Brains were collected from different sources (Table 1). Samples were obtained during field work for other experimental purposes, collected from animals euthanized due to the severe injury, or were available from previous work (Amrein et al., 2007).

Brains were removed from the cranial cavity 5 min–2 h postmortem and immersion-fixed in 4% phosphate-buffered paraformaldehyde containing 15% picric acid. For immunohistochemistry, one hemisphere was cryoprotected in a 30% sucrose solution, frozen and cut horizontally at a thickness of 40 μm using a sliding microtome (Microm, HM325). Serial sections were collected and preserved in cryoprotectant at −20°C. For the quantitative assessment of hippocampal neuron numbers, one hemisphere was embedded in 2-hydroxyethyl-methacrylate (2-HEMA; Technovit 7100, Heraeus Kulzer GmbH, Wehrheim/Ts, Germany) following the manufacturer’s instructions. Tissue sections were cut horizontally at 20 μm, mounted, and dried at 60°C for 1 h. Giemsa, a modified Nissl stain, was performed according to the protocol of Iñiguez et al. (1985).

For the qualitative assessment of hippocampal features, immunohistochemical staining were performed in all species except in Pipistrellus pygmaeus, Plecotus auritus and Sturnira lilium where only HEMA-embedded material was available. We used free-floating sections of series that spanned the entire hippocampus. Negative controls for non-specific binding were routinely performed by omitting the primary antibody. Stained neuronal elements resembled those observed in mouse sections processed in the same batches and reported for other species. However, without knockout controls, all immunoreactivity that we report here should be read as calbindin-like, calretinin-like or parvalbumin-like.

Epitopes were retrieved with 0.5% sodium borohydride in phosphate-buffered saline (PBS) for 30 min. Endogenous peroxidase was blocked with 0.6% hydrogen peroxide in Tris-Triton-buffer (1:10 of Tris base in dH₂0 + 0.05% of Triton, ph 7.4) for 15 min. Subsequently, sections were incubated for 1 h in 2% normal serum (calbindin and calretinin: goat, parvalbumin: horse) with 0.2% Triton in Tris-Triton buffer. Next, sections were incubated overnight at room temperature using rabbit anti-calretinin (Swant, CR7697, Lot 1893-0114, dilution 1:500), rabbit anti-calbindin (Swant, CB-38a, Lot 9.03, dilution 1:500) and mouse anti-parvalbumin (Sigma Aldrich, P3088, Lot 104780. Dilution 1:1000). Afterward, sections were washed in Tris-buffered saline (TBS) and incubated with goat anti-rabbit (1:300, Vector labs, BA-1000, Lot X11041) or horse anti-mouse (1:300, Vector labs, BA-2000, Lot ZF0521) for 40 min at room temperature. After this step, sections were incubated for 20 min in avidin-biotin complex (Vector Labs, Lot ZJ0909) in TBS. Sections were 3,3′-Diaminobenzidine stained, mounted, dehydrated and cover-slipped using Eukitt.

Timm stained sections of Seba’s short-tailed bat (Carollia perspicillata), pale spear-nosed bat (Phyllostomus discolor) and Pallas’s long-tongued bat (Glossophaga soricina) were available from a previous study (Amrein et al., 2007). From Glossophaga soricina, only the Timm stain was available for this study. In short, anesthetized animals were perfused transcardiacally in a series of PBS, 0.6% sodium sulfide solution and 4% paraformaldehyde. Horizontal cryostat sections (40 μm) were mounted and developed at 37°C in darkness for ∼60 min in a mixture of gum Arabic, hydroquinone and citric acid containing silver nitrate according to the protocol of Danscher and Zimmer (1978). Afterward, slides were rinsed in 1% sodium thiosulfate for 1 min, counterstained with neutral red, dehydrated and cover-slipped.

The naming and boundaries for hippocampal neuron populations followed conventions detailed in Maliković et al. (2023). In this study, the boundary between the CA3 and CA1 subregions was identified by the termination of the mossy fiber zone visible in Giemsa-stained sections. Definitions of neuron population boundaries were cross-checked using the stains for calcium-binding proteins, which often mark interregional boundaries.

Neuron number estimation was performed in the HEMA embedded, Nissl-stained sections of all 15 species (Table 1) using StereoInvestigator 10 Software (MBF Bioscience, Williston, VT, United States). In the Nissl stain, cytoplasm of neurons stains blue, nucleoli dark blue while the nucleus is largely unstained, differentiating neurons from glia as the cytoplasm of glia cells is not stained and nuclei stain light to dark blue without a distinct nucleolus (Fitting et al., 2010; García-Cabezas et al., 2016). Neuron counts were obtained using the optical fractionator method (West et al., 1991; Slomianka, 2021), with 10 μm high disector and a 2 μm top guard zone. Section thickness was measured at every fifth sampling site. Neuron counting was conducted under a × 63 oil immersion lens (NA 1.4). All sampling parameters and neuron counts are listed in the Supplementary Table 1.

