Fourteen male ferret pups naturally delivered by six pregnant ferrets were purchased from Japan SLC (Hamamatsu, Japan). The pups were reared with lactating mothers (4–6 pups/mother) in stainless-steel cages (80 cm × 50 cm × 35 cm) maintained at 21.5 ± 2.5°C under 12-h artificial illumination in the Facility of Animal Breeding, Nakaizu Laboratory, Japan SLC. All lactating mothers were fed a pellet diet (High-Density Ferret Diet 5L14; PMI Feeds, Inc., St. Louis, MO, United States) and tap water ad libitum.

Seven ferrets were intraperitoneally injected VPA on PDs 6 and 7 at 200 μg/0.01 ml/g body weight. 5-Bromo-2-deoxyuridine (BrdU; Sigma-Aldrich, St. Louis, MO, United States) was injected at 30 μg/0.01 ml/g body weight simultaneously with the last injection of VPA. As controls, seven ferrets were administered BrdU on PD 7. Two hours after BrdU injection, three pups in each group were perfused with 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS) under deep anesthesia with ∼2% isoflurane gas. The remaining four ferrets in each group were reared until PD 20 and then perfused with PFA under deep anesthesia.

Three-dimensional (3D) magnetic resonance (MR) images were acquired from the fixed brains of PD 20 ferrets using a 7.0-T magnetic resonance imaging (MRI) system (magnet: 400 mm inner diameter bore; Kobelco and Jastec, Kobe, Japan; console: AVANCE-I; Bruker BioSpin, Ettlingen, Germany) according to the procedure reported previously (Sawada and Aoki, 2017).

The hippocampus was segmented semiautomatically on coronal (transaxial) MRIs based on image contrast as well as user knowledge of anatomy using the “Morpho” tool of the SliceOmatic software ver. 4.3 (TomoVision, Montreal, QC, Canada), according to the procedure reported previously (Sawada et al., 2013). The segmented images were used to calculate the hippocampal volume and to render the hippocampus in 3D using SliceOmatic software, as reported previously (Sawada et al., 2013).

Cerebral hemispheres were immersed in 30% sucrose-PBS solution overnight and then embedded in optimal cutting temperature compound (Sakura Finetek Japan Co., Ltd., Tokyo, Japan) at −70°C. Coronal cryosections of the hemispheres at 100 μm thickness were made from PD 7 and PD 20 ferrets with or without neonatal VPA exposure using a Retratome (REM-700; Yamato Kohki Industrial Co., Ltd., Saitama, Japan) equipped with a refrigeration unit (Electro Freeze MC-802A, Yamato Kohki Industrial).

Immunofluorescence staining was performed on floating sections with identical staining conditions using the same sets of the solution by the slight modified procedures in the previous report (Sawada, 2019). The primary antibodies used were highly specific for ferret tissues and the endogenous antigen (BrdU) (Sawada, 2019; Kamiya and Sawada, 2021) and included rabbit antibodies to calbindin D-28k (1:1,000; CB38; Swant, Bellinzona, Switzerland), NeuN (1:1,000; ABN78; Millipore, Billerica, CA, United States), and S100 protein (1:200; 942001; Immunostar, Hudson, WI, United States); mouse antibodies to glial fibrillary acidic protein (GFAP) (1:1,000, G3893, Sigma-Aldrich); a goat antibody to Sox2 (1:1,000; AF2018, R&D Systems Minneapolis, MN, United States); a rat antibody to phospho-histone H3 (PH3) (1:1,000; ab10543; Abcam, Cambridge, United Kingdom); and a sheep antibody to BrdU (1:500; ab1893, Abcam). Secondary antibodies included Alexa 488 donkey anti-rabbit IgG (1:500; A21206, Thermo Fisher Scientific, Waltham, MA, United States), Alexa 555 donkey anti-mouse IgG (1:500; A31570, Thermo Fisher Scientific), Alexa 555 donkey anti-goat IgG (1:500; A-21432, Thermo Fisher Scientific), Alexa 647 donkey anti-sheep IgG (1:500; ab150179, Abcam), and Alexa 555 goat anti-rat IgG (1:500; A21434, Thermo Fisher Scientific). All sections were stained using Hoechst, mounted on slides, and coverslipped with glycerin.

Serial digital sectioning images (10 sections at 1 μm plane thickness) were obtained under a 20 × objective using an Axio Imager M2 ApoTome.2 microscope equipped with an AxioCam MRm camera (Zeiss, Gottingen, Germany) with Zen 2.3 blue edition software (Zeiss). Densities of immuno- and BrdU-labeled cells were calculated by the disector method using systematic random sampling, according to a previous report (Sawada, 2019). To estimate cell density, three consecutive sections at the coronal plane, including the caudal end of the splenium of the corpus callosum, were analyzed per animal. Frames with six square boxes (box size = 40 μm × 40 μm) were used to select systematically the region of interest (ROI) superimposed randomly on subregions of dorsal and/or ventral hippocampi. The percentage of immuno- or BrdU-labeled cells was estimated by summing the cells counted within all ROIs from six cerebral hemispheres of VPA-treated and control groups on PD 7, and eight hemispheres of both two groups on PD 20.

