Dissecting Gene Expression Networks in the Developing Hippocampus through the Lens of NEIL3 Depletion

Gene regulation in the hippocampus is fundamental for its development, synaptic plasticity, memory formation, and adaptability. Comparisons of gene expression among different developmental stages, distinct cell types, and specific experimental conditions have identified differentially expressed genes contributing to the organization and functionality of hippocampal circuits. The NEIL3 DNA glycosylase, one of the DNA repair enzymes, plays an important role in hippocampal maturation and neuron functionality by shaping transcription. While differential gene expression (DGE) analysis has identified key genes involved, broader gene expression patterns crucial for high-order hippocampal functions remain uncharted. By utilizing the weighted gene co-expression network analysis (WGCNA), we mapped gene expression networks in immature (p8-neonatal) and mature (3m-adult) hippocampal circuits in wild-type and NEIL3-deficient mice. Our study unveiled intricate gene network structures underlying hippocampal maturation, delineated modules of co-expressed genes, and pinpointed highly interconnected hub genes specific to the maturity of hippocampal subregions. We investigated variations within distinct gene network modules following NEIL3 depletion, uncovering NEIL3-targeted hub genes that influence module connectivity and specificity. By integrating WGCNA with DGE, we delve deeper into the NEIL3-dependent molecular intricacies of hippocampal maturation. This study provides a comprehensive systems-level analysis for assessing the potential correlation between gene connectivity and functional connectivity within the hippocampal network, thus shaping hippocampal function throughout development.


Introduction
The hippocampal formation is an important brain structure, extensively studied for its crucial roles in memory consolidation, information processing, and emotional regulation (Buzsáki and Moser, 2013;Moser and Moser, 1998;Scoville and Milner, 1957).Anatomically, the hippocampal structure includes the following subfields: cornus ammonis area (CA1-CA3) and dentate gyrus (DG).These subregions contain distinct cell types and connectivity patterns, forming the core of the hippocampal trisynaptic circuit (Amaral and Witter, 1989).In rodents as in other mammals, the hippocampus has an extended developmental trajectory with refinements occurring in the trisynaptic circuit until adolescence (Gomez and Edgin, 2016).During the first weeks after birth, a large part of the hippocampal neurons is still in an immature state.This state of immaturity could be defined by the expression of specific anatomical markers, the size of dendritic projections, and the number and composition of synaptic connections (Michael and László, 2006).In mice, postnatal hippocampal development progresses through sequential maturation, with CA3 maturing at p20, CA1 at p23, and DG at p26 (Donato et al., 2017).This progression is evaluated by assessing the expression of anatomical markers such as DCX and NeuN.A growing body of evidence demonstrates that the hippocampal architecture and connectivity are sculpted by early development and the dynamics of hippocampal network are prewired before the brain fully matures (Cossart and Khazipov, 2022).This prewiring is essential for the basic functioning of the hippocampal circuitry and serves as a foundation upon which further neural plasticity and learning can build.
Many of the molecular determinants involved in neurodevelopment and network formation are well-characterized in the hippocampus (Goncalves et al., 2016;Urban and Guillemot, 2014).New factors governing epigenome regulation and thus the spatiotemporally controlled transcription of these determinants are constantly being added to the picture (Fan et al., 2018;Salinas et al., 2020).The DNA glycosylase NEIL3 is highly expressed in the developing central nervous system, particularly within neurogenic niches such as the hippocampus and the subventricular zone (Hildrestrand et al., 2009;Rolseth et al., 2008).In mice, the highest expression of NEIL3 has been observed in early postnatal days and decreases gradually with age (Sejersted et al., 2011a).Previous studies suggested the role of NEIL3 in neural progenitor cell survival, post-stroke neurogenesis, hippocampal adult neurogenesis, and shaping complex behavioral phenotypes (Regnell et al., 2012;Sejersted et al., 2011a).Our recent work revealed delayed CA1 maturation, altered CA1 transcriptome, and impaired functional plasticity of CA1 spatial cells in Neil3 -/-mice (Kunath et al., 2021).This highlights the significance of NEIL3 in orchestrating the formation of the hippocampal network.
RNA sequencing (RNA-seq) has emerged as a valuable tool for dissecting hippocampal development.It offers an in-depth perspective on gene expression patterns, regulatory networks, and the molecular mechanisms shaping the formation and maturation of this crucial brain area.A growing body of research compares the gene expression profiles of the hippocampus within distinct cell types (Cembrowski et al., 2016), at different developmental stages (Lee et al., 2017;Mody et al., 2001;O'Reilly et al., 2015;Olsen et al., 2023), or under specific experimental conditions (e.g., disease models) (Nashed et al., 2016).A remarkable transcriptional diversity has been revealed in the developing dorsal and ventral hippocampal subregions of rat brains at different postnatal stages, demonstrating the dynamic changes of gene expression during hippocampal maturation (Lee et al., 2017;O'Reilly et al., 2015;Olsen et al., 2023).Gene ontology and pathway enrichment analyses elucidate the biological functions and pathways associated with differentially expressed genes, which provides insights into the molecular processes and cellular functions underlying hippocampal development (Greene et al., 2009).
However, differential gene expression (DGE) analysis focuses on individual genes with significant expression changes in distinct groups.It does not consider broader gene expression patterns and networks that could reveal broader functional insights and potential regulatory relationships within the hippocampal network.Weighted gene co-expression network analysis (WGCNA) provides a systems biology perspective by exploring relationships of the co-expressed gene clusters across samples that may be co-regulated and participate in common biological processes or pathways (Langfelder and Horvath, 2008;Zhang and Horvath, 2005).WGCNA generates a complex network structure, as a dendrogram and a module-trait relationship heatmap (Langfelder and Horvath, 2008;Zhang and Horvath, 2005).Analyzing these networks can reveal key modules associated with specific traits, such as developmental stages, and provide broader functional insights and potential regulatory relationships.This correlation network methodology has been successfully applied to describe the conservation and evolution of gene networks in primate brains (Oldham et al., 2006), for the prediction of novel key genes in Alzheimer's Disease (Liang et al., 2018;Sun et al., 2022) J o u r n a l P r e -p r o o f as well as in cancer biomarkers identification and prognosis (Di et al., 2022;Rezaei et al., 2022;Yin et al., 2020).
In this study, we explored the postnatal maturation of hippocampal subregions in both wildtype and Neil3 -/-mice, utilizing the advanced capabilities of WGCNA.This approach allows us to decode gene expression networks in both the immature (p8-neonatal) and mature (3m-adult) stages of hippocampal circuits.Leveraging a substantial dataset of 42 transcriptome samples from both genotypes across CA1, CA3, and DG at key developmental stages, we investigated the complexities of gene co-expression networks, highlighting features integral to hippocampal development and synaptic functionality.Our detailed visualization of co-expressed gene modules uncovered hub genes that corresponded to the functional maturity and region-specific attributes of the hippocampal network.By investigating the module differences in NEIL3-deficient mice, we identified NEIL3associated hub genes within specific modules, suggesting an important role of NEIL3 in hippocampal maturation and function.Combining the insights from WGCNA and DGE analysis, we provided a comprehensive view of the intricate molecular mechanisms, tethered to NEIL3, that underpinned hippocampal development.This study provides a systems-level analysis revealing potential contexts between gene connectivity and functional connectivity in the hippocampal network.

