A newly identified trigeminal brain pathway in a night-migratory bird could be dedicated to transmitting magnetic map information

Abstract

Night-migratory songbirds can use geomagnetic information to navigate over thousands of kilometres with great precision. A crucial part of the magnetic ‘map’ information used by night-migratory songbirds is conveyed via the ophthalmic branches of the trigeminal nerves to the trigeminal brainstem complex, where magnetic-driven neuronal activation has been observed. However, it is not known how this information reaches the forebrain for further processing. Here, we show that the magnetically activated region in the trigeminal brainstem of migratory Eurasian blackcaps (Sylvia atricapilla) represents a morphologically distinctive neuronal population with an exclusive and previously undescribed projection to the telencephalic frontal nidopallium. This projection is clearly different from the known trigeminal somatosensory pathway that we also confirmed both by neuronal tracing and by a thorough morphometric analysis of projecting neurons. The new pathway we identified here represents part of a brain circuit that—based on the known nidopallial connectivities in birds—could potentially transmit magnetic ‘map’ information to key multisensory integration centres in the brain known to be critically involved in spatial memory formation, cognition and/or controlling executive behaviour, such as navigation, in birds.

1. Introduction

(a). Avian magnetoreception

In birds, three sensory systems have been suggested to be involved in magnetoreception [1]: the visual system relying on light-dependent, radical-pair-based reactions of sensor molecules, probably cryptochromes, located in the retina [2] providing directional, i.e. ‘compass’ information [3]; the trigeminal system with currently unknown receptors providing input to a magnetic ‘map’ [4–8]; and the vestibular system with currently unknown sensors [9], potentially based on an electromagnetic induction mechanism [10] located in the inner ear [11,12].

The necessity of trigeminal input for a functional magnetic map has been shown in behavioural studies, in which migratory Eurasian reed warblers (Acrocephalus scirpaceus) with bilaterally ablated ophthalmic branches of the trigeminal nerves (V1) failed to compensate for geographical [7] and ‘virtual’ [8] magnetic displacements (figure 1a–d).

Figure 1.

Summary of the behavioural and neurobiological evidence which indicate the involvement of the ophthalmic branch of the trigeminal nerves (V1) in the perception of positional information based on previous studies. (a) Map showing magnetic field parameters (green isolines, total intensity; brown isolines, declination) of the capture site in Rybachy and the displacement site in Zvenigorod [4,5,7,8]. (b) Several previous studies [4,5,7,8] have shown that Eurasian reed warblers (Acrocephalus scirpaceus) orient northeast towards their breeding areas (yellow) during spring migration. (c) After both real and virtual magnetic displacements, intact birds shifted their orientation towards the northwest to compensate for the displacement [4,5]. (d) Birds with bilaterally ablated V1s could, however, not perceive their position based on magnetic cues and thus retained the same orientation as before the displacement [7,8]. (e) Magnetic field-driven activation of the trigeminal brainstem nucleus PrVv as revealed by the expression of the immediate-early gene Egr-1 (redrawn based on [13]). (f) Neuronal activation in PrVv significantly dropped down after bilateral V1 ablation or removal of magnetic stimuli (redrawn based on [13], see [13] for detailed statistical analyses). 5M, motor nucleus of the trigeminal nerve; d, dorsal; l, lateral; m, medial; N.V, trigeminal nerve; PrVd, dorsal part of the principal sensory trigeminal nucleus; PrVv, ventral part of the principal sensory trigeminal nucleus; v, ventral. Scale bars: (a) 400 km; f (for d,f), 200 µm.

In the brain, primary V1 afferents were shown to terminate in the spinal (SpV) and principal (PrV) sensory trigeminal brainstem nuclei [14]. The latter consists of a dorsal (PrVd) and a ventrally adjacent subcompartment (PrVv) [13–15]. Notably, previous studies [13,15,16] consistently showed increased neuronal activation only in PrVv in response to a strongly changing magnetic field stimulus (sCMF), but not to a zero magnetic field stimulus or to the sCMF stimulus when V1 was cut (figure 1e,f).

