Zebra finches have a light-dependent magnetic compass similar to migratory birds

It is well known that birds can detect the Earth's magnetic field (reviewed by Wiltschko and Wiltschko, 1995) and use information from the sun (reviewed by Schmidt-Koenig, 1990), the skylight polarization pattern (Muheim et al., 2006a) and stars (Emlen, 1975) for orientation. Despite extensive research, however, the sensory mechanisms involved in avian magnetoreception are still not fully understood.

It has been suggested that animals may perceive magnetic compass information by a light-dependent radical-pair process taking place in cryptochromes in the avian retina (Schulten et al., 1978; Ritz et al., 2000; reviewed by Rodgers and Hore, 2009; Wiltschko et al., 2010; Solov'yov et al., 2010; Mouritsen and Hore, 2012; Hore and Mouritsen, 2016). Cryptochromes are the only known vertebrate photopigments to form long-lived, spin-correlated radical pairs upon light excitation (Ritz et al., 2000). They have been reported in several animal groups, including insects (Emery et al., 1998; Gegear et al., 2010), amphibians (Eun et al., 2003), birds (Mouritsen et al., 2004; Möller et al., 2004; Wiltschko et al., 2007a; Liedvogel and Mouritsen, 2010; Nießner et al., 2013; Fusani et al., 2014; Du et al., 2014) and mammals (van der Horst et al., 1999; Avivi et al., 2004). The light-absorbing cofactor of cryptochromes, flavin adenine dinucleotide (FAD), exists in three redox states with different absorption spectra (Fig. 1): a fully oxidized state, FAD_ox, with an absorption spectrum in the UV and blue light (peaks at about 360 and 470 nm); a semi-reduced state, the neutral semiquinone radical FADH^• (peaks at about 495 and 580 nm); and a fully reduced state, FADH⁻, with no absorption in the visible part of the spectrum (Liu et al., 2010; reviewed by Wang et al., 2015). Magnetic compass orientation in diverse animal groups has been observed under both blue and green monochromatic light (reviewed by Muheim et al., 2002; Johnsen et al., 2007; Phillips et al., 2010a; Wiltschko et al., 2010). Recent evidence from European robins (Erithacus rubecula) suggests that the magnetically sensitive radical pair in cryptochromes involved in light-dependent magnetoreception in birds includes the semiquinone FADH^•, which is formed during the light-independent re-oxidation of the fully reduced FADH⁻ to FAD_ox, and an as yet unknown partner radical, possibly O₂^•− (Solov'yov and Schulten, 2009; Müller and Ahmad, 2011; Nießner et al., 2013, 2014; Wiltschko et al., 2016; reviewed by Wang et al., 2015).

Behavioural evidence from various animals suggests that access to light of specific characteristics (wavelength and intensity) is necessary for magnetic compass reception (reviewed by Wiltschko et al., 2010; Phillips et al., 2010b). Insects and amphibians trained and tested under wavelengths of light shorter than 450 nm displayed oriented behaviour towards a trained magnetic compass direction, but shifted their orientation by ∼90 deg when tested under longer wavelengths (Phillips and Borland, 1992a,b; Phillips and Sayeed, 1993; Vacha, 2004; Dommer et al., 2008; Diego-Rasilla et al., 2010, 2013). Young, inexperienced homing pigeons (Columba livia) were unable to orient towards home when transported to the release site in complete darkness or under red light (650 nm), but were well oriented when transported under green (565 nm) or full-spectrum light (Wiltschko and Wiltschko, 1981, 1998). Domestic chickens (Gallus gallus), trained to relocate a social stimulus using their magnetic compass, showed oriented behaviour under blue light (465 nm), but not under red light (645 nm) (Wiltschko et al., 2007a). European robins, Australian silvereyes (Zosterops lateralis) and garden warblers (Sylvia borin) were found to be oriented towards the expected migratory direction when tested for magnetic compass orientation in funnels under full-spectrum, UV, blue, turquoise or green light (373–565 nm) (e.g. Wiltschko et al., 1993, 2010, 2014; Wiltschko and Wiltschko, 1999; Rappl et al., 2000; Wiltschko and Wiltschko, 2001; Muheim et al., 2002; for a complete list of experiments testing magnetic compass orientation in migratory birds carried out under monochromatic light, see Tables S1 and S2; Fig. S1). When tested at longer wavelengths, however, the birds were either disoriented or showed a ∼90 deg shifted orientation from the expected direction (e.g. Wiltschko et al., 1993, 2008; Muheim et al., 2002), which at least in some cases was found to be a fixed direction, not based on a radical-pair mechanism (Wiltschko et al., 2008). This change in behaviour under specific wavelengths of light happens very abruptly. European robins tested under narrowband green (561 nm) and green–yellow (568 nm) light were oriented under the shorter wavelength, but disoriented under the longer wavelength (Muheim et al., 2002).