Tissue collection.
Mice were anesthetized using isoflurane (Baxter, Oslo, Noway) and subsequently killed by an intraperitoneal overdose of pentobarbital (>200mg/kg body weight).An intracardial perfusion with 0.9% saline (B.Braun, Melsungen, Germany) with a constant, steady, hand-controlled flow for 3-4 minutes was followed by fixation with 4% paraformaldehyde in phosphate-buffered saline (PBS), also administered intracardially.The brain was collected and put into 4% paraformaldehyde/PBS solution for a minimum of 48 hrs at 4°C for definite fixation.Brains were cut at a thickness of 30µm using a cryostat (Leica CM3050S, Nussloch, Germany) and subsequently stored at 4°C in a PBS-solution containing 0.05% of Proclin (Merck, Darmstadt, Germany) as a preservative.We chose a sagittal cutting direction and started collecting brain slices at a mediolateral depth of 900µm and continued until the end of the tissue block.Two medial and two lateral brain slices were used for the immunostaining to avoid a tissue collection bias alongside the mediolateral axis.

Confocal imaging and 3D image analysis
All fluorescent images were taken by confocal microscope (Zeiss LSM880).We took individual images (z-stack 2 µm) in hippocampal CA1, CA3, and DG, using a Plan-Apochromat 40x/1.4Oil DIC M27 objective (Carl Zeiss, Jena, Germany).NeuN-positive mature neurons were quantified using the 3D image analysis software Imaris 9.3 (Bitplane) as published (Kunath et al., 2021).Briefly, we detected NeuN+ cells using the Imaris spot detection tool.A 5µm fixed seed point diameter proved most reliable in detecting all cells of interest.We set the threshold for low-intensity versus high-intensity NeuN+ cells at the point where the rise in average intensity between both cell populations was steepest, using the Imaris intensity histogram to guide the decision of where to set the threshold.The mean intensity value of DCX-reactivity within the nuclei was used as a threshold for DCX-detection, assuming that no DCX-reactivity should be present across the nuclear area.We detected the entire population of NeuN+ cells and determined the average intensity of DCXreactivity within these cells.Only cells with an average intensity higher than the within-nuclei-DCX-intensity were considered for counting.
Three two-way-ANOVAs were performed to compare results between genotypes across timepoints, specifically within the CA1, CA3 and DG regions respectively.The threshold for significant results was set at p<0.05.A correction for multiple testing was performed in all cases, generally employing Šídák's method.All comparisons were conducted with 1 animal equaling 1 statistical unit.In general, this was done by averaging values of several samples (e.g., several hippocampal slices) taken from the same specimen.All statistical analysis was conducted using GraphPad Prism Version 10 (GraphPad Software, San Diego, USA).

