Eyes are essential for magnetoreception in a mammal

Abstract

Several groups of mammals use the Earth's magnetic field for orientation, but their magnetosensory organ remains unknown. The Ansell's mole-rat (Fukomys anselli, Bathyergidae, Rodentia) is a microphthalmic subterranean rodent with innate magnetic orientation behaviour. Previous studies on this species proposed that its magnetoreceptors are located in the eye. To test this hypothesis, we assessed magnetic orientation in mole-rats after the surgical removal of their eyes compared to untreated controls. Initially, we demonstrate that this enucleation does not lead to changes in routine behaviours, including locomotion, feeding and socializing. We then studied magnetic compass orientation by employing a well-established nest-building assay under four magnetic field alignments. In line with previous studies, control animals exhibited a significant preference to build nests in magnetic southeast. By contrast, enucleated mole-rats built nests in random magnetic orientations, suggesting an impairment of their magnetic sense. The results provide robust support for the hypothesis that mole-rats perceive magnetic fields with their minute eyes, probably relying on magnetite-based receptors in the cornea.

1. Introduction

Magnetoreception, the ability to perceive magnetic fields, is a sensory modality that occurs in all major vertebrate groups as well as in a range of invertebrate taxa [1–3]. The hunt for a magnetic sense in mammals has gained pace since the late 1980s. By now magnetoreception has been shown in numerous mammal groups (reviewed in [4]). Still, even though rodents are readily available for experiments under laboratory conditions, only a few attempts have been made to explore the underlying physiology of this sensory modality in mammals. Past studies employed methods spanning electrophysiology [5], pharmacological inhibition [6] and neural activity mapping [7,8], but their paucity contrasts with the great number of methodologically diverse experiments performed with birds [9–11]. Therefore, the location, structure and functional properties of mammalian magnetoreceptors are still elusive [4].

A popular hypothesis proposes that magnetite crystals (Fe₃O₄) linked to mechanosensitive ion channels could enable magnetically modulated neuronal excitation, rendering magnetoreception possible [3,12]. Such magnetite receptors would function light independently and could be located anywhere in the body. In mammals, there is evidence consistent with magnetite-mediated magnetoreception in bats [13,14] and in Ansell's mole-rats (Fukomys anselli), a subterranean rodent from the woodlands of Central Zambia [15].

Ansell's mole-rat is a model species in the study of magnetoreception because of its strong innate preference to build nests in the south-eastern sector of a circular arena [15]. This orientation bias is species-specific, with other African mole-rats displaying deviating magnetic preferences in analogous experimental settings [16]. Nest-building assays have revealed several properties of the Ansell's mole-rat's magnetic compass: it works in total darkness and responds to changes in field polarity but not in inclination [17]; it is affected by strong magnetic pulses [18] but is insensitive to radiofrequency magnetic fields [19]. Responsiveness to radiofrequencies was tested by applying a broadband field of 0.1–10  MHz with an intensity of 85  nT or 1.315  MHz fields of either 480 nT or 4800 nT intensity, but none affected the magnetic orientation of the mole-rats [19]. The mentioned characteristics of the mole-rat's magnetic sense fit magnetite-based receptors but seem incompatible with the radical pair mechanism of magnetoreception [19], which is prominently hypothesized to be present in birds [10,11] and also in murid rodents [20]. However, the anatomical locus of the corresponding receptor cells remains unknown.

The eye has a crucial role in the avian magnetic compass response ([10,11] but see [21]) and has therefore also attracted attention as a potential location of magnetoreceptors in mammals (e.g. [6]). Although Ansell's mole-rat spends most of its life in darkness and visual input is of marginal behavioural relevance for the species (cf. [22]), its minute eyes show no sign of qualitative regression. Instead, they conform to the mammalian basic pattern in displaying all typical anatomical features [23]. Wegner et al. [6] reported that local lidocaine anaesthesia of the cornea led to a loss of directional nest building in Ansell's mole-rats. The treatment did not impair light–dark discrimination, suggesting that at least some functions of the retina were unaffected. The authors concluded that the mole-rat magnetoreceptors are located in the cornea, but the utility of lidocaine for behavioural testing has significant drawbacks. First, induced anaesthesia, even after repeated applications, only lasts up to 15 min [24]. Second, lidocaine is lipophilic and can diffuse widely, potentially affecting untargeted tissues. Lidocaine further binds to serum proteins in the blood and rapidly crosses the blood–brain barrier, where it may induce non-specific effects on the central nervous system [25]. Therefore, ablation experiments should be preferred over local anaesthesia in experiments aimed to narrow down the site of magnetoreceptors [26]. To test the hypothesis that the magnetoreceptors of Ansell's mole-rats are indeed located in the eye, we assessed magnetic orientation during nest building in subjects with surgically removed eyes.