Apart from cerebellar projections, SpV has been shown to form a feedback loop projecting to the PrV complex [17,18]. It is thus reasonable to assume that the only region to send trigeminally perceived magnetic information to the forebrain is the PrV complex [17–20]. Therefore, PrVv is a prime candidate to relay trigeminally perceived magnetic information to the forebrain.

To understand trigeminal magnetoreception and multisensory integration of navigation-relevant cues, it is essential to map the onward connection between PrVv and the forebrain [21]. Therefore, the aim of the present study was to discover this connection in a night-migratory songbird, the Eurasian blackcap (Sylvia atricapilla).

It has long been known that PrVd relays somatosensory information from the beak to the telencephalic nucleus basorostralis (NB) [19]. By contrast, the connectivities of PrVv have remained elusive [20]. We started our work based on two mutually exclusive hypotheses. Either magnetic information reaching PrVv (i) is further processed along the same neural pathway as the adjacent somatosensory part and projects to NB, or (ii) it follows a separate, currently unknown pathway to the forebrain.

To test these hypotheses, we combined neuronal tract tracing, detailed morphometric analyses and analyses of magnetic field-induced brain activation patterns in the PrV complex of Eurasian blackcaps to reveal its exact anatomical features and connectivities.

2. Methods

(a). Animals and housing

All animal procedures were approved by the local and national authorities for the use of animals in research (Niedersächsisches Landesamt für Verbraucherschutz und Lebensmittelsicherheit/LAVES, Oldenburg, Germany, Az.: 33.12-42502-04-10/0121; 33.19-42502-04-11/0423; 33.19-42502-04-15/1865; 33.19-42502-04-17/2724). Twelve adult Eurasian blackcaps (S. atricapilla) of both sexes were caught by mist-netting in the vicinity of Oldenburg University. Birds were housed in pairs in wire cages and experienced a circadian and circannual light regime matching the local natural conditions. Food and water were provided ad libitum.

(b). Neuronal tract tracing

To visualize the afferent connectivities of the primary magnetosensory nucleus PrVv, we stereotactically injected neuronal tracer into NB (one individual), telencephalic frontal nidopallium (NFT) (four individuals) and/or NFT and V1 simultaneously (one individual). We also performed six attempts to inject tracer into PrVv. However, owing to the small size of PrVv in Eurasian blackcaps, we failed to trace PrVv and decided to stop owing to the strictly limited number of experimental, wild caught animals granted in our animal experiment permit.

To inject the tracer substances, fully anaesthetized birds were head-fixed in a custom-built stereotactic apparatus. After incision of the scalp, neuronal tracer was administered through a small window in the skull above the target region by minute pressure injections of 9.6–23 nl Choleratoxin B subunit (CtB, 1% in distilled water; Sigma, Germany) using a microinjector (WPI-2000, World Precision Instruments, USA) and bevelled glass capillaries. Coordinates (NB: A4.1 mm, L2.5 mm, 3.85 mm below brain surface; NFT: A4.1–4.3 mm, L2.8–3.0 mm, 3.5–3.65 mm below brain surface) were determined relative to the confluence of the superior sagittal and cerebellar ‘Y’ blood sinus. For V1 tracings, an injection of 0.5 µl CtB was made into the nerve as it passes medial to the eye within the orbit. After the surgery, the skull and skin were repositioned and sealed with cyanoacrylate surgical glue. Each bird was given 3–6 days to recover from the surgery and to let the tracer transport.

(c). Magnetic stimulation

Single birds were placed in a round plexiglas cage within a double-wrapped Merrit four-coil system [22] inside a radiofrequency field shielded [23] non-magnetic building [24] illuminated with an intensity of approximately 2.3 mW m⁻² [3]. The birds experienced a magnetic stimulus containing randomized small and large variations in field intensity (18 500–111 000 nT), horizontal direction (0–359°) and inclination (−84.9° to +76.6°). This stimulus was identical to the one which resulted in the activation of the trigeminal brainstem complex in other bird species in previous studies [13,15,16]. Although previous studies have shown no effect of motor behaviour on magnetic activation in PrVv [13], we excluded any potential effects of excessive mechanical stimulation of the beak [25] by monitoring the behaviour of each bird in real time using infrared cameras and by taking a bird for brain analyses only when it was constantly awake and sitting still during magnetic exposure for at least 90 min.

