Human hippocampus niche and circuit neurotypical and major depression profiles

Human hippocampus niche and circuit neurotypical and major depression profiles | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Human hippocampus niche and circuit neurotypical and major depression profiles Madeline B. Mariani, Cheick A. Sissoko, Alexandra M Wamalwa, Anthony Ramnauth, and 15 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8593764/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Hippocampal neuroplasticity regulates memory and emotional responses, which are altered in major depression, a leading cause of disability worldwide 1 . Molecular underpinnings of reduced hippocampal neuroplasticity 2 , fewer neurons, and smaller volume 3,4 in MDD are unknown and may involve blunted adult hippocampal neurogenesis (AHN), a debated phenomenon in human brain 5–8 . Dynamic changes in gene expression (GEX) are mediated, in part, by cis-regulatory elements (CREs), in response to developmental cues, environmental signals, and cellular contexts. However, the accessibility of CREs and their relation to GEX in specific hippocampus cell types in MDD is unknown. To this end, we simultaneously profiled chromatin accessibility and GEX in 349,847 human hippocampal nuclei, used machine learning 9,10 , RNA Velocity 11 and pseudotime trajectory inference 12 to investigate the presence of progenitor cells and their progeny, and spatial transcriptomics in intact hippocampus sections to sequence cytoplasm and neuronal processes providing anatomical transcriptome mapping. We identified Type I, II and III neural progenitors and immature granule neurons. We find hyperactivated inflammation and apoptosis in immature and mature granule neurons, possibly affecting AHN and cell survival in MDD. Proteomics and transcriptomics convergent dysregulations in MDD, point to lipid metabolism, neurogenesis, synaptogenesis, cell adhesion, and plasticity. Identified molecular markers may drive biomarker and drug discovery for MDD and hippocampus resilience. Cellular & Molecular Neuroscience Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Main Major depressive disorder (MDD) is recurrent and episodic, often treatment resistant 2 , a major cause of disability worldwide, and the most frequent diagnosis in those who die by suicide 1 . Human brain postmortem studies found smaller hippocampus dentate gyrus (DG) and fewer granule neurons (GNs) 3,13 in MDD, which may reflect impaired neuroplasticity resulting in MDD symptoms like impaired emotional dysregulation and memory 2 . Moreover, hippocampus neuronal and neuropil loss were correlated with MDD severity and illness duration 3,13 . In vivo brain imaging confirmed smaller hippocampal volume in MDD compared to controls 14 . It remains unknown whether smaller hippocampal volume and diminished neuroplasticity in MDD are the result of increased apoptosis, decreased AHN, or other cellular and molecular changes. Deficits in AHN have been linked to depressive-like behavior in rodents 15 . Conversely, adult born GNs support stress recovery and emotional regulation 16 , enhance hippocampal plasticity, and circuit rewiring 17 . Transcriptomics studies agreed that rodent markers are inadequate to identify progenitor cells and their trajectory in humans and non-human primates 5,7,8 . In humans, early studies showed proliferating cells becoming GNs 18 and estimated neuron birth date based on 14 C decay indicated substantial AHN in people who had been subjected to radioactive exposure 19 . Nevertheless, human AHN remains debated 20,21 , as immunofluorescence (IF), immunohistochemistry (IHC) 22-24 , in situ hybridization (ISH) 23,25 , and transcriptome sequencing studies 5-8 led to opposite conclusions, likely driven by tissue processing methods, subjects’ medical history, mode of death, and psychotropic drugs exposure 26 . To resolve this debate, provide new insights into the molecular characteristics of human hippocampus cell types, and discover the molecular basis of cellular disruption in MDD, we assessed 106 subjects and for the first time profiled almost 350,000 cells in untreated MDD subjects (MDDs) and neurotypical controls (CTRL) with multi-omic single-nucleus RNA sequencing (snRNA-seq) and assay for transposase-accessible chromatin with sequencing (snATAC-seq), spatial RNA-seq, and extensive molecular marker validation (Fig. 1a). Here we present results from the largest and most deeply sequenced single-nucleus multiome and spatial transcriptome resource dataset of neurotypical and MDD human hippocampus cell types to date 27 . Single-nucleus multiome and spatial transcriptome profiling of adult human hippocampus We simultaneously generated snRNA-seq and snATAC-seq data on 349,847 nuclei from deeply clinically phenotyped sudden death subjects with high-quality brain tissue (52 samples from 18 subjects, 9 male MDDs and 9 male CTRL, with 53% of nuclei from MDDs; Fig. 1a-d, Supplementary Tables 1-3; Methods). Number of samples per donor, cells profiled per donor, RNA integrity number, post-mortem interval, and age did not differ between MDDs and CTRL (Extended Data Fig. 1a-e). The medium number of unique molecular indices, chromatin regions, and genes per donor did not differ between MDDs and CTRL (Extended Data Fig. 1f-i). We used Harmony 28 to correct for sample, donor, and batch in RNA and ATAC data separately (Extended Data Fig. 2). To integrate snRNA-seq and snATAC-seq data, we used weighted nearest neighbor (WNN) analysis 29 (Fig. 1b-d), an unsupervised method in Seurat (v5) to obtain a unified representation of snRNA-seq and snATAC-seq data. We identified 31 unsupervised clusters, representing 12 broader cell types (Fig. 1b-c), that were not driven by batch or donor effects (Fig. 1b-d; Extended Data Fig. 3a-d). In each cluster and cell type, average GEX and predicted GEX inferred by accessible chromatin (or gene activity) data were strongly correlated with each other (Fig. 1e; Extended Data Fig. 4e) and used to annotate cell types based on their top-expressed genes (Fig. 1f-g; Extended Data Fig. 4c-d; Methods; Supplementary Table 4). Clusters included oligodendrocyte progenitor cells (OPC), oligodendrocytes (Olig), endothelial cells (Endo), pericytes (Peri), vascular leptomeningeal cells (VLMC), microglia (Micro), macrophages (Macro), T-cells (Tcell), choroid plexus cells (Ch.Pl), subventricular zone (SVZ) ependymal cells (Epe), Cajal Retzius cells (CR), three astrocyte (Astro1-3), nine inhibitory neuron (InN.SST, InN.LHX6, InN.PVALB, InN.VIP, InN.LAMP5, InN.SLC17A8, InN.PENK, InN.TUBB3, InN.PROX1), six non-granule excitatory neuron (ExN1-6), and two GN (GN1, GN2) clusters (Fig. 1d, Extended Data Fig. 4a-d). These 31 cluster annotations aligned with those from recent human hippocampus snRNA-seq studies that did not include ATAC sequencing 6-8 (Extended Data Fig. 5; Supplementary Table 5). The 12 broader cell types included: OPC, Olig, Ch.Pl, Epe, and CR (identical to the clusters), and merged clusters of vasculature cells (Vasc; including Endo, Peri, and VLMC), immune system cells (Immune; including Micro, Macro, and Tcell), astrocytes (Astro; including all three astrocyte clusters), granule neurons (GN, including both GN clusters), inhibitory neurons (InN, including all nine InN clusters), CA excitatory neurons (CA.ExN, including ExN1, ExN 3) and other excitatory neurons (O.ExN, including ExN2, ExN4-6 that did not map on CA regions, Fig. 2e-f; Supplementary Table 6). Spatial GEX is concordant with single-nucleus cell type annotations To investigate GEX in whole tissue, including cytoplasm and neuronal processes, and provide anatomical mapping of GEX onto hippocampal subfields, we used Visium (10X Genomics) on intact hippocampus proper tissue sections (6 subjects, 2 samples each; 3 male MDDs and 3 male CTRL; Supplementary Table 7). First, we used histology (Fig. 2a) to guide hippocampus manual anatomical annotation into granule cell layer (GCL), subgranular zone (SGZ), molecular layer (ML), CA1-4, and white matter (WM, Fig. 2b). Then we performed unsupervised clustering of Visium spots (55x55 µm in size) which provided a more detailed spot-type annotation than manual annotation of hippocampal subfields and delivered 13 clusters that expressed genes concordant with expected cell types based on spot location in the respective subfields (Fig. 2c-d; Supplementary Table 8). Unsupervised spot clusters included (Methods): 2 hilus clusters (hilus1, hilus2), CA3 stratum pyramidales (ca3), stratum oriens-stratum lucidum (so-slu), so-slu of the CA3 (ca3-so-slu), stratum radiatum (sr), stratum oriens-stratum lacunosum (so-sla), stratum moleculare (ml), stratum granulosum (gcl), SGZ-polymorphic layer (sgz-pl), stratum moleculare (sgz-ml). We further used Visium for validation of single-nucleus cell type annotations, performing anchor-based integration of sn-multiome-based cell types (reference) with whole-transcriptome spatial GEX (query), and found that Olig and Astro cell types mapped in WM, GN in GCL (Fig. 1h), and CA.ExN in CA regions, while O.ExN had low prediction score in CA regions, and vasculature was spread across the tissue (Fig. 2e). Different ExN clusters mapped on specific subfields, with ExN1 mapping on CA1, ExN3 on CA3, while Ch.Pl and Epe correctly mapped on choroid plexus and SVZ (Extended Data Fig. 4f). We then used Xenium multiplex ISH (10X Genomics, 266 probes, 10.45x22.45 mm capture area) for additional anchor-based integration of single-nuclei cell types. While CA.ExN mapped on CA1-4, O.ExN mapped on the subiculum (Fig. 2e-f), which was not possible to verify with Visium because the smaller capture area could not include the subiculum. Xenium confirmed the expected location of Astro, Olig, GN (Fig. 1i), and Vasc (Fig. 2f). Spatial RNA Velocity identifies a cell trajectory from the SGZ to the GCL To infer cell trajectories based on proportions of spliced and unspliced RNAs, we applied RNA Velocity analysis to spatial RNA-seq rather than to snRNA-seq data because the method relies on both cytoplasmic and nuclear RNAs to infer transcription and degradation rates 11 . Since we hypothesized that, if a neurogenic trajectory exists, it would be in the DG, to infer RNA Velocity, we selected spots located in the SGZ, GCL, and ML (Fig. 2g; Extended Data Fig. 6; Methods). The partition-based graph abstraction (PAGA) analysis identified a lineage from spatial cluster sgz.ml to cluster sgz.pl and then to gcl (Fig. 2h), indicating that sgz.ml could have the most immature cells with cell bodies in the SGZ and apical dendrites in ML, likely radial-glia-like quiescent neural stem cells 30 . The next cluster on the trajectory was sgz.pl likely containing intermediate progenitors located in SGZ. Ultimately the gcl spatial cluster would comprise GN at different dynamic functional states. Latent time (Fig. 2i) and pseudotime calculated from RNA Velocity (Fig. 2j) agreed with time zero corresponding to cluster sgz.ml and increasing time latency going towards clusters sgz.ml and then gcl. GEX-by-latent-time analysis showed high expression of PCNA , HESX1 , MKI67 , and ENOX1 at the earliest stage (sgz.ml spatial cluster), DCX , CALB2 , SOX11 , GAD1 , GAD2 , and STMN2 at the intermediate stage (sgz.pl spatial cluster), and CAMK2A , NEUROG3 , CAMK2A , SNAP25 , PPFIA2 , and RBFOX3 at the final stage of maturation (gcl spatial cluster, Fig. 2k). Supervised machine learning characterization of neurogenic cell trajectory To determine the presence of a human hippocampus neurogenic niche at single-nucleus resolution, we used a supervised machine learning approach. Supervised models have previously been used to distinguish transcriptionally ambiguous cell subtypes 9,10 , including those of the neurogenesis trajectory 8 . We scored all 349,847 high quality nuclei using normalized counts of genes known to be expressed in mammalian hippocampus quiescent neural stem cells (Type I), amplifying neural progenitors (Type II), intermediate neural progenitors (Type III), and immature GN (ImGN; Fig. 3a) as per rodent and human literature 5,8,20,30,31 . A percentage of high confidence cells were selected (Methods). High confidence cells with high expression of required marker genes and low expression of genes typically not expressed in that cell type 5,7,8,20,30 were used to train the models: for Type I we used GFAP , SOX2 , PAX6 , NES , FABP7 , HES5 , and ETNPPL ; Type II had ASCL1 , FOXO3 , NR2E1 , EOMES , NEUROG2 , NES , MCM2 , PCNA , MKI67 , TOP2A , and not MBP or MOBP , to distinguish them from oligodendrocyte progenitors; for Type III we used MYT1L , PROX1 , DCX , CALB2 , RELN , ST8SI2 , ST8SIA4 , and not mature neuron marker RBFOX3 , or inhibitory neuron marker VIP ; for ImGN, RBFOX3 , PROX1 , NEUROD1 , DCX , BHLHE22 , COL25A1 , and POSTN . Cells with 85% prediction score or higher were annotated as TypeI (4,413 cells, 7.47% of GN) TypeII (1,680 cells, 2.83% of GN), TypeIII (4,423 cells, 7.49% of GN) and ImGN (5,611 cells, 9.50% of GN) and clustered separately (Fig. 3b). The LASSO approach extracted these cells from the original clusters (Supplementary Table 9). TypeI derived from Astro, as radial-glia-like progenitors are most like astrocytes 30 . TypeII derived mostly from OPC, Astro, and Endo, as those clusters contain proliferating cells. TypeIII derived from InN and CR clusters. CR cells have a prominent role in brain development 32 and express CALB2 and TUBB3 and not SYT1 (Extended Data Fig. 4c-d), not typical of adult excitatory CR. InN clusters expressed immature neuron markers: CALB2 (InN.VIP, InN.PROX1), TUBB3 (InN.TUBB3), RELN (InN.LAMP5), DCX (all InN subtypes), and PROX1 (InN.PROX1, InN.VIP, InN.LAMP5; Fig. 1f-g, Extended Data Fig. 4c,d). Moreover, GAD1 and GAD2 , classically used as InN markers, have transcript variants expressed during brain development until neurons fully differentiate 33 . Using Xenium and RNAscope (ACDbio), we examined the expression of GAD1 in the DG, finding GAD1 co-expressed with PROX1 and DCX in the GCL (Fig. 3h). Finally, ImGN were pulled mostly from GN, in line with a recent publication 8 . To validate spatial location of ImGN, we used anchor-based integration of single-nuclei GEX (reference) and Xenium ISH (query) and found that ImGN are in SGZ and GCL (Fig. 3g). The maturity stages of the machine learning-identified cell types were reflected by gene ontology terms showing TypeI enriched for cell fate commitment, TypeII for stem cell differentiation and neural precursor cell proliferation, TypeIII for neuron migration and axon development, and ImGN for cell growth and learning or memory (Supplementary Table 10) . In a larger sample (Supplementary Table 1), using RNAscope, we validated expression of DCX in human SGZ and GCL as previously shown 21 , and demonstrated DCX co-localizing with NEUROD1, PROX1, TUBB3 and RBFOX3 (Extended Data. Fig. 7a-g). Using IHC, we confirmed expression of Nestin and Ki67 proteins in human SGZ, DCX protein in human SGZ and GCL, and NeuN protein in GCL, as previously reported 18 . Using double IF, we replicated co-localization of DCX with PSA-NCAM 18 , and showed DCX co-labeling with neurofilament, calretinin, calbindin, and NeuN, and no DCX co-localization with glial markers GFAP and IBA1 (Extended Data Fig. 7-8). Single-nucleus pseudotime trajectory inference supports a human neurogenic lineage To test if machine learning-identified neurogenic cell types TypeI, TypeII, TypeIII, and ImGN belonged to a differentiation trajectory, we used Palantir 12 to model differentiation capturing the continuity in cell states and cell fate determination (Fig. 3c-e). We included TypeI, TypeII, TypeIII, ImGN, and the cells remaining in the GN cell type after applying the LASSO model (Fig. 3b). Palantir assigned pseudotime in concordance with our hypothesis and ordered cell types from TypeI, TypeII, TypeIII to ImGN and GN (Fig. 3c). GEX along the neurogenic lineage included PAX6 , ETNPPL , and SOX2 in TypeI quiescent progenitors; MKI67 , TOP2A , PCNA in TypeII proliferating progenitors; DCX , GAD1 , GAD2 , EOMES , TUBB3 , PROX1 , STMN1 in TypeIII neuroblasts and ImGN; and finally, SCL17A7 , CALB1 , CAMK2A , SYT1 , and RBFOX3 in mature GN, also corresponding to open chromatin at those genes (Fig. 3d-f). Moreover, Palantir lineage cell GEX is concordant with progressive GEX in Visium spots ordered by latent time based on RNA Velocity (Fig. 2k), and they both suggest a phase in which GAD1 and GAD2 are expressed in TypeIII. In fact, all InN expressed DCX ; InN.LAMP5 expressed RELN ; InN.VIP and InN.PROX1 expressed CALB2 ; InN.TUBB3 expressed TUBB3 ; and InN.PROX1, InN.VIP, and InN.LAMP5 expressed PROX1 (Fig. 1f; Extended Data Fig. 4c). Cell-type specific gene expression and chromatin accessibility are altered in MDD To understand molecular underpinnings of hippocampus anatomical and functional disruption in MDD 2,3,13,14 , we analyzed cell-type specific differentially expressed genes (DEGs), differentially accessible chromatin regions (DARs), and predicted differentially active genes (DAGs) both in the 31 clusters and the 16 cell types (OPC, Olig, Ch.Pl, Epe, CR, Vasc, Immune, Astro, TypeI, TypeII, TypeIII, ImGN, GN, InN, CA.ExN, O.ExN, Supplementary Tables 11-13). The clusters with the highest number of DEGs were GN2, GN1, ExN2, Epe, OPC, and Oligo, (Fig. 4a-b). GN1 and GN2 had 75 and 43 DEGs, contributing to 13.2% and 