All animals were anesthetized in a sealed chamber containing 4% isoflurane and subsequently placed on an autoregulated heat pad within a stereotactic frame. They were then maintained under 2% isoflurane via a stereotactic nose mask (Kopf Instruments, no: 906). Analgesia was provided with buprenorphine (0.03 mg/kg) and 1.5 mL of Ringer’s acetate, administered subcutaneously, along with a 50/50 mixture of Lidocaine and Marcaine (0.250 mg/mL) applied to the scalp. A midline incision was made through the skin and periosteum, followed by a 2.75 mm burr hole drilled 2 mm posterior and 2 mm lateral to bregma, exposing the dura. The secondary projectile was positioned adjacent to the dura. The secondary projectile consisted of a 30 mm long aluminum probe with a spherical tip (diameter 2 mm), weighing 0.66 g. A modified air rifle (CNC-Process AB, Hova, Sweden) was used to fire a lead pellet, which impacted the secondary projectile that accelerated to a speed of 110 m/s, penetrating the meninges and brain parenchyma. The spherical tip causes a direct tissue laceration but also produces a surrounding injury zone due to transient cavity expansion, propagating a pressure wave into adjacent tissue and resulting in damage that extends beyond the immediate laceration site (33). A cone-shaped section of the secondary probe served as a stopper, limiting the mean penetration depth to 5.075 mm (n = 4, SD = 0.05 mm) and ensuring survivability. In the sham surgery group, a blank bullet was used, thus no penetration occurred.

Lastly, the skin was sutured, and animals were placed in a recovery box on an autoregulated heat pad and supplied 100% O₂ until mobile. Buprenorphine (0.03 mg/kg) was administered for analgesia to all animals, regardless of intervention, three times daily for three days post-surgery. Previous studies describing the penetration model in further depth are available (33–35).

Rats were sacrificed by guillotine decapitation under isoflurane anesthesia 72 h after pTBI or sham induction. Brains were harvested and flash-frozen on dry ice, then embedded in OCT and stored at −80 °C. No tissue perfusion was performed, in order to minimize the interval between animal sacrifice and tissue freezing to preserve RNA quality. 14 μm sections were acquired at Bregma: – 3.60 mm (Paxinos 6th edition) using a CryoStar NX70 cryostat treated with RNaseZap (Invitrogen, Cat no: AM9780) to eliminate RNase contamination. Sections intended for LCMD and RNA-sequencing were placed on MMI membrane slides (MMI, Cat no: 50102), and directly adjacent sections used for multiplex immunofluorescence on SuperFrost Plus slides (Thermo Fisher Scientific).

All ethanol solutions were diluted from 100% using RNase-free water (Invitrogen, cat no: 10977035) in a RNase-free environment. Sections intended for LCMD and subsequent RNA sequencing were fixed for 5 min in −20 °C 70% ethanol and then rinsed twice for 10 s in RNase-free PBS (Invitrogen, cat no: AM9624) before nuclear staining with a modified cresyl violet solution (1% w/w in 75% ethanol) for 20 s. Afterward, sections were dehydrated in 70% ethanol and 100% ethanol for 30 s and 1 min, respectively. Following dehydration, sections were incubated at room temperature for 15 min. Sections were prepared one at a time and dissected immediately after the 15-min incubation period.

LCMD was conducted using a Leica DM6000B microscope equipped with a Leica LMD7000 laser-capture system. Hippocampal subregions were dissected at predefined locations described in Figure 1. Dissections were performed at 20x magnification, carefully delineating the pyramidal layer for CA1-CA3 and the granular layer for the DG to minimize glial contamination. Approximately 100 neurons were collected from each sample and subregion. The dissected tissue was collected into 0.2 mL adhesive caps (Zeiss, Cat no: 415190–9,181-000), and 8 μL of nuclease-free water, along with 1 μL of 10x Reaction Buffer (Takara Bio, Cat no: 635013), was added for initial lysis. Following a 1-min centrifugation, the samples were placed on dry ice and subsequently stored at −80 °C. In order to minimize potential batch effects, tissue sections were thawed, stained, and microdissected individually one at a time using a standardized workflow to ensure comparable handling times from thawing to lysis and freezing, with samples processed in randomized order. Figure 2 presents a schematic overview of the sample processing pipeline and images illustrating the specificity of the LCMD procedure.

