Cognitive maps and the magnetic sense in vertebrates

Navigation requires a network of neurons processing inputs from internally generated cues and external landmarks. Most studies on the neuronal basis of navigation in vertebrates have focused on rats and mice and the canonical senses vision, hearing, olfaction, and somatosensation. Some animals have evolved the ability to sense the Earth ’ s magnetic field and use it for orientation. It can be expected that in these animals magnetic cues are integrated with other sensory cues in the cognitive map. We provide an overview of the behavioral evidence and brain regions involved in magnetic sensing in support of this idea, hoping that this will guide future experiments.


Introduction
Spatial navigation depends on a complex underlying network of neurons that encode a map of the environment.Initially hypothesized by Tolman in 1948 [1], the first direct evidence of a cognitive map came in 1971, with the discovery of neurons in the brains of rats that increased their firing rate when the animal visited a specific location in the environment, hence, termed place cells [2].Many other spatial cell types were discovered in the following years, including grid cells that fire at regularly spaced distances, head direction cells that fire at specific head orientation angles, and border cells that fire at the edges of the environment [3].The ensemble activity of spatially tuned neurons represents the neural correlate of a map of the external space: a cognitive map.
Forming and maintaining a cognitive map requires sensory and motor cue inputs.Internal self-motion cues (idiothetic cues) like vestibular signals, sensory flow, proprioception, and motor efference copies keep track of travelled distances and directions.Over time, however, these accumulate errors, which are minimized by consulting external environmental cues (allothetic cues) [4].The predominantly studied organisms for cognitive maps are rats and mice, which effectively navigate with external cues perceived by the canonical senses -vision, olfaction, hearing, and somatosensationused flexibly depending on availability and task requirements (Figure 1).However, within the diverse animal kingdom some species have evolved sensory specializations to navigate in their unique habitats, such as the ability to detect the Earth's magnetic field [5].Do magnetic cues integrate into the cognitive map in these animals?
The geomagnetic field The Earth's magnetic field provides an omnipresent set of allothetic cues to any navigator able to sense it (Figure 2): The polarity (direction), inclination (tilt), declination (angle between magnetic and geographic north) and intensity (density) of the field lines.In addition to directional (compass) information, inclination, declination and intensity gradients potentially provide positional (map) information [6,7].Moreover, magnetic anomalies created by magnetic minerals in the Earth's crust could serve as local landmarks [7].

The magnetic compass sense
Many organisms use the direction of the Earth's magnetic field for orientation [8].An intensively studied model organism is the European robin, a long-distance migratory bird that uses a magnetic inclination compass on its biannual migrations [9,10].Fishes, amphibians, and turtles also possess a magnetic compass sense.Even among mammals a magnetic compass has been demonstrated in species that inhabit aphotic habitats, such as bats and subterranean mole-rats [11].Mole-rats use their compass to navigate through mazes of tunnels [12], bats to return to their home roost after nightly foraging bouts [13], and presumably for long-distance migration [14].

The magnetic map sense
In addition to the magnetic compass, evidence from long-distance migrants supports the use of geomagnetic cues for true navigation, i.e. to determine position based on magnetic parameters [6,7].The global gradients of geomagnetic field intensity, inclination and declination often form grids that provide information for a bicoordinate map.Virtual displacement experiments in which animals displayed compensatory reorientation behavior after sole exposure to the magnetic field parameters of a remote location (without being physically displaced) provide the strongest evidence.For example, reed warblers exposed to the magnetic field parameters of a location w2700 km northeast from their southern migratory route displayed a corresponding southwest deflection of their directional preference in a laboratory orientation assay [15].Similarly, hatchling green turtles use magnetic signatures to identify critical locations along their migratory route, where they must change their swimming direction to stay in the Atlantic gyre and prevent drifting into colder waters [7].Migratory fishes such as eel and salmon also possess a magnetic map sense [16,17].It is important to note, however, that the geomagnetic field gradients are very shallow, rendering a magnetic map most feasible for long-distance migrants.Intensity and inclination change w3e5 nTand w0.009 per km, respectively, corresponding to a relative change of 0.01% per km [18,19].Absolute sensitivities and discrimination thresholds of the magnetic sense are not well studied, but published thresholds for the inclination compass of 2e5 for birds [19] and 0.5 for newts [18] suggest that the use of magnetic maps over distances <50e200 km is unlikely (others have suggested a lower limit of >10 km [6]).The lower threshold might be determined by receptor properties [20] or background noise, such as the daily fluctuations of the geomagnetic field, which are regularly of the same magnitude as the field would change over a distance of 10 km.Indeed, even in homing pigeons which home from distances of up to several hundreds of kilometers, there is no strong evidence for the use of a magnetic map sense [21].Similarly, magnetic manipulations did not impede homing success in shearwaters that were displaced 400 km [22].Accordingly, geomagnetic gradients are likely not helpful for animals orienting over short distances (<2 km), such as rodents.However, local magnetic anomalies caused by magnetic minerals in the Earth's crust could steepen gradients and serve as landmarks [7].