Data from the left and right hippocampi were considered to be independent samples, since no significant the left- and right-side differences in any measurements were demonstrated by a paired sample t-test. One-way analysis of variance (ANOVA) followed by Student’s t-tests was used to assess statistically the volume of the hippocampus between VPA-treated and control groups. A repeated-measures two-way ANOVA was used to assess the hippocampal region (dorsal or ventral)-related differences in immuno- and BrdU-labeled cell density. Scheffe’s test was performed as post hoc testing when significant interactions were revealed by two-way repeated-measures ANOVA, followed by simple main effects at P < 0.05.

The Chi-square test was conducted to compare the proportions of immunolabeled cells positive for each marker to BrdU-labeled cells between two groups of ferrets. The total number of BrdU-labeled cells counted was defined as “n” for the Chi-square test.

The volume of the hippocampus was estimated based on MR images. The hippocampal volume was 27.3 ± 0.6 mm³ in VPA-treated ferrets, not significantly different from that in control ferrets (27.4 ± 0.6 mm³) (Table 1). Three-dimensional rendered images of the hippocampus with the cerebral cortex in the left hemisphere are depicted in Figure 1A. The hippocampus was situated in the deep portion of the posterior ectosylvian gyrus, corresponding to the auditory cortex. Gross structures of the hippocampus were not different between the VPA-treated and control ferrets. Coronal MR images of the hippocampus of VPA-treated ferrets and controls are shown in two identical planes, that is, the plane including the caudal end of the splenium of the corpus callosum (Figure 1B) and in the plane including the posterior commissure (Figure 1C). In coronal MR images at the caudal end of the splenium of the corpus callosum, the dorsal hippocampus began to descend in both groups (Figure 1B). A connection from the dorsal to ventral hippocampus was found in coronal MR images at the posterior commissure in VPA-treated and control ferrets (Figure 1C).

The dorsal hippocampus in the coronal plane including the caudal end of the splenium of the corpus callosum was depicted by calbindin D-28k immunofluorescence with BrdU labeling and Hoechst staining in both VPA-treated and control ferrets (Figure 2A). In the dentate gyrus (DG), the subgranular layer was distinguishable by an array of calbindin D-28k-negative Hoechst-stained cells beneath a layer consisting of calbindin D-28k-positive granular neurons. To assess the effect of neonatal VPA exposure, the densities of cells defined by marker antigens for neuronal progenitors, neurons, and glial cells were estimated. There were no differences in the densities of calbindin D-28k-positive granular neurons in the DG granular layer and calbindin D-28k-negative Hoechst-stained cells in the DG subgranular layer between VPA-treated and control ferrets (Supplementary Figures 1A,B).

5-Bromo-2-deoxyuridine -labeled cells were found throughout the hippocampus of PD 20 ferrets and were particularly abundant in the DG (Figure 2A). A significant effect on treatments [F_(1,14) = 24.756, P < 0.001], but not hippocampal regions (dorsal/ventral), was observed in the density of BrdU-labeled cells in the DG subgranular layer by repeated-measures two-way ANOVA. Post hoc testing indicated significantly denser BrdU-labeled cells in VPA-treated ferrets compared to controls in both the dorsal and ventral hippocampi (Figure 3A). Significantly denser BrdU-labeled cells in VPA-treated ferrets compared to controls were also detected in the DG granular layer of the dorsal hippocampus (P < 0.001; Figure 3A), following a significant effect on treatment [F_(1,14) = 8.612, P < 0.05] by repeated-measures two-way ANOVA. This meant a slight disarrangement of the DG granular layer of the dorsal hippocampus by mismigration of the tracked neonatal DG progenitors, although this disarrangement was difficult to define optically. In the CA1 and CA3 fields, there was no difference in the BrdU-labeled cell density between VPA-treated and controls (Supplementary Figures 2A,C).

Sox2-positive progenitors were distributed widely throughout the hippocampus in both VPA-treated ferrets and controls on PD 20. Multiple immunofluorescence staining revealed the presence of BrdU labeling in Sox2-positive progenitors (Figure 2B). More than 70% of BrdU labeling appeared in Sox2-positive progenitors in the DG (Table 2) and CA1 and CA3 fields (Supplementary Tables 1, 2) in both groups. The density of BrdU-labeled Sox2-progenitors was estimated in the subregions of the dorsal and ventral hippocampi. Repeated-measures two-way ANOVA revealed a significant effect of treatment [F_(1,14) = 99.127, P < 0.001] and hippocampal regions (dorsal/ventral) [F_(1,14) = 13.057, P < 0.005], and these two interactions [F_(1,14) = 6.018, P < 0.05] in the DG subgranular layers. Post hoc testing indicated significantly denser BrdU-labeled Sox2-progenitors in VPA-treated ferrets than in controls in the DG subgranular layer of both the dorsal and ventral hippocampi (Figure 3B). In the DG granular layer, significantly denser BrdU-labeled Sox-2-positive progenitors in VPA-treated ferrets compared to controls were detected in the dorsal hippocampus (P < 0.001; Figure 3B), following a significant effect on treatment [F_(1,14) = 7.118, P < 0.05] by repeated-measures two-way ANOVA. In other hippocampal regions examined, including the CA1 and CA3 fields, there was no difference in the BrdU-labeled Sox-2-positive progenitor density between VPA-treated and control ferrets (Supplementary Figures 2B,D).

