MEMRI for visualizing brain activity after auditory stimulation in frogs

Abstract

Anuran amphibians are common model organisms in bioacoustics and neurobiology. To date, however, most available methods for studying auditory processing in frogs are highly invasive and thus do not allow for longitudinal study designs, nor do they provide a global view of the brain, which substantially limits the questions that can be addressed. The goal of this study was to identify areas in the frog brain that are responsible for auditory processing using in vivo manganese-enhanced magnetic resonance imaging (MEMRI). We were interested in determining if the neural processing of socially relevant acoustic stimuli (e.g. species-specific calls) engages a specific pattern of brain activation that differs from patterns elicited by less- or non-relevant acoustic signals. We thus designed an experiment, in which we presented three different types of acoustic stimuli (species-specific calls, band-limited noise, or silence) to fully awake Northern Leopard frogs (Rana pipiens) and then conducted MEMRI T1-weighted imaging to investigate differences in signal intensity due to manganese uptake as an indication of brain activity across all three conditions. We found the greatest change in signal intensity within the torus semicircularis (the principal central auditory region), the habenula, and the paraphysis of frogs that had been exposed to conspecific calls compared to noise or silence conditions. Stimulation with noise did not result in the same activation patterns, indicating that signals with contrasting social relevance are differentially processed in these areas of the amphibian brain. MEMRI provides a powerful approach to studying brain activity with high spatial resolution in frogs.

Introduction

Acoustic signals play a vital role in animal life-history. Many animals make use of sound or vibrational cues to localize conspecifics such as mating partners or kin (Simmons, Popper, & Fay, 2003; Suthers, Fitch, Fay, & Popper, 2016), during foraging (Bernal, Page, Rand, & Ryan, 2007; Fitzsimmons, Foote, Ratcliffe, & Mennill, 2008; Narins, Lewis, Jarvis, & O’Riain, 1997), to identify predators (Anton, Evengaard, Barrozo, Anderson, & Skals, 2011), or to retrieve information about the emotional state from group members (Blumstein & Recapet, 2009; Szipl, Ringler, & Bugnyar, 2018). However, in their everyday life, animals are exposed to a continuous barrage of environmental sound cues (Caldart, Iop, Lingnau, & Cechin, 2016; Forrest, 1994; Narins & Zelick, 1988; Ruhl & Dicke, 2012), and therefore, for efficient communication, attention should primarily be paid to biologically relevant stimuli (Brumm, Voss, Köllmer, & Todt, 2004).

Acoustic communication is a key element of social interactions in anuran amphibians. Males often produce advertisement signals to attract female mating partners and to repel male competitors (Gerhardt & Huber, 2002; Simmons, Popper, & Fay, 2003). In several studies it has been shown that females preferentially respond to species-specific signals (Braaten & Reynolds, 1999; Ryan & Rand, 1993), using either spectral (Schwartz & Gerhardt, 1998) or temporal (Lopez & Narins, 1991) characteristics for cue recognition. However, the cognitive mechanisms by which individuals differentiate between different sound sources are poorly understood. While much research has been done on signal processing across multiple sensory domains in mammals including humans (Petersen & Posner, 2012), highlighting the prominent role of the neocortex in attention and perception (Corbetta & Shulman, 2002), relatively little is known about signal processing in species that lack such cortical structures (Krauzlis, Bogadhi, Herman, & Bollimunta, 2018).

Anuran amphibians (frogs and toads) have long been recognized as excellent model organisms in bioacoustics and neurobiology (Bee, 2015; Capranica, 1965; Dicke & Roth, 2009; Gerhardt & Huber, 2002; Narins, Feng, Fay, & Popper, 2007), thanks to their relatively simple brain organization and stereotypic calls. Therefore, much is known about the anatomy of the amphibian ear and ascending auditory pathways involved in signal processing; for reviews of the amphibian ear, see (Simmons, Meenderink, & Vassilakis, 2007; van Dijk, Mason, Schoffelen, Narins, & Meenderink, 2011). To date, however, most available methods for studying auditory processing in frogs are highly invasive and often require euthanizing animals, and thus do not allow for a longitudinal study design. Nor do they provide a global view of the brain, which substantially limits the types of questions that can be addressed. Thus in this study, we aimed to evaluate the suitability of magnetic resonance imaging (MRI) for studying brain activity in frogs, in particular for identifying areas that are responsible for auditory processing. Specifically, we wanted to determine if neural processing of socially relevant and irrelevant acoustic stimuli engage different patterns of brain activation.

