2006; Ito et al.,

2009). Thus, it is reasonable to expect that LFPs play a role in the large ensembles that make up the CX. Given the size and importance of the CX neuropils, it may be surprising that a description of spontaneous LFP activity specifically in the insect CX is not available.

Additionally, the network dynamics corresponding with head direction and spatial context discrimination in the insect brain are completely unknown. We predict that oscillations in the CX may underlie the organization of the head direction signal. That hypothesis can only be addressed after an analysis of LFPs associated with the CX has been completed.

To begin to address this hypothesis, we, therefore, provide a description of spontaneous LFPs in the cockroach CX. We then monitored changes in CX LFP spectral bands during manipulation of the animals head direction and spatial context. We found that delta-band and in some cases theta-band activity significantly encoded head direction, independent of the sensory cues available to the animal. Additionally, significant changes in beta-band power were linked to changes in spatial sensory context, such as landmark removal or changes in rotation direction. These results indicate that the CX may directly participate in attention-like processes and through orientation coding it may have the capacity to provide spatial memory networks with essential navigational information.

Materials and Methods

Experimental Procedures

Experimental procedures for this paper have been described in detail elsewhere

(Varga and Ritzmann, 2016) and single-unit data from these animals has been previously published (Varga and Ritzmann, 2016). Briefly, cockroaches were implanted with tetrodes

(Guo et al., 2014) and head- and body-fixed in a restraint apparatus. The tetrodes were used to record both extracellular multi-unit activity and local field potentials. The restrained animals were placed on an Arduino-controlled rotating platform in the middle of a black cylindrical recording arena with a single solid white landmark on the wall (60° in extent).

Because of the restraint apparatus the subject’s field of view was only 180°, thus in any position throughout an ~120°-150° portion of the arena, no parts of the visual cue were visible to the animal. Data analyses were performed using custom Spike2 scripts, the

MATLAB (MathWorks, Natick, MA, USA) Circstat toolbox (Berens, 2009), Origin

(Northampton, MA, USA) and in Microsoft’s Excel (Redmond, WA, USA).

Experimental subjects

Subjects were 15 adult male cockroaches (Blaberus discoidalis) that came from our laboratory colony. The subjects were housed together in 5 gallon plastic bins in a room with a constant temperature of 27C and a 12:12-h dark-light cycle. The original group of subjects (27 adult males) provided the foundation for a previous paper. The 15 animals discussed in this paper, were included in that previous dataset (Varga and Ritzmann, 2016).

Here, we only focused on these 15 animals, because these were the individuals with single units that significantly encoded head direction.

Recording procedures

Detailed surgical procedures can be found in (Guo et al., 2014) and (Varga and

Ritzmann, 2016). We used a reference electrode bundle (7/46 SPSN Litz wire, MWS Wire

Industries, CA, USA) and a tetrode composed of four copper-plated 12µm NiCr wires

(Kanthal RO-800, Sandvik, Hallstahammar, Sweden) for every recording. The reference electrode was lowered behind the brain, while the tetrode was lowered into the CX using a micromanipulator. The tetrode was fixed in the brain at the location that had the best extracellular signal with high signal:noise ratio. The reference electrode and the tetrode were fixed in the head capsule with light curable clear glue (Loctite 3555 transparent Light

Cure Adhesive). We also fixed the recording and reference electrodes to the neck supporting plastic collar.

After tetrode implantation, the subjects were placed in a custom 3D-printed polylactic acid (PLA) restraint apparatus that was marked with green and red tape for body axis tracking. The restraint apparatus consisted of a 27mm i.d. PLA tube measuring 75mm in length. Since, we fixed the subject’s head and neck with a plastic collar, the body axis values are consistently referred to as head angle values. A 5mm layer of sponge served to minimize leg movements in the restraint apparatus. After recovery, the restraint tube was placed onto an Arduino-controlled (Arduino Uno by Arduino Italy/USA) gear platform

93 with the animal’s head located exactly in the midpoint of the platform. The restraint apparatus was fixed on the platform with a piece of magnet, which also served as a control for excluding any magnetic fields. The platform was placed in the middle of the uniformly painted, matte black cylindrical recording arena (d=40cm; height=30cm) with a single visual landmark (white paper, width=21.5cm; height=28cm; angular extent=60) fixed on the arena wall. The contrast ratio of the background and landmark was 22.1 (background luminance: 0.07 candela/m2; landmark luminance: 1.55 candela/m2). A white curtain was used to block out conflicting visual cues from above and around the arena. A custom- written Arduino motor stepper script controlled the platform rotations through a DC motor.

All physiological recordings were acquired with a Neuralynx Cheetah system at 30 kHz (Bozeman, MT, USA). Band-pass filtered activity (0.1-325 Hz) was collected and used as the LFP signal. The subject’s movements and head direction were recorded with synchronized video-tracking (30Hz; JAI CV-S3200 camera). Most control experiments (11 out of 12) and all head-covered experiments (n=3) started with a recording of a 5min period of spontaneous activity wherein the animal was stationary in the arena. Following this period, the subject was exposed to 4 experimental paradigms. In each paradigm the cockroach was rotated around the arena 4-6x (4-6 trials/paradigm). Each trial consisted of

12x30° rotations (2 sec; rotation speed 15°/sec) followed by a 10 sec immobile period when the subject was able to sample its environment. Only the middle 8sec of these stationary periods were used for analyses. The duration of each recording was ~1-1.5 hours. At the end of the experiment we marked the recording location by lesioning the brain and depositing copper from the electrodes (see Appendix Figure S2 for recording sites; (Guo et al., 2014)).

In this paper we focused on 15 subjects that had single units displaying significant head direction-modulated firing rates during at least one experimental condition (Varga and

Ritzmann, 2016). In more detail, subjects in Cohort 1 (12 animals) were exposed to 4 conditions – ‘landmark in control position + CW rotations’; ‘landmark in control position

+ CCW rotations’; ‘no landmark + CW rotations’ and ‘no landmark + CCW rotations’.

Subjects in Cohort 2 (3 animals) were exposed to 4 conditions – ‘head-covered + CW rotations’; ‘head-covered + CCW rotations’; ‘landmark in control position + CW rotations’; and ‘landmark in control position + CCW rotations’ respectively. Rotation direction (CW vs. CCW) was randomized. During head-covered trials, the animal’s entire head was carefully covered with foil to block all visual cues (both the ocelli and compound eyes were covered) and most antenna movements.

Data analysis and statistics

Extraction and analysis of the LFP signal were performed off-line in Spike2 v7.15

(CED, Cambridge UK). The raw LFP data was filtered (0.1-200 Hz) using a second-order band pass Butterworth filter. Next, line noise was removed from the signal with a second- order band-stop Butterworth filter (59-62 Hz). These filtered data were used in all further analysis. Throughout all of our analyses LFP spectral bands were defined as delta (0.1-4

Hz), theta (4-10 Hz), beta (10-30 Hz) and gamma (30-100 Hz) (Buzsáki and Draguhn,

2004; Buzsáki et al., 2013; Paulk et al., 2013).

For the analysis of spontaneous LFP activity, an ~100sec time epoch of Fast Fourier transform (FFT) power spectrum (0.54sec Hanning window, 1.85 Hz resolution) was taken from each cockroach from the ~ 5min immobile period the experiments started with (n =

11 subjects; 4 outliers were excluded from this analysis due to the FFT power being too large). The averages of the power spectra for each animal were normalized to the minimum/maximum and averaged across all animals to obtain a single average power spectrum ± SD for the spontaneous stationary state. These data were then used to quantify the average normalized power in the delta-, theta-, beta-, and gamma-bands.

To compare the power modulation by head direction in each experimental condition

(control CW, control CCW, no landmark CW, no landmark CCW; head-covered CW, head-covered CCW) the data had to first be normalized. Tthe average power spectra for each 30° angle bin (12 bins) for each animal were normalized to the minimum/maximum of the individual’s ~100sec spontaneous power spectrum. FFT power was then averaged within the above defined LFP spectral bands. The average normalized band power per angle bin was used to determine the degree of angle modulation with the Rayleigh test

(significance = p<0.05; MATLAB (MathWorks, Natick, MA) using the Circstat toolbox

(Berens, 2009)). We also used the Rayleigh test results to determine the characteristics of angle-modulation in the four experimental conditions where different sensory cues were available to the animals ((control CW, control CCW, no landmark CW, no landmark CCW; or head-covered CW, head-covered CCW, control CW and control CCW). By normalizing the values to the maximum and aligning the responses to the same maximum value (at bin

6, marked Max. resp. in Figures 4 and 5), we demonstrated the relative response magnitudes within each spectral band for every experimental condition. Last, we looked at

96 changes in the average (not normalized) power spectra across the four sensory experimental conditions for each subject. We tested for statistical differences between the four condition’s overall power (0.1-100 Hz), and separately for delta-, theta-, beta-, and gamma-band FFT power averaged across angles with an ANOVA followed by Fisher’s

PLSD.

Results

In the present study we recorded the spontaneous and head direction modulated

LFP activity, to our knowledge for the first time, specifically in the CX of insects. The cockroaches were restrained and recordings were performed using an implanted tetrode wire bundle (Guo et al., 2014; Varga and Ritzmann, 2016).

Description of spontaneous LFPs in the cockroach central complex

We began our investigation by characterizing the spontaneous LFPs from 15 adult male cockroaches that were head-fixed and slightly immobilized in a restraining apparatus placed in a dark cylindrical arena. All subjects were restrained in the recording arena for

5min to allow them to acclimate to this novel environment before exposing them to the head direction paradigms. We analyzed ~100sec of these 5 min spontaneous periods.

Figure 3.1A shows an example trace of spontaneous, full-band LFP activity in the CX, as well as filtered activity in the delta-, theta-, beta-, and gamma-bands. In this trace, peaks

(i.e., oscillations) in delta-band power occur intermittently (Figure 3.1A, see star indicators). In contrast, theta-, beta-, and gamma-band powers are largely stable throughout the recording with only minor variations over time. The corresponding example FFT power spectrum in Figure 3.1B shows the distribution of energy within this same epoch from

Figure 3.1A. We found CX LFP activity to consist of coexisting slow and fast rhythms.

The CX LFP is qualitatively dominated by slow-wave and delta oscillations, with activity between 0.1-4 Hz, which we will refer to as delta-band activity (Buzsáki et al., 2013;

Chauvette et al., 2011). Theta-band activity (4-10 Hz) was visibly intermittent and its

98 power was more prominent than beta- (10-30 Hz) and gamma-band (30-100 Hz) activity

(Buzsáki and Draguhn, 2004; Buzsáki et al., 2013; Paulk et al., 2013).

Figure 3.1: Example spontaneous LFPs in the cockroach CX. (A) An example trace of spontaneous, full- band LFP activity, and filtered activity in the delta, theta, beta, and gamma spectral bands. Peaks in delta- band power (viz., oscillations) occurred intermittently as indicated by stars. (B) The corresponding example FFT power spectrum showing the quantified distribution of spectral energy within this same epoch from (A).

Theta-band activity (4-10 Hz) is more prominent in power than beta- (10-30 Hz) and gamma-band (30-100 Hz) activity.

To quantify the spontaneous LFPs in the CX, we used the FFT power analysis across data from 11 subjects (Figure 3.2A and B, 4 subjects were excluded due to their FFT power being larger than 50% of the overall standard deviation in the entire dataset). The average normalized FFT power spectrum in Figure 3.2A indicates seemingly increased variability across subjects in lower frequency band power (~0.1- 10 Hz) and in the beta- band, especially between ~20-30 Hz. Delta-band activity was significantly greater than theta- [F(1,20) = 22.041, p = 0.000139], beta- [F(1,20) = 25.125, p = 0.00006] or gamma- band activity [F(1,20) = 26.779, p = 0.00004] (Figure 2A). Theta-band activity was statistically similar to beta- [F(1,20) = 2.130, p = 0.1599], and significantly more dominant than gamma-band power [F(1,20) = 5.986, p = 0.02378]. Meanwhile, gamma-band power was significantly less prominent than beta-band activity [F(1,20) = 8.557, p = 0.00837].

Thus, spontaneous LFP in the CX is dominated by lower frequency activity, including in spectral bands considered important for navigational coding in other animals (Buzsáki and

Moser, 2013; Ulanovsky and Moss, 2007).

100

Figure 3.2: Average spontaneous LFP in the CX is dominated by delta-band activity. (A) Normalized mean FFT power ± SD for 11 subjects. (B) Quantification of the normalized mean FFT power ± SD showed in (A), across spectral bands. Delta-band activity was significantly more prominent than activity in any other spectral band. * p<0.05; ** p< 0.001; *** p<0.0001; two-tailed t-test.