Sensory basis of the magnetic sense
The cellular and molecular basis of the magnetic sense remains obscure [23], but several physically plausible sensory mechanisms have been suggested [24].The radical pair hypothesis describes a quantum mechanism based on the influence of weak magnetic fields on the recombination of radicals [25].There is evidence for magnetic field effects on radicals produced in photoexcited cryptochromes in the retina [26].A second proposed mechanism is based on magnetic nanocrystals (e.g.magnetite), but such particles are still awaiting their discovery in a sensory cell of any vertebrate [23].Finally, recent evidence from pigeons supports a compass mechanism based on electromagnetic induction in the semicircular canals of the inner ear [27].It is conceivable that an animal species possesses two magnetic sensors tuned to fulfil different functions (e.g.compass þ map).The geomagnetic field as an allocentric cue for navigation.a. Magnetic cues from the Earth's magnetic field can be used to derive directional information by using polarity, inclination or declination angle (angle between magnetic north' mN and geographic north' gN) or positional information by using declination changes or the gradients of inclination angle or intensity of the field.b.Example of a bat migrating a distance of 2,224 km in southwest direction from Pape, Latvia to Pitillas, Spain [14].The bat could either follow a magnetic direction over a specified distance or follow magnetic field gradients until the parameter of the local magnetic field matches the desired location (which the animal remembered or interpolated).

Brain regions involved in the magnetic sense
Our understanding of the neuronal circuits responsible for processing magnetic information remains incomplete.Various species appear to engage different circuits for this task, although direct neurophysiological recordings supporting this assertion are scarce [28].Instead, in many experiments animals were exposed to magnetic stimuli and immediate early genes (IEGs such as c-fos, ZENK/EGR-1, pERK) served as proxies for neuronal activity [29].The IEG expression was compared to sham-exposed control animals to identify activated brain regions.Stimulus-response properties of magnetosensory systems are unknown, thus often complex changing magnetic fields served as stimuli to maximize the likelihood of activating the receptors.Based on such studies, brainstem trigeminal and/or vestibular nuclei are likely primary processing centers for magnetic information in rodents, birds, and fishes [30e32].From the primary brainstem areas, magnetic information distributes across many brain areas, including the thalamus, superior colliculus, hippocampal formation, and cortical/hyperpallial areas.A separate thalamofugal pathway from the retina to a forebrain region termed Cluster N appears to process magnetic compass information coming from the retina in nightmigratory birds [33].Cluster N densely connects with the hippocampal formation [34], a key brain region for spatial navigation and cognitive maps.