SMART-Seq libraries were synthesized from total RNA using SMART-Seq v4 Ultra Low Input RNA Kit (Takara Bio) according to the manufacturer’s instructions. The first-strand cDNAs were synthesized from the total RNA using nontemplated nucleotides added by the SMARTScribe Reverse Transcriptase, then, the double-strand cDNAs were synthesized by template switching. PCR was directly performed and samples purified. Quality control of cDNAs was performed on Agilent Tapestation according to the manufacturer’s instructions. 1 ng of cDNA was used for the tagmentation reaction. During tagmentation, samples are fragmented and tagged with adapters, followed by PCR amplification and index addition. The indexed libraries were purified, normalized, combined, and sequenced on the Illumina NovaSeq X Plus 10B-300 lane generating 2×150 bp paired-end reads. Basecalling and demultiplexing was performed using bcl2fastq software with default settings generating Fastq files for further downstream mapping and analysis. Library preparation was performed in a blinded and randomized manner, and all libraries were sequenced together in a single run, minimizing batch-related effects on subfield-level clustering.

Sections used for immunofluorescent staining were directly adjacent to the sections used for LCMD and subsequent RNA-sequencing. Sections were thawed for 20 min at room temperature, fixed in 4% paraformaldehyde for 10 min at 4 °C, and washed in PBS for 15 min. Slides were incubated with blocking solution (0.01 M PBS + 0.1% NaN3 + 0.3% Triton + 5% BSA) for 45 min at room temperature, followed by incubation with primary antibodies toward microglia (rabbit anti-Iba1, FUJIFILM, cat no: 019–19,741, 1:500), neurons (chicken anti-NeuN, Sigma, cat no: ABN91, 1:1000), astrocytes (rabbit anti-GFAP, Abcam, cat no: ab33922, 1:1000), oligodendrocytes (goat anti-OLIG2, R&D Systems, AF2418, 1:500), leukocytes (mouse anti-CD45, Abcam, cat no: ab33923, 1:50), oligodendrocyte precursors (mouse anti-NG2, Millipore, cat no: MAB5384-I, 1:50), epithelial tight junctions/vessels (rabbit anti-Occludin, Abcam, cat no: ab216327, 1:200), in blocking solution for 16 h at 4 °C. The next day, slides were washed in PBS for 30 min and incubated with secondary antibodies (Alexa Fluor, all cross-adsorbed, donkey anti-mouse 555 & 647, anti-goat 488, anti-rabbit 555 & 647, all 1:500) in a buffer (0.01 M PBS + 0.1% NaN3 + 0.3% Triton) for 45 min at room temperature. Slides were washed 30 min in PBS, incubated with DAPI for 10 min, washed again in PBS for 30 min, and mounted with Prolong Diamond antifade mounting medium (Invitrogen).

Images for immunofluorescent analysis were acquired using a Zeiss LSM800 confocal microscope at 20x magnification, with four images captured per subregion and subsequently arranged as 2 × 2 tiles. Images were processed in Fiji/ImageJ, with automated contrast adjustment followed by channel merging and background subtraction as no intensity quantification was performed. Images were analyzed qualitatively, assessing infiltration and presence of non-neuronal integration of the pyramidal and granule cell layer.

RNA-seq data was processed using a standard analysis pipeline. Basecalling and demultiplexing was performed with bcl2fastq (v2.20.04.422). Reads were aligned to the Ensembl Rattus norvegicus reference genome (mRatBN7.2) using STAR (v2.7.9a). Gene-level counts were generated with featureCounts (v1.5.1) using Ensembl annotations. Downstream normalization, annotation, visualization and differential gene expression (DGE) analysis were conducted using DESeq2 (v1.44.0). For DGE, significance was assessed using a Wald test between experimental groups. Log₂ fold-change shrinkage was applied with apeglm, and p-values were adjusted for multiple comparisons using the Benjamini–Hochberg method. Genes were considered significantly differentially expressed when adjusted p < 0.05 and the log₂ fold-change > 1.

Gene set enrichment analysis (GSEA) was conducted using fgsea with gene sets derived from the Gene Ontology (GO) knowledgebase. Parameters included a minimum gene set size of 15 and a maximum of 500, with significance set at p < 0.005 after Benjamini–Hochberg correction. Over-representation analysis (ORA) was likewise performed using GO terms, using ClusterProfiler. Input genes were limited to those meeting the significance criteria (adjusted p < 0.05, log₂ fold-change > 1), up to 1,000 genes. Gene sets containing between 15 and 500 members were tested, and enrichment p-values were adjusted via the Benjamini–Hochberg method.