Head direction modulation of central complex network activity

We next asked whether changes in CX LFPs have the capacity to encode the animal’s head direction. Specifically, does activity in any of the recorded spectral bands correspond with the animal’s head direction? To test this, we analyzed the band-filtered

LFP of animals that were rotated around in a dark environment in 30° increments. All data

101 analyses were restricted to the 10 second stationary periods between the rotational steps.

We tested the uniformity of average spectral band power throughout all rotations in each condition with the Rayleigh test. Significant Rayleigh test results (p < 0.05) demonstrate a divergence from circular uniformity in power. Such circular divergence would indicate that power in that specific frequency band increased significantly when the animal was facing a specific, preferred angle, while the band power significantly decreased when the animal was facing a non-preferred direction. Example histograms showing head direction modulation in the delta-and theta-bands are illustrated in Figure 3.3 A and B. In contrast,

Figure 3.3 C shows the results of a different recording where gamma-band activity was not modulated by the animal’s head direction.

We found that head direction modulated CX LFP activity and that this may occur in delta-, theta-, beta-, and gamma-range powers (Figure 3.3D). We organized these data further based upon the recording location (FB or EB). LFP power modulation by head direction in all bands but beta was independent of the location of the recording within the

CX, with both FB and EB recordings displaying significant changes in the delta-, theta- and gamma-bands (Figure 3.3D). Significant beta-band modulation by head direction only occurred in the EB (Figure 3.3D). These changes occurred in all experimental conditions, including the control condition, no landmark condition, head-covered condition and control after head-covered condition (Figure 3.3E and F). These data show that head direction similarly influences LFP activity in both the FB and EB, with the exception of the beta- band.

102

A B C

0.35 0.03

SD 0.035 ±

0.3 SD SD ± 0.025 ± 0.03 0.25 0.02 0.025

0.2 band band power - 0.02

band band power 0.015 -

0.15 band power - 0.015 0.1 0.01 0.01 0.05 0.005

Mean Mean FFT delta 0.005

0 Mean FFT theta 0

0 Mean Mean FFT gamma

Angles Angles Angles

D 20 EB 15 FB

LFP band power power band LFP

# of experiments where where experiments of # encoded head direction head encoded 0 DELTA THETA BETA GAMMA

E F

100% 100% Head-covered

Control 83% experiments experiments 80% 80% 71% No landmark Landmark exposed experiments after head-covered 60% experiments 60% 50%

40% 40% 33% 33% 25% 21% 21% 17% 20% 20% 8% 8% 4% 4% 0% 0% 0% 0% power encoded power head direction 0%

% of % experiments LFP where band DELTA THETA BETA GAMMA DELTA THETA BETA GAMMA

Figure 3.3: Delta-band activity reflects head direction more than any other frequency band. (A) and (B) Example histogram showing significant head direction modulation in the delta-band (A) and the theta- band (B). p< 0.05, Rayleigh test. Both examples came from the same experiment. Teal bar marks the significant preferred angle encoded by delta and theta, which occurred in the same 0-30° angle bin. (C) 103

Example histogram showing no head direction modulation in the gamma-band (from a different experiment; p> 0.05, Rayleigh test). (D) # of experiments where LFP band power was significantly modulated by head direction (p< 0.05; Rayleigh test) in the EB and the FB. LFP power modulation was independent of the recording site in the delta-, theta- and gamma-bands. (E) and (F) % of experiments where LFP band power was significantly modulated by head direction (p<0.05; Rayleigh test) during control and no landmark experiments (D) and during head-covered and landmark exposed after head-covered experiments (E). Delta- band modulation significantly decreases during no landmark experiments (E) and landmark exposed after head-covered experiments, indicating that the first exposure to an environment determines delta-band modulation by head direction rather than the presence/absence of the landmark.

Delta-band activity encodes head direction independent of the underlying sensory cues

During control experiments, modulation of delta-band power was the most prominent, with delta-band activity significantly encoding an angle in 71% of the experiments (Figure 3.3E, light gray). Head direction driven changes in the theta-band were the second most prominent, with theta encoding an angle in 25% of all control experiments (Figure 3.3E, light gray). Delta-band modulation by head direction was significantly more common than theta-band modulation (χ2 (1) = 14.295; p=0.0002; two- tailed with Yates correction). Activity in the beta-band was significantly modulated during only 4% of the control experiments, which is significantly less than the amount of delta- or theta-band modulation (compared to delta: χ2 (1) = 44.397; p<0.0001; compared to theta:

χ2 (1) = 11.973; p=0.0005; two-tailed with Yates correction). Meanwhile, gamma-band power significantly encoded an angle during 8% of the same experiments (Figure 3.3E, light gray). The amount of gamma-band modulation by angle was similar to the amount of beta-band modulation (χ2 (1) = 0.680; p=0.4097; two-tailed with Yates correction), but significantly lower than delta- or theta-band modulation (compared to delta: χ2 (1) =

36.288; p<0.0001; compared to theta: χ2 (1) = 6.557; p=0.0104; two-tailed with Yates correction).

104

Following the control condition, the landmark was removed from the environment, thus the animals had to rely on a memory of this reference point or update the head direction system using motion cues (Figure 3.3E, dark gray). In the ‘no landmark’ condition, changes in the delta-band significantly encoded a preferred angle in 21% of the experiments, which was a significant reduction compared to the control condition (χ2 (1) =

19.075; p<0.0001; two-tailed with Yates correction; Figure 3.3E, dark gray). Results in the theta-, beta- and gamma-bands were similar to the control, despite the lack of reference point for the cells to lock onto conditions (theta: χ2 (1) = 0.938; p=0.3327; beta: χ2 (1) =

0.062; p=0.8033; gamma: χ2 (1) = 0.326; p=0.5680; all two-tailed with Yates correction

Figure 3.3E, dark gray). These results suggest that delta-modulation by head direction may benefit from having a landmark as a reference point for orientation coding.

Experiments conducted on head-covered, landmark naïve animals resulted in effects similar to those described in the previous section (Figure 3.3F, light gray). Changes in the delta-band significantly encoded an angle in 83% of the experiments. Theta-band activity significantly encoded a preferred angle in 50% of the experiments, which is significantly less compared to the amount of delta-band modulation (χ2 (1) = 4.478; p=0.0343; two-tailed with Yates correction). Activity in the beta-band was significantly modulated in 17% of the head-covered experiments, which is significantly less than the amount of delta- or theta-band modulation (compared to delta: χ2 (1) = 29.146; p<0.0001; compared to theta: χ2 (1) = 11.384; p=0.0007; two-tailed with Yates correction).

Meanwhile, gamma-band power did not significantly encode an angle in any one of the experiments (Figure 3.3F).

105

Compared to these results, when the head-cover was removed and landmark was exposed to the animals, only activity in the theta- and beta-bands significantly encoded a preferred angle, both in 33% of the experiments which was similar to the percent of angle- modulated experiments in the head-covered condition (theta: χ2 (1) = 2.076; p=0.1497; beta: χ2 (1) = 3.495; p=0.0615; both two-tailed with Yates correction; Figure 3.3F, dark gray). Thus, delta-band modulation by head direction significantly dropped when the landmark was revealed to the animals (χ2 (1) = 62.008; p<0.0001; two-tailed with Yates correction). These results, combined with the results of control experiments followed by landmark removed experiments, indicate that landmark presence or absence is not a determinant of the angle modulation of delta-band power.

Delta-band activity is significantly more likely to encode the animal’s head direction than any other frequency band ([F(1,7) = 17.494, p = 0.00918] one-way ANOVA followed by Fisher’s PLSD). The significant modulation of delta-band activity during both the control condition and head-covered experiments strongly suggests that the head direction signal is encoded by LFPs independent of the sensory sources that the animal uses as a navigational frame of reference.

Sensory context does not affect relative response magnitudes in the delta-band

Our previous results indicated that LFPs, especially in the delta-band, are modulated by head direction independent of the sensory reference frame used to establish the heading signal. To investigate how different spatial sensory contexts might affect LFP response magnitudes (thus, the relative increase in power at the preferred angle), we

106 adjusted the orientation of the data relative to 12 angle bins around the arena by setting the maximum response at bin 6 regardless of the angle bin where the actual maximum response occurred. We then normalized the average power in each band across all experiments to the maximum (independent of Rayleigh test results) and plotted the data separately for control CW and CCW and no landmark CW and CCW experiments (Figure 3.4A). We found that the relative magnitudes of change in delta-band power are statistically similar for all conditions ([F(3,47) = 0.9436, p = 0.427] one-way ANOVA followed by Fisher’s

PLSD). Thus, the relative increase in delta-band power at the preferred angle is independent of the landmark’s presence or absence (control vs. no cue conditions: p=0.592; paired two-tailed t-test) and also independent of rotation direction history (CW vs. CCW; p=0.339; paired two-tailed t-test). Nevertheless, these data are not in disagreement with the previous results that showed a significant decrease in the amount of experiments with delta- band head direction modulation after landmark removal.

Similarly, the relative response magnitudes in the theta- and beta-bands did not reflect any changes in the sensory environment (theta: ([F(3,47) = 0.0411, p = 0.9887]; p=0.770; p=0.636; beta: ([F(3,47) = 0.3245, p = 0.8075]; p=0.085; p=0.217; ANOVA followed by Fisher’s PLSD; CW vs. CCW comparison with two-tailed t-test and control vs. no cue comparison with paired two-tailed t-test respectively; Figure 3.4A). On the other hand, gamma-band activity was significantly different during CW and CCW trials

([F(3,47) = 2.2988, p = 0.0905]; p=0.002; p=0.5521; ANOVA followed by Fisher’s PLSD;

CW vs. CCW comparison with two-tailed t-test and control vs. no cue comparison with paired two-tailed t-test respectively; Figure 3.4A) . Thus, with the exception of the gamma- band, the relative amount of increase in power at the preferred angle is the same no matter

107 how the environmental context changes, at least in animals that were first exposed to an environment with a conspicuous visual cue in it.

The previously observed lack of environmental effects may have been the result of the inclusion of experiments where band-power in some frequency bands were not modulated by head angle. Therefore, we next examined the changes in relative response magnitude separately for the experiments with significant Rayleigh test results. Based on the fact that delta-band activity encoded head direction in significantly more experiments than any other band and because theta-band modulation was more prominent than beta or gamma (see Figure 3.3), we only analyzed data for these two frequency bands. Figure 3.4B shows that there is no overall statistical difference between the control conditions and no landmark conditions during experiments where delta-band power significantly encoded head direction (p=0.305; two-tailed t-test). However, there is a difference between normalized power in the 8th bin between the two conditions, corresponding to 60° away from the preferred angle (p=0.035; two-tailed t-test). The normalized response magnitude in the theta-band was also statistically similar in both control and no landmark conditions

(p=0.240; two-tailed t-test) and in all angle bins (Figure 3.4C). Because both delta- and theta-band activity encoded head direction similarly independent of whether a landmark was available to the animal and independent of the rotational context, the above results indicate that CX LFPs in the delta- and theta-bands do not directly encode the contexts surrounding the orienting animal.

108

B C

Figure 3.4. Sensory context and landmark exposure does not affect relative LFP response magnitudes when the animals can initially rely on a landmark. (A) Normalized mean FFT power ± SEM for all of the subjects was similar in all four bands in all four experimental conditions (p< 0.05; ANOVA followed by Fisher’s PLSD). (B) Subjects that significantly encoded head direction in the delta and theta spectral bands during control experiments. Relative LFP power response magnitude in the delta- and theta-bands was similar during control CW and CCW experiments and no landmark CW and CCW experiments (p>0.05; two-tailed t-test) with the exception of the 8th bin where delta power significantly decreased after landmark removal (p=0.035; two-tailed t-test).

Our initial results on LFP modulation by head direction shown in Figure 3.3 indicate that the animal’s first experience with an environment may shape the angle-

109 modulated responses in CX LFPs. If this is true, we would expect that the relative response magnitudes should be similar in the experiments where we first exposed head-covered, landmark naïve animals to the arena and only removed the head-cover later and those that were first exposed to the control conditions with a landmark.

We found that the relative magnitudes of change in delta-band power are statistically similar for all conditions ([F(3,47) = 1.2257, p = 0.3116] one-way ANOVA followed by Fisher’s PLSD, Figure 3.5). However, Figure 3.5 also illustrates that the relative response magnitudes were different in some frequency bands depending on the sensory context. Normalized theta-band power was significantly different during all experiments ([F(3,47) = 9.2075, p < 0.0001] one-way ANOVA followed by Fisher’s

PLSD), except for head-covered CW and control CCW and control CW and control CCW, which were similar. Response magnitudes in the beta-band significantly differed in some cases ([F(3,47) = 6.0502, p = 0.0015] one-way ANOVA followed by Fisher’s PLSD), but were similar during head-covered CW and CCW experiments, head-covered CW and control CW experiments and control CW and CCW experiments. Responses in the gamma- band were statistically similar, with the exception of the last, control CW experiments that differed from every other condition ([F(3,47) = 6.7345, p < 0.001] one-way ANOVA followed by Fisher’s PLSD). While such variable results could stem from the limited number of experiments (n=3), the lack of delta-band modulation by contextual information further supports our hypothesis that delta-band activity encodes head direction independent of the sensory context that the signal is established in.