Dissection of hippocampal subregions (CA1, DG, CA3)
A needle-scratch method was established in our lab to precisely isolate hippocampal subregions with little prone to contamination from adjacent anatomical areas (Kunath et al., 2021).The brain was taken out from the mouse without intracardial perfusion, mounted on a cryostat metal socket (Leica CM3050S) using a drop of mounting media (Tissue-Tek OCT compound), immediately frozen using pulverized dry ice (101 Cold Spray) and kept on dry ice until further processing.Coronal brain sections were frozen sectioned at a thickness of 100μm and mounted immediately onto gelatin-coated slides (SuperFrost Plus, ThermoFisher).After drying for 10-15min, subregions of the hippocampus (CA1, CA3, or DG) were scratched with a 30G needle (B.Braun) and collected in a dissection cap (MMI).The samples were lysed in RLT lysis buffer (AllPrep Kit, Qiagen) using a bead homogenizer (MagNAlyser, Roche) and frozen at −80°C until further processing.

RNA sequencing
RNA was extracted using the RNeasy Mini Kit from Qiagen according to the manufacturer's instructions.RNA samples typically yielded >100ng of RNA with a RIN value of >7 as determined by Bioanalyzer (Agilent Technologies).Whole-transcriptome sequencing was done by BGI Genomics Co., Ltd., Hong Kong, China.The samples were sequenced with the BGISEQ-500 platform with a read depth of 20M clean reads per sample, averagely generating about 4.92 Gb bases per sample.The sequencing reads containing low-quality, adaptor-polluted, and high content J o u r n a l P r e -p r o o f of unknown base (N) reads were processed and removed.The clean data were provided by BGI and the bioinformatic analysis was performed in our lab.

Data and preprocessing
The RNAseq dataset consisted of a total of 42 samples, including wildtype samples (4 of p8-CA1, 4 of p8-CA3, 4 of p8-DG, 3 of 3m-CA1, 3 of 3m-CA3 and 3 of 3m-DG) and Neil3 -/-samples (4 of p8-CA1, 4 of p8-CA3, 4 of p8-DG, 3 of 3m-CA1, 3 of 3m-CA3 and 3 of 3m-DG).The raw data of gene expression was obtained from BGI.To reduce computational burden and possibly enhance the signal in our data, DESeq2 (Anders and Huber, 2010;Love et al., 2014) was used to identify the subset of 15,000 genes that were most highly variated in mature and immature hippocampal subregions in wildtype mice.The normalized expression levels were transformed by the variance stabilizing function DESeq2::vst and then used for further analysis in both genotypes.DESeq2 was further used to test for differential expression between genotypes in each region and for each age.Differential gene expression was determined by the threshold of adjusted p-value < 0.05 and ABS(log 2 fold change) >0.6 (Kunath et al., 2021).
The principal component analysis was performed using PCA tools in the R package (Blighe K., 2022).The 10% of variables with the lowest variance were removed and principal component analysis was performed.The eigenvector plot was created to explore how the variance among treats (genotype, age, and brain region) is explained.

Weighted gene co-expression network construction
Weighted gene co-expression network analysis (WGCNA) package and the set of Network Functions (Oldham et al., 2006) were used for the co-expression network construction (Langfelder and Horvath, 2008).To identify outliers sample clustering tree was created using the hclust function and no outliers were detected.Then, the function pickSoftThreshold was used to choose an appropriate soft-thresholding power (β) based on a scale-free topology criterion.Weighted adjacency matrices were calculated using the power of 6 (Zhang and Horvath, 2005) for both wildtype and mutant mice (Fig. S3B and Fig. S4A-B).Minimum connections of 0.2 were set based on median connectivity which resulted in distance matrices based upon topological overlap of 9630 genes with the highest connectivity.The trait information (genotype, age, and region) was imported, binarized, and correlated with the expression data.