The pyramidal layers of CA1–CA3 and the granule layer of the DG were predominantly composed of NeuN-positive neurons. Limited glial staining was detected within the pyramidal layer, mainly corresponding to distal processes of GFAP-positive astrocytes. No hypertrophy of astrocytic processes was observed. Iba1- and CD45-positive cells were absent from the pyramidal layer, Olig2 staining was sparse overall, and occludin expression was confined to regions outside the pyramidal layer. NG2-positive cells were observed only in co-localization with occludin and were therefore not classified as oligodendrocytes; no NG2-positive but occludin-negative cells were detected. A representative panel illustrating the cellular distribution across hippocampal subregions is shown in Figure 3.

The mean number of total reads per sample ranged from 43.4 to 51.9 million, with consistently high mapping and assignment rates. Three samples were excluded during quality control based on objective sequencing and assignment metrics prior to analysis: one DG TBI sample due to insufficient sequencing depth (5.1 million total reads), and two sham samples due to markedly low numbers of assigned reads despite adequate total read counts, consistent with suspected technical or contamination-related issues (CA3 sham: 46.0 million reads, 3.49 million assigned; CA2 sham: 46.8 million total reads, 5.19 assigned). These excluded samples originated from different subregions and experimental groups and therefore did not disproportionately affect any specific region.

In CA1, sham samples yielded an average of 43.5 (SD = 3.8) million total reads, with 87.1% (SD = 1.8) uniquely mapped and 60.1% (SD = 2.2) assigned to annotated transcripts. TBI samples in CA1 showed comparable sequencing depth, with 43.9 (SD = 6.3) million total reads, 82.9% (SD = 4.2) uniquely mapped and 54.3% (SD = 3.2) assigned. In CA2, total reads averaged 50.0 (SD = 6.4) million for sham and 50.5 (SD = 13.4) million for TBI. In sham, 85.2% (SD = 1.4) were uniquely mapped and 59.6% (SD = 1.4) assigned. For TBI, 83.0% (SD = 1.2) were uniquely mapped and 56.1% (SD = 5.2) assigned. CA3 yielded 45.7 (SD = 1.47) million reads in the sham group and 50.5 (SD = 6.1) million in the TBI group. Mapping and assignment percentages remained consistent across interventions, in sham 86.8% (SD = 1.7) were uniquely mapped and 63.4% (SD = 2.0) assigned, after TBI uniquely mapped and assigned transcripts were 83.3% (SD = 3.0) and 62.1% (SD = 3.8), respectively. In DG, sham had 48.1 (SD = 10.5) million total reads with 85.8% (SD = 2.2) uniquely mapped and 56.9% (SD = 2.7) assigned. TBI samples in DG exhibited similar metrics 49.7 (SD = 1.8) million total reads, 85.5% (SD = 3.2) uniquely mapped, and 53.4% (SD = 5.1) assigned. No substantial differences were observed in sequencing depth or mapping efficiency between sham and TBI groups within any subregion, supporting consistent RNA quality and library preparation across conditions. A summary of dissected tissue size and sequencing quality indicators are presented in Table 1, and principal component analysis (PCA) clustering in Figure 4.

Selection of differentially expressed genes (DEGs) and significance thresholds were set at adjusted p-value < 0.05 and a log₂ fold change (LFC) > 1 or < −1.

