Introduction

Use of the geomagnetic field for spatial orientation has been documented in all five classes of vertebrates [1,2]. Among mammals, bats [3,4], two families of subterranean molerats [5–7], as well as Siberian hamsters and C57BL/6 mice [8–10], have been shown to rely on magnetic cues to determine compass headings. Magnetic cues have also been implicated in the tendency of free-living mammals as diverse as cattle, deer, and foxes to preferentially align their body axes along the North–South axis [11–13].

Evidence for learned magnetic compass orientation in Siberian hamsters and C57BL/6 mice is of particular interest because to date the large literature on rodent spatial behavior has yielded no other evidence for the use of magnetic cues in these, or other, epigeic rodents (i.e., species that are active above ground and, typically, have well-developed visual systems). It is noteworthy that all of the studies demonstrating the use of magnetic cues by rodents, with the exception of [14], have used assays that involve spontaneous or learned nest positioning behaviors [6,7,9,15]. Consequently, one possible explanation for the lack of evidence in other hamster and mouse studies is that rodents only utilize magnetic compass cues in certain behavioral contexts such as nest-building. However, given the obvious utility of a global reference system provided by the Earth’s magnetic field [1,16], the failure to use magnetic cues in other spatial tasks would be surprising, all the more so because magnetic responses in molerats appear quite robust, having been demonstrated in multiple species belonging to two different families and in studies carried out by at least four different laboratories [5–7,14,17,18].

Findings obtained over the last ~35 years indicate that there are two magnetoreception mechanisms in terrestrial vertebrates, based on fundamentally different biophysical processes [17,19–27].

Magnetite-based mechanisms involve single domain or interacting super-paramagnetic particles of biogenic magnetite believed to produce mechanical deformation of, or torque on, membrane structures that activate coupled membrane channels [28] or, in the case of freely-rotating single domain particles, to secondarily affect the rate of free-radical reactions that, in turn, influence the opening or closing of membrane channels [29]. Although in theory a magnetite-based mechanism could mediate responses that are sensitive to only the axis but not polarity, or to both the axis and polarity, of the magnetic field [28], to date all responses in which magnetite has been implicated are polarity sensitive [30–32].

A second class of mechanism is thought to involve a light-dependent biochemical reaction that forms long-lived, spin-correlated radical pair intermediates, i.e., the radical pair mechanism or RPM. To date, all of the responses in which a light-dependent, putatively radical pair-based mechanism has been implicated are sensitive to the axis, but not polarity, of the magnetic field [19,33], consistent with magnetic field effects on radical pair systems [34–36]. Animals with this type of ‘inclination’ compass use the slope or inclination of the magnetic field lines, instead of polarity (i.e., north vs. south), to distinguish between ’poleward’ and ’equatorward’ directions [33]. However, the most compelling evidence for a RP-based magnetic compass has come from behavioral experiments showing disruption of magnetic compass orientation by low-level radio frequency fields (1-10 MHz) that can alter the dependence of the radical pair reaction on magnetic field alignment [37–40].

Cryptochromes are the only animal photopigments known to form radical pair intermediates and are widely believed to play a role in the radical pair mechanism [34–36]. Absorption of a photon of light causes the flavin chromophore of cryptochrome to undergo an electron transfer reaction with an as yet unidentified electron donor resulting in the formation of the radical pair. The effect of an earth-strength magnetic field on the spin dynamics of the radical pair [34–36] is proposed to alter the formation or persistence of the cryptochrome’s signaling state, or of another downstream processes involved in phototransduction [34,36,38,39]. As a consequence, the response of photoreceptors containing an ordered array of light-activated cryptochrome molecules may show a complex dependence on magnetic field alignment [34,41,42].

Molerats are subterranean rodents adapted to live in aphotic habitats and exhibit spontaneous magnetic nest building responses that are consistent with a magnetite-based mechanism, i.e., independent of light, sensitive to the polarity of the magnetic field, and affected by brief magnetic pulses strong enough to remagnetize single domain particles of magnetite [5,17,43]. In contrast, the magnetoreception mechanism underlying learned nest building responses of C57BL/6 laboratory mice exhibits a complex pattern of response that is axially symmetrical (i.e., independent of polarity), consistent with the involvement of a radical pair mechanism [10] [16,21,25,33,39,44].

Reliance on magnetic cues may be especially important in novel surroundings, as shown in some species of migratory songbirds that use magnetic field cues for the initial calibration of star patterns [45]; but see 46. If epigeic rodents exposed to novel surroundings use the magnetic compass to ‘calibrate’ visual landmarks and then transfer control of spatial behavior from the magnetic reference to the visual landmarks, subsequent experimental manipulations such as rotation of visual landmarks, or changing the alignment of the magnetic field, will be unlikely to reveal the involvement of magnetic cues (see Discussion).