Multiple immunofluorescence staining was conducted using neuronal and glial markers, that is, granular neurons in the DG and pyramidal neurons in the CA fields (calbindin D-28k-positive), neurons (NeuN-positive), and glial cells (S100-positive or GFAP-positive). NeuN-positive neurons were distributed sparsely in the hilus and molecular layer of the DG (Figure 2C), the strata radiatum and oriens of the CA1 and the CA3, and also aligned into the DG granular layer and the CA pyramidal layer. Calbindin D-28k immunostaining also appeared predominantly in DG granular neurons and CA pyramidal neurons (Figures 2A,B). While 5.9% of BrdU-labeled cells were unexpectedly NeuN-positive in the CA3 stratum radiatum, almost all NeuN-positive neurons and calbindin D-28k granular/pyramidal neurons in the hippocampus were not labeled with BrdU in both VPA-treated and control ferrets (Figures 2B,C, Table 2 and Supplementary Tables 1, 2).

While S100-positive cells were distributed throughout the hippocampus, their alignment at the bottom of the DG subgranular layer was striking (Figure 2D). S100 immunostaining was found in BrdU-labeled cells in the range of 0–15.3% in VPA-treated ferrets and from 0 to 13.3% in controls (Table 2). There was no difference in the incidence of S100 immunostaining in BrdU-labeled cells in any DG subregion between VPA-treated and controls. In contrast, no BrdU-positive cells expressed S100 immunostaining in the CA1 and CA3 fields in either group (Supplementary Tables 1, 2).

GFAP immunostaining mainly appeared in astrocytes sparsely distributed throughout the hippocampus (yellow closed arrowhead in Figure 2D) and was not observed in radial glia-like cells aligned in the DG granular through subgranular layers (Figure 2D), as seen in the adult ferret hippocampus (Supplementary Figures 3A,B). There were no BrdU-labeled GFAP-positive astrocytes in the hippocampal subregions, including the DG, except the stratum oriens in the CA1 and CA3 fields. GFAP-positive astrocytes were found in BrdU-labeled cells of the CA1 stratum oriens at 5.9% in the ventral hippocampus of VPA-treated ferrets and 12.5% in the dorsal hippocampus of control ferrets (Supplementary Table 1). In the CA3 stratum oriens, 21% of BrdU-labeled cells were GFAP-positive astrocytes in the dorsal hippocampus of VPA-treated ferrets (Supplementary Table 2). In any case, there was no difference in the incidence of GFAP-positive astrocytes in BrdU-labeled cells in all hippocampal subregions examined between VPA-treated and control ferrets.

Notably, immunostaining for PH3, a specific marker for mitosis, was found in a very small population of BrdU-labeled cells in the hippocampus (Figure 2C). Only 0.2 and 0.5% of BrdU-labeled cells distributed throughout the hippocampus were PH3-immunopositive in VPA-treated and control ferrets, respectively. The co-presence of PH3 immunostaining with BrdU labeling indicates the possibility that Sox2-positive progenitors distributed throughout the hippocampus may have mitotic potency.

Immunofluorescence staining for Sox2 with BrdU labeling was performed in the premature hippocampus of VPA-treated ferrets and controls on PD 7, 2 h following BrdU injection simultaneously with the last VPA injection. BrdU-labeled Sox2-positive progenitors were abundant, particularly in the granular and subgranular layers and hilus of the DG in the premature hippocampus (Figure 4). As it was difficult to distinguish the border between granular and subgranular layers in the premature DG, the density of BrdU-labeled Sox2-positive progenitors was estimated in areas including either the granular layer or subgranular layer or the hilus. BrdU-labeled cells were significantly denser in both the granular/subgranular layers and hilus in VPA-treated ferrets than in controls (Figure 5). Sox2-positive progenitors were found in BrdU-labeled cells at 97.5% and 88.2% in the granular/subgranular layers and 94.6% and 91.7% in the hilus of VPA-treated and control ferrets, respectively (Table 3). The density of BrdU-labeled Sox2-positive progenitors was significantly greater in both the granular/subgranular layers and hilus of VPA-treated ferrets than those of controls (Figure 5). Thus, VPA exposure facilitated the proliferation of Sox2-positive progenitor pools in the DG of the premature hippocampus.