Light intensity (photon irradiance) is also an important parameter affecting how birds perceive the magnetic field. European robins and Australian silvereyes have been shown to be oriented towards the expected migratory direction when tested under low-intensity green light (total irradiance ∼0.06×10¹⁶–2.1×10¹⁶ quanta s⁻¹ m⁻²; see Table S1, Fig. S1A). However, when tested under green light at higher intensities (total irradiance ∼2.6×10¹⁶–7.2×10¹⁶ quanta s⁻¹ m⁻²), they were either disoriented or showed various orientation responses either unimodally or axially along the North–South or East–West axis (e.g. Wiltschko and Wiltschko, 2001; Muheim et al., 2002; Wiltschko et al., 2005; Wiltschko et al., 2010) (see Table S2, Fig. S1B). Some of these responses appear to be fixed directions, as they were independent of the migration season and did not change as a result of inversions of the inclination angle of the magnetic field (Wiltschko et al., 2003, 2005, 2007b). This suggests that the birds were unable to use their light-dependent inclination compass under these conditions, but instead relied on an iron–mineral-based magnetic polarity sense.

Evidence is growing that birds use a magnetic compass for various spatial tasks unrelated to migratory orientation. Chickens (Gallus gallus) have been shown to be able to orient towards a specific magnetic direction associated with a social stimulus (Freire et al., 2005; Denzau et al., 2013). Zebra finches (Taeniopygia guttata) are able to orient towards a magnetic compass direction associated with a food reward when trained in an open arena (Voss et al., 2007) or in a 4-arm ‘plus’ maze (Muheim et al., 2016). When tested in the presence of a vertically aligned radio-frequency electromagnetic (RF)-field, which is a diagnostic test to assess the involvement of a radical pair mechanism in the primary magnetoreception process (Timmel and Hore, 1996; Ritz et al., 2004; Henbest et al., 2004), zebra finches were no longer able to use their magnetic compass (Keary et al., 2009; Muheim et al., 2016). This suggests that the magnetic compass of non-migratory birds is very likely to be based on a radical pair-based reception process, as has been suggested for migratory birds (e.g. Ritz et al., 2004).

In the present study, we trained zebra finches to find a food reward in a 4-arm plus maze under different wavelengths and intensities of light to investigate whether their magnetic compass showed light-dependent properties similar to that of other birds, and to examine whether the behavioural evidence fits the suggested model of cryptochrome as magnetoreceptor, taking into account the spectral and biophysical characteristics of its different redox states.

All behavioural experiments were carried out between 2014 and 2016 at Stensoffa Field Station, near Lund (Sweden). The experimental animals belonged to a permanent captive breeding colony of zebra finches (Taeniopygia guttata Reichenbach 1862) at the station. All experiments were done according to the ethical guidelines from the Malmö-Lund Animal Ethics Committee (permits M 158-11, M 423-12 and M 24-16).

Individual birds were trained to find a reward (millet seeds) in a plus maze using directional magnetic compass information as the only orientation cue (no visual, sound or olfactory cues; for details on the setup, see Muheim et al., 2016). The maze was centred on a magnetic coil (Merrit design), which allowed the experimenter to align an Earth-strength magnetic field (intensity ∼50.6 µT, inclination 69.8 deg) towards any of the four cardinal directions [magnetic North (mN) at geographic North (gN), South (gS), East (gE) or West (gW)], without changing magnetic inclination or total intensity (Muheim et al., 2016).

