Latexin and calretinin together define a novel excitatory neuron subclass in the claustrum of the short-tailed fruit bat, Carollia perspicillata

Abstract

The claustrum is a telencephalic structure with inputs from and outputs to many other brain structures. This central arrangement has motivated research on the claustrum’s role in cognition and highlights the need to understand its intrinsic connectivity. In the fruit bat, Carollia perspicillata, the large size of the claustrum offers access to its intrinsic structure. Previously, we defined the structure of the C. perspicillata claustrum with antibodies against latexin as an excitatory cell marker and against calcium-binding proteins as inhibitory cell markers. Using this immunohistochemical method, we have now identified an unexpected cell type with concurrent latexin and calretinin immunoreactivity. The calretinin⁺ neurons of the claustrum, including those that coexpress GAD67 (another inhibitory cell marker) and those that coexpress latexin, are located in the claustral shell subregion. Neuronal latexin⁺/calretinin⁺ somata are smaller than either latexin⁻/calretinin⁺ or latexin⁺/calretinin⁻ somata. Since latexin labels glutamatergic neurons in multiple brain areas and has never been found to colocalize with GAD, we conclude that the latexin⁺/calretinin⁺ neurons in the claustral shell are excitatory. They represent one of three excitatory cell types that are identifiable in the claustral shell and demonstrate that calretinin can label both inhibitory and excitatory cells in the C. perspicillata claustrum.

INTRODUCTION

Latexin, a member of the carboxypeptidase A inhibitor family, is an intracellular marker that has been well-defined in several brain regions, including the claustrum and dorsal endopiriform nucleus of rats, cats, and bats.^1–6 In these species, the densest population of latexin immunoreactive claustral neurons exists within the claustral core subregion. Other markers (e.g., NURR1, NTNG2, GNB4) have also been shown in cells of the claustrum and the endopiriform nucleus, and several of them (e.g., GNG2, GNB4, parvalbumin) preferentially label the claustral core subregion too (reviewed in Ref. 7 and see also Ref. 8). Some of this apparent preference for the claustral core subregion in rodent studies may be due to the relatively slender or compressed shell relative to the core in these animals.

The claustrum and the dorsal endopiriform nucleus of Seba’s short-tailed fruit bat (Carollia perspicillata) are relatively larger structures than in rodents, and this size advantage has permitted the identification of multiple subregions in each structure based on differing levels of immunoreactivity for various markers, including latexin and calretinin.⁴ As we have described in detail previously, the claustral core subregion in C. perspicillata is surrounded by a shell subregion that can itself be subdivided based on the pattern of latexin immunoreactivity.⁴ There is a dorsal shell zone that contains latexin immunoreactive neurons in lower density than the claustral core, but higher density than other claustral shell or dorsal endopiriform nuclear zones. There is a central shell zone with the lowest density of latexin immunoreactive neurons, and there is a ventral shell zone that lies between the claustral core subregion and the dorsal endopiriform nucleus with latexin immunoreactive neurons of moderate density and whose dendritic orientations are distinct from the latexin immunoreactive neurons of other subregions.

In addition to the claustrum, latexin-immunoreactive (-ir; latexin⁺) neurons are also found in several neocortical regions, including the auditory, somatosensory, secondary visual, insular, and perirhinal cortices^9,10 and in transitional cortices such as retrosplenial cortex.¹¹ When studied in these different cortical regions, the vast majority of latexin-ir neurons were demonstrated to be glutamatergic pyramidal projection neurons.¹² In the claustrum and endopiriform nucleus of the rat and bat, coexpression of latexin with markers of inhibitory neurons has never been reported, suggesting that the latexin-ir neurons of these regions are excitatory neurons as well.^4,6,12

Calretinin is a calcium-binding protein that, along with parvalbumin and calbindin, is traditionally used to describe subpopulations of inhibitory interneurons producing GABA throughout the brain (e.g., see Refs. ^13–19). Calretinin has also been helpful in describing the boundaries of the claustrum.^4,20–28 GABAergic calretinin-containing neurons have been shown to have the unique property of targeting other classes of inhibitory neurons, thus being disinhibitory neural elements in regions of the brain like the hippocampus and neocortex.^18,29,30

As an extension of our work describing the relationships between the subregions and cell types of the C. perspicillata claustrum, we found a small subset of neurons that were immunoreactive for both latexin and calretinin. In this paper, we distinguish these neurons from neurons that express either latexin or calretinin, but not both, estimate the relative size of this cell population, and define their distribution in the C. perspicillata claustrum.

