The auditory cortex of the bat Phyllostomus discolor: Localization and organization of basic response properties

Background The mammalian auditory cortex can be subdivided into various fields characterized by neurophysiological and neuroarchitectural properties and by connections with different nuclei of the thalamus. Besides the primary auditory cortex, echolocating bats have cortical fields for the processing of temporal and spectral features of the echolocation pulses. This paper reports on location, neuroarchitecture and basic functional organization of the auditory cortex of the microchiropteran bat Phyllostomus discolor (family: Phyllostomidae). Results The auditory cortical area of P. discolor is located at parieto-temporal portions of the neocortex. It covers a rostro-caudal range of about 4800 μm and a medio-lateral distance of about 7000 μm on the flattened cortical surface. The auditory cortices of ten adult P. discolor were electrophysiologically mapped in detail. Responses of 849 units (single neurons and neuronal clusters up to three neurons) to pure tone stimulation were recorded extracellularly. Cortical units were characterized and classified depending on their response properties such as best frequency, auditory threshold, first spike latency, response duration, width and shape of the frequency response area and binaural interactions. Based on neurophysiological and neuroanatomical criteria, the auditory cortex of P. discolor could be subdivided into anterior and posterior ventral fields and anterior and posterior dorsal fields. The representation of response properties within the different auditory cortical fields was analyzed in detail. The two ventral fields were distinguished by their tonotopic organization with opposing frequency gradients. The dorsal cortical fields were not tonotopically organized but contained neurons that were responsive to high frequencies only. Conclusion The auditory cortex of P. discolor resembles the auditory cortex of other phyllostomid bats in size and basic functional organization. The tonotopically organized posterior ventral field might represent the primary auditory cortex and the tonotopically organized anterior ventral field seems to be similar to the anterior auditory field of other mammals. As most energy of the echolocation pulse of P. discolor is contained in the high-frequency range, the non-tonotopically organized high-frequency dorsal region seems to be particularly important for echolocation.


Background
During the last decade, the bat P. discolor has been used increasingly for psychophysical and neurophysiological studies of echolocation [1][2][3][4]. P. discolor is medium-sized and forages for fruit, nectar, pollen and insects in a neotropical forest habitat. Its vocal emissions are brief (< three milliseconds), broadband multi-harmonic, downward frequency modulated (FM) echolocation pulses with a frequency range of about 40 to 90 kHz. In contrast, the rich repertoire of communication calls used for speciesspecific social interaction covers the lower frequency range from 11 to 54 kHz [5,6]. The responses of cortical neurons to complex stimuli relevant for echolocation in P. discolor have been compared to the behavioral performance of the bat [7,8]. Therefore, it is important to gain knowledge of the detailed organization of the AC with respect to basic response properties.
The auditory cortex (AC) of mammals is composed of distinct fields, which can be characterized by physiological and cytoarchitectural features and their specific thalamocortical connections (for review see [9,10]). The functional organization of the AC in bats has been extensively studied physiologically in several species (e.g. Pteronotus parnellii [11,12], Rhinolophus spec. [13,14], Eptesicus fuscus [15,16], for review see [17]). Among the best studied ACs so far are those of the mustached bat P. parnellii and the horseshoe bat, Rhinolophus rouxi, both belonging to the group of the so-called CF/FM bats whose echolocation pulses consist of a constant frequency (CF) and a frequency modulated (FM) component. As common to all mammals studied so far, their ACs contain a tonotopically organized primary auditory field (AI) with the frequency gradient running from caudal to rostral. However, in both CF/FM bats frequencies of the CF component of the calls are largely overrepresented in AI while frequencies of the FM component are only weakly represented [18]. The AI is surrounded by cortical regions with neurons that show facilitated responses to specific spectral and temporal combinations of the CF and FM parts of the different harmonics of the echolocation pulses.
CF/FM bats are rather specialized echolocators in that they hunt almost exclusively insects on the wings, whereas other bat species display more varied feeding ecology (insects, vertebrates, nectar, fruits) and very commonly use short downward FM echolocation pulses often with several harmonic components. In these bats the functional specialization of the AC is often not so clearly apparent, but still cortical fields can be segregated based on neurophysiological criteria like best frequency (BF; frequency at which threshold is lowest) representation and response threshold [15,19]. In the phyllostomid FM bat Carollia perspicillata for example, two dorsal fields containing mainly neurons with BFs in the high-frequency range have been reported in addition to the tonotopically organized fields AI and AAF (anterior auditory field) [20]. In these high-frequency fields some neurons exhibited pulse-echo delay sensitivity as in CF/FM bats [6], but without topographical organization.
Except for a short autoradiographic labeling study [21] the topography of the AC of P. discolor has not been studied. In general, the AC of only one other phyllostomid FM bat, C. perspicillata, has been investigated, so far [20]. Therefore, the aim of the present study was to investigate neuroanatomical and neurophysiological properties of the AC of P. discolor in order to delineate its subdivisions.