We designed an experiment in which we presented three types of acoustic conditions (species-specific calls, band-limited noise, silence) to fully awake Northern Leopard frogs (Rana pipiens) and subsequently examined neuronal brain activity using manganese-enhanced magnetic resonance imaging (MEMRI; Silva, Lee, Aoki, & Koretsky, 2004). We predicted that signal enhancement in the primary midbrain auditory region, the torus semicircularis (TS), would be significantly elevated in the “socially relevant” condition, but not when frogs were exposed to noise or silence. Alternatively, perception of socially relevant cues may also be facilitated via the activation of additional, non-auditory brain areas, not specifically associated with auditory processing, for example the amygdala (Hall, Ballagh, & Kelley, 2013).

Most studies using fMRI in rodents use BOLD-fMRI, which does not need the application of any contrast agent, however, it is often hampered by its comparatively low spatial resolution and high background noise produced by the scanner. Manganese-enhanced MRI (MEMRI) is a powerful tool with which to study in-vivo neuronal activity in small animals (Koretsky & Silva, 2004; Silva et al., 2004; Herberholz et al., 2011). This technique makes use of the fact that paramagnetic manganese ions (Mn²⁺) shorten the proton relaxation time constant (T1) in tissue where they accumulate, leading to signal enhancement in the respective areas on T1-weighted MRI images (Silva et al., 2004). Furthermore, Mn²⁺ acts as a biological Ca²⁺ analogue, as it can enter neurons via active voltage-gated calcium channels during neuronal depolarization and it can be transported to adjacent neurons (Narita, Kawasaki, & Kita, 1990; Silva et al., 2004; Watanabe, Frahm, & Michaelis, 2004). Mn²⁺ can be administered as a Manganese-chloride solution (MnCl₂) either systemically or directly into the tissue of interest (Watanabe et al., 2004). Tissue contrast enhancement usually reaches its final pattern 24 h after MnCl₂ administration, and due to the slow degradation time of Mn²⁺, the relative distribution of contrast remains relatively constant for several hours. Highly active neurons will lead to a greater accumulation of Mn²⁺ ions within tissue responding to experimental sensory stimulation, resulting in a change in magnetic resonance (MR) signal intensity relative to surrounding tissue that is not activated (Lin & Koretsky, 1997; Silva et al., 2004). Thus the observed signal intensity directly reflects neuronal activity in the brain from the prior period of sensory experience. This methodology is of particular value for investigating auditory-related activity (Lee, et al., 2007; Watanabe, Frahm, & Michaelis, 2008; Yu et al., 2008; Yu, Wadghiri, Sanes, & Turnbull, 2005), since compared to evoked activation elicited inside the MR spectrometer (e.g. blood-oxygen-level-dependent-functional MRI; Amaro et al., 2002), MEMRI images are not confounded by scanner noise.

MEMRI has proven particularly useful in several previous studies that investigated the function of the auditory system. For example, it enabled to delineate auditory pathways in free-moving rodents (Kim et al., 2014; Lee et al., 2007; Lee, Joong et al., 2012; Watanabe et al., 2008; Yu et al., 2005; Yu et al., 2008), to demonstrate tonotopic organization of the auditory cortex (Yu et al., 2005), and to map sound-evoked activity in the inferior colliculus of young mice during development (Watanabe et al., 2008; Yu et al., 2008). Moreover, by directly injecting Mn²⁺ into the cochlea, Lee et al. (2007) demonstrated differential signal enhancement in auditory centers between mice with and without high-level acoustic stimuli. To date, nearly all published functional MRI protocols have been restricted to birds and mammals. The aim of the present study was to (1) assess the suitability and applicability of MEMRI for research on auditory processing in the anuran brain; and (2) investigate differences in brain activity in response to socially relevant and irrelevant sound, respectively.

Materials and Methods

Ethics

All animal procedures were performed in accordance with NIH standards and followed the Public Health Service Policy on Humane Care and Use of Laboratory Animals. The experimental protocol was approved by the Animal Research Committee of the University of California Los Angeles (ARC-protocol 1994-086-73, PI: Peter Narins) and the Ahmanson-Lovelace Brain Mapping Center at UCLA. All efforts were made to minimize the number of animals studied. Animals did not exhibit any abnormal behavior after scanning and/or manganese administration; thus additional analgesic treatment was not necessary. At the end of the experiments, all frogs were euthanized via immersion in a 1% solution of MS-222 (Tricaine-S, Western Chemical Inc.) followed by the double pithing procedure.