Estimated total neurons were calculated based on number-weighted section thickness (Dorph-Petersen et al., 2001) and the precision of number estimates was assessed by the coefficient of error CE (Gundersen et al., 1999) with a conservative smoothness factor of m = 0.

Species classification for foraging habitat and dietary preferences (Table 1) were based on previous studies (Fleming, 1982; Norberg and Rayner, 1987; Ratcliffe, 2009; Denzinger and Schnitzler, 2013; Graf and Fischer, 2021).

Quantitative relations of each hippocampal neuron population (in percent) were visualized on the extracted phylogenetic tree (Álvarez-Carretero et al., 2022) using the R packages ape (Paradis and Schliep, 2019), phytools (Revell, 2024), ggtree (Yu et al., 2018) and ggplot2 (Wickham, 2016). Hippocampal neuron numbers of each specimen were log₁₀-transformed and z-scored (mean of 0 and standard deviation of 1 across all neuron populations), resulting in values representing the relative numeric contribution of each neuron population within the hippocampal circuitry. Z-scored neuron numbers were then visualized with a correspondence analysis as described before (Slomianka et al., 2013; Maliković et al., 2023) using the R package made4 (Culhane et al., 2005), including factors such as species, family, diet preference and foraging habitat.

Unless stated otherwise in the figure legends, images represent the intermediate (mid-dorsoventral) hippocampus and nearby structures in horizontal sections. Images were captured using a Zeiss Axio Imager.M2 microscope in the slide scanning mode of Stereo Investigator version 10 (MBF Bioscience, Williston, VT, RRID), using a × 20 objective.

The description of cytoarchitectural traits in vespertilionid and phyllostomid bats follows the classical hippocampal tri-synaptic loop of information processing in the hippocampus (Johnston and Amaral, 1998). The dentate gyrus granule cell layer (gcl) shows the usual densely packed appearance, with an elongated upper (suprapyramidal) blade in the dorsal and intermediate hippocampus (Figures 1A–C, 2A–C). The dorsally short lower (infrapyramidal) blade becomes longer beyond the intermediate level, surpassing the length of the upper blade ventrally (Figures 1D–F, 2D–F). Myotis nigricans was the only species in which length differences between blades was not extensive (Figures 1B, E). The hilar polymorphic cell layer (hpcl) is separated from the granule cell layer by a wide cell-poor subgranular zone, which is even wider in phyllostomid (Figure 2) than in vespertilionids bats (Figure 1). In vespertilionid bats, hilar polymorphic neurons are dispersed, whereas in phyllostomid bats, they form a dense band. The CA3 pyramidal neurons are similarly organized in both families, with a dorsal dense band of large pyramidal neurons (Figures 1A–C, 2A–C) becoming loosely arranged ventrally (Figures 1D–F, 2D–F). A tendency for proximally small CA3 pyramidal neurons to become larger distally was observed in some species (see example of Phyllostomus discolor in Supplementary Figure 1). Beyond the tip of the mossy fiber zone, the transition from CA3 to CA1 pyramidal neurons is relatively short in phyllostomid bats (Figure 2), in contrast to a more gradual shift seen in vespertilionid bats (Figure 1). This transition zone contains a blend of large and small pyramidal cells, and is likely to correspond to Lorente de Nó’s (1934) CA2. The wide CA1 pyramidal neuron layer is characterized by deep pyramidal neurons dispersing into stratum oriens (Figures 1, 2). Dispersal of CA1 pyramidal cells is most prominent in the tailed tailless bat (Anoura caudifer), in which also superficial CA1 pyramids extend much further toward the hippocampal fissure than those in CA3 (Figures 2B, E). Cytoarchitectural arrangements of CA1 vary between species. Superficial CA1 pyramids in Vespertilio murinus (Figure 1A) and species of the myotis group are arranged into a compact layer one to two neurons wide. The condensation of superficial pyramids is present at various degrees in all other species, but least prominent in Desmodus rotundus (Figures 2C, F) where superficial and deep pyramids appear largely homogenously distributed. Small species-specific variations in the arrangements of deep versus superficial CA1 pyramidal neurons can be seen along the proximodistal and dorsoventral (septotemporal) axis. The transition from CA1 to the subiculum is gradual in all bats. The subiculum itself is small in both families, recognized by a condensation of horizontally oriented neurons towards the alveus. Additionally, the boundary is marked by a more heterogeneous population of neurons, varying in both shape and size, and superficial neurons do not reach as far into stratum radiatum as in CA1 (see for example Figures 1C, F). The presubiculum is marked in both bat families by increased cell density and reduced cell size across all layers.