Auditory responses in the cortex of Phyllostomus discolor
As shown in Fig. 1A, units responding to acoustic stimuli are found at parieto-temporal portions of the neocortex of P. discolor. The distribution of neurophysiological recording sites defines the functional location of the bat's AC. External features roughly delineating the AC are the pseudocentral sulcus [22], which is located at the rostro-dorsal border and the fissura rhinalis, which is located at the ventral border. The auditory cortical area covers a rostro-caudal distance of about 4800 μm and has a dorso-ventral extension of about 5100 μm in the lateral view as shown in Fig. 1A. To obtain a realistic estimate of the cortical surface containing auditory units, the locations of the recorded units were projected on an unrolled and flattened surface projection along the medio-lateral coordinate (Fig. 1B, see Methods). In this projection, the lateral extension of the auditory responsive area is roughly 7000 μm.

Neuroanatomy
In the area responsive to acoustic stimuli four major fields are recognized based on cyto-and myeloarchitectural features and zinc staining pattern, i.e. the anterior and posterior dorsal fields (ADF and PDF), an anterior ventral field (AVF) and a posterior ventral field (PVF). The PVF maybe further subdivided into a dorsal and a ventral part (PVFd and PVFv) and a border zone (PVFbz), due to minor modifications of the neuroarchitectural characteristics. The topographic position of these fields is depicted in Fig. 1C. Reliable borders of cortical fields are indicated by solid lines. Dashed lines represent the more variable outlines of the AC itself and for PVF possible anatomical subdivisions that are not corroborated by neurophysiological data of the study.
Frontal sections in Fig. 2 give showcase characteristics at two rostro-caudal levels (as indicated in Fig. 1A) to get a general idea of field differences. Total cortical thickness, relative thickness of the different layers, composition of cell types, cell density, content of myelinated fibers and zinc are considered as parameters for the distinction of the different fields. Cut-outs of frontal sections stained for cells (Nissl), myelinated fibers (Gallyas) and for zinc (sulphide/silver histochemical method) from the centers of the different cortical fields are mounted in Fig. 3 for detailed comparison.
Total cortical thickness varies between 1600 μm (ADF) and 1200 μm from dorso-rostral to ventro-caudal locations (PVFv) in the AC. Despite the high density of granular elements in layer III and IV of the dorsal fields, the cortical layers V and VI take more of the total thickness of the cortex. Therefore, they are addressed as parietal cortical fields. The ventral fields are thought to belong to the temporal cortex, as layers III and IV dominate layers V and VI in thickness although granular elements are sparse and the cell density is lower in layers III/IV compared to the dorsal fields.
Dorsal and ventral fields also differ with respect to layer I, which is generally thinner in the dorsal than in the ventral fields. The myelinated horizontal fibers in the superficial part of layer I form a narrow dense band in the dorsal fields, but a broader one in the ventral fields, which is paralleled by a nearly zinc-free stripe in the ventral fields.
The most characteristic features used for delimitation of cortical fields are described in detail as follows. ADF has the greatest cortical width that seems to be caused by a doubled layer V: layer V of the non-auditory cortical area dorsal to the AC seems to continue beneath layer V of ADF (see frontal section stained for zinc in Fig. 2, left column, bottom, and Fig. 3, upper left panel, stars marking the two components of layer V).
In AVF, the rostral beginning coincides with the most caudal part of the claustrocortex without a sharp border. The myelin content in AVF is comparably low, whereas the stain for zinc is generally intense (as is the stain in the rostrally adjacent claustrocortex and the ventrally bordering perirhinal cortex). Even the paler staining band corresponding to layer IV is relatively dark.
The posterior fields are more homogeneous. Cortical width in PDF is smaller than in ADF and characteristic differences to the neighboring PVF are obvious at higher magnification (see Fig. 3). In the zinc stain, layer IV is narrower and more heavily stained in PDF than in PVF, whereas layer V is thicker and of higher staining intensity.
PVF has the most conspicuous wide and pale zinc staining band in layer IV, encompassing deep layer III. Layer V shows three subdivisions in the dorsal part (PVFd), whereas in the ventral part (PVFv) only two subdivisions are recognizable. Layer Vb, the most intense staining part Recording sites and subfields in the auditory cortex of Phyllos-tomus discolor Figure 1 Recording sites and subfields in the auditory cortex of Phyllostomus discolor. A) Lateral view of the P. discolor brain. Recording sites of all 849 units are indicated as black dots. Superimposed black outlines are neuroanatomically determined borders. Solid black lines represent reliable borders, whereas stippled black lines represent more variable borders. Rostro-caudal positions of frontal sections shown in Fig. 2 are indicated by the white vertical lines. Colored lines represent equal medio-lateral distances from the midline in 1000 μm steps as shown in the flattened cortical surface projection in 1B. B) Projection of recording sites (black crosses) and neuroanatomical borders (black lines) on an unrolled and flattened cortical surface. Lateral distances on the cortical surface are indicated in 1000 μm steps by corresponding colors as in the side view (1A). The origin used for the flattening process was fixed at 2000 μm lateral from the midline of the brain (upper dark blue line). C) Schematic of the auditory cortical subfields: anterior dorsal field (ADF), posterior dorsal field (PDF), anterior ventral field (AVF) and posterior ventral field (PVF) with dorsal (PVFd), ventral (PVFv) parts and a border zone (bz) reconstructed on the flattened cortical surface. The neuroanatomically determined borders are indicated by black lines.  of layer V is strikingly darker in PVFv. In PVFbz a gradual change of the characteristic features from PVF toward the ventrally adjacent perirhinal cortex takes place, e.g. the layered organization fades as well in the cell stain as in the zinc stain, and so do the myelinated fibers.
Layer IV appears as a pale band with the lowest staining intensity in the zinc stain and the staining intensity of layer IV varies characteristically between the fields. It is faint in layer IV of PDF, comparably lighter in ADF, relatively dark in AVF and faintest in PVF. Layer IV (and deep layer III) contains a high number of granular elements in the dorsal fields, whereas the neuronal somata in the ventral fields are larger and the cell density is lower. Layer IV is heavily myelinated in all fields except in the sparsely myelinated AVF.