2. Materials and methods

2.1. Subjects

We studied 40 (series 1: six males, six females; series 2: 16 males, 12 females) adult Ansell's mole-rats (Fukomys anselli, Bathyergidae, Rodentia) from the breeding stock of the Department of General Zoology at the University of Duisburg-Essen. All subjects were born in captivity and socially housed as either pairs or family groups of variable size. In enucleated mole-rats (series 1: 12 subjects; series 2: 14 subjects), the eyes were removed entirely. Surgical removal of the miniscule eyes is uncomplicated, as no bony orbit is present and the ocular musculature, apart from the musculus retractor bulbi, is severely reduced [23]. Excision of the eye was accomplished under general anaesthesia and after application of a local muscle relaxant by a single cut with curved microsurgical scissors. This procedure sectioned the optic nerve and the three oculomotor cranial nerves as well as the part of the ophthalmic branch of the trigeminal nerve that innervates the eye, including the cornea. Other portions of the ophthalmic branch of the trigeminal nerve, which innervate the snout region, and other cranial nerves remained intact so that the neurological effects of the surgery remained precisely localized and impairment of animal welfare minimal. The enucleation procedure is further described in [27]. The recovery period between surgery and testing was a minimum of three weeks and 18 months in the first and second series of experiments, respectively. During recovery and following the experiments, the animals were socially housed in their home terraria, analogous to control group subjects. All surgeries and experiments conformed to the relevant ethical standards and were approved by the animal welfare officer of the University of Duisburg-Essen and the LANUV NRW, Germany (series 1: 50.05-230-37/06; series 2: 84-02.04.2015.A387).

2.2. Ethograms

Using video recordings, we compared the behaviour of enucleated (n = 6) and control (n = 8) adult mole-rats of four families from experimental series 2 in their home terraria for a minimum of 435 min (maximum 1338 min). The analysis of the videos was performed manually and blindly with respect to the experimental group (the resolution of the videos was not sufficient to resolve the mole-rats' minute eyes) by one observer (Katharina Schröer) who was not further involved in the project. Behaviours were evaluated separately for each individual on a minute by minute basis. The behaviour that lasted longest during the respective minute was noted; comparatively rare events (grooming, sniffing, and social play) were noted even if they occurred only shortly during the respective minute interval. The time budgets of the following routine behaviours were determined for each animal: resting, feeding (including food transport), locomotion (including digging), sniffing, grooming (auto- and allogrooming) and social play (e.g. sparring with teeth, play fighting). The fraction of each behavioural category per total amount of recording time was calculated. Statistical differences between the experimental groups were tested in R version 3.6.0 [28] with a binomial general linear model with experimental group and type of behaviour as factors.

2.3. Nest-building assay

We used the established nest-building assay to study magnetic compass orientation in Ansell's mole-rat (e.g. [6,15,17,19]). When being offered nesting material, mole-rats will typically start constructing nests within a short period of time, irrespective of their sex and reproductive status. The disposition to build nests also does not differ between animals tested in pairs, in groups or alone. Each pair was treated as a statistical unit. Animals were placed singly or in pairs (table 1) inside an opaque circular plastic arena (80 cm diameter, 30 cm height) and were given approximately 60 min to build a nest from paper scraps, which were evenly distributed across the arena. A white PVC cylinder (diameter 15 cm; height 20 cm) was placed in the middle of the arena to prevent nest construction in the centre. When paper scraps were used to construct a clearly demarcated nest mound (rather than a diffuse, potentially random aggregation of material), its position was documented photographically. In case no unequivocal nest was constructed (n = 10 for controls, n = 8 for enucleated subjects in series 2), animals repeated trials under the respective magnetic condition until a valid nest was built.

Table 1.

Basic information on study animals used in series 1 (2006) and series 2 (2018/2019). Data on age refer to the time of testing. µ and r represent the direction and lengths of the angular mean vector regarding magnetic north, respectively.