Brain regions involved in navigation (cognitive maps)
The formation of a cognitive map requires integration of multiple sensory cues, both internally generated and external landmarks, distributed across several brain regions [35,36].Information from these sensory systems is first encoded in an egocentric reference frame, i.e. with respect to the animal's position and orientation in space.This is then transformed into allocentric information to form a coherent cognitive map of the external space, independent of the animal's head and body position [37].
Upstream brain regions like the medial vestibular nucleus, nucleus prepositus, dorsal tegmental nucleus, lateral habenula and the lateral mammillary nucleus encode angular head velocity using inputs from selfmotion cues [38,39].This idiothetic information on speed and direction of head turning is integrated with allothetic sensory cues, leading to the directional firing of head direction cells that are found in many brain regions, including the dorsal tegmental nucleus, lateral mammillary nuclei, anterodorsal thalamic nucleus, laterodorsal thalamic nucleus, postsubiculum, postrhinal-retrosplenial-, and medial entorhinal cortex (MEC) [39].Edges of the environment are encoded in border cells, also called boundary vector cells, in the subiculum, postrhinal cortex, retrosplenial cortex and MEC [40e42].Moreover, object vector cells in the MEC encode information on the distance to objects in the environment [43].Cells that encode the speed of movement are also found in the MEC [44].The combination of movement speed, distance and direction leads to the formation of multiple hexagonally arranged periodically firing fields characteristic of grid cells in the MEC [44] and the pre-and parasubiculum [45].These cells likely provide metrics for navigation by laying a grid on the external space which acts as a reference for measuring distances.Ultimately, inputs from this distance metric together with spatial information from other cell types feed into place cells in the hippocampus CA1, CA3, and dentate gyrus [46,47].Apart from the animal's location in space, the hippocampus also encodes the position of objects in landmark cells and their sensory features in object recognition cells [48].Moreover, spatially selective neurons resembling place cells were also discovered outside the hippocampus, e.g. in the visual cortex [49], somatosensory cortex [50], the subiculum [47], and superior colliculus [51], suggesting a wide distribution of spatial signals in the brain with the hippocampus playing the role of a central hub [52].

Possible integration of magnetic information into the cognitive map
Animals with a magnetic sense likely use directional (compass) and perhaps positional (map) information of the geomagnetic field together with other sensory inputs when forming a cognitive map.It is conceivable that they possess neurons that respond to specific values or combinations of magnetic field parameters, but where could these neurons be located?
In birds and mammals, several brain regions have been identified to exhibit both spatially firing cells and responses to magnetic fields (Figure 3).An early step in the formation of a cognitive map requires the integration of head position and movements using input from brainstem vestibular nuclei [53].Interestingly, in head-fixed pigeons these nuclei are also activated by magnetic field rotations [27,54] and electrophysiological recordings in head-fixed pigeons revealed vestibular brainstem neurons tuned to specific magnetic field directions [55].Interestingly, the cells appeared to be tuned to the intensity of the Earth's magnetic field, as their response increased from 20 mT to 50 mT, but did not increase further at 100 mT or 150 mT [55].These results are consistent with a population of magnetic head direction cells that transmit allocentric directional information to downstream spatial brain areas.Where and how this information is integrated is currently unknown.
A study in mole-rats reported a reduction in c-fos expression in the anterodorsal (AD) and laterodorsal (LD) thalamic nuclei, the dorsal tegmental nucleus, the postsubiculum, the retrosplenial and entorhinal cortices, and the hippocampus when the ambient magnetic field was perturbed.While this does not prove that all these areas process magnetic information, it suggests a correlation between stable magnetic fields and the cognitive map in these subterranean rodents [56].It is known from other rodents that spatially tuned cells often stop firing when their sensory anchors are unreliable [57] and the expression of c-fos has been linked to their activity in a given environment [58].The reported magnetic responses in the anterior thalamus in mole-rats are particularly noteworthy, as recently the firing of head direction cells in the anterior thalamus of mice were found to increase their average firing rate in response to auditory and somatosensory stimulation [59].This demonstrates that multimodal information from external cues already enters the head direction network at the level of the thalamus.Furthermore, magnetic field effects on c-fos expression in the dorsal thalamus were also found in head-fixed pigeons [54].It would be interesting to test if magnetic stimuli modulate the activity of head direction cells in this brain region.Consistent with the notion that magnetic cues are integrated with other modalities to form a multisensory representation of space, magnetic responses were also found in the superior colliculus of mole-rats [32].
Further downstream, the retrosplenial cortex which contains head direction cells and border cells is centrally involved in egocentric to allocentric transformation [37].Irrespective of the upstream magnetic information processing brain regions, we postulate that this hub for transformation of external sensory information into the spatial navigation circuitry could integrate magnetic information into the cognitive map.Once incorporated into the allocentric map, the spatial representations in downstream brain regions like the MEC and the hippocampus can also be expected to exhibit activity modulation by magnetic cues.
Finally, the hippocampal formation, which harbors place and head direction cells in bats, birds and rodents is an especially intriguing brain region [60e63].Magnetic field effects on hippocampal c-fos expression were found in birds and mammals.In freely hopping zebra finches, a changing magnetic field increased c-fos expression in the dorsomedial anterior hippocampus [64], and a similar effect has been reported for the hippocampus in head-fixed pigeons [54].The pigeon hippocampus is involved in spatial navigation [65,66] and contains place cells [67], as does the anterior hippocampus of zebra finches [61].Overall, these findings are consistent with the integration of magnetic cues in the hippocampal formation.