(d). Immunohistochemistry

Birds were deeply anaesthetized and transcardially perfused with saline followed by paraformaldehyde (4%/phosphate buffered saline (PBS), pH 7.4). Brains were extracted, cryoprotected overnight and serially sectioned in 40-µm thick slices in the frontal plane. Parallel series of slices were sequentially incubated free-floating with primary antibodies against either CtB (polyclonal rabbit anti-CtB, working dilution 1 : 3000, Sigma, Deissenhofen, Germany, or polyclonal goat anti-CtB for fluorescent stainings, working dilution 1 : 1000, Cayman Chemical, USA) overnight, Egr-1 (polyclonal rabbit anti-Egr-1, dilution 1 : 1000, Santa Cruz, CA, USA) for 72 h or the general neuronal marker HuC/HuD (monoclonal mouse anti-HuC/HuD, dilution 1 : 500, Invitrogen, Carlsbad, CA, USA) overnight, biotinylated secondary antibodies and an avidin-coupled peroxidase complex (Vectastain ABC Kits, Vector Laboratories, CA, USA). The signal was visualized using a 3′-3-diaminobenzidine (DAB) reaction using glucose oxidase instead of hydrogen peroxide [26]. We used comparatively high antibody working dilutions owing to a more stringent buffer system, i.e. tris-buffered saline instead of PBS, in order to improve the signal to background ratio.

For fluorescent immunolabellings, the primary antibodies were detected by Alexa-conjugated secondary antibodies (Alexa-488 donkey anti-rabbit, Alexa-488 donkey anti-goat, Alexa-568 donkey anti-mouse, Thermo Fisher Scientific, Germany, dilution 1 : 500) raised against the primary antibodies' host. After the reaction, brain slices were mounted on glass slides, coverslipped and analysed using light microscopy (Leica MDG36, Leica Z6 APO, Leica MC 170 HD, Wetzlar, Germany; Zeiss AxioScan Z1, Jena, Germany) or covered with antifade mounting medium with 4',6-diamidin-2-phenylindol (Vectashield, Vector Laboratories, CA, USA) and analysed using confocal microscopy (Leica TCS Sp8 MP, Leica, Germany).

For double stainings against Egr-1 and CtB, we performed two subsequent peroxidase substrate reactions: first, Egr-1 was visualized by the Ni-enhanced DAB reaction described above. Afterwards, CtB was visualized using an ImmPACT^® NovaRED™ kit (SK-4805, Vector Laboratories, CA, USA).

(e). Morphometric analysis/statistics

Retrogradely labelled neurons in the PrV complex were manually counted and their maximum soma diameters were measured using Fiji [27]. Statistical analyses of soma diameters were performed in R [28] using repeated measurements ANOVA: Soma size∼Region (PrVd/PrVv) * Group (HuC/HuD stainings/tracing) + Error (individual identity). The density distribution was depicted as a kernel distribution of soma sizes for corresponding regions using the R function geom_density from the ‘ggplot2’ package (for HuC/HuD stained neurons) and a histogram (for retrogradely traced neurons; figure 1f). The Gaussian mixture model (mclust package in R) was fitted to the distribution of soma sizes from the entire PrV complex (based on HuC/HuD stainings), and the model with the lowest Bayesian information criterion (ΔBIC) was selected. This statistical procedure is a standard approach for model selection that does not depend on the p-value and shows the relative fitness of one model over the other (ΔBIC between 2 and 6 is considered as a positive evidence [29]).

3. Results

(a). PrV connectivities

To test if the magnetically activated PrVv shares the neural pathway with the neighbouring somatosensory system, we first focally applied a retrograde neuronal tracer into NB (figure 2a, magenta), which, if working hypothesis (i) would be true, should label cells in PrVv. Surprisingly, we observed labelled neurons only in PrVd (figure 2d) and had no tracer signal in PrVv at all. The axons of some PrVd neurons are known to bifurcate into two branches [20] so that the same neurons also project to the contralateral NB (figure 2b). Therefore, the occurrence of labelled fibre terminals in NB of the opposite brain hemisphere served as a diagnostic tool to validate that we had actually injected the tracer into NB. These results exclude working hypothesis (i), namely that PrVv projects to NB, and therefore favour hypothesis (ii), suggesting that a previously unknown pathway is potentially dedicated to transmit the magnetic ‘map’ information to the forebrain.