110

Figure 3.5: Sensory context and landmark exposure affects relative LFP response magnitudes during head covered and control after head-covered experiments in all bands with the exception of the delta- band. Normalized mean FFT power ± SEM for all of the subjects varied in the LFP spectral bands depending on the sensory context of the condition (n=3 animals). ** p < 0.005; *** p < 0.0001; One-way ANOVA.

Sensory context modulates the average power of central complex network activity in the theta-, beta- and gamma-bands, but not in the delta-band

It is possible that LFPs encode changes in the sensory context by increasing or decreasing the overall power, and that changes in power may be obscured when analyzing normalized responses. Such changes in overall power would not affect the angular response of the tested frequency band, thus any possible changes caused by different sensory contexts would be conveyed simultaneously with the head direction code within the same signal. Therefore, we next investigated how the average LFP band power (not normalized to the maximum) reflects changes in the sensory context by only focusing on experiments that started with the control condition and ended with the no landmark condition (n=12).

111

We compared the overall average full band power between 0.1-100 Hz, as well as power in the delta-, theta-, beta-, and gamma-bands separately in each experiment across the four sensory conditions (control CW; control CCW; no landmark CW; no landmark CCW).

Figure 3.6 illustrates an example analysis where we compared the average activity in the control CCW condition to the average activity in the no landmark CCW condition.

Our analysis indicates that the full band power (0.1-100 Hz) is statistically similar in both conditions (p = 0.07; paired two-tailed t-test). Similarly, we found no changes in delta- band power in this experiment (p = 0.144; paired two-tailed t-test). On the other hand, activity in the theta-, beta- and gamma-bands was significantly modulated by changes in the sensory context (theta: p = 0.023; beta: p = 0.010; gamma: p = 0.00009; paired two- tailed t-test).

Across all subjects we found that the overall power between 0.1-100 Hz was statistically similar throughout all conditions in every experiment (one-way ANOVAs followed by Fisher’s PLSD, p > 0.05). Activity in the delta-band alone did not reflect any of the changes in the sensory context in any one of the 12 experiments. In some cases theta-

(4/12 experiments) and gamma-band power (5/12 experiments) significantly increased/decreased in response to changes in the sensory context. Most importantly, we found that beta-band power significantly changed in response to manipulations to the sensory context in 9/12 experiments. Such changes were not only induced by the presence or absence of the visual cue, but also by different rotation direction history, and more importantly, they were independent of head direction in every experiment. These results

112 show that CX LFPs in the theta-, beta- and gamma-bands encode information about the sensory context independent of the animal’s head direction.

Figure 3.6: Modulation of average FFT power by sensory context is independent of head direction coding by CX LFPs. Example analysis comparing the average FFT power in the control CCW condition to the average FFT power in the no landmark CCW condition. The full band power (0.1-100 Hz) is statistically similar in both conditions (p = 0.07; paired two-tailed t-test). Delta-band power was significantly modulated by head direction (p<0.05; Rayleigh test). We found no sensory context dependent changes in delta-band power in this experiment (p = 0.144; paired two-tailed t-test). Theta-band power also significantly encoded head direction (p<0.05; Rayleigh test). The average power in the theta-band was significantly modulated by changes in the sensory context (p = 0.023; paired two-tailed t-test). Beta-band power was not modulated by head direction, but significantly encoded changes in the sensory context (Rayleigh test; p> 0.05 and p=0.010; paired two-tailed t-test). Activity in the gamma-band was significantly modulated by head direction (p<0.05; Rayleigh test) and changes in the sensory context as well (p = 0.00009; paired two-tailed t-test).

113

Discussion

LFPs are considered to be important mediators of insect nervous system function

(e.g. Broome et al., 2006; Ito et al., 2009; Laurent and Davidowitz, 1994; Marder et al.,

2015; Nitz et al., 2002; Paulk et al., 2015; van Swinderen and Greenspan, 2003; Wehr and

Laurent, 1996). Here we examined LFPs in the cockroach CX. We found that spontaneous

CX LFP activity consist of coexisting slow and fast rhythms (Figures 3.1 and 3.2). More importantly, we found that all LFP frequency bands are occasionally modulated by head direction and changes in the environmental context in a passive rotation paradigm, suggesting that network-level oscillations may be important for encoding navigation- relevant information. These data provide important insights into network-level aspects which may be important for adaptive navigation as discussed herein.

Table 3.1: Summary of LFP analysis results.

Average Significantly modulated by head direction Relative response magnitude power (p < 0.05; Rayleigh test) changed (p < 0.05; ANOVA) changed in different sensory Head- contexts Landmark Control to No covered to No Head- (p < 0.05; Control exposed after landmark landmark landmark covered t-test) head-covered condition exposed condition DELTA ● ● ● X X X X THETA ● ● ● ● X ● ● BETA ● ● ● ● X ● ● GAMMA ● ● X X X ● ●

● significant change occurred; X no significant difference occurred.

114

Studies examining navigational circuits in rats and bats highlight that in spite of the similarities on the cellular level, there are important differences between the underlying network activity (Hemberger et al., 2016). Rat head direction cells may be internally organized as an attractor network and also regulated by theta oscillations (Jeffery et al.,

1995; Tsodyks et al., 1996; Tsanov et al., 2011; Bjerknes et al., 2015; Peyrache et al.,

2015), while the bat navigation system does not appear to be affected by the theta-rhythm

(Ulanovsky and Moss, 2007; Yartsev et al., 2011). On the other hand, it has been suggested that head direction cells in Drosophila may be organized as a ring attractor (Seelig and

Jayaraman, 2015), but whether the network’s activity is influenced by rhythmic oscillations is unknown. Based on previous results from our laboratory and a detailed comparison of the cockroach and rat head direction systems, we hypothesized that the head direction signal in cockroaches is possibly established through similar mechanisms as those employed by rat navigation circuits (Varga and Ritzmann, 2016). Thus, a major goal of the present study was to determine whether LFPs in the cockroach CX have the capacity to encode head direction.

Our results support our main hypothesis that LFP oscillations in the insect CX encode the animal’s head orientation (Figure 3.3 and Table 3.1). We found that LFP power was modulated by head direction in both the EB and FB of the cockroach CX (Figure 3.3C).

Specifically, we observed significant changes in the delta-, theta-, beta- and gamma-bands that encoded the cockroach’s heading relative to a sensory reference frame throughout all experiments.

Compared to the described dominant theta-rhythmicity in the rat head direction network (Huxter et al., 2008; Tsanov et al., 2011; Brandon et al., 2013), we observed

115 significantly more delta-band modulation by head direction in the CX than angle coding by any other frequency band (Figure 3.3). Delta and slow-wave modulation is coupled with the rotational speed of a visual panorama in the fruit fly’s central brain (van Swinderen and

Greenspan, 2003). In contrast, we found that head direction coding by delta-band activity occurred significantly more often upon the first exposure to an environment, regardless of whether a visual landmark was available to the animals (Figures 3.3 – 3.5). It is important to note that our previous study (Varga and Ritzmann, 2016) as well as the study of head direction in Drosophila (Seelig and Jayaraman, 2015), indicated that these insects could establish head direction with non-visual cues. Delta-band modulation by head angle significantly decreased during the second condition of the experiments, even when it meant that the animal’s blind-fold was removed and a conspicuous visual landmark was revealed

(Figure 3.3D and E). This suggests that the head direction signal encoded by the delta-band was independent of the source of the sensory cues. We hypothesize that changes to the first, familiar environment possibly lead to a change in the signal’s reference frame that causes remapping, and searching for a new reference point in the head direction system.

That stated, it is possible that our experiments were too short for the delta-band signal to stabilize after the environmental manipulations and encode head direction relative to the new sensory reference frame. Future studies with longer trials and recurring environmental features may be helpful to address the mechanisms whereby delta-band activity encodes head direction.

While less prevalent than delta-band modulation effects, we also found evidence that CX LPFs in the theta-band encode head direction (Figure 3.3). This is not surprising in the context of our passive, restrained preparation, considering that theta oscillations are

116 linked to locomotion and active behaviors in all mammalian species (Robinson, 1980;

Vanderwolf, 1969; Winson, 1972). Although, one could argue that theta-band activity might have a different role in the insect brain, several studies suggest that frequency bands, with slight variations, are functionally similar in insects and mammals (e.g. Kay, 2015;

Kirszenblat and van Swinderen, 2015; Paulk et al., 2015). Thus, it is possible that similarly to rodent navigation systems, theta oscillations play an important role in organizing the activity of CX spatial cells. Specifically, theta-modulation in the CX did not change upon landmark removal and it was constant during head-covered and control experiments, suggesting that the head direction signal encoded by theta is independent of the environment’s features and may be relying on internally available sensory cues only

(Figure 3.3). Importantly, our present results provide the first evidence that head direction coding by the cockroach theta-band activity might be similar to theta-driven rhythmicity in the rat head direction system.

Additional analyses on the relative response magnitudes and the average changes in power evoked by changes in the sensory context further support the hypothesis that the activity patterns in the delta-band are not directly modulated by sensory information processes, as long as the environment remains stable enough for the network to reliably encode head direction (Figures 3.4 and 3.5, Table 3.1). Meanwhile, changes in relative response magnitude, as well as average power in the theta-, beta-, and gamma-bands indicated that CX LFPs may encode information about the spatial sensory context (Figures

3.4 – 3.6). CX single neurons encode sensory information that arrives indirectly from several primary sensory areas (Kathman et al., 2014; Milde, 1988; Phillips-Portillo, 2012;

Ritzmann et al., 2008; Ritzmann et al., 2012; Seelig and Jayaraman, 2013). These sensory

117 cues are then assumed to be used in downstream brain regions for action selection and to fine-tune motor commands (Fiore et al., 2015; Martin et al., 2015). Our results indicate that

CX LFPs also encode sensory cues, and in addition to action selection and maintenance, they may also play a role in navigation.

The basis of adaptive navigation is the brain’s ability to encode spatial information, and to compare the current navigational context to a past or future context. Spatial context discrimination requires the transformation of sensory information and its implementation in associative functions, such as attention-like processes and spatial memory. Evidence for the central brain’s involvement, and therefore, the CX’s possible role in associative processes that potentially underlie navigation has come from LFP recordings in the fruit fly (Nitz et al., 2002; van Swinderen and Greenspan, 2003; van Swinderen et al., 2009).

High frequency beta-band oscillations (~20-30 Hz) in the central protocerebrum of

Drosophila have been suggested to play a role in signaling sensory salience and participate in attention-like processes (van Swinderen and Greenspan, 2003). Genetic studies linked the evoked activity in the beta-band to the expression of short-term and long-term memory genes (van Swinderen, 2007; van Swinderen et al., 2009; van Swinderen and Brembs,

2010). Here, we found that CX LFP activity, especially in the beta-band, significantly encodes changes in the sensory environment (Figure 6). These data together, combined with the findings of the van Swinderen laboratory, indicate that the changes in CX LFPs could serve as a neural substrate for attention-like processes during spatial context discrimination (van Swinderen and Greenspan, 2003). These data might also indicate that the CX participates in spatial memory in parallel or in conjunction with the mushroom bodies (van Swinderen, 2007; van Swinderen and Brembs, 2010; van Swinderen et al.,

118

2009). Future studies examining LFP spectral coherence between the two structures will be necessary to elucidate the exact relationship between the two structures. In mammals, spatial memory and possibly navigational context discrimination, is linked to the theta- rhythmic segregation of hippocampal place cell firing (Buzsáki and Moser, 2013;

Mizumori et al., 2009; Penner and Mizumori, 2012). Whether the observed beta- and/or theta-band modulation might serve similar functions in the insect brain will need to be resolved perhaps by employing LFP recordings in freely walking insect preparations.

In summary, our results indicate that oscillations in the CX may participate in adaptive navigation by encoding the animal’s head direction and signaling spatially relevant contextual changes in the environment. We predict that analysis of oscillations underlying insect navigation will contribute to a broad comparative approach to understanding the general principles of navigation. At the same time, comparative studies will also help us uncover the unique network components and neural mechanisms that different species may employ to successfully navigate. Future studies building upon these findings will be crucial in uncovering the neural processes governing navigation and spatial memory in insects and may also provide the mammalian navigation field with new insights.

