Subtype- and Layer-Specific Developmental Gradients of Postnatal GABAergic Neurons in the Rodent Barrel Cortex

Abstract

Introduction

The developing brain exhibits rostro-caudal gradients that align with the maturation of functionally organized circuits. The barrel cortex, a spatially precise sensory structure, serves as an ideal model to examine such gradients within a confined functional domain. GABAergic interneurons, characterized by subtype-specific developmental trajectories and pivotal roles in early cortical dynamics, provide a strategic cellular entry point for this investigation.

Methods

To explore cellular and functional gradients in the developing barrel cortex, we combined transgenic mouse lines, immunohistochemical analyses with in vivo electrophysiological recordings of whisker-evoked activity during early postnatal stages. We also employed a model of ethanol exposure to assess potential differences in apoptotic vulnerability along the rostro-caudal axis.

Results

Immunohistochemistry revealed distinct layer- and subtype-specific gradients of GABAergic neurons. Notably, we observed a widespread rostro-caudal gradient in 5-HT3AR-expressing cells and a localized gradient of somatostatin-positive (SST+) interneurons in the deep layers. These gradients diminished in a subtype-specific manner from postnatal day (P)5 to P10, indicating transient developmental features. In vivo recordings showed that caudal whisker stimulation elicited stronger responses, while rostral stimulation produced a broader spatial spread of activity, suggesting region-specific functional properties. Furthermore, rostral regions exhibited higher expression of the maturation marker KCC2, supporting the notion of more advanced maturation in the rostral barrel cortex. Ethanol exposure induced greater apoptosis in caudal layer 5 compared to its rostral counterpart, revealing layer- and region-specific vulnerabilities.

Conclusions

These findings highlight spatially regulated trajectories of cortical maturation and underscore how regional differences in development may influence sensory processing and contribute to the heterogeneity of symptoms observed in neurodevelopmental disorders.

Introduction

The development of the cerebral cortex, a hallmark of mammalian evolution, plays a fundamental role in complex cognitive abilities and sensory processing, making it a long-standing focus of interest for neuroscientists. Cortical development is orchestrated by both intrinsically encoded information [1] and extrinsic inputs from the thalamus and the external environment [2, 3]. Key developmental processes, such as area patterning and neuronal migration, are regulated by gradients of transcription factors and guidance molecules [4–6]. For instance, the rostro-caudal gradients of Emx2 and Pax6 transcription factors regulate cortical arealization, and mutations in the genes encoding these proteins can impair the development of motor, somatosensory, and visual cortices [5].

Within the somatosensory system, the lateral-to-medial axis dictates the sequential development of areas representing different body parts. Facial and mystacial representations emerge first, followed by areas for the forelimbs, trunk, and hind paws [7–9]. This hierarchy aligns with the survival needs of neonatal mammals as the mystacial and facial areas provide the most vital information for the pups’ survival during early life [10–12]. Similar developmental gradients are observed in humans: somatosensory evoked potentials exhibit age-specific patterns in neonates, with body-part-specific responses maturing progressively [13].

Not all gradients are transient, as shown by Fazzari et al. [14] who demonstrated persistent gradients in GABAergic neuron density across motor, visual, and somatosensory cortices of 1-month-old mice. Indeed, gradients can reflect not only ongoing development but also functional specializations that are critical for network composition and operation.

While the abovementioned studies compare cellular and functional gradients across cortical regions or somatosensory subfields, few studies have explored gradients within specific brain areas in the adult brain. In the whisker-to-barrel system, notable differences along the rostro-caudal axis include variations in whisker morphology [10, 15, 16], barrel size, and mean number of axons innervating specific vibrissal follicles [17]. Functionally, caudo-dorsal whiskers elicit stronger responses [18], whereas rostral whiskers exhibit higher sensitivity in texture discrimination [16]. These differences imply distinct roles of rostral and caudal whiskers in sensory processing.

The barrel cortex undergoes major cellular and structural alterations during the first postnatal week [19]. Given the significant ongoing changes, the presence of developmental gradients is expected. However, many studies of barrel cortex development neglect to consider or report these gradients, potentially overlooking critical developmental mechanisms that shape the functional maturation of the cortex. For instance, tactile experience has an impact on critical developmental processes, such as neuronal migration, apoptosis, and synaptogenesis [20]. Particularly, the role of activity in the development of GABAergic neurons – the cells mediating inhibitory transmission through γ-aminobutyric acid (GABA) secretion [21] – has been studied extensively during recent years. Indeed, GABAergic neurons play a pivotal role in shaping network activity maturation [22–24]. Moreover, several studies have linked impaired development of GABAergic neurons with neurodevelopmental disorders, such as autism spectrum disorder and schizophrenia [25–27]. However, the disease mechanisms remain largely unknown.

A gradient – a change in the value of a given variable per certain unit – can indicate ongoing development, serve as fundamentally important factor regulating developmental processes, or it may represent an underlying feature of differently operating neuronal networks. In this study, we aimed to elucidate possible cellular gradients along the rostro-caudal axis of the barrel cortex at postnatal day (P)5 and 10 by utilizing transgenic mouse lines and immunohistochemical approaches. Particular attention was given to GABAergic neurons, which undergo significant developmental processes between these ages. Functional differences between rostral and caudal regions of the barrel cortex were investigated with in vivo electrophysiological recordings in mice and rats. Finally, the possible role of rostro-caudal sensitivity for apoptosis was investigated.

Materials and Methods

Animals

Animal experiments were approved by the Animal Experimental Board of Finland (KEK23-009, ESAVI/3183/2022) and the French National Institute of Health and Medical Research (INSERM, provisional approval N007.08.01). All efforts were made to minimize the number and suffering of animals. The tests were carried out on P4–P10 pups. The date of birth was denoted as P0. Sex was not considered due to the presumed similarity of the target processes. Animals were housed under standard laboratory conditions (12:12 h light/dark cycle, ad libitum access to food and water).

Transgenic mouse lines were chosen to enable reliable identification and quantification of GABAergic neurons and their subtypes and targeting barrel cortex and specific barrels for in vivo electrophysiological recordings. The experiments including immunohistochemical analyses utilized Gad67-GFP knock-in [28], Tg(Htr3a-EGFP)DH30Gsat/Mmnc (RRID: MMRRC_000273-UNC), and wild type mice (B/6, ICR). Since the brain size of mouse pups limited the in vivo electrophysiological recordings in cases when several barrels were recorded from the same animal, we utilized Wistar rats (P5–P8) in these functional recordings. We presume that the observations of this study can be extrapolated inter-species since the explored questions regard fundamental developmental processes. Longitudinal in vivo extracellular recordings were conducted with transgenic mice with calcium indicator GCaMP6f expression under the Lhx6 promoter (Lhx6-Cre+/− [RRID: IMSR_JAX:026555], GCaMP6f+/− [RRID: IMSR_JAX:028865]) with C57BL/6J background.

Ethanol Exposure

To assess sensitivity within the barrel cortex, we used ethanol (EtOH) exposure treatment, a procedure largely based on the Britton and Miller study [29]. We allocated P4 mouse pups into 2 groups. The allocation was non-randomized but balanced by litter. The ethanol group received EtOH (3.2 mL/kg per body weight, 96% EtOH) and the control group was injected with comparable volumes of saline (0.9% NaCl). Ethanol was diluted to 25% in saline, to reach injection volumes of 50–60 µL. The injections were given intraperitoneally 2 times, 2 h apart. Injections were administered during the evening period (approximately 9–12 PM), though exact times were not recorded, and followed by transcardial perfusion of P5 pups, 8–12 h after the first injection. The weights of the pups varied between 3.76 g and 4.46 g.

Brains

The pups were anesthetized by hypothermia or 1:1 Euthasol vet (Dechra, injection dosage 150 µg/g) – lidocaine (Baxter, 10 mg/mL) mixture. The anesthesia was followed by transcardial perfusion with ice-cold phosphate-buffered saline (PBS) and fixation in ice cold 4% paraformaldehyde. Next, the brains were dissected, post-fixed in 4% paraformaldehyde at +4°C, and immersed in 30% sucrose. Once sunken, the brains were embedded in Tissue-Tek (O.C.T. Compound, Sakura, 4583). Leica CM3050 S Research Cryostat (cutting angle 5, chamber temperature −20°C, Accu-Edge carbon steel blades [C35, 4810]) was used in preparation of 50 µm free-floating coronal sections. In some experiments, during sectioning, an incision was made in the ventral part of the left hemisphere. This allowed subsequent separation of the hemispheres. Brain slices were collected sequentially across wells: the first slice placed in the first well, the second in the second well, and so on proceeding row by row through the plate. Once the plate was filled, the sequence restarted. This method ensured an even representation of slices from the rostro-caudal axis in each well. Finally, the sections were transferred to a cryoprotectant solution (0.3 g/mL sucrose, 0.01 g/mL polyvinyl-pyrrolidone [PVP-40], 0.1 M phosphate buffer [PB], and 30% ethylene glycol) and stored at −20°C.

Immunohistochemistry

We employed two immunostaining protocols: long immunofluorescence and methanol (MeOH) protocols. All steps were conducted at room temperature unless otherwise stated.

Long immunofluorescence protocol: this staining protocol has longer primary antibody incubation as well as an increased number of washing steps. These alterations improve the signal-to-background ratio. The brain sections were washed in PBS (1 × 5 min, 2 × 10 min) and quenched in 50 mm NH4Cl-tris-buffered saline solution (2 × 30 min). Next, the slices were incubated in a blocking solution (2% bovine serum albumin [BSA, Sigma A2153-50 G] – 3% Goat Serum – 0.4% Triton-X in PBS) for 3 h. After blocking, the primary antibody (in 1% BSA – 1% Goat Serum – 0.4% TX-100 in PBS) incubation was carried out for 60–72 h at +4°C followed by washing in PBS – 0.4% Triton-X (2 × 5 min, 4 × 10 min, 8 × 15 min). After washing steps, secondary antibody (in 1% BSA – 1% Goat Serum – 0.4% TX-100 in PBS) incubation took place for 2 h. The sections were then washed in PBS (2 × 5 min, 4 × 10 min, 4 × 15 min), stained with Hoechst 33342 nuclear stain (1:10,000, Invitrogen C10340) for 20 min, and briefly rinsed in 1x PB. Brain slices were mounted on microscope slides (Thermo Scientific Menzel-Gläser Superfrost Plus, J1800AMNZ) and coverslipped (Menzel-Gläser, 24 × 50 mm #1.5) using Fluoromount-G (SouthernBiotech, 0100-01).