7.6% of all DEGs respectively. Epe and ExN2 had 182 DEGs, both contributing to 32% of DEGs. Regarding DARs, ExN1 had the most with 497 hits (87.5% of all), Oligo and GN1 had 39 and 36, contributing to 6.9% and 6.3% of DARs respectively (Fig. 4c-d). As for predicted DAGs, 97.6% were identified in OPCs, and GN1, GN2, and Oligo had 15, 13, and 13 respectively (Fig. 4e-f). When analyzing cell types, GN had 101 DEGs, O.ExN had 38, ImGN had 20, TypeI and TypeIII each had one DEG: LINC02315 (upregulated) and ZNF257 (upregulated) respectively (Fig. 4a-b). DARs were the highest in Immune (14227 DARs), followed by GN (41 DARs; Fig. 4c-d). DAGs were 60 in GN, 41 in O.ExN, 2 in ImGN, and one in TypeI and TypeIII ( ACOT1, upregulated in MDDs, Fig. 4e-f). Taken together, GN had the most consistently dysregulated genes across modalities. PsyGeNET Analysis To explore which of the identified dysregulated genes had previously been associated with MDD, we used PsyGeNET text mining database 34 . In all clusters, we found 43 genes with DEGs and DAGs overlapping with psychiatric disease-associated genes in PsyGeNET, and significant for depressive disorders (p=0.033017458) with 15 overlapping genes ( ADCY8 , ATF3 , DISC1 , EMP1 , ENPEP , FOLH1 , FOS , GRIN2A , GRK5 , IL18 , IL19 , IL6R , SOD2 , SYNDIG1 ), and for alcohol use disorder (pl=0.009390476), with 83 overlapping genes (Fig. 5a-b, Supplementary Table 14). When analyzing cell types, GN had 7 DEGs previously linked to psychiatric disorders, including depression, schizophrenia, and mood disorders ( ADCY8 , CCND2 , DNAH8 , GABBR1 , HLA-DQB1 , PRKCB , THEMIS ; Fig. 5c), without any significant disease-specific associations. Biological pathways of cell-type specific DEGs in MDD To infer pathways associated with cell-type specific multi-omics changes in MDD, we performed pre-ranked gene set enrichment analysis (GSEA, Supplementary Table 15). We focused on GN that had the most dysregulated genes across modalities, and ImGN to understand why GN are fewer in MDD 3,4,13 . In ImGN, pathways related to interferon signaling, cellular stress, Golgi, MTORC1, sex hormone response and apoptosis were enriched, while lipid homeostasis and glycosylation were negatively regulated in MDD (Fig. 5d-e). In GN, pathways related to PTEN, MTORC1, apoptosis, sex hormone and immune response, and thyroid, Parkinson, Huntington and Alzheimer’s diseases were enriched, and lipid homeostasis was negatively regulated in MDD (Fig. 5f-g). Notably, in both ImGN and GN, apoptosis was enriched in MDD with top genes CCND2 , JUN , MADD , BCAP31 , DNAJA1 , NEFH , PEA15 , DAP , ENO2 , BMP2 (Supplementary Table 15). Weighted gene co-expression networks dysregulation in MDD While differential GEX analysis provides information on single-gene expression differences, MDD is highly polygenic 35 and disease state may be driven by gene network changes rather than single-gene alterations 36 . Therefore, we used pseudo-bulk GEX in GN and ImGN to perform weighted-gene co-expression network analysis (WGCNA) 37 and determine clusters of highly correlated genes that have an association with MDD. Each cluster was represented by a color-coded module eigengene (ME), the first principal component of the module representing the overall expression level of the module. In ImGN, six out of seven modules had correlations with MDD (Fig. 5h-j; Extended Data Fig. 9a-d). Four ME were positively associated and two were negatively associated with MDD. MEblue, enriched for RNA metabolism, splicing, translation and response to cellular stress, was positively correlated with MDD (correlation 0.99, p=1× 10 −40 ), in line with GSEA findings described above. The MEturquoise module, enriched for lipid modification and metabolism, was negatively correlated with MDD (correlation -0.99, p=2× 10 −47 ) underscoring GSEA findings of lipid homeostasis being negatively regulated in MDD ImGN and GN (Fig. 5e-g). In GN, 8 modules out of 9 had correlations with MDD (Fig. 5k-m; Extended Data Fig. 9e-h). Three ME were positively associated and five were negatively associated with MDD. The MEturquoise module, enriched for lipid modification and metabolism, was positively correlated with MDD (correlation 0.99, p=2 × 10 −43 ) and showed significant overlap (p=2 × 10 −04 ) with upregulated DEGs in GN, and with GSEA findings (Fig. 5e-g). The MEblue module, enriched for development and cellular morphogenesis, was negatively correlated with MDD (correlation -0.99, p=1× 10 −50 ) and showed significant overlap ( p =7 × 10 −06 ) with downregulated DEGs in GN, and with GSEA findings (Fig. 5e-g). Quantification of target marker RNA and protein expression in MDD In separate and enlarged samples, using unbiased stereology, we quantified number of cells expressing neurogenic lineage RNAs and proteins labeled by multiplex ISH and IF, and single IHC, and used quantitative real-time polymerase chain reaction (qPCR) to quantify RNA in bulk tissue (Methods; Supplementary Table 16, Extended Data Fig. 7-8). RNAscope DCX+/TUBB3+ cells were fewer in MDDs (N=11) vs. CTRL (N=11) in GCL (t=2.294; p=0.04) but not SGZ (Extended Data Fig. 7n). DCX RNA measured by qPCR was lower in MDDs (N=5) vs. CTRL (N=5) in anterior (t=3.371, p=0.0145) but not posterior DG (Extended Data Fig. 7o). Single-IHC Ki67+ cells were fewer in MDDs (N=17) vs. CTRL (N=23) in anterior (t=2.780; p=0.0084) but not mid or posterior DG (Extended Data Fig 8p). MDDs had fewer Nestin+ cells in anterior (34 MDDs, 38 CTRL, t=4.360; p<0.0001) and posterior DG (31 MDD, 28 CTRL, t=3.654; pval=0.0008) but not in mid DG (Extended Data Fig. 7q). Nissl+ glial cells were not different between MDDs and CTRL in any region (Extended Data Fig. 7r). Single-IHC DCX+ cells were fewer in MDDs (N=22) vs. CTRL (N=36) in anterior (t=3.547; p=0.0008), but not in mid or posterior DG (Extended Data Fig. 7s). MDD had fewer single-IHC NeuN+ GNs in anterior (35 MDDs, 32 CTRL, t=2.951; p=0.004) and mid DG (28 MDDs, 26 CTRL, t=2.410; p=0.01), but not in posterior DG (Extended Data Fig. 7t). IF-stained DCX+/NeuN- cells were fewer in MDDs (N=8) vs. CTRL (N=8) in SGZ (welch-corrected t [t]= 2.318; p=0.04) but not GCL, and DCX+/NeuN+ cells did not differ between groups (Extended Data Fig. 8l-o). Proteomics We further used high resolution shotgun proteomics liquid chromatography with tandem mass spectrometry (LC/MS/MS) 38 , to quantify expression of proteins corresponding to DEGs in MDDs vs. CTRL as per snRNA-seq data (Methods; Extended Data Fig. 9i-m; Supplementary Table 17). After running proteomics as we described 38 , Elucidator (Methods) detected 80,504 peptides across 48 runs and identified 1811 proteins with protein score at a 4% false discovery rate (Extended Data Fig. 9i-m; Supplementary Table 17). The 1131 proteins with two or more peptides and protein scores greater than 250 were used for the analyses. We found 308 differentially expressed proteins between MDDs and CTRL, of which four corresponded to detected DEGs: CPNE8, HSDL2, NCH1, and RAC1 (Extended Data Fig. 9n). HSDL2, a key factor of fatty acid regulation in lipid metabolism 39 , was upregulated at both the RNA level in GN and the protein level in bulk tissue. RAC1, important in neurodevelopmental function including neurogenesis, synaptogenesis, and plasticity and expressed in migrating cortical neurons and adult neurons 40 , was lower at the protein level in MDDs vs. CTRL. RAC1 is negatively regulated by ARHGAP15 41 , whose genomic sequence overlaps with long non-coding RNA AC079793.2 , also downregulated in ImGN in MDD. Conversely, CPNE8 regulating cell adhesion 42 showed downregulated RNA expression in MDD GN and upregulated protein expression in bulk MDD tissue. Similarly, NCEH, a protector from α-synuclein neurotoxicity 43 , was downregulated at the RNA level in Epe but the protein was upregulated in bulk hippocampus tissue in MDD. Discussion Employing the first single-nuclei multiome and spatial transcriptome profiling of human anterior hippocampus cells, combined with shotgun proteomics in anterior hippocampus tissue, we compared sudden death neurotypical adult subjects and unmedicated patients with MDD. Implementing a machine learning approach for rare cells identification, we detected quiescent (Type I), amplifying (Type II), and intermediate (Type III) progenitor cells, and replicated the identification of immature neurons 8 . Our integrated analysis identified distinct clusters representing all hippocampal cell types, in agreement with other publications 6–8 , identified a neurogenic lineage from early progenitors to immature granule neurons, and defined cell type-specific gene expression and chromatin accessibility dysregulation in MDD. Importantly, the