119

Chapter 4

Conclusion

120

Navigation is a complex feat for an animal to accomplish. In contrast to the wealth of information on the behaviors displayed by insects during navigation and patterns of navigation across vast terrains, very little is known regarding how the brain orchestrates navigation (Fiore et al., 2015; Heinze, 2015; Ritzmann et al., 2012; Webb and Wystrach,

2016). Therefore, the overall goal of this thesis was to extend our knowledge on the neural dynamics which serve navigation, specifically the mechanisms underlying two important foundations of navigation – head direction coding and context discrimination. There has been an especially major void in our understanding of brain mechanisms of navigation in insects. To begin to address this, throughout this research I utilized the cockroach as a model system and as detailed below, the results of this model provide new general insights into the brain basis of navigation. Moreover, this research has contributed significant advances in our specific understanding of how the arthropod brain accomplishes navigation.

Considerations of the neural basis of navigation

For successful navigation to occur, the nervous system must continuously update behavioral responses based upon both incoming sensory information and previously acquired sources of associative information. As an example, consider the desert ant navigating its native sandy terrain. The smooth sandy floor of the desert is mostly void of visual landmarks and also lacking of chemical and tactile cues to guide the animal, yet the ant somehow manages to navigate across large distances. To accomplish navigation, the desert ant keeps track of its movement and orientation through continuous updates from idiothetic sensory systems (e.g., Müller and Wehner, 1988; Collett et al., 1998; Wehner,

121

2003; Srinivasan, 2015). The use of polarized light to accomplish navigation is also well established in several insect models and these studies provide us with more information about the brain systems which are involved in navigation (e.g., Heinze and Homberg, 2007;

Heinze et al., 2009; Heinze and Reppert, 2011; Homberg et al., 2011; Merlin et al., 2014;

Pfeiffer and Homberg, 2014; el Jundi et al., 2014; el Jundi et al., 2014; Bockhorst and

Homberg, 2015; el Jundi et al., 2015). For instance, neurons in the central complex (CX) of the Monarch butterfly encode the directional cues embedded in the polarized skylight and may adjust the butterfly’s heading accordingly (Heinze and Reppert, 2011; Reppert et al., 2004; Reppert et al., 2016). These strategies are just a few examples of how the brain may facilitate adaptive navigation in these animals.

In contrast to insects, far more is known about the brain systems of navigation in mammals, especially in rodents. The rodent brain encompasses several navigation circuits that contain a variety specialized cell types which keep track of the direction and speed of the animal’s movements, as well as encode relative position and orientation. For instance, the speed of movements is encoded by speed cells (Kropff et al., 2015), while the animals distance and direction are continuously updated through grid cells (Hafting et al., 2005;

Rowland et al., 2016). Head direction cells are responsible for encoding the animal’s orientation relative to a starting point or certain landmarks (Taube, 2007; Taube et al.,

1990a; Taube et al., 1990b). Meanwhile, place cells represent abstract information about the animal’s position and inner state in its proximal environment while also encoding this information in spatial memory (McNaughton et al., 2006; O’Keefe and Dostrovsky, 1971).

The exact relationship between these cell types is not completely known, however data suggest that head direction cells provide a crucial input to grid cells (Winter et al., 2015a),

122 which have reciprocal connections with place cells. A diagram demonstrating the anatomical connections between mammalian navigation circuits and the types of spatial cells that can be found within these brain regions is illustrated in Figure 1.1. For the most part it is entirely unknown if these types, or similar systems are operating in the insect brain.

Among other insects, cockroaches provide an excellent model for studying the neural dynamics of navigation. First, cockroaches have relatively large brains allowing for the utilization of high throughput extracellular recordings. Also of importance, cockroaches are ecologically similar to rats given that they are nocturnal and frequently navigate subterranean environments. Thus, comparisons can be readily made with the more thoroughly studied rat model. In this thesis, I exploited these similarities and advantages and adopted the methods of rat head direction studies to elucidate the neural mechanisms governing head direction coding in the cockroach CX.

Broad discussion of major findings and their significance to the field of neurobiology

While casually viewed simply as a sensory and motor behavior, navigation is ultimately a cognitive process. Although some may argue lower animals, including insects are not ‘cognitive’ creatures (e.g., James, 1890; Wehner and Menzel, 1990; Cruse and

Wehner, 2011), recent breakthroughs in insect neuroscience suggest that insects possess some of the necessary building blocks of cognition (Giurfa, 2013; Giurfa et al., 2001;

Haberkern and Jayaraman, 2016; van Swinderen and Greenspan, 2003; Webb, 2012).

Cognition may be broadly defined as the sum of higher order processes, such as attention,

123 working memory, and internal model-based decision making (Haberkern and Jayaraman,

2016; Menzel and Giurfa, 2001). The insect brain’s potential for supervising such complex computations has been indicated by a range of behavioral and physiological studies (Collett and Land, 1975; Esch et al., 2001; Gould, 1986; Krashes et al., 2009; Liu et al., 1999;

Mischiati et al., 2015; Neuser et al., 2008; Ofstad et al., 2011; Paulk et al., 2014; Reinhard et al., 2004; Tang and Juusola, 2010; van Swinderen and Greenspan, 2003; Von Frisch,

1967; Wiederman and O’Carroll, 2013). The results of these studies point in the direction of the insect central brain’s involvement in attention-like processes, context dependent action selection and maintenance, and the utilization of working memory during guided behaviors. Directly relevant to this thesis, these studies thematically also hint at a role for the insect brain in orchestrating adaptive navigation and spatial memory, which together synthesize all the previous processes listed. Although, the results of this thesis do not directly support the cognitive functioning of insect neural circuits, they provide a foundation for future studies attempting to elucidate the higher order neural functions that may underlie cognitive processes.

Single neurons and LFPs encode head direction in the insect CX

My extracellular tetrode recordings in a restrained, passively rotated preparation revealed that single units in the cockroach CX have the capacity to encode the animal’s head direction (Varga and Ritzmann, 2016). The observed response characteristics of CX cells were similar to the described rat head direction cells (Taube, 2007). For instance, some of the CX neurons encoded a preferred orientation with narrow tuning schemes that closely resembled rat head direction tuning curves. Additionally, I found through careful

124 manipulations of the proximal environment that, just like head direction cells in the rat brain, CX neurons establish their preferred angles by primarily relying upon salient visual landmarks, but can update the signal based on internally available motion cues when such visual information is not available.

These data corroborated and extended the results of the only other head direction study in the insect literature (Seelig and Jayaraman, 2015). Seelig and Jayaraman conducted experiments in head-fixed fruit flies that were able to walk on an air-suspended ball in a closed-loop paradigm. They also found that flies primarily rely upon visual cues to encode head direction. However, in darkness, the fly head direction system quickly accumulated directional error and the encoded orientation started to shift. This did not occur in the cockroach experiments. There could be at least two possible explanations for this major difference between the Drosophila study and my findings. First, Drosophila and cockroaches are ecologically and behaviorally distinct species. Specifically, it is likely that since the diurnal fruit flies mostly navigate during the day, their head direction system is not as reliable in the dark as the nocturnal cockroach’s system. Differences in CX control of navigation have been noted in nocturnal and diurnal species of dung beetles (el Jundi et al., 2014b; el Jundi et al., 2015). Identifying whether this difference contributed to the distinctions between these two studies would require recordings from Drosophila species which are nocturnal and flightless. Alternatively, the difference between these studies could stem from the experimental design. The fruit flies received continuous feedback from the legs (both proprioceptive and motor feedback) since they were actively walking on the air-suspended ball. However, since their heads were fixed, they could not integrate motion- related head angular velocity signals. On the other hand, due to the passive rotation design

125 of my experiment, the only idiothetic motion cue available to the cockroaches had to be derived from angular velocity inputs mostly likely arriving to head sensors. These data are in accordance with the findings of rat studies, wherein head direction cells primarily rely upon visual landmarks, but can update the system using idiothetic cues when a familiar landmark disappears or becomes unreliable (Taube, 2007). Also, similarly to insects, rat spatial cells accumulate error when the animal only receives feedback from the limbs, but work properly when the only input indicating movement arrives through the vestibular system (Taube, 2007). Indeed, it has been shown that lesioning the inner ear completely abolishes the head direction signal (Muir et al., 2009; Stackman and Taube, 1997;

Stackman et al., 2002; Valerio and Taube, 2016).

Neural activity in the mammalian navigation circuits, including the head direction system, is subject to interactions with LFP oscillations (e.g., Tsodyks et al., 1996; Buzsáki,

2002; Brandon et al., 2011, 2013; Tsanov et al., 2011; Buzsáki and Moser, 2013). The use of extracellular recordings in the passive rotation experiments described above allowed me to monitor the LFPs that may underlie head direction coding and context discrimination in the insect CX. In contrast with the dominant theta-modulation in rodent head direction circuits (Huxter et al., 2008; Tsanov et al., 2011; Brandon et al., 2013), my data revealed significantly more delta-band modulation by head direction in the CX, than orientation coding by any other frequency band. Van Swinderen and Greenspan (2003) found that delta-modulation in the central brain of fruit flies is coupled with the rotational speed of a visual stimulus projected onto a circular LED wall. In contrast, my results indicate that delta-band modulation is independent of the visual panorama of the animal, because activity in the delta-band also significantly encoded head direction in blind-folded animals.

126

Environmental manipulations led to a decrease in the amount of delta-band modulation by orientation, which might suggest that delta-band activity relies on a sensory reference frame to establish a preferred angle, and environmental changes lead to remapping in the system. While there are numerous methodological differences between van Swinderen and

Greenspan (2003) and my recent work (Chapter 3), it is interesting to highlight that both studies uncovered evidence that LFPs may be distinctly modulated by cognitive factors.

While less prevalent than delta-band modulation effects, I also found evidence that

CX LPFs in the theta-band encode head direction. The amount of theta-band modulation by head angle is not surprising, since theta-rhythmicity is linked to active behaviors in all mammals explored to date (Robinson, 1980; Vanderwolf, 1969; Winson, 1972), and my experiments were conducted with restrained, passively rotated animals. The results of these experiments indicate that similarly to rodent head direction systems (Brandon et al.,

2013; Hartley et al., 2014; Huxter et al., 2008), activity in the theta-band may also be important in the CX navigation circuits. Although, theta oscillations are well known to participate in head direction coding in rodents, their role in synchronizing the activity of place cells and grid cells, and encoding spatial memory, has been studied more extensively

(e.g. Brandon et al., 2011; Brandon et al., 2013; Buzsáki, 2002; Buzsáki and Moser, 2013;

DeCoteau et al., 2007; Dragoi and Buzsáki, 2006; Ekstrom et al., 2005; Hasselmo, 2005;

Jeewajee et al., 2008; Koenig et al., 2011; Maurer et al., 2006; Ray et al., 2014; Tsanov et al., 2011; Tsodyks et al., 1996).The presence of theta oscillations in the insect brain might indicate that the CX is responsible for more navigational functions than orientation coding alone.

127

The results of this thesis together support the hypothesis that the CX participates in encoding head direction and suggest that the head direction systems of possibly all animals, from insects to humans, rely upon the same general coding principles. Since the head direction signal is established and continuously updated in reference with a salient sensory reference point, I hypothesize that the mechanisms underlying orientation coding and the utilization of the head direction signal requires the interplay between attention-like processes and working memory. Future studies utilizing electrophysiological techniques in freely behaving animals performing spatial and cognitive behavioral tasks may help us get closer to understanding the neural computations supporting cognitive processes in the insect brain.

Single neurons and LFPs encode spatial context cues in the insect CX

In addition to encoding the orientation and position of the animal, successful navigation also requires memories and updates about context-dependent cues. To test whether single neurons in the CX encode such spatial information, the cockroaches were rotated around in clockwise and counterclockwise directions during the passive rotation experiments. I found that similarly to “path cells”, “complex-spike cells”, and other context-sensitive cells in mammalian navigation circuits (Chadwick et al., 2015; Ekstrom et al., 2003; Jacobs et al., 2010; McNaughton et al., 1983; Mizumori et al., 2009), single neurons in the CX also encoded the directional history of past rotations, independent of the head direction signal (Varga and Ritzmann, 2016). The exact role of these spatial context cells in the mammalian navigation circuits is not known, however it is suggested that they may underlie context discrimination processes (Mizumori et al., 2009; Penner and

128

Mizumori, 2012), which in turn contribute to spatial memory and adaptive navigation.

Similarly, as discussed in Chapter 2, I predict that the encoding of past directional movement by CX single neurons is essential for navigation (Varga and Ritzmann, 2016).