To evaluate the light dependency of the magnetic compass in zebra finches, we followed the training and testing procedure described in detail in Muheim et al. (2016). Individual male zebra finches of at least 6 months of age were trained to find a food reward in either the mN or mS direction under one of the following experimental conditions: monochromatic light with peak wavelengths at 430 nm indigo, 461 and 463 nm blue, 521 nm green or 638 nm red (see Table S3 for details on the light conditions used in this study and Fig. S1 for comparison of light spectra with previous studies), produced by an array of LEDs (OF-BLR5060RGB300, OPTOFLASH and 5 W 430 nm Actinic Indigo High Power; Shenzhen Weili Optical Co. Ltd, Shenzhen, China). We used two light intensity conditions with a total light irradiance of 0.80×10¹⁶ quanta s⁻¹ m⁻² (low-intensity light) and 15×10¹⁶–18×10¹⁶ quanta s⁻¹ m⁻² (high-intensity light), respectively (measured with IL 1400 radiometer with SHD033 detector; International Light Technologies, Peabody, MA, USA). The training procedure was as follows: the bird was put inside the release device in the centre of the maze in complete darkness. Once the experimenter had left the arena, the lights were switched on and the bird was released into the arena after 30 s. The bird was allowed to search the maze until it located the reward. Whenever the bird hopped inside an unrewarded feeder, the lights were turned off for 8–10 s for punishment. When the bird found the food, it was allowed to feed for 10 s, after which it was removed from the maze and put back into the holding cage. A training trial was considered successful when the bird found the reward within 3 min, without visiting more than seven arms in total. Each bird was trained twice during an afternoon, and then tested once the next morning. Between the two training trials, the location of the food reward and the alignment of the magnetic field were shifted by 90 deg clockwise or counter-clockwise. Once a bird had been rewarded in at least three different maze arms under a training condition, which usually was the case after the second training day, its orientation was tested in a probe trial. In the probe trial, the bird was tested once in a magnetic field aligned towards one of the four geographic directions. The bird was allowed to search the maze (without food reward) for 60 s, and its behaviour was recorded with a video camera centred above the maze. For each experimental condition, the number of birds tested in each of the four cardinal directions was kept balanced as far as possible. Individual birds were repeatedly trained under the same light condition; 32 out of the 54 individuals used in this study were re-trained under a different wavelength after a break of at least 2 months from any experimental work (e.g. some birds trained under low-intensity blue light were later trained under low-intensity green light, but were never re-trained under low-intensity blue light again).

To examine whether the magnetic compass orientation was based on a radical-pair mechanism, we also tested the birds in the presence of a low-intensity, RF-field at 1.403 MHz (Larmor frequency at the study site) with ∼260 nT peak intensity (see Fig. S2) (Muheim et al., 2016), aligned vertically with respect to gravity, i.e. at an angle of ∼20 deg relative to the static magnetic field vector (see Ritz et al., 2004). Individual birds were tested in the presence of a RF-field under the same monochromatic light under which they were trained. For most of the birds, a probe trial without the RF-field was carried out prior to a probe trial with the RF-field. A few birds experienced the probe trial with the RF-field before the probe trial without the RF-field, but they did not exhibit a different behaviour from the other birds in the same experimental group.

The movements of the birds during the probe trials were automatically tracked using custom-written tracking software (Muheim et al., 2016). The number of frames a bird spent in each of the four maze arms was used to calculate the mean orientation of each individual bird. Thereby, unit vectors for the geographic directions of each of the four maze arms (0, 90, 180, 270 deg) were summed by vector addition with the number of frames the bird spent in each arm, irrespective of its exact position in the arm. The mean orientation relative to gN and mean vector length of each bird was then obtained by dividing the resultant vector by the total number of frames (Batschelet, 1981). The orientation of the individual birds relative to gN was then recalculated relative to mN and relative to the trained magnetic compass direction (this being the only one where orientation was expected). The mean orientation of groups of birds was calculated by vector addition, disregarding the individual mean vector lengths, and tested for significance with the Rayleigh test (Batschelet, 1981). Axial orientation was calculated by doubling the angles, and groups with an axial mean vector length larger than the unimodal vector length were considered axially oriented. We used the 95% confidence interval test to determine whether the birds were significantly oriented towards the trained magnetic compass direction. All circular statistics were performed with Oriana version 4 (Kovach Computing Services, Anglesey, UK).

