Neuroethology of bat navigation

Current Biology

Magazine a large repertoire of compass cues (Figure 1A,B), dynamically adjusting the relative weights of these cues according to their availability and the time of day. Behaviourally, this can most easily be demonstrated by the use of a mirror. When positioned so as to change the position of the sun, or the moon, by 180° (while the experimenter simultaneously shields the real sun or moon from the beetle’s view), diurnal beetles swiftly change their bearings by about 180°. Nocturnal beetles, on the other hand, carry on as if nothing has changed. The explanation for these contrasting behaviours can be found by simulating the same situation in the lab, while simultaneously recording from the TL (Figure 2C) and CL1 neurons in the central complex of either of the two species of beetles. From these recordings, it is obvious that the diurnal beetle’s sky-compass network codes the sun as the main reference for straight-line orientation. In their nocturnal cousins, the same types of compass neurons are primarily tuned to the polarization pattern of the moon, rather than to the moon itself. Consequently, as the sky-wide polarization pattern remains unaffected by the mirror in our experiment above, the beetles do not follow the ‘displaced moon’. But if the nocturnal beetles are instead made to roll during the day (which is possible through their abrupt exhumation from the ground followed by the addition of some tempting, fresh dung) they can now be observed to quickly and accurately change their bearing in response to the mirrored sun. An identical switch, from broad polarization tuning to point source tuning, can be recorded from their central-complex neurons as we raise the light levels in the lab from dim to bright. This dynamic modulation of the relative impact of different compass cues most likely serves to allow the animals to follow the most reliable compass cue at any moment in time. On moonless nights, the beetles fall back on the Milky Way for directional guidance (Figure 1B). Even though we do not yet know how this celestial cue is encoded, we expect it to be processed by the same compass network as presented above (Figure 2B).

Taken together, dung beetles rely on the same neural network to stabilise their bearings as path-integrating and migratory insects do, with the central complex as the main cockpit for steering. For tenacious dung beetles that push their balls through highly challenging terrains at all hours of the day, a dynamic and straightforward snapshot compass system seems to be the most efficient way to roll as the crow flies. FURTHER READING Dacke, M., Baird, E., Byrne, M., Scholtz, C.H., and Warrant, E.J. (2013). Dung beetles use the Milky Way for orientation. Curr. Biol. 23, 298–300. Dacke, M., Byrne, M., Baird, E., Scholtz, C.H., and Warrant, E.J. (2011). How dim is dim? Precision of the celestial compass in moonlight and sunlight. Phil. Trans. R. Soc. Lond. B 366, 697–702. Dacke, M., Byrne, J.M., Smolka, J., Warrant, E.J., and Baird, E. (2013). Dung beetles ignore landmarks for straight-line orientation. J. Comp. Physiol. A 199, 17–23. Dacke, M., Nilsson, D.E., Scholtz, C.H., Byrne, M., and Warrant, E.J. (2003). Animal behaviour: insect orientation to polarized moonlight. Nature 424, 33. 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. USA 112, 11395–1140. el Jundi, B., Foster, J.J., Khaldy, L., Byrne, M.J., Dacke, M., and Baird, E. (2016). A snapshotbased strategy for celestial orientation. Curr. Biol. 26, 1456–1462. Foster, J.J., el Jundi, B., Smolka, J., Khaldy, L., Nilsson, D-E., Byrne M.J., and Dacke, M. (2017). Stellar performance - mechanisms underlying Milky Way orientation in dung beetles. Philos. Trans. R. Soc. B 372, 20160079. Immonen, E.-V., Dacke, M., Heinze, S., and el Jundi, B. (2017). Anatomical organization of the brain of a diurnal and a nocturnal dung beetle. J. Comp. Neurol. 525, 1879–1908. 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. Pfeiffer, K., and Homberg, U. (2014). Organization and functional roles of the central complex in the insect brain. Annu. Rev. Entomol. 59, 165–184. Seelig, J.D., and Jayaraman, V. (2015). Neural dynamics for landmark orientation and angular path integration. Nature 521, 186–191. Stone, T., Webb, B., Adden, A., Weddig, N.B., Honkanen, A., Templin, R., Wcislo, W., Scimeca, L., Warrant, E.J., and Heinze, S. (2017). An anatomically constrained model for path integration in the bee brain. Curr. Biol. 27, 3069–3085. Turner-Evans, D.B., and Jayaraman, V. (2016). The insect central complex. Curr. Biol. 26, R453–R457.

1

Lund Vision Group, Department of Biology, Lund University, 22362 Sweden. 2Emmy Noether Animal Navigation Group, Zoology II, Biocenter, University of Wuerzburg, 97074 Germany. E-mail: [email protected] (M.D.), [email protected] (B.e.J.)