Basic neuronal response properties
Because it was not always possible to discriminate the activity of a single neuron, the term 'unit' will be used in the following to describe the activity of one neuron to clusters of three neurons recorded at a distinct recording site. Extracellular recordings were derived from a total of 849 units from both hemispheres of ten bats.  Fig. 4B).

Latency and response duration
The first spike latency was measured in 681 cortical units and ranged between five and 127 ms with a bias toward short latencies (Fig. 4D). Median latency of all units was 9 ms (interquartile range: 7 to 20 ms).
Units often exhibited onset responses to pure tones frequently followed by a sustained response component that generally exhibited a considerable amount of variability. Other response types e.g. on-off responses were not observed in the AC of P. discolor. The response duration to a 20 ms pure tone was determined in 730 units and ranged from three to 385 ms (median: 25 ms; interquartile range: 14 to 76 ms, Fig. 4E). Short phasic responses with durations below 20 ms (  The first row shows the number of electrode penetrations in the AC of each bat, the total number of recorded units in each of the ten experimental animals is shown in the second row. units, the FRAs constituted closed areas in the frequencyintensity field with spike rates dropping to zero at all frequencies with increasing stimulus level (Fig. 5E). The remaining units (13 %, 93 of 745) featured complexshaped FRAs (Fig. 5F).

Binaural response properties
In 394 units the binaural response properties were measured and units were classified following their contra versus ipsilateral response characteristics (see Methods).

Representation of neuronal response properties in cortical fields
The local representation of selected response properties on the cortical surface was visualized using the Voronoi tessellation method (see "Methods"). The tessellation field around the recording site has its limits at halfway distance to the next surrounding recording sites and its color displays the respective response strength of the unit at the center. This forcibly entails large tessellation fields in areas in which recording density is lower or in border regions. As a consequence, there is no proper meaning in the area of single tessellation fields, but rather the clustering or systematic trends of tessellation fields within the AC are important information. Response properties of cortical units were analyzed for the four major subfields. Possible anatomical subdivisions of PVF were not corroborated by the neurophysiological data, and thus, PVF was analyzed as a whole.

Representation of best frequency and Q 10dB
The organization of BFs within the AC of P. discolor is shown in Fig. 6A. Units in the dorsal fields had a relatively restricted range of mainly high BFs, which covered frequencies in the range of the dominant harmonics of the echolocation pulse of this species. Most units in the ADF had BFs above 45 kHz, whereas units in the PDF had mainly BFs above 60 kHz. Only at the most rostral positions of the ADF units with low BFs were found. No tonotopic arrangement of BFs was found in the dorsal fields. In contrast, fields in the ventral part of the AC showed a tonotopic organization of BFs. The frequency gradient in AVF developed along the rostro-lateral to caudo-medial axis with low BFs at rostro-lateral positions, whereas in PVF the BFs decreased from rostral to caudal cortical positions. Thus, PVF and AVF shared a common high-frequency border. The rough direction of BF gradients in the ventral fields is schematically shown in Fig. 6B. As shown in Fig. 6D, the median BF (65 kHz) in PDF was significantly higher than the median BFs of the other cortical fields (Kruskal-Wallis test, p < 0.05). The lowest median BF (50 kHz) was found in PVF.
Q 10dB values of units were also not uniformly distributed on the cortical surface but showed a tendency to increase from anterior to posterior locations (Fig. 6C). In both anterior fields, units showed broader frequency tuning with significantly lower Q 10dB values (Kruskal-Wallis test, p < 0.05) than found in units in the posterior fields (Fig.  6E). Figure 6F shows the cortical representation of response thresholds at BF within the different fields. Thresholds of units in the dorsal fields were significantly higher than thresholds of units in PVF (Kruskal-Wallis test, p < 0.05) but only slightly higher than those of units in AVF (Fig.  6I). Units in the ADF had highest median response threshold (50 dB SPL), whereas the lowest median response threshold was found in units in the PVF (40 dB SPL).