METHODS

Animals

Fruit bats (C. perspicillata) were housed in an AAALAC-accredited facility in groups of 10–20 animals per cage with continuous access to water and with daily feedings of a prepared diet that have been described in detail previously.^31,32 Cages contain males only, females only, or a mix of males and females, whose ages range from newborn up to 13 years old. Distinguishing between mature adult and adolescent/young animals is straightforward and, for studies specifically comparing animals of different ages, DNA methylation analysis permits age estimation.³³ Room lights go on at 3:00 a.m. and off at 3:00 p.m. daily. The target temperature and humidity ranges are: 75 ± 2 degrees °F; 50 ± 10% relative humidity. Cages comprise a 20 cubic foot light half and a 16.7 cubic foot dark half (see Refs. ^{31, 34}, and ³⁵ for additional details).

Tissue collection

Eight adult bats (six female, two male) were deeply anesthetized with urethane (≥ 2 mg urethane/mg body weight given as a 20% w/v urethane solution in water injected subcutaneously) and perfused transcardially with cold (4°C) phosphate buffer (0.1 M Na_xH_3-xPO₄, pH 7.4), followed by cold, buffered paraformaldehyde (4% w/v in phosphate buffer). Subsequent tissue handling was performed in the cold with cold solutions. The perfused brains were removed from the skulls and post-fixed for 18 h. After two rinses with phosphate-buffered saline (PBS: NaCl 1.37 × 10⁻¹ M, KCl 2.68 × 10–3 M, Na₂HPO₄ 1.014 × 10⁻² M, KH₂PO₄ 1.76 × 10⁻³ M; pH 7.4), the brains were immersed in cryoprotectant (30% wt/vol sucrose in PBS) until they sunk. Frozen sections (brains were sectioned at −33°C) were cut into 35 μm thick slices in the sagittal (six brains) or coronal (two brains) plane with a freezing microtome (Microm HM 430, Thermo Scientific) and stored in PBS in the cold until labeling.

Immunolabeling

Immunolabeling and basic imaging was performed as described previously in Refs. ^{4, 5, 11}, and ³⁶. For immunofluorescence, free-floating sections were permeabilized for 10 min at room temperature (0.4% vol/vol Triton X-100, approximately 23°C), rinsed in PBS, and blocked for 30 min (1% wt/vol bovine serum albumin, 5% vol/vol normal goat serum, and 0.1% vol/vol Tween 20 in PBS). Following blocking, the free-floating sections were incubated with primary antibodies diluted in blocking buffer overnight (16–20 h) at 4°C. The following day, sections were rinsed with PBS and incubated with secondary antibodies diluted in a blocking buffer for 2 h at room temperature. Secondary antibody solution was removed, sections were counterstained with 4’,6-diamidino-2-phenylindole (DAPI; 1 μg/mL) (Invitrogen, D357) for 20 min, washed, and mounted with ProLong Diamond Antifade Mountant (Invitrogen, P36961).

The primary antibodies were anti-latexin (mouse monoclonal IgG1, 1/500 working dilution, gifted from Y. Arimatusu, AB_2571634), anti-calretinin (rabbit polyclonal antiserum, 1/3000, Swant CR7697, AB_2619710), and anti-GAD67 (mouse monoclonal IgG2a, 1/1000, Sigma-Aldrich MAB5406, AB_2278725). The secondary antibodies were goat anti-mouse IgG (1/500, Jackson 115–545–003, AB_2338840) and goat anti-rabbit IgG (1/500, Jackson 111–295–003, AB_2338022).

Imaging

Single-plane confocal images were acquired with a 63×/1.4 NA plan-apochromatic oil lens on a Zeiss Axio Observer 7/LSM 800 inverted compound microscope and Zen Blue version 2.3 software (Carl Zeiss Microscopy). LED lighting (405, 493, and 575 nm) and a quad-band bandpass filter were used for widefield imaging. Confocal tiled 63× mosaic images were collected as 512 × 512 pixel tiles with a pixel dwell time of 1 μs, 2 × line averaging, and pinhole size at 1 Airy unit. Before a tiling series, the laser intensity was adjusted for each label to between 0.5% and 2%, and master gain was adjusted to between 650 and 850 V. Image stitching was performed in Zen 2.3 software with the Stitching function on the Alexa Fluor 488 channel (5% minimum overlap, 10% maximum shift).