Animals

In the present study, we used eight (six females, two males) Northern leopard frogs (Rana pipiens), with body weights of 52.4–93.9 g (mean ± sd = 75.3 ± 14.3 g). All frogs were housed in the vivarium of the Division of Laboratory and Animal Medicine, UCLA. The animals were maintained at an ambient temperature of 21°C, and in a 12-h light/12-h dark cycle. Frogs were group-housed in single plexiglass tanks that were equipped with stones and artificial plants for hiding and climbing, and an automatic water-flow system. Frogs were fed three times a week with mealworms. Individual frogs could be identified via their unique dorsal pattern. Changes in body weight of the frogs over the course of the study were tested using paired t-tests.

Mn²⁺ administration

For the MEMRI scans, frogs were imaged before and after receiving a systemic dose (0.4 mmol/kg body weight) of MnCl₂ solution via a single intraperitoneal injection. MnCl₂ clears from the system in endotherms in approximately 2–3 weeks (songbirds: van der Linden, A. et al., 2002, mice: Yu et al., 2005). No reference data were available for ectotherms; however, we reasoned that retention times might be longer in ectotherms, due to their generally slower metabolic rates. Nevertheless, similar pause intervals to those used in previous studies of endotherms were chosen and these yielded satisfactory results of 109.3% signal change on average. For each trial, frogs were initially scanned pre-treatment to acquire baseline images that account for any remaining MnCl₂ from previous tests.

Stimulus presentation

After administration of the MnCl₂, frogs were transferred to custom-made, acoustically isolated chambers (57x39x39 cm) where they received either no stimulus (“Silence” - control condition), conspecific calls (“Call”), or band-limited noise (“Noise”). To avoid desiccation, frogs were placed in plastic boxes (29.8x19.7x20.3 cm), in which the floor was covered with paper towels and filled with 500 ml of dechlorinated tap water. The box was covered with a thin nylon mesh to prevent frogs from escaping, and to minimally affect the sound signal measured at the frog. An audio speaker (Pioneer, TS-G1645R) was attached to the center of the lid of the acoustic chamber.

Recordings of R. pipiens calls were obtained from the Macaulay Library at the Cornell Laboratory of Ornithology (ML 182027; Ithaca, NY), and four different calls of a single individual were selected for the stimulation. The four calls were broadcast in a pseudo-random order until all four calls had been presented. Pauses between calls lasted 5, 10, 15, or 20 s and were also presented pseudo-randomly to minimize behavioural habituation to the calls (Figure 1). Band-limited noise (0.1-2.5 kHz) covering the frequency range of the selected R. pipiens calls was presented in the same temporal pattern and pseudo-randomized order as the calls. After completion of a full sequence (four calls/noise and four pauses), a newly generated full sequence was presented to the frogs. These pseudo-randomized sequences of sound were generated using a script written in Matlab (Mathworks, Natick, MA, USA). Animals were exposed to auditory stimulation (or silence) for 30 h, with peak sound levels adjusted to 95 dB SPL measured at the position of the frog. All eight frogs were tested in all three experimental conditions in a pseudorandomized design. Consecutive tests were separated by a minimum of 21 days, to assure that contrast levels had returned to baseline and to allow animals to recover from any stress due to the experimental procedure. To obtain baseline measurements of signal intensity and to also account for any remaining MnCl₂ in the brain, we also conducted control pre-Mn²⁺ scans before each experimental stimulation (including the control condition). Differences in contrast between the pre-treatment and post-treatment images show the overall signal enhancement due to Mn²⁺ uptake in the brain. Comparing differences in Mn²⁺ uptake (post- minus pre-images) across the three test conditions (call, noise, silence) allowed us then to elucidate the areas of the frog brain involved in processing of the presented auditory stimuli.

Figure 1.

Oscillograms (top) and corresponding spectrograms (botttom) of the presented stimuli. A) the species’ own call, B) noise filtered to the frequency range of the species’ own call. The figure displays one example of a randomized sequence of four calls (Call 1–4) or noise (Noise 1–4) and pause intervals (5 s, 10 s, 15 s, 20 s).