The present study was carried out to determine whether C57BL/6 mice: (1) can use magnetic cues to solve a water maze task carried out in an electromagnetically shielded enclosure that reduces levels of background radio frequency interference, and (2) can rapidly learn the relative position of a submerged platform with respect to the alignment of the magnetic field when they are first introduced to the novel surroundings. Experiments were carried out in a modified 4-arm (plus) water maze [47] enclosed within a pair of coils used to position an earth-strength magnetic field in one of four horizontal alignments coinciding with the four arms of the maze. To maximize the likelihood that mice would rely on magnetic cues, the testing apparatus was housed in a radially symmetrical room without directional visual, auditory, vibratory or olfactory cues.

Materials and Methods

Water maze

Mice were trained and tested in a plus water maze ([47]; Figure 1), modified to include a centrally located release device (Figure S1). The plus maze was aligned so that the arms coincided with the cardinal compass directions. During training, a submerged 10.5 x 6.5 cm platform was located at the end of one of the four arms (~0.8 cm below the water surface). The water was made opaque with white tempera paint (Crayola Premier, non-toxic) so that the location of the submerged platform was not visible. Previous work has shown that water temperature has a strong effect on water maze performance [48]. In early experiments (Figure 2; Table S1) water temperatures were maintained between 28.5–30.0 °C, and in later experiments (Figure 3, Table S2) between 27.5–29.0 °C; at progressively higher water temperatures the trained response relative to the magnetic field deteriorated and then reversed (unpublished data). In both training and testing, the holding room was maintained at a temperature of 25.5-30 °C and the room containing the water maze was maintained at 27-31 °C and relative humidity of 60-90%.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Training and testing protocol.

Mice with no prior experience with the water maze were given two training trials in different arms of the water maze with the submerged platform in the same relative alignment to the magnetic field (A & B). In this example, the mouse is being trained to orient to magnetic south. For testing the following morning, the submerged platform was removed (C). The mice were released individually from a central release device (Figure S1) and had free access to all four arms of the maze. Magnetic field alignment was changed between trials, and data pooled across testing groups, so an equal number of mice were tested in each of the four magnetic field alignments, i.e., magnetic North (mN) at geographic North, East, South or West; each mouse tested only once. Orientation direction was calculated by the tracking software as the vector sum of the times spent in the four arms during the 60 sec testing trial (D).

https://doi.org/10.1371/journal.pone.0073112.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. C57BL/6 mice rapidly learn the magnetic direction of a submerged platform in the plus water maze (data in Table S2).

Directional responses from mice given two training trails in the late afternoon (Figure 1A & B), and then tested the following morning (Figure 1C). A) The distribution of topographic bearings, i.e., deviations from the north arm of the maze (topN), was indistinguishable from random (p > 0.10, Rayleigh test). B) The same was true of the distribution of magnetic bearings, i.e., deviations from the alignment of magnetic north in testing (magN). C) In contrast, the distribution of bearings relative to the trained magnetic direction (black triangle) was non-randomly distributed, and the 95% confidence interval for the mean vector bearing contained the trained direction. Each data point is the directional response of a single mouse tested in one of the four magnetic field alignments (see Methods & Methods). Arrow in the center of (C) is the mean vector for distributions of bearings that are non-randomly distributed. The length of the arrow is proportional to the mean vector length (r), a measure of the clustering of bearings ranging from 0 to 1; radius of the circle corresponds to r = 1. Dashed lines show the 95% confidence interval for the mean vector bearing [52]. ‘n.s.’- not significant (p > 0.10; Rayleigh Test).

https://doi.org/10.1371/journal.pone.0073112.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Replication of learned response (data in Table S3).

A) The distribution of topographic bearings was indistinguishable from random (p > 0.10, Rayleigh test). B) The distributions of magnetic bearings from north- and south-trained mice were significantly different (p < 0.02, U²=0.253 Watson U² test). Due to the unequal sample sizes (Table S3), however, these responses did not cancel out and the overall distribution of magnetic bearings was non-randomly distributed. C) The distribution of bearings relative to the trained magnetic direction (black triangle) was non-randomly distributed, and the 95% confidence interval for the mean vector bearing contained the trained direction.

https://doi.org/10.1371/journal.pone.0073112.g003

The water maze was illuminated with white light from a ~1m diameter ring of rope lights (total length ~ 29 m consisting of ~1200 miniature bulbs; Utilitech Extra-Bright Clear Rope Light) coiled ~ 10 cm above a white 1.2 x 1.2 m Plexiglas diffuser. The rope lights were located above the electromagnetic shielding that enclosed the testing room. The video camera used to record the movements of the mice was also located above the electromagnetic shielding centered above a ~20 cm hole in the center of the diffuser that permitted an unobstructed view of the water surface. Light intensity at the water surface in the center of the plus maze was 28 lux.