annotation of these cell types, including those of the neurogenesis trajectory, was validated by spatial GEX profiling using Visium, RNAScope, and Xenium, and by protein marker expression identification by double IF and IHC, providing robust evidence of cell identity and relative cellular and spatial distribution within the hippocampus. Cell trajectory inference using both single-nuclei and spatial transcriptome sequencing, supported the neurogenic cell-lineage trajectory, and gene expression corresponding to maturational stages described in mice 5,7 . Findings suggest transcriptional and epigenetic similarities of astroglia-like (TypeI) progenitors with astrocytes, amplifying progenitors (TypeII) with other dividing cells (OPC, Astro and Endo). These data do not demonstrate when cells of the neurogenic trajectory were generated but the expression of proliferation markers ( PCNA , MKI67 , TOP2A ; Fig. 2k, 3e) in TypeII suggests the presence of mitosis. TypeIII origination from GAD + neurons initially classified as InN, is largely due to DCX expression in GAD + cells, and warrants further investigation on transitional GAD expression in maturing cells. We demonstrated for the first time GAD1 expression in TypeIII, colocalizing with DCX , TUBB3 and PROX1 , and DCX mapping onto SGZ and GCL, in line with findings that GAD1 alternative transcripts are expressed in human hippocampus during neuronal development 33,44 . Additionally, GABAergic basket cells could be generated in human DG as demonstrated in adult rats 45 . ImGN coming from the GN cluster has been previously observed. Notably the number of ImGN is similar to what previously estimated (close to 10% of all GNs), suggesting cells might remain in this immature status for a long time, which has profound implications for the functional significance of AHN in the human brain. The presence of immaturity markers suggests a regenerative potential of the human SGZ. It remains unknown how prolonged the immature state is and how long it takes to make a human neuron. Different sequencing technologies, including epigenome analysis and deeper sequencing, may explain our finding the full neurogenic lineage when others have not 6–8 . This validity of our trajectory reconstruction is strengthened by RNA Velocity and pseudotime analysis inferring of cell trajectory, and its identified molecular signatures that characterize neural stem cell activation, proliferation, differentiation, and ultimately, neuronal maturation, matching those found in the literature 20,26,30 . These findings provide a valuable resource for future studies investigating the regulation of AHN and its potential modulation for therapeutic purposes. In MDD, elevated inflammation, response to cellular stress and apoptosis pathways in ImGN and GN suggests the molecular basis of smaller in vivo hippocampal volume 14 and postmortem findings of fewer GNs 4,13 and agree with finding in MDD PFC 27 and patient blood 46–48 . Overlap between DEGs and PsyGeNET MDD and alcohol use disorder genes, may explain the comorbidity between the two. Upregulation of LINC02315 in TypeI and OPC in MDD could prevent proliferation or transition to the amplifying stage because this long-intergenic non-coding RNA is protective in lung squamous cell carcinoma by blocking proliferation 49 . In TypeIII, InN and Olig, upregulation of ZNF257 which has over 1,000 binding sites in gene promoters and is believed to regulate transcription in the developing brain 50 , may contribute to increased nucleic acid binding, transcription and Vitamin D receptor pathway activity in MDD affecting developmental processes. Dysregulation of MTORC1 in ImGN and GN supports altered cell growth and metabolism 2 . Using both proteomics and transcriptomics, we find genes dysregulated at both levels. Upregulated HSDL2, a key factor of fatty acid regulation in lipid metabolism 39 , at the RNA level in GN and protein level in bulk tissue suggests lipid metabolism dysregulation in MDD. Downregulation of both long non-coding RNA AC079793.2 in ImGN, overlapping in genomic sequence with ARHGAP15 , and possibly negatively regulating it, and its target protein RAC1 in bulk tissue, suggest reduced neurogenesis, synaptogenesis, and plasticity 40 . The fact that CPNE8 , regulating cell adhesion 42 , and NCEH, protecting from α-synuclein neurotoxicity 43 , have downregulated RNA expression in GN and Epe respectively, while the proteins are upregulated in bulk tissue, may be due to tissue-specific transcriptomic differences involving DG and Epe, or compensatory post-transcriptional mechanisms. Although this is the first study combining single-nucleus multi-omic (RNA and ATAC) profiling, with spatial transcriptomic and bulk proteomics in human hippocampus, and comparing sudded death neurotypical controls with unmedicated depressed subjects that had MDD, limitations include small sample, not including women, possible clinical variability, and few subjects having received benzodiazepines prior to death – although we excluded comorbidity and patients on antidepressants or antipsychotics. Results should be replicated in larger samples that include women and low testing effects of clinical and sociodemographic variables, because in animal models stress greatly affects neurogenesis. Overall, this comprehensive single-nuclei multi-omic, spatial transcriptomic, and proteomic profiling of brain tissue supports the presence of a neurogenic niche and suggests molecular underpinnings of anterior hippocampus atrophy in MDD. Dysregulated markers will drive future mechanistic studies investigating their cellular, molecular, and behavioral effects, and will lead blood and brain imaging biomarker discovery and drug development for MDD and hippocampus trophic support. Declarations Data availability The single cell and spatial genomics data will be available at the GEX Omnibus database upon publication. Specimen information and sequencing statistics are described in Supplementary Tables 1-3 and 7. Sources of the published sequencing datasets used in this study are described in Supplementary Table 5. Raw mass spectroscopy data are available at the MassIVE repository entry under dataset identifier MSV000094862 (reviewer password: reviewer052424). The final datasets for this study, which will be archived in our servers, will include demographic data, psychiatric assessment data, psychological ratings, and postmortem studies data. Because the data we collect are sensitive in nature, we carefully ensure that appropriate privacy safeguards are in place. All data will be stripped of identifiers prior to release for sharing. In addition, we will provide relevant protocols and published genetic and phenotypic data upon request. Material transfers will be made with no more restrictive terms than in the Simple Letter Agreement (SLA) or the Uniform Biological Materials Transfer Agreement (UBMTA) and without reach through requirements. Should any intellectual property arise which requires a patent, we will ensure that the technology (materials and data) remains widely available to the research community in accordance with the NIH Principles and Guidelines. Code availability The computational code used in this study is available at GitHub or upon request. Acknowledgements We thank the families for donating the brain tissue used in this study. This work is supported by NIH grants AI164769, AG076949, MH133561, AG080790. We thank Stefano Marenco and the Human Brain Collection Core at the National Institute of Mental Health, and Mark Underwood at the New York State Psychiatric Institute for providing some of the postmortem human brain tissue samples. We thank Erin Bush and Peter Sims and the Columbia Single Cell Analysis Core and JP Sulzberger Columbia Genome Center for their services. We thank Jeremy Worley and the Columbia Genome Engineering Core for Xenium (10X Genomics) experimentation and analysis services. We thank Vinod Gupta and the Columbia Center for Computational Biology and Bioinformatics for services in data storage and high-performance computing infrastructure usage. We thank Tao Su and the Molecular Pathology Shared Resource at Columbia for RNA extraction and quantification services. This study used the Confocal and Specialized Microscopy Shared Resource of the Herbert Irving Comprehensive Cancer Center at Columbia University, funded in part through the NIH/NCI Cancer Center Support Grant P30CA013696. Some schematic illustrations were created or modified using BioRender.com. Contributions MBM and CAS performed sequencing experiments, single-nuclei and spatial library generation, and bioinformatics. AMW performed WGCNA analysis. AR worked on spatial data analysis and