Network module detection
For module detection, we limited our analysis to the 9630 most connected genes, which were selected using the threshold connectivity k = 0.2 based on median connectivity in wildtype (k = 0.26) and mutant (k = 0.23).By definition, module genes are highly connected with the genes of their module.Thus, for module detection, restricting the analysis to the most connected genes should not lead to major information loss (Fuller et al., 2007).Hierarchical clustering was used to group genes with coherent expression levels into modules.The height cut-off value of the dendrogram was chosen by inspection and set to 0.7, and the minimum size of the module was set to 30.

Connectivity, hub gene detection, and module characterization
For each gene, the connectivity (also known as degree) is defined as the sum of connection strengths with the other network genes: k i = ∑u≠ia ui .In co-expression networks, the connectivity measures how correlated a gene is with all other network genes.Hub genes are defined as genes with the sum of connection weights with all other nodes (Dong and Horvath, 2007;Langfelder et al., 2013).The key hub genes were identified using wgcna::chooseTopHubInEachModule function and the TOP highest connected genes were exported to Visant, a free software for biological network analysis (Hu et al., 2004;Hu et al., 2005).Visant is a Java-based, platform-independent tool suitable for a J o u r n a l P r e -p r o o f wide range of biological applications, including studies of pathways, gene regulation and systems biology.
The main hub gene and the top 10 most connected genes were identified for each module and exported to Visant.

Differential gene expression in immature and mature hippocampal subregions
Next, we dissected the neuronal layers in hippocampal subregions at p8 and 3m (Kunath et al., 2021), and profiled subregion-specific gene expression in immature and mature hippocampal neurons.The RNAseq dataset consists of 42 samples, including wildtype and Neil3 -/-CA1, CA3, and DG at p8 (4 replicates of each subregion per genotype) and adult (3 replicates of each subregion per genotype) stages.The principal component analysis (PCA) (Blighe K., 2022) reveals that age (PC1), region (PC2), and genotype (PC3) were the main variance determinants, accounting for 80% of the total variation (Fig. 1B).As illustrated by the PCA plots (Fig. 1C), around 10% of the overall J o u r n a l P r e -p r o o f variation was mainly attributed to the genotype difference at p8, not in adults, with almost no differences observed among hippocampal subregions.This result suggests that NEIL3 deletion markedly affected the overall gene expression in the neonatal hippocampus, supporting a role of NEIL3 in hippocampal maturation.

Gene co-expression networks in wildtype and Neil3 -/-hippocampus differ.
To comprehensively elucidate gene expression dynamics during the postnatal hippocampal subregional development, especially under the effects of NEIL3 depletion, we applied the weighted gene co-expression network analysis (WGCNA) (Langfelder and Horvath, 2008;Oldham et al., 2006).The weighted gene co-expression network was constructed based on the subset of 15000 genes with the highest variation selected from the RNAseq dataset of the immature (p8) and mature (3m) hippocampal subregions in wildtype mice (Fig. S3A).Although gene expression levels were highly correlated in wildtype and Neil3 -/-mice networks, the correlation of network connectivity was reduced (0.99 vs 0.82, Fig. S4C).Analysis of the 9630 most connected genes in wildtype identified 11 distinct modules (Fig. 2A, the tree cut-off 0.7), with PCA showing the first principal component accounted for 73% to 86% of the variance in these modules (Fig. S5).The grey module, lacking coherent expression patterns (PC1 explaining only 42% of variance), was excluded from further analysis (Fig. S5A).To represent the extent of module similarity between the wildtype and Neil3 -/-gene co-expression networks, the 9630 most-connected genes in wildtype were analyzed in the Neil3 -/-network, assigned with the same wildtype module colors.Strikingly, the module distribution was completely alerted in the Neil3 -/-network (Fig. 2C).The MDS plot shows a different organization of module genes in the Neil3 -/-network, compared to the wildtype (Fig. 2D), underscoring NEIL3's importance in maintaining gene network structure and dynamics.
Module analysis reveals specific gene co-expression networks in immature and mature hippocampal circuits.
Our analysis of the wildtype hippocampal transcriptome revealed modules of varying sizes, each representing a network of highly interactive genes (Fig. 3A).Hub genes in each module, identified as those with the highest cumulative connection weights across all other nodes (Dong and Horvath, 2007;Langfelder et al., 2013), were determined through network analysis using the VisANT software (Hu et al., 2004;Hu et al., 2005).In line with the module-trait analysis (Fig. S5B), the expression of top 10 hub genes in each module, visually represented in heat plots, represents the module specificity in immature (p8) and mature (3m) hippocampal subregions (Fig. 3B and Fig. S6).An intriguing observation in our examination of hippocampal gene expression is the divergent regulation of hub genes across developmental stages.Notably, the 1-brown, 2-yellow, and 6magenta modules contained highly connected hub genes up-regulated in immature DG and mature CA1/CA3 (Fig. 3B, left panels).The 8-red, 5-black, 3-blue, and 9-purple modules mainly consisted of up-regulated hub genes in immature hippocampal subregions (Fig. 3B, middle panels).The 7pink, 10-greenyellow, and 4-green modules represented gene co-expression networks specific for adult CA1, CA3, and DG, respectively (Fig. 3B, right panels).The 11-turquoise module represented the most differentially regulated genes during postnatal hippocampal development but without subregion specificity.The expression patterns of example hub genes show similar trajectory as confirmed by in situ hybridization (ISH) (Fig. S7, ISH images obtained from the Allen Developing Mouse Brain Atlas), supporting that our analysis robustly clustered coordinated genes with subregion-and age-specific expression patterns.In addition, we performed interactive Gene Ontology (GO) analysis and identified module-specific GO terms and pathways (Table S1 and S2).Specifically, the 1-brown, 3-blue, and 8-red modules represented enriched GO-terms in neurodevelopmental processes; the 2-yellow module in synaptic regulation; and the 6-magenta module in postsynaptic density and GABAergic signaling.
NEIL3 impacts module connectivity and specificity in immature and mature hippocampus.