MeOH protocol: this procedure is optimized for cleaved caspase-3 antibody and thus used when this antibody was included in the staining. The sections were washed in PBS (3 × 10 min) and dehydrated with increasing concentrations of MeOH (30%, 50%, and 80%, 30 min each). For antigen retrieval, sections were then treated with Dent’s fixative (80% MeOH, 20% dimethyl sulfoxide) for 1 h. The sections were washed for 30 min with TBSTD (tris-buffered saline with 0.1% Tween-20 [Sigma P1379-500 mL] and 5% dimethyl sulfoxide [Fisher Bioreagents BP-151-100]) and blocked for 3 h in TBSTD with 5% BSA and 0.4% serum. Next, primary antibody (in TBSTD with 5% BSA and 0.4% serum) incubation took place for 48 h at +4°C. The sections were then washed throughout the day with TBSTD followed by secondary antibody (in TBSTD with 5% BSA and 0.4% serum) incubation overnight at +4°C. Next, the sections were washed (3 × 20 min) with TBSTD. Slices were stained with the nuclear stain Hoechst 33342 (1:10,000, Invitrogen C10340) for 10 min, taken into PB, mounted on slides, and coverslipped using Fluoromount G.

Antibodies: primary antibodies were as follows: rabbit anti-cleaved caspase-3 (D175) (Cell Signaling, #9661S, 1:250), chicken anti-Gfp (Aves Labs, AB_2307313, 1:2,000), rabbit anti-Kcc2 (A) C-terminus [30] (1:2,000), rabbit anti-Kcc2 (Millipore, 07-432, 1:2,000), guinea pig anti-Vglut2 (Millipore, AB2251-I, 1:2,000), rat anti-Sst (Millipore, MAB354, 1:200), rabbit anti-Tbr2 (Abcam, ab23345, 1:1,500), rabbit anti-Cux1 (M-222) (Santa Cruz, sc-13024, 1:200), and rat anti-Ctip2 (25B6) (Abcam, ab18465, 1:2,000). The secondary antibodies were conjugated with Cy5 (1:300, Molecular Probes, A10523), Alexa Fluor^® 568 (1:400, Invitrogen, A10042), Alexa Fluor^® 568 (1:400, Invitrogen, A11075), Alexa Fluor^® 488 (1:400, Invitrogen A11039), Alexa Fluor^® 555 (1:400, Invitrogen, A21434), and Alexa Fluor^® 647 (1:500, JacksonImmuno, 712-605-153).

Microscopy

Image acquisition: images for quantifications of cell densities were captured with an Inverted Zeiss Axio Observer V1 microscope (Fluar 10x/0.50 M27 objective, AxioCam MR R3 camera) equipped with a motorized stage and ZEN 3.1 software. Representative images and images for Kcc2 intensity quantification were taken with the Zeiss LSM 980 confocal microscope (C-Apochromat 40x/1.20 W Korr objective, Plan-Apochromat 20x/0.8 M27 objective). The image capture settings, including exposure time, gain, and laser intensity, were consistent between the images compared. After acquisition, image tiles were stitched together using ZEN 3.1 software.

Rostro-Caudal Location

Allocation of the slices in the rostro-caudal axis was determined using the Allen Mouse Brain Atlas (adult mouse, coronal sections) [31] or the Developing Mouse Brain Atlas [32] (P6). Location in rostro-caudal axis was set to 0 mm for the slice that was the first one included in the analysis (the most rostral slice analyzed). For the quantification of average density, we separated the data in rostral and caudal compartments as follows: slices with location equal or less than 1.5 mm were considered rostral, whereas slices with location more than 1.5 mm were considered caudal. For KCC2 quantification, two slices from rostral and two slices from caudal areas from each animal were selected manually using a microscope without exact quantification of location in the rostro-caudal axis.

Cell Counting and KCC2 Intensity Quantification

Further image processing and quantifications were conducted in Fiji ImageJ software 1.54f. A TrackMate plugin [33] or manual cell counting was used in a case-dependent manner. Cell density was obtained by dividing the cell count by the area. KCC2 mean intensity was quantified by taking the channel’s mean gray value (confocal, raw images). For cell-specific KCC2 intensity quantification, we first identified the 5-HT3AR-GFP-labelled interneurons in ImageJ by thresholding the channel, creating ROIs and measuring the mean gray value of KCC2 within each ROI. Same threshold was used in all images.

In vivo Electrophysiological Recordings

Below is described the procedure of the recordings done in rats. View the online supplementary Material and Methods (for all online suppl. material, see https://doi.org/10.1159/000549093) for the procedure of surgery and recordings in mice (shown in online suppl. Material and Methods S1).

Surgery: the surgery was performed under isoflurane anesthesia (5% for the induction and 1.5% during the surgery). The skin over the skull was gently removed and the skull was cleaned. Head plate [34] was glued on top of the barrel cortex and additionally fixed using the dental cement. Next, the pup was fixed to the frame of the stereotaxic apparatus and warmed and left to recover from anesthesia for 1 h. Animals were surrounded by a cotton nest and heated via a thermal pad (35–37°C). A chloride-coated silver wire was placed in the cerebellum or visual cortex and served as the ground electrode. All recordings were made from rat pups anesthetized by ip injection of urethane (1 g/kg).

Whisker stimulation: before the experiment, whiskers were trimmed to a length of 3 mm. The principal whisker was stimulated by piezo deflection. The tip of the whisker was inserted 2 mm into a wire ring attached to the piezo bender (Noliac, USA) so that the whisker rested snugly inside. To induce deflection of the piezo bender, square 50–70 V pulses of 10 ms were applied. To avoid depression of the evoked response, whiskers were stimulated every 30 s.

Extracellular recordings: extracellular recordings of evoked cortical activity in the cortical whisker representations of interest were performed using a single shank 16-channel silicon probe (Neuronexus Technologies, USA) with a separation distance of 0.1 mm between electrodes. Two recording electrodes were placed into the somatosensory cortex. The corresponding whiskers projected to the recorded loci were determined based on several criteria: (i) presence of a short latency evoked local field potential (LFP) deflection and multiple-unit activity following stimulation of the corresponding whisker [35–37] and (ii) predominance of the gamma frequency in the evoked LFP and multiunit activity (MUA) [38]. The signals were amplified (×10,000), filtered (0.1 Hz–10 kHz) using a 128-channel amplifier (Neuralynx, USA) or 64-channel amplifier (DIPSI, France), and analyzed post hoc.

Data analysis: analysis was performed using custom-written functions in Python using the libraries: NumPy (version 1.23.1), Elephant (version 1.0.1b1), and SciPy (version 1.13.1). First, the raw data were analyzed to detect MUA, followed by downsampling the raw data to 1,000 Hz for further analysis of LFPs. MUA was detected in a band-pass-filtered signal between 400 and 4,000 Hz, in which all negative events exceeding 3.5 standard deviations (SDs) below the mean (calculated over the entire trace) were considered as spikes (>99.9% confidence) [37].

The onset of the evoked response was calculated as a time between the moment of stimulation and significant increase in spike rate. For that, the number of spikes was calculated in 5 ms bins with an overlap of 2.5 ms. The bin with amplitude exceeding 3.5 SD over the mean spike rate (calculated over the entire trace) was considered a bin showing a significant spike rate increase. Activity following stimulation with an onset delay of less than 50 ms was considered as evoked [38]. For characterization of spectral properties of the evoked response, the signal was high pass filtered (3 Hz) using a Butterworth filter, followed by spectral density estimation using multi-taper methods with 5 tapers over a 500 ms duration of the evoked response. The peak frequency and its amplitude in alpha/beta (8–28 Hz) and gamma (30–70 Hz) frequency bands were calculated from the averaged power spectral densities calculated for stimulated rostral and caudal whiskers in each animal. To characterize the propagation of the evoked response between the cortical representations of rostral and caudal whiskers, the cross-correlation of the evoked signals was calculated.

Study Design and Statistical Analysis

All statistical analyses were performed with RStudio software (RStudio 2023.06.0+421, “Mountain Hydrangea” Release). Before quantifications, all images were blinded with a custom-made R script, and the blinded images were organized alphabetically, resulting in a randomized order for data analysis. After cell counting, the data were unblinded, and the outcome assessment was conducted with an R script, which handles all the data similarly. The scripts used in analysis can be viewed in the online supplementary material (shown in online suppl. Material and Methods S2, and data analyzed are available in online supplementary data file (online suppl. Data S1). In this study, we performed three types of experiments, where study design and statistical analyses are described below.

Association between cell density and rostro-caudal location: in this exploratory approach, we collected several slices along the rostro-caudal axis from each mouse. We first aimed to collect about 10 observations per animal but increased the amount to 15–20 observations due to a greater variance than expected. In each figure legend, we report the exact value of sample size (n), number of mice, and number of litters used. Slices derived from the same mice were used in several experiments. For instance, brain slices derived from the Gad67-GFP mice were used for the quantification of both GAD67+ neurons and the density of cleaved caspase-3+ cells. Exclusion criteria included damaged anatomy of the brains or slices, or poor quality of histology disabling cell counting in the barrel cortex, such as slice folding or bubbles on top of the area of interest. Moreover, influential points indicated by the residual vs. leverage diagnostic plot were individually inspected to see whether they should be removed as outliers.

