Male individuals of the house wren (Ta) (n = 9) and the rufous-tailed hummingbird (At) (n = 4) were collected at the University of Antioquia – Campus Sede Oriente, Carmen de Viboral. The same sampling effort was made for both species in two reproductive seasons; however, collecting house wren was much easier due to their territorial behavior and abundance. Juvenile and female birds were avoided to preserve their population index and to discount any neurobiological differences due to age. These species are considered common and abundant, classified as “Least Concern” in Colombia’s birds red list (Echeverry-Galvis et al., 2022). Additionally, they are widely distributed and relatively common in both urban areas and agricultural landscapes. Furthermore, these species have a broad distribution and are relatively housed in urban areas and agroscapes (Scott, 2005). The collection of biological specimens is covered by the framework permit for the collection of wild specimens for non-commercial purposes, which was issued by ANLA in resolution 1,461 of December 3, 2014. The size sampling permission and the procedure was approved by the Ethics Committee for Animal Experimentation at the University (CEEA), resolution 139 of March 29, 2021.

Birds were captured using the “Manual of Methods for Biodiversity Inventory Development” (Álvarez et al., 2004). They were collected using four 12 × 2 m mist nets with 30 mm mesh eye nets, each checked every 20 min from 6:00 to 13:00 h. Captured individuals were identified as sacrificed, and various morphometric parameters were measured, including sex, mass, exposed and total culmen length, beak height and width, closed wing length, and tarsus length. For euthanasia, isoflurane was administered at a dose of 2 mL per 125 g B.W. or less in a sealed box containing a cotton ball soaked with the anesthetic. The box was allowed to saturate with the anesthetic for at least 3 min, and then the bird was placed in the anesthesia chamber, the lid was closed, and the birds were left undisturbed for 30–60 s until they became immobile. If they remained active after this time, the dose was re-administered. After the last breath of the bird, a 30-s waiting period was observed, then decapitation was performed, and the skull was dissected to extract the brain. During the procedure, a portion of the metatarsus, breast, liver, and gonads were also collected for pathological analyzes and hormonal status. The brain of one individual of the house wren was not extracted but was preserved in a 4% paraformaldehyde solution, along with the bony part of the head. Table 1 depicts the collected individuals’ information.

After extracting the brain, it was washed with PBS 1X solution and then fixed in 4% paraformaldehyde in a cytoskeleton buffer with changes every 24 h for 3 days (Posada-Duque et al., 2013). Subsequently, a sucrose gradient was performed with concentrations of 7, 25, and 30% over three consecutive days, increasing the concentration daily. The brains that were not used immediately were stored at −20°C in a cryopreservative solution; the rest were sectioned using a cryostat.

One brain was selected for each of the two species. Once the sucrose gradient process was completed, the brains were washed with phosphate Buffer (PB) (0.1 M pH: 7.4) and a sagittal cut of the hemispheres was made using a blade. Then, using a 0.5 mm needle, four lateral-to-medial perforations were made along each hemisphere. This was done to subsequently locate the perforations in the slices and align the tissues for the 3D reconstruction of the main vocal areas. Next, parasagittal sections of the entire hemisphere were prepared using the cryostat (LEICA CM1850 UV). Each hemisphere embedded in OCT was positioned laterally on the specimen holder, with a weight on top, and allowed to freeze (approximately 3–5 min). Then, the entire hemisphere was sliced using a blade (LEICA 819 - Low Profile) with blade changes every 20 cuts, and the sections stored in 48-well plates with PB 0.1 M solution (pH: 7.4) + 0.1% Azide. Later, the sections were mounted on labeled glass slides (SuperFrost PLUS-0006E) in the correct order and left to dry in a dust-protected tray.

A gradual rehydration was performed by slightly tilting each slide to quickly impregnate the tissues with 100, 70, and 50% alcohol. Then, excess alcohol was removed, through rinsing with distilled water, and allowed to drain. Next, the tissues were stained with Toluidine Blue dye [1:10] and left to act, followed by rinsing with distilled water and draining. Subsequently, dehydration was carried out using 70, 96, and 100% alcohol. The final dehydration and clearing were done in an extraction chamber using xylene, and Shandon Consul-Mount (Thermo Scientific; 9,990,440) was added to adhere the glass cover, and then the slides were left to dry.