J o u r n a l P r e -p r o o f
NEIL3 depletion influenced the overall gene co-expression network in the hippocampus (Fig. 2 and Fig. S4C) and the correlation of gene connectivity in each module was low (average rho = 0.56, excluding the grey module) (Fig. 4A).To further explore these disruptions, we performed the differential gene expression (DGE) analysis in NEIL3-deficient hippocampal subregions at immature (p8) and mature (3m) stages and determined the number of differentially expressed genes (DEGs) in distinct modules (Fig. 4B).
To assess the impact of NEIL3 on the connectivity and specificity of individual modules, we calculated the ratio of the gene connectivity in wildtype to Neil3 -/-network, named as similarity ratio (SimRatio) as described in Method (Oldham et al., 2006).Briefly, pairs of genes were identified with high topological overlap (TO) in wildtype and low TO in Neil3 -\-in each module, and the pairs consisting of the top 10 hub genes were examined to identify the most affected hub genes.SimRatio exceeded 0.7, indicating a significant decrease in connectivity upon NEIL3 depletion.Two hub genes (Casp2 and Cdk6), highly expressed in p8-DG, displayed the most reduced connectivity in the 1-brown module (Fig. 5A), suggesting that NEIL3 influences the brown module specificity by targeting the Casp2 and Cdk6 co-expression network in the immature DG.The hub gene Pcdh7, highly expressed in p8-CA3, showed the most reduced connectivity in the 5black module (Fig. 5B), suggesting that NEIL3 influences the 5-black module specificity by targeting the Pcdh7 co-expression network in the immature CA3.Strikingly, the main hub gene Fbxo34, highly expressed in p8-DG, had reduced connectivity to all genes in the 6-magenta module (Fig. 5C, mean connectivity difference = 0.24, p = 1.7852E-32), suggesting that NEIL3 influences the 6-magenta module specificity by targeting the Fbxo34 co-expression network in the immature DG.All hub genes (top 10), highly expressed in immature hippocampal subregions, showed reduced connectivity in the 3-blue module (Fig. 5D), suggesting that NEIL3 affects the overall gene co-expression network in the 3-blue module driving its specificity in the immature hippocampus.Of note, most of these hub genes did not show differential regulation in Neil3 -/-hippocampus, except Fbxo34 that had reduced expression in Neil3 -/-p8-DG, specifically (Fig. 5).Other differentially connected genes (e.g., Tst in the 1-brown module, Coro1b in the 5-black module, Cfap20, Gabra2, etc. in the 3-blue module) did not show the consistent up and down dysregulation direction (Fig. 5).These results suggest that the gene co-expression network is disrupted by a synergistic effect of up-and down-regulated genes caused by NEIL3 depletion.
The 11-turquoise module represented the general specificity between the immature (p8) and mature (3m) hippocampus (Fig. 3B).All top 10 hub genes showed very high simRatio paired with 48 genes in Neil3 -/-network (Fig. 7A), indicating a considerable reduction in module connectivity.While all hub genes demonstrated similar expression levels in both wildtype and Neil3 -/-hippocampus (Fig. 7B), the genes differently connected to the hubs displayed patterned dysregulation in either immature or mature HPC, respectively (Fig. 7C).
In summary, these results highlight an important role of NEIL3 in shaping gene networks crucial for the development and function of the hippocampus.NEIL3 depletion leads to impaired connectivity and specificity of gene expression network in both immature (affected hub genes such as Casp2, Cdk6, Pcdh7, Fbxo34, and others) and mature (affected hub genes such as Bag5, Sgk1, MPP5, Syt3, Calr, and Rbm28) hippocampal stages.The overarching specificity revealed in the 11-J o u r n a l P r e -p r o o f turquoise module underscores NEIL3's influence on gene co-expression networks, spanning both immature and mature hippocampal stages.