Representation of threshold, latency and response duration
The distribution of first spike latencies in the AC of P. discolor is shown in Fig. 6G. As a general trend, short first spike latencies were preferentially represented in units in the two anterior fields, whereas in units in the posterior fields also long latencies were found. The median latency of PVF (13 ms; interquartile range: 8 to 28 ms) was significantly longer (Kruskal-Wallis test, p < 0.05) than the median latencies of ADF (median: 8 ms; interquartile range: 7 to 11 ms), AVF (median: 8 ms; interquartile range: 7 to 24 ms) and PDF (median: 9 ms; interquartile range: 7 to 20 ms, Fig. 6J).
The representation of response durations within the AC (Fig. 6H)  Examples of the six different FRA types of cortical units tical representation of double tuned FRAs with harmonically related components did not show a specific clustering within certain subfields. Figure 7B shows the representation of the different binaural response types in the AC of P. discolor. The distribution of binaural response types in the dorsal cortical fields was significantly different (Chi-Square test, p < 0.05) compared to the ventral cortical fields (Fig. 7D). Units of the dorsal fields were mainly driven by input from both ears and were classified as EE (ADF: 62 %; PDF: 38 %). In contrast, units of the ventral fields were predominantly characterized by EI input type. In detail, 53 % of units in AVF and 47 % of units in PVF were excited only through the contralateral ear and were inhibited by ipsilateral stimulation.

Discussion
This study investigated the localization, neuroarchitecture and basic physiological response properties of the auditory cortex of the bat P. discolor. The large area of neocortex responsive to acoustic stimulation comprises four anatomically distinguishable fields: a ventral part of the AC with an anterior and a posterior subdivision (AVF and PVF) and a dorsal part of the AC further divided into anterior and posterior subfields, ADF and PDF. The anatomical parcellation of the AC was consistent with the differentiated distribution of basic acoustical response properties of units in the cortical subfields. Location and extension of the AC of P. discolor as defined by neurophysiological and neuroarchitectural methods in this study roughly corroborated the findings from an earlier 2-DG study in the same species by Esser [21].
Spatial representation of FRA types and binaural response types and distribution in different cortical subfields

Comparison with the auditory cortex of other bats
The distribution of BFs in the AC of P. discolor (see Fig. 4A) showed an overrepresentation of frequencies in the spectral range of the bat's echolocation pulse (40 to 90 kHz). This overrepresentation might be due to the fact that the dorsal fields are mainly containing units with BFs in the high-frequency range. Thus, a larger amount of cortical surface is devoted to units with high BFs as compared to units with low BFs. In addition, more units were recorded from the dorsal fields and this bias might also be reflected in the distribution of BFs. A similar overrepresentation of BFs in the spectral range of the echolocation pulses has been found in the AC of other microchiropteran bats e.g. E. fuscus [23], Myotis lucifugus [17] and C. perspicillata [20]. In the CF/FM bats P. parnellii [18], R. rouxi [13] and R. ferrumequinum [14] the overrepresentation of a narrow callspecific spectral range of the CF-part is reflected throughout the ascending auditory pathway [24][25][26][27][28][29] and is based on specializations already implemented on the cochlear level [30][31][32].
In CF/FM bats, a large part of the AC is involved in processing of specific combinations of the different harmonics in the echolocation pulse and echo. In P. parnellii and slightly less distinctive in R. rouxi, several fields can be segregated by virtue of selective responses of neurons to different combinations of FM-parts or CF components of call and echo. Neurons are tuned to delays between the components and are topographically ordered [12,17,33]. Such a complex representation of pulse and echo delay has generally not been found in FM-bats so far, although pulse-echo delay sensitive neurons exist in the AC of e.g. M. lucifugus [34] and in the dorsal parts of the AC of the phyllostomid bat C. perspicillata [6], which is closely related to P. discolor. In the present study, only pure tones but no echolocation related stimuli (e.g. FM/FM pairs) were used for acoustic stimulation (see Methods). Consequently, no statements concerning the functional involvement of cortical subfields other than pure tone processing (e.g. delay sensitivity) can be made, so far. Besides that, the organization and parcellation of the AC of P. discolor is generally in accordance with that of C. perspicillata. The dorsal fields in the AC of Carollia were similarly characterized by neurons with BFs in the high-frequency range, and the ventral part of the AC was also composed of two tonotopically organized subfields, which shared a common high-frequency border [20]. In contrast, the AC of other non-phyllostomid FM bats like M. lucifugus and E. fuscus showed differentiation into subfields to a lesser degree [15,19]. In both species tonotopically organized fields with frequency gradients running in opposite directions were found, but no fields with characteristics similar to the dorsal cortical fields in P. discolor were described.
A high percentage (55 %) of units in the AC of P. discolor had first spike latencies below 10 ms. As other studies in FM-bats reported mean latencies measured at 10 to 20 dB above threshold around 8 to 10 ms already on the level of the IC [35,36], this seems to be unusually short. However, short latencies in the AC of FM-bats were also reported [37]. In addition, Haplea et al. showed that on the level of the IC latencies in E. fuscus still could be as short as about three milliseconds [35]. Although most cortical neurons in R. rouxi had latencies in the range between 10 to 15 ms, short latencies below 10 ms were found in a substantial 18 % of neurons [13]. The bias toward short latencies below 10 ms in P. discolor might be explained by the fact that a large number of recordings in our study were derived from dorsal regions of the AC where short latencies are predominately represented.
In contrast to findings in other mammals where the AI has been described to have the shortest latencies [38], the lowest median first spike latency in P. discolor was found in the ADF, whereas in the PVF, which might be equivalent to the AI, the longest first spike latencies were found. Our findings in P. discolor are supported by the fact that in the AC of R. rouxi shortest latencies were also found in the dorsal fields [13]. In R. rouxi, ventral and dorsal cortical fields are not sequentially connected but rather receive connections from different parts of the auditory thalamus via different projection pathways [39][40][41]. In P. discolor, a similar connection pattern could explain the shorter latencies in the dorsal fields compared to the ventral fields. However, this hypothesis needs to be proved by future tracer studies in this species.