All the Nissls slides were scanned with a NanoZoomer-XR brightfield tissue scanner microscope (Hamamatsu), equipped with a 20x objective (NA 0.75; UPlanSApo; Olympus) using a 40x digital zoom at a single layer. They were opened with QuPath software (Bankhead et al., 2017) to perform quality control (excluding tissues with tears or folds) and export each tissue as TIFF format, considering the order of each tissue for subsequent analysis with Stereo Investigator software (MBF Bioscience - MicrobrightField, Version: 2021). One hemisphere from each species was selected to draw and obtain the 3D reconstruction of the vocal areas and perform volumetric quantification using Neurolucida Explorer software (MBF Bioscience - MicrobrightField, Version: 2021).

For the delimitation and reconstruction of the vocal areas, the following parameters were considered: i. A detailed review of the research by Karten et al. (2013) and Stokes et al. (1974) for the house wren, and the research by Jarvis et al. (2000) and Gahr (2000) for the rufous-tailed hummingbird, in order to locate the vocal areas and identify the parameters they had used. ii. The sequential order of the tissues for each species. Iii. Histological description of the shape of each area, cell labeling, and cell clustering; type of cellular labeling that distinguishes the center or interior from the border or exterior of each area. iv. Sequential and complete drawing of each of the vocal areas. v. Location of markers (icons) on the perforations made with the needle for 3D rotation and reconstruction. vi. Rotation of the tissues to fit into the tracings and markers made at each hole created with the needle. Vii. Validation of the vocal areas in a different individual from the 3D reconstruction using immunohistochemistry for the nuclear marker mouse anti-NeuN (Sigma-Aldrich; MAB377; [1:250]) and with mouse anti-MAP2 (Sigma-Aldrich; M9942; [1:250]), using different tissues that included the initial and final bregmas of the main areas of the song (Supplementary Figure S1A).

After recognizing the anatomical vocal brain nucleus, its most medial and lateral limits (in μm from midline) were determined. To compare the location and width of a given song nucleus between both species, the nucleus’s relative mediolateral limits for each species were calculated, considering 0 as the midline and 1 as the brain maximal extension. The song nuclei’s maximal width was calculated by subtracting each nucleus’s most lateral and medial limits.

From the Nissl-stained tissues used for 3D reconstruction, 50% of the bregmas encompassing each of the four vocal areas were selected, and cell counting was performed. Cellular density analysis was conducted using a fractionated scanning method with equidistant grids to count cells in the area of interest. Cell counts were performed within these grid points and then extrapolated to the entire area of interest. The “Area Fraction Fractionator” tool in the Stereo Investigator software allowed for the estimation of cell counts within an area of interest using predetermined parameters for grid size and distance between grid points, which were then extrapolated to the total measured area (Supplementary Figure S1B). Using the “icon” option, each cell within the grid was marked without touching the lower or left boundary for each grid point in the entire area of interest. At the end of this process, the cell counts were extrapolated to the size of the total area, ensuring that the Cruz-Orive/Geiser, Schmitz-Hof, and Scheaffer error coefficients were below 0.1 to verify the accuracy of the sampling.

Images scanned from Nissl stained slices within the vocal brain area per specimen were processed and analyzed using QuPath software (Bankhead et al., 2017). Whole nuclei were delineated, and cells per 1 mm² from three slices were quantified from whole delineated nuclei of the house wren (n = 7) and rufous-tailed hummingbird (n = 4).