The distribution of calcium binding proteins showed many commonalities within the two taxonomic groups, Vespertilionidae and Phyllostomidae, represented in the sample. They are described in the species best representing the two groups, the parti-colored bat (Vespertilio murinus) and the pale spear-nosed bat (Phyllostomus discolor). Shorter notes provide observations in two additional species from each group (Myotis nigricans, Pipistrellus pipistrellus, Desmodus rotundus and Anoura caudifer). Unless noted otherwise, descriptions are applicable to all dorsoventral levels of the hippocampus.

To explore the unusual staining patterns of calbindin and calretinin in the CA3 region of phyllostomids, we evaluated mossy fiber stains in Glossophaga soricina (Figure 6A), Phyllostomus discolor (Figure 6B), and Carollia perspicillata (Figure 6E). In Phyllostomus discolor, Timm-positive staining of zinc-containing mossy fiber boutons in the proximal CA3 were adjacent to the CA3 pyramidal cell layer, while in the distal CA3 (Figure 6B) the zinc-positive band of mossy fibers separated from the pyramidal layer. The same proximal to distal pattern of change was observed for mossy fiber zone calbindin immunoreactivity (Figure 6C). The zinc- and calbindin-negative suprapyramidal area was in size and distribution similar to the calretinin-positive plexus seen in this species (Figure 6D). This gap between zinc- and calbindin-positive mossy fiber zone was more pronounced in Glossophaga soricina (Figure 6A) than in Phyllostomus discolor (Figures 6B, C). The gap was absent in Carollia perspicillata (Figure 6E), where the Timm-stained mossy fiber zone borders the CA3 pyramidal layer as typical in other mammals.

Absolute numbers of hippocampal neurons (Table 2) varied considerably between species, as did body and brain sizes (Table 1, see also Supplementary Figure 2). For comparisons, neuron numbers of each principal neuron populations were therefore presented in percentage of total hippocampal neuron numbers (Figure 7). Within vespertilionids (Figure 7A), the proportion of each neuron population within the hippocampus was relatively uniform, with the exception of Vespertilio murinus where the percentage of granule cells was smaller in favor of more CA1 pyramidal neurons (Figure 7B). Proportions of neuronal populations were more variable in phyllostomids, even between closely related species. Granule cell numbers differed by 10% between the two frugivore bats (Carollia perspicillata and Sturnira lilium), and by 18% between the two vampire bats (Diphylla ecaudata and Desmodus rotundus). The proportion of subicular neurons was small in all bats.

For the correspondence analysis (Figure 8), estimated total neuron numbers were z-transformed. The two axes in Figure 8 represent 85.5% of the variance in the data (1st factor, x-axis: 52.6%, 2nd factor, y-axis: 32.9%). Hilar and subicular neurons, the smallest neuron populations, caused the largest separation between species and dominated the first axis, while the second axis was dominated by granule cells and CA1 pyramidal neurons (Figure 8A). CA3 pyramidal neurons did not have much differentiating power. Species distribution (Figure 8B) already indicated that there is considerable overlap between the two families (Figure 8C), but separation of the two clusters suggested an increased emphasis on input neuron populations of the hippocampus (granule and hilar neurons) in vespertilionids, while the output side (CA1 pyramidal and subicular neurons) appeared more dominant in phyllostomids. Within the family of phyllostomid bats, preferred diet (Figure 8E) segregated the omnivore and frugivore bats (increased weight on HIL) from the nectivorous and hematophagous vampire bats (increased weight on SUB) along the first axis of the correspondence analysis. Foraging habitat (Figure 8F) was nearly congruent with families, again indicating a tendency to an input dominant hippocampus in edge space foragers, while narrow space foragers had more weight in output neuron populations of the hippocampus. The only open space forager (Vespertilio murinus) was separated from all other species in this analysis due to its increased weight on CA1.