Discussion
By the detection of the adult-like neuronal marker, we identified that depletion of NEIL3 led to delayed maturation of all three hippocampal subregions (CA1, CA3, and DG), with no discernible impact on the overall number of mature neurons in adulthood (3m) (Fig. 1A).Further, RNAseq analysis from relatively homogenized neuronal samples of the neonatal and adult hippocampal subregions demonstrated heterogenous gene expression profiles and revealed a prominent effect of NEIL3 depletion at the p8 immature state (Figure 1B-C).This aligns with our previous findings that NEIL3 influences CA1 maturation by shaping transcription (Kunath et al., 2021).The internal dynamics of hippocampal cells and circuits are prewired during development which is essential for the functional organization of the adult hippocampus (Cossart and Khazipov, 2022).Notably, NEIL3-depleted neurons in adulthood, following a delayed maturation process, exhibited functional impairment, as evidenced by a reduction in the spatial stability of CA1 place cells (Kunath et al., 2021).Hippocampal place cells, the critical components of the brain's spatial navigation system and memory formation processes, are generated within a network of interconnected brain regions, including the entorhinal cortex, DG, CA3, and CA1 (Witter et al., 2014).Our work suggests that NEIL3 impacts the pre-configuration of the functional hippocampal network by shaping transcription during development, highlighting its potential relevance to neurological conditions associated with hippocampal dysfunction.Using a constitutive Neil3 knockout mouse model presents challenges in distinguishing the direct from indirect roles of NEIL3 in shaping the hippocampal transcriptome and cellular functions.Further research is necessary to elucidate the NEIL3-targeted molecular mechanisms underlying hippocampal development and function.
Gene expression profiling is an important technique to decode the functional dynamics and structural organization of hippocampal cells and networks (Cembrowski et al., 2016;Greene et al., 2009;Lee et al., 2017;Mody et al., 2001;Olsen et al., 2023;Thompson et al., 2008).Differential Gene Expression (DGE) analysis reveals the heterogeneity of gene expression levels corresponding to the variability and functions of different cell types in the hippocampal network.The DGE analysis focuses on individual genes with significant expression changes, which is fundamental for understanding how gene expression patterns change in response to different biological states, treatments, or experimental factors.In this study, we used the weighted gene co-expression network analysis (WGCNA) to elucidate the large-scale organization of gene expression networks in immature (p8-neonatal) and mature (3m-adult) hippocampal circuits, specifically investigating the impact of NEIL3 depletion (Fig. 2).WGCNA aids in viewing these networks through a systems biology perspective, identifying modules of genes with coherent expression patterns that are potentially linked to certain functions or characteristics (Langfelder and Horvath, 2008;Zhang and Horvath, 2005).Highly connected genes in gene co-expression networks are shown to play an important, organizing role in biological networks (Carlson et al., 2006;Han et al., 2004).The connectivity measure was successfully used to pinpoint biologically significant genes in cancer (Horvath et al., 2006) and to identify conserved gene networks in primates (Oldham et al., 2006).
The developmental WGCNA has been elucidated in the embryonic and postnatal mouse diencephalon using voxels of the gene expression data derived from ISH (Thompson et al., 2014).Our study marks the application of WGCNA in the focused context of hippocampal development using region-specific RNA-seq data.We characterized distinct gene expression networks in immature and mature hippocampal circuits and the module specificity was driven by the specific expression pattern of highly connected hub genes (Fig. 3).Notably, the expression of hub genes in individual modules showed an divergent regulation across hippocampal subregions during maturation, e.g., upregulated hub genes in immature DG vs mature CA1/3 (shown in the 1-brown, J o u r n a l P r e -p r o o f 2-yellow and 6-magenta modules), in immature CA1 and/or CA3 (shown in the 8-red module, 5black, 3-blue, and 9-purple modules), and in mature CA1, CA3, DG, respectively (shown in the 7pnk, 10-greenyellow, and 4-green modules) (Fig. 3).The 11-turquoise module represented most differentially regulated genes during postnatal hippocampal development but without subregion specificity (Fig. 3).The expression patterns of example hub genes were found to be similar, as observed by in situ hybridization (ISH) images obtained from the Allen Developing Mouse Brain Atlas (Fig. S7).Our gene profiling analysis is sensitive and quantitative, enabling robust clustering of coordinated genes with subregion-and age-specific expression patterns.The dynamics of gene co-expression networks in immature and mature hippocampal subregions may reflect the functional difference of immature and mature hippocampal networks.Immature hippocampal networks exhibit high levels of synaptic plasticity, including long-term potentiation (LTP) and long-term depression (LTD), which is crucial for the establishment of synaptic connections during development (Leinekugel, 2003).Indeed, three modules (1-brown, 3-blue, and 8-red) represented enriched GO-terms in neurodevelopmental processes; 2-yellow module in synaptic regulation; and 6-magenta module in postsynaptic density and GABAergic signaling, suggesting that these modules represent specific gene co-expression networks contributing to the functional maturation of the hippocampal network.