Influence of anesthesia and multi unit recordings on response patterns
A large proportion of units in the P. discolor AC showed onset-type responses. It is known that the response pattern of cortical neurons is influenced by anesthetics. For example, the use of barbiturates like pentobarbital that enhances GABA A -mediated inhibition biased the distribution of response types toward short onset responses in the rat auditory cortex [42]. The drug Midazolam used in our study, like other benzodiazepines, also enhances action of GABA on GABA A receptors. Thus, the high abundance of phasic onset-type responses in our study could be a consequence of the use of Midazolam. However, a high percentage of onset-type response patterns (68 %) were also found in neurons in the AC of awake R. rouxi [13] and awake mustached bats [43], indicating that even in the un-anesthetized bat phasic responses are common. This minimizes the potential effect of anesthetics on response types of cortical units in the present study. Another point that could have influenced the response pattern is that recordings in our study were not always derived from single units (see Methods). Seshagiri and Delgutte [44] showed that adjacent neurons recorded simultaneously with tetrodes in the inferior colliculus of cats could differ significantly in their temporal discharge pattern although BFs and threshold were highly correlated. Therefore, blurring of response pattern is possible in some cases in multi unit recordings, but markedly dependent on the number of contributing neurons with dissimilar response types. As multi unit recordings in this study were generally derived from only up to three neurons (see Methods), and most units showed phasic responses the pooling of spikes in multi units should not have severely effected the response pattern in most cases.

Parcellation of the AC in P. discolor
One way of defining different fields of AC is by comparing the afferent connections with the auditory thalamus. Common to all mammals is a strong tonotopically organized projection from the ventral division of the medial geniculate body (MGBv) to AI [45]. In contrast to this, non-tonotopically organized non-primary cortical areas receive major projections from the dorsal division of the MGB (MGBd) [45]. CF/FM bats like horseshoe bats (e.g. R. rouxi, [39,41]) and P. parnellii [46] follow this connectivity pattern. In both animals, AI contains a tonotopic map where frequencies of the CF-components of the calls are overrepresented while frequencies of the FM-components are largely absent [18]. In R. rouxi, neurons specialized for the processing of acoustic parameters relevant for echolocation (e.g. echo delay) are mainly represented in dorsally located cortical areas that are targeted by projections from the MGBd [41]. The view of a congeneric projection from MGBv to AI covering the entire frequency range has been challenged by recent findings in the gleaning bat Antrozous pallidus. In this bat, the low frequency portion of a coherent frequency representation in the AC is innervated by the MGBv, whereas the high-frequency part received major projections from the supra-geniculate nucleus, a part of MGBd [47]. Neurons in the high-frequency range were moreover of a distinct binaural input type (EO) and preferred FM-sweeps, whereas neurons in the low frequency part were classified as EI and responded preferentially to noise. Two functionally distinct pathways evidently exist in this bat species: First, a low frequency pathway serving for passive prey location and second, a high-frequency pathway serving for obstacle avoidance during active echolocation [47,48]. The approach that part of AI may not receive input from the MGBv challenges one of the traditional definitions for AI via connectivity that, however, does not constitute the only valid criterion.
The distribution of binaural response types in P. discolor resembles that in A. pallidus, whereas neuroarchitectonic and neurophysiological features argue for a cortical fields pattern of the AC in P. discolor similar to the arrangement in R. rouxi and other mammals. Tracer experiments were not included in the present study in P. discolor, and consequently data on thalamo-cortical connections are not available. However, neuroarchitectonic features provide valuable tools for the definition of subdivisions in AC when combined with neurophysiological outcomes. Undoubtedly, the combination of all three approaches would yield optimal possibilities to classify cortical subdivisions. The specific usefulness of different neurohistological stains to determine cortical fields will be discussed in the following: The staining for zinc proved to be specifically helpful, complementing staining for cells and myelinated fibers. The zinc stain reveals horizontal bands in the neocortex that show characteristic variation in width and staining intensity among cortical regions, sharply defining the borders between cortical areas and their subdivisions, and are generally coincident with the limits defined in the Nissl stained sections. The banding patterns are stable among the species studied so far, allowing for an interspecies comparison (e.g. rats: [49][50][51] ; mice: [52] ; cats: [53] ; wallabies: [54] and humans [55]).
The PVF shows several features in the zinc stain characterizing primary sensory regions in the mammalian neocortex: conspicuous pale zinc staining of the outer portion of layer I, the differentiation of layer V in sublayers and, most characteristic, a wide, light band corresponding to layer IV encompassing the lower portion of layer III. In PDF, most distinctive to the PVF is the zinc staining of layer IV, which is narrower and more heavily stained, whereas layer V is thicker and of higher staining intensity. Zinc staining in AVF is generally intense. These features of PDF and ADF have been described to be characteristic for secondary sensory regions (rats: [49][50][51] ; mice: [52] ; cats: [53] ; wallabies: [54] and humans [55]). Furthermore, the structure of ADF seems unique in that it could be an amalgam of auditory and neighboring non-auditory, probably somatosensory areas. Mixed auditory-somatosensory fields neighboring the AC rostrally have been reported for several mammalian species (flying fox and gray squirrel: [56,57] ; mouse: [58] ; cat: [59] ; rat: [60]).
The accordance of response properties of cortical units in AVF and PVF with the anatomical partitioning of ventral auditory fields in the cortex is a further strong support, that the PVF might correspond to AI in P. discolor. PVF is tonotopically organized with BFs increasing from caudal to rostral. The AVF shows also a tonotopic organization but in contrast to the PVF, BFs increase from rostro-ventral to caudo-dorsal locations. This organization is in accordance to findings in other mammals where tonotopic gradients with roughly opposite directions have also been identified in the AI and the AAF (e.g. cat: [61]; ferret: [62]; gerbil: [63]; bat: C. perspicillata [20]). Because of the common high-frequency border of the AVF and PVF, high fre-quencies are represented all the way along the medial border of the PVF down to the most ventral border of this field (see Fig. 6A). In addition, no clear tonotopic gradient can be seen in the PDF that could be a tonotopic continuation of the frequency representation in the ventrally located PVF. Thus, it is unlikely that the PDF just represents the high-frequency portion of a contiguous frequency representation starting with low frequencies in the PVF as assumed for the AC of A. pallidus [47]. It is more likely that the PDF represents a non-tonotopically organized field like the dorsal fields found in other bats before (e.g. R. rouxi [13] and C. perspicillata [20], see above). However, also in C. perspicillata information about the thalamo-cortical connections is lacking.
In addition to BF, the distribution of several response properties like duration and first spike latency showed significant differences between the PDF and the PVF. To further test the possibility that high-frequency units in the PDF and PVF belong to a contiguous frequency representation like in A. pallidus [47], we additionally tested the differences of latency and response duration only for the high-frequency units (BF > 50 kHz) of both cortical fields. Latency and response duration for high-frequency units alone were still significantly different (Mann-Whitney U test, p < 0.05) between PDF and PVF. These findings further strengthen our parcellation of the AC based on the anatomical findings.