We focus here on two features in the hippocampal cytoarchitecture of small echolocating bats. The first observation is the length difference of the upper and lower blade of the granule cell layer that shifts along the dorsoventral axis. Anatomical asymmetry in the dentate gyrus is not uncommon and has been described in detail for rats early on [reviewed by Amaral et al. (2007)], length difference between dentate blades in rodents appear however on a smaller range than what we observed in bats. In rodents, the upper and lower blade differ in their morphological (Gallitano et al., 2016), physiological (Mishra and Narayanan, 2020; Strauch et al., 2025) and functional (Chawla et al., 2005; Ramírez-Amaya et al., 2005; Satvat et al., 2011; Schmidt et al., 2012) characteristics. Whether these properties extend to the bat dentate gyrus too is currently not known, but one might wonder about the advantage of a prominent upper blade in the dorsal and intermediate dentate gyrus. In rats, granule cells in the upper blade are more activated during spatial tasks than those in the lower blade (Chawla et al., 2005; Ramírez-Amaya et al., 2005; Schmidt et al., 2012), thus a prominently elongated upper blade as seen in bats might increase precise spatial processing, which is the domain of the dorsal hippocampus (Strange et al., 2014). Specific functional attributes to the lower blade are, to our knowledge, currently unknown. Spatial exploration induces more Arc-expression in the upper than lower blade in the dorsal dentate gyrus, however the lower blade responds to the experimental exposure with a delayed and transient Arc-expression (Ramírez-Amaya et al., 2005). We found in bats that the lower blade is prominently elongated in the most ventral part of the dentate gyrus, a region of the hippocampus traditionally associated with emotional responses (Strange et al., 2014). To elucidate the specific function of lower blade granule cells, one may have to venture into behavioral tests for fear, motivation or reward, and small echolocating bats with their prominent lower blade ventrally might become especially suitable models to do so.

The second feature is the radial expansion of the CA1 pyramidal cell layer in all bats presented here. The distribution of the superficial and deep CA1 pyramidal cells in bats is reminiscent of that seen in primates including humans (Lorente de Nó, 1934; Slomianka et al., 2011). Deep and superficial CA1 pyramidal cells differ in their development, gene expression profile and connectivity (Slomianka et al., 2011; Valero and de la Prida, 2018; Cembrowski and Spruston, 2019) and process different information within the hippocampal circuits (Geiller et al., 2017; Soltesz and Losonczy, 2018; Harvey et al., 2023). In the context of the bat CA1, findings by Sharif et al. (2021) in mice might be most interesting, as they report more place cell activity in cue-rich environment in deep CA1 pyramids, while a cue-poor environment is linked to increased neuronal activity in superficial CA1 pyramids. Most bats commute relatively long distances to their feeding grounds, and one can assume this behavior corresponds to a distinct shift between cue-poor navigation in open space and cue-rich navigation once animals start feeding. Previous studies have confirmed CA1 place cells in large fruit bats (Yartsev and Ulanovsky, 2013) and echolocating bats (Ulanovsky and Moss, 2007), without differentiating between deep and superficial cells. The well-differentiated CA1 in bats would be convenient to explore flexible navigation in response to spatial maps of different scale in a natural environment.

With the verspertillionid and phyllostomid small echolocating bats, some information is now available for the second most speciose group of mammals. Together with the information already available from many other clades, it permits not only to (yet again) point out species differences and similarities, but also to outline some basic ideas about how the distribution of calcium-binding proteins may serve in understanding hippocampal function.

A highly unusual feature revealed by the distribution of calbindin, calretinin and the Timm stain was the gap between the superficial boundary of the pyramidal cell layer and the zone staining for the classical mossy fiber markers calbindin and zinc in phyllostomids. Tract tracing studies will be needed to resolve the cause of the gap. We considered two possibilities. First, extra- or intrahippocampal CR+ afferents may intervene between pyramidal cells and the mossy fiber zone. Although a likely source does not come to mind, the gap itself is such an unlikely feature that the possibility should not be dismissed. Afferents important enough to displace the mossy boutons from close to the somal center would tell us much about hippocampal information processing in these species. A second possibility is the presence of neurochemically distinct granule cell and/or mossy bouton populations that segregate radially where the gap is visible. A subset of temporal granule cells was CR+, and calbindin was absent from granule cells and mossy fibers in the phyllostomid Desmodus rotundus and unusually weak in verspertilionid species. Other species show a heterogenous (Cavegn et al., 2013), sometimes laminar (Slomianka et al., 2013) expression of calbindin in granule cells, i.e., calbindin can be absent or undetectable in (a subset) of granule cells and mossy fiber boutons. Also, a subset of rat mossy boutons did not express the transporter ZnT3 (Rekart and Routtenberg, 2010) responsible for vesicular zinc loading and, consequently, the Timm staining (Cole et al., 1999). Although the two bouton types did not segregate radially in rats, they occurred more frequently in distal CA3 [Figure 3 in Rekart and Routtenberg (2010)], where the gap is widest in bats. The concept of segregated information streams across subsets of CA1 pyramidal cells has become well-established (Cid et al., 2021; Sharif et al., 2021; Ding et al., 2022; Huszár et al., 2022) and evidence is accruing also for CA3 (Slomianka et al., 2011; Marissal et al., 2012; Balleza-Tapia et al., 2022). But streams were originally suggested to apply already to the dentate granule cells (Deguchi et al., 2011), which is supported by a (often birthdate-related) diversity of dentate granule in gene expression and physiology (Shah et al., 2016; Imura et al., 2019; Save et al., 2019; Erwin et al., 2020). Neurochemical diversity and spatial segregation of mossy boutons in phyllostomid bats would provide an early anatomically defined window into hippocampal information streams. In general, the histoarchitectural differentiation of CA1 and the prominent definition of subsets of CA3 cells by calretinin point toward a stronger definition of such streams in small echolocating bats than in many other species. Given that phyllostomid bats are amenable to laboratory colony breeding, they harbor the potential to gain insights into hippocampal function similar to those that were gained from large fruit-eating bats.