To assess the association between a dependent variable (cell density) and location in the rostro-caudal axis, we used linear mixed-effects regression analysis. In addition to the location, we included other potential explanatory factors, including litter and subject (mouse), and in some cases hemisphere, to the model. Location in the rostro-caudal axis and litter were included as fixed effects, hemisphere and mouse as random terms. Hemisphere was nested within mouse. The model is as follows:

$$lmer\left(density\sim location+litter+\left(1+location|subject/hemisphere\right)\right),data=mydata)$$

We constructed several models with different sets of explanatory variables and selected the best-fit model based on the Akaike information criterion (AIC). If the chosen model included only fixed effects, a linear regression analysis was used. If random terms were included in the selected model, we completed statistical analysis by comparing the selected model to the null model with anova. Additionally, we obtained 95% confidence intervals (CIs) for the estimated effect of location. We also report adjusted R² with linear regression and marginal R² with linear mixed-effects regression. Marginal R² represents variance explained by the fixed effect. Herein, a positive regression coefficient corresponds to a higher density associated with a more caudal location in the rostro-caudal axis. We consider a p value <0.05 as statistically significant. Residuals vs. fitted, residual Q-Q, and residual vs. leverage diagnostic plots were used to visually evaluate validity of linearity assumptions, heteroskedasticity, and to detect outliers and influential observations. Scatterplots along with a regression line are used for visualization of the associations. Moreover, we report descriptive statistics (mean together with SD) in the whole barrel cortex but also separately in the rostral and caudal sites.

Evoked activity and KCC2 intensity between rostral and caudal locations: in these experimental designs, we gathered data from rostral and barrel locations and employed paired analysis due to dependent observations recorded from the same brain. Evoked activity data were recorded from 6 animals. For KCC2 intensity quantifications, from each mouse (n = 5), we stained two rostral and two caudal slices. We performed microscopic imaging on both hemispheres, giving a total of four observations per location per mouse. From some mice, only two observations per location were collected due to damaged brain slices (exclusion criteria). The images were blinded (see above), and mean intensity per location per mouse was quantified. Those mean intensities were compared between rostral and caudal locations.

Due to low sample size, data distribution could not be reliably confirmed as normal, which is why we used a nonparametric paired Wilcoxon signed-rank exact test. The observations derived from the same animal (rostral and caudal) are dependent, and thus pairing was done based on the subject (pup). A p value of less than 0.05 was considered statistically significant.

In addition to p value and test statistics, we report effect sizes to estimate the magnitude of differences with 95% CI to indicate the precision of the effect size. Effect sizes were interpreted as follows: no effect or small effect from 0 to 0.3, medium effect from 0.3 to 0.5, and large effect if higher than 0.5 [39].

Sensitivity quantification by comparing apoptosis in ethanol and saline exposed mice: the same number of mice were allocated in ethanol- and saline-exposed groups. Each group had 3 pups from two different litters. The images were blinded for the quantification of cleaved caspase-3 expressing cells, which helped to reduce subjective bias – being particularly important since no randomization or blinding was done in the group allocation or injections administration.

To evaluate the effects of environmental manipulation, we collected slices from rostral and caudal parts of the barrel cortex and determined their location with the help of atlases. Slices from each mouse were ordered along the rostro-caudal axis, then paired by matching positions within animal in the rostral and caudal halves of the series (e.g., the first rostral slice with the first caudal slice, second with second, etc.). This approach ensured relatively uniform spacing between the paired slices.

From each experimental animal, we analyzed 5–9 rostral-caudal slice pairs (n = number of pairs). Differences between the pairs were calculated. Shapiro-Wilk’s normality test and Levene’s tests were performed to check normality and variance equality. We then compared the rostro-caudal difference of apoptosis between ethanol- and saline-injected mice: group means with a two-sided Welch two-sample t test, or medians with a Wilcoxon rank-sum test based on data normality. Moreover, we quantified effect sizes (Wilcoxon effect size for non-normally distributed data and Hedge’s g for normally distributed data) together with 95% CIs. This experiment sought to provide initial evidence regarding the sensitivity differences, and thus the number of animals collected was relatively low. Future studies should increase the sample size to obtain stronger support for sensitivity at a population level. The same statistical approach was used in analysis of longitudinal in vivo recordings.

Results

GABAergic Neuron Density Varies with Rostro-Caudal Location within the Barrel Cortex at P5

Cellular densities shift dramatically during early postnatal development. For instance, GABAergic neurons undergo vital processes – including migration, apoptosis, and network integration – during this period [40, 41]. To investigate GABAergic neuron distribution, we used Gad67-GFP transgenic mice (shown in Fig. 1c) to quantify their density along the rostro-caudal axis in the barrel cortex of P5 mice (shown in Fig. 1a). We aligned each brain slice along the rostro-caudal axis using anatomical landmarks identified with Hoechst staining. To ensure that our analysis was restricted to the barrel cortex, we labeled the target region using the well-established marker VGLUT2 and the less conventional indicator KCC2 (shown in Fig. 1b).

Fig. 1.

Quantification of the association between the density of GAD67+ neurons and the rostro-caudal location in the barrel cortex at P5. a Whisker-to-barrel cortex system with indication of rostro-caudal axis (a1) and illustrative image of coronal brain slices (a2). Hoechst nuclear stain was used to indicate brain structures, based on which the location in rostro-caudal axis was defined. Magenta dashed line indicates the incision, which enabled determining the hemisphere. Scale bar, 1,000 µm. b Immunostainings with CUX1, CTIP2, and TBR2 together with Hoechst (b1, b2) confirming accurate layer recognition with Hoechst nuclear stain. Scale bar, 50 µm. Representative images of VGLUT2 (magenta) and KCC2 (green) immunostainings (b3) as barrel markers. Scale bar, 50 µm. c Representative image of the GAD67-GFP (white) immunostaining. Hoechst (blue) is used to define the cortical layers. VGLUT2 (magenta) labels the barrels. Scale bar, 100 µm. d Association between the GAD67+ cell density and location in the rostro-caudal axis of the barrel cortex: scatter points show the original cell densities, and the black line represents the marginal effect. The points are color-coded by mouse and shape-coded by litter. The data were collected from 4 mice and 2 litters. From each mouse, there are 8–12 observations (slices). Linear regression models included GAD67+ cell density as dependent variable and location in the rostro-caudal axis as fixed effect (L1: F(1, 37) = 11.23, p = 0.00186; L2/3: F(1, 37) = 16.89, p = 0.000211; L4: F(1, 38) = 2.082, p = 0.1573). The selected mixed-effects models included GAD67+ cell density as dependent variable, location in the rostro-caudal axis as fixed effect, and mouse as random term (L5: χ²(1) = 5.6282, p = 0.01767; L6: χ²(1) = 3.0317, p = 0.08165). The 95% CIs of the estimated effect of location are presented in the figure.

We validated cortical layer identification with Hoechst co-staining against layer-specific markers: CUX1 (layer [L]2–4), CTIP2 (L5, low in L6), and TBR2 (L6) (shown in Fig. 1b). To examine how GABAergic cell density is associated with rostro-caudal position, we applied linear mixed-effects regression analysis, incorporating litter, subject (pup), and hemisphere, as additional explanatory factors. We constructed several models that included different sets of variables and selected the best-fitting model based on the lowest AIC.

At P5, the more caudal locations were associated with higher GABAergic neuron densities in L1, L2/3, and L5 but not in L4 or L6 (Table 1, shown in Fig. 1d). On average, the density in the caudal location was about 7% greater, although it varied between 2 and 12% depending on the layer. This gradient was observed in both barrel column and septal regions (shown in online suppl. Fig. S1). We report descriptive statistics, including total, rostral, and caudal means along with SDs in online supplementary Table S1 as a reference for future studies performing GABAergic neuron quantification in the barrel cortex.

Table 1.

Linear (mixed-effects) regression analyses of the cortical layer-specific association between GABAergic (GAD67+) cell density and explanatory variables in the mouse barrel cortex at P5 and P10

Age	Region analyzed	Model	Additional explanatory variables	Estimate (location)	95% CI (location)	Test statistics	p value	Adjusted/marginal R2
P5	L1	lm	​	128.77	50.92, 206.61	F(1, 37) = 11.23	0.00186**	0.2121
P5	L2/3	lm	​	131.95	66.90, 197.00	F(1, 37) = 16.89	0.0002106***	0.2949
P5	L4	lm	​	48.14	−19.41, 115.68	F(1, 38) = 2.082	0.1573	0.02698
P5	L5	lmer	Mouse	107.582	36.92, 178.24	χ2(1) = 5.6282	0.01767*	0.2000545
P5	L6	lmer	Mouse	58.747	−3.45, 120.95	χ2(1) = 3.0317	0.08165	0.1135695
P10	L1	lm	​	28.44	−9.45, 66.33	F(1, 97) = 2.219	0.1396	0.01229
P10	L2/3	lm	Litter	32.02	−0.89, 64.94	F(2, 99) = 4.75	2.467e−06***	0.214
P10	L4	lmer	Litter, mouse	10.903	−55.69, 77.50	χ2(1) = 0.1309	0.7175	0.1579525
P10	L5	lmer	Mouse	39.568	−12.18, 91.31	χ2(1) = 2.2389	0.1346	0.06078363
P10	L6A	lmer	Mouse, hemisphere	47.235	−40.11, 134.59	χ2(1) = 1.196	0.2741	0.04968001
P10	L6B	lm	​	6.492	−14.44, 27.43	F(1, 101) = 0.3784	0.5398	−0.006131
P5	Septal column, all layers	lm	​	131.82	42.68, 221.00	F(1, 39) = 8.946	0.0048**	0.1657
P5	Barrel column, all layers	lm	​	117.44	32.49, 202.39	F(1, 39) = 7.82	0.007982**	0.1457

Location in the rostro-caudal axis is included as a fixed effect in all models. The best-fit model is selected based on the lowest AIC. The variables included in the selected model, in addition to the rostro-caudal location and cell density, are shown in the table. The estimated coefficient (location in the rostro-caudal axis) and its 95% CI are presented along with the test statistics and p value. Statistically significant (p < 0.05) results are indicated with stars. Goodness-of-fit measures reflecting how much variance of the dependent variable can be explained by the fixed effect are also shown: adjusted correlation coefficient (R²) for linear regression (lm) and marginal R² for linear mixed-effects regression (lmer) analyses. Marginal R² reflects only the variance of the fixed effects excluding the random terms.