LFP recordings revealed that CX network activity reflects contextual changes as well. We found that CX LFP activity in the theta-, beta- and gamma-bands significantly encodes changes in the sensory environment, such as landmark removal or rotation direction history. Similarly, results from LFP recordings in the fruit fly’s central brain has led to the prediction that changes in the beta-band power signal sensory salience and participate in attention-like processes (van Swinderen, 2007; van Swinderen and Brembs,

2010; van Swinderen and Greenspan, 2003). Additionally, these changes in the beta-band were directly linked to the expression of memory-genes in the CX and mushroom bodies of the Drosophila brain (van Swinderen, 2007; van Swinderen and Brembs, 2010; van

Swinderen et al., 2009).

These results support the hypothesis that the CX participates in navigational processes by encoding information about spatial context. Such contextual cues are not only important for navigation circuits, but also provide crucial information about the sensory environment during virtually all adaptive behaviors and may lead to context-dependent action selection and maintenance. This process also requires the utilization of memory pools, including working memory, and attention. I predict the CX is involved in processing of all of these factors, perhaps with the assistance of the mushroom bodies, and that these processes are essential to numerous adaptive behaviors displayed by the cockroach as well as other insects.

129

Conclusions and Future Directions

The results of this thesis provide significant advances in our understanding of how the insect CX contributes to navigation. I described the neural mechanisms whereby the animal’s orientation and some components of the spatial context are encoded in the insect brain. These results provide a foundation for future studies examining insect navigation.

The similarities I uncovered between the insect and mammalian head direction circuits raise the exciting possibility that other types of spatial cells are also present in the insect brain. In rodents, grid cells rely upon crucial input from the head direction network to function properly (Winter et al., 2015a). In line with this, my uncovering of head direction cells in the insect CX lead me to hypothesize there are grid cells in the insect brain. Speed cells and place cells are also hypothesized to have a direct connection with the rat head direction system (Hartley et al., 2014; Taube, 2007). Is there a demand for such specialized spatial cells in insect navigation circuits or are there simpler solutions to encoding and utilizing such spatial information? Certainly CX cells have been found in which activity is correlated with stepping speed (Bender et al., 2010; Martin et al., 2015). Could the described spatial context cells function as place cells by not only encoding current context, but also providing information about past and future contexts? Future studies on freely navigating insects will likely bring us closer to the answers of these questions (Martin et al., 2015). Further, incorporating cell-specific manipulations in combination with physiology in navigating insects will allow causal relationships to be established between

CX neural activity and navigation.

130

Appendix

This appendix includes additional data as a supplement to Chapter 2.

Appendix Figure S1. Related to Figure 2.1. Outline of the experimental design showing the order and number of conditions the animals were exposed to. As described in the results, the order of the rotation directions (CW, red arrow-clockwise rotation; CCW, blue arrow – counterclockwise rotation) were varied in some cases. The white arch resembles the landmark’s position on the recording arena’s wall. (A) “Landmark control” conditions were those wherein the landmark started within and was maintained in its original position. In later recordings, from the same animals, the landmark may be “rotated” 180º or 90º (LR trials). (B) In separate recordings the landmark may be removed from the recording arena following LR trials. (C) In the third set of recordings a landmark naïve animal’s head was covered with aluminum foil to obstruct visual input.

131

Appendix Figure S2. Related to Figure 2.2. Locations of recording sites within two neuropils of the cockroach central complex. The figure shows the schematic of the frontal view of the central complex (gray: protocerebral bridge; blue: fan-shaped body, green: ellipsoid body, gray lines indicate the start of the outline of the lateral accessory lobes). Left and right indicate directions from the animal’s point of view. Black squares mark the recording sites within the fan-shaped body where no units encoded head direction. Purple squares mark the recording sites where fan-shaped body units significantly encoded head direction (p<0.05, Rayleigh test). Black circles mark the recording sites in the ellipsoid body with no angle-modulated units. Purple circles mark recording sites within the ellipsoid body where significantly angle-modulated units were found (p<0.05, Rayleigh test).

132

14 12 10 8 6

4 Number of units of Number 2 0 45 90 135 180 225 270 315 360 Preferred angle (degrees)

Appendix Figure S3. Related to Figure 2.2 and 2.3. Central complex units possess the capacity to represent any angle relative to the animal, just like a compass. The histogram shows the distribution of preferred angles encoded by CX units.

133

Appendix Figure S4. Related to Figures 2.2, 2.4 and 2.5. CX units encode head direction with similar precision under control conditions with a landmark and in head-covered landmark-naïve animals. We investigated the autocorrelation of each single unit across trials and calculated their average directional stability scores to measure how much error the units accumulate from trial to trial. The distribution of average stability scores are plotted for (A) Control experiments, (B) Landmark rotated experiments, (C) ‘No landmark’ experiments and (D) Head-covered landmark-naïve experiments. (E) To measure the average amount of error accumulation in each experiment, we averaged the directional stability scores during the control, landmark rotated, no landmark and head-covered trials. We tested for group differences with a Kruskal-Wallis nonparametric test (* = significance at p<0.05). Control vs. Landmark rotated (p = 0.055); Control vs. No landmark (p = 0.040); Control vs. Head-covered (p = 0.084).

134

Bibliography

Alphen, B. van, Yap, M. H. W., Kirszenblat, L., Kottler, B. and Swinderen, B. van (2013). A Dynamic Deep Sleep Stage in Drosophila. J. Neurosci. 33, 6917–6927.

Arleo, A., Déjean, C., Allegraud, P., Khamassi, M., Zugaro, M. B. and Wiener, S. I. (2013). Optic flow stimuli update anterodorsal thalamus head direction neuronal activity in rats. J. Neurosci. 33, 16790–5.

Batschelet, E. (1981). Circular Statistics in Biology.

Bech, M., Homberg, U. and Pfeiffer, K. (2014). Receptive fields of locust brain neurons are matched to polarization patterns of the sky. Curr. Biol. 24, 2124–2129.

Bell, W. J., Roth, L. M., Nalepa, C. A. and Wilson, E. O. (2007). Cockroaches: Ecology, behavior, and natural history.

Bender, J. A., Pollack, A. J. and Ritzmann, R. E. (2010). Neural Activity in the Central Complex of the Insect Brain Is Linked to Locomotor Changes. Curr. Biol. 20, 921–926.

Berens, P. (2009). CircStat: A MATLAB toolbox for circular statistics. J. Stat. Softw. 31, 1–21.

Bjerknes, T. L., Langston, R. F., Kruge, I. U., Moser, E. I. and Moser, M. B. (2015). Coherence among head direction cells before eye opening in rat pups. Curr. Biol. 25, 103–108.

Blair, H. T. and Sharp, P. E. (1996). Visual and vestibular influences on head-direction cells in the anterior thalamus of the rat. Behav. Neurosci. 110, 643–660.

Boccara, C. N., Sargolini, F., Thoresen, V. H., Solstad, T., Witter, M. P., Moser, E. I. and Moser, M.-B. (2010). Grid cells in pre- and parasubiculum. Nat. Neurosci. 13, 987–994.

Bockhorst, T. and Homberg, U. (2015). Compass Cells in the Brain of an Insect Are Sensitive to Novel Events in the Visual World. PLoS One 10, e0144501.

Brandon, M. P., Bogaard, A. R., Libby, C. P., Connerney, M. A., Gupta, K. and Hasselmo, M. E. (2011). Reduction of Theta Rhythm Dissociates Grid Cell Spatial Periodicity from Directional Tuning. Science (80-. ). 332, 595–599.

Brandon, M. P., Bogaard, A. R., Schultheiss, N. W. and Hasselmo, M. E. (2013). Segregation of cortical head direction cell assemblies on alternating θ cycles. Nat. Neurosci. 16, 739–48.

135

Broome, B. M., Jayaraman, V. and Laurent, G. (2006). Encoding and Decoding of Overlapping Odor Sequences. Neuron 51, 467–482.

Bülthoff, H., Götz, K. G. and Herre, M. (1982). Recurrent inversion of visual orientation in the walking fly, Drosophila melanogaster. J. Comp. Physiol. □ A 148, 471–481.

Buzsáki, G. (2002). Theta oscillations in the hippocampus. Neuron 33, 325–340.

Buzsáki, G. and Draguhn, A. (2004). Neuronal oscillations in cortical networks. Science 304, 1926–9.

Buzsáki, G. and Freeman, W. (2015). Editorial overview: Brain rhythms and dynamic coordination. Curr. Opin. Neurobiol. 31, 5–9.

Buzsáki, G. and Moser, E. I. (2013). Memory, navigation and theta rhythm in the hippocampal-entorhinal system. Nat. Neurosci. 16, 130–8.

Buzsáki, G. and Watson, B. O. (2012). Brain rhythms and neural syntax: Implications for efficient coding of cognitive content and neuropsychiatric disease. Dialogues Clin. Neurosci. 14, 345–367.

Buzsáki, G., Anastassiou, C. A. and Koch, C. (2012). The origin of extracellular fields and currents -- EEG, ECoG, LFP and spikes. Nat Rev Neurosci 13, 407–420.

Buzsáki, G., Logothetis, N. and Singer, W. (2013). Scaling brain size, keeping timing: Evolutionary preservation of brain rhythms. Neuron 80, 751–764.

Cardin, J. A., Carlén, M., Meletis, K., Knoblich, U., Zhang, F., Deisseroth, K., Tsai, L.-H. and Moore, C. I. (2009). Driving fast-spiking cells induces gamma rhythm and controls sensory responses. Nature 459, 663–7.

Chadwick, M. J., Jolly, A. E. J., Amos, D. P., Hassabis, D. and Spiers, H. J. (2015). A goal direction signal in the human entorhinal/subicular region. Curr. Biol. 25, 87– 92.

Chauvette, S., Crochet, S., Volgushev, M. and Timofeev, I. (2011). Properties of slow oscillation during slow-wave sleep and anesthesia in cats. J. Neurosci. 31, 14998– 5008.

Collett, T. S. and Land, M. F. (1975). Visual control of flight behaviour in the hoverfly Syritta pipiens l. J. Comp. Physiol. 99, 1–66.

Collett, M., Collett, T. S., Bisch, S. and Wehner, R. (1998). Local and global vectors in desert ant navigation. Nature 394, 269–272.

Cruse, H. and Wehner, R. (2011). No need for a cognitive map: Decentralized memory for insect navigation. PLoS Comput. Biol. 7,.

136

Dacke, M., el Jundi, B., Smolka, J., Byrne, M. and Baird, E. (2014). The role of the sun in the celestial compass of dung beetles. Philos. Trans. R. Soc. Lond. B. Biol. Sci. 369, 1–7.

De Bivort, B. L. and Van Swinderen, B. (2016). Evidence for selective attention in the insect brain. Curr. Opin. Insect Sci. 15, 9–15.

DeCoteau, W. E., Thorn, C., Gibson, D. J., Courtemanche, R., Mitra, P., Kubota, Y. and Graybiel, A. M. (2007). Learning-related coordination of striatal and hippocampal theta rhythms during acquisition of a procedural maze task. Proc. Natl. Acad. Sci. U. S. A. 104, 5644–5649.

Deshmukh, S. S., Yoganarasimha, D., Voicu, H. and Knierim, J. J. (2010). Theta modulation in the medial and the lateral entorhinal cortices. J. Neurophysiol. 104, 994–1006.

Dill, M. and Heisenberg, M. (1995). Visual pattern memory without shape recognition. Philos. Trans. R. Soc. Lond. B. Biol. Sci. 349, 143–52.

Dill, M., Wolf, R. and Heisenberg, M. (1993). Visual pattern recognition in Drosophila involves retinotopic matching. Nature 365, 751–3.

Dill, M., Wolf, R. and Heisenberg, M. (1995). Behavioral analysis of Drosophila landmark learning in the flight simulator. Learn. Mem. 2, 152–60.

Dragoi, G. and Buzsáki, G. (2006). Temporal Encoding of Place Sequences by Hippocampal Cell Assemblies. Neuron 50, 145–157.

Ekstrom, A. D., Kahana, M. J., Caplan, J. B., Fields, T. A., Isham, E. A., Newman, E. L. and Fried, I. (2003). Cellular networks underlying human spatial navigation. Nature 425, 184–188.

Ekstrom, A. D., Caplan, J. B., Ho, E., Shattuck, K., Fried, I. and Kahana, M. J. (2005). Human hippocampal theta activity during virtual navigation. Hippocampus 15, 881–889. el Jundi, B., Pfeiffer, K., Heinze, S. and Homberg, U. (2014a). Integration of polarization and chromatic cues in the insect sky compass. J. Comp. Physiol. A Neuroethol. Sensory, Neural, Behav. Physiol. 200, 575–589. el Jundi, B., Smolka, J., Baird, E., Byrne, M. J. and Dacke, M. (2014b). Diurnal dung beetles use the intensity gradient and the polarization pattern of the sky for orientation. J. Exp. Biol. 217, 2422–2429. el Jundi, B., Warrant, E. J., Byrne, M. J., Khaldy, L., Baird, E., Smolka, J. and Dacke, M. (2015). Neural coding underlying the cue preference for celestial orientation. Proc. Natl. Acad. Sci. U. S. A. 112, 11395–400.