Quantification

Cells were counted from tiled images of sections. Cells were identified and labeled (the process of cell segmentation) on the Alexa Fluor 488 (latexin) and Rhodamine Red-X (calretinin) channels with Cellpose.³⁷ Cells were classified as immunoreactive (-ir) as described in detail in Ref. 4. The cell diameter parameter was set to 40 pixels and the cell probability threshold to 50%. Images were processed in 2048 × 2048 pixel tiles with 5% edge overlap. Detections were exported in GeoJSON format (https://geojson.org/) and imported into QuPath (https://qupath.github.io/). Cell marking for plots was performed with QuPath 0.2.2 and Fiji/ImageJ software (https://imagej.net/software/fiji/). Latexin⁺/calretinin⁺ double-ir cells were identified by sequentially applying the latexin and calretinin classifiers on Alexa Flour 488 segmentations.

Section alignment

Sections were registered via affine transformation (preserving parallel lines but allowing scaling and shearing) to a single reference section to correct for (1) differences in section rotation on mounting and (2) normal variation in the shape of the claustrum. This alignment was accomplished by annotating each section for the claustrum based on latexin immunoreactivity and finding the affine transformation between that claustrum and a reference image. The resulting transformation matrices were applied to cell segmentations of the respective images. Pre- and post-alignment examples and the resulting post-alignments of all sections are given in the Results. Cell areas and circularities were calculated in QuPath (https://qupath.github.io/).

Statistics

Statistics were calculated in Python using the SciPy and Statsmodels packages.^38–40 Counts of double-ir (latexin⁺/calretinin⁺) cells, latexin⁺ single-ir, and calretinin⁺ single-ir neurons were compared between regions (claustrum core, claustrum shell, and neocortex) with Fisher’s exact test. Area values (lower bounded at 0) were transformed with the log function and circularity values (bounded between 0 and 1) were transformed with the logit function. Differences in area or circularity between cell types were tested by one-way analysis of variance (ANOVA) and Tukey’s honest significant difference test (Tukey’s HSD). A p value less than 0.05 was considered to be statistically significant.

RESULTS

Latexin⁺/calretinin⁺ double-immunoreactive (ir) neurons were identified, counted, and mapped in the claustrum and dorsal endopiriform nucleus. Figure 1 illustrates a sagittal brain section from C. perspicillata with the claustrum boundaries drawn as they were established in Ref. 4. As described in that original work, the claustrum regional boundaries were defined using latexin labeling with confirmation from labeling with parvalbumin or calretinin on the section in question or an adjacent section. The claustral shell subregion wraps the claustral core; located rostral to (as shown in the figure) and lateral to the core. The claustral shell can be further subdivided based on the density of latexin⁺ neurons into zones dorsal or ventral to the core with a relatively high density of latexin⁺ neurons and a central shell zone that is rostral (as shown in the figure) and lateral to the core with a relatively low density of latexin⁺ neurons. Note that the claustral shell is analyzed in this report as a single region (i.e., the specific subregions of the claustral shell as identified in Morello et al.⁴ are not distinguished here).

FIGURE 1.

Latexin and calretinin immunoreactivity in the claustrum of Carollia perspicillata. (A) Tiled confocal image (5×) of a sagittal section of the C. perspicillata brain labeled for latexin (green) and calretinin (red). Rostral is to the left and caudal to the right; dorsal is up and ventral is down. (B) Tiled confocal image (63×) to highlight the claustrum (Cl) and dorsal endopiriform nucleus (DEn) in (A). The overall boundary of the claustrum is shown as a solid black line, and the claustral core subregion is circled with a dotted line (based on Ref. 4). The claustral shell subregion is thus the part of the claustrum that surrounds the core subregion. Orange circles surround cells identified as latexin⁺/calretinin⁺ double-ir. Scale bar is 1 mm in both (A) and (B). Abbreviations: CA3, area CA3 of hippocampus; Cl, claustrum; DEn, dorsal endopiriform nucleus, DG, dentate gyrus, La, lateral nucleus of the amygdala, P, putamen.