Primer

Neuroethology of bat navigation Daria Genzel1,2, Yosef Yovel3,4, and Michael M. Yartsev1,2,* Once a year about 15 million Mexican free-tailed bats (Tadarida brasiliensis) migrate up to 1,500 kilometers from wintering grounds, seamlessly flying over the Mexican border to enter the United States. Their destination is the Bracken Cave in southern Texas, which will be their summer home between the months of March through October. While residing there, these bats emerge every night at dusk from the narrow 100-foot-wide opening of this enormous cave and begin their nightly commute to foraging grounds located up to 50 kilometers away. Upon arrival, they will spend the night hunting for insects in mid-air while providing a valuable service to local farmers by keeping crop pests in check. Close to the break of dawn, as the night of hunting comes to an end, these bats will begin making their trip back to the roost. How bats achieve these feats of navigation we are only beginning to understand. Since the discovery of position-coding neurons in the brains of rodents, considerable progress has been made in unraveling the neural mechanisms underlying how the brain guides navigation through complex environments, but much remains unknown. The study of bats with their amazing navigational abilities, coupled to recent technological advances that allow for cellular-resolution measurements and manipulation of brain activity during free flight, has begun to provide important insight into how the mammalian brain might support complex three-dimensional navigation. Yet, for such insight to be effective it is important to consider the fine details of navigation behaviors exhibited by different bats and remember that while all bats navigate while flying, not all navigate the same way. In this Primer, we will explore the diversity of navigational abilities exhibited by bats and discuss how considering this diversity can

Current Biology 28, R952–R1008, September 10, 2018 © 2018 Elsevier Ltd.

R997

Current Biology

Magazine Spatial working memory Prey detection

Frequency (kHz)

B A

Angle of sound incidence (°)

Prey localization

D 20 km 80kHz

C glints

Long-range visual navigation

20ms

Current Biology

Figure 1. Navigational challenges for different bat species. Illustration of the different functional challenges bat species face during diverse forms of navigation. (A) Prey detection: Noctilio leporinus successfully trawls for fish by identifying the ripples fish make as they disturb the water’s surface (photo credit Dietmar Nill). (B) Spatial working memory: Glossophaga soricina can successfully remember a large number of visited flowers in dense vegetation (photo credit Harry Colquhoun). Right: spectrogram showing an example of a distinct echo-acoustic signature of a flower that is open and ready for pollination. This species uses these acoustic signatures to facilitate foraging success for nectar (spectrogram image adapted from Simon et al. 2011, reprinted with permission from AAAS). (C) Long-range spatial navigation: Rousettus aegyptiacus is possibly guided through a visual spatial map for its many km of nightly commute. This is reflected in its very stereotypical flight routes from roost to feeding ground. (D) Prey detection: Rhinolophus bats successfully forage for moths by monitoring a narrow frequency band around their constant frequency echolocation calls (photo credit Frank Greenaway). The moth’s wing beats (illustrations on bottom right) are embedded in the echoes as frequency and amplitude modulations (glints) resulting in optimal target detection (data used for generating the spectrogram are from Koselj et al. 2011).

lead us to a better and more refined understanding of the brain mechanisms that can support navigation as a whole. A unique and diverse navigator Bats are the only mammal capable of self-propelled flight, a feature that enables them to freely explore their three-dimensional environment. Their body is optimally suited for an aerial lifestyle, allowing highly dynamic and adaptable flight maneuvering with speeds reaching up to 50 km/h and possibly even more. Furthermore, bats are renowned for their ability to extract information about the environment through their echolocation system (biosonar). About 90% of all bat species emit ultrasonic calls and utilize the echoes reflected from their surroundings to orient, navigate and capture prey in complete darkness. A common misconception, however, is that because all bats (>1,300 species) are capable of flight, they navigate in a similar fashion. Often R998

studies in one species have been generalized to all others, and thought to be sufficient to reveal the behavioral and neural mechanisms supporting three-dimensional aerial navigation, or navigation in general — an erroneous assumption. As one of the most diverse mammalian groups on our planet (one of every four mammalian species is a bat), these remarkable creatures inhabit diverse ecological niches, facing distinct navigational challenges and constraints therein. While some bat species only travel locally within hundreds of meters from their roosts, others commute over 100 kilometers to remote feeding grounds every single night. Some species of bats even migrate thousands of kilometers every year to reach better foraging regions or refuge sites (hibernacula). Furthermore, bats exploit a large miscellany of food sources ranging from fruit, nectar and pollen to various insects, arthropods, fish/other small vertebrates and even blood, the

Current Biology 28, R952–R1008, September 10, 2018

obtainment of which often requires very different navigation strategies (Figure 1). The sensory modalities used for navigation are also diverse between different species of bats, each of which faces different navigation challenges for obtaining its food. For example, species that navigate over large distances cannot rely on echolocation, because echolocation largely operates in the ultrasonic range and is highly affected by atmospheric attenuation, limiting its range to less than 100 meters for very large objects. Hence, such bats probably heavily rely on visual information and perhaps on additional modalities such as magnetic sensing and olfaction for long-range navigation. Conversely, bats that hunt for insects in dark confined spaces may have very poor visual acuity and thus predominantly rely on echolocation to orient. With echolocation being ideal for small scale resolution and vision being more reliable for distal landmarks, many bats probably also effectively utilize both, switching back and forth between these different sensory modalities to optimize their overall navigation capacities in specific behavioral contexts. Of course, additional inputs from other sensory systems, such as olfaction or touch, can play a role as well. Thus, it is clear that the large diversity of bats necessitates considering the many navigational constraints and challenges each specific species faces if we are to accurately decipher the behavioral mechanisms and neural codes that facilitate their navigational capacities. Even more so, a single bat can continuously switch between different navigational strategies depending on the precise navigational task it has to perform and the scale of that task (Figure 2). Thus, it is crucial to consider not only how species differ from one another, but also the specific challenges each species faces in particular situations. Here, we will address the challenges bats face when performing specific navigational tasks: first, we will consider the challenges encountered while orienting inside the roost and around it, then discuss the navigational strategies employed travelling from roost to foraging grounds (commuting) as well as longdistance migration. We will also discuss potential strategies that bats apply for