Sections of 30 μm were obtained using a cryostat, similar to those used for the 3D reconstruction of vocal brain areas. Antigen retrieval was performed using 1X citrate buffer (pH: 6) (Master-diagnostic; MAD-004071R/D) at 85°C for 20 min with constant dripping onto the tissues. Subsequently, endogenous peroxidase activity was blocked with a peroxidase blocker (Master-diagnostic; MAD-021540Q-125) for 20 min to prevent nonspecific antibody binding. Samples were incubated in 1% bovine serum albumin (BSA; Sigma-Aldrich, A9647), 0.3% Triton-X100 (Sigma-Aldrich, T9284-500ML), and 0.1 M PB (pH: 7.4) for 1 h at room temperature with constant agitation. This was followed by a 72-h incubation at 4°C in a primary antibody solution containing anti-GFAP mouse (Sigma-Aldrich; G3893 – RRID: AB_477010; [1:250]), anti-S100β rabbit (Dako-Agilet; Z0311 – RRID: AB_10013383; [1:250]), anti-MAP2 mouse (Sigma-Aldrich; M9942 – RRID: AB_477256; [1:250]), or anti-NeuN mouse (Sigma-Aldrich; MAB377 – RRID: AB_2298772; [1:250]) diluted in 0.3% BSA, 0.3% Triton-X100, and 0.1 M PB with constant agitation. After 20 min of washing off excess antibodies, the tissues were incubated for 2 h at room temperature with biotinylated goat anti-mouse secondary antibody (Invitrogen; 31,800 – RRID: AB_228305; [1:250]) or biotinylated goat anti-rabbit secondary antibody (Invitrogen; B-2770 – RRID: AB_2536431; [1:250]) diluted in 0.3% BSA, 0.3% Triton-X100, and 0.1 M PB with constant agitation. Subsequently, the tissues were vigorously washed in 0.1 M PB solution three times for 5 min each, followed by a 1-h incubation in the avidin-biotin peroxidase standard staining kit (Thermo Scientific; 32,020) at room temperature. Development was carried out for 3–4 min using 3,3′-diaminobenzidine tablets (Sigma-Aldrich; D4293). The tissues were then mounted on glass slides and subjected to sequential dehydration with ethanol (70, 96, 100%) and xylene. Finally, without allowing the tissues to dry with xylene, coverslips were mounted using Shandon Consul-Mount (Thermo Scientific; 9,990,440) for drying and observation under the microscope.

Whole brain imaging for GFAP slides [Ta (n = 6) At (n = 4)] was scanned with a Ventana-DP 200 brightfield tissue scanner microscope (Roche), equipped with a 40x objective (NA 0.75; UPlanSApo; Olympus) using a 40x digital zoom at a single layer.

Unique cell imaging for fractal morphological analysis of astrocytes was performed using brightfield microscope imaging. For this purpose, five cells were recorded per telencephalic region in the pallium (Laminar edge of pallium (LEP) and vascular pallium), the pallidum, and the mesencephalon for Ta (n = 6) and At (n = 3). These cells were randomly selected and captured using an Olympus CX35 microscope (Model X31RBSFA) equipped with a 100x oil immersion objective (NA 1.25; PlanC N; Olympus) and a Swift camera (SC1803R).

Image processing of Nissl and GFAP was performed in QuPath 0.3.2 software for Ta (n = 6) and At (n = 4) (Bankhead et al., 2017). Quantifying the percentage of GFAP area in pallium was performed using a particle detector using the same threshold, and the GFAP area was shown by heatmap.

The single-cell images were processed, segmented, and analyzed using FIJI software (NIH ImageJ) (Schindelin et al., 2012). For the morphological analysis of astrocytes, the images were converted to 8-bits, and astrocyte somas were manually segmented using the brush selection tools. The Simple Neurite Tracer plugin was utilized to segment the cellular processes, and then a merge of the soma and processes of each astrocyte was obtained (Longair et al., 2011). Subsequently, morphological parameters of the soma, cellular processes, and the entire cell were obtained using the FracLac plugin (Karperien, n.d.) and FIJI tools. Finally, based on Fernández-Arjona et al. (2017), morphological characteristics related to complexity (fractal dimension, lacunarity, roughness, density, and compactness), size (area, perimeter, and major axes), shape (circularity and aspect ratio), and spatial domain (scaling and shape features of the convex hull, bounding rectangle, and fitted ellipse) were analyzed.