Results

In the first test series, six groups of mice were trained to symmetrical directions relative to the magnetic field, i.e., trained direction = north (1 group), = south (1), = west (2), = east (2), and each mouse was tested once in only one of the four magnetic field alignments (magnetic north aligned to geomagnetic North, East, South, or West; see Materials & Methods). As in earlier nest-building experiments [9], this resulted in a total of 16 experimental conditions, i.e., 4 trained directions relative to the magnetic field x 4 alignments of the magnetic field in testing. Both the distribution of topographic bearings (i.e., absolute bearings, ignoring the alignment of the magnetic field in testing; Figure 2A) and the distribution of magnetic bearings (i.e., deviations of bearings from magnetic north, ignoring the trained direction; Figure 2B) were indistinguishable from random. In contrast, the distribution of bearings in Figure 2C was clustered in the trained magnetic direction (Table S2).

A similar response was obtained with a larger sample of mice trained to the north and south (Figure 3; Table S3). Evidence from preliminary experiments suggested that female mice may be able to recognize and follow odors left by their littermates in the plus water maze (Figure S3; Table S1). In the present experiments, ‘olfactory following’ was eliminated by scrubbing the inside walls of the maze and thoroughly mixing the water from the four arms between testing trials (see Materials & Methods). However, subsequent analyses of data pooled from both series of experiments (Figures 2 & 3) indicated that the remaining non-directional olfactory cues from littermates in the same testing group (i.e., the group of mice trained and tested together) may have influenced female behavior. The distribution of bearings relative to the trained magnetic direction from mice tested in littermate groups of 1 or 2 showed significantly less scatter than that from mice in littermate groups of 3 or more (Figure S4; p < 0.05, Watson U² test). Ongoing research has confirmed the testing groups made up of subgroups of 1-2 littermates show strong magnetic compass responses comparable to those shown in Figure S4A.

Discussion

The current findings show that C57BL/6 mice can use magnetic cues to solve a 4-armed (plus) water maze task. Given the large number of studies using water maze assays, it is surprising that earlier studies have not reported evidence for the use of magnetic cues. There are several possible reasons why previous studies have failed to find evidence for the use of magnetic cues. Magnetic cues may only be used when mice are presented with a task that favors the use of directional, rather than spatial, information, as is the case in the plus water maze. Magnetic cues may have less salience when alternative, non-magnetic cues such as visual landmarks (including the observer) are available or, more generally, when the problem or task the animal is confronting is more readily solved using local rather than global cues. Alternatively, mice may only attend to magnetic cues when first exposed to novel non-magnetic directional cues (i.e., unfamiliar landmarks) or surroundings.

Studies of migratory birds have shown that the Earth’s magnetic field can be used both as a primary source of compass information, and as a calibration reference for other directional cues, i.e., the sun and star compasses [1]. In nocturnal migrants, the magnetic compass is used to calibrate star patterns, but the ‘star compass’ then assumes primary control of migratory orientation even with the magnetic field present [45,46]. A similar phenomenon could occur in epigeic rodents if visual landmarks are initially calibrated relative to the magnetic field, but then take over primary control of directional and spatial behavior. For example, in water maze experiments, mice are often given acclimation trials prior to the onset of training to familiarize them with the task [53,54]. Any involvement of magnetic cues in the initial encoding (‘‘calibration’’) of non-magnetic spatial/directional cues may be missed unless manipulation of the relationship between magnetic and non-magnetic cues occurs when experimental subjects are first introduced into the water maze, as occurred in the present experiments. Further research is needed to determine if any of these factors, as well as background levels of RFI (see below), affect the use and relative importance of magnetic cues.

Despite these unresolved issues, there are several important conclusions to be drawn from the present study:

1. 1. The findings confirm the results of earlier experiments showing that C57BL/6 mice, and in all likelihood other epigeic rodents, have a well-developed magnetic compass [8,9].
2. 2. Use of magnetic cues is not limited to nest-building behavior [9]. Given evidence presented here for the use of magnetic compass cues in a water maze assay (Figures 2, 3), it is likely that magnetic cues will be found to be involved in other directional/spatial behaviors.
3. 3. Mice can learn the magnetic direction of an important feature of the environment very quickly. Mice required only two brief (< 60 sec) training trials to learn the compass direction of the submerged platform in the plus water maze.