Current Biology

Magazine

Echo-acoustic route guidance

Path integration

Visual beaconing

Echo-acoustic landmark

Echo-acoustic flow field

Smell

Current Biology

Figure 2. Context-dependent navigation challenges in a single bat. Here we illustrate a theoretical timeline for one bat as it emerges from its cave roost and navigates to its final foraging site. The employed navigational strategy depends on the behavioral context and available sensory cues. Navigation in the roost is dependent on echo-acoustic information and route guidance. The path from the cave opening to the open space is easiest guided through path integration due to confounding echoes from the dense vegetation. A nearby mountain range serves as a visual beacon in open air and a group of conifers act as an echo-acoustic landmark. Echo-acoustic flow fields aid the navigational course along the hedge to an orchard, where olfaction guides foraging.

orienting at smaller scales, upon arrival at the feeding grounds or during prey capture. Finally, we will describe the recent progress made in understanding how the mammalian brain might support navigation in three-dimensional space. Navigation inside the roost Bat roosts differ from species to species and can range from small tree holes to huge caves. Some bats construct their own tent-like roosts by folding large leaves while others use human-made structures such as roof tiles or highway bridges. Roosts can vary seasonally, for example spring versus winter quarters, and can be inhabited by different social compositions, such as female and pup maternity colonies versus male colonies. Roost navigation therefore requires precise and varied strategies for success: for example, solitarydwelling males must determine whether a potential roost is occupied by a competitor; bats living in hierarchical groups must find their precise perch location within the roost as dictated by the social structure of the colony; females must precisely locate and retrieve nursing pups left behind while out foraging. A common feature of many roosts, however, is a visually dark environment, so navigation is unlikely to use visual cues alone, but rather to be carried out predominantly via echolocation and spatial memory. In fact, it has been suggested that lingual echolocation evolved only in one genus of old-world fruit-bats (Rousettus) in order to enable

orientation within dark roosts (other genera in this family either do not echolocate or use a primitive form of wing-clicking). The recent development of large laboratory flight tunnels has made it possible to study the neural mechanisms that might support such a form of roost navigation in this species. Finer-scale orientation can then be guided by acoustic or olfactory cues, as pups or conspecifics can be located (and possibly even identified) through communication calls or smell. Individuals could therefore encode absolute space or the location of other individuals with whom they are sharing a specific volume of the environment. This strategy could also apply to species that emerge from their roosts in swarms, where navigation might be aided by encoding not space per se but the location of each neighbor or object within a restricted volume (Figure 3F). Commuting Many bat species commute daily from the roost to their foraging grounds, with distances ranging from a few to many dozens or more than 100 kilometers. How do these bats reliably navigate to a specific location day after day? A feasible navigational strategy is beaconing: here, distal cues or landmarks are perceived from the starting location and usually are attended to from further distances, though this has not been shown conclusively for bats. Because beaconing is driven by long-distance perception, employing echolocation cues is made impractical by atmospheric attenuation of the sound

signals. Vision, on the other hand, is an ideal sensory modality for perceiving long-range cues and could be utilized by bats commuting in open airspace. Indeed, homing in the greater spearnosed bat (Phyllostomus hastatus), an omnivorous bat which is found in the tropical regions of South America, is suggested to rely on a prominent mountain range as a visual beacon. In large forested areas, visual cues are more difficult to perceive, but low frequency (<20 kHz) acoustic cues, which are less affected by attenuation, could act as local beacons in these visually crowded environments. The big brown bat, Eptesicus fuscus, and the fringe-lipped bat, Trachops cirrhosus, indeed seem to navigate towards the distant mating calls of frogs, but for different reasons: E. fuscus follows the calls to reliably locate ponds where it then forages for insects using echolocation, whereas T. cirrhosus uses frog calls to detect the frogs which it then attacks while relying on both passive hearing and echolocation. The latter species even possesses specialized anatomical features to enhance the perception of such lowfrequency (<5 kHz) sound signals. A second navigational strategy that might support longer range navigation is route guidance: here, familiar trajectories are followed to distinct decision points recognized through landmarks. Like beaconing, route guidance has not yet been indisputably demonstrated in bats. Some myotis and pipistrelle species that commute to their foraging grounds to hunt for insects in flight have been suggested to navigate

Current Biology 28, R952–R1008, September 10, 2018

R999

Current Biology

Magazine D

F

Firing rate (Hz)

A

4 Place field Switch

0 0 Pat h di

180 0 (°) stan –12 –180 direction l ce ( a o G m)