Manganese-enhanced magnetic resonance imaging (MEMRI)

High-resolution, T1-weighted, two-dimensional (2D), spin-echo images were obtained before and 27-30 hours after intraperitoneal injection of MnCl₂ solution and sound stimulation. For scanning, animals were anesthetized by immersion in a 0.2% solution of MS-222 for 10 min. After induction of anesthesia, frogs were transferred to a custom-built cradle for correct positioning in the center of the magnet (Figure 2). Animals were wrapped in wet gauze to facilitate cutaneous respiration and prevent physiological stress and overheating. A sealed glass capillary containing 0.05 mM MnCl₂ solution was placed on top of the surface coil, for use as an external standard for signal calibration (subsequently referred to as “phantom”). This facilitated signal comparison between scans for each frog and between different frogs within and between each group. The surface coil was covered with a finger piece cut from a nitrile glove to prevent any signal distortion caused by the moisture of the frogs. All imaging was performed on a 7T Bruker MRI horizontal small animal spectrometer (Oxford Instruments, Carteret, NJ, USA) using a 72 cm birdcage transmit coil and an actively decoupled receive-only, single-turn surface coil. Data acquisition was conducted using Paravision 5.2 software. Following a multi-slice, gradient echo pilot scan to confirm correct positioning of the brain within the magnet isocenter, localized shimming was performed on the head to improve B0 homogeneity. T1-weighted images were acquired using a modified driven equilibrium Fourier transform (MDEFT) sequence with Echo Time = 4 ms, Repetition Time = 4000 ms, Inversion Time = 1000 ms, Flip Angle = 20°, bandwidth = 50 kHz and 6 averages per phase-encoding increment. Sixty-four, 0.125 mm-thick sagittal slices were acquired using a data matrix of 192 x 192 x 64 within a 24 x 24 x 8 mm field-of-view, resulting in scans with an isotropic resolution of 125 µm. Total scanning time was 1 h 42 min.

Figure 2.

Experimental setup of frogs inside the small-animal MRI scanner. A) Frogs were transferred to a custom-built cradle and placed on wet paper towels. B) frogs were then wrapped in wet gauze to facilitate cutaneous respiration and prevent physiological stress and overheating. C) A sealed glass capillary tube containing 0.05 mM MnCl₂ solution was placed on top of the surface coil, for use as an external standard for signal calibration. D) The surface coil was covered with a finger piece cut from a nitrile glove to prevent any signal distortion caused by the moisture of the frogs.

At the end of the imaging session the animals were immediately removed from the magnet and allowed to awake in a recovery box. After full recovery, frogs were returned to their home tank. Frogs were retested with another stimulus condition after a minimum recovery period of three weeks. All eight frogs were tested in all three experimental conditions in a pseudo-randomized order, allowing each individual to act as its own control. No individual received the same stimulus condition twice.

Image processing and analysis

All image processing and whole-brain voxel-wise analysis was conducted using FSL tools (Jenkinson, Beckmann, Behrens, Woolrich, & Smith, 2012; Smith et al., 2004). We normalized all images regarding spatial orientation and intensity differences resulting from different scanning sessions as follows: the greyscale of each brain volume was adjusted so that the average signal intensity of the pixels within the phantom volume of interest was 1000. Then all brain-extracted image data were spatially co-registered to one brain that was randomly selected from the population of 8 frogs and used as a template volume. The FSLtools ‘flirt’ function was used to achieve this using affine transformations with 12 degrees of freedom and using a correlation ratio cost function weighting and sinc interpolation (Jenkinson, Bannister, Brady, & Smith, 2002). After this normalization procedure, all “pre” images were subtracted from the corresponding “post” images, to create standardized images of relative signal enhancement. These normalized data were then used for voxel-wise statistical comparison. A binary mask inclusive of the whole brain was manually created from the template brain in order to limit a voxel-wise statistical comparison to the brain.

We first ran an ANOVA to determine statistical comparison of group-level differences (silence, call, noise). The FSLtool ‘randomise’ function (FSL's tool for nonparametric permutation inference on neuroimaging data, Winkler et al., 2014) was used to do this with 500 permutations and threshold-free cluster enhancement to detect significant clusters of signal difference (p<0.05). After achieving overall group-level significance, a design matrix was constructed to test for statistical inference of post-hoc differences between groups (silence-call, call-noise, silence-noise) using voxel-based thresholding. All tests were conducted in a paired design in order to account for repeated measures of single individuals. A variance smoothing function of 5 mm was used to improve statistical sensitivity to detect differences spanning multi-voxel clusters rather than merely single voxel spikes of high intensity differences. A p-value <0.05 corrected for multiple voxel comparisons and multiple testing was accepted as significantly different. The regions of statistically different voxels were displayed as overlays on to the template brain image for comparison to the brain atlas of R. pipiens (Wada et al., 1980). Only clusters of statistically significant voxels that matched distinct brain areas outlined in the brain atlas were interpreted as significant changes in brain activity. Single or scattered voxels that did not match any distinct brain region were interpreted as artefacts due to minor defects in image co-registration.