Series 1 (2006)
ID	sex	age (months)	µ (control)	r (control)	µ (enucl.)	r (enucl.)
0287 3854	F M	30 30	119°	0.99	285°	0.32
7556 0919	F M	82 60	89°	0.68	250°	0.11
2668 3900	F M	47 60	182°	0.59	41°	0.38
2848 3369	F M	72 32	127°	0.46	150°	0.31
9546 3919	F M	56 59	134°	0.58	37°	0.32
8118 0466	F M	91 106	105°	0.43	106°	0.67

Series 2 (2018/2019)
ID	sex	age (months)	enucleated/control	µ	r
2654 0328	F M	55 55	both enucleated	47°	0.06
6125 6752	F M	66 39	both enucleated	125°	0.21
5496 0861	F M	69 54	both enucleated	71°	0.60
1334 0772	F M	75 78	both enucleated	250°	0.27
0772 5813	F M	31 43	both control	111°	0.98
6767 0816	F M	26 51	both control	142°	0.80
4780 0726	M M	69 21	both control	196°	0.58
0724 0717	F M	102 23	both control	150°	0.51
4295	F	57	enucleated	241°	0.60
2175	M	65	enucleated	30°	0.21
9074	F	90	control	245°	0.34
1641	M	58	enucleated	118°	0.37
5683	F	44	enucleated	218°	0.68
0528	F	151	control	216°	0.87
3360	M	184	enucleated	270°	0.08
6126	M	42	enucleated	214°	0.09
2731	M	27	control	18°	0.06
0737	F	25	control	185°	0.74
5222	M	26	control	121°	0.19
3493	M	24	control	195°	0.69

In experimental series 1, mole-rat pairs were tested under ambient local geomagnetic field conditions (51°27'50.3″ N 7°00'18.9″ E, 48.6 μT, 66° inclination). Tests were performed in 2006 and took place in a non-magnetic unshielded greenhouse (made of plastic and aluminium) at Essen University campus, Germany. Animals were manually introduced into the arena from random directions. During testing, the arena was covered with a light-impervious lid to exclude possible visual orientation. A potential influence of ambient auditory stimuli from the campus environment could not be fully mitigated, but we made no observations potentially reflective of acoustically induced biases. Furthermore, Ansell's mole-rats have limited hearing capabilities especially at frequencies above 1 kHz [29]. Each mole-rat pair underwent four tests under control conditions and six tests after enucleation.

In experimental series 2, individual animals or pairs were tested in Essen-Haarzopf, a rural area under ambient geomagnetic field conditions (49 µT, 66° inclination) in the periphery of the city of Essen, Germany (51°25'09.6″ N 6°56'50.9″ E). Tests were performed in 2018–2019 and took place in a windowless wooden hut constructed completely of non-magnetic materials. The interior of the hut was shielded from radiofrequencies by a grounded Faraday cage made of aluminium mesh. Radiofrequency noise within the hut was low, peaking at 0.4 MHz with intensities of approximately 10 nT, but well below 1 nT for frequencies up to 100 MHz (discussed and visualized in [30]). To be able to distinguish topographic biases from magnetic orientation, each individual/pair was tested in the ambient magnetic field (mN = 360°) and three shifted fields with magnetic north at geographic east (mN = 90°), south (mN = 180°) and west (mN = 270°), respectively; the inclination and the intensity of the shifted magnetic fields remained unchanged (intensity: 48.5 ± 0.18 µT; inclination: 65.5 ± 0.05°). The intensity has been measured with a three-axial magnetic field sensor (FGM3D, Sensys, Bad Saarow, Germany). The sequence of magnetic conditions was randomized. The local magnetic field was manipulated by a double-wrapped three-axial four coil Merritt system (edge length: 3 m × 3 m × 3 m) powered by a programmable multichannel power supply (HMP4040, Rohde & Schwarz, Munich, Germany). Importantly, during the ambient magnetic field tests, the identical current was run antiparallelly through the coil system, so that possible side effects of the operational coils (noise, vibrations, heat) were identical for all sessions. The experimental arena was placed within a square wooden box (edge length: 120 cm) equipped with four windows (100 cm length × 80 cm height), one on each cardinal direction, which were used to place animals into the arena from different topographical positions in a randomized sequence. These windows were closed during testing. The arena was fixed on a pedestal seated in a sandbox to minimize substrate vibrations.