Compared to a diverse sample of mammals (van Dijk et al., 2016), the overall quantitative makeup of hippocampal neuronal populations in echolocating bats shows a mixture of known and unique features. The relatively small granule cell population (on average 45%) was seen in rodents before, while the large hilar neuron population (5%) does not fit the rodent trait. In bats, the potentially stronger feedback control by many hilar neurons on relatively few granule cells may strengthen pattern separation in the dentate gyrus via a gate or filter function [reviewed by Borzello et al. (2023)]. The well-developed CA3 (20%) is a feature observed in mole-rats too, while many CA1 pyramidal neuron (24% in bats) is not a common feature in rodents, but was seen in dog and human (van Dijk et al., 2016), thus, CA1 pyramidal neurons are not only less densely packed in echolocating bats than in many other mammals, they are also quite numerous. Quantitatively, there is no indication that the loose arrangement of CA1 pyramidal neurons as seen in bats, humans and other primates might be a consequence of high numbers of CA1 pyramids within the hippocampal circuitry. While humans do have a high percentage (33%) of CA1 pyramidal neurons, other primates have less than bats (Rhesus monkey 13%, marmoset 17%, van Dijk et al., 2016). Drivers for the loose arrangement of CA1 pyramidal neurons are, to our knowledge, unknown. Striking in comparison to other species such as rodents (on average 12%) or humans (13%) is the small subicular neuron population in bats (7%). The function of the subiculum within the hippocampal network has been extended from a mere relay unit for information transfer between the hippocampus and cortical and subcortical target areas to a unique role for integrating and compressing spatial information of large, complex environments [reviewed by Matsumoto et al. (2019)], with specialized cells such as boundary vector cells (Lever et al., 2009) and vector trace cells (Poulter et al., 2021) responding to objects, boundaries and cues. In this study, we present the subicular neuron population as a homogenous entity. Yet, at least two types of pyramidal cells with different electrophysiological properties and spatial distribution have been described (Ishihara and Fukuda, 2016), neurons in the subiculum also segregate along the proximo-distal axis, e.g., vector trace cells are nearly exclusively found in the distal subiculum of rats (Poulter et al., 2021). How bats navigate successfully in a three-dimensional environment with relatively few subicular neurons still needs to be explored, further investigations would be required to verify if the bat-specific small subiculum stems from a global reduction, or a reduction of specific cell types and/or subregions.

How are ecological factors such as foraging habitat and diet reflected in the neuronal composition of the hippocampus? In the present species sample, we found not a clear separation between edge space and narrow space foragers in terms of hippocampal neuronal composition, which stands in contrast to correlations between hippocampal volume and foraging habitat reported before (Baron et al., 1996b; Safi and Dechmann, 2005). In our analysis, a similar distribution in the correspondence analysis is however seen if bats are grouped by family, thus, space use cannot be separated from phylogenetic traits in the numerical composition of hippocampal neuron numbers in this sample of bats. Diet preferences however did separate phyllostomid bats in their hippocampal cellular composition, which was not found on hippocampal volumetric analysis in this family (Ratcliffe, 2009). The sample of bat species presented here is relatively small, and Vespertilio murinus with a unique ecological trait already indicate that the inclusion of more species might reveal structure-function associations between hippocampal neuron numbers and ecological factors that we cannot see at present. Yet, our study presents the first quantitative data set of the bat hippocampus at circuit level providing ample evidence of unique and interesting hippocampal features, and we believe this report will help guiding future studies into this large group of mammals.