137

Ernst, R. and Heisenberg, M. (1999). The memory template in Drosophila pattern vision at the flight simulator. Vision Res. 39, 3920–3933.

Esch, H. E., Zhang, S., Srinivasan, M. V and Tautz, J. (2001). Honeybee dances communicate distances measured by optic flow. Nature 411, 581–583.

Fazelpour, S. and Thompson, E. (2015). The Kantian brain: Brain dynamics from a neurophenomenological perspective. Curr. Opin. Neurobiol. 31, 223–229.

Fell, J. and Axmacher, N. (2011). The role of phase synchronization in memory processes. Nat. Rev. Neurosci. 12, 105–118.

Finkelstein, A., Derdikman, D., Rubin, A., Foerster, J. N., Las, L. and Ulanovsky, N. (2014). Three-dimensional head-direction coding in the bat brain. Nature 517, 159– 164.

Fiore, V. G., Dolan, R. J., Strausfeld, N. J. and Hirth, F. (2015). Evolutionarily conserved mechanisms for the selection and maintenance of behavioural activity. Philos. Trans. R. Soc. B Biol. Sci. 370, 20150053.

Frye, M. a and Dickinson, M. H. (2003). A signature of salience in the Drosophila brain. Nat. Neurosci. 6, 544–546.

Giurfa, M. (2013). Cognition with few neurons: Higher-order learning in insects. Trends Neurosci. 36, 285–294.

Giurfa, M., Zhang, S., Jenett, A., Menzel, R. and Srinivasan, M. V (2001). The concepts of “sameness” and “difference” in an insect. Nature 410, 930–933.

Goodridge, J. P., Dudchenko, P. a, Worboys, K. a, Golob, E. J. and Taube, J. S. (1998). Cue control and head direction cells. Behav. Neurosci. 112, 749–761.

Gould, J. L. (1986). The locale map of honey bees: do insects have cognitive maps? Science 232, 861–863.

Griffin, A. L., Eichenbaum, H. and Hasselmo, M. E. (2007). Spatial representations of hippocampal CA1 neurons are modulated by behavioral context in a hippocampus- dependent memory task. J. Neurosci. 27, 2416–2423.

Guo, P. and Ritzmann, R. E. (2013). Neural activity in the central complex of the cockroach brain is linked to turning behaviors. J. Exp. Biol. 216, 992–1002.

Guo, P., Pollack, A. J., Varga, A. G., Martin, J. P. and Ritzmann, R. E. (2014). Extracellular wire tetrode recording in brain of freely walking insects. J. Vis. Exp. 1–8.

Haberkern, H. and Jayaraman, V. (2016). Studying small brains to understand the building blocks of cognition. Curr. Opin. Neurobiol. 37, 59–65.

138

Hafting, T., Fyhn, M., Molden, S., Moser, M.-B. and Moser, E. I. (2005). Microstructure of a spatial map in the entorhinal cortex. Nature 436, 801–6.

Hansen, L., Jonsson, N. and Jonsson, B. (1993). Oceanic migration in homing Atlantic salmon. Anim. Behav. 45, 927–941.

Harley, C. M. and Ritzmann, R. E. (2010). Electrolytic lesions within central complex neuropils of the cockroach brain affect negotiation of barriers. J. Exp. Biol. 213, 2851–2864.

Hartley, T., Lever, C., Burgess, N. and O’Keefe, J. (2014). Space in the brain: how the hippocampal formation supports spatial cognition. Philos. Trans. R. Soc. Lond. B. Biol. Sci. 369, 20120510.

Hasselmo, M. E. (2005). What is the function of hippocampal theta rhythm? - Linking bahavioral data to phasic properties of field potential and unit recording data. Hippocampus 15, 936–949.

Hasselmo, M. E. and Eichenbaum, H. (2005). Hippocampal mechanisms for the context-dependent retrieval of episodes. Neural Networks 18, 1172–1190.

Heinze, S. (2015). Neuroethology: Unweaving the Senses of Direction. Curr. Biol. 25, R1034--7.

Heinze, S. and Homberg, U. (2007). Maplike representation of celestial E-vector orientations in the brain of an insect. Science (80-. ). 315, 995–997.

Heinze, S. and Homberg, U. (2008). Neuroarchitecture of the central complex of the desert locust: Intrinsic and columnar neurons. J. Comp. Neurol. 511, 454–478.

Heinze, S. and Homberg, U. (2009). Linking the input to the output: new sets of neurons complement the polarization vision network in the locust central complex. J. Neurosci. 29, 4911–4921.

Heinze, S. and Reppert, S. M. (2011). Sun compass integration of skylight cues in migratory monarch butterflies. Neuron 69, 345–358.

Heinze, S., Gotthardt, S. and Homberg, U. (2009). Transformation of polarized light information in the central complex of the locust. J. Neurosci. 29, 11783–11793.

Heinze, S., Florman, J., Asokaraj, S., El Jundi, B. and Reppert, S. M. (2013). Anatomical basis of sun compass navigation II: The neuronal composition of the central complex of the monarch butterfly. J. Comp. Neurol. 521, 267–298.

Hemberger, M., Pammer, L. and Laurent, G. (2016). Comparative approaches to cortical microcircuits. Curr. Opin. Neurobiol. 41, 24–30.

139

Homberg, U. (1985). Interneurones of the central complex in the bee brain (Apis mellifera, L.). J. Insect Physiol. 31,.

Homberg, U. (1994). Flight-correlated activity changes in neurons of the lateral accessory lobes in the brain of the locust Schistocerca gregaria. J. Comp. Physiol. A 175, 597–610.

Homberg, U., Wurden, S., Dircksen, H. and Rao, K. R. (1991). Comparative anatomy of pigment-dispersing hormone-immunoreactive neurons in the brain of orthopteroid insects. Cell Tissue Res. 266, 343–357.

Homberg, U., Hofer, S., Pfeiffer, K. and Gebhardt, S. (2003). Organization and neural connections of the anterior optic tubercle in the brain of the locust, Schistocerca gregaria. J. Comp. Neurol. 462, 415–430.

Homberg, U., Heinze, S., Pfeiffer, K., Kinoshita, M. and el Jundi, B. (2011). Central neural coding of sky polarization in insects. Philos. Trans. R. Soc. Lond. B. Biol. Sci. 366, 680–687.

Huber F (1960). Untersuchungen über die Funktion des Zentralnervensystems und insbesondere des Gehirnes bei der Fortbewegung und der Lauterzeugung der Grillen. Z. Vgl. Physiol. 60–132.

Huxter, J., Burgess, N. and O’Keefe, J. (2003). Independent rate and temporal coding in hippocampal pyramidal cells. Nature 425, 828–32.

Huxter, J. R., Senior J, T., Allen, K. and Csicsvari, J. (2008). Theta phase-specific codes for two-dimensional position, trajectory and heading in the hippocampus. Nat. Neurosci. 11, 587–594.

Igarashi, K. M., Lu, L., Colgin, L. L., Moser, M.-B. and Moser, E. I. (2014). Coordination of entorhinal-hippocampal ensemble activity during associative learning. Nature 510, 143–7.

Ito, I., Bazhenov, M., Ong, R. C., Raman, B. and Stopfer, M. (2009). Frequency Transitions in Odor-Evoked Neural Oscillations. Neuron 64, 692–706.

Jacobs, J., Kahana, M. J., Ekstrom, A. D., Mollison, M. V. and Fried, I. (2010). A sense of direction in human entorhinal cortex. Proc. Natl. Acad. Sci. 107, 6487– 6492.

James, W. (1890). The principles of psychology (Vols. 1 & 2). New York Holt 118, 688.

Jankowski, M. M., Ronnqvist, K. C., Tsanov, M., Vann, S. D., Wright, N. F., Erichsen, J. T., Aggleton, J. P. and O’Mara, S. M. (2013). The anterior thalamus provides a subcortical circuit supporting memory and spatial navigation. Front. Syst. Neurosci. 7, 45.

140

Jeewajee, A., Barry, C., O’Keefe, J. and Burgess, N. (2008). Grid cells and theta as oscillatory interference: Electrophysiological data from freely moving rats. Hippocampus 18, 1175–1185.

Jeffery, K. J., Donnett, J. G. and O’Keefe, J. (1995). Medial septal control of theta- correlated unit firing in the entorhinal cortex of awake rats. Neuroreport. 6, 2166– 2170.

Kamikouchi, A., Inagaki, H. K., Effertz, T., Hendrich, O., Fiala, A., Göpfert, M. C. and Ito, K. (2009). The neural basis of Drosophila gravity-sensing and hearing. Nature 458, 165–171.

Kathman, N. D., Kesavan, M. and Ritzmann, R. E. (2014). Encoding wide-field motion and direction in the central complex of the cockroach Blaberus discoidalis. J. Exp. Biol. 217, 4079–90.

Kaun, K. R., Riedl, C. A. L., Chakaborty-Chatterjee, M., Belay, A. T., Douglas, S. J., Gibbs, A. G. and Sokolowski, M. B. (2007). Natural variation in food acquisition mediated via a Drosophila cGMP-dependent protein kinase. J. Exp. Biol. 210, 3547–58.

Kay, L. M. (2014). Circuit oscillations in odor perception and memory. Prog. Brain Res. 208, 223–251.

Kay, L. M. (2015). Olfactory system oscillations across phyla. Curr. Opin. Neurobiol. 31, 141–147.

Kay, L. M. and Beshel, J. (2010). A Beta Oscillation Network in the Rat Olfactory System During a 2-Alternative Choice Odor Discrimination Task. J. fo Neurophysiol. 104, 829–839.

Kent, C. F., Daskalchuk, T., Cook, L., Sokolowski, M. B. and Greenspan, R. J. (2009). The Drosophila foraging gene mediates adult plasticity and gene- environment interactions in behaviour, metabolites, and gene expression in response to food deprivation. PLoS Genet. 5,.

Kirschfeld, K. (1992). Oscillations in the insect brain: do they correspond to the cortical gamma-waves of vertebrates? Proc. Natl. Acad. Sci. U. S. A. 89, 4764–4768.

Kirszenblat, L. and van Swinderen, B. (2015). The Yin and Yang of Sleep and Attention. Trends Neurosci. 38, 776–786.

Knierim, J. J. and Zhang, K. (2012). Attractor Dynamics of Spatially Correlated Neural Activity in the Limbic System. Annu. Rev. Neurosci. 35, 267–285.

Knierim, J. J., Kudrimoti, H. S. and McNaughton, B. L. (1995). Place cells, head direction cells, and the learning of landmark stability. J. Neurosci. 15, 1648–1659.

141

Knierim, J. J., Kudrimoti, H. S. and McNaughton, B. L. (1998). Interactions between idiothetic cues and external landmarks in the control of place cells and head direction cells. J. Neurophysiol. 80, 425–446.

Koenig, J., Linder, A. N., Leutgeb, J. K. and Leutgeb, S. (2011). The Spatial Periodicity of Grid Cells Is Not Sustained During Reduced Theta Oscillations. Science (80-. ). 332, 592–595.

Kopell, N., Kramer, M. A., Malerba, P. and Whittington, M. A. (2010). Are different rhythms good for different functions? Front. Hum. Neurosci. 4, 12.

Krashes, M. J., DasGupta, S., Vreede, A., White, B., Armstrong, J. D. and Waddell, S. (2009). A Neural Circuit Mechanism Integrating Motivational State with Memory Expression in Drosophila. Cell 139, 416–427.

Kropff, E., Carmichael, J. E., Moser, M.-B. and Moser, E. I. (2015). Speed cells in the medial entorhinal cortex. Nature 523, 419–424.

Labhart, T. and Meyer, E. P. (1999). Detectors for polarized skylight in insects: a survey of ommatidial specializations in the dorsal rim area of the compound eye. Microsc. Res. Tech. 47, 368–379.

Lakatos, P., Karmos, G., Mehta, A. D., Ulbert, I. and Schroeder, C. E. (2008). Entrainment of neuronal oscillations as a mechanism of attentional selection. Science 320, 110–3.

Laurent, G. (1996). Dynamical representation of odors by oscillating and evolving neural assemblies. Trends Neurosci. 19, 489–496.

Laurent, G. (1999). A systems perspective on early olfactory coding. Science 286, 723– 728.

Laurent, G. and Davidowitz, H. (1994). Encoding of olfactory information with oscillating neural assemblies. Science 265, 1872–5.

Laurent, G. and Naraghi, M. (1994). Odorant-induced oscillations in the mushroom bodies of the locust. J. Neurosci. 14, 2993–3004.