Figure 2.

(a) The magnetically activated PrVv (green) forms a previously unknown connection upon the NFT, whereas PrVd (magenta) projects upon NB. NB (b), but not NFT (c) displays contralaterally projecting axon collaterals. (d) NB tracings retrogradely label neurons in PrVd, whereas (e) NFT tracings label PrVv neurons. (f) Morphometric analyses of the PrV neurons' diameters based on HuC/HuD staining show that PrVv neurons (green line; 9.70 ± 2.41 µm (s.d.); n = 640) are significantly smaller (repeated measurements ANOVA, p < 0.001) than neurons in PrVd (magenta line; 12.27 ± 2.71 µm (s.d.); n = 1050). Soma sizes of retrogradely labelled neurons from NFT (green bars; 9.2 ± 2.5 µm (s.d.); n = 99) and NB (magenta bars; 12.32 ± 3.47 µm (s.d.); n = 624) strikingly match (repeated measurements ANOVA, p = 0.54) the soma sizes of PrVv and PrVd, respectively. (g) HuC/HuD stainings reveal morphological differences between PrVd and PrVv. (h) Magnetic fields activate PrVv neurons (depicted as nuclear expression of the immediate-early gene Egr-1) only. 5M, motor nucleus of the trigeminal nerve; Cb, cerebellum; d, dorsal; H, hyperpallium; l, lateral; m, medial; M, mesopallium; N, nidopallium; NB, nucleus basorostralis; NFT, trigeminal frontal nidopallium; OT, optic tectum; PrVd, dorsal part of the principal sensory trigeminal nucleus; PrVv, ventral part of the principal sensory trigeminal nucleus; Str, Striatum; Tel, telencephalon; v, ventral. Scale bars: b (for b,c), 200 µm; d (for d–h), 100 µm.

Therefore, we asked where the magnetically activated PrVv does project to. In one of the pioneering studies, Wild et al. [19] reported retrogradely labelled somata in both dorsal and ventral PrV parts after retrograde tracer injections into the NB–NFT complex. Because PrVd is known to appear as an ovoid-shaped structure, the shown retrograde tracing pattern in the aforementioned study very likely included PrVv as well (fig. 5 in [19]). Could the NFT projections represent the next upstream relay and processing station for trigeminally transmitted magnetic navigational information?

Consequently, we proceeded to perform spatially restricted focal tracer injections into NFT. These NFT injections retrogradely labelled approximately 15% of neurons exclusively located in PrVv (624 neurons (HuC/HuD)/99 neurons (CtB); figure 2e; electronic supplementary material, figure S1a–d), which essentially matched the spatial distribution of the magnetically activated neurons (figure 2h). An entire lack of any projections to the contralateral hemisphere (figure 2c) excluded the possibility that any tracer injected into NFT had spread into NB. To further verify that we indeed hit NFT, we analysed target brain regions of putative NFT projections that are known from the existing literature [30–32]. As expected, we found tracer signal in the posterior pallial amygdala [30], the caudolateral nidopallium (NCL) [31] and the ventral part of the frontal mesopallium [32] (see the electronic supplementary material, figure S1F–H).

Because the NB–NFT complex has also been shown to receive input from the superior vestibular hindbrain nuclei [20], we had to exclude the possibility that our identified pathway represented parts of the vestibular system. To do so, we performed additional double tracings, in which we injected neuronal tracer into both NFT and V1. We could show that PrVv, where we found neuronal somata retrogradely labelled from NFT, was also innervated by the trigeminal V1 nerve (figure 3a–d). This result confirms that the retrogradely labelled neurons in PrVv belong to the trigeminal system.

Figure 3.