0°,+135°

Relevant volume

Tadarida brasiliensis exiting bracken cave

0°,+45°

B

135°,0°

Tim e

E

0°

Heading direction

0°,0°

90°,0°

Place field

Place field

Magnetic field

0°,–45°

Place field

G C

Firing rate

Place fields

-90

-45

0

45

90

Deviation to magnetic orientation (°)

Rousettus aegyptiacus: roost navigation/commute Pteronotus parnellii: commute Rhinopoma microphyllum: foraging Plecotus auritus: foraging

Current Biology

Figure 3. Observed and hypothesized neural codes for spatial navigation in bats. (A–D) Observed spatial coding in hippocampal formation of the Egyptian fruit bat (Rousettus aegyptiacus). (A) Two-dimensional grid and border cells (left and right, respectively; image adapted from Yartsev et al. 2011, reprinted by permission from Springer Nature). (B) Three-dimensional head directional coding (image adapted from Finkelstein et al. 2015, reprinted by permission from Springer Nature). (C) Three-dimensional place cells (image adapted from Yartsev and Ulanovsky 2013). (D) Vectorial representation of goal location (image adapted from Sarel et al. 2017, reprinted with permission from AAAS). (E–G) Hypothesized spatial coding that has yet to be demonstrated. (E) Neurons with firing field representing deviations from the Earth’s magnetic field could be integrated with heading direction coding to guide long-range navigation along the gradient. (F) Many bat species navigate in groups through complex environments and often need to constrain their navigation based on the location of proximal bats or objects. For example, during roost emergence bats flying in swarms (for example Tadarida brasiliensis exiting the Bracken Cave; photo credit USFWS/Ann Froschauer) may not encode space per se, but rather vectorial distances of nearby neighbors or objects within a restricted volume. Such constraints may necessitate flexible representation of the volume around them as conspecifics and objects are replaced. This is indicated with a ‘place field switch’ when one individual moves out of the volume and a different individual moves into this space and is compatible with recent observations of individuals and, even more strongly, of objects within a stable space (Omer et al. 2018). (G) Different scales of spatial coding may exist for different bat species and might depend on the navigational strategies and sensory information that guides the navigation. Smaller and larger scaled place cells are depicted for the Egyptian fruit bat during two different behavioral tasks (roost and long-distance navigation). Other species of bats likely exhibit different scales of spatial coding due to the different challenges they face during navigation.

along stereotypical flight paths (flyways) which are anchored to familiar linear structures (houses, bridges, hedges or groves) while ignoring possible shorter routes. This does not mean, however, that the bats are flying ‘blind’. On route, bats most likely continuously acquire and compare sensory inputs (echoacoustic or visual flow fields along these linear structures) to generate a spatial representation of their flight routes, allowing them to adjust their trajectories accordingly. Greater horseshoe bats (Rhinolophus ferrumequinum) and mustached bats (Pteronotus parnellii), for example, rapidly adapt their flight and echolocation behavior when novel objects are introduced near to or in their commuting flight paths, suggesting that they are constantly accumulating sensory information. R1000

These species are found in Europe and South America, respectively, and hunt for moths with distinct constant frequency echolocation calls. The moths are detected as unique signatures of amplitude and frequency modulations, so called ‘glints’, embedded in the echoes. These result from the upward and downward wing beats of the flying moths (Figure 1D). In some cases though, such as during heavy fog, bats may have to overcome sparse or even a lack of acoustic or visual information and therefore navigate segments of their path in another way. A potential way of doing this is by path integration, which solely depends on internally-generated movement cues that correspond to a specific length of distance traveled and is independent of external landmarks

Current Biology 28, R952–R1008, September 10, 2018

(or even independent of any external sensory input in the extreme case). Because of the variability inherent in muscle activity from one repetition to the next, distance ‘calculation’ errors may accumulate quickly without spatial recalibration to external cues at regular intervals. Bats can monitor their speed and use that as a more reliable measure for determining the distance traveled. Studies of the insect hunting bat E. fuscus and the short-tailed fruit bat (Carollia perspicillata) have shown that the wing membranes of bats are covered with miniscule subcortical sensory hairs which are highly sensitive to air flow and could be used to estimate flight speeds. These bats could apply path integration on their internally measured speed, and recent results for Kuhl’s pipistrelles (Pipistrellus kuhlii), an

Current Biology

Magazine insect-hunting bat that is characterized by its poor vision, seem to confirm the existence of path integration for distances up to 70 meters. The reliability of path integration based on an internal speed measurement only is questionable though as external wind would confound velocity estimations. Future studies, where possibly the bats’ flight speed is manipulated, should be carried out to corroborate these findings. A major open question is whether some bat species possess a cognitive map. Such a memorized map can be based on spatially distinct landmarks and it could enable the navigator to move through the mapped area using triangulation to navigate in novel trajectories. There is some evidence suggesting the existence of a cognitive map from studies of the Egyptian fruit bat, Rousettus aegyptiacus. This bat flies many dozens of kilometers from roost to specific fruit trees along very stereotypical flight routes (Figure 1C). Displacement experiments suggest that a visual cognitive map, defined by conspicuous distal landmarks, enables these bats to navigate home from an unfamiliar terrain (dozens of kilometers away from their home-range). Migration Some bats, particularly those living in the temperate regions of the world, perform annual long-distance migration flights. These seasonal migrations range from a few hundred kilometers to 3,000– 4,000 kilometers per leg for the yearly trip from summer breeding grounds to the winter habitat and back. How these species successfully navigate such long distances back and forth every year is still completely unclear, as until recently the bats’ smaller size made it almost impossible to track them on such long routes. There are abundant migratory studies for birds (though not always with conclusive results), and we can assume the same hypotheses debated in the bird literature are also relevant for bats. For example, migration could be achieved by route following in which each segment in the route is very long, on the order of 100 kilometers, and is completed using beaconing. A further strategy could be gradient following: olfactory gradients could allow successful navigation from one location to another. Differently oriented