Lavoie, A. M. and Mizumori, S. J. Y. (1994). Spatial, movement- and reward-sensitive discharge by medial ventral striatum neurons of rats. Brain Res. 638, 157–168.

Li, W., Pan, Y., Wang, Z., Gong, H., Gong, Z. and Liu, L. (2009). Morphological characterization of single fan-shaped body neurons in Drosophila melanogaster. Cell Tissue Res. 336, 509–519.

Liu, L., Wolf, R., Ernst, R. and Heisenberg, M. (1999). Context generalization in Drosophila visual learning requires the mushroom bodies. Nature 400, 753–756.

142

Liu, G., Seiler, H., Wen, A., Zars, T., Ito, K., Wolf, R., Heisenberg, M., Liu, L., District, C., Hubland, A., et al. (2006). Distinct memory traces for two visual features in the Drosophila brain. Nature 439, 551–6.

Malhotra, S., Cross, R. W. A. and Van Der Meer, M. A. A. (2012). Theta phase precession beyond the hippocampus. Rev. Neurosci. 23, 39–65.

Marder, E., Goeritz, M. L. and Otopalik, A. G. (2015). Robust circuit rhythms in small circuits arise from variable circuit components and mechanisms. Curr. Opin. Neurobiol. 31, 156–163.

Martin, J. R., Raabe, T. and Heisenberg, M. (1999). Central complex substructures are required for the maintenance of locomotor activity in Drosophila melanogaster. J. Comp. Physiol. - A Sensory, Neural, Behav. Physiol. 185, 277–288.

Martin, J. P., Guo, P., Mu, L., Harley, C. M. and Ritzmann, R. E. (2015). Central- complex control of movement in the freely walking cockroach. Curr. Biol. 25, 2795–2803.

Matsuo, E., Yamada, D., Ishikawa, Y., Asai, T., Ishimoto, H. and Kamikouchi, A. (2014). Identification of novel vibration- and deflection-sensitive neuronal subgroups in Johnston’s organ of the fruit fly. Front. Physiol. 5 MAY,.

Maurer, A. P., Cowen, S. L., Burke, S. N., Barnes, C. A. and McNaughton, B. L. (2006). Organization of hippocampal cell assemblies based on theta phase precession. Hippocampus 16, 785–794.

McGeorge, A. J. and Faull, R. L. M. (1989). The organization of the projection from the cerebral cortex to the striatum in the rat. Neuroscience 29, 503–537.

McNaughton, B. L., Barnes, C. A. and O’Keefe, J. (1983). The contributions of position, direction, and velocity to single unit activity in the hippocampus of freely- moving rats. Exp. Brain Res. 52, 41–49.

McNaughton, B. L., Battaglia, F. P., Jensen, O., Moser, E. I. and Moser, M.-B. (2006). Path integration and the neural basis of the “cognitive map”. Nat. Rev. Neurosci. 7, 663–678.

Mendoza-Angeles, K., Cabrera, A., Hernández-Falcón, J. and Ramón, F. (2007). Slow waves during sleep in crayfish: A time-frequency analysis. J. Neurosci. Methods 162, 264–271.

Mendoza-Angeles, K., Hernández-Falcón, J. and Ramón, F. (2010). Slow waves during sleep in crayfish. Origin and spread. J. Exp. Biol. 213, 2154–64.

Menzel, R. and Giurfa, M. (2001). Cognitive architecture of a mini-brain: The honeybee. Trends Cogn. Sci. 5, 62–71.

143

Merlin, C., Heinze, S. and Reppert, S. M. (2012). Unraveling navigational strategies in migratory insects. Curr. Opin. Neurobiol. 22, 353–361.

Milde, J. J. (1988). Visual responses of interneurones in the posterior median protocerebrum and the central complex of the honeybee Apis mellifera. J. Insect Physiol. 34, 427–436.

Mischiati, M., Lin, H.-T., Herold, P., Imler, E., Olberg, R. and Leonardo, A. (2015). Internal models direct dragonfly interception steering. Nature 517, 333–338.

Mizumori, S. J. Y., Ragozzino, K. E. and Cooper, B. G. (2000). Location and head direction representation in the dorsal striatum of rats. Psychobiology 28, 441–462.

Mizumori, S. J. Y., Puryear, C. B. and Martig, A. K. (2009). Basal ganglia contributions to adaptive navigation. Behav. Brain Res. 199, 32–42.

Mizunami, M., Weibrecht, J. M. and Strausfeld, N. J. (1998). Mushroom bodies of the cockroach: Their participation in place memory. J. Comp. Neurol. 402, 520–537.

Mizuseki, K., Sirota, A., Pastalkova, E. and Buzsáki, G. (2009). Theta Oscillations Provide Temporal Windows for Local Circuit Computation in the Entorhinal- Hippocampal Loop. Neuron 64, 267–280.

Morris, R. G. M. (1981). Spatial Localization Does Not Require Local Cues the Presence of. Learn. Motiv. 12, 239–260.

Morris, R. G. M., Garrud, P., Rawlins, J. N. and O’Keefe, J. (1982). Place navigation impaired in rats with hippocampal lesions. Nature 297, 681–683.

Muir, G. M., Brown, J. E., Carey, J. P., Hirvonen, T. P., Della Santina, C. C., Minor, L. B. and Taube, J. S. (2009). Disruption of the head direction cell signal after occlusion of the semicircular canals in the freely moving chinchilla. J. Neurosci. 29, 14521–33.

Müller, M. and Wehner, R. (1988). Path integration in desert ants, Cataglyphis fortis. Proc. Natl. Acad. Sci. U. S. A. 85, 5287–5290.

Neuser, K., Triphan, T., Mronz, M., Poeck, B. and Strauss, R. (2008). Analysis of a spatial orientation memory in Drosophila. Nature 453, 1244–1247.

Nitz, D. A., Van Swinderen, B., Tononi, G. and Greenspan, R. J. (2002). Electrophysiological correlates of rest and activity in Drosophila melanogaster. Curr. Biol. 12, 1934–1940.

O’Keefe, J. and Dostrovsky, J. (1971). The hippocampus as a spatial map. Preliminary evidence from unit activity in the freely-moving rat. Brain Res. 34, 171–175.

144

O’Keefe, J. and Recce, M. L. (1993). Phase relationship between hippocampal place units and the EEG theta rhythm. Hippocampus 3, 317–330.

Ofstad, T. A., Zuker, C. S. and Reiser, M. B. (2011). Visual place learning in Drosophila melanogaster. Nature 474, 204–7.

Okada, R., Sakura, M. and Mizunami, M. (2003). Distribution of dendrites of descending neurons and its implications for the basic organization of the cockroach brain. J. Comp. Neurol. 458, 158–174.

Palop, J. J. and Mucke, L. (2010). Amyloid-beta-induced neuronal dysfunction in Alzheimer’s disease: from synapses toward neural networks. Nat. Neurosci. 13, 812–818.

Pan, Y., Zhou, Y., Guo, C., Gong, H., Gong, Z. and Liu, L. (2009). Differential roles of the fan-shaped body and the ellipsoid body in Drosophila visual pattern memory. Learn. Mem. 16, 289–295.

Patel, J., Fujisawa, S., Berényi, A., Royer, S. and Buzsáki, G. (2012). Traveling Theta Waves along the Entire Septotemporal Axis of the Hippocampus. Neuron 75, 410– 417.

Paulk, A. C., Zhou, Y., Stratton, P., Liu, L. and van Swinderen, B. (2013). Multichannel brain recordings in behaving Drosophila reveal oscillatory activity and local coherence in response to sensory stimulation and circuit activation. J. Neurophysiol. 110, 1703–21.

Paulk, A. C., Stacey, J. a, Pearson, T. W. J., Taylor, G. J., Moore, R. J. D., Srinivasan, M. V and van Swinderen, B. (2014). Selective attention in the honeybee optic lobes precedes behavioral choices. Proc. Natl. Acad. Sci. U. S. A. 111, 5006–11.

Paulk, A. C., Kirszenblat, L., Zhou, Y. and van Swinderen, B. (2015). Closed-Loop Behavioral Control Increases Coherence in the Fly Brain. J. Neurosci. 35, 10304– 15.

Paulsen, O. and Moser, E. I. (1998). A model of hippocampal memory encoding and retrieval: GABAergic control of synaptic plasticity. Trends Neurosci. 21, 273–278.

Paxinos, G. and Watson, C. (1997). The rat brain in stereotaxic coordinates.

Penner, M. R. and Mizumori, S. J. Y. (2012). Neural systems analysis of decision making during goal-directed navigation. Prog. Neurobiol. 96, 96–135.

Peyrache, A., Lacroix, M. M., Petersen, P. C. and Buzsáki, G. (2015). Internally organized mechanisms of the head direction sense. Nat. Neurosci. 18, 569–75.

145

Pfeiffer, K. and Homberg, U. (2014). Organization and functional roles of the central complex in the insect brain. Annu. Rev. Entomol. 59, 165–84.

Pfeiffer, K., Kinoshita, M. and Homberg, U. (2005). Polarization-sensitive and light- sensitive neurons in two parallel pathways passing through the anterior optic tubercle in the locust brain. J. Neurophysiol. 94, 3903–3915.

Phillips-Portillo, J. (2012). The central complex of the flesh fly, Neobellieria bullata: Recordings and morphologies of protocerebral inputs and small-field neurons. J. Comp. Neurol. 520, 3088–3104.

Poeck, B., Triphan, T., Neuser, K. and Strauss, R. (2008). Locomotor control by the central complex in Drosophila - An analysis of the tay bridge mutant. Dev. Neurobiol. 68, 1046–1058.

Ramón, F., Hernández-Falcón, J., Nguyen, B. and Bullock, T. H. (2004). Slow wave sleep in crayfish. Proc. Natl. Acad. Sci. U. S. A. 101, 11857–11861.

Ranck, J. B. (1985). Head direction cells in the deep cell layer of dorsal presubiculum in freely moving rats. In Electrical Activity of the Archicortex (ed. Buzsaki, G.) and Vanderwolf, C. H.), pp. 217–220. Budapest: Akademia Kiado.

Ray, S., Naumann, R., Burgalossi, A., Tang, Q., Schmidt, H. and Brecht, M. (2014). Grid-layout and theta-modulation of layer 2 pyramidal neurons in medial entorhinal cortex. Science 343, 891–6.

Reinhard, J., Srinivasan, M. V and Zhang, S. (2004). Olfaction: scent-triggered navigation in honeybees. Nature 427, 411.

Reppert, S. M., Zhu, H. and White, R. H. (2004). Polarized Light Helps Monarch Butterflies Navigate. Curr. Biol. 14, 155–158.

Reppert, S. M., Guerra, P. A. and Merlin, C. (2016). Neurobiology of Monarch Butterfly Migration. Annu. Rev. Entomol. 61, annurev-ento-010814-020855.

Ridgel, A. L., Alexander, B. E. and Ritzmann, R. E. (2007). Descending control of turning behavior in the cockroach, Blaberus discoidalis. J. Comp. Physiol. A Neuroethol. Sensory, Neural, Behav. Physiol. 193, 385–402.

Ritzmann, R. E., Ridgel, A. L. and Pollack, A. J. (2008). Multi-unit recording of antennal mechano-sensitive units in the central complex of the cockroach, Blaberus discoidalis. J. Comp. Physiol. A Neuroethol. Sensory, Neural, Behav. Physiol. 194, 341–360.

Ritzmann, R. E., Harley, C. M., Daltorio, K. a, Tietz, B. R., Pollack, A. J., Bender, J. a, Guo, P., Horomanski, A. L., Kathman, N. D., Nieuwoudt, C., et al. (2012). Deciding which way to go: how do insects alter movements to negotiate barriers? Front. Neurosci. 6, 97.

146

Robertson, R. G., Rolls, E. T., Georges-Francois, P. and Panzeri, S. (1999). Head direction cells in the primate pre-subiculum. Hippocampus 9, 206–219.

Robinson, T. (1980). Hippocampal rhythmic slow activity (RSA; theta): a critical analysis of selected studies and discussion of possible species-differences. Brain Res. 203, 69–101.

Rodriguez, E., George, N., Lachaux, J. P., Martinerie, J., Renault, B. and Varela, F. J. (1999). Perception’s shadow: long-distance synchronization of human brain activity. Nature 397, 430–433.

Rowland, D. C., Roudi, Y., Moser, M.-B. and Moser, E. I. (2016). Ten Years of Grid Cells. Annu. Rev. Neurosci. 39, annurev-neuro-070815-013824.

Rubin, A., Yartsev, M. M., Ulanovsky, N., Smith, D. M. and Mizumori, S. J. Y. (2014). Encoding of head direction by hippocampal place cells in bats. J. Neurosci. 34, 1067–80.

Sakura, M., Lambrinos, D. and Labhart, T. (2008). Polarized skylight navigation in insects: model and electrophysiology of e-vector coding by neurons in the central complex. J. Neurophysiol. 99, 667–682.