Latexin⁺/calretinin⁺ neurons (marked with orange circles in Figure 1) occur in the claustral shell subregion (predominantly the ventral part of the claustral shell) and the neocortex adjacent to the claustral shell, but not in the claustral core. Double-labeled cells were also rare in the dorsal endopiriform nucleus, even though latexin-ir were common in the endopiriform nucleus (see also Ref. 4), and calretinin⁺ neurons were found in both the dorsal endopiriform nucleus and the neighboring anterior amygdaloid area, cortical amygdaloid nucleus, and deep pyriform area.⁴¹ Latexin⁺/calretinin⁺ neurons were not detected in these areas. Figure 2 illustrates cells that are positive for latexin, calretinin, or both at high magnification. Figure 3 illustrates labeling for GAD67, calretinin, and their combination with other high magnification examples.

FIGURE 2.

Latexin and calretinin immunoreactivity in the claustral shell subregion to illustrate single- and double-labeling. Panels show latexin (left), calretinin (center), and merged (right) images; nuclei labeled with DAPI appear blue in all three images. The arrowhead in the left and right panels marks a latexin⁺/calretinin⁻ neuron. Double arrows in the center and right panels mark a latexin⁻/calretinin⁺ neuron. The single arrow in all three panels marks a latexin⁺/calretinin⁺ neuron. The latexin⁺ neuron near the left edge of the left and right panels is not considered to be double-ir. The calretinin labeling in this cell was not sufficient to count the cell among the calretinin⁺ neurons. Scale bar is 100 μm.

FIGURE 3.

GAD67 and calretinin immunoreactivity in the claustral shell subregion to illustrate single- and double-labeling. (A) GAD67-ir (green; left panel), calretinin-ir (red; center panel), and merged (right panel) views of C. perspicillata claustral neurons. Neurons marked with arrowheads (B), (C), and (D) are shown at higher power in rows labeled (B), (C), and (D), respectively, below. Scale bar for the top row (A) is 50 μm. Scale bars in rows (B), (C), and (D) are 10 μm. Row (B) shows a GAD67⁻/calretinin⁺ neuron. Row (C) shows a GAD67⁺/calretinin⁺ double-ir neuron. Row (D) shows a GAD67⁺/calretinin⁻ neuron. The cells shown in rows (C) and (D) would be considered inhibitory. The cell in row (B) might be an excitatory calretinin neuron of the type identified by colocalization of latexin-ir, but latexin labeling in this cell was not tested. Note that we do not show preparations that are triple-labeled to exclude latexin. The nonoverlapping labeling of latexin and GAD67 was illustrated in Morello et al. (Fig. 12 in Ref. 4).

Calretinin-containing neurons were found in the claustrum, dorsal endopiriform nucleus, neocortex, and amygdala. Within the claustrum, calretinin neurons were distributed throughout the shell but rarely within the core. For localization purposes, we used an alignment method (see Methods) that enabled the registration of data across sections. All quantified sections were taken from approximately the same level in sagittal sections (e.g., Figure 1). Figure 4 illustrates the alignment method (Figure 4A) and registration results for eight sections (Figure 4B). Figure 5 illustrates the resulting mapping of cell identities overlaid on composite maps for latexin⁺, calretinin⁺, and latexin⁺/calretinin⁺ double-ir neurons. This figure highlights the density of latexin labeling in the claustral core subregion and the contrasting preferred localization of calretinin⁺ neurons in the claustral shell subregion. The latexin⁺/calretinin⁺ double-ir neurons are a subpopulation located primarily in the claustral shell.

FIGURE 4.

Illustration of the method used to align sections and the results of example transformations. (A) Overlay of two sections pre- and post-transformation. Cyan and magenta represent latexin-ir neurons from two separate sections. (B) Post-registration of sections used in the analysis. Magenta and cyan markers represent latexin⁺ cells. Magenta markers represent a single reference section (same as in A). Cyan markers represent latexin-ir from each other section. The alignment of the reference with itself is shown in the upper left section of panel (B). As shown, latexin⁺ neurons and the claustrum boundaries were well-aligned after transformation.

FIGURE 5.

Localization summary for all latexin and calretinin in the Carollia perspicillata claustrum for all animals studied. Plots of neuronal somata labeled for latexin (latexin⁺/calretinin⁻; green; left), calretinin (latexin⁻/calretinin⁺; red; center), and both (latexin⁺/calretinin⁺; orange; right). The boundary of the claustrum is shown as a solid black line, and the claustral core subregion is circled with a dotted line. The heavy density of latexin⁺ neurons in the claustral core and the relative absence of calretinin⁺ neurons from the core are evident.