The choice of bulk RNA-sequencing over single-cell methods presented both advantages and limitations. RNA-seq provides high sensitivity and broad dynamic range, while laser-capture microdissection preserved precise anatomical resolution and generated neuron-enriched samples, though minimal glial contamination was unavoidable. Hippocampal neuronal layers within each subregion are composed of multiple excitatory principal neuron subtypes with distinct transcriptional and functional profiles, which may influence bulk RNA-seq-derived signatures (52, 53). While LCMD enriches for pyramidal and granule neurons, residual heterogeneity among excitatory subtypes may contribute to averaged expression patterns. Importantly, approximately 95% of the pyramidal layer consists of neurons, a dominance confirmed by our immunofluorescent analyses (Figure 3) (54, 55). The examined regions showed predominantly neuronal staining, with only sparse GFAP-positive astrocytic processes extending into the pyramidal layer, while other glial and vascular markers were confined to adjacent areas. These GFAP-positive elements represent distal astrocytic processes rather than astrocyte somata, as no cell bodies were observed within the captured areas. Fine astrocytic processes contain minimal cytoplasmic volume and low total mRNA abundance (56, 57). Moreover, perisynaptic astrocytic processes are enriched for synaptic and metabolic functions rather than inflammatory gene programs, limiting their potential contribution to immune-related signatures (57). Notably, markers commonly regarded as glial, including GFAP and TSPO, are also expressed at low abundance in neurons, particularly following injury, and inflammatory mediators such as Toll-like receptors and TNF-α contribute to neuron-intrinsic inflammatory signaling (58–61). Taken together, these observations indicate that the sequenced material primarily reflects neuronal populations. Nevertheless, sparse astrocytic processes may contribute low-level astrocytic transcripts, and the resulting profiles therefore represent integrated, neuron-enriched subregional responses rather than purely neuronal transcriptional programs, an important consideration when interpreting glia-associated gene expression signatures. While single-cell RNA-seq could theoretically achieve cell-type–specific resolution, the limited amount of dissected tissue and the high neuronal content of these regions made bulk RNA-seq the most robust approach, ensuring deep coverage while preserving regional specificity.

Differential expression analysis applied strict thresholds (log₂ fold change > 1 and adjusted p < 0.05). While this increased confidence in the robustness of detected genes, it may underrepresent more subtle yet biologically relevant transcriptional changes. Modest regulation of key signaling or regulatory genes, as well as coordinated but small shifts across gene networks, can nevertheless play important roles in post-traumatic processes such as inflammation, metabolic adaptation, and synaptic remodeling. To address this, we complemented ORA, which relies on predefined cutoffs, with GSEA, which considers ranked expression data independent of thresholding. The combined use of ORA and GSEA therefore allowed us to capture both strongly dysregulated genes and more nuanced, pathway-level trends, providing a more comprehensive view of the transcriptional responses across hippocampal subregions. The concordance between ORA and GSEA supports the robustness of the identified biological themes while capturing both strong gene-level effects and more nuanced pathway-level regulation.

Although the hippocampal subregions exhibited fundamentally distinct transcriptomic profiles after penetrating TBI, a small set of stress-associated genes—TSPO, S100a4, Hspb1, Mt2A, Timp1, and Igfbp2—was consistently upregulated across all fields, indicating a conserved core response to injury. These genes represent central elements of the secondary injury cascade, reflecting activation of mitochondrial stress, oxidative imbalance, cytoskeletal instability, and ECM regulation. TSPO, a canonical marker of mitochondrial dysfunction and oxidative stress present in both neurons and glia, was elevated throughout the hippocampus, suggesting a global disturbance of mitochondrial homeostasis (60, 62). Hspb1 and S100a4, both induced in all subregions, mediate cytoskeletal stabilization and chaperone-assisted repair, counteracting protein denaturation and calcium-driven structural collapse (63, 64). The consistent increase of Mt2A reflects an antioxidant defense aimed at sequestering metal ions and limiting oxidative injury, while Timp1 and Igfbp2 indicate coordinated extracellular and trophic adaptations that help preserve matrix integrity and promote cell survival signaling (65–67). Collectively, this gene set defines a unified hippocampal cellular stress module integrating mitochondrial, cytoskeletal, redox, and extracellular protection mechanisms during the early secondary phase of injury. Despite this conserved transcriptional foundation, few enriched pathways were shared between subregions in the overall ontology analyses, underscoring that while the hippocampus mobilizes a common stress-defense machinery at the gene level, the broader biological context diverges sharply. CA1, CA2, CA3, and DG each extend this core program along distinct trajectories, ranging from structural stress signaling to inflammatory, metabolic, or reparative activation. These networks converge on core components of the secondary injury cascade, including inflammation, oxidative stress, and gliosis, which together shape the distinct molecular landscape of each hippocampal subregion and are examined in detail in the following sections.