ratio gradients of volatile hydrocarbons (for example, butane, benzene and isoprene) do in fact exist and are fairly resistant to, though not independent of, wind direction changes. Thus, they could be implemented as a directional reference scale, but no work in this domain exists for bats. There is evidence that bats might use magnetic information for homing. The Chinese noctule bat (Nyctalus plancyi), which has a habitat distributed over a large area, has a high sensitivity to magnetic fields and appears to align itself to the magnetic north. E. fuscus and greater mouse-eared bat (Myotis myotis) seem to calibrate a magnetic compass through polarized patterns of sunlight during sunset or sunrise. As M. myotis commonly navigate over open areas and travel up to 100 kilometers between summer and winter roosts, navigation could be theoretically carried out by such a global gradient, independent of any local landmarks (Figure 3E). Foraging In general, navigational strategies employed during foraging are the same as discussed for long range navigation, but on a smaller scale. Some bats use beaconing, like an insectivorous bat that heads toward a street light where insects are available or a nectar feeding bat that uses the acoustic beacon evolved by the Mucuna plant, the leaves of which act like a satellite dish and conspicuously reflect a bat’s echolocation signal largely independent of the ensonification angle (Figure 1B). Other bats navigate via route guidance from one potential foraging site to another (as seen for fruit bats examining different trees). Naturally, a bat’s foraging behavior depends on the spatiotemporal pattern of its food distribution. Food sources for nectarivorous and frugivorous bats are typically stationary and available over multiple nights, so these bats can rely on spatial memory for guidance. Studies investigating the foraging behaviors of both the nectar feeding Glossophaga soricina, which visits 300–1200 flowers per night (Figure 1B), and the fruit-eating C. perspicillata, which is the main seed dispersalist of the Pepper elder, Piper amalago, show that their search strategy is dominated by spatial

memory rather than cue-based search. Quality control — choosing the ripest fruit or best flower — is aided by either olfaction (C. perspicillata have fairly low thresholds for fruit odors through which they detect the fruit) or echolocation (some nectar-producing flowers exhibit a very distinct echo-acoustic signature when open and ready for pollination). As nectar sources are highly timevariant due to fast depletion rates, nectarivorous bats would additionally benefit from memorizing previous visits. G. soricina indeed displays a highly effective spatial and time-dependent working memory by holding more than 40 past behavioral actions (feeder visits) without indication of decay — a memory feat not shown for any other bat species, or even more generally any other taxa, to date. In contrast to frugivorous and nectarivorous bats, predatory bats feeding on insects or small vertebrates must constantly evaluate their environment to detect highly spatiallyvariable and often unpredictable prey. Many species know where to find insects, such as the Pipistrellus species that hunt under street lights where insects tend to accumulate, and thereby restrict their foraging area. Others move through the foraging area in a pseudoBrownian motion in constant search with the aim of optimizing their foraging success. Such is the case of the greater mouse-tailed bat (Rhinopoma microphyllum), which spends the entire night foraging in flight with many directional switches, travelling up to 90 kilometers yet still able to keep track of its spatial position and reliably return to its roost. While clearly it is crucial for the bat to attend to its own location with respect to the environment, it is important to consider that predatory foraging does not occur in isolation. Rather, this process is often described as a dynamic interplay between the predator and its prey which necessitates keeping track of the spatial position of the prey — as indeed many bats do. For example, the brown long-eared bat (Plecotus auritus) and the false vampire bat (Megaderma lyra) remain perched slightly elevated from the ground and have specialized in listening for rustling sounds generated by insects or smaller vertebrates moving in the vegetation. These are then gleaned by actively echolocating