The brains of songbirds were extracted and fixed as previously described. Subsequently, the samples were sectioned into parasagittal slices of 30 μm thickness using a Leica cryostat for songbirds. Before immunostaining, antigen retrieval was performed on the songbird slices using citrate buffer at 95°C for 20 min, respectively. Autofluorescence was blocked using 50 mM NH₄Cl prepared in H₂O. The samples were incubated in 1% bovine serum albumin (BSA, Sigma-Aldrich, A9647) for 1 h at room temperature to prevent nonspecific antibody binding. The brain sections were incubated for 72 h at 4°C in primary antibodies, rabbit anti-GFAP (Sigma-Aldrich; ab5804 – RRID: AB_305124; [1:250]) with mouse anti-NeuN (Sigma-Aldrich; MAB377 – RRID: AB_2298772; [1:250]), diluted in an antibody solution containing 0.3% BSA, 0.3% Triton-X100, and PB (0.1 M, pH 7.4). After removing excess antibodies through a 20-min wash, the sections were incubated for 2 h at room temperature in a secondary antibody solution with goat anti-mouse Alexa Fluor 488 (Invitrogen; A-11001; 1:250) and goat anti-rabbit Alexa Fluor 594 (Invitrogen; A-11012; [1:250]); the nuclear marker Hoescht (Vector Labs; DL-1068; [1:2500]) was also incubated. Subsequently, the samples were vigorously washed in PB (0.1 M) three times for 5 min each. Finally, the sections were mounted on glass slides with FluorSave Reagent (Millipore; 345,789).

Astrocytes in the telencephalon and mesencephalon for bird song were captured using confocal microscopy. Three high-magnification images per slide were obtained from the triple immunostaining, and they were captured using an Olympus FV1000 scanning confocal microscope equipped with a 60X oil immersion objective (NA 1.42; PLAPON; Olympus) with a zoom factor of 4, three lasers at 488, 594 nm, and DAPI in FluoView 3.1.1.9 software (Olympus). Sixteen-bit TIFF images of 1,024 × 1,024 pixels (105.47 × 105.47 μm) were obtained with an XY pixel size of 103 nm and a 400 nm spacing between Z sections. 26 optical sections were captured from each field (with a thickness of 10 μm). In addition, we created a confocal mosaic image for each bird to generate a complete representation of the brain slide in GFAP areas, using a 10X air objective (NA 0.4; APLANPOS; Olympus). For the house wren, we used a 9×9 grid, while for the hummingbird, we used a 7 × 7 grid, and the final reconstructed images were achieved using FluoView 3.1.1.9 software (Olympus).

The confocal images were deconvoluted, processed, segmented, and 3D reconstructions were created. Image deconvolution was performed using Huygens Essential 23.3 software (Scientific Volume Imaging B.V.). The classic maximum likelihood estimation (CMLE) algorithm was used for image deconvolution with a signal-to-noise ratio of 21, and the bird images were deconvoluted using the deconvolution assistant. The images were converted to 8 bits and subsequently processed and analyzed in FIJI software. Immunofluorescence signals were segmented using intensity thresholding with the Huang algorithm to standardize fluorescence signals across all images. Z-projections using the standard deviation of segmented stacks were employed to assess astrocyte structure. These Z-projections were segmented and used to quantify astrocyte processes. Finally, for illustrative purposes, surface rendering was performed to display the 3D projections of deconvoluted images using the 3D Surface rendering option in Huygens Essential 23.4 (Scientific Volume Imaging B.V.).

The astrocyte processes were counted using deconvoluted images. Excess background illumination was subtracted from a value of 25 using the Image-Pro Plus Subtract tool. The images were processed, segmented, and analyzed using FIJI software (NIH ImageJ) (Schindelin et al., 2012). For crossing number of astrocyte processes, the images were converted to 8-bits and segmented using intensity thresholding by the Huang algorithm. Skeletonization was carried out, and the processes stemming from the soma were quantified using the Simple Neurite Tracer plugin, employing Sholl analysis to count the number of intersecting processes. Finally, the images were processed to obtain the representative figures.