Conclusion
In this study, four subfields were identified in the AC of the bat P. discolor. These differed in their neuroanatomical attributes and in the response properties of their units to pure tone stimulation. The neuroanatomical and neurophysiological properties of the tonotopically organized PVF reflected common characteristics of AI in the mammalian AC (e.g. very wide and pale zinc staining band in layer III/IV, BF gradient ascending from caudal to rostral). The AVF located rostrally to the PVF showed a tonotopic BF gradient running roughly in opposite direction to the gradient of the PVF. Thus, the AVF of P. discolor might represent the AAF described in other mammals. The dorsal part of the AC of P. discolor was non-tonotopically organized with BFs mainly in the high-frequency range. As most energy of the echolocation pulse of this bat species is contained in the frequency range between 60 to 80 kHz, the dorsal region of the AC of P. discolor seems to be particularly important for echolocation. With regard to the relative position of cortical fields and their basic properties of BF representation, the AC of P. discolor seems to follow the organizational pattern seen in other phyllostomid bats.
To further compare the parcellation of the AC of P. discolor with those of other bats and mammals, the specific connectivity between subdivisions of the AC and different divisions of the auditory thalamus of P. discolor must be investigated in future studies.

Experimental animals
All experiments were performed in agreement with the principles of laboratory animal care and under the regulations of the current version of German Law on Animal Protection (approval 209.1/211-2531-68/03 Reg. Oberbayern).
Five male and five female adult spear-nosed bats (Phyllostomus discolor, body weight: 30 to 45 g) were used in this study. The animals originated from a breeding colony in the Department Biology II of the Ludwig-Maximilians-University in Munich. For experiments, animals were kept separated from other bats under semi-natural conditions (12 h day/12 h night cycle, 65 to 70 % relative humidity, 28°C) with free access to food and water.