GABAergic Gradient Patterns Reflect Subtype-Specific Features at P5

Distinct temporal profiles characterize interneuron subtypes. Medial ganglionic eminence (MGE)-derived interneurons undergo neurogenesis [42, 43], migration [40], and apoptosis [41, 44] earlier than caudal ganglionic eminence (CGE)-derived interneurons. To further explore whether the GABAergic cell gradient is subtype specific, we separately quantified the MGE-derived somatostatin (SST)-expressing interneurons and CGE-derived 5-hydroxytryptamine receptor 3A (5-HT3AR)-expressing interneurons at P5 (shown in Fig. 2). These cell types were selected due to their distinct developmental origins and temporal profiles.

Fig. 2.

a, b Quantification of the association between the density of interneuron subtypes and the rostro-caudal location in the barrel cortex at P5. a1, b1 Representative images of the SST+ and 5-HT3AR-GFP+ (white) immunostainings. Hoechst (blue) is used to define the cortical layers. VGLUT2 (magenta) labels the barrels. Scale bar, 100 µm. a2, b2 Linear mixed-effects regression analysis. The scatter points show the original cell densities, and the black line visualizes the marginal effect. The points are color-coded by mouse and shape-coded by litter. Models with different terms were created and the best was chosen based on the lowest AIC. The 95% CIs of the estimated effect of location are presented in the figure. a2 Association between the SST+ cell density and location in the rostro-caudal axis of the barrel cortex. The data were collected from 5 mice and 2 litters. There are 7–10 observations from each mouse. Comparison of the AICs supported choosing the linear regression model where location in the rostro-caudal axis is the only explanatory variable. Linear regression analysis: L5: F(1, 41) = 5.114, p = 0.029; L6A: F(1, 41) = 0.03282, p = 0.8571; L6B: F(1, 41) = 0.06474, p = 0.8004. b2 Association between the 5-HT3AR+ cell density and location in the rostro-caudal axis of the barrel cortex. The data were collected from 5 mice and 2 litters. There are 6–18 observations from each mouse. Linear regression model was selected for the analysis of L4 (F(1, 48) = 9.597, p = 0.003253), L5 (including also litter as fixed effect, F(2, 47) = 45.1, p = 1.164e−11), L6A (F(1, 48) = 45.69, p = 1.719e−08), and L6B (F(1, 48) = 58.38, p = 7.726e−10). Linear mixed-effects models were selected for the analysis of L1 (χ²(1) = 0.0102, p = 0.9196) and L2/3 (χ²(1) = 10.174, p = 0.001424), both including location and litter as fixed effects and mouse as random term. L1, L2/3, and L5 contain 2 regression lines because there are two fixed effects included in the selected model.

SST+ interneurons are among the first functional inhibitory neuron populations in the cortex [24, 45]. For SST+ interneurons, we focused the analysis on infragranular layers (L5–6), where our immunodetection approach successfully spotted these cells (shown in Fig. 2a). A higher density of SST+ interneurons was associated with more caudal location in cortical L5 but not in L6 (Table 2, shown in Fig. 2a). For the quantification of 5-HT3AR+ interneurons, we utilized a transgenic mouse line with 5-HT3AR+ cells labelled with GFP. Rostro-caudal gradients in 5-HT3AR+ cell densities were shown in all cortical layers except in L1 (Table 2, shown in Fig. 2b). Similar to GAD67+ cell distribution, more caudal locations were associated with higher densities, with SST+ cells about 13% difference in L5 and for 5-HT3AR+ cells up to 30% difference in the deep layers of rostral and caudal averages (online suppl. Table S1). Together, these findings show that the GABAergic neuron density gradients in the barrel cortex at P5 are subtype and layer specific.

Table 2.

Linear (mixed-effects) regression analyses of the cortical layer-specific association of SST- and 5-HT3AR-expressing cell density with explanatory variables in the mouse barrel cortex at P5 and P10

Cell type	Age	Cortical layer	Model	Additional explanatory variables	Estimate (location)	95% CI (location)	Test statistics	p value	Adjusted/marginal R2
SST	P5	L5	lm	​	20.804	2.23, 9.38	F(1, 41) = 5.114	0.0291*	0.08922
SST	P5	L6A	lm	​	1.071	−10.86, 13.01	F(1, 41) = 0.03282	0.8571	−0.02357
SST	P5	L6B	lm	​	1.710	−11.86, 15.29	F(1, 41) = 0.06474	0.8004	−0.02278
5-HT3AR	P5	L1	lmer	Litter, mouse	−1.609	−92.14, 88.92	χ2(1) = 0.0102	0.9196	0.1098245
5-HT3AR	P5	L2/3	lmer	Litter, mouse	136.799	77.09, 196.51	χ2(1) = 10.174	0.001424**	0.370579
5-HT3AR	P5	L4	lm	​	43.81	15.38, 72.25	F(1, 48) = 9.597	0.003253**	0.1493
5-HT3AR	P5	L5	lm	Litter	103.32	81.01, 125.64	F(2, 47) = 45.1	1.164e−11***	0.6429
5-HT3AR	P5	L6A	lm	​	50.481	35.47, 65.50	F(1, 48) = 45.69	1.719e−08***	0.477
5-HT3AR	P5	L6B	lm	​	127.44	93.90, 160.97	F(1, 48) = 58.38	7.726e−10***	0.5394
SST	P10	L1	lmer	Mouse	−5.846	−11.24, −0.46	χ2(1) = 4.2017	0.04038 *	0.04019181
SST	P10	L2/3	lmer	Mouse	0.2581	−15.39, 15.90	χ2(1) = 0.0019	0.9649	0.0000403
SST	P10	L4	lm	​	−18.69	−28.24, −9.13	F(1, 86) = 15.13	0.000197***	0.1397
SST	P10	L5	lmer	Mouse	15.838	−5.18, 36.86	χ2(1) = 2.1853	0.1393	0.02419
SST	P10	L6	lmer	Mouse	4.406	−6.34, 15.16	χ2(1) = 0.7679	0.3809	0.007212942
5-HT3AR	P10	L1	lmer	Mouse	53.806	16.43, 91.19	χ2(1) = 5.6786	0.01717*	0.07294445
5-HT3AR	P10	L2/3	lmer	Mouse	81.397	62.93, 99.87	χ2(1) = 15.397	0.00008711***	0.3715596
5-HT3AR	P10	L4	lmer	Mouse	32.397	16.18, 48.61	χ2(1) = 7.942	0.00483**	0.2237684
5-HT3AR	P10	L5	lmer	Mouse	36.891	26.55, 47.23	χ2(1) = 13.901	0.0001927***	0.3759763
5-HT3AR	P10	L6	lm	​	26.538	15.67, 37.40	F(1, 86) = 23.57	0.000005329***	0.206

Location in the rostro-caudal axis is included as a fixed effect in all models.

The best-fit model is selected based on the lowest AIC. The variables included in the selected model, in addition to the rostro-caudal location and cell density, are shown in the table. The estimated coefficient (location in the rostro-caudal axis) and its 95% CI are presented along with the test statistics and p value. Statistically significant (p < 0.05) results are indicated with stars. Goodness-of-fit measures reflecting how much variance of the dependent variable can be explained by the fixed effect are also shown: adjusted correlation coefficient (R²) for linear regression (lm) and marginal R² for linear mixed-effects regression (lmer) analyses. Marginal R² reflects only the variance of the fixed effects excluding the random terms.

Temporal Refinement of Interneuron Gradients from P5 to P10

To determine whether these early developmental gradients are transient or persist with cortical maturation, we next examined the GABAergic cell distribution at P10 (Table 1, shown in Fig. 3). By this age, the GABAergic population has already undergone extensive apoptosis, although the period of cell death does continue until ∼P15 [41]. The overall model showed a strong relationship in L2/3, but further analysis revealed that litter (t(99) = 5.035, p = 2.15e−06), rather than rostro-caudal location (t(99) = 1.930, p = 0.0564), was the primary factor driving this effect. The rest of the layers showed no statistically significant associations. From P5 to P10, the average density of GAD67+ cells dropped 38–50% depending on the cortical layer (online suppl. Table S2), consistent with previous study showing about 40% reduction of cortical GAD67+ cells after developmental apoptosis [41].

Fig. 3.

Quantification of the association between the density of GAD67+ neurons and the rostro-caudal location in the barrel cortex at P10. a Representative image of the GAD67-GFP (white) immunostaining. Hoechst (blue) is used to define the cortical layers. Vglut2 (magenta) labels the barrels. Scale bar, 100 µm. b Association between the GAD67+ cell density and location in the rostro-caudal axis of the barrel cortex. Scatter points show the original cell densities, and the black line represents the marginal effect. The points are color-coded by mouse and shape-coded by litter. The data were collected from 4 mice and 2 litters. There are 20–30 observations from each mouse. Models with different terms were created and the best was chosen based on the lowest AIC. Linear regression model included GAD67+ cell density as dependent variable and location in the rostro-caudal axis as fixed effect (L1: F(1, 97) = 2.219, p = 0.1396; L6B: F(1, 101) = 0.3784, p = 0.5398). L2/3 analysis contained two fixed effects: location and litter (F(2, 99) = 14.75, p = 2.467e−06, see further description in the Results section). Mixed-effects models were selected for the analysis of L4, L5, and L6A, which included, in addition to the location, the following explanatory variables: mouse (random) and litter (fixed) in L4 (χ²(1) = 0.1309, p = 0.7175), mouse (random) in L5 (χ²(1) = 2.2389, p = 0.1346), and mouse (random) and hemisphere (random, nested within mouse) in L6A (χ²(1) = 1.196, p = 0.2741). L2/3 and L4 contain 2 regression lines because there are two fixed effects included in the selected model. The 95% CIs of the estimated effect of location are presented in the figure.