For the analysis of cell density in each of the vocal brain regions of the house wren (n = 7) and the rufous-tailed hummingbird (n = 4), cells were quantified in three sections, blind to the conditions. The GFAP area percentage was quantified in two sections, blind to the conditions. We performed a descriptive analysis of the distribution of each cell detected in the vocal areas through box and whisker plots, showing the median (middle line), 25th to 75th percentiles (box limits), and Min to Max value (whiskers limits) of house wren (green), and rufous-tailed hummingbird (orange), depicting cellular area quantifications performed in QuPath. Data normality was assessed using the Shapiro–Wilk test. A non-parametric Mann–Whitney test was conducted to compare cell density between the house wren and the rufous-tailed hummingbird.

Fractal morphology analysis and astrocyte intersections (Sholl analysis) were performed using five images per brain region for each specimen, house wren (n = 6) and hummingbird (n = 3). Data normality was assessed using the Shapiro–Wilk test. For parametric univariate data, a one-way analysis of variance (ANOVA) was employed, followed by Tukey’s multiple comparison tests to compare astrocytes within LEP, FLP, and FRL of the house wren. A non-parametric Mann–Whitney test was conducted to compare the fractal morphology and Sholl analysis of astrocytes in LEP between the house wren (n = 7) and the rufous-tailed hummingbird (n = 3). All groups were processed simultaneously to mitigate potential experimental variations. Statistical analyzes were performed using GraphPad Prism software, version 8.0. Significance levels were determined as follows: * for p < 0.05, ** for p < 0.01, and *** for p < 0.001.

Body condition reflects the quality of life related to survival, reproduction, and behavior. The wild birds collected during the breeding season and the morphometric patterns obtained were related to adult birds based on established criteria (Table 1) (Scott, 2005; Johnson and Clark, 2020). The brain volume-to-body weight ratio was 19,008 mm³/g for the house wren and 17,323 mm³/g for the rufous-tailed hummingbird.

Among bird species, there are variations in the location, shape, and volume of the vocal areas (Jarvis et al., 2000; Mooney, 2009b). From the manually delimited contours for the entire hemisphere and all the vocal areas, we generated a 3D brain reconstruction to describe the position and volume of representative vocal areas. In the house wren, the vocal areas were closer to the midline, whereas in the rufous-tailed hummingbird, they were at an intermediate location between the lateral portion and midline (Figure 2; Table 2). The house wren showed LMAN and Area X in an anterior hemisphere position, while RA and HVC were in a posterior location (Figure 2A; Supplementary Figures S2A,B; Table 2). Despite the rufous-tailed hummingbird showing similarity in the location of LMAN, Area X, and RA; the HVC area spans both the anterior-dorsal and posterior regions of the hemisphere, being longer but narrower compared to the house wren (Figure 2B; Supplementary Figures S2C,D). These findings suggest a differential spatial localization of vocal brain areas between the house wren and rufous-tailed hummingbird.

Starting from the 3D construction of each nucleus, we determined the most medial and lateral limits and, hence, were able to calculate their mediolateral extensions (widths). Then, their widths were relativized to the maximal mediolateral extension of the hemisphere (Table 2). Interestingly, relative to the hemisphere extension, the house wren showed a wider HVC compared to the rufous-tailed hummingbird, while Area X was wider for the rufous-tailed hummingbird, compared to the house wren, suggesting differences in the size of production and learning vocal areas between both species.

We described the cellular morphology and distribution of vocal areas for the house wren and the rufous-tailed hummingbird based on whole and unique cellular imaging of the vocal areas and the quantitative area registration of each cell contained in each region (Figure 3). In the house wren and the rufous-tailed hummingbird, the LMAN area showed a volume of 0.0993 and 0.2018 mm³, respectively (Figure 3C). We observed large and elongated cells with a pyramid shape and intense staining, surrounded by scattered cells forming a sphere in the house wren; contrastingly in the rufous-tailed hummingbird, the LMAN area showed smaller elongated cells, closely packed together, surrounded by scattered cells forming an oval shape (Figures 3C,G.I; Supplementary Figure S3A). Interestingly, the LMAN of the rufous-tailed hummingbird showed an intense patch of blue color, suggesting a possible higher cellular density compared to the house wren (Figure 4A; Supplementary Figure S3A).