Possible encoding of magnetic landmarks
The hippocampus, MEC, and LEC [68] store information on both the spatial location and non-spatial sensory features of objects in the environment.Interestingly, mole-rats inspect magnetic objects longer than similar non-magnetic ones [69].We speculate that magnetosensitive animals encode the magnetic properties of objects in object tuned cells to map them as magnetic landmarks (anomalies) during navigation, comparable to the use of acoustic landmarks in bats [70].

Feasibility of recording brain activity in animals with a magnetic sense
Although there is some evidence for a magnetic sense in standard laboratory rodents [71e73], there is a lack of evidence for the use of magnetic cues in navigation in these animals [74].It is, therefore, advisable to study sensory specialists.These are primarily non-model organisms inhabiting specific niches, but recent advances in recording technology enable freely moving electrophysiological recordings in any animals able to carry a .*Magnetic responsive cells are yet to be discovered in MP where head direction cells were discovered in shear water chicks and DL where grid-like cells along with other spatially selective cells were very recently found in the tufted titmouse in a region lateral to the dorsal hippocampus seen as a functional homolog to the mammalian MEC (dashed black lines).✝DT and VN were found to have a magnetic response.The vestibular nucleus contains angular head velocity (AHV) cells while the dorsal thalamus contains head direction cells in mammals.However, presence of these directionally selective cells is yet to be established in the two brain regions in bird brains (dashed red lines).few grams on their head.Wireless transmission or logging of acquired neural data has made it possible to record from freely flying bats [62] and barn owls [60], swimming goldfish [75], freely moving quails [76] and migratory seabirds [77].High-density silicon probes enable the acquisition of stable neural data from hundreds to thousands of neurons over months and in deep brain regions [78].In addition to advances in electrophysiology, modern imaging technologies such as functional ultrasound imaging [79] and multi-photon quantification of calcium reporters enable functional assessment of large neuronal populations in freely moving animals [80].
Considerations for studies on the role of magnetic cues in cognitive maps When studying magnetic orientation behavior, the stimuli should mimic natural situations, i.e. they should have Earth's field strength and be spatially homogeneous.Furthermore, when integrating magnetic stimuli into electrophysiological experiments, one needs to be aware of artefacts due to electromagnetic induction [81].Controlling for other artefacts, such as heating or vibrations, by using double-wrapped coils is crucial [82].Moreover, as the compass of birds and possibly some mammals is sensitive to interfering radiofrequencies [71,83,84], minimizing and characterizing the electromagnetic fields in the recording setup is imperative.
Attention states and task requirements influence the dynamics of spatial representations [85].Hence, it is crucial to study the contribution of magnetic cues to the cognitive map when the magnetic information is relevant to the task [77].Long-distance migrants rely on their magnetic sense primarily during the migratory seasons and in specific migration phases [10].Furthermore, some animals might use their magnetic sense only during certain times of the day, e.g.around sunset, either to calibrate a sun compass [86] or because the magnetic field fluctuations are minimal during this time [87].A key challenge is, therefore, to find a species and behavior with a robust magnetic response at a spatial (<1e2 m arena) and temporal scale (<30e60 min per session) suitable for freely moving neural recordings inside a magnetic coil system.
An exciting example are streaked shearwater chicks just before their first migration, in which HDCs were found to fire preferentially towards the seasonally appropriate northern direction [77].The recordings were made during 10 min sessions in a small arena (diameter 120 cm), providing an ideal opportunity to study the effect of magnetic field manipulations on HDC firing.Other magnetosensitive species in which spatial cell recordings have been achieved include bats [62], fishes [75], pigeons [67], and zebra finches [61].To screen for magnetic compass information in spatial brain regions of these species, several sessions could be recorded from an animal inside a magnetic coil system that changes magnetic directions between the sessions (Figure 4a).Rotations of spatial representations (e.g. the preferred firing direction of a head direction cell) would indicate a compass response, similar to the response observed in rats when a prominent visual cue has been rotated [88].
It is currently technically impossible to study the cognitive map at the scale relevant for the magnetic map sense.It could be addressed, however, by recording from longdistance migrants before and after a virtual displacement.Place cells and grid cells are known to exhibit environment-specific firing patterns that can be stable over several months [90].In different environments, however, the cells reorganize their firing activity [91] in form of global remapping by place cells wherein the place fields of place cells change their respective firing locations, and in the form of grid realignment wherein the entire grid pattern of a grid cell moves and rotates.If an animal indeed perceives an environment as novel after a magnetic virtual displacement (with all cues except magnetic cues unchanged), one would expect to observe global rearrangement of activity in these cells (Figure 4b).
To promote the use of magnetic cues, other available cues should be minimized [20], although presenting the magnetic field along with additional visual cues as a reference could yield insights in cue conflicts induced by magnetic field rotations as demonstrated in bogong moths [89].Finally, global remapping has also been demonstrated to occur in the hippocampus and subiculum of bats when they were forced to switch from one sensory modality (vision) to another one (echolocation) within the same environment [92].It is therefore conceivable that remapping takes place in magnetosensitive animals when the magnetic field is not present.