Anesthesia and surgical preparation
In a pre-recording surgical session, the bat was prepared under anesthesia to provide for stable stereotaxic fixation in well defined coordinates and proper access to the target area. For anesthesia, a combination of Medetomidin (Domitor ® , Novartis, Mississauga, Canada), Midazolam (Dormicum ® , Hoffmann-La Roche, Mississauga, Canada) and Fentanyl (Fentanyl-Janssen ® , Janssen-Cilag, Neuss, Germany) was injected subcutaneously (MMF, 0.4, 4.0 and 0.04 μg/g body weight). Skin and muscles covering the upper part of the cranium were cut rostro-caudally along the midline and shifted aside laterally. The cranial bone was completely cleaned of all remaining tissue and a small metal rod was fixed onto the caudal part of the bat's skull using light-curing dental cement (Charisma ® , Heraeus Kulzer, Wehrheim Germany). After surgery, the anesthesia of the animal was antagonized by injecting a weight-dependent dose of a mixture of atipamezole hydrochloride (Antisedan ® , Novartis), Flumazenil (Anexate ® , Hoffmann-La Roche) and Naloxon (DeltaSelect ® , DeltaSelect, Dreieich, Germany) subcutaneously (AFN: 8.1, 0.034 and 0.32 μg/g bodyweight). To alleviate postoperative pain, the analgesic Meloxicam (Metacam ® , Boehringer-Ingelheim, Ingelheim, Germany, 0.2 mg/kg bodyweight) was administered after full recovery of the bat.

Stereotaxic procedure and verification of recording sites and correlation with neuroarchitectural features
A detailed description of the stereotaxic procedure, determination of brain orientation and reconstruction of recording sites has been already published elsewhere [64]. In brief, the animal's head was fastened with the surgically mounted rod to a stereotaxic frame in a well defined orientation. The alignment of the animal's skull and conse-quently the brain within the stereotaxic coordinate system was measured by scanning characteristic profile lines of the skull. The profile lines were measured in parasagittal and frontal planes (500 μm steps) relative to a fixed reference point and compared to a profile defining a standardized orientation of the skull and brain of P. discolor (Fenzl and Nixdorf, unpublished data), that also determines the normal axes of the available brain atlas of the animal. Thus, deviations of the actual alignment of the bat's head from the standardized position could be corrected by appropriate tilt and pitch of the animal within the experimental setup. Any reorientation of the skull is thus under accurate stereotaxic control. This procedure allowed the pooling and comparison of all electrophysiologically measured data, brain lesions and tracer deposits within one and among different experimental animals. Locations of lesions or marker (HRP, WGA-HRP) deposits were determined after transcardial perfusion and subsequent histological processing of the brain based on the brain atlas. These atlas-oriented coordinates were compared with those of the physiologically determined reconstruction of the specific location. This allowed a verification of the validity of recording coordinates, or eventually indicated errors that could typically be corrected after retracing the mistake in localization.
For the analysis of the neuroarchitectural features of the auditory cortex, series of frontal sections stained for cresyl violet, myelin [65] and zinc were available. For the detection of zinc, brains were processed according to a modification of the Timm method [66]. Zinc plays an important role in cell physiology [67] and has synaptic signaling functions in the mammalian brain [68,69]. In the neocortex, the stain for zinc reveals horizontal bands that show marked variation in width and staining intensity among functionally different cortical regions. These characteristic banding patterns are generally coincident with the limits defined in the Nissl stained sections, but have the advantage that they are readily noticeable at low magnification.
Additional series of frontal sections stained for the calcium-binding proteins, acetylcholine-esterase (ACHe), cytochrome-oxidase were not included in the analysis presented here, as they did not allow the differentiation between core and belt regions as clearly as described for other mammals with a more highly evolved neocortex (e.g., ACHe: [70,71]; Cytochromoxidase: [72]; calciumbinding proteins: [73,74]). Therefore, although involving a complicated perfusion procedure, the stain for zinc was chosen as histochemical marker to delineate cortical areas.

Acoustic stimuli and recording of neuronal responses
All experiments were conducted in a heated (36°C), electrically shielded and anechoic chamber. They normally lasted 4 hours per session and were performed at four days a week for up to six weeks. For each session, the bat was anaesthetized using MMF (see above). Throughout the whole experiment the animals were supplied with oxygen. To lower electrodes into the brain areas for recording of neural responses, small holes of about 500 microns in diameter were drilled into the animal's skull above the auditory cortex and the dura was perforated. Electrode penetrations were generally carried out in tangential direction, roughly parallel to the surface of the brain (bat #1-5, #8 and #9) or in a roughly perpendicular direction to the brain surface (bat #10). Two mapping experiments were done with orthogonal electrode penetrations (bat #6 and #7) to cover the entire auditory cortex in single individual bats and to neurophysiologically outline the borders and reveal possible tonotopic trends of the AC as a whole. Only basic properties like tuning and threshold were recorded in these two experiments.
Search stimuli for neuronal activity were pure tones with 20 ms duration produced with a frequency generator (Wavetek model 136, 186, FG-5000) and a custom-made pulse former. The attenuation of these stimuli could be modified manually (external attenuator AP401, adret electronique, France). The stimuli were presented via custom-made ultrasonic earphones [75] with a flat frequency response (± 3 dB SPL between 10 to 100 kHz). As it was not always possible to discriminate the activity of a single neuron, the term 'unit' is used in the text to describe the activity of one neuron to clusters of up to three neurons recorded at a distinct recording site. Once a unit was detected, its BF was determined audio-visually. For most units, the frequency response area (FRA) was determined in more detail. Therefore, pure tone stimuli (20 ms duration, 2 ms rise and fall time), in various frequency and sound pressure level combinations were presented contralaterally. Although P. discolor emits FM-pulses for echolocation, pure tone stimuli were used to allow comparison of the data to a previous study in a closely related phyllostomid bat, C. perspicillata, [20]. The stimulus duration of 20 ms is roughly intermediate between the short echolocation pulses (< 3 ms) and longer social calls (about 50 ms, [76]) of P. discolor.
The binaural response properties were determined at the unit's BF by using ABI (averaged binaural intensity, [77]). 20 ms long pure tones were presented binaurally with increasing intensity at contralateral and decreasing intensity at ipsilateral side and vice versa. The interaural intensity difference (IID) was changed from -20 dB SPL to +20 dB SPL in steps of 5 dB SPL. The level of the signals was chosen so, that both ears were stimulated at 20 dB SPL above the unit's auditory threshold. Thereafter, the same stimuli were presented monaurally to the contralateral and ipsilateral ear.
All stimuli for neuronal recordings were computer-generated (Matlab ® 6.1; Mathworks, Natick, USA), digital-analog converted (RX6; sampling rate 260 kHz, Tucker Davis Technologies, Gainsville, USA) and attenuated (PA5, Tucker Davis Technologies, Gainsville, USA). The acoustical stimulus parameters (frequency/sound pressure level combinations) were presented pseudo-randomly with a repetition rate of 1.3 Hz, a silent interval of 10 to 50 ms leading stimulus onset and with 10 repetitions of the entire parameter set.