NEIL3 depletion interfered with the overall gene co-expression network in the hippocampus (Fig. 2).By calculating the ratio of the gene connectivity in wildtype to Neil3 -/-network, we assessed the impact of NEIL3 on the connectivity and specificity of individual modules.Strikingly, in the 3blue module (Fig. 5D) and the 11-turquoise module (Fig. 7A), all top 10 hub genes showed a significant decrease in gene connectivity in response to NEIL3 depletion, demonstrating an important role of NEIL3 in shaping the immature hippocampal network by modulating specific gene co-expression networks.In other modules, 1-2 hub genes that drove the module specificity to distinct hippocampal subregions were specifically targeted by NEIL3.For example, in the 1-brown module (Fig. 5A), the connectivity of two hub genes (Casp2 and Cdk6) was affected by NEIL3 depletion.CASP2 plays an important role in stress-induced apoptosis, DNA repair, and tumor suppression (Hu et al., 2015;Puccini et al., 2013;Shalini et al., 2015), and CDK6 plays an important role in the progression and regulation of the cell cycle (Malumbres and Barbacid, 2005;Meyerson and Harlow, 1994).In the 5-black module (Fig. 5B), the targeted hub gene by NEIL3 was Pcdh7 (mean connectivity difference = 0.07, p = 6.8244E-08), which has been implicated in neuronal survival (Xiao et al., 2018).In the 6-magenta module (Fig. 5C), the connectivity of the main hub gene Fbxo34 was dramatically reduced upon NEIL3 depletion (mean connectivity difference = 0.24, p = 1.7852E-32).Fbxo34 belongs to the F-box family proteins, known as subunits of the protein-ubiquitin ligase E3 complex (Randle and Laman, 2016).Protein ubiquitination plays an important role in regulating synapse development, formation, maturation, and plasticity in the brain (Jarome and Helmstetter, 2013;Yamada et al., 2013).These results suggest that NEIL3 influences specific gene co-expression networks by targeting distinct hub genes in the immature hippocampal subregions.In the 2-yellow, 4-green, 7-pink, and 10-greenyellow modules, NEIL3 targeted specific hub genes (e.g., Syt3, MPP5, Bag5, and Sgk1) that displayed reduced connectivity in adult hippocampal subregions (Fig. 6).SYT3 has been implicated to play a role in dendrite extension, dopamine secretion, and synaptic vesicle-mediated transport (Awasthi et al., 2019;Lu et al., 2023;Weingarten et al., 2022), MPP5 in neuronal progenitor cell survival, cell polarity formation, and cerebral histogenesis (Kim et al., 2010;Ozcelik et al., 2010), BAG5 in protein metabolic processes (Gupta et al., 2022), and SGK1 in neuronal excitability and renal sodium excretion (Lang et al., 2010).These results suggest that NEIL3-affected gene co-expression networks may influence the function of the mature hippocampus by modulating processes of synaptic communication and plasticity.Of note, most of these hub genes did not show differential expression in Neil3 -/-hippocampus, suggesting that the disruption of gene co-expression networks is driven by a synergistic effect of differentially connected genes to the hubs.This observation J o u r n a l P r e -p r o o f underscores the complexity of gene network dynamics in the hippocampus and highlights the need for comprehensive research to delineate the specific roles and mechanistic contributions of NEIL3.Such investigations are particularly critical for understanding the broader implications of NEIL3 function in hippocampal integrity and addressing its potential links to neuropathological states stemming from hippocampal malfunction.
Gene co-expression networks and hippocampal neural networks are interconnected frameworks critical to understanding hippocampal functions.Gene co-expression networks are composed of nodes representing individual genes, with edges denoting significant co-expression correlations (Zhang and Horvath, 2005).These correlations may be quantified, with edge weights reflecting the strength of gene co-expression signals.Genes clustered within the same network module often share functional relationships or contribute to similar biological processes (van Dam et al., 2012).In the hippocampal neural networks, nodes symbolize neurons, and edges signify synaptic connections, which are fundamental for neural computations and are the basis for memory encoding and spatial processing (Strange et al., 2014).Crucially, the hippocampal network exhibits functional distinctions between its immature and mature states, reflecting the dynamic gene expression alterations during developmental phases.Dissecting the gene co-expression network in distinct hippocampal subregions with molecular intervention (e.g., NEIL3 depletion) in neonatal and adult mice provides new insights into the functional regulation of the hippocampal network that supports cognitive processes throughout an individual's life.