data interpretation. YX advised on bioinformatics. ANT, CF, and AMW completed IHC, IF and ISH experiments. YL designed the machine learning approach. WH advised on spatial transcriptomics approaches. YYH worked on nuclei isolation. MBD, GR, JJM, VA and AD provided the human postmortem specimens and clinical characterization of the subjects. STMD and LB worked on proteomics. RH helped with the interpretation of the neurogenesis findings. HG designed the biostatistics approach. MBD secured funding for this project, guided experimental design and approaches, worked on result interpretation and reporting strategy. MBM, and MBD wrote the first draft of the manuscript. All authors reviewed the manuscript. Competing Interests HG and her family own stocks in Illumina, Inc.; JJM receives royalties from Columbia University for the Columbia Pathways App. and from the Research Foundation of Mental Hygiene for commercial use of the C-SSRS. Corresponding author Correspondence to [email protected] References James, S. L. et al. Global, regional, and national incidence, prevalence, and years lived with disability for 354 diseases and injuries for 195 countries and territories, 1990–2017: a systematic analysis for the Global Burden of Disease Study 2017. The Lancet 392 , 1789-1858 (2018). Tartt, A. N., Mariani, M. B., Hen, R., Mann, J. J. & Boldrini, M. Dysregulation of adult hippocampal neuroplasticity in major depression: pathogenesis and therapeutic implications. Mol Psychiatry (2022). https://doi.org:10.1038/s41380-022-01520-y Cobb, J. A. et al. Hippocampal volume and total cell numbers in major depressive disorder. J Psychiatr Res 47 , 299-306 (2013). https://doi.org:10.1016/j.jpsychires.2012.10.020 Boldrini, M. et al. Resilience Is Associated With Larger Dentate Gyrus, While Suicide Decedents With Major Depressive Disorder Have Fewer Granule Neurons. Biological Psychiatry 85 , 850-862 (2019). https://doi.org:https://doi.org/10.1016/j.biopsych.2018.12.022 Wang, W. et al. Transcriptome dynamics of hippocampal neurogenesis in macaques across the lifespan and aged humans. Cell Research 32 , 729-743 (2022). https://doi.org:10.1038/s41422-022-00678-y Ayhan, F. et al. Resolving cellular and molecular diversity along the hippocampal anterior-to-posterior axis in humans. Neuron 109 , 2091-2105.e2096 (2021). https://doi.org:10.1016/j.neuron.2021.05.003 Franjic, D. et al. Transcriptomic taxonomy and neurogenic trajectories of adult human, macaque, and pig hippocampal and entorhinal cells. Neuron 110 , 452-469. e414 (2022). Zhou, Y. et al. Molecular landscapes of human hippocampal immature neurons across lifespan. Nature 607 , 527-533 (2022). https://doi.org:10.1038/s41586-022-04912-w Habib, N. et al. Massively parallel single-nucleus RNA-seq with DroNc-seq. Nature methods 14 , 955-958 (2017). Shekhar, K. et al. Comprehensive Classification of Retinal Bipolar Neurons by Single-Cell Transcriptomics. Cell 166 , 1308-1323 e1330 (2016). https://doi.org:10.1016/j.cell.2016.07.054 La Manno, G. et al. RNA velocity of single cells. Nature 560 , 494-498 (2018). https://doi.org:10.1038/s41586-018-0414-6 Setty, M. et al. Characterization of cell fate probabilities in single-cell data with Palantir. Nat Biotechnol 37 , 451-460 (2019). https://doi.org:10.1038/s41587-019-0068-4 Boldrini, M. et al. Hippocampal Granule Neuron Number and Dentate Gyrus Volume in Antidepressant-Treated and Untreated Major Depression. Neuropsychopharmacology 38 , 1068-1077 (2013). https://doi.org:10.1038/npp.2013.5 Nolan, M. et al. Hippocampal and Amygdalar Volume Changes in Major Depressive Disorder: A Targeted Review and Focus on Stress. Chronic Stress 4 , 2470547020944553 (2020). https://doi.org:10.1177/2470547020944553 Surget, A. et al. Antidepressants recruit new neurons to improve stress response regulation. Mol.Psychiatry (2011). Anacker, C. et al. Hippocampal neurogenesis confers stress resilience by inhibiting the ventral dentate gyrus. Nature 559 (2018). https://doi.org:https://doi.org/10.1038/s41586-018-0262-4 Temprana, S. G. et al. Delayed coupling to feedback inhibition during a critical period for the integration of adult-born granule cells. Neuron 85 , 116-130 (2015). https://doi.org:10.1016/j.neuron.2014.11.023 Eriksson, P. S. et al. Neurogenesis in the adult human hippocampus. Nat.Med. 4 , 1313-1317 (1998). Spalding, K. L. et al. Dynamics of hippocampal neurogenesis in adult humans. Cell 153 , 1219-1227 (2013). https://doi.org:10.1016/j.cell.2013.05.002 Gage, F. H. Adult neurogenesis in mammals. Science 364 , 827-828 (2019). Tosoni, G. et al. Mapping human adult hippocampal neurogenesis with single-cell transcriptomics: Reconciling controversy or fueling the debate? Neuron 111 , 1714-1731. e1713 (2023). Boldrini, M. et al. Human hippocampal neurogenesis persists throughout aging. Cell stem cell 22 , 589-599. e585 (2018). Sorrells, S. F. et al. Positive controls in adults and children support that very few, if any, new neurons are born in the adult human hippocampus. Journal of Neuroscience 41 , 2554-2565 (2021). Moreno-Jiménez, E. P., Terreros-Roncal, J., Flor-García, M., Rábano, A. & Llorens-Martín, M. Evidences for adult hippocampal neurogenesis in humans. Journal of Neuroscience 41 , 2541-2553 (2021). Tartt, A. N. et al. Considerations for Assessing the Extent of Hippocampal Neurogenesis in the Adult and Aging Human Brain. Cell Stem Cell 23 , 782-783 (2018). https://doi.org:10.1016/j.stem.2018.10.025 Kempermann, G. et al. Human adult neurogenesis: evidence and remaining questions. Cell stem cell 23 , 25-30 (2018). Maitra, M. et al. Cell type specific transcriptomic differences in depression show similar patterns between males and females but implicate distinct cell types and genes. Nat Commun 14 , 2912 (2023). https://doi.org:10.1038/s41467-023-38530-5 Korsunsky, I. et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nature methods 16 , 1289-1296 (2019). Hao, Y. et al. Integrated analysis of multimodal single-cell data. Cell 184 , 3573-3587.e3529 (2021). https://doi.org:https://doi.org/10.1016/j.cell.2021.04.048 Ming, G.-l. & Song, H. Adult neurogenesis in the mammalian brain: significant answers and significant questions. Neuron 70 , 687-702 (2011). Abbott, L. C. & Nigussie, F. Adult neurogenesis in the mammalian dentate gyrus. Anatomia, Histologia, Embryologia 49 , 3-16 (2020). https://doi.org:https://doi.org/10.1111/ahe.12496 Gil-Sanz, C. et al. Cajal-Retzius cells instruct neuronal migration by coincidence signaling between secreted and contact-dependent guidance cues. Neuron 79 , 461-477 (2013). https://doi.org:10.1016/j.neuron.2013.06.040 Tao, R. et al. GAD1 alternative transcripts and DNA methylation in human prefrontal cortex and hippocampus in brain development, schizophrenia. Mol Psychiatry 23 , 1496-1505 (2018). https://doi.org:10.1038/mp.2017.105 Gutiérrez-Sacristán, A. et al. Text mining and expert curation to develop a database on psychiatric diseases and their genes. Database (Oxford) 2017 (2017). https://doi.org:10.1093/database/bax043 Kendall, K. M. et al. The genetic basis of major depression. Psychol Med 51 , 2217-2230 (2021). https://doi.org:10.1017/s0033291721000441 Perez, J. M. et al. Hippocampal subfield transcriptome analysis in schizophrenia psychosis. Molecular Psychiatry 26 , 2577-2589 (2021). https://doi.org:10.1038/s41380-020-0696-6 Li, J. et al. Application of Weighted Gene Co-expression Network Analysis for Data from Paired Design. Scientific reports 8 , 622 (2018). https://doi.org:10.1038/s41598-017-18705-z Werth, E. G., Rajbhandari, P., Stockwell, B. R. & Brown, L. M. Time Course of Changes in Sorafenib-Treated Hepatocellular Carcinoma Cells Suggests Involvement of Phospho-Regulated Signaling in Ferroptosis Induction. Proteomics 20 , e2000006 (2020). https://doi.org:10.1002/pmic.202000006 Yang, Y. et al. Lipid metabolism regulator human hydroxysteroid dehydrogenase-like 2 (HSDL2) modulates cervical cancer cell proliferation and metastasis. J Cell Mol Med 25 , 4846-4859 (2021). https://doi.org:10.1111/jcmm.16461 Liaci, C. et al. Loss of ARHGAP15 affects the directional control of migrating interneurons in the embryonic cortex and increases susceptibility to epilepsy. Front Cell Dev Biol 10 , 875468 (2022). https://doi.org:10.3389/fcell.2022.875468 Zamboni, V. et al. Disruption of ArhGAP15 results in hyperactive Rac1, affects the architecture and function of hippocampal inhibitory neurons and causes cognitive deficits. Scientific Reports 6 , 34877 (2016). https://doi.org:10.1038/srep34877 Zhang, P. et al. CPNE8 Promotes Gastric Cancer Metastasis by Modulating Focal Adhesion Pathway and Tumor Microenvironment. Int J Biol Sci 18 , 4932-4949 (2022). https://doi.org:10.7150/ijbs.76425 