Regional colocalization of retrogradely labelled neurons from NFT, anterogradely labelled fibre terminals of V1 and magnetically induced Egr-1 expression in PrVv. (a) HuC/HuD labels neuronal somata; (b) exemplary case for a neuron within PrVv retrogradely traced from NFT (indicated by an arrow) and fibre terminals anterogradely traced from V1 (indicated by an arrowhead); and (c) nucleic marker DAPI labelling both neurons and glial cells. Therefore, not all cells showing nucleic staining by DAPI (c) are labelled by the neuronal marker HuC/HuD (a); (d) CtB signal (green), HuC/HuD labelling (magenta) and DAPI (blue) colocalize within one cell; (e) exemplary case for a neuron (indicated by an arrow) within PrVv retrogradely traced from NFT (red) displaying nuclear Egr-1 expression (black). Ventrally adjacent, an Egr-1-expressing neuron that is not retrogradely traced is shown (indicated by an arrowhead). Scale bars: d (for a–d), 10 µm; e, 5 µm.

Double immunostainings revealed that approximately 12% of the neurons in PrVv back-filled from NFT also showed neuronal activation upon magnetic stimulation (figure 3e).

(b). PrV morphology

To further validate that PrVv comprises a distinct neuronal subpopulation with a separate pathway to the forebrain, we performed comprehensive morphometric analyses of the PrV complex.

In frontal sections of Eurasian blackcap brains stained against the general neuronal marker HuC/HuD, the prominent ovoid-shaped PrVd appeared to consist of large, densely packed neurons, whereas the ventrally adjacent, crescent-shaped PrVv contained smaller, more sparsely distributed neurons (figure 2g; electronic supplementary material, figure S1I,K).

Morphometric analyses based on the HuC/HuD stainings revealed that the neuronal somata within PrVd (12.27 ± 2.71 µm (s.d.); n = 1050; magenta line in figure 2f) were significantly larger (repeated measurements ANOVA, p < 0.001) than those in PrVv (9.70 ± 2.41 µm (s.d.); n = 640; green line in figure 2f) and thus indicating the presence of two morphologically distinct subpopulations. Strikingly, the measured soma sizes of retrogradely labelled neurons from NFT (green bars; 9.2 ± 2.5 µm (s.d.); n = 99) and NB (magenta bars; 12.32 ± 3.47 µm (s.d.); n = 624) matched the soma sizes of PrVv and PrVd neurons, respectively (repeated measurements ANOVA, p = 0.54; electronic supplementary material, figure S1I–L).

We could thus independently confirm the actual presence of two morphologically distinct subpopulations by fitting a Gaussian mixture model to the distribution of soma sizes within the entire PrV complex. The model with two distinct Gaussian distributions resulted in the lowest Bayesian information criterion (ΔBIC = 4.9 relative to a single distribution, figure 4), which implies strong evidence against the null hypothesis of a single neuronal population.

Figure 4.

A Gaussian mixed model was used to fit the soma size distribution of all neurons in the PrV complex. The model with two independent Gaussian distributions (black lines) resulted in the lowest BIC value (ΔBIC = 4.9 relative to a single Gaussian distribution). The modelled distributions based on all PrV neuron soma sizes coincide closely with the soma size distributions found when the neuron sizes within PrVd (magenta) and PrVv (green) were analysed separately. This analysis independently confirms that the PrV complex consists of at least two morphologically distinct neuronal subpopulations.

4. Discussion

(a). A potential magnetic map processing pathway in Eurasian blackcaps

Both behavioural [4,5,7,8] and neurobiological evidence [13,15,16] have indicated an involvement of the ophthalmic branch of the trigeminal nerve in magnetoreception with PrVv representing a likely primary relay to process trigeminally perceived magnetic information.

In the present study, connectivity analyses show that the magnetically activated PrVv in Eurasian blackcaps comprises a morphologically distinct neuronal subpopulation, which forms the origin of a previously unknown projection to NFT and thus markedly differs from the originally described trigeminal projections towards NB (which we here show to exclusively originate from PrVd).

Our assumption, that PrVd and PrVv actually belong to different neuronal pathways, is strongly supported by morphometric analyses of the neuronal soma sizes retrogradely labelled from NFT and NB. The soma sizes of neurons back-traced from NFT (located in PrVv) and NB (located in PrVd) very closely matched the respective soma sizes within PrVv and PrVd as defined by HuC/HuD stainings (figure 2f; electronic supplementary material, figure S1I–L). Could this newly found neuronal connection between the magnetically activated PrVv and NFT be part of a novel magnetoreception pathway in birds?