Conclusions
In summary, our study revealed intricate gene network structures underlying hippocampal maturation, identified modules of co-expressed genes, and pinpointed highly interconnected hub genes specific to the maturity of hippocampal subregions.Through our focused investigation of gene co-expression networks and the molecular depletion of NEIL3, we elucidated NEIL3associated molecular determinants underlying the maturation of hippocampal subregions.This study enhances our understanding of the hippocampal network's complexity and underscores NEIL3's role in hippocampal development.Our work with neonatal and adult mouse models provides a foundation for continued research into how genetic factors influence hippocampal function during development.

Declaration of interests
☒ The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
J o u r n a l P r e -p r o o f

Figure 2 .Figure 3 .
Figure 2. Gene co-expression networks in wildtype and Neil3 -/-hippocampus (A) The dendrogram illustrates the average linkage of hierarchically clustered genes from the wildtype hippocampus based on the topological overlap (TO).The red line represents the cut-off (0.73) to define modules.Modules are assigned colors and numbers as shown in the horizontal bar under the dendrogram.(B) The gene co-expression network in the wildtype hippocampus is represented by classical multi-dimensional scaling plots.Each gene is represented by a dot, where the color of the dot corresponds to the gene module to which that gene belongs.(C) The dendrogram illustrates the average linkage of hierarchically clustered genes from the Neil3 -/-hippocampus based on TO.Genes in Neil3 -/-mice are depicted using the same module color as defined in the wild type to show the extent of module similarity.(D) The gene co-expression network in the Neil3 -/-hippocampus is represented by classical multi-dimensional scaling plots.J o u r n a l P r e -p r o o f

Figure 4 .
Figure 4. Altered module connectivity in the Neil3 -/-hippocampus.A. Scatter plots represent Spearman's correlation (rho) between wildtype and Neil3 -/-gene co-expression networks in each module.A rho value deviated from 1 suggests a reduced correlation between the wildtype and Neil3 -/-gene co-expression networks.The red line represents the regression line, quantifying the disruption severity in the network connectivity post-NEIL3 depletion.B.A table summarizes the number and the percentage (in brackets) of differentially expressed module genes in NEIL3deficient hippocampal subregions at immature (p8) and mature (3m) stages.The table columns represent gene modules categorized by color, and each cell indicates the number of differentially expressed genes in a specific module for a given developmental stage/region.The included percentages offer insights into the proportion of genes affected by Neil3 knockout relative to the total number of genes in each module for the wildtype.J o u r n a l P r e -p r o o f

Figure 6 .
Figure 6.NEIL3 depletion influenced gene co-expression networks in the mature (3m) hippocampal subregions.SimRatio in the 7-pink module (A), the 10-greenyellow module (B), the 4-green module (C), and the 2-yellow module (D) is represented with colors showing the ratio of the topological overlap (representing connectivity strength) of a given gene pair in wildtype to the gene's pair topological overlap in Neil3 -/-mice.The y-axis shows the top 10 hub genes, and the xaxis shows genes with connectivity affected in Neil3 -/-mice.The expression counts of hub gene Bag5 in the 7-pink module, Sgk1 in the 10-greenyellow module, Mpp5 in the 4-green module, and Syt3 in the 2-yellow module are represented in dot plots (blue dots for wildtype samples, and the grey triangles for Neil3 -/-samples).J o u r n a l P r e -p r o o f

Figure 7 .
Figure 7. NEIL3 depletion influenced the gene co-expression networks in the 11-turquoise module A. SimRatio in the 11-turquoise module is represented with colors showing the ratio of topological overlap (representing connectivity strength) of a given gene pair in wildtype to the gene's pair topological overlap in Neil3 -/-mice.B. Heatmap represents expression levels (normalized using z-score by row) of top hub genes in the 11-turquoise module.C. Heatmap represents expression levels (normalized using z-score by row) of genes differentially connected to hub genes.Genes with different expression levels at the immature stage were marked in red and genes with different expression levels at the mature stage were marked in blue.