In order to make a secondary, univariate comparison of distinct regions of interest (ROIs) revealed by the whole-brain analysis, we calculated average signal intensities of clusters identified from the whole brain analysis. Manual brain extraction and segmentation of certain regions-of-interest as identified by the whole-brain analysis, for example the torus semicircularis (TS), habenula (HB), and paraphysis (PP) were performed using the software ITK-Snap (Yushkevich, Gao, & Gerig, 2016), following the detailed brain atlas for R. pipiens (Wada, Urano, & Gorbman, 1980). Results were then compared using Kruskal-Wallis in RStudio (RStudio Team, 2015). Significant results in these omnibus tests were then compared using post-hoc tests corrected for multiple-ROI comparisons using the false-discovery rate (fdr) correction (Benjamini & Hochberg, 1995). ROI differences in normalized MRI signal were compared using the Wilcoxon-signed-rank test.

Results

All frogs remained in good physiological condition during the entire study period (July 2017 to January 2018), with no significant decrease in body weight over time (paired t-test, n = 8, t(7) = -0.147, p = 0.887). Anesthesia typically lasted for 4–5 h, and none of the frogs awoke during imaging. We did not observe any adverse effects on the health of the frogs due to either manganese injections or to long-term auditory stimulation.

The F-tests revealed significant differences in relative signal intensities across test conditions (p<0.05). The post-hoc tests then corroborated that there were statistically significant differences in relative signal intensity only between the conditions “call” and “silence” (p<0.05). Compared to signal enhancement without explicit sound stimulation, we found significantly stronger signal enhancement in the TS, HB, and the PP, in frogs that received socially-relevant calls compared to frogs that were exposed to silence (Figure 3 & 4). The difference in signal enhancement across different brain areas after noise stimulation compared to either silence or calls did not survive statistical correction (i.e. not statistically significant).

Figure 3.

Brain sections of R. pipiens showing significant increase of MEMRI signal in response to calls compared to silence. The greyscale image represents the highest quality frog brain that was used as an image co-registration template. The color pixel overlays indicate statistically significant differences (p<0.05) of normalized voxel signal intensities between the two conditions “call” and “silence”. A) Coronal section; B) Sagittal section; C) Axial section; D) Axial section (anatomical image). Significant signal enhancements within the torus semicircularis (TS) and the paraphysis (PP) are highlighted.

Figure 4.

T1-weighted MR images showing relative increase in MEMRI signal in A) the torus semicircularis (TS), B) the habenula (HB), and C) the paraphysis (PP) after pairwise comparisons of the test conditions “call” vs. “silence” (color pixels: p<0.05).

Figure 5 shows the difference in average signal intensity (arbitrary value due to image normalization) in the TS, HB, and PP across all three conditions (Kruskal-Wallis test, TS: X²(2) = 9.555, p = 0.008, HB: X²(2) = 10.535, p = 0.005, PP: X²(2) = 9.06, p = 0.01). Pairwise post-hoc comparisons corroborated significant signal intensity enhancement in frogs that received the socially relevant call compared to silence across all three ROIs (n₁ = n₂ = 8, TS: p = 0.006, HB: p = 0.04, PP: p = 0.006), while no statistically significant difference was found between “call” and “noise” (all p>0.3). The comparison between “noise” and “silence” revealed significant differences in signal intensity only in the TS (p = 0.02) and the HB (p = 0.04), but not the PP (p = 0.1).

Figure 5.

Boxplots showing differences in mean signal intensity (arbitrary value due to image normalization) measured in the torus semicircularis (TS), habenula (HB), and paraphysis (PP) across all three experimental conditions.