Current Biology 28, R952–R1008, September 10, 2018

R1001

Current Biology

Magazine the prey’s precise location utilizing frequency modulated echolocation calls which highly resolve space. Another example is the common vampire bat (Desmodus rotundus), which most often feeds on the blood of a single individual prey animal every night, possibly by listening for its prey’s breathing sounds. Prey identification might then become a matter of distinguishing the prey’s unique sound signature, much as we can identify specific people by their voices. Bulldog bats (Noctilio leporinus) and Daubenton’s bats (Myotis daubentonii) also rely on prey-generated signals, but in a different manner: these bats hunt above water and echo-acoustically detect their prey (fish and insects, respectively) by the ripples the prey makes as it disturbs the water’s surface (Figure 1A). In general, different bat species deal with different navigational challenges during foraging. Some bats, for example aerial hawkers, need to continuously update their position in reference to their home roost in spite of many directional switches and longer travel distances, whereas others, such as fruit or nectar bats, need to memorize multiple foraging sites. Other species employ different types of information and modalities to detect background objects while searching for their prey, for example bats hunting near vegetation, and most bats need to switch between modalities to optimize not only their foraging success but their navigation in its entirety. Spatial representation Different navigational strategies require different spatial representations. For some strategies — beaconing, route guidance and cognitive map navigation — landmarks must be correctly identified and the correct mental map must be consulted, sometimes, on dramatically different scales (for example, commuting versus foraging). We also expect to find differences in space representation among commuting bats, which apply the same navigation strategy, because some navigate over hundreds of kilometers, while others do so over only a few kilometers. Route guidance relies on exact turns at distinct landmarks, whereas in cognitive map navigation, bats must identify landmarks and R1002

resolve the spatial relationships between them. Both strategies, however, require a representation of distinct landmarks or specific spatial locations. The resolution and accuracy of spatial representation will always be limited by the sensory modality used: for example, the location of an insect detected via echolocation is encoded with spatial accuracy far better than the location of a fruit detected through olfactory cues. Path integration and gradient following, where decisions might be based on measured deviations (of gradient orientation or motion), could be represented in a vectorial manner, rather than in distinct spatial locations as landmarks are for cognitive map navigation, although these strategies too must represent some locations such as the starting point (such as the roost) and the goal. Potential food sources could be further pinpointed through association with spatial locations and stored in spatial memory or identified via comparison of characteristic features or patterns with memorized signatures (such as the echo-acoustic signature of a flower), allowing for target categorization. Spatial representations should therefore differ between different types of navigators (Figure 3G) and can sometimes differ even within an individual bat during different behavioral contexts (Figure 2). To illustrate this point, let us consider the following scenario (Figure 2): a bat emerges from the tunnels of its cave roost after flying several hundreds of meters in complete darkness using route guidance cued through echolocation. It then integrates its path for a few meters from the mouth of the cave to open air. A nearby mountain range serves as a visual beacon and it flies towards it for 10 kilometers, until it echo-acoustically recognizes a group of conifers beside a lake. Here, it turns left to fly along a hedge towards an orchard, where it hovers around one of the apple trees, navigating towards the ripest fruit by smell. On a single night, this bat switched between distinct challenges and navigational strategies several times, and its nervous system must be equipped to handle all different strategies — underscoring the importance of understanding the precise navigation behavior the bat is engaged in at each specific moment in time. This leads us to the

Current Biology 28, R952–R1008, September 10, 2018

obvious question: what features of the mammalian nervous system support these different forms of navigation? Neural representation Neural investigations in freely behaving and flying bats have thus far utilized electrophysiological recording methods to study the firing pattern of neurons in the hippocampal formation (Figure 3). This work has begun to reveal the similarities and, importantly, the differences between mammals navigating in two-dimensional versus three-dimensional environments. Research in bats navigating in small laboratory two-dimensional environments revealed that bats have cell types with distinct spatial receptive fields, similar to those that have been previously reported in rodents, such as place, grid, border and head-direction cells (Figure 3A). Yet, in striking contrast to rodents, though similar to primates, it was found that the firing patterns of hippocampal neurons in bats persist in the absence of the theta oscillations that were believed to play an important role in generating these hippocampal firing patterns in rodents, thus underscoring the importance of the comparative approach. Later studies have leveraged the development of wireless neural recording methods in freely flying bats to begin describing the firing properties of these spatial neurons in threedimensional space. These recordings revealed that three-dimensional place-cells in flying bats appear to encode three-dimensional space isotropically, as the spatial activity fields of these neurons were equally tuned to all orthogonal dimensions of the environment, a required feature for an animal that navigates in threedimensional space (Figure 3C). This necessity was also revealed in the threedimensional encoding properties of the bats’ head direction cells which respond not only to directionality in the horizontal or azimuthal plane, as reported previously in rodents, but also to pitch, roll and yaw. This results in a toroidal (doughnut-shaped) representation which appears ideally suited to guide movements in flight (Figure 3B). Studies in rodents have revealed that the tuning and firing fields of almost all spatial cells in the hippocampal formation are dependent on many

Current Biology

Magazine environmental and internal parameters, such as available cues, attention, motivational state, novelty and the behavioral context the subject is engaged in. It is hypothesized that this allows for a dynamic computation of space in correlation with behavioral context and sensory modality providing spatial information. Similar modulation of spatial firing fields has been reported in the flying bat with particular dependency on navigation goals and sensory modalities, again compatible with the natural complexity of the bat’s natural navigation conditions for which its nervous system has evolved. Most previous neural studies focused on the encoding of the animal’s location and direction within the environment, while the encoding of the location and direction of points of interest in ego-centric coordinates has not been sufficiently examined, although it is clear that such a representation is essential for navigation. Recently, it was reported that a subset of hippocampal neurons in the Egyptian fruit-bat (R. aegyptiacus) possess a vectorial representation of a stationary target (Figure 3D), an observation that is compatible with the natural behavior of this species that follows extremely stereotyped spatial navigation patterns every night. While these studies have provided valuable insight, all neural investigations in freely flying bats to date have focused on a single brain region (the hippocampal formation), in very few bat species using a single method (electrophysiology). As we will address next, future studies utilizing a broader range of methodologies applied to a broader range of species and brain regions will be transformative in deciphering the neural computations in the mammalian brain that facilitate the full complexity of navigation. Outstanding questions As the field of navigation, especially bat navigation, is still filled with unproven hypotheses and riddles, the following questions include several major open research directions that we think need to be investigated and followed up on. What are the mechanisms guiding migration? With miniature GPS techniques now available, studies need to be conducted to finally describe migration in bats in detail and to