Inclusion of hemisphere did not significantly improve the model fit in any of the analyses except in L6A at P10 age-group (shown in Fig. 3b). In this case, there were four slices with very low cell density compared to the other slices. Further inspection revealed that these slices were from the same pup and same hemisphere. Since other mice did not show the same trend, it is likely that these low-density slices are of lower quality. Thus, our results show hemispheric symmetry in the developmental progression of GABAergic densities.

Although no significant rostro-caudal gradient was observed in GAD67+ cells at P10, we assessed whether this pattern persisted in interneuron subtypes (shown in Fig. 4a). SST+ interneurons’ quantification showed that the gradient observed in L5 at P5 was absent at P10 (Table 2, shown in Fig. 4b). Surprisingly, at P10, after the peak of developmental apoptosis of GABAergic neurons, SST+ interneuron average density was higher than 5 days earlier (P10: 377.1 cells/mm², P5: 284.6 cells/mm²) (online suppl. Tables S1, S2). These unexpected results may reflect low Sst expression at P5, rendering some cells undetectable by our immunodetection protocol. Indeed, fluctuations in Sst expression are previously shown in a developing hippocampus [46]. Additionally, at P10 our immunochemical approach was sufficient to detect SST+ neurons also in superficial layers. Quantification revealed lower cell densities associated with caudal locations in L1 and L4 (shown in Fig. 4b).

Fig. 4.

Quantification of the association between the density of interneuron subtypes and the rostro-caudal location in the barrel cortex at P10. a Representative images of the SST+ and 5-HT3AR-GFP+ (white) immunostainings. Hoechst (blue) is used to define the cortical layers. VGLUT2 (magenta) labels the barrels. Scale bar, 100 µm. b, c Linear (mixed-effects) regression analysis. The scatter points show the original cell densities, and the black line visualizes the marginal effect. The points are color-coded by mouse and shape-coded by litter. Models with different terms were created and the best was chosen based on the lowest AIC. The 95% CIs of the estimated effect of location are presented in the figure. b Association between the SST+ cell density and location in the rostro-caudal axis of the barrel cortex at P10. The data were collected from 5 mice from 1 litter. There are 15–20 observations from each mouse. In L4, the comparison of the AICs supported choosing the linear regression model where location in the rostro-caudal axis is the only explanatory variable (L4: F(1, 86) = 15.13, p = 0.000197). In other layers, the chosen model included location as fixed effect and mouse as random term (L1: χ²(1) = 4.2017, p = 0.04038; L2/3: χ²(1) = 0.0019, p = 0.9649; L5: χ²(1) = 2.1853, p = 0.1393; L6: χ²(1) = 0.7679, p = 0.3809). c Association between the 5-HT3AR+ cell density and location in the rostro-caudal axis of the barrel cortex at P10. The data were collected from 5 mice from 1 single litter. There are 15–20 observations from each mouse. Linear regression model was selected for the analysis of L6 (F(1, 86) = 23.57, p = 0.000005329). Linear mixed-effects models were selected for the analysis of L1 (χ²(1) = 5.6786, p = 0.01717), L2/3 (χ²(1) = 15.397, p = 0.00008711), L4 (χ²(1) = 7.942, p = 0.00483), and L5 (χ²(1) = 13.901, p = 0.0001927), which included location as fixed effect and mouse as random term.

Quantification of 5-HT3AR+ interneuron densities still exhibited location-dependent gradients in all cortical layers (Table 2, shown in Fig. 4c). Moreover, this subtype showed expected reductions in cell densities from P5 to P10. During those 5 days, the densities dropped almost 60% in layers 2–5, about 47% in L6, and 34% in L1. In superficial cortical layers (L1–4), the percentual drop seems greater in the rostral mean density, whereas in deep layers the reduction was larger in the caudal sites (online suppl. Table S2).

Interestingly, in all layers except in L1, the differences between rostral and caudal 5-HT3AR+ cell density averages were smaller at P10 than at P5, meaning that the gradient became less pronounced (online suppl. Table S3), possibly due to the developmental apoptosis. However, whether apoptosis completely removes the gradient of this subtype would require a later timepoint to be explored. A likely explanation for the absence of a gradient in the GAD67-GFP line, despite its presence in 5-HT3AR+ cells, is that 5-HT3AR expressing cells comprise only ∼30% of all GABAergic interneurons [43]. To sum up, gradients of GABAergic neurons along the rostro-caudal axis of the barrel cortex seem to either disappear or dilute from P5 to P10 in a subtype-specific manner.

Developmental Apoptosis Occurs Asynchronously within the Barrel Cortex at P5 but Does Not Explain the Gradient in GABAergic Cell Density

We asked if the observed location-dependent GABAergic cell density reflects ongoing developmental processes. Since there was still a clear gradient of 5-HT3AR interneurons at P10, it is unlikely that migration would even up these numbers. Thus, we hypothesized more advanced developmental apoptosis in the rostral barrel cortex. To examine gradients in GABAergic neurons’ cell death, we co-localized cleaved caspase-3 (CC3) with GAD67 (shown in Fig. 5a).

Fig. 5.

Quantification of the relationship between the density of cleaved caspase-3 (CC3)-expressing GAD67+ neurons and the rostro-caudal location in the barrel cortex at P5. a Representative image of the CC3+ cell (magenta) and GAD67+ cell (white) co-localization. Hoechst (blue) is used in cortical layer identification. The yellow arrowheads point to the co-localizing GAD67+ CC3-cells. Scale bar, 30 µm. b Quantification of the association between apoptotic (CC3+) GABAergic (GAD67+) cells and location in the rostro-caudal axis of the barrel cortex. The scatter points show the original cell densities, and the black line visualizes the marginal effect. The points are color-coded by mouse and shape-coded by litter. The data were collected from four mice and two litters. There are 10–12 observations from each mouse. Location is set to zero for the most rostral slice analyzed. Based on the lowest AIC, linear regression model with cell density as dependent variable and location in the rostro-caudal axis as fixed effect was applied on L5 data: F(39) = 0.0883, p = 0.7679. The 95% CIs of the estimated effect of location are presented in the figure. The large amount of zero values in the rest of the layers did not allow to apply regression analysis to these datasets. c Proportions of apoptotic (CC3+) GABAergic (GAD67+) and non-GABAergic cells show an increasing percentage of apoptotic GABAergic neurons from superficial to deeper cortical layers. d Same as in c, but including only L5 data. Quantification separated into rostral, middle, and caudal locations.

In L5, no sign of rostro-caudal gradient was detected for apoptotic GABAergic neurons at P5 (Table 3, shown in Fig. 5b). Since many samples exhibited no apoptotic GABAergic cells, we did not perform regression analysis beyond L5 but instead relied on qualitative inspection indicating no gradients (shown in Fig. 5b).

Table 3.

Linear regression analysis of the association between (GAD67+) cleaved caspase 3 (CC3)-positive cell density and location in the rostro-caudal axis in the mouse barrel cortex at P5

Cell type	Cortical layer	Model	Estimate (location)	95% CI (location)	Test statistics	p value	Adjusted R2
CC3	L1	lm	−2.928	−9.57, 3.71	F(1, 40) = 0.7939	0.3783	−0.005053
CC3	L2/3	lm	0.624	−3.42, 4.67	F(1, 40) = 0.09704	0.757	−0.02252
CC3	L4	N/A	–	–	–	–	–
CC3	L5	lm	7.878	2.79, 12.96	F(1, 39) = 9.827	0.003262**	0.1808
CC3	L6	N/A	–	–	–	–	–
GAD67+ CC3	L1	N/A	–	–	–	–	–
GAD67+ CC3	L2/3	N/A	–	–	–	–	–
GAD67+ CC3	L4	N/A	–	–	–	–	–
GAD67+ CC3	L5	lm	0.4412	−2.56, 3.44	F(1, 39) = 0.0883	0.7679	−0.02332
GAD67+ CC3	L6	N/A	–	–	–	–	–

The AIC advocated choosing simple linear regression; thus, no additional variables were included in the chosen models. The estimated coefficient (location in the rostro-caudal axis) and its 95% CI are presented along with the test statistics and p value. Statistically significant (p < 0.05) results are indicated with stars. Goodness-of-fit measure (adjusted R²) reflecting how much variance of the dependent variable can be explained by the fixed effect is also shown.

We also showed an increasing fraction of apoptotic GABAergic neurons among all apoptotic cells from superficial to deeper cortical layers (shown in Fig. 5c). Furthermore, because of the observed gradient of total apoptosis in L5 (Table 3, shown in online suppl. Fig. S2), we present the proportions of apoptotic GABAergic and apoptotic non-GABAergic cells in rostral, middle, and caudal parts in this layer. The percentage of apoptotic GABAergic neurons decreases from ∼70% in rostral regions to less than 40% in caudal areas (shown in Fig. 5d).

Together, these results demonstrate that apoptosis of GABAergic neurons constitutes a larger fraction of total apoptosis in rostral L5. Given that cell death alone could not explain the spatial differences in GABAergic neuron density, we reasoned that these gradients might instead reflect, or even drive, regional differences in early cortical circuit function.

KCC2 Expression Differs between Rostral and Caudal Barrel Cortex

While apoptosis did not seem to explain GABAergic cell gradients, we explored whether differences in network maturation might be associated with the observed spatial gradients. GABAergic neurons are known to fine-tune network functionality in mature and developing cortices [47–49]. During development, when the expression of potassium-chloride co-transporter KCC2 is low, GABA binding causes depolarization. Thus, KCC2 mediates the maturation of inhibitory transmission. In the somatosensory cortex, KCC2 expression remains low at embryonic stages but strongly increases within the first postnatal weeks [50].