Zhang, S., Glukhova, S. A., Caldwell, K. A. & Caldwell, G. A. NCEH-1 modulates cholesterol metabolism and protects against α-synuclein toxicity in a C. elegans model of Parkinson's disease. Hum Mol Genet 26 , 3823-3836 (2017). https://doi.org:10.1093/hmg/ddx269 Ramnauth, A. D. et al. Spatially-resolved transcriptomics of human dentate gyrus across postnatal lifespan reveals heterogeneity in markers for proliferation, extracellular matrix, and neuroinflammation. bioRxiv (2023). https://doi.org:10.1101/2023.11.20.567883 Liu, S. et al. Generation of functional inhibitory neurons in the adult rat hippocampus. Journal of Neuroscience 23 , 732-736 (2003). Felger, J. C. et al. What does plasma CRP tell us about peripheral and central inflammation in depression? Molecular Psychiatry 25 , 1301-1311 (2020). https://doi.org:10.1038/s41380-018-0096-3 Enache, D., Pariante, C. M. & Mondelli, V. Markers of central inflammation in major depressive disorder: A systematic review and meta-analysis of studies examining cerebrospinal fluid, positron emission tomography and post-mortem brain tissue. Brain, Behavior, and Immunity 81 , 24-40 (2019). https://doi.org:https://doi.org/10.1016/j.bbi.2019.06.015 Zhan, Y. et al. Alterations of multiple peripheral inflammatory cytokine levels after repeated ketamine infusions in major depressive disorder. Translational Psychiatry 10 , 246 (2020). https://doi.org:10.1038/s41398-020-00933-z Liu, J., Yao, Y., Hu, Z., Zhou, H. & Zhong, M. Transcriptional profiling of long‐intergenic noncoding RNAs in lung squamous cell carcinoma and its value in diagnosis and prognosis. Molecular genetics & genomic medicine 7 , e994 (2019). Farmiloe, G., Lodewijk, G. A., Robben, S. F., van Bree, E. J. & Jacobs, F. M. J. Widespread correlation of KRAB zinc finger protein binding with brain-developmental gene expression patterns. Philos Trans R Soc Lond B Biol Sci 375 , 20190333 (2020). https://doi.org:10.1098/rstb.2019.0333 Methods Brain Collection Brain tissue was obtained from the Brain Collection of the New York State Psychiatric Institute (NYSPI) at Columbia University, which includes brain samples from the Republic of Macedonia 1 , and the Human Brain Collection Core (HBCC) of the National Institute of Mental Health (NIMH). Brain tissue collection was conducted with IRB approval and informed consent obtained from next of kin who agreed to donate the brains and participate in psychological autopsy interviews 2 which provide diagnoses according to the Diagnostic Statistical Manual of Mental Disorders (DSM) through structured proxy-based interviews comparable in reliability to the Structured Clinical Interview for DSM diagnoses (SCID) in living patients, and has been extensively used in postmortem studies 3-7 . Subjects We included 106 subjects with sudden death, no neuropathology including microvascular disease, short agonal state, no resuscitation with prolonged (>10 min) hypoxia, no alcohol or drug use disorder, or chronic illness that may affect CNS function (e.g. epilepsy, renal failure, metastatic malignancy, HIV, Supplementary Table 1). All 51 subjects with Major Depressive Disorder (MDD) included in this study had a diagnosis of MDD, a major depressive episode within 4 months of death, were not in remission at death, had no history of treatment during the three months before death, and had negative toxicology for psychoactive drugs, except for 15 subjects with positive toxicology for benzodiazepines that could have been given at the time of death in the emergency room, or previously prescribed. All 55 controls included in this study had no neurological or psychiatric diagnoses, no pathological or traumatic condition affecting the brain, clear toxicology reports, and had no history of psychiatric treatments or suicide attempts. Brain tissue processing Right hemispheres were sliced at 2-cm intervals, flash frozen and stored at -80ºC at the time of death. Routine neuropathological examination and toxicology exams were performed. If the subject was 45 years old or older, hippocampal formation and neocortex are stained for senile plaques and neurofibrillary tangles. The whole hippocampal formation was dissected from two or three consecutive 2 cm-thick frozen coronal blocks. Frozen adjacent sections of 50 m thickness were collected in tubes for snRNA/ATAC-seq, and two adjacent sections of 14 m were mounted on barcoded slides for Visium spatial transcriptomics (10X Genomics) per donor. Frozen tissue sections (16 m) were also used for qPCR, RNAscope® Duplex, Multiplex and Hiplex in situ hybridization (ACDbio), and for Western Blots (WB). Tissue fixed in 4% paraformaldehyde (length of tissue fixation was 5’ per each mm of block thickness) was used for immunohistochemistry (IHC) and immunofluorescence (IF). Supplementary Table 1 describes experiments performed in each sample. Nuclei isolation We adapted methods from Maitra et al. 2021 8 for isolating high-quality nuclei from human frozen brain tissue. Using a glass douce tissue homogenizer (Wheaton, 357542), approximately 90 mg of tissue were homogenized in 5mL lysis buffer (10mM tris-HCl pH 7.4, 10mM NaCl, 3mM MgCl 2 6H 2 O, 0.05% NP-40) and incubated on ice for 5 minutes. Then, 5mL nuclei wash buffer (NWB: 5% BSA, 0.25% glycerol, 0.001% protector RNase inhibitor (MilliporeSigma, cat. no. 333539901) in 1x PBS) were added to sample to quench the lysis. Homogenate was filtered through a 40 m Corning cell strainer (Millipore Sigma, cat. no. CLS431750). Filtrate was centrifuged three times at 500g for 5 minutes at 4°C. After each incubation, pellet was resuspended in NWB. Final resuspension was ultracentrifuged at 10,000g for 30 minutes at 4°C over iodixanol cushion (Millipore Sigma, cat. no. D1556-250). Nuclei pellet was resuspended in less than 100 L nuclei wash buffer (10X Genomics). Nuclei concentration and cell viability were estimated using trypan blue and DAPI in Countess II Hemocytometer (Applied Biosystems). Nuclei concentrations varied between 2-9 10 6 nuclei/mL. Library preparation and sequencing Approximately 16,000 nuclei for each sample were loaded into the 10X Chromium Controller immediately after quantification. Gene expression (GEX) and Assay for Transposase assessable chromatin (ATAC) libraries were prepared according to Chromium Next GEM Single Cell Multiome ATAC + GEX (CG000338) User Guide. We generated 2-4 libraries per sample. The libraries were sequenced on an Illumina NovaSeq 6000 using v4 chemistry to an average sequencing depth of 80,611 reads per cell for RNA libraries and 37,119 reads per cell for ATAC libraries (Supplementary Tables 3 and 7). Pre-processing and filtering multiome data A count matrix and a fragment file were generated by 10X CellRanger (v2.0.0) pipeline with the chemistry parameter set to Single Cell Multiome ATAC + GEX v1 (Supplementary Table 3) . The reads were aligned to the GRCh38 human genome. RNA and ATAC data for each sample were preprocessed in parallelly using Seurat 9,10 (V5) and Signac 11 (V1)R packages. The count matrix was filtered for cells that have greater than 500 total number of unique transcripts and 200 genes expressed detected to remove dead or dying cells, less than 10% mitochondrial DNA, less than 150,000 UMIs, and less than 20,000 genes expressed to remove multiplets, an artifact that occurs when two or more cells are encapsulated in the same GEM (Gel beads-in-EMulsion) and are tagged with the same barcode (Extended Data Fig. 3e). Multiplets were then removed by considering canonical markers of expected cell types. Clusters that expressed multiple markers of more than one cell type will be removed. Remaining cells were pre-processed again using the standard Seurat workflow. For all samples, peaks BED (Browser Extensible Data) files output from CellRanger were converted to GenomicRanges 12 , an R/Bioconductor package for manipulating genomic intervals and combined to create a unified peak set representing all samples in the dataset. Peaks longer than 10,000 and shorter than 20 bases were removed. Barcodes for cells that were previously identified to be multiplets, dying, or quality were removed. Cells with transcription start site (TSS) enrichment score less than 1, nucleosome signal greater than 3, ATAC read counts less than 200 and greater than 100,000 were removed (Extended Data Fig. 3e). Cells that passed quality control (QC) in both modalities were retained for further analysis. Multi-modal integration In RNA and ATAC modalities, 52 samples were merged. Genes and peaks expressed in less than 0.01% of cells that passed QC were removed. For chromatin accessibility data, we followed the Signac workflow for merging multiple objects. Peaks on nonstandard chromosomes and