investigate the underlying mechanisms. This includes the sensory input guiding migration and the sensorimotor strategy then translating sensory input into movement. These questions are also relevant for all smaller scale types of navigation, for example commuting, described above. How do bats learn to navigate? Data have been accumulating revealing the different types of navigational skills bats engage in. But how these skills are acquired remains unknown. Are they innate or learned through following and copying conspecifics? How do bats navigate socially? A further important aspect of an animal’s spatial behavior which should not be neglected is social space. Bats are highly social animals, sometimes living in colonies numbering in the hundreds of thousands to millions of individuals. They can live up to approximately 40 years and are likely highly dependent on social learning for their fitness. It is plausible that bats encode social space. Neural recording in hippocampal circuits in R. aegyptiacus has shown that cells fire in relation to the spatial positioning of both objects and conspecifics. Yet to distinguish between responses to an item (be it a conspecific or inanimate object) it would be interesting to test whether such coding is also dependent on the conspecific’s behavior and the information that needs to be extracted from it, for example in species showing social learning such as the Honduran tent-making bat (Uroderma bilobatum). Which neural computations are necessary for bat navigation and for what specific aspects of it? To date, it still remains to be determined whether the spatial signals reported in the flying bat are necessary for supporting its navigation abilities and, if so, then which specific aspects of the behavior are these neural signals required for. Importantly, we also need to work out what happens when an individual switches from one navigation strategy to another. Addressing these fundamental questions can be approached via causal manipulation experiments and more sophisticated behavioral designs that have yet to be performed in freely navigating bats. What signals in the bat nervous system support different aspects of

navigation and in what neural circuits are these embedded? Research in other model organisms has taught us that the ability to navigate depends on a very broad range of computations that are distributed throughout the brain. For example, while the hippocampal formation is predominantly tied to world-centered (allocentric) spatial representations, the representation of position in body-centered coordinates (egocentric) is essential for efficient navigation and is believed to be represented in structures such as the inferior colliculus where echoacoustic flow-field information has been reported to be encoded in several bat species (E. fuscus, C. perspicillata, Phyllostomus discolor). Studies in other species have revealed that higher-level cognitive processing centers such as the orbitofrontal and parahippocampal brain regions (such as the retrosplenial cortex) contain and process information about spatial planning, landmark analysis and goal values and thus are quite likely to be key components in the neural computations that facilitate the bat’s navigation behavior. These and many other brain structures (both cortically and subcortically) have never been explored in the bat brain and stepping outside of the hippocampal formation will be indispensable if we are to understand the complexity of bat navigation in particular and of animal navigation in general. How do neural signals in different species of bats differ? As mentioned above most of the aforementioned neurobiological studies have been conducted in one species, R. aegyptiacus. Taking into account the tremendous differences in lifestyle and navigational needs between bat species, it would not be surprising that spatial representations in the brain differ as well. Thus, to enrich our understanding of how the brain encodes navigation in threedimensional space, a broader set of studies harnessing the diversity of bat species and the specific navigational challenges each has solved would be of great benefit. On the other hand, it is exactly these variations that make bats such an ideal model organism. They afford us the unique opportunity to investigate and compare varied and distinct navigational strategies within the same animal order.