To further assess the GABAergic neuron and inhibitory system gradients, we quantified KCC2 mean intensity in the barrel cortex using immunohistochemistry and confocal imaging. Previous studies have utilized immunohistochemical assessment in determination of KCC2 expression levels, including demonstrating its developmental increase [30, 51].

The average total mean gray intensity of KCC2 labeling was compared between rostral and caudal regions using a two-sided paired Wilcoxon signed-rank exact test (Table 4, shown in Fig. 6). Although the statistical significance level was not quite reached, effect sizes suggest higher KCC2 intensity in rostral locations, particularly in L2/3, L4 (septa), and in L6 (p = 0.0625, Wilcoxon effect size = 0.905). In all layers, the rostral medians (L1: 17.76, L2/3: 8.75, L4 septa: 10.52, L4 barrel: 13.89, L5: 12.2, L6A: 7.05, L6B: 7.12) were higher compared to the caudal medians (L1: 15.44, L2/3: 6.86, L4 septa: 6.74, L4 barrel: 11.67, L5: 7.24, L6A: 4.96, L6B: 4.87).

Table 4.

Total KCC2 intensity quantified from immunostainings

KCC2 intensity
​	difference between rostral and caudal barrel cortex					rostral		caudal
	test statistics (V)	p value	95% CI (Wilcox. test)	Wilcoxon effect size	95% CI (Wilcox_effsize)	median	range	median	range
L1	4	0.4375	−5.24, 4.71	0.422	0.06, 0.93	17.76	3.73	15.44	13.24
L2/3	0	0.0625	−3.15, −0.10	0.905	0.9, 0.95	8.75	2.44	6.86	5.5
L4, septa	0	0.0625	−4.55, −0.29	0.905	0.9, 0.95	10.52	3.67	6.74	6.07
L4, barrel	3	0.3125	−5.21, 2.01	0.543	0.06, 0.93	13.89	3.51	11.67	10.73
L5	1	0.125	−6.19, 0.05	0.784	0.18, 0.95	12.2	3.89	7.24	6.62
L6A	0	0.0625	−3.35, −0.05	0.905	0.9, 0.93	7.05	1.98	4.96	3.72
L6B	0	0.0625	−4.00, −1.51	0.905	0.9, 0.95	7.12	3.13	4.87	2.76

A paired Wilcoxon signed-rank exact test was used to compare the mean gray value of KCC2 intensity between rostral and caudal locations. n (pairs) = 5 mice from two different litters. Statistical reporting: test statistics along with p value, Wilcoxon 95% CI, Wilcoxon effect size, and its 95% CI, medians, and ranges separately for rostral and caudal locations.

Fig. 6.

Mean intensity of total KCC2 in the rostral and caudal ends of the barrel cortex at P5. a Representative images of immunostainings with VGLUT2 (magenta), KCC2 (white), and Hoechst (blue). Scale bar, 50 µm. b A paired Wilcoxon signed-rank exact test was used to compare the mean gray value of KCC2 intensities between rostral and caudal locations in a layer-specific manner. L1: V = 4, p = 0.4375, r = 0.422; L2/3: V = 0, p = 0.0625, r = 0.905; L4, barrel: V = 3, p = 0.3125, r = 0.543; L4, septa: V = 0, p = 0.0625, r = 0.905; L5: V = 1, p = 0.125, r = 0.784; L6A: V = 0, p = 0.0625, r = 0.905; L6B: V = 0, p = 0.0625, r = 0.905. The boxplots represent the median of the averaged (per location/animal) mean gray value of KCC2. n (pairs) = 5 mice from 2 different litters. Statistical reporting: paired Wilcoxon signed-rank exact test including test statistics and p value and Wilcoxon effect size (r).

Since 5-HT3AR+ interneurons exhibited the strongest cellular gradients, we further examined KCC2 intensity in this subtype (online suppl. Fig. S3). L5 was selected for the analysis due to three reasons: utilization of L4-VGLUT2 labelling helps with defining the area, appearance of 5-HT3AR+ interneurons in this layer but not as densely as in more superficial L2/3 (Fig. 2b). The averaged mean gray value of KCC2 inside 5-HT3AR+ interneurons compared between rostral and caudal sites. Effect size indicated a large effect between rostral and caudal locations, but statistical significance was not reached, possibly due to low sample size (V = 0, p = 0.0625, r = 0.905, median_rostral = 29.47, median_caudal = 19.17) (online suppl. Fig. S3b).

To conclude, KCC2 expression might be higher in the rostral barrel cortex. Interestingly, the expression in the caudal part seems lower, where more GABAergic neurons were observed. As KCC2 is a marker for the maturation of inhibitory transmission, these results suggest that rostral barrel cortex may be more mature than the caudal cortex, although further investigation is needed to conclude this.

Rostral and Caudal Whisker Stimulations Elicit Different Responses in Their Corresponding Barrels

We next asked if the observed gradients are associated with differing functionality between rostral and caudal parts of the area. Functional gradients along the rostro-caudal axis were explored with in vivo extracellular recordings in P5–P8 rat pups (Table 5, shown in Fig. 7). We used rats instead of mice due to the difficulty in reliably separating simultaneous rostral and caudal signals in the smaller mouse cortex.

Table 5.

Descriptive statistics of the barrel network activity recorded after whisker stimulation and statistical inferences of the evoked activity between rostral and caudal barrels

Whisker stimulation-evoked activity in the corresponding barrel
variable	difference between rostral and caudal barrels					rostral		caudal
	test statistics (V)	p value	95% CI (Wilcox. test)	Wilcoxon effect size	95% CI (Wilcox_effsize)	median	range	median	range
Onset latency	15	0.05906	–	0.769	0.5, 0.89	33 ms	11	36.5 ms	12
Peak power (alpha/beta freq.)	36	0.007813	8.48, 122.79	0.891	0.89, 0.9	26.5 µV2/Hz	109.29	58.09 µV2/Hz	205.25
Peak power (gamma freq.)	36	0.007813	3.54, 18.03	0.891	0.89, 0.9	4.12 µV2/Hz	14.56	16.33 µV2/Hz	38.76

Variable	​					Caudal → rostral		Rostral → caudal
	test statistics (V)	p value	95% CI (Wilcox. test)	Wilcoxon effect size	95% CI (Wilcox_effsize)	median	range	median	range
Cross-correlation	0	0.007813	−0.22, −0.097	0.891	0.89, 0.9	0.23	0.41	0.4	0.27

Recordings were done at P5–P8 rats. Rostral and caudal barrels recorded from the same mouse were paired. Analysis includes peak power in alpha/beta and gamma frequency ranges, onset latencies, and cross correlations. Paired Wilcoxon signed-rank exact test was used to compare the locations. In this table, we report test statistics, p value, Wilcoxon CI (if achievable), Wilcoxon effect size, and its 95% CI. Moreover, we separately show medians and ranges for rostral and caudal locations and median and range of cross-correlation. The data were collected from 6 mice, two of them providing 2 rostro-caudal barrel pairs recorded, giving 8 for the total sample size (n = 8).

Fig. 7.

In vivo electrophysiological recordings of whisker stimulation evoked responses in corresponding barrels (caudal, rostral) in P5–8 rat pups. a Representative examples of cortical activity evoked by A1 and A3 whisker stimulations and recorded activity in corresponding cortical barrels at P7. b Pooled data of synaptic delays of cortical responses evoked by rostral and caudal whisker stimulations. Comparison of the caudal and rostral onset latencies almost reached statistical significance (V = 15, p = 0.05906, r = 0.769). c Representative examples of power spectral densities of barrel cortex responses evoked by rostral and caudal whisker stimulations (left). Pooled results of peak power in alpha/beta (middle) and gamma (right) frequency ranges of the barrel cortex responses evoked by rostral and caudal whisker stimulations. Power was normalized to the values of caudal whiskers. Power ratio is presented separately for alpha-beta and gamma frequency ranges (C = caudal, R = rostral) (alpha-beta: V = 36, p = 0.007813, r = 0.891; gamma: V = 36, p = 0.007813, r = 0.891). d Representative examples of simultaneous barrel cortex recordings in A1 and A3 barrels following the A3 and A1 whiskers stimulations (left and right columns, respectively). Cross-correlograms for these experiments are shown below. Pooled data of correlation coefficients at lag 0 are shown on the right (cross-correlation: V = 0, p = 0.00781, r = 0.891). Statistical reporting: paired Wilcoxon signed-rank exact test including test statistics, p value, and Wilcoxon effect size (r). The data were collected from 6 mice, two of them providing 2 rostro-caudal barrel pair recordings, giving 8 for the total sample size (n).

Single whisker stimulation elicited responses in the corresponding cortical barrels of the contralateral hemisphere (shown in Fig. 7a). Onset latencies were similar for rostral and caudal responses (V = 15, p = 0.05906, n = 8, r = 0.769) (shown in Fig. 7b), with the median delay of 33.0 and 36.5 ms, respectively. Notably, the effect size indicates a large difference, and the statistical significance is close to the chosen significance level. Although not statistically significant, the shorter delay in rostral responses suggests faster signal propagation, potentially reflecting greater maturation [36].

Cortical responses exhibited two frequency peaks: ∼19 Hz and ∼39 Hz (shown in Fig. 7c). The low frequency alpha-beta range represents spindle bursts, and the high frequency characterizes gamma burst activity, both important in neuronal network synchronization [36, 52, 53]. Paired two-sided Wilcoxon signed-rank exact test for comparison of caudal and rostral responses suggest that caudal whisker (A1) stimulation evokes stronger spindle and gamma bursts in the corresponding barrel than the rostral (A3) whisker-barrel pair (V = 36, p = 0.007813, n = 8, r = 0.891) (shown in Fig. 7c).