Latexin⁺/calretinin⁺ double-ir neurons were found distributed throughout the claustrum shell and overlying cortex but rarely within the claustrum core (Figure 5). They clustered in latexin-ir areas with the highest calretinin⁺ neuron density. These double-ir neurons were not common in the dorsal endopiriform despite the dual presence of calretinin⁺ and latexin⁺ neurons in this region (Figure 6). The few latexin⁺/calretinin⁺ double-ir cells found in the claustrum core were located at the periphery of the core.

FIGURE 6.

Localization summary for calretinin⁺ and latexin⁺/calretinin⁺ neurons in the Carollia perspicillata claustrum for all animals studied. The latexin⁻/calretinin⁺ neurons are represented as red circles and the latexin⁺/calretinin⁺ neurons as orange triangles. The claustrum boundary is as described earlier. The boundary for dorsal endopiriform nucleus is drawn below the claustrum. Note in the dorsal endopiriform nucleus, the presence of many calretinin⁺ neurons contrasting with a near complete absence of latexin⁺/calretinin⁺ double-ir cells.

Neurons double immunoreactive for latexin and calretinin were prominent in the claustrum shell and in the cortex immediately outside the claustrum (Table 1). These neurons made up a small percentage of the total population of latexin⁺ neurons (0.9% in shell, 2.2% in cortex). Within the claustrum shell, latexin⁺/calretinin⁺ neurons comprised 25% of all calretinin⁺ neurons of the claustral shell. Within the claustrum core, the total number of calretinin⁺ cells was much smaller, and latexin⁺/calretinin⁺ neurons comprised 33% of this population of claustral core calretinin⁺ neurons. There were no statistical differences between anatomical regions (claustrum core, claustrum shell, neocortex) for the proportions of latexin⁺/calretinin⁺ neurons relative to calretinin⁺ neurons (core to shell, core to cortex, shell to neocortex: p = NS; counts and proportions are shown in Table 1). However, statistically significant differences in the proportions of latexin⁺/calretinin⁺ double-ir cells or calretinin⁺ cells relative to latexin⁺ cells between anatomical regions were found (p values from Fisher exact test comparisons of counts of double-labeled to latexin-labeled cells were 0.0003 core vs. shell, <0.00001 core vs. neocortex, and 0.0001 shell vs. neocortex. p Values from comparisons of counts of calretinin-labeled to latexin-labeled cells were <0.00001 core vs. shell, core vs. neocortex, and shell vs. neocortex. Degrees of freedom for Fisher exact test = 1. If we apply a Bonferroni correction for multiple testing by dividing alpha by 8, 0.05/8 = 0.0083 becomes the corrected alpha.).

TABLE 1.

Cell types based on latexin and calretinin labeling by region.

Region	Latexin (LXN) total cell count	Calretinin (CR) total cell count	LXN-CR cell count	$\frac{\text{LXN}-\text{CR}}{\text{TOTAL}\text{LXN}}\%$	$\frac{\text{LXN}-\text{CR}}{\text{TOTAL}\text{CR}}\%$
Claustrum core (pooled)	3242	16	8	0.2%	33%
Claustrum core (avg. single section; mean± SEM)	405.3±27.6	2.0±0.7	1.0±0.7	0.2%	50%
Claustrum shell (pooled)	3593	105	34	0.9%	25%
Claustrum shell (avg. single section)	449.1±25.4	13.1±2.5	4.3±1.8	0.9%	32%
Cortex (pooled)	1914	202	44	2.2%	18%
Cortex (avg. single section)	239.3±27.7	25.2±6.3	5.5±2.1	2.3%	21.8%
Endopiriform nucleus* (pooled)	2003	837	7	0.3%	0.8%
Endopiriform nucleus* (avg. single section)	250.4±21.2	104.6±17.3	0.9±0.4	0.3%	0.8%

Note: Values shown for each brain region are the pooled total counts and the fractions of latexin⁺/calretinin⁺ double-versus single-immunoreactive neurons in eight brain sections, as well as the mean ± SEM per single section (percentages at the right are based on the mean values). Note that the claustral shell is analyzed as a single region (i.e., the specific subregions of the shell as identified in Morello et al.⁴ are not distinguished here).

^*The entry counts for dorsal endopiriform nucleus include small portions of the anterior amygdaloid area, cortical amygdaloid nucleus, and deep pyriform area⁴¹ where calretinin⁺ neurons can occur. Double labeled cells were not detected in these areas.