A key finding of the present study was the region-specific inflammatory response across hippocampal subfields following pTBI. This aligns with the well-established subacute neuroinflammatory peak typically observed within the first week of injury (8). However, our results reveal distinct subregional patterns: CA2 and CA3 displayed strong and diverse inflammatory signatures, whereas CA1 and DG were devoid of enriched inflammatory pathways (Figure 11). The lack of inflammatory activation in CA1 and DG contrasts with previous reports describing widespread microglial and astrocytic responses in these subfields within 1–3 days post-injury (68, 69). The pyramidal and granule cell layers, which were selectively isolated in this study, are densely populated by neurons and contain relatively few glial or vascular cells (54, 55). As a result, our neuron-enriched design largely excluded transcripts from microglia, astrocytes, and endothelial cells, the primary mediators of cytokine signaling, leukocyte adhesion, and complement activation. Accordingly, the absence of inflammatory enrichment in CA1 and DG likely reflects the limited inflammatory response in CA1 and DG neurons rather than a true absence of immune activity within the tissue. Collectively, these findings indicate that the inflammatory milieu reported in previous studies originates mainly from non-neuronal compartments.

Although both CA2 and CA3 exhibited robust transcriptional activation of immune pathways, their inflammatory programs diverged markedly, indicating subfield-specific mechanisms of neuronal immune engagement. Increasing evidence suggests that injured neurons can express complement components, cytokine modulators, and adhesion molecules, enabling them to participate actively in early innate immune responses following trauma or ischemia, consistent with our findings (70–73). In CA2, the transcriptional response was dominated by genes involved in innate immune activation, complement signaling, and cell-adhesion processes. Enriched ontologies such as myeloid leukocyte activation, leukocyte–cell adhesion, and complement cascade point to a primarily innate, chemotactic inflammatory state within the pyramidal neuronal population. This transcriptional pattern is consistent with early complement engagement and adhesion-mediated recruitment of immune effectors, mechanisms known to facilitate microglial contact, synaptic surveillance, and localized remodeling rather than widespread inflammatory damage (74–79). The upregulation of complement components together with integrin signaling molecules and Fc-receptor family genes, suggests that CA2 neurons participate in the early immune–synaptic interactions. Such activity-dependent complement tagging has been linked to selective pruning of weakened synapses and synaptic homeostasis during both development and injury and may thus represent an attempt to stabilize circuitry after TBI. This interpretation aligns with emerging evidence that CA2 neurons possess intrinsic resilience to injury compared to neighboring CA1 and CA3 regions (43).

In contrast, the CA3 subfield exhibited a broader and more complex inflammatory signature, encompassing cytokine signaling, TNF-superfamily activation, and MHC class II antigen-presentation pathways, indicating a shift toward a mixed innate–adaptive neuronal immune response. This pattern is consistent with previous observations of robust hippocampal inflammation and microglial activation in the subacute phase after TBI, although most prior studies have focused on glial rather than neuronal populations (80–82). The present findings extend this literature by suggesting that CA3 neurons transcriptionally engage in immune signaling, implying that neurons are not merely passive targets of glial cytokines but actively shape the inflammatory environment through expression of immune mediators and antigen-processing machinery. Although blood–brain barrier (BBB) disruption occurs in our injury model, and regional differences in BBB permeability have been reported, the pronounced neuronal immune activation observed in CA3 likely reflects intrinsic subregional properties rather than a primarily vascular-driven effect (34, 83). CA3 pyramidal neurons form dense recurrent excitatory networks and receive powerful glutamatergic input from the DG, conferring a high degree of synaptic plasticity and metabolic demand. These same features also render CA3 highly susceptible to excitotoxic and cytokine-induced stress, where heightened excitatory drive amplifies inflammatory cascades (84–86). Functionally, such neuronal engagement in cytokine and TNF signaling could have implications for synaptic stability and hippocampal output. TNF-α and IL-1β are known to modulate glutamatergic transmission and dendritic spine morphology, shifting the excitatory–inhibitory balance and altering network synchronization (87–89). In the context of CA3’s recurrent circuitry, these processes could lead to hyperexcitability, impaired pattern completion, and reduced circuit resilience, ultimately contributing to cognitive and memory dysfunction observed after TBI. Thus, while CA2 appears to engage in complement-mediated synaptic refinement, CA3 demonstrates a pro-inflammatory and cytokine-driven activation pattern that may disrupt normal circuit communication and amplify secondary injury progression.