Our double tracings from NFT and V1 (figure 3a–d) confirmed that PrVv receives input from the ophthalmic branch of the trigeminal nerve (figure 3a–d) that has been shown to transmit magnetic information [7,8]. Moreover, NFT-projecting neurons in PrVv regionally colocalized with magnetically activated neurons, and, occasionally, tracer-filled somata and Egr-1 expressing nuclei could be assigned to the very same neurons (figure 3e). Our tracer labelled only approximately 15% of PrVv neurons which might be explained by the fact that only parts of NFT received tracer injections. We also cannot rule out any possible interneuronal connectivities within PrVv which leaves the possibility that further, unlabelled PrVv neurons could also be involved in processing trigeminally perceived magnetic information. Future studies will be needed to investigate if any unlabelled PrVv neurons might be involved in processing magnetic/non-magnetic stimuli and if they innervate other, currently unknown brain regions.

(b). Multisensory integration of navigational information

Studies on the trigeminal sensory-motor circuit in the avian forebrain have shown that the PrVv-recipient NFT is connected to two top-level integration centres in the avian forebrain, both of which were recently proposed as ideal sites which could integrate navigational information [21].

The first integration centre in the avian forebrain is the hippocampal formation (HF) [21,33]. Although its role in long-distance navigation has remained elusive to date [34], HF neurons in pigeons have been reported to respond to magnetic stimuli [11,35]. This makes HF a potential candidate structure for integrating magnetic information from various sensory systems (vestibular [11,36,37], visual [21,33] and trigeminal [21,33]).

Another major candidate for integrating magnetic information in the avian forebrain is the NCL. It is the avian analogue of the mammalian prefrontal cortex, where various sensory inputs are weighed against each other in order to calculate the optimal motor output [38]. The NCL has been shown to play a crucial role in executing navigational behaviour, such as homing from unfamiliar locations in pigeons [39].

Although the putative role of the HF and/or the NCL in integrating magnetic information is not proven, the connectivities shown here could provide an important neuroanatomical basis for future functional analyses on the magnetic senses of migratory birds. To sum up, our newly discovered connection between PrVv and NFT represents a previously unknown brain pathway in birds connecting the magnetically activated trigeminal brainstem with telencephalic higher order brain centres involved in controlling navigational behaviour. It is thus tempting to speculate that PrVv and NFT might process important aspects of magnetic ‘map’ information and that the expected onward projections to the hippocampal formation, and NCL might be used to integrate trigeminal magnetic ‘map’ information with other orientation cues in order to calculate navigational goals in a migratory bird.

Ethics

All animal procedures were approved by the local and national authorities for the use of animals in research (Niedersächsisches Landesamt für Verbraucherschutz und ebensmittelsicherheit/LAVES, Oldenburg, Germany, Az.: 33.12-42502-04-10/0121; 33.19-42502-04-11/0423; 33.19-42502-04-15/1865; 33.19-42502-04-17/2724).

Data accessibility

All data related to this study have been included in the original manuscript.

Authors' contributions

D.H. and H.M. designed research; D.H., D.K., S.S. and B.M. performed experiments; D.H., D.K., S.S. and B.M. analysed the data; D.K. and M.W. performed statistical analyses; H.M. provided reagents and facilities; D.H., D.K. and H.M. wrote the first draft of the paper, which all coauthors commented on.

Competing interests

We declare we have no competing interests

Funding

Generous financial support was provided by the Deutsche Forschungsgemeinschaft (Projektnummer 395940726—SFB 1372 ‘Magnetoreception and Navigation in Vertebrates’ to D.H., H.M. and M.W.; He6221/1-1 to D.H.; Mo1408/1-2 to H.M.; GRK 1885 to H.M and M.W.), a stipend from ‘Landesgraduiertenkolleg Nano-Energieforschung’ funded by ‘Ministerium für Wissenschaft und Kultur’ of Lower Saxony to D.K. via H.M.