Individual zebra finches were initially trained and tested under low-intensity (0.80×10¹⁶ quanta s⁻¹ m⁻²) light of one of three peak wavelengths: 461 nm blue, 521 nm green or 638 nm red light (Fig. 2, top; Table 1). Birds trained and tested under green light showed a significant axial orientation along the trained magnetic compass direction (Fig. 2B). To test whether the orientation observed under green light was dependent on a radical-pair mechanism, we tested the birds under the same monochromatic light in the presence of a weak 1.4 MHz RF-field (Fig. 2E). Birds tested under 521 nm green light in the presence of the RF-field were completely disoriented (Fig. 2E), supporting the involvement of a radical-pair mechanism. Birds trained and tested under 461 nm blue light were not significantly oriented, but showed a tendency to orient axially along the trained magnetic compass direction (Fig. 2A), similar to the axial orientation observed under green light. This tendency disappeared when the birds were tested in the presence of a RF-field (Fig. 2D). Under 638 nm red light, the zebra finches tended to orient unimodally about 45 deg clockwise to the trained magnetic direction (Fig. 2C). In the presence of a RF-field, the birds were no longer oriented along the trained magnetic compass direction, but instead oriented significantly along the topographic North–South axis (P=0.014; Table 1).

Considering that cryptochromes are blue-light photosensitive proteins (Wang et al., 2015), the lack of orientation under 461 nm blue light was quite unexpected. Because we could not exclude that the failure to orient was a result of the relatively low light intensity in the maze, compared with the full-spectrum light used by Muheim et al. (2016), we carried out a series of experiments under monochromatic light of higher intensity (15×10¹⁶–18×10¹⁶ quanta s⁻¹ m⁻²) (Fig. 2G–J, Table 2). However, birds trained and tested under high-intensity 463 nm blue light were totally disoriented (Fig. 2H), which could indicate that cryptochromes are not involved in the magnetoreception process studied here. Alternatively, the disorientation of the zebra finches under 461 nm blue light could also be related to the spectral properties of the blue LEDs. The peak wavelength of our blue light was shifted about 20–40 nm towards longer wavelengths than the blue light spectra used previously in orientation experiments with European robins, garden warblers and Australian silvereyes (peak wavelengths at 424 and 443 nm) (Wiltschko et al., 1993, 2003, 2007b; Wiltschko and Wiltschko, 1999, 2001; Rappl et al., 2000; for an exception, see Wiltschko et al., 2007a) (Tables S1 and S2). Therefore, we trained and tested the zebra finches under high-intensity 430 nm indigo light. The birds oriented significantly along the trained magnetic compass axis (Fig. 2G), similar to the birds trained and tested under low-intensity green light (521 nm). In the presence of a RF-field, the birds were no longer oriented (Fig. 2I), indicating that magnetic compass orientation under high-intensity indigo light is likely to be mediated by a radical-pair mechanism.

We initially used three different monochromatic lights in the blue (461 nm), green (521 nm) and red (638 nm) spectrum to evaluate whether magnetic compass orientation in zebra finches was wavelength dependent, as has previously been reported for migratory songbirds, homing pigeons and chickens (Muheim et al., 2002; Johnsen et al., 2007; Wiltschko et al., 2007a, 2010). We expected oriented behaviour towards the trained magnetic compass direction in zebra finches trained and tested under blue and green light, and disoriented behaviour or 90 deg-shifted orientation in birds tested under red light. To our surprise, the light-dependent magnetic compass of the zebra finches followed that prediction only partially. Birds trained and tested under low-intensity 521 nm green light oriented as expected along the trained magnetic compass axis, as has previously been shown under full-spectrum white light (Muheim et al., 2016). Axial orientation along the trained magnetic compass axis, instead of unimodal orientation towards the trained magnetic direction, has been documented in previous magnetic training studies in birds (Freire et al., 2005; Voss et al., 2007; Muheim et al., 2016). The reasons for this axiality are currently unknown, but could be related to the birds’ motivation, the experimental procedure, or the perception of the magnetic field, i.e. how the birds perceive and interpret the magnetic modulation pattern proposed to be produced by a radical-pair-based magnetic compass (Ritz et al., 2000; Phillips et al., 2010b). The disorientation of the zebra finches tested under low-intensity 521 nm green light in the presence of a RF-field supports the involvement of a radical-pair mechanism in the magnetic compass response under this light condition (see Ritz et al., 2004; Keary et al., 2009; Muheim et al., 2016).