Our findings in the CA2 subregion revealed a strong signature of mitochondrial dysfunction, detected by GSEA but not ORA, suggesting that this response results from coordinated, moderate transcriptional shifts across multiple mitochondrial pathways rather than isolated, high-amplitude gene changes (90). The enriched GSEA terms included oxidative phosphorylation, mitochondrial respiratory chain complex assembly, ATP metabolic process, mitochondrial translation, and NADH dehydrogenase complex assembly, all showing negative enrichment scores. This indicates a broad suppression of mitochondrial bioenergetic and structural functions, encompassing both reduced oxidative metabolism and downregulation of genes related to mitochondrial organization and biogenesis. These features suggest that CA2 neurons enter a state of metabolic and structural restraint rather than activating repair or biogenic pathways.

This widespread downregulation aligns with established models of secondary injury after TBI, in which excitotoxic calcium influx and ischemia disrupt electron transport at mitochondrial complexes I and III, resulting in reduced ATP synthesis, membrane depolarization, and increased ROS formation (91–93). The coordinated suppression of oxidative-phosphorylation and NADH-linked pathways in CA2 likely reflects an adaptive reduction in metabolic flux, aimed at minimizing oxidative injury and stabilizing intracellular calcium homeostasis during the acute phase.

This transcriptional signature was unique to CA2, highlighting the region’s distinctive metabolic and physiological profile. CA2 neurons are enriched for mitochondrial and calcium-handling machinery, including the mitochondrial calcium uniporter complex, and exhibit dense dendritic mitochondrial networks that tightly couple synaptic activity to oxidative metabolism (43, 94, 95). This architecture sustains high energetic throughput, but following injury it may predispose CA2 mitochondria to calcium-driven overload and depolarization. The observed downregulation of mitochondrial and respiratory-chain pathways therefore likely represents a controlled metabolic rebalancing, a temporary suspension of oxidative phosphorylation to limit ROS generation and preserve organelle integrity. Together with its high mitochondrial capacity and robust calcium-buffering mechanisms, these features likely provide CA2 with a larger metabolic reserve, allowing neurons to withstand energy downscaling without immediate degeneration (43, 94).

While CA2 is often considered one of the most resilient hippocampal subfields, direct evidence of its oxidative vulnerability remains limited. Under physiological conditions, its enhanced antioxidant and calcium-buffering potential likely confer resistance to excitotoxic and ischemic injury (43). However, these same features may render CA2 highly responsive to mitochondrial stress, as its dense mitochondrial population and tight calcium–metabolic coupling are energetically demanding (95). The observed suppression of mitochondrial structural and functional pathways therefore likely represents a regulated protective mechanism, aimed at conserving energy and preventing oxidative damage rather than indicating irreversible dysfunction. Nevertheless, prolonged energetic depression would inevitably constrain ATP production, impair ion homeostasis, and reduce synaptic efficacy (96). Given CA2’s crucial role in social memory and temporal coding, such bioenergetic limitations could attenuate its excitatory output to CA1 and disrupt hippocampal network synchrony, leading to subtle but persistent cognitive deficits even without significant cell loss.

The hippocampal subregions displayed distinct transcriptional profiles reflecting differences in reparative capacity after injury. The DG showed selective activation of processes related to gliosis, tissue regeneration, and vascular remodeling, whereas CA1 exhibited a comparatively muted, cell-autonomous response dominated by cytoskeletal organization and cell-cycle regulation. No comparable activation of reparative pathways was detected in CA2 or CA3. As the laser-captured material was enriched for neurons, these results primarily represent neuron-derived transcriptional programs, suggesting that the apparent gliogenesis-associated and angiogenic activity in the DG reflects neuronal engagement of repair-associated signaling networks rather than widespread glial transcription.