Conclusion
In recent decades, our understanding of cognitive maps and the magnetic sense has evolved substantially, albeit independently.At the same time, neuroscience has undergone a technological revolution that begins to relax the restriction to a handful of standard model organisms.Novel tools allow bridging both fields to record neural activity in animals exploring naturalistic environments [93] and investigate the role of the Earth's magnetic field in the cognitive map of sensory specialists.We anticipate that as technology progresses, the number of recordings of spatial cells across diverse species will increase and simultaneously advance our understanding of the cognitive map and the role that magnetic cues might play in it.
. Zapka M, Heyers D, Hein CM, Engels S, Schneider NL, Hans J, Weiler S, Dreyer D, Kishkinev D, Wild JM, et al.: Visual but not trigeminal mediation of magnetic compass information in a migratory bird.Nature 2009, 461:1274-1277.Zapka et al. present the results from an elegant series of lesion experiments in European robins demonstrating i) that a forebrain region called Cluster N is crucial for magnetic compass orientation but not other compasses and ii) that the trigeminal nerve opthalmic branch is dispensable for magnetic compass responses.Together with other studies reporting magnetic responses in the trigeminal system, this work strongly supports the existence of separate magnetosensory systems for a magnetic compass and map sense.* .Nagelhus A, Andersson SO, Cogno SG, Moser EI, Moser MB: Object-centered population coding in CA1 of the hippocampus.Neuron 2023, 111:2091-2104 e2014.Nagelhus et al. studied the influence of objects introduced in an environment on population code of CA1 place cells.They recorded from object coding cells that encode both spatial and non-spatial features of the environment as well as non-object place cells in the CA1 region.The population of place cells, both object coding and non-object coding, were found to reorganize their activity around the object, hence, highlighting the influence of objects in the environment on the ensemble spatial code in the hippocampus.49.Saleem AB, Diamanti EM, Fournier J, Harris KD, Carandini M: Coherent encoding of subjective spatial position in visual cortex and hippocampus.Nature 2018, 562:124-127.