Area X had a volume of 0.6764 and 0.0405 mm³ in the house wren and the rufous-tailed hummingbird, respectively (Figure 3D). In the house wren, Area X showed various types of cells, including small, round, elongated, and big cells dispersed throughout the area, with intense-staining pyramid-shaped cells. Its shape resembled a water droplet, narrower at the upper dorsal and rounder at the lower ventral part. While in the rufous-tailed hummingbird, Area X was spherical and composed of cells of various sizes, with a predominance of big cells in clusters (Figures 3D,G.II; Supplementary Figure S3A).

The RA area in the house wren and the rufous-tailed hummingbird had volumes of 0.1694 and 0.0432 mm³, respectively (Figure 3E). In the house wren, the RA area was mainly comprised of pyramid-shaped cells spaced apart, with intense staining distinguishing them from the surrounding environment. An oval shape on the area’s outer part was formed by elongated and closely spaced cells. In contrast, in the rufous-tailed hummingbird, the RA area mainly showed round cells close to each other, with lighter staining compared to the elongated and separate surrounding cells, forming a slight oval shape around them (Figures 3E,G.III; Supplementary Figure S3A).

The HVC area showed 0.5431 and 0.1451 mm³ volumes for the house wren and the rufous-tailed hummingbird, respectively (Figure 3F). In the house wren, the HVC area displayed different types of cells distributed variably (Supplementary Figure S3A). Sections close to the midline showed small and closely packed cells forming a thin band near the hippocampus, while in sections further away from the midline, cells were a more dispersed and with a more intense staining. The HVC area took on an oval shape with flattened ends and was surrounded by small, elongated cells close to each other. And in the rufous-tailed hummingbird, the HVC area mainly exhibited round cells close to each other. At the upper-dorsal part of the area, elongated cells with intense staining were found, extending towards the anterior nidopallium, making its morphology similar to that of the house wren (Figures 3F,G.IV).

We performed the 3D rendering of each hemisphere and each vocal area and calculated the percentage of occupancy of each nucleus relative to the total size of the hemisphere. We analyzed the relation between the volume of the vocal areas and the total volume of the hemisphere in the house wren and the rufous-tailed hummingbird, and the LMAN showed an occupancy percentage of 0.0655 and 0.0924; Area X of 0.4459 and 0.4605; RA of 0.1117 and 0.0986; and HVC of 0.3580 and 0.3311, respectively (Table 3).

In order to explain the volumetric differences between both species, we performed a cellular density analysis of the vocal areas. For this purpose, a fractionated scanning method was used in the area of interest, overlaying equidistant grids (as shown in Supplementary Figure S1B). Although the cellular density in RA area was similar for both species, we observed that the house wren exhibited a higher cellular density in Area X, while in the rufous-tailed hummingbird, this higher density was found in the LMAN and HVC areas (Table 3) (Figure 4A; Mann–Whitney test, * p value <0.05). This suggests a differential occupation and cellular density of LMAN, HVC and Area X between the house wren and rufous-tailed hummingbird.

The contours of the vocal areas were delineated through a meticulous analysis of their location and cellular and histological description. We used specific neuronal proteins, NeuN and MAP2, to validate the location of the vocal areas. Specifically, we used representative slices at the bregma where each vocal song area began and ended. The corresponding contours for LMAN, Area X, RA, and HVC were drawn in the specific location using both markers (Figure 4; Supplementary Figures S3B–E). The nuclear pattern and the location of all the vocal areas were found between positions 270–1920 μm, from the midline, for the house wren (Figures 4B,C; Supplementary Figures S3B,C) and 870–2040 μm for the rufous-tailed hummingbird (Figures 4D,E; Supplementary Figures S3D,E). This nuclei pattern from neuronal markers reproduces the 3D location of the vocal areas at the proposed bregma in the house wren and the rufous-tailed hummingbird.