Zebra finches trained and tested under 638 nm red light were not significantly oriented, but showed a weak tendency to orient about 45 deg clockwise of the trained magnetic compass direction. Similar shifts in orientation under red light have been observed in European robins (Muheim et al., 2002; Stapput et al., 2008; Wiltschko et al., 2008). Such responses have been suggested to be mediated (1) by two antagonistic spectral mechanisms, resulting in orientation responses aligned roughly perpendicular to each other (Phillips and Borland, 1992a,b; Muheim et al., 2002), or (2) by fixed magnetic directions mediated by iron–mineral-based receptors (Stapput et al., 2008; Wiltschko et al., 2008). As we trained and tested the zebra finches under the same wavelength of light, we expected them to also be oriented towards the trained magnetic compass direction in the presence of an antagonistic mechanism (see Deutschlander et al., 1999). The shifted orientation under red light might alternatively be a fixed magnetic direction, as has been observed in migratory songbirds (Wiltschko et al., 2008). Fixed magnetic orientation, or magnetic alignment, has been described in a diverse range of animals, ranging from insects to mammals, and refers to a spontaneous, non-goal-oriented preference to align the body axis or orient towards a specific magnetic direction (reviewed by Begall et al., 2013), the adaptive significance of which is not fully understood.

Under 461 nm blue light, the zebra finches showed a tendency to orient along the expected trained magnetic compass axis. We found the lack of significant orientation very puzzling, as animals using a light-dependent magnetic compass based on cryptochromes should be oriented under this wavelength (e.g. Phillips and Borland, 1992a; Phillips and Sayeed, 1993; Wiltschko and Wiltschko, 1998, 1999; Muheim et al., 2002; Vacha, 2004; Wiltschko et al., 2007b; Fig S1A). We noticed that the birds moved more slowly in the maze when trained and tested under the monochromatic light conditions compared with the full-spectrum light used in Muheim et al. (2016). Thus, we reasoned that the lack of orientation under low-intensity (0.80×10¹⁶ quanta s⁻¹ m⁻²) blue light could be the result of the light intensity being too low for the birds to perceive the magnetic field [for comparison, light irradiances used by Muheim et al. (2016) were ∼0.4×10¹⁶–65×10¹⁶ quanta s ⁻¹ m⁻²] or that the dim light negatively affected the birds' motivation to move around in the maze. Surprisingly, the orientation of birds trained and tested under higher intensity (15×10¹⁶ quanta s⁻¹ m⁻²) 463 nm blue light was even more scattered, as evident from a completely homogeneous distribution of directional choices between individuals. Similar effects of light irradiance on magnetic orientation have been observed in nocturnally migrating birds, which showed oriented behaviour under low-intensity monochromatic green light (∼0.06×10¹⁶–2.1×10¹⁶ quanta s⁻¹ m⁻²), but were disoriented or showed various orientation responses unrelated to migratory orientation under higher intensity green light (∼2.6×10¹⁶–7.2×10¹⁶ quanta s⁻¹ m⁻²) (Wiltschko and Wiltschko, 2001; Muheim et al., 2002; Wiltschko et al., 2010; Tables S1 and S2). The irradiance levels at the transition between orientation and disorientation, however, appear to be one order of magnitude higher in zebra finches compared with migratory birds. This could be related to intrinsic differences between the diurnal zebra finches and the nocturnal migratory birds, or to the difference in the behavioural assay used to test magnetic compass orientation (food searching versus migratory orientation).