in blacklist regions from the hg38 genome were removed. Merged gene expression data were processed as follows: to account for different number of counts per gene in each cell, the gene counts for each cell were divided by the sum of counts for that cell, multiplied by 10,000 and natural log transformed using Seurat function NormalizeData with default parameters. The intersection of the top 2,000 variable genes and a list of 93 genes involved in neurogenesis and neural maturation (sum: 2,065 genes) was passed to Seurat function ScaleData for linear transformation, to center mean expression across cells for each gene to zero, and scale expression of each gene, so that highly expressed genes were not given more weight in downstream analysis. Principal component analysis (PCA) was run (RunPCA) with features set to these 2,065 genes, to prevent overfitting. Otherwise, default parameters were used. Fifty PCA cell embeddings were passed to Seurat’s wrapper for Harmony 13 with group.by.vars set to sample, donor, and batch. The Harmony-corrected PCA embedding was used for UMAP (Uniform Manifold Approximation and Projection) and Louvain clustering (Extended Data Fig. 2). Similarly, accessible chromatin data are processed using Signac workflow. We used latent semantic indexing (LSI) first by term frequency-inverse document frequency normalization (TF-IDF) and then singular value decomposition (SVD) dimension reduction with Signac functions RunTFIDF and RunSVD . TF-IDF accounts for differences in sequencing depth between cells and gives higher values to rarer peaks, so they are not overlooked in downstream analysis. All features in the TF-IDF matrix are passed to SVD. A total of 50 LSI components, excluding the first, were passed to RunHarmony with group.by.vars parameter set to sample, donor, and batch. The Harmony-corrected PCA and LSI reductions were passed to Seurat FindMultiModalNeighbors function for Weighted Nearest Neighbor (WNN) analysis 9 . The weighted nearest neighbor and weighted shared nearest neighbor graphs were passed to UMAP and Smart Local Moving (SLM) clustering algorithm according to Seurat WNN vignette. Unsupervised clustering implemented with FindClusters function in Seurat with resolution set to 1.6 yielded 113 clusters. Clusters with less than 20 cells and clusters driven by one donor were merged with the nearest cluster using BuildClusterTree with slot set to ‘scale.data’ resulting in 31 clusters (Fig. 1d, Extended Data Fig. 3). We then created a gene activity matrix inferred from single nuclei ATAC sequencing (snATAC-seq) data with the GeneActivity function in Signac which quantifies ATAC-seq counts in the 2 kb-upstream region and gene body. Cell cluster identification The 31 cell clusters were annotated using canonical gene expression in accordance with published single-cell transcriptomics studies of mammalian hippocampus 14-18 , markers identified by a Wilcoxon rank-sum test implemented by FindAllMarkers function in Seurat (parameter setting RNA: only.pos = TRUE, min.pct = 0.2, min.diff.pct = 0.1, logFC_threshold = 1; parameter setting ATAC: only.pos = T, min.pct = 0.05, test.use = ‘LR’, latent.var = ‘nCount_ATAC’), and prediction scores from integration with spatial data implemented with TransferData function in Seurat. The following canonical cell types were annotated using the respective recognized markers (ordered as in Fig. 1f-g): oligodendrocyte progenitor cell (OPC): PDGFRA 14,16,19-21 , OLIG1 21 ; oligodendrocyte (olig) MBP 19,22 ; vasculature cells including endothelial cells (Endo): FLT1 , PECAM1 (protein name: CD31), CLDN5 , NES ; vascular leptomeningeal cells (VLMC): CEMIP , COL1A2 ; pericyte (Peri): PDGFRB , ABCC9 ; immune cells including microglia (Micro) APBB1IP , PTPRC 14,20 , P2RY12 14 , macrophage (Macro): CD163 (alias is macrophage associated protein); T-cell: CD247 (alias is T-cell surface glycoprotein); ependymal cells of the subventricular zone (Epe): CFAP54 , DNAH9 , FOXJ1 ; choroid plexus cells (Ch.Pl): HTR2C , CLIC6 , TTR ; astrocytes (Astro, 3 clusters): AQP4 , SLC1A2 , ALDH1L1 , GLUL , GFAP , with subtypes Astro1: GPC5 , WIF1 , GFAP; Astro2: CD44 , TNC , GFAP ; Astro3: TNC , GFAP ; Cajal Retzius cells (CR): RELN 14,16 ; granule neurons (GN, 2 clusters): RBFOX3 , SYT1 , MYT1L , SLC17A7 , CAMK2A , PROX1 , CALB1 , TLL1 , PPFIA2 , SEMA5A , BDNF , with subtypes POSTN - (GN1) and POSTN + (GN2); excitatory neurons (ExN, 6 clusters): with RBFOX3 , MYT1L , CAMK2A , SLC17A7 , SV2B expressed by all ExN and specific subtype markers for ExN1: CADPS2 , NEUROD6 ; ExN2: SATB2 , SLC22A10 ; ExN3: NEUROD6 , TRHDE , CPNE4 , HTR7 ; ExN4: SATB2 , SLC22A10 , HS3ST2 , SYNPO2 ; ExN5: HS3ST2 , NTNG1 ; ExN6: HS3ST2 , CHRM2 , PCP4 ; inhibitory neurons (InN, 9 clusters): all expressing RBFOX3 , MYT1L , GAD1 , GAD2 , SLC32A1 , DCX , with InN subtypes expressing: VIP , SST , PVALB , LHX6 , LAMP5 , SLC17A8 , TUBB3 , PROX1 , PENK . Out of the 31 cell clusters (Extended Data Fig. 4), 12 cell types were identified by merging all astrocyte clusters, all inhibitory neuron clusters, all immune clusters, all vasculature clusters, the two GN clusters, ExN1 and ExN3 merged to CA.ExN ( EPHA5 16 , NEUROD6 19 ), and ExN2 and ExN4-6 merged to Other ExN (O.ExN; expressing TSHZ2 23,24 ). Single nuclei integration with other datasets We integrated our database with publicly available databases to achieve a better characterization of the molecular landscapes defining cell types and states in the human hippocampus in neurotypical controls and MDD. We downloaded the count matrix of datasets generated by recently published studies focusing on the hippocampus 14,16,25,26 . Independently generated datasets were preprocessed using our standard pipeline using Seurat 10 and Harmony 13 in R. Then, following a published approach 14 , for the top 2000 variably expressed genes, Pearson’s R was calculated for the average expression across cell types (Extended Data Fig. 5). Tissue and library preparation for Visium spatial gene expression Two adjacent sections per donor were sectioned at 14 µm at -20C and mounted within capture areas on Visium slides. Slides were stored at -80C, then fixed, stained, and imaged according to 10X User Guide CG000160 Rev C. Spatial gene expression libraries were prepared according to 10X User Guide CG000239 Rev F. The libraries were sequenced on an Illumina NovaSeq 6000 using v4 chemistry to an average sequencing depth of about 67,000 reads/spot and 1965 median genes/spot (Supplementary Table 7). SpaceRanger (v1.3.0) with chemistry set to Spatial 3’ v1 used to generate count matrices. Tissue preparation and loading for Xenium in-situ hybridization Fresh frozen tissue was sectioned at 10µm onto 10.45mm x 22.45mm sampling area on 10X Genomics Xenium Slide according to 10X Demonstrated Protocol CG000579 Rev C. Tissue was fixed and permeabilized according to 10X Demonstrated Protocol CG000581 Rev C. Probe hybridization, ligation, and amplification were carried out according to 10X User Guide CG000582 Rev C. The output from the Xenium analyzer was loaded into Xenium Explorer 1.3 or RStudio for integration with single-nuclei gene expression data. Spatial gene expression data pre-processing, integration, and clustering Visium (10X Genomics) count matrices and spatial data were loaded into Seurat using the Load10X_Spatial command with the default parameters. Feature counts for each spot were divided by total counts for that spot and multiplied by scale factor (10,000) and natural log transformed using function NormalizeData. Features that were outliers to mean variability were found with local polynomial regression of log(variance) and log(mean) which gives the observed mean and expected variance of feature expression implemented using the FindVariableFeatures function. Then, data were scaled and centered using ScaleData function. Spatial data from replicates were merged using the merge function. Dimensionality reduction and clustering were calculated using RunPCA, FindNeighbors, FindClusters, and RunUMAP with dimensions parameter set to 1:30. Spatial data were integrated with multiome data using FindTransferAnchors and TransferData functions in Seurat 10 (Fig. 1h, Fig. 2e). Hippocampalsubfields, including granule cell layer (GCL), subgranular zone (SGZ), molecular layer (ML), Cornu Ammonis (CA), white matter (WM), vasculature (vasc), chorioid plexus (chor.plex), and subventricular zone (SVZ) were manually annotated in Loupe Browser within the Visium fiducial frame (Fig. 2b). Folds in the tissue were labeled as such and excluded from downstream analysis. Spots with less than 500 unique molecular identifiers were excluded from downstream analysis. Unsupervised clustering of