Sale, A. and Luschi, P. (2009). Navigational challenges in the oceanic migrations of leatherback sea turtles. Proc. Biol. Sci. 276, 3737–3745.

Seelig, J. D. and Jayaraman, V. (2013). Feature detection and orientation tuning in the Drosophila central complex. Nature 503, 262–266.

Seelig, J. D. and Jayaraman, V. (2015). Neural dynamics for landmark orientation and angular path integration. Nature 521, 186–191.

Shinder, M. E. and Taube, J. S. (2011). Active and passive movement are encoded equally by head direction cells in the anterodorsal thalamus. J. Neurophysiol. 106, 788–800.

Skaggs, W. E., Knierim, J. J., Kudrimoti, H. S. and McNaughton, B. L. (1995). A model of the neural basis of the rat’s sense of direction. Adv. Neural Inf. Process. Syst. 7, 173–180.

Srinivasan, M. V. (2015). Where paths meet and cross: navigation by path integration in the desert ant and the honeybee. J. Comp. Physiol. A Neuroethol. Sensory, Neural, Behav. Physiol. 201, 533–546.

Stackman, R. W. and Taube, J. S. (1997). Firing properties of head direction cells in the rat anterior thalamic nucleus: dependence on vestibular input. J. Neurosci. 17, 4349–4358.

147

Stackman, R. W., Clark, A. S. and Taube, J. S. (2002). Hippocampal spatial representations require vestibular input. Hippocampus 12, 291–303.

Stackman, R. W., Golob, E. J., Bassett, J. P. and Taube, J. S. (2003). Passive transport disrupts directional path integration by rat head direction cells. J. Neurophysiol. 90, 2862–74.

Steriade, M. (2001). Impact of network activities on neuronal properties in corticothalamic systems. J. Neurophysiol. 86, 1–39.

Stewart, M. and Fox, S. E. (1991). Hippocampal theta activity in monkeys.

Stopfer, M., Bhagavan, S., Smith, B. H. and Laurent, G. (1997). Impaired odor discrimination on desynchronization of odor–encoding neural assemblies. Nature 390, 70–74.

Strausfeld, N. J. (1976). Atlas of an Insect Brain. Springer 52, 1096–1109.

Strausfeld, N. J. (1999). A brain region in insects that supervises walking. Prog. Brain Res. Res. 123, 273–284.

Strausfeld, N. J. and Hirth, F. (2013). Deep Homology of Arthropod Central Complex and Vertebrate Basal Ganglia. Sci. (New York, NY) 340, 157–161.

Strausfeld, N. J., Sinakevitch, I., Brown, S. M. and Farris, S. M. (2009). Ground plan of the insect mushroom body: Functional and evolutionary implications. J. Comp. Neurol. 513, 265–291.

Strauss, R. (2002). The central complex and the genetic dissection of locomotor behaviour. Curr. Opin. Neurobiol. 12, 633–638.

Strauss, R. and Heisenberg, M. (1993). A higher control center of locomotor behavior in the Drosophila brain. J. Neurosci. 13, 1852–61.

Strauss, R., Hanesch, U., Kinkelin, M., Wolf, R. and Heisenberg, M. (1992). No- bridge of Drosophila melanogaster: portrait of a structural brain mutant of the central complex. J. Neurogenet. 8, 125–155.

Tang, S. and Juusola, M. (2010). Intrinsic activity in the fly brain gates visual information during behavioral choices. PLoS One 5,.

Tang, S., Wolf, R., Xu, S. and Heisenberg, M. (2004). Visual pattern recognition in Drosophila is invariant for retinal position. Science 305, 1020–1022.

Taube, J. S. (1995). Head direction cells recorded in the anterior thalamic nuclei of freely moving rats. J Neurosci 15, 70–86.

148

Taube, J. S. (2007). The head direction signal: origins and sensory-motor integration. Annu. Rev. Neurosci. 30, 181–207.

Taube, J. S. (2010). Interspike Interval Analyses Reveal Irregular Firing Patterns at Short, But Not Long, Intervals in Rat Head Direction Cells. J. Neurophysiol. 104, 1635–1648.

Taube, J. S. and Burton, H. L. (1995). Head direction cell activity monitored in a novel environment and during a cue conflict situation. J. Neurophysiol. 74, 1953–1971.

Taube, J. S., Muller, R. U. and Ranck, J. B. (1990a). Head-direction cells recorded from the postsubiculum in freely moving rats. I. Description and quantitative analysis. J. Neurosci. 10, 420–435.

Taube, J. S., Muller, R. U. and Ranck, J. B. (1990b). Head-direction cells recorded from the postsubiculum in freely moving rats. II. Effects of environmental manipulations. J. Neurosci. 10, 436–447.

Tiesinga, P. H. E. and Sejnowski, T. J. (2010). Mechanisms for phase shifting in cortical networks and their role in communication through coherence. Front. Hum. Neurosci. 4, 5.

Tort, A. B. L., Kramer, M. a, Thorn, C., Gibson, D. J., Kubota, Y., Graybiel, A. M. and Kopell, N. J. (2008). Dynamic cross-frequency couplings of local field potential oscillations in rat striatum and hippocampus during performance of a T- maze task. Proc. Natl. Acad. Sci. U. S. A. 105, 20517–20522.

Träger, U., Wagner, R., Bausenwein, B. and Homberg, U. (2008). A novel type of microglomerular synaptic complex in the polarization vision pathway of the locust brain. J. Comp. Neurol. 506, 288–300.

Triphan, T., Poeck, B., Neuser, K. and Strauss, R. (2010). Visual Targeting of Motor Actions in Climbing Drosophila. Curr. Biol. 20, 663–668.

Trongnetrpunya, A., Nandi, B., Kang, D., Kocsis, B., Schroeder, C. E. and Ding, M. (2015). Assessing Granger Causality in Electrophysiological Data: Removing the Adverse Effects of Common Signals via Bipolar Derivations. Front. Syst. Neurosci. 9, 189.

Tsanov, M., Chah, E., Vann, S. D., Reilly, R. B., Erichsen, J. T., Aggleton, J. P. and O’Mara, S. M. (2011). Theta-modulated head direction cells in the rat anterior thalamus. J. Neurosci. 31, 9489–502.

Tsodyks, M. V, Skaggs, W. E., Sejnowski, T. J. and McNaughton, B. L. (1996). Population dynamics and theta rhythm phase precession of hippocampal place cell firing: a spiking neuron model. Hippocampus 6, 271–280.

149

Ulanovsky, N. and Moss, C. F. (2007). Hippocampal cellular and network activity in freely moving echolocating bats. Nat. Neurosci. 10, 224–233.

Valerio, S. and Taube, J. S. (2012). Path integration: how the head direction signal maintains and corrects spatial orientation. Nat. Neurosci. 15, 1445–53.

Valerio, S. and Taube, J. S. (2016). Head Direction Cell Activity Is Absent in Mice without the Horizontal Semicircular Canals. J. Neurosci. 36, 741–754. van der Meer, M. A. A., Knierim, J. J., Yoganarasimha, D., Wood, E. R. and van Rossum, M. C. W. (2007). Anticipation in the rodent head direction system can be explained by an interaction of head movements and vestibular firing properties. J. Neurophysiol. 98, 1883–1897. van Swinderen, B. (2007). Attention-like processes in Drosophila require short-term memory genes. Science 315, 1590–1593. van Swinderen, B. and Brembs, B. (2010). Attention-like deficit and hyperactivity in a Drosophila memory mutant. J. Neurosci. 30, 1003–1014. van Swinderen, B. and Greenspan, R. J. (2003). Salience modulates 20-30 Hz brain activity in Drosophila. Nat. Neurosci. 6, 579–586. van Swinderen, B., McCartney, A., Kauffman, S., Flores, K., Agrawal, K., Wagner, J. and Paulk, A. (2009). Shared visual attention and memory systems in the Drosophila brain. PLoS One 4,.

Van Swinderen, B., Nitz, D. A. and Greenspan, R. J. (2004). Uncoupling of brain activity from movement defines arousal States in Drosophila. Curr. Biol. 14, 81–7.

Vanderwolf, C. H. (1969). Hippocampal electrical activity and voluntary movement in the rat. Electroencephalogr. Clin. Neurophysiol. 26, 407–418.

Varga, A. G. and Ritzmann, R. E. (2016). Cellular Basis of Head Direction and Contextual Cues in the Insect Brain. Curr. Biol. 26, 1816–1828.

Vass, L. K. and Epstein, R. A. (2013). Abstract representations of location and facing direction in the human brain. J. Neurosci. 33, 6133–6142.

Von Frisch, K. (1967). The dance language and orientation of bees. Heredity (Edinb). 90, 212.

Waespe, W. and Henn, V. (1977). Vestibular nuclei activity during optokinetic after- nystagmus (OKAN) in the alert monkey. Exp Brain Res 30, 323–30.

Walcott, C. (1996). Pigeon homing: observations, experiments and confusions. J. Exp. Biol. 199, 21–7.

150

Ward, L. M. (2003). Synchronous neural oscillations and cognitive processes. Trends Cogn. Sci. 7, 553–559.

Watson, B. O. and Buzsáki, G. (2015). Sleep, Memory & Brain Rhythms. Daedalus 144, 67.

Webb, B. (2012). Cognition in insects. Philos. Trans. R. Soc. B Biol. Sci. 367, 2715– 2722.

Webb, B. and Wystrach, A. (2016). Neural mechanisms of insect navigation. Curr. Opin. Insect Sci. 15, 27–39.

Wehner, R. (2003). Desert ant navigation: How miniature brains solve complex tasks. In Journal of Comparative Physiology A: Neuroethology, Sensory, Neural, and Behavioral Physiology, pp. 579–588.

Wehner, R. and Menzel, R. (1990). Do insects have cognitive maps? Annu. Rev. Neurosci. 13, 403–414.

Wehr, M. S. and Laurent, G. (1996). Odour encoding by temporal sequences of firing in oscillating neural assemblies. Nature 384, 162–6.

Whitlock, J. R., Sutherland, R. J., Witter, M. P., Moser, M.-B. and Moser, E. I. (2008). Navigating from hippocampus to parietal cortex. Proc. Natl. Acad. Sci. U. S. A. 105, 14755–14762.

Wiederman, S. D. and O’Carroll, D. C. (2013). Selective attention in an insect visual neuron. Curr. Biol. 23, 156–161.

Winson, J. (1972). Interspecies differences in the occurrence of theta. Behav. Biol. 7, 479–487.

Winter, S. S., Clark, B. J. and Taube, J. S. (2015a). Disruption of the head direction cell network impairs the parahippocampal grid cell signal. Science (80-. ). 347, 870– 4.

Winter, S. S., Mehlman, M. L., Clark, B. J. and Taube, J. S. (2015b). Passive transport disrupts grid signals in the parahippocampal cortex. Curr. Biol. 25, 2493– 2502.

Womelsdorf, T., Schoffelen, J.-M., Oostenveld, R., Singer, W., Desimone, R., Engel, A. K. and Fries, P. (2007). Modulation of neuronal interactions through neuronal synchronization. Science 316, 1609–12.

Yartsev, M. M., Witter, M. P. and Ulanovsky, N. (2011). Grid cells without theta oscillations in the entorhinal cortex of bats. Nature 479, 103–107.

151

Yoder, R. M. and Taube, J. S. (2009). Head direction cell activity in mice: robust directional signal depends on intact otolith organs. J. Neurosci. 29, 1061–76.

Yoder, R. M., Clark, B. J., Brown, J. E., Lamia, M. V, Valerio, S., Shinder, M. E. and Taube, J. S. (2011). Both visual and idiothetic cues contribute to head direction cell stability during navigation along complex routes. J. Neurophysiol. 105, 2989– 3001.

Yoganarasimha, D., Yu, X. and Knierim, J. J. (2006). Head Direction Cell Representations Maintain Internal Coherence during Conflicting Proximal and Distal Cue Rotations: Comparison with Hippocampal Place Cells. J. Neurosci. 26, 622–631.

Yorozu, S., Wong, A., Fischer, B. J., Dankert, H., Kernan, M. J., Kamikouchi, A., Ito, K. and Anderson, D. J. (2009). Distinct sensory representations of wind and near-field sound in the Drosophila brain. Nature 458, 201–205.

Zars, T., Fischer, M., Schulz, R. and Heisenberg, M. (2000). Localization of a Short- Term Memory in Drosophila. Science. 288, 672–675.

Zugaro, M. B., Tabuchi, E., Fouquier, C., Berthoz, a and Wiener, S. I. (2001). Active locomotion increases peak firing rates of anterodorsal thalamic head direction cells. J. Neurophysiol. 86, 692–702.

152