During testing, an LED table lamp emitting monochromatic red light (Parathom R50 80.337 E14 Red 617 nm, 6 W, Osram, Munich, Germany) illuminated the room outside of the wooden box to allow the experimenters to operate. African mole-rats are not capable of perceiving red light [22,31]. Photon density (light level) within the wooden box under testing conditions was less than 0.001 µmol s⁻¹ m⁻² (measured by LI-COR photometer, model LI-250). The temperature of the hut was controlled by an air conditioning system (Model FTXS50, Daikin) and was measured before (mean: 26.37°C; s.d.: ± 1.74) and after (mean: 23.53°C; s.d.: ± 2.06) each experimental run. The air conditioner was deactivated during experiments to minimize directional acoustic cues and prevent disruption of the magnetic environment within the hut.

2.4. Data analysis

Nest orientations were scored (to the closest 5°) with reference to topographic north from photos by an experimenter (S.B.) blind to the magnetic condition and experimental group. We refer to topographic north as the direction of magnetic north before the field had been artificially rotated. After scoring, data from the four magnetic conditions were back-transferred by subtracting 90° (mN = east), 180° (mN = south) or 270° (mN = west) to obtain values with reference to magnetic north (for details see [32]). Combining the data collected under different magnetic conditions but with reference to topographic north will be non-random if the animals use non-magnetic cues for orientation. On the other hand, if the pooled data with reference to magnetic north will be non-random, the animals orient by using magnetic cues. Mean vectors of the four trials of each individual/pair were calculated with respect to magnetic or topographic north via vector addition. Second-order statistics (Moore's modified Rayleigh test) using mean vectors and lengths of the respective mean vectors were used to detect significant deviations from a random distribution, with α = 0.05. To test for a symmetric bimodal distribution, we employed Moore's modified Rayleigh test on doubled angles. All calculations were performed with Oriana 4.02 (Kovach Computing, UK). We performed model-fitting statistics in R using the package CircMLE, including all 10 models available in the package [33]. Mann–Whitney U tests were used to test for differences in the length of the mean vectors of both experimental groups (GraphPad Prism V. 8.4.3).

3. Results and discussion

In the first series of experiments performed in 2006, we tested six mole-rat pairs before and after enucleation using the nest-building assay in the ambient magnetic field (table 1). Before enucleation, the animals displayed a strong preference for the south-eastern sector of the arena (figure 1a; mean direction: 124°, 95% confidence interval (CI): 82°–188°, mean vector length r = 0.549, n = 6, Moore's modified Rayleigh test: R* = 1.231, p < 0.01), in line with published magnetic preferences in Ansell's mole-rats [6,15,17,19]. By contrast, the distribution of nests built after enucleation was random (figure 1b, r = 0.140, n = 6, Moore's modified Rayleigh test: R* = 0.647, p > 0.1). These results suggested an impairment of the magnetic sense but did not allow us to exclude topographic factors or series effects.

Figure 1.

Pilot experiments showing a loss of directional preferences in the Ansell's mole-rat after enucleation. Nest distribution of six pairs of Ansell's mole-rats (a) before (control) and (b) after surgery (enucleation). Control trials (n = 4 per pair) exhibited a significant preference to build nests in the magnetic southeast, while the nests built by the same pairs after surgery (n = 6 per pair) were distributed randomly. The small blue arrows represent the mean vectors of each pair of Ansell's mole-rats and the arrow lengths reflect the r-values, a measure of the concentration of the nests. The red arrows are the weighted mean vector calculated over the mean vectors of tested pairs when significant; 95% confidence intervals of the weighted grand mean are indicated by dashed lines. The dots outside of the circles indicate the positions of all nests of each experimental group (controls: 24, enucleated: 36). The p-values indicate the results from Moore's modified Rayleigh tests performed on the mean vectors.

To address these questions, we conducted a second series of experiments in 2018–2019, in which we tested a new cohort of enucleated (n = 10) and control (n = 10) animals with a coil set-up that precisely controlled the magnetic field (table 1). Furthermore, to minimize unspecific effects of the surgery, the behavioural experiments were performed more than 1.5 years after enucleation. We ascertained whether the enucleated animals behaved normally by recording ethograms of the experimental and control subjects in their home enclosures. The measurements of the six quantified behaviours revealed no significant differences between enucleated subjects and controls (figure 2; binomial GLM: t = −0.536, p = 0.594), indicating that enucleation does not impact routine behaviour in captive settings.

Figure 2.

Enucleation does not affect the general behaviour of Ansell's mole-rats. Mean time budgets of six behaviours of enucleated and control Ansell's mole-rats observed in their home enclosures. All behaviours are expressed as the percentage of time spent per observation period. There were no significant differences between enucleated subjects and controls (binomial GLM for proportional data (behaviour × treatment): t = −0.536, p = 0.594). n = 6 (enucleated) and n = 8 (controls). Error bars represent the standard deviation.