Abbreviations: CR, cells immunoreactive for calretinin only; LXN, cells immunoreactive for latexin only; LXN/CR, cells double-immunoreactive for latexin and calretinin.

Analysis of cell morphometrics (area and circularity) revealed differences between each cell type. In the claustrum and neocortical regions, latexin⁻/calretinin⁺ neurons had the largest areas, followed by latexin⁺/calretinin⁻ neurons, and finally latexin⁺/calretinin⁺ neurons (Table 2). These differences in areas between cell types were statistically significant (F(9,391) = 19.37, p =< 0.001, Tukey HSD test statistic = 25.698) See Table 3 for the individual post hoc comparisons and p values. Differences in circularity values between all cell types were < 2%. The statistical comparisons of cells by soma size (Tables 2 and 3) are intended to reinforce the immunohistochemical classes based on the presence or absence of latexin or calretinin. In other words, the soma sizes support the idea that there are three cell types: latexin⁺/calretinin⁻, latexin⁻/calretinin⁺, and latexin⁺/calretinin⁺. The comparisons of cells by subregional localization (Table 4) are intended to offer relative abundance comparisons for the different cell types.

TABLE 2.

Area and circularity measures for cell types based on latexin and calretinin immunolabeling.

Class	Area (μm2)	Circularity
Calretinin	120.3 ± 37.3	0.60 ± 0.04
Latexin	102.4 ± 33.6	0.58 ± 0.04
Latexin-calretinin	91.9 ± 37.0	0.57 ± 0.04

Note: Morphometric comparison of all latexin and calretinin double- and single-immunoreactive neurons listed in Table 1. Values presented as mean ± standard deviation. Circularity calculated as: $4\pi \times (\frac{Area}{Perimete{r}^2})$.

TABLE 3.

Mean differences in cell soma area by cell type.

Cell type 1	Cell type 2	mean difference in area (cell type 2 – cell type 1)	p value	95% Confidence interval
Calretinin	Latexin	−12.3	<0.0001	[−16.7, −7.8]
Calretinin	Latexin-calretinin	−23.2	<0.0001	[−32.8, −13.7]
Latexin	Latexin-calretinin	−11.0	0.0071	[−19.5, −2.5]

Note: Results of Tukey’s HSD Test for multiple comparisons of mean area between cell types. F = 19.37, p < 0.001; Tukey HSD test statistic = 25.698, degrees of freedom = 9391.

TABLE 4.

Immunohistochemical cell types in the Carrolia perspicillata claustrum.

Cell type	Latexin	GAD	Calretinin	Parvalbumin	Calbindin	Claustral localization
Excitatory
E1	+	−	−	−	−	Both
E2	−	(−)	(−)	(−)	(−)	Shell
E3	+	−	+	−	−	Shell
Inhibitory
I1	−	+	−	+	−	Core
I2	−	+	+	−	−	Shell
I3	−	+	−	−	+	Both

+ labeled.

− not labeled.

(−) individual cells maybe difficult to classify, but total cell counts establish this class.

Note: E1, E2, E3: excitatory cell types based on labeling patterns, for example, cell type E1 is latexin immunoreactive, but not immunoreactivefor the other listed markers. I1, I2, I3: inhibitory cell types based on labeling patterns, for example, cell type I1 is labeled with GAD and parvalbumin. Gray shading identifies the cell class that is the focus of this manuscript.

Sources: Morello et al.⁴ and this manuscript.

DISCUSSION

We describe the distribution and morphometric features of a novel, putative excitatory claustral cell type in C. perspicillata brain that is defined by its dual immunoreactivity for latexin and calretinin. Within the claustrum, these neurons were localized to the claustral shell and not the claustral core. Latexin⁺/calretinin⁻ neurons were also scattered in the neocortex immediately superficial to the claustrum. The significance of our findings is that we show that there are multiple types of excitatory neurons in the claustrum core and shell—the discovery of one of these is the subject of this paper—just as there are multiple types of inhibitory neurons.

Whereas latexin and calretinin have been shown to typically mark excitatory and inhibitory cells, respectively,^12,17 we suggest that the latexin and calretinin coexpressing claustral cells are excitatory neurons. As described in the Introduction, the vast majority of claustral latexin-ir neurons colocalize with glutamate, and no claustral latexin-ir neurons have been found to be colabeled with GABA or GAD67, two markers of inhibitory neurons^4,12 (see fig. 12 in Morello et al.;⁴ see also Figure S1). Therefore, we believe that the latexin⁺/calretinin⁺ double-ir neuron represents an excitatory neuron subtype.