To observe how evoked activity changes as maturation progresses, we performed in vivo extracellular recordings in mice at P5–6 and P8–10 (shown in online suppl. Fig. S4). Baseline-normalized power of evoked activity showed that responses were lower in both alpha-beta (W = 36, p = 0.002165, n = 6, r = 0.832) and gamma (W = 35, p = 0.004329, n = 6, r = 0.786) frequency ranges in the more mature cortex. These findings show that a decrease in evoked activity occurs as a function of maturation.

Additionally, simultaneous recordings in A1 (caudal) and A3 (rostral) barrels of rats revealed that stimulation of rostral whisker evokes evident responses in both caudal and rostral barrels. In contrast, caudal whisker stimulation activates mainly the corresponding caudal barrel (V = 0, p = 0.00781, n = 8, r = 0.891) (shown in Fig. 7d). Inter-columnar spread of evoked activity is previously associated with broader activity spread with cortical maturation [36, 54].

These results indicate that (1) sensory stimulation elicits stronger responses in caudal than rostral barrels, (2) the signal may be transmitted slightly faster in the rostral whisker-barrel pair than in the caudal path, and (3) the evoked responses spread into larger area from the rostral principal barrel. Together, these findings show that the rostral and caudal barrels possess differing functionality.

Rostral and Caudal Locations of the Barrel Cortex Show a Differing Sensitivity to Ethanol Exposure-Induced Apoptosis

Given these intrinsic developmental differences along the rostro-caudal axis, we next asked whether these regions differ in vulnerability to environmental stressors. We sought to examine this question by exposing pups to ethanol through intraperitoneal injections because fetal and early postnatal exposure to alcohol is shown to trigger rapid and widespread apoptosis in the brain [55]. The brains were collected for immunostaining 8 h after the first injection at P5.

Initially, we compared the mean density of apoptotic (CC3+) cells in the barrel cortex between saline-treated controls and ethanol-exposed mice (shown in online suppl. Fig. S5a). Layers 1–3 were analyzed together due to the high proportion of apoptotic cells located near the L1–2 border, where imprecise layer distinction could obscure results. All analyzed layers showed a substantial increase in the density of apoptotic cells following ethanol exposure (shown in online suppl. Fig. S5b), with L5 exhibiting the largest increase (L5: median_{caudal, EtOH} = 612 cells/mm², median_{rostral, EtOH} = 480 cells/mm²; median_{caudal, saline} = 40 cells/mm², median_{rostral, saline} = 22 cells/mm²). Notably, many apoptotic cells located in L5 extended CC3+ processes to the L1–2 border (shown in Fig. 8a; online suppl. Fig. S5a).

Fig. 8.

Caudal and rostral parts of the barrel cortex exhibit sensitivity differences for ethanol exposure. a Representative image of cleaved caspase-3 (white) staining in ethanol- and saline-treated samples. VGLUT2 (magenta) labels the barrels and Hoechst (blue) is used as a cortical layer marker. Scale bar, 50 µm. b Quantification workflow. Sensitivity differences of rostral and caudal areas were tested by comparing the changes in the densities of apoptotic cells (CC3+) along the rostro-caudal axis between saline- and ethanol-exposed groups. c The boxplots show the median of the Δ[CC3+ cells] for EtOH and saline groups. Two-samples Wilcoxon test or Welch t test were used in comparison of the group median/mean depending on the data distribution and variance homogeneity. Moreover, effect sizes (Wilcoxon effect size (r) for non-normally distributed data and Hedge’s g (g) for normally distributed data) along with 95% confidence intervals (CIs) are shown. L1–3: W = 273, p = 0.6396, r = 0.069, 95% CI: [0.003, 0.38]; L4: t(41.811) = 0.1263, p = 0.9001, g = 0.03, 95% CI: [−0.50, 0.57]; L5: W = 461, p = 0.0007275, r = 0.471, 95% CI: [0.17, 0.72]; L6A: t(32.893) = −0.52563, p = 0.6027, g = −0.14 m, 95% CI: [−0.67, 0.39]; L6B: W = 256.5, p = 0.5485, r = 0.088, 95% CI: [0.006, 0.38]. The data are derived from three mice from two different litters. From each mouse, there are 5–9 rostral-caudal slice pairs. n = number of rostral-caudal slice pairs; n_saline = 22, n_EtOH = 27.

We next examined the rostro-caudal differences in CC3+ cell densities between ethanol- and saline-treated groups (shown in Fig. 8). If regional vulnerability were uniform, ethanol would increase apoptosis equally across rostro-caudal locations. However, ethanol exposure caused a pronounced change in the rostro-caudal difference in L5 (p = 0.0007275, r = 0.471; density change: median_EtOH = 165 cells/mm², median_saline = 18 cells/mm²). The increase was markedly higher in the caudal area, indicating greater sensitivity to ethanol in this region. These findings demonstrate that early postnatal ethanol exposure significantly increases apoptotic cell death in the developing barrel cortex, with L5 of the caudal region displaying particularly heightened vulnerability – highlighting a spatially specific sensitivity along the rostro-caudal axis.

Discussion

Our study reveals transient cellular gradients in GABAergic cell distribution and functional differences across rostro-caudal axis of the neonatal barrel cortex. The cellular gradients are more pronounced at P5 and diminish by P10, suggesting a critical window of spatially organized inhibitory development.

Transient Gradients of GABAergic Neurons Exhibit Subtype- and Cortical Layer-Specific Patterns

The reducing gradients of SST- and 5-HT3AR interneuron densities indicate that the gradients represent ongoing developmental processes. 5-HT3AR+ interneurons still exhibit a clear gradient at P10, which suggests that additional mechanisms beyond unfinished migration likely contribute to these gradients. In support of this notion, previous research suggests that the migration of MGE- and CGE-derived interneurons is largely completed by approximately postnatal day 5 and postnatal day 7, respectively [40]. A second process crucial for regulating neuronal numbers during development is apoptosis [56]. Lack of association between cell death of GABAergic neurons and rostro-caudal location at P5 indicates that caspase-dependent apoptosis does not play a significant role in the formation of distributional gradient of GABAergic neurons.

These insights indicate temporally greater accumulation of GABAergic neurons in caudal areas, which might be associated with their migratory route from subcortical regions via several streams [57, 58]. In particular, CGE-derived interneurons are shown to arrive earlier in caudal cortical regions compared to rostral regions [43]. Their further dispersion to more rostral areas may rely on repulsive signaling from other interneurons or attractive guidance cues secreted by glutamatergic neurons [57, 58]. If caudal areas provide attractive signals, migrating interneurons may be more likely to stop there, resulting in temporary accumulation. Future studies employing live imaging could further clarify how migration dynamics contribute to these subtype-specific gradients.

Programmed cell death of GAD67+ neurons peaks between P7 to P11, with CGE-derived cells possibly peaking slightly later (around P9), than MGE-derived cells [41, 44]. Extrapolating from these insights, we hypothesize that developmental apoptosis contributes to reducing the gradients later on. Notably, while our study focuses on early postnatal development, it does not exclude the possibility that 5-HT3AR interneurons maintain a persistent, though milder, gradient beyond P10.

Our observations of interneurons expressing cleaved caspase-3 at P5 exhibited greater apoptosis in deep cortical layers (∼50%) compared to only 15–20% in upper layers. This may reflect the characteristics of interneuron subtypes: the MGE-derived population primarily localizes in deep layers (>75%) and enters apoptosis earlier than CGE-derived interneurons, which settle mainly in upper layers (almost 100% in L1, about 50% in L2/3) [41, 43, 44, 59, 60]. Taken together, the gradients seem to arise from the caudal accumulation of GABAergic neurons, possibly due to migratory characteristics, with a more even distribution achieved later through apoptosis.

Rostral Barrel Cortex Exhibits Signatures of More Mature Network Activity

Cellular allocation serves as a basis for assembling neuronal circuits. Cortical network dynamics undergo substantial changes during early postnatal development [61], regulated by GABAergic neurons [23, 24, 62, 63]. With functional recordings, we showed that whisker stimulation elicited more advanced network signatures in the rostral barrel in comparison to the caudal barrel. Our longitudinal recordings showed an age-related decline in evoked amplitude, reflecting that lower amplitude associates with more advanced network activity maturation. These findings parallel observations in postnatal mice [64] and in human preterm EEG studies, which indicate developmental modifications in network activity [65, 66].

Lower evoked amplitude in the rostral barrel may reflect greater network decorrelation – a hallmark of cortical maturation – where fewer neurons are activated simultaneously [67, 68]. Previously, decorrelation has been demonstrated in the mouse prefrontal cortex already at the end of first postnatal week [69], although admittedly the most critical transition occurs during the second postnatal week, shown in several cortical areas [24, 67, 69]. However, since most studies focus on the peak of the transition, it remains possible that sparsification begins earlier.

Higher expression of KCC2 in the rostral barrel cortex may contribute to differing functionality by facilitating chloride extrusion and enhancing hyperpolarizing GABAergic signaling [70]. We provided some indication that a strong cellular gradient of a specific interneuron population may associate with differing KCC2 levels. However, further investigation is needed to provide support for direct causal relationship between KCC2 and activity differences, including possible sparsification. Future studies may also investigate the contribution of NKCC1, the Na-K-2Cl cotransporter, in gradual cortical maturation. During normal development, NKCC1 has been suggested to reduce its neuronal expression. Supporting this, high levels of NKCC1 are shown in Fmr1 knockout mice modeling Fragile X syndrome, and inhibition of the protein restores some of the abnormal associated phenotype [71]. However, there are inconsistent data regarding NKCC1’s role in neuronal activity development, possibly due to cell type-specific splice variant expression, and potentially compensatory mechanisms provided by other proteins involved in ion transportation [72, 73]. Notably, NKCC1b variant expressed in neurons is shown to exhibit constant expression levels during early postnatal stages [73]. These studies suggest that KCC2, rather than NKCC1, is sufficient marker for neuronal maturation during postnatal stages. However, both proteins should be investigated systematically for high-resolution spatial expression during early postnatal stages.