*
. Long X, Zhang SJ: A novel somatosensory spatial navigation system outside the hippocampal formation.Cell Res 2021, 31: 649-663.The spatial navigation circuitry in mammals has so far focused on a subset of spatially selective cells in the hippocampal formation.However, space was predicted to be an a priori to overlay experiences and hence was expected to have representations distributed in multiple brain regions.Long & Zhang provide evidence for the prevalence of the spatial code by demonstrating the presence of the entire subset of spatially tuned cells like place cells, grid cells, border cells, and head direction cells in the somatosensory cortex wherein the spatial map could enable representation of body in space.Wu & Dickman exposed head-fixed pigeons to a rotating magnetic field stimulus and mapped c-fos expression across the brain.They identified a set of brain regions with increased c-fos compared to controls kept in a zero magnetic field, including the hippocampus, dorsal thalamus, lateral hyperpallium and vestibular brainstem nuclei.Interestingly, the response was reduced to baseline levels after ablation of the inner ear lagena in all regions but the hyperpallium.These findings suggests that the lagena plays a critical role, but that a second magnetoreceptor exists which might be trigeminally innervated as they also found increased c-fos in the principal sensory nucleus of the trigeminal system.
Science 2012, 336:1054-1057.Wu & Dickman exposed head-fixed pigeons to a rotating magnetic stimulus and recorded single units in the vestibular brainstem.They discovered neurons tuned to the different components of the magnetic field, direction, intensity and polarity.Sensitivity to polarity is in line with magnetoreceptors based on ferrimagnetic particles or an induction based mechanism, while it contradicts a mechanism based on radical pairs.56 * * .Burger T, Lucova M, Moritz RE, Oelschlager HH, Druga R, Burda H, Wiltschko W, Wiltschko R, Nemec P: Changing and shielded magnetic fields suppress c-Fos expression in the navigation circuit: input from the magnetosensory system contributes to the internal representation of space in a subterranean rodent.J R Soc Interface 2010, 7:1275-1292.Burger et al. compared the c-fos expression in mole-rats that explored a novel arena.They found c-fos expression in many regions of the brain's navigation circuit.Interestingly, when the animals were exploring the arena in a near zero magnetic field or when the magnetic field was unstable, c-fos expression was suppressed in these regions.The findings are the most direct evidence so far that magnetic cues are part of the cognitive map in a vertebrate.The authors in this study demonstrate modulation of head direction cell activity in the anterior thalamus by behavioral states.Sensory cues like auditory stimuli and whisker stimulation were found to cause an increase in firing rate of HD cells.Additionally, monitoring of behavior states like pupil dilation and whisker pad movement was also found to influence the firing rate of HD cells during arousal states.Social interactions with conspecifics was also found to increase the firing rate of HD cells.These responses were absent in non HD cells in the anterior thalamus.This study provides evidence for relay of relevant sensory inputs by the mechanism of gain modulation of head direction cell activity in anterior thalamus, hence enabling flexible encoding of information by the spatial navigation system.Payne et al. compared spatial representations between the hippocampus of the food caching tufted tit mice and non-food caching zebra finches by recording in freely moving birds.They found place cells in both birds comparable to those found in mammalian species.However, the recruitment of the hippocampal cells to encode space and the precision of the spatial code was stronger in the food caching bird than in the zebra finches.This study provides keen insights into the dependence of spatial codes on ethological needs of the animal.

*
. Eliav T, Maimon SR, Aljadeff J, Tsodyks M, Ginosar G, Las L, Ulanovsky N: Multiscale representation of very large environments in the hippocampus of flying bats.Science 2021, 372.Neural encoding of space in all animal species investigated so far was carried out in smaller environments which do not replicate the naturalistic large and complex environments that these animals inhabit.Eliav et al. recorded neural activity from the hippocampus of freely flying fruit bats in a 200m long tunnel to study the spatial representations in naturalistic environments.They discovered that place cells employ a multiscale strategy wherein individual place cells encode multiple locations of the large environment with varying field sizes within the cells which makes the spatial code more efficient and precise.* * .Takahashi S, Hombe T, Matsumoto S, Ide K, Yoda K: Head direction cells in a migratory bird prefer north.Sci Adv 2022, 8, eabl6848.Takahashi et al. recorded neural activity from head direction cells in the medial pallium of shearwater chicks just before their migration period when they highly rely on magnetic compass for orientation.The population of head direction cells recorded was found to have a significant preference for encoding the northern magnetic direction of migration.This response was found to be independent of geolocation of the animals and was robust across different environmental conditions.This study highlights the importance of context and behavioral demands in shaping the cognitive maps.