Neuroanatomical studies of the vocal areas have focused on neuronal circuits (Gahr, 2000; Poirier et al., 2008; Vellema et al., 2011). However, astrocytes might participate in vocal areas (Kafitz et al., 1999; Haim and Rowitch, 2016). We labeled GFAP astrocytes with an antibody previously used in birds [Anti-GFAP (Rb) (Merk-millipore; ab5804)] (Polomova et al., 2019), and we also probed human and murine-tested anti-GFAP and anti-S100β antibody [anti-GFAP (Ms) (Sigma-Aldrich; G3893) and anti-S100β (Rb) (Dako-Agilet; Z0311)] (Rantamäki et al., 2013; Laddach et al., 2023). Consistently, we found that anti-GFAP tested for birds labeled effectively for both species, even for recognition in humans (Supplementary Figure S4A). Therefore, we examined the distribution of GFAP and S100β astrocytes in the vocal areas LMAN, Area X, RA, and HVC. Although these areas did not show classical astrocytic morphology, we found a GFAP and S100β punctate pattern in both species (Supplementary Figure S4B). Consistently, we found GFAP astrocytes in other brain regions, such as the telencephalon in the pallium located in the laminar edge of pallium (LEP) and vascular portions for house wren, as well as in the pallidum and mesencephalon, specifically, in the lateral prosencephalic fascicle (FPL) and the lateral mesencephalic reticular formation (FRL) for house wren (Figures 5, 6A,C). These regions were located at the same bregma point where the vocal areas were situated (Supplementary Figure S2), as was referenced in the known zebra finch atlas brain (Lovell et al., 2020). We did not observe an apparent label of S100β astrocytes in the same areas in the house wren (Figures 5A,B, 6A,B). In the rufous-tailed hummingbird, we found GFAP-positive astrocytes in the LEP region but not in the FPL and FRL, and consistently, we found a varicose pattern of s100β processes resembling astrocytes in the three regions (Figures 5C,B, 6C,D, and Supplementary Figure S5). We quantified the total GFAP in the pallium and mesencephalon and showed it on a heat scale for each species, finding that GFAP astrocytes were more abundant in the house wren than in the rufous-tailed hummingbird (Figure 5E).

The GFAP astrocytes of each area were segmented and analyzed for fractal parameters to determine their complexity. The house wren showed a typical morphology of GFAP astrocytes and cells with low levels of somatic S100β (Figures 5A,B). Comparing GFAP astrocytes within house wren, the GFAP astrocytes of the FPL and FRL showed a greater complexity, with more extended and more filamentous structures compared to astrocytes at LEP (Figures 6A,E, and Supplementary Figure S6A; ANOVA and Tukey’s test, * p < 0.05, ** p < 0.01, and *** p < 0.001). Interestingly, in the rufous-tailed hummingbird, S100β astrocytes exhibited a more straightforward punctate pattern in the LEP and a pattern of long filamentous in the FPL and FRL (Figure 5S). In addition, we found typical GFAP astrocytes in the LEP rufous-tailed hummingbird (Figure 5D). Therefore, we characterized the fractal morphology of GFAP astrocytes in the LEP of both species, resulting in a greater perimeter, width, and process length in the house wren compared to the rufous-tailed hummingbird (Figures 6A,C,F; Supplementary Figure S6B; ANOVA and Mann–Whitney test, * p < 0.05).

Subsequently, we performed a morphological analysis comparing GFAP astrocytes between the two species using immunofluorescence and confocal microscopy. This analysis would support observations by immunohistochemistry and delve into the three-dimensional detail of the astrocytes. Specifically, we identified astrocytes within the regions above (Figures 7A,B). Furthermore, the Sholl analysis, which shows the branching of astrocytic processes and GFAP astrocytes in the house wren, exhibited greater complexity across all examined areas, specifically in FPL and FRL, as shown above (Figures 7C,D; ANOVA and Tukey’s test, ** p < 0.01, and *** p < 0.001 7; 7E; Mann–Whitney test, * p < 0.05). Consistently, LEP GFAP astrocytes were compared; they showed more significant branching in the house wren compared to those in the rufous-tailed hummingbird. These findings indicate that astrocytes are considerably more complex in the house wren compared to the rufous-tailed hummingbird.