Data analysis
All non-commercial computer programs used for data analysis were written in Matlab ® . Data analysis was done based on peri-stimulus time histograms (PSTH, 1 ms bin width) constructed from the spike responses to different parameter sets. The time window for analysis was adjusted to the total response duration of the unit. It started when the first bin exceeded the level of spontaneous activity and ended when the response reached steady spontaneous level again. The level of spontaneous activity was derived from the silent period preceding each stimulus onset. First spike latency was determined for stimuli presented at the unit's BF at a level of 20 dB SPL above threshold (see below for details on BF and threshold). The latency of the first spike to each presentation was used to calculate the median value of first spike latency. Only spikes occurring within the time window for analysis were included in the calculation.
The response characteristics of units that were tested at different combinations of frequency and sound pressure level, were visualized as FRA constructed from the summed activity within the given time window. Responses at different frequency level combinations were considered to be significant if the spike rate exceeded 20 % of the absolute maximum response of the unit. Basic characteristic values of the units like best frequency and tuning quality (Q 10dB and Q 30dB ) were directly derived from the FRA. The data basis for best frequency thresholds is the same as used for a previous publication [78]. However, data was re-analyzed in the present paper regarding the topographical distribution of thresholds within the AC of P. discolor. FRAs of units were classified as V-shaped, circumscribed, double-tuned or complex following roughly the categories used by Heil and Irvine [79] and Sutter [80]. The spike rate of units with circumscribed FRAs decreased at all frequencies to zero with increasing stimulus level. Double tuned units had FRAs that showed two clearly separable response regions, whereas complex FRAs consisted of multiple activity patches separated by regions of low activity. Cortical units classified as Vshaped responded over a contiguous range of frequency level combinations. V-shaped and double tuned FRAs were further distinguished according to their spike rate level function as monotonic or non-monotonic. The unit was labeled as non-monotonic if the spike-rate decreased again to below 75 % of the maximum at BF at higher levels. Thus, units with circumscribed FRAs always featured non-monotonic spike-rate-level functions by definition. Q 10dB values as a measure of sharpness of frequency tuning were calculated using the following equation: Q 10dB = BF/bandwidth measured 10 dB SPL above threshold. Double tuned and complex units did not generally allow determining a Q 10dB value. If the shape of FRAs allowed, Q 30dB values were also measured.
To determine a unit's binaural properties, the mean number of spikes was plotted as a function of IID for binaural, monaural contralateral and monaural ipsilateral stimulation. Comparison of these curves allowed the grouping of the binaural properties into four categories. Units that were excited by monaural stimulation of each ear were named EE (excitatory-excitatory). Units that only received excitatory input from the contralateral ear are named E0 (excitatory input only from the contralateral ear). In case the unit was excited by monaural contralateral stimulation, did not respond to monaural ipsilateral stimulation and showed inhibition in binaural stimulation, it was classified as EI (excitatory-inhibitory). The last group ("other") contained units that could not be ranked into one of the previous groups.
The locations of recording sites were projected to the flattened surface of the auditory cortex. For a detailed description of the surface projection method, see [81]. Briefly, each recording site was projected to the location at the cortical surface that had the shortest distance to the recording site. For the flattening process, the distance of the projected recording site to the origin fixed at 2000 μm lateral from the midline of the brain (upper blue line in Fig. 1B) was calculated. Thus, distortions due to irregularity of the cortical surface (e.g. pseudocentral sulcus) were avoided.
Distribution and trends of parameters on the cortical surface are represented with the help of the Voronoi tessellation procedure in two dimensions implemented with Matlab ® . In detail, recording sites of the chosen parameter are connected with all neighboring recording sites. Cells characterizing the value of a special parameter around recording sites are constructed as polygons whose sides pass equidistantly between recording sites and cross the connection lines perpendicularly.