Discussion

In this study, we used MEMRI for detecting activity patterns elicited by sound-stimulation in brains of anuran amphibians. Specifically, we asked how the frog brain responds to both socially relevant and irrelevant acoustic cues. Our results clearly demonstrate manganese-related signal enhancements in the auditory and reward processing regions of the brain after intraperitoneal administration of MnCl₂ and sound stimulation with the species-specific advertisement call.

As expected, we found the TS, the principal auditory structure in the amphibian brain (Wilczynski & Endepols, 2007), to be significantly activated in individuals exposed to conspecific calls (Figure 4A). The dominant role of this brain region in encoding temporal and spectral properties of auditory signals has been demonstrated, mainly by electrophysiological approaches, in a wide range of anuran species (Arch, Grafe, Gridi-Papp, & Narins, 2009; Penna, Lin, & Feng, 1997; Walkowiak, 1980). Activation appeared mainly restricted to the principal nucleus of the TS (cf. Feng, 1983). The laminar and magnocellular nuclei showed comparatively lower activation patterns. Response selectivity of different torus neurons when presented with sound stimuli of different social relevance is known mainly from electrophysiological studies (Penna, Lin, & Feng, 2001). Two studies in Physalemus pustulosus using immediate early gene expression analysis found a functional heterogeneity of processing of mating calls within the torus. While Hoke et al. (2004) show differential induction in response to the acoustic stimuli in the laminar, midline, and principal, but not the ventral, nuclei of the torus semicircularis, Mangiamele & Burmeister (2011) found that the behavioral relevance of presented stimuli was the best predictor of egr-1 expression in the laminar nucleus, but not elsewhere. Future comparative studies across different anuran species are needed to elucidate the functional role of different regions in the torus semicircularis.

We also found that the HB as well as the PP showed significant signal enhancement after auditory stimulation with the calls compared to silence. The bilaterally paired habenular nuclei are located in the dorsal diencephalon and project to midbrain areas that are involved in both serotonin and dopamine release (Beretta, Dross, Gutierrez-Triana, Ryu, & Carl, 2012). Despite its widespread presence in nearly all vertebrate species, which points to a highly conserved network structure (cf. Bianco & Wilson, 2009), the HB previously received only limited research interest. Only recently, was it suggested that the HB might play a prominent role in cognition and reward processing in rats (Andres, Düring, & Veh, 1999; Bianco & Wilson, 2009). Studies using fMRI in humans (Ullsperger & Cramon, 2003) and single unit electrophysiology in rhesus monkeys (Bromberg-Martin & Hikosaka, 2011; Matsumoto & Hikosaka, 2007) have closely linked the function of the lateral HB with the encoding of motivational values, particularly negative ones. In line with these results are findings in zebrafish where the HB was found to be involved in social conflict resolution (Chou et al., 2016). Also in mice, the lateral HB was shown to be strongly activated when animals were exposed to bitter taste, pain, social attack by aggressors, and social defeat (Wang et al., 2017). Further support for the idea of the HB might be involved in conspecific call recognition in amphibians comes from a study in water frogs, Rana esculenta (Kemali, Guglielmotti, & Fiorino, 1990). This study showed that the HB of frogs were larger in spring, when frogs are reproductively active, compared to winter months, and this seasonal change in HB size was most noticeable in females. The authors suggest that hearing calls of conspecifics induces hormonal changes in females that might finally lead to this change in HB size (Kemali, Guglielmotti, & Gioffre, 1980). But despite these few examples of the mechanistic knowledge of this brain structure, the HB remains a little-known structure compared to other areas of the vertebrate auditory pathway, whose functions are yet to be fully elucidated (Hikosaka, 2010; Mizumori & Baker, 2017). The elevation of the HB in our experiment could indicate that hearing conspecific calls continuously for 30 h without the possibility of seeing or approaching that sound source may have led to individual distress that triggered higher activation in this brain region. Further research is needed to clarify the eventual function of this brain region regarding signal evaluation.