An alternative explanation for the disorientation of the zebra finches under 461–463 nm blue light could lie in the spectral properties of the light. Studies testing magnetic compass orientation in migratory birds under ‘blue’ light used relatively broad-spectrum light (half-bandwidths of 56–70 nm, compared with 20–30 nm of our blue spectra), with peak wavelengths at 424 and 443 nm, thus much shorter than our blue light (Fig. S1, Tables S1 and S2 and references therein; note, however, that chickens can orient under 465 nm blue light; Wiltschko et al., 2007a). If light-dependent magnetoreception of the avian magnetic compass is based on two or more antagonistic, spectral mechanisms, as suggested by Muheim et al. (2002) and Wiltschko et al. (2004), the disorientation under 461–463 nm blue light could be the result of two spectral mechanisms being exited equally and cancelling out each other. This has been suggested to occur in insects and amphibians at around 475 nm (Phillips and Borland, 1992a; Deutschlander et al., 1999; reviewed by Phillips et al., 2010a), which is very close to our 461–463 nm blue light. Thus, if the disorientation of our zebra finches was the result of two antagonistic spectral mechanisms cancelling out each other, we would expect the birds to be oriented at wavelengths of light shorter than 461–463 nm. Indeed, birds trained and tested under 430 nm indigo light oriented along the trained magnetic compass axis, as we had observed in birds trained and tested under green light. The birds were no longer oriented when they were tested in the presence of a RF-field, indicating the involvement of a radical-pair mechanism in the primary reception process. However, further experiments are needed to conclusively establish whether an antagonistic spectral mechanism is responsible for the disorientation under 460 nm blue light.

As suggested by Ritz et al. (2000), the most promising molecules for involvement in magnetoreception are cryptochromes, as they are the only known vertebrate photopigments to form long-lived, spin-correlated radical pairs upon light excitation. Such radical pairs exist only in the transient semiquinone form in the redox cycle of cryptochrome, formed either during photoreduction of the fully oxidized form by UV to blue light or during re-oxidation of the fully reduced form without the need of light (Solov'yov and Schulten, 2009; Nießner et al., 2013, 2014; Wiltschko et al., 2016; reviewed by Wang et al., 2015; Fig. 1B). The orientation of our zebra finches under 430 nm indigo light agrees with the involvement of the semiquinone FADH^• formed either during the reduction of FAD_ox to FADH^• or during the light-independent re-oxidation of the fully reduced FADH⁻ to FAD_ox (Müller and Ahmad 2011; Nießner et al., 2013, 2014; Wiltschko et al., 2016; reviewed by Wang et al., 2015). The well-directed orientation of the zebra finches under 521 nm green light, in contrast, can only be explained by the involvement of the semiquinone FADH^• formed during the light-independent re-oxidation of the fully reduced FADH⁻ to FAD_ox, as suggested by Nießner et al. (2013, 2014) and Wiltschko et al. (2016). The alternative radical pair formed during the reduction of FAD_ox to FADH^• after illumination by UV to blue light cannot explain the orientation of the birds under green light (Fig. 1B). Given that calculated and measured lifetimes of cryptochrome lit states in vivo, in both plants and insects, are in the order of minutes (Herbel et al., 2013) and that our birds were exposed to full-spectrum white light in the holding room until between 30 s and a maximum of 1 min before the start of the training and probe trials, it is reasonable to assume that enough FADH⁻ was available for reoxidation to form FADH^•.

Taken together, our findings support previous evidence for the involvement of a light-dependent mechanism mediating magnetic compass orientation in zebra finches (Voss et al., 2007; Keary et al., 2009; Muheim et al., 2016). Our data support the hypothesis that the magnetic compass of zebra finches is based on a radical-pair mechanism and exhibits similar wavelength characteristics to those described earlier in homing pigeons (Wiltschko and Wiltschko, 1981, 1998) and migratory songbirds (Wiltschko et al., 1993, 2010, 2014; Wiltschko and Wiltschko, 1999, 2001; Muheim et al., 2002) (Tables S1 and S2). The understanding of the primary magnetoreception processes of light-dependent magnetic compass orientation is far from complete. Nevertheless, this and earlier studies with non-migratory birds (Freire et al., 2005; Wiltschko et al., 2007a; Voss et al., 2007; Denzau et al., 2013; Muheim et al., 2016) support the idea that light-dependent magnetoreception is not an exclusive feature of migratory birds, but instead is a general capability of all birds. Animals thus use a light-dependent magnetic compass not only for large-scale migrations but also for small-scale orientation tasks in their familiar environment (e.g. Muheim et al., 2006,b; Painter et al., 2013; Phillips et al., 2013; see also Phillips et al., 2010b).

We thank John B. Phillips for invaluable input into the development of the behavioural assay and critical discussions, and two anonymous reviewers for their constructive comments on the manuscript. We thank Johan Bäckman and Arne Andersson from the Centre of Animal Movement Research, CAnMove, for help with the experimental setup.