In the present experiments, stable magnetic compass responses were only observed when experiments were carried out in isolated buildings where ambient radio frequency fields were relatively low; the testing room was equipped with a single layer of grounded aluminum window screen, and electrical lines entering the shielded enclosure were equipped with in-line RFI/EMI filters that reduced levels of radio frequency interference in the water maze to < 0.1 nT at frequencies from 0.2 to 200 MHz (see also Figure S2). In the absence of such precautions, magnetic compass responses of mice in the water maze experiments, as well as mice and hamsters in earlier nest building experiments [8,9], varied in both strength and direction (Phillips unpubl. data). These findings suggest that sensitivity to low-level radio frequency fields, at intensity levels found in typical laboratory environments, may be a source of unexplained variability in studies of spatial behavior in animals that rely on radical pair-based magnetic compass.

Whether or not use of magnetic compass cues requires a low-RFI environment [55], understanding the role(s) of these cues in rodent spatial behavior and cognition is clearly of first importance. In contrast to visual, olfactory, auditory or vibrational ‘landmarks’, the geomagnetic field provides a global reference frame that can be used to help organize local landmark arrays [56] into a global map of familiar space. Use of such a global reference contrasts with the prevailing view that epigeic rodents use inertial guidance, and other ideothetic cues, to place visually isolated regions of the environment into register [57–60]. Consistent with this view, lesions of the vestibular nuclei have been shown to alter or eliminate directional responses at both behavioral and neural levels when rats move from familiar to unfamiliar surroundings [61–63]. In birds, however, neurophysiological responses have been recorded from the vestibular nuclei that are sensitive to both magnetic and gravitational input, suggesting that this may be a site where the integration of magnetic and gravitational information necessary for an inclination compass occurs [64]. Recent evidence from birds also suggests that the vestibular nuclei receives input from a magnetite-based receptor in the lagena [65,66] but see 67, or inner ear [67]. It remains to be determined if lesions of the vestibular nuclei in rodents eliminate input from an inclination magnetic compass or prevent integration of magnetic cues with other egocentric and allocentric cues used for path integration [68,69]. In the latter case, simple directional information provided by either a magnetite-based or radical pair-based mechanism would suffice [28].

If the magnetic compass of epigeic rodents is mediated by a radical pair mechanism, as appears to be the case in amphibians and birds [21,25,34,38,39,70,71], magnetic cues could play a variety of previously unforeseen roles in rodent spatial behavior and cognition. When photo-magnetoreceptors mediating a radical pair-based magnetic compass are located in the retina, as may be the case in birds and epigeic rodents [15,72–75], the magnetic field may be perceived as a complex, 3-dimensional pattern of light intensity or color superimposed on the animal’s surroundings and fixed in alignment with respect to the magnetic field [34,36,41]. The type of pattern produced by a radical pair mechanism could provide not only a global source of directional information, but also a simple spherical grid or coordinate system that could help to integrate spatial information from multiple sensory modalities [55].

In experiments with rats carried out in total darkness in an unshielded laboratory environment, Tryon et al. [76] found no evidence for magnetic field effects on the response of head direction cells, or the associated behavior, both during spontaneous activity and during attempts to condition the animals to directional magnetic stimuli. Unlike subterranean molerats [17,18], there is no evidence that epigeic rodents derive compass information from a light-independent, magnetite-based mechanism, although the involvement of this type of mechanism in a ‘map’ or geographic position sense is also a possibility.

Sensitivity to low-level radio frequency fields is a diagnostic property of the radical pair mechanism [34,37–39,42]. Studies of birds and cockroaches suggest that responses to earth-strength (~50 µT) magnetic fields can be disrupted by radio frequency fields tuned to the Larmor Frequency (the precession frequency of the electron spin dipole in the ambient magnetic field) at intensities as low as 10-15 nT [38,39,77,78]. Evidence that magnetic compass orientation in Siberian hamsters and C57BL/6 mice can be disrupted by background levels of radio frequency interference present in laboratory settings suggests that thresholds could be even lower (i.e., < 5 nT), although this level of sensitivity is difficult to reconcile with current theory [36,79].

Conclusion

C57BL/6 mice have a well-developed magnetic compass sense that they readily use in nest-building [9] and water-maze (present study) tasks, both of which are learned responses. Given the value of the magnetic field as a global reference, it is difficult to imagine any directional or spatial behavior in which magnetic cues would not play at least some role, provided that the conditions necessary for detection of the magnetic field are met. The type of biophysical process that underlies the magnetic sense (i.e., either magnetite-based, or radical pair-based) may determine both the conditions under which the magnetic compass can operate, and the potential role(s) of magnetic cues in spatial behavior and cognition. Determining whether the magnetic compass of epigeic rodents is sensitive to low-level radio frequency fields and, if so, what the threshold is for such effects, has the potential to reveal fundamental properties underlying the detection and use of the geomagnetic field for spatial orientation.

Supporting Information

Acknowledgments

We would like to thank David Dommer for his help in the design and construction of research facilities.