Our experiments also revealed a significant increase in signal intensity after stimulation with socially relevant calls compared to silence in the PP of frogs. The PP is a glandular structure found in the third ventricle of lower vertebrates and is well developed in amphibians and reptiles. The function of the gland is not well understood, but it may play a vital role in calcium metabolism (Nelson, Foltz, Camarata, & Sarras, 1985). Given that paramagnetic manganese was used as a calcium analogue in this study, the significant increase in signal intensity in the PP after stimulation with conspecific calls could possibly be due to an overall increase in blood pressure, and thereby increased Mn²⁺ transport into the PP, caused by a generalized increase in arousal in treatments with call stimulation – which would also be in line with the elevated signal found in the HB. Although eventual differences in the permeability of the blood-brain-barrier (BBB) close to the PP could have led to an accumulation of Mn²⁺ in this area, our experiment only detected increased signal enhancement in the PP in conditions with explicit sound stimulation. Therefore, we can exclude a sole methodological or anatomical explanation for the increased signal in the PP. Further studies are needed to investigate specific features of the BBB in the anuran brain. The few voxels in the optic tectum that show significant differences in signal intensity between the call stimulation and silence (Figure 3 & 4), are likely to be artefacts due to image registration rather than differences in brain activity between the two conditions.

Surprisingly, we did not find any significant enhancement in signal intensity in the amygdala, although this brain region is generally known to be important for processing of memory-based, decision-making and emotional responses in vertebrates (Emery & Amaral, 2000; Newman, 1999) and was also previously found to be important for discrimination of acoustic stimuli regarding their social contexts in Xenopus (Hall et al., 2013).

Interestingly, significant signal enhancements in TS, HB, and PP were only observed when comparing frogs exposed to “call” versus “silence”. Stimulation with noise did not result in statistically significant signal enhancement in the voxel-based whole-brain analysis, compared to call or silence conditions (Walkowiak, 1980). Theoretically, this could have been due to differences in the acoustic properties of the signals. The “call” stimulus had a bimodal distribution with the energy concentrated in 0.2-0.7 kHz and 1-1.5 kHz, while in the Noise stimulus the energy was evenly distributed from 0.1 to 2.5 kHz. In our experiments, the peak amplitude of each stimulus was 95 dB SPL, but because of the differences in their spectral structure, the amplitude of the frequency bands had less energy (6-12 dB) when presenting the “noise”. However, we don’t think that these subtle differences in spectral energy can explain the significant increase in signal intensity in response to conspecific vocalizations compared to noise stimuli, as in general very high levels of sound (95 dB SPL) were set for all stimulus conditions. Moreover, the less conservative test of comparing average signal intensities in the different ROIs showed statistically significant increase in brain activation also after noise stimulation compared to silence in both the TS and the HB, although differences between “call vs. silence” were much more pronounced.

A higher selectivity for ecologically relevant signals compared to other acoustic stimuli has been also described in the central auditory system of birds, frog and bats (Fuzessery & Feng, 1983; Theunissen et al., 2004; Wohlgemuth & Moss, 2016). Specifically, in the thalamus of R. pipiens, (Fuzessery & Feng, 1983) identified a group of neurons that exhibit selectivity for behaviorally relevant tone combinations. Most of those multi-tone responder neurons did not respond to broadband noise but showed increased excitation when tone combinations coincide with the frequencies of the call. These authors suggested that the energy in additional frequencies may exert an inhibitory influence over some auditory neurons. In support of this idea are findings from playback experiments with mating calls of R.pipiens and a sympatric species, Rana blairi (Kruse, 1981). Unlike the bimodal distribution of spectral energy in the calls of R. pipiens, the mating calls of R. blairi have a unimodal distribution of energy from 0.1-1 kHz. Phonotatic responses of female R. pipiens were more frequently observed to the conspecific call than to the call of Rana blairi. Stimulus frequencies between 0.8-1 kHz could have exerted a suppressive influence on the response to the acoustic stimulation.

Further support of the idea that frogs have reduced physiological response to irrelevant stimuli are findings from a study that investigated hormonal changes in males of the green tree frog, Hyla cinerea, after exposure to different sound cues (Burmeister & Wilczynski, 2000). The circulating androgen levels in frogs increased after being broadcast conspecific calls, but not when presented with random tones matched in time and amplitude. Frogs receiving random tones also did not differ from frogs that were kept in silence for the duration of the experiment. The same effect has also been reported in southern leopard frogs, Rana sphenocephala (Chu & Wilczynski, 2001). Differences in neuronal responses when presented with either conspecific calls or noise, were recently demonstrated for the green treefrog (Gall & Wilczynski, 2015). Exposure to a simulated male chorus lowered auditory thresholds in males, while exposure to random tones had no effect, indicating that the peripheral auditory responses can change in threshold or other neural response measures according to signal relevance (Gall & Wilczynski, 2015, 2016). Moreover, prior exposure to conspecific signals significantly increased immediate early gene responses in the auditory midbrain when presented novel conspecific advertisement calls (Gall & Wilczynski, 2014).