In the DG, both GSEA and ORA analyses revealed enrichment of gliogenesis, astrocyte differentiation, ECM organization, regulation of angiogenesis, and enzyme inhibitor activity, together defining a reparative transcriptional state characterized by matrix stabilization, vascular support, and proteolytic control. The enrichment of protease-inhibitory and ECM-related pathways implies that DG neurons contribute to limiting secondary tissue breakdown, maintaining BBB-stability, and creating an environment conducive to microvascular remodeling and neurogenesis (97, 98). Functionally, these processes suggest that the DG enters a protective restructuring phase, aimed at stabilizing the injured microenvironment and facilitating subsequent neuronal and vascular recovery. Although the DG showed gliogenic enrichment, the magnitude of this response was modest, consistent with a low-to-moderate activation pattern rather than a large-scale gliotic shift. This restrained transcriptional activity aligns with recent single-cell and spatial transcriptomic studies, which similarly report limited astroglial proliferation in the DG after injury, alongside prominent neurogenic and niche-stabilizing responses (31, 99, 100). TBI biases neural stem cell fate toward neurogenesis and away from astrogliogenesis, highlighting that early DG remodeling primarily reflects neurogenic and vascular niche activation rather than astrocyte expansion (31). The moderate gliogenesis terms observed in this study are therefore best interpreted as neuron-driven signaling that interfaces with glial and vascular pathways, consistent with a reparative but not proliferative response. Importantly, many genes annotated to gliogenesis, angiogenesis, and ECM-pathways are also induced in neurons following injury, where they contribute to neuron–glia-vascular signaling and tissue repair rather than astrocyte proliferation per se (101, 102). The presence of moderate GFAP enhancement in the DG warrants cautious interpretation. While GFAP splice variants in hippocampal neurons have been described, neuronal GFAP expression remains debated (58, 59, 103, 104). Consequently, the observed GFAP signal may partly be explained by minor astrocytic contamination of the laser-captured tissue rather than genuine neuronal transcription. Nonetheless, such limited contamination does not account for the full range of enriched processes—particularly ECM organization, angiogenesis regulation, and protease inhibition—which indicate a coordinated, neuron-driven reparative response that partially overlaps with glial signaling domains. This composite signature likely represents the early molecular phase of neuron–glia–vascular coordination, consistent with the DG’s established role as the hippocampal hub of post-injury remodeling.

Our study has several limitations. The modest sample size per hippocampal subregion reflects the experimental design, which required LCMD of multiple regions from each animal, thereby constraining the number of biological replicates and reducing power to detect transcripts with small effect sizes. In addition, only male animals were included to avoid further reducing the sample size per group, which limits inference regarding sex-specific transcriptional responses. The low RNA input inherent to LCMD-based sampling may also reduce sensitivity for low-abundance transcripts. These limitations were addressed by applying conservative statistical criteria and by emphasizing consistent, biologically coherent transcriptional patterns across subregions.

In summary, the present transcriptomic analysis reveals striking subregional specialization in the hippocampal response to penetrating TBI. CA2 and CA3 displayed robust yet divergent neuronal inflammatory programs consistent with their distinct circuit properties, with CA2 engaging an innate, complement-associated response suggestive of synaptic stabilization within a metabolically constrained and structurally stable network, whereas CA3 exhibited a cytokine-driven activation profile aligned with its densely recurrent excitatory circuitry and known susceptibility to activity-dependent amplification and excitotoxic stress (43, 44, 105). In parallel, CA2 showed coordinated downregulation of oxidative phosphorylation and mitochondrial biogenesis pathways, consistent with protective metabolic suppression aimed at limiting oxidative injury and preserving organelle integrity. By contrast, the DG mounted a modest but distinct reparative transcriptional program characterized by neuronal engagement of gliogenic, angiogenic, and ECM-pathways, in line with its established role in plasticity, neurogenesis, and circuit remodeling after injury (102). Together, these findings delineate a heterogeneous hippocampal response in which CA2 prioritizes metabolic restraint and synaptic stabilization, CA3 engages pro-inflammatory and stress-amplifying signaling, and the DG initiates neuron-driven repair, collectively highlighting region-specific strategies that shape post-traumatic recovery and vulnerability across the hippocampal circuit. Functionally, these region-specific responses suggest differential circuit consequences, with CA3 inflammation contributing to network instability and impaired pattern completion, CA2 metabolic suppression favoring neuronal survival at the cost of reduced excitatory throughput and DG-driven repair and stabilization of circuitry. From a translational perspective, this subfield-specific framework provides a foundation for developing targeted neuroprotective and immunomodulatory interventions to mitigate neuronal dysfunction and cognitive decline. Future studies including both sexes, integrating single-cell and spatial transcriptomics with validation of key pathways such as mitochondrial suppression in CA2 and extensive inflammatory activation in CA3 will be crucial to link these molecular programs to functional outcomes and to guide precision strategies for limiting secondary injury and promoting circuit restoration after TBI.