Calretinin-ir excitatory projection cells^42–44 and excitatory interneurons^45,46 have been described previously in other brain areas. Calretinin has been shown to colocalize with VGLUT2, for example, in the nucleus accumbens,⁴⁷ entorhinal cortex,^48,49 and cerebellum.⁵⁰ An additional possibility exists with reports of neurons capable of coreleasing glutamate and GABA.^51–55

Whereas colocalization of calretinin with GAD67 was used to identify calretinin-ir cells as inhibitory neurons,⁴ we are not able to determine if the population of neurons that is GAD67⁻/calretinin⁺ is the same population as the latexin⁺/calretinin⁺ population. Immunolabeling for latexin as well as for calretinin is robust, and thus the double labeling is typically clear (an example of a cell whose calretinin labeling was questionable and thus not counted as calretinin-ir is shown in Figure 2). The double labeling with calretinin and GAD, however, can be less clear (due mainly to the properties of GAD labeling), especially when attempting to make the determination that labeling is absent.

Based on our work, the immunohistochemically identified cell types of the C. perspicillata claustrum can be summarized in Table 4. The differential distribution of the various claustral cell types between core and shell of the claustrum further supports the notion of a core and shell division of the claustrum.^4,56,57 Interestingly, although latexin⁺ and calretinin⁺ neurons were codistributed in the dorsal endopiriform nucleus, no latexin⁺/calretinin⁺ double-ir neurons were found in the endopiriform nucleus. latexin⁺/calretinin⁺ double-ir neurons were also not found in retrosplenial cortex, where we reported an abundance of latexin⁺ neurons in area 29ab.¹¹ Colocalization of latexin with parvalbumin or calbindin has also not been seen in Ref. 4.

While the function of the latexin⁺/calretinin⁺ double-ir cell remains unknown, the main message is that the claustral shell is structurally complex. All three types of excitatory cells exist in the shell and E2 and E3 are predominantly elements of the shell (not the core). The inhibitory elements present in the claustral shell are also different from the core subregion.

In our opinion, the existing data from this and previous studies do not yet allow the cross-association of immunohistochemical cell types with the electrophysiological cell types,^58–60 morphological cell types,^25,61–64 projection patterns,^8,65 and/or the cell types defined by RNA sequencing⁵⁶ or gene expression.⁸ The place where it is most likely possible to draw associations from the existing data is in the claustral core subregion. The E1 class of cells (latexin⁺/calretinin⁻) are the main, if not exclusive, excitatory cell type of the claustral core and the I1 (parvalbumin+) class of inhibitory cells is the predominant inhibitory cell type of the claustral core.

The latexin⁺/calretinin⁺ cells were found in both hemispheres and both biological sexes in the C. perspicillata claustrum, but we did not build a dataset for the purpose of testing for sex-based or hemispheric differences in cell frequency or localization. No obvious differences were noted. Figure S2 illustrates cells from both hemispheres of both sexes. We do not have data in this paper to address questions about properties, proportions, or localization of this cell type in other species (e.g., rodents). Given the multiple similarities between the C. perspicillata brain and primate brain,³³ it will be interesting to explore phylogenetic conservation or divergence in the claustrum’s intrinsic circuitry. Recognizing that the claustral shell is quite thin in the rodent brain, the cell types identified in the bat claustral shell set a target for comparison to identify cells (and circuits) that are phylogenetically conserved.

We recognize that our study is limited to identifying the existence of a class of latexin⁺/calretinin⁺ double-ir neurons in the claustrum, their structural similarities and differences with latexin⁺ and calretinin⁺ single-ir neurons, and the spatial localization of this newly identified cell type. Cataloging the cell types of the claustrum is a necessary step toward defining its intrinsic circuitry and ultimately understanding its function. We note, however, that the electrophysiological properties and connectivity patterns for these cells have not yet been studied by us in C. perspicillata or in any species to our knowledge. Data on counts, connectivity, and functional expectations for the various claustral cell types will be necessary to determine the functional consequences of any differences in these metrics. Changes in cell number are perhaps most likely to be significant and functionally disruptive in aging or disease where one or another cell type may be preferentially impacted. C. perspicillata provides an extraordinary opportunity to address these questions.

Supplementary Material

SUPPORTING INFORMATION

Additional supporting information can be found online in the Supporting Information section at the end of this article.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.