In the nest-building assay, all animals were tested in four different magnetic field alignments to distinguish magnetic from topographic orientation responses. As in the first experimental series, control animals displayed a significant preference for the magnetic south-eastern sector of the arena (figure 3a; mean direction: 172°, CI: 119°–222°, r = 0.441, n = 10, Moore's modified Rayleigh test: R* = 1.268, p < 0.01). By contrast, the magnetic distribution of nests built by enucleated mole-rats was indistinguishable from random (figure 3b; r = 0.129, n = 10, Moore's modified Rayleigh test: R* = 0.519, p > 0.5). This difference was also expressed by the significantly shorter mean vectors of enucleated compared to control animals (figure 3c; median r_enucleated: 0.243, median r_control: 0.637, n = 10, one-tailed Mann–Whitney test: U = 27.5, p = 0.045). With respect to topographic north, the nest directions of the controls did not deviate from a random distribution (figure 3d; r = 0.062, n = 10, R* = 0.396, p > 0.5). The nest directions of enucleated animals with respect to topographic north, however, did not appear random, so we used a model-fitting approach to identify the most likely underlying distribution. Three distributions were almost equally likely, with two of them being axial distributions (electronic supplementary material, table S1). Indeed, testing for an axial distribution using Moore's modified Rayleigh test on doubled angles revealed a preference to build nests along the topographic north–south axis in the enucleated group (figure 3e; mean direction = 176°/356°, CI: 272°–58°, r = 0.429, n = 10, R* = 1.302, p < 0.01). Further, the mean vectors with respect to topographic north were significantly longer in the enucleated animals (figure 3f; median r_enucleated: 0.569, median r_control: 0.287, one-tailed Mann–Whitney test: U = 18, p = 0.007). These findings demonstrate that magnetic cues guided nest building in control animals, whereas enucleated animals appeared unable to perceive the magnetic field, leading them to fall back on individual topographic preferences. How such a topographic bias could emerge in the stimulus-deprived environment of the testing arena remains puzzling. Still, since there was no sign of a similar topographic preference in the controls, magnetic stimuli evidently represent more salient cues for animals with intact eyes.

Figure 3.

Enucleation results in a loss of magnetic directional preferences in the Ansell's mole-rat. (a,b) Nest distribution with respect to magnetic north (mN) of controls (a) and enucleated Ansell's mole-rats (b). (c) Mean angular vectors (calculated for each animal separately with respect to magnetic north) for controls and enucleated mole-rats. (d,e) Nest distribution with respect to topographic north (tN) in control (d) and enucleated mole-rats (e). (f) Mean angular vectors (with respect to topographic north) for controls and enucleated mole-rats. Small blue arrows represent the mean vectors of four nests built by individuals or pairs of Ansell's mole-rats and the arrow lengths reflect the r-values, a measure of the concentration of the nests. Red arrows represent weighted mean vector calculated over the mean vectors of tested individuals/pairs when significant. Dashed lines indicate 95% confidence intervals of the weighted grand mean. Dots outside the circles: positions of all 40 nests of each experimental group. The p-values indicate the results from Moore's modified Rayleigh tests (a,e) or one-tailed Mann–Whitney tests (c,f) performed on the mean vectors.

In summary, we found a loss of magnetic directional preferences in two independent series of nest-building experiments, yet we did not detect other behavioural consequences of enucleation in mole-rats. All enucleated animals were fully immersed members of their respective family groups and many successfully bred and raised offspring. Enucleated subjects were equally motivated to build nests, arguing against stress or pain-related side effects of surgery, since these reduce nest-building behaviour in rodents [34]. As we carefully controlled for the influence of other sensory stimuli, we conclude that the removal of the eyes led to a permanent impairment of the magnetic sense. The effect of enucleation on magnetic orientation is comparable to lidocaine application onto the cornea, corroborating the conclusion by Wegner et al. [6] that this type of anaesthesia affected the eyes rather than adjacent tissues or the central nervous system. Thus, future screens for light independent, probably magnetite-based magnetoreceptors in this species should focus on the minute eyes of Ansell's mole-rat which are small enough (approx. 2 mm in diameter) to be investigated entirely by techniques such as high-throughput electron microscopy [35,36].