One remaining question concerns the functional interpretation of our findings. MEMRI is based on the concept that Mn²⁺ acts as a physiological Ca²⁺ analogue and thus differences in signal intensity of MEMRI images reflect differences in neuronal activity of a given brain area. However, we cannot discern whether this is due to a higher number of neurons being activated simultaneously, or the result of frequency-dependent synaptic activity (i.e. higher firing rates by single neurons). Empirical as well as theoretical studies suggest that pooled neural responses from many auditory neurons might be necessary to achieve an enhanced response to sound (Jin et al., 2013). It has also been shown that sound stimulation can accelerate Mn²⁺ trans-synaptic transport through neuronal circuits (van der Linden, van Meir, Tindemans, Verhoye, & Balthazart, 2004). Therefore, the sound stimulation with socially relevant calls may have induced the acceleration of Mn²⁺ trans-synaptic transport, leading to stronger signal enhancement due to more Mn²⁺ accumulation in these areas of the brain. While further research is needed to disentangle the underlying molecular mechanisms that lead to final differences in signal processing of socially relevant and irrelevant sound stimulation, our findings clearly show differential physiological processing of signals with contrasting social relevance in the amphibian brain.

Our study demonstrates that MEMRI is a highly powerful approach to study activity patterns in the brains of small animals, such as frogs, with high resolution. MEMRI allows the study of the same subject over time and addresses the entire brain at once. Neural activity is imaged after the accumulation of paramagnetic Mn²⁺ ions within synaptically activated neurons of awake, free-moving frogs. MEMRI images are not impacted by scanner noise, and therefore particularly useful for bioacoustical studies, as stimulus presentation and manganese accumulation takes place outside the scanner. Compared to standard experimental models used for MRI (i.e. rodents), the handling of frogs for scanning was relatively simple. They could be anesthetized via a preceding bath, without the need for continuous supplying anesthesia inside the scanner. Body temperature was easily kept at optimal conditions as air temperature inside the scanner is relatively cool and humidity was maintained via the wrapping of wet gauze. We did not have to intervene to increase the permeability of the BBB to enable Mn²⁺ ions to be transported into the brain (Rapoport, 2000). Although MnCl₂ was found to considerably alter behavioral and neurochemical parameters in larvae and adult zebrafish (Altenhofen et al., 2017), no noticeable adverse effects due to manganese administration were found in any subject of our study (e.g. no decrease in foraging, movement, or body tension). One reason might be that the dosage applied in our study was far below the amounts reported by Altenhofen et al. (2017), and thus probably far below the critical amount needed for permanently altering an individual’s behavior. To the best of our knowledge, there is only one report to date that used MEMRI in frogs, in order to study optic nerve regeneration after injury (Sandvig et al., 2011).

Surprisingly, despite the technical advances over the last years in developing more powerful magnets, which would allow scanning of smaller-brained animals, and the recent trend of integrating basic cognition research and medical brain imaging (e.g. van der Linden et al., 2002; Vuilleumier, Armony, Driver, & Dolan, 2001; Ziegler et al., 2011), the application of MEMRI, and fMRI in general, has been largely restricted to laboratory rodents (but see Herberholz et al., 2011). Other common animal models in neuro- and molecular biology such as chicks, frogs, and zebra fish have brains that differ considerably from those of mammals. Although several brain regions common to all vertebrates might be structurally, functionally as well as evolutionarily homologous (O'Connell & Hofmann, 2011, 2012), the cerebral cortex, for example, is unique to mammals. In recent years, the urgent need for more and novel model-organisms in scientific research has been raised repeatedly (Bolker, 2012; Taborsky et al., 2015). The small number of common model organisms cannot nearly fulfill the requirements of providing universally valid explanations for the immense diversity in nature around us (Laine & van Oers, 2017). FMRI research in non-standard model organisms brings with it the possibility of widening our understanding of brain function in basic research fields such as cognition, bioacoustics, and physiology. Functional MRI experiments may also provide a new and convenient in vivo imaging tool for studying cognitive processes of rare or endangered species, e.g. common marmosets (Okano & Mitra, 2015). Furthermore, given the substantial difference between brains of different vertebrate taxa, comparative approaches will help us investigate architecture and functional connectivity of vertebrate neuronal structures, and thus provide us with a more complete understanding of important but poorly understood biomechanical mechanisms of perception and attention.