There is an intriguing contrast between the hypothesis that the inhibition matures earlier in the rostral barrel cortex and evidence showing higher density of GABAergic neurons in the caudal regions. The mere presence of interneurons does not seem to sufficiently reflect the maturational state of inhibitory transmission. In line with this, previous research showed that the population size alone is not the main explanator behind functional inhibition [41]. Thus, GABAergic neuron allocation on cortical areas might be relatively independent of the maturation of network activity. However, future studies should exploit transgenic mouse lines to explore the cellular gradients of other interneuron subtypes, including SST- and parvalbumin-expressing subtypes, particularly due to their demonstrated roles in network pattern development [22–24].

Interestingly, in slices derived from the rat somatosensory cortex at the end of first postnatal week, GABA is shown to promote giant depolarizing potentials, important in synchronizing neuronal assemblies [74]. These insights raise the possibility that GABAergic signaling could support network synchronization in caudal areas while providing hyperpolarization in rostral regions. However, whether this assumption holds true in vivo remains uncertain, given the ongoing debate regarding the functional effects of early GABA_A receptor responses in vivo [75, 76].

We also showed that rostral whisker stimulation elicited responses not only in its principal barrel but also in a caudal barrel, whereas response to caudal whisker stimulation remained localized. This broader response spread aligns with enhanced columnar integration across whiskers as maturation progresses [36, 54] providing support for the idea that the rostral barrel cortex could be functionally more integrated than caudal barrels. Taken together, although our functional recordings show signs of more advanced maturation in the rostral barrel cortex, further investigation including more timepoints are needed for drawing conclusions.

Do Gradients Persist into Maturity?

It is worth noting that previous studies have indicated sustained gradients in a mature barrel cortex. However, despite the fact that whisker thickness and number of peripheral axons per vibrissal follicle follow rostro-caudal gradient [17], evoked response amplitude does not directly correlate with whisker diameter [18]. Furthermore, Hubatz et al. [18] demonstrated no gradient in evoked amplitude along the A row of the mature barrel map suggesting that the functional gradient observed in our study is specific to developmental stages. Additionally, the disappearance of SST+ cells’ and dilution of 5-HT3AR+ interneurons’ gradients by P10 support the view that these gradients are transient and developmentally regulated.

Sensitivity of the Barrel Cortex to Ethanol during Development

Given the distinct characteristics observed between rostral and caudal regions of the barrel cortex, we investigated whether these areas differ in their sensitivity to external perturbations [77]. We employed exposure to ethanol to rapidly induce widespread apoptosis in the developing brain [78, 79]. Depending on the timing of ethanol exposure, different brain regions and cell types are the most susceptible to alcohol-induced cell death [55, 80]. For instance, exposure of P7 mouse pups to alcohol strongly decreased the number of GABAergic cells, particularly parvalbumin- and calretinin-expressing subtypes [81].

The observed nine-fold increase of apoptotic cell density difference between rostral and caudal sites in L5 as a result of alcohol exposure suggests different sensitivity to ethanol-induced apoptosis between the locations rather than mere greater apoptosis due to higher cell densities. It is interesting to speculate that rostral barrel cortex is somehow, perhaps due to more advanced developmental stage, more protected against environmental disturbances, which may confer some neuroprotective benefits during early development.

Indeed, at early postnatal stages rodent pups rely on macrovibrissae to maintain close contact with littermates and the dam [10–12]. Specifically, rostral whiskers make contact with surfaces earlier than caudal whiskers. Snout contact involving rostral whiskers has been associated with huddling behaviors that facilitate access to the dam’s nipple for suckling [12] suggesting vital role of rostral vibrissae in supporting key early-life behaviors. In contrast, caudal whiskers, despite their larger diameter [17] and higher evoked amplitude, become more relevant only later as exploratory behaviors emerge during the second postnatal week [12]. Region-specific maturation trajectories may prioritize areas essential for immediate survival.

Our study does not exclude the possibility that the effects of ethanol would be due to gradients of development of the vascular system instead of sensitivity of the neurons. During embryonic stage of mice, brain periventricular vessels develop in a ventral-to-dorsal gradient regulated by specific transcription factors [82]. During postnatal stages, the vascular network still undergoes extensive alterations [83, 84]; however, whether it exhibits spatial gradients and enables delivery of ethanol more efficiently to the caudal sites, and thus contributes to impacts of ethanol observed in our study, remains to be investigated.

Timing of alcohol exposure determines the specific manifestations of alcohol spectrum disorders, which are also associated with symptoms of autism spectrum disorders [85]. Understanding the spatio-temporal profiles of the developmental gradients, and thus vulnerability of the transient circuits, could provide better insights into the effects of (short-term) environmental challenges. Thus, future research should pay close attention to the timing and duration of perturbations to better understand their developmental consequences and potential therapeutic interventions.

Limitations

We defined rostro-caudal location based on anatomical landmarks, which were easier to identify in slices containing the hippocampal formation, as these could be reliably positioned along the rostro-caudal axis. In contrast, more rostral slices lacked distinct anatomical markers, potentially introducing variability in assigning their exact rostro-caudal location.

In some experiments, where multiple linear regression modeling was utilized, the cell densities varied between pups considerably. In some of these cases, a low number of observations per animal increased uncertainty in assumption validation. To promote transparency, we present the original data points in graphical presentations.

Finally, even though both mice and rats are weaned around P21 before which they undergo the same developmental events, the exact timing of specific processes can differ. Due to fast development during the first postnatal week, slight temporal mismatch may become highly relevant. The species should not be used interchangeably without considering species-specific timescales [86]. In our experimental design, we gathered immunohistochemical data from P5 mice and performed functional recordings on P5–P8 rats. Since mice are known to develop slightly faster [87], these experimental designs could match the developmental stages of these two different species. However, technical developments enabling reliable separation of simultaneously recorded multiple barrels in mice would remove the need for extrapolating interpretations interspecies.

Technical Considerations

Our study emphasizes the ethical and methodological importance of considering both spatial and temporal aspects in developmental research. Specifically, the rostro-caudal location must be treated as a critical explanatory factor. To avoid false results and to enhance reproducibility, we recommend the following measures: (1) collect slices uniformly across the area of interest, (2) match regions between experimental groups, (3) leverage brain atlases or AI-based tools for precise localization of the region, and (4) apply specific markers for area identification. Furthermore, the experimental designs should include strategies to accurately identify the subtype of interest. Moreover, inclusion of litter as an explanatory factor appeared important in explaining cell density in several cases, highlighting the importance of including pups from several different litters.

Conclusions

Our study shows subtype- and cortical layer-specific developmental gradients of GABAergic neurons. Additionally, functional recordings showed that the rostral and caudal barrels had different characteristics in evoked responses, possibly deriving from differential maturational stage. The possible advanced maturation of the rostral barrel cortex may be linked to the early survival needs of pups, where sensory modalities essential for survival mature earlier to support critical behaviors. Whether the differences between rostral and caudal sites stem from activity-dependent mechanisms or are genetically programmed to prioritize rostral whisker function remains an open question. Future studies could dissect the role of activity-dependent versus intrinsic genetic programs in determining this asynchronous development.

From an evolutionary standpoint, the asynchronous maturation of functional units (columns) would offer an adaptive advantage by protecting against short-term perturbations. Such a feature could mitigate the behavioral and neurobiological consequences of acute damage. Furthermore, it is intriguing to speculate that the differences in developmental trajectories and sensitivity between cortical columns may underlie both regional vulnerabilities and interindividual differences in sensory processing.

Acknowledgments

The authors acknowledge the Neuroscience Center research unit within Helsinki Institute of Life Science (HiLIFE) and the In Vivo Brain Imaging – Unit of the Helsinki In Vivo Animal Imaging Platform (HAIP) for their facilities. The Biostatistics Unit’s consulting service of the University of Helsinki provided guidance in statistical analyses.

Statement of Ethics

Protocols were approved by the Animal Experimental Board of Finland (Approval No.: KEK23-009 and ESAVI/3183/2022) and the French National Institute of Health and Medical Research (INSERM, provisional approval No. N007.08.01).

Conflict of Interest Statement

The authors have no conflicts of interest to declare.

Funding Sources

This work was supported by Magnus Ehrnrooth Foundation to H.K., S.L., and A.L. (Ludwig), Research Council of Finland (341361, 308265) and ANR (GABGANG) to C.R., and PRIORITY-2030 program to M.M., A.L. (Logashkin), and V.S. The funders had no role in the design, data collection, data analysis, and reporting of this study.

Author Contributions

H.K. conceived the project, designed and performed all experiments except electrophysiological recordings, did statistical analysis, and interpreted the data, as well as prepared the figures. V.S., A.L. (Logashkin), and M.M. performed the in vivo electrophysiological recordings in rats, as well as analysis and visualization of these data. S.L. and A.L. (Ludwig) performed the in vivo electrophysiology in mice and analyzed and visualized these results. H.K. and C.R. wrote and edited most of the manuscript. S.L., A.L. (Ludwig), and M.M. wrote the methodology of electrophysiological recordings and edited the manuscript.

Funding Statement

This work was supported by Magnus Ehrnrooth Foundation to H.K., S.L., and A.L. (Ludwig), Research Council of Finland (341361, 308265) and ANR (GABGANG) to C.R., and PRIORITY-2030 program to M.M., A.L. (Logashkin), and V.S. The funders had no role in the design, data collection, data analysis, and reporting of this study.

Data Availability Statement

All data generated or analyzed during this study are included in this article and its online supplementary material files. Further inquiries can be directed to the corresponding author (contact corresponding author: Claudio Rivera, claudio.rivera@helsinki.fi).