Due to the fact that enucleation has no perceivable influence on mole-rats' behaviours apart from affecting magnetoreception, one might be tempted to speculate that the small but structurally complex eyes of these animals are still retained because of their involvement in sensing magnetic fields. It is unlikely, however, that ocular magnetoreception necessitates the eyes to display such a high degree of structural complexity. First, in contrast with light-dependent radical pair-based receptors [11], light-independent magnetite-based receptors do not require specific spatial receptor arrangements to function [12,37]. Second, the Middle East blind mole-rat (Spalax ehrenbergi), a distantly related underground-dwelling rodent of the muroid superfamily, has already been shown to respond to magnetic stimuli in total darkness, despite its highly regressed, subcutaneous eyes [38]. Whether impairments due to eye loss in free-living Ansell's mole-rats would be more severe, for example during surface activity, remains speculative since little is known about the behavioural ecology of this species in the wild. Why African mole-rats retain a structurally complex eye, while many other subterranean mammals do not, continues to be elusive [23].

There have been past attempts to detect magnetite in various rodent tissues [39] but, to our knowledge, no detailed surveys for ocular magnetite have been conducted in any mammal species so far, including African mole-rats. Electron-dense crystalloid bodies were coincidentally noticed in retinal photoreceptors of Ansell's mole-rats, assumedly consisting of magnetite [40]. However, neither were the size and distribution of these particles quantified, nor has it been investigated whether they indeed contain iron. Later, ferric aggregates were identified by Prussian blue staining in the cornea of a single mole-rat eye, which have also been interpreted as magnetite crystals [6]. This intriguing, but non-replicated finding should be interpreted with caution because Prussian blue staining does not specifically detect magnetite. For example, iron-rich cells in the avian upper beak, once hypothesized to be magnetoreceptors and discovered via this method, were subsequently identified as macrophages, immune cells that accumulate ferric iron [41]. Although macrophages are typically absent from the rodent cornea, they invade corneal tissue during inflammation [42]. Besides that, the finding could be attributed to iron contamination from the laboratory environment, a familiar confounder in histological screens for magnetoreceptors [43].

If magnetite receptors are located in the cornea, they would most likely be innervated by the ophthalmic branch of the trigeminal nerve [44]. This would be consistent with the finding of magnetically induced neural activity in a part of the Ansell's mole-rat's superior colliculus that predominantly receives trigeminal input [7]. The trigeminal nerve has further been demonstrated to be involved in the magnetic sense in birds [9,45–48], but the exact location and structure of avian trigeminal magnetoreceptors is equally unknown [26]. The presence of similar trigeminal magnetoreception systems in birds and mammals appears conceivable, but ophthalmic nerve ablation experiments coupled with behavioural assays are needed to establish the role of trigeminal input for magnetoreception in mole-rats. Our study highlights the significance of the eye for mammalian magnetoreception, which could facilitate future research on its cellular basis. A thorough screen for magnetite in the mole-rat eye using electron-microscopic and spectroscopic methods is a warranted future experiment. A further candidate group to search for ocular magnetoreceptors besides mole-rats would be insectivorous bats, which are also microphthalmic and rely on a similar magnetic polarity compass system, probably based on magnetite [13]. By contrast to mole-rats, several bat species are widespread laboratory models in sensory biology and neurophysiology (e.g. Carollia perspicillata), their nervous systems are well characterized and they are readily available for a wide range of physiological methodologies. Ultimately, such experiments will contribute to characterize the function and significance of this enigmatic sensory channel within the mammalian radiation.

Ethics

All surgeries and experiments conformed to the relevant ethical standards and were approved by the animal welfare officer of the University of Duisburg-Essen and the LANUV NRW, Germany (series 1: 50.05-230-37/06; series 2: 84-02.04.2015.A387).

Data accessibility

The data are included in table 1.

Authors' contributions

R.E.M. and P.N. designed the experiments of series 1; S.B. designed the experiments of series 2; P.N. conducted the surgeries (series 1); R.E.M., K.M., K.R.C. and S.B. conducted the experiments; K.R.C., E.P.M., R.E.M., K.M. and S.B. analysed the data; E.P.M. created the figures; K.R.C., E.P.M. and S.B. drafted the manuscript with input from all others authors.

Competing interests

The authors declare no competing interests.

Funding

K.R.C. was supported by a PhD fellowship of the German National Academic Foundation (Studienstiftung des deutschen Volkes). This project was partly funded by the grant ‘EVA4.0', no. CZ.02.1.01/0.0/0.0/16_019/0000803 financed by OP RDE of the European Union and the Ministry of Education, Youth and Sport of the Czech Republic (to S.B.).