Current Biology 28, R952–R1008, September 10, 2018

R1003

Current Biology

Magazine Aharon, G., Sadot, M., and Yovel, Y. (2017). Bats use path integration rather than acoustic flow to assess flight distance along flyways. Curr. Biol. 27, 3650–3657. Barchi, J.R., Knowles, J.M., and Simmons, J.A. (2013). Spatial memory and stereotypy of flight paths by big brown bats in cluttered surroundings. J. Exp. Biol. 216, 1053–1063. Finkelstein, A., Derdikman, D., Rubin, A., Foerster, J.N., Las, L., and Ulanovsky, N. (2015). Three-dimensional head-direction coding in the bat brain. Nature 517, 159–165. Genzel, D., Geberl, C., Dera, T., and Wiegrebe, L. (2012). Coordination of bat sonar activity and flight for the exploration of three-dimensional objects. J. Exp. Biol. 215, 2226–2235. Genzel, D., Hoffmann, S., Prosch, S., Firzlaff, U., and Wiegrebe, L. (2015). Biosonar navigation above water II: exploiting mirror images. J. Neurophysiol. 113, 1146–1155. Hulgard, K., and Ratcliffe, J.M. (2014). Niche-specific cognitive strategies: object memory interferes with spatial memory in the predatory bat Myotis nattereri. J. Exp. Biol. 217, 3293–3300. Jensen, M.E., Moss, C.F., and Surlykke, A. (2005). Echolocating bats can use acoustic landmarks for spatial orientation. J. Exp. Biol. 208, 4399–4410. Koselj, K., Schnitzler, H.U., and Siemers, B.M. (2011). Horseshoe bats make adaptive prey selection decisions, informed by echo cues. Proc. R. Soc. Lond. Ser. B 278, 3034–3041. Omer, D.B., Maimon, S.R., Las, L., and Ulanovsky, N. (2018). Social place-cells in the bat hippocampus. Science 359, 218–224. Sarel, A., Finkelstein, A., Las, L., and Ulanovsky, N. (2017). Vectorial representation of spatial goals in the hippocampus of bats. Science 355, 176–180. Schnitzler, H.U., Moss, C.F., and Denzinger, A. (2003). From spatial orientation to food acquisition in echolocating bats. Trends Ecol. Evol. 18, 386–394. Simon, R., Holderied, M.W., Koch, C.U., and Von Helversen, O. (2011). Floral acoustics: Conspicuous echoes of a dish-shaped leaf attract bat pollinators. Science 333, 631–633. Tsoar, A., Nathan, R., Bartan, Y., Vyssotski, A., Dell’Omo, G., and Ulanovsky, N. (2011). Largescale navigational map in a mammal. Proc. Natl. Acad. Sci. USA 108, E718–E724. Voigt, C.C., Frick, W.F., Holderied, M.W., Holland, R., Kerth, G., Mello, M.A.R., Plowright, R.K., Swartz, S., and Yovel, Y. (2017). Principles and patterns of bat movements: from aerodynamics to ecology. Q. Rev. Biol. 92, 267–287. Winter, Y., and Stich, K.P. (2005). Foraging in a complex naturalistic environment: capacity of spatial working memory in flower bats. J. Exp. Biol. 208, 539–548. Yartsev, M.M. (2017). The emperor’s new wardrobe: Rebalancing diversity of animal models in neuroscience research. Science 358, 466–469. Yartsev, M.M., and Ulanovsky, N. (2013). Representation of three-dimensional space in the hippocampus of flying bats. Science 340, 367–372. 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. 1 Department of Bioengineering, UC Berkeley, Berkeley, CA 94708, United States. 2Helen Wills Neuroscience Institute, UC Berkeley, Berkeley, CA 94708, United States. 3Department of Zoology, George S. Wise Faculty of Life Sciences, Tel Aviv University, Tel Aviv 69978, Israel. 4Sagol School of Neuroscience, Tel Aviv University, Tel Aviv 69978, Israel. *E-mail: [email protected]

R1004

Primer

Tolman’s experiment challenged the view that stimulus–response associations could explain all animal behaviour, and thus was an early harbinger of the cognitive revolution. The cognitive map idea has continued to be prominent in neuroscience research, associated with the Nobelworthy discovery of neurons that code location (place cells and grid cells), as well as head direction cells, border cells and boundary cells. Recent neuroimaging work suggests that, as regards spatial cognition, the human brain is substantially similar to the rodent brain. But do we (and other species) really form ‘cognitive maps’? There are many criticisms of the idea, based on findings of various kinds of imprecision and irrationality. For example, human participants seem to remain unaware of distortions in a virtual environment that includes wormholes that transport them to a different section of a maze. They do not become lost or disoriented, and in fact are perfectly happy to take the wormhole ‘shortcut’ (Warren et al., 2017). If the gaps between wormholes are not represented, the cognitive map does not seem very Euclidean. The alternative is that human spatial representations may be like a labeled graph — a mathematical structure in which

Individual variation in human navigation Nora S. Newcombe The cognitive map view of navigation posits that humans and other species represent space in a format that encodes the distances and directions among locations in relation to external reference frames such as boundaries and landmarks. Such an allocentric representation is combined with egocentric representations that track a navigator’s position over movement. Jointly, these systems enable the flexible planning of routes, including shortcuts, detours, and paths between two locations not formerly travelled. Tolman’s (1948) work on the sunburst maze initiated this tradition. In the sunburst maze, rats who had followed a circuitous route to find a reward, as shown at the left of Figure 1, could take a shortcut to the goal when their route was blocked and other routes were offered, as shown at the right of Figure 1. That is, to use one set of terms, they were place learners as well as response learners, or to use another set of terms, they formed survey representations as well as route representations.

10

9

11

G

F

8

12

7

Block

FURTHER READING

6 5

D E

4 13 14

C

15

3

16 17

2

18 1

B

A Apparatus used in preliminary training

A Apparatus used in the test trial

Figure 1. Setup for Tolman’s (1948) shortcut experiment. The rats initially traverse a route with several turns to get to a reward and were later offered a sunburst array of routes. They mostly chose the route that would take them directly to the reward, or a route not too far off. Reprinted with permission of the APA from Tolman et al. Studies in spatial learning. I. Orientation and short-cut. J. Exp. Psychol., 1946, 36, p.17.

Current Biology 28, R952–R1008, September 10, 2018 © 2018 Elsevier Ltd.