merged spatial gene expression data was driven largely by sample and donor identify, so spatial gene expression data were integrated using canonical correlation analysis (CCA) in Seurat (Extended Data Fig. 6a-b). Spatial cluster identification Using FindClusters in Seurat, 13 clusters were identified at 0.5 resolution: two hilus clusters, hilus1 comprising mostly excitatory neurons, and hilus2 showing markers of inhibitory neurons, mossy cell markers, serotonin receptor 2c ( HRT2C ) and tryptophan hydroxylases 2 ( TPH2 ); CA3 stratum pyramidales (ca3) comprising excitatory neurons; stratum oriens-stratum lucidum (so-slu) representing WM; so-slu of the CA3 (ca3-so-slu) expressing parvalbumin and markers involved in cell-cell adhesion, extracellular matric, methylation, calcium signaling, and trophic factors; stratum radiatum (sr) comprised of axonal markers; stratum oriens-stratum lacunosum (so-sla) made up of dendrites; stratum moleculare (ml) comprised of apical dendrite-associate markers and HTR2B ; stratum granulosum (gcl) made up mostly of GNs; SGZ-polymorphic layer (sgz-pl) expressing markers involved in neurogenesis, development and differentiation, TPH1 and TPH2 ; scattered spots in SGZ and stratum moleculare (sgz-ml) expressing stem cell, differentiation and neurogenesis markers and HTR3A and HTR1D ; vasculature (vasc) expressing endothelial and collagen markers, and HTR2B and TPH1 ; and choroid plexus (ch.pl) expressing cilia and flagella associated markers, HTR2C , and astrocyte markers (Supplementary Table 8). Integration of spatial and single-nucleus gene expression datasets We used Seurat v3 anchor-based integration to transfer annotations from single nucleus data to spatial Visium data. Single-nucleus annotations were used as the reference dataset and both whole transcriptome Visium spatial gene expression and Xenium in-situ hybridization were used as query data. Spots were assigned probability scores based on transcriptional congruence with cell types from single nucleus data (Fig. 1h-i, Fig. 2e-f, Extended Data Fig. 4f). RNA Velocity inference on Visium spatial transcriptomic data To understand the kinetics of cell state in human hippocampus, we computed RNA velocity using velocyto 27 based on the relative abundance of nascent (unspliced) mRNAs and mature (spliced) mRNAs on Visium spots (55x55 m). Samtools 28 (v1.9) was used to sort bamfiles and velocyto run 27 was used to run velocity analysis and output a loom file. Pseudotime lineage analysis was performed by RNA velocity (scVelo Python package 29 , Fig. 2g-k, Extended Data Fig 6). The proportion of spliced/unspliced RNAs did not differ between unsupervised spot clusters, and we did not have power to test differences between MDD and CTRL. We used unsupervised clusters located in SGZ and GCL (Supplementary Table 8) that expressed proliferation, differentiation, maturation and neurogenesis markers among their top 100, to subset spots hypothesized to be part of neurogenesis trajectory (Extended Data Fig. 6c-h). We selected spatial clusters sgz.ml, sgz.pl, and gcl (Supplementary Table 8) and recalculated UMAP embedding using the top 5000 variably expressed genes and additional genes known to be involved in neurogenesis (total of 5074 genes) and the first 15 PCs for FindNeighbors and RunUMAP. Seurat object was translated to AnnData using SeuratDisk functions SaveH5Seurat and Convert with ‘dest’ set to ‘h5ad.’ The ‘data’ layer from the Seurat object was used for gene expression calculations in scvelo. The AnnData object was merged with loom files output of velocyto run on the command line. The dynamical model was run with the default parameters. Identification of neurogenic niche stem cell progeny using machine learning on single nuclei transcriptome data We apply a supervised learning approach to identify neurogenic cell types, as recently published 16 . Prototypes for each cell type were discovered by first retaining the sparse counts matrix for the top 4000 variable genes and an additional 54 genes involved in neurogenesis and neuronal maturation. The Seurat function NormalizeData was rerun on this matrix with the default parameters. For each gene, normalized counts that were above the sum of the third quartile and the product of 1.5 times the inter-quartile range were brought down to this value to reduce the effect of extreme outliers. Prototype scores were assigned to cells for identification of the following cell types: TypeI progenitors (quiescent neural stem cells), TypeII progenitors (intermediate amplifying neural progenitors), TypeIII (differentiating neurons), and ImGN (immature granule neurons, which are maturing and integrating into the DG circuit) using normalized count data. The distribution of summed normalized counts of selected genes expected to be expressed at each maturation stage was used to determine a cutoff for defining prototypical cells, considering that, based on previous studies in rodents and primates, progenitors are expected to be a small proportion of the total number of adult human dentate gyrus cells (<1% of all cells). We summed normalized counts of the following stage-specific genes: for TypeI, we summed normalized counts for GFAP , SOX2 , PAX6 , NES , FABP7 , HES5 , and ETNPPL ; for TypeII, normalized counts for ASCL1 , FOXO3 , NR2E1 , EOMES , NEUROG2 , NEUROD1 , NES , MCM2 , PCNA , MKI67 , and TOP2A were summed, and because oligodendrocyte progenitors can express markers in common to other progenitors, normalized counts for MBP and MOBP were subtracted so that cells that could be on the oligodendrocyte differentiation trajectory were less likely to be determined to be TypeII prototypes; for TypeIII, we summed normalized counts for MYT1L , PROX1 , DCX , CALB2 , RELN , ST8SIA2 , and ST8SIA4 and subtracted normalized counts for RBFOX3 and VIP to decrease the likelihood that TypeIII prototypes were at a later stage of maturation or inhibitory neurons; for ImGN prototypes, normalized counts of RBFOX3 , PROX1 , NEUROD1 , DCX , BHLHE22 , COL25A1 , and POSTN were summed. Cells with high prototypes scores were identified as prototypical, or high confidence cells. We applied the LASSO regression (implemented with glmnet in R 30 ) for each cell type individually. For each cell type, the prototypes and a randomly sampled quantity of cells required to reach 50,000 were used for training and testing the model. The trained model uses a list of positively and negatively ranked coefficients to rank genes in their ability to predict cell types. This method uses many genes to characterize these cell types (for TypeI, TypeII, TypeIII, and ImGN it used 372, 410, 555, and 422 genes respectively), instead of just a few arbitrarily chosen genes. Few cells (approximately 150 cells) were assigned two identities. For these cells, we manually assigned the identity with the highest predicted probability from predict.glm with type set to ‘response’ and the tuning parameter s was chosen by 10-fold cross validation by cv.glmnet. For final counts of cells in each LASSO identified cell type, see Supplementary Table 9. Cell trajectory inference based on single nucleus transcriptomic data In the absence of cytoplasmic mRNA and thus degradation rate in the single-nuclei gene expression data, we used Palantir 31 to characterize the trajectory of cells identified as TypeI, TypeII, TypeIII, ImGN, and GN from LASSO regression. TypeI, TypeII, TypeIII, ImGN, and GN cell types were subset, and re-preprocessed using Seurat and Harmony as described in Multi-modal integration. The first 5 Harmony batch, sample, and donor corrected PCs were used for FindNeighbors and RunUMAP (Fig. 3b). Seurat object was converted to AnnData using SeuratDisk functions SaveH5Seurat and Convert with ‘dest’ set to ‘h5ad.’ Palantir function core.run_palantir was used to calculate psueodtime ordering of each cell (Fig. 3c), terminal state probabilities, and entropy (a quantitive measure of the differentiation potential of each cell). Generalized Additive Models 31 were used to determine the gene expression trends along the neurogenesis lineage using the function presults.compute_gene_trends (Fig. 3e). Differential gene expression analysis in MDD versus CTRL To determine cell-type specific differentially expressed genes (DEGs) between subjects with MDD and neurotypical controls, count, peak, and gene activity matrices were aggregated on the sample and cell type level. Data were normalized using center log ratio transformation implemented using the clr function in the compositions R package. Then, we used the differential expression pipeline from R package limma 32,33 . Samples from the same donor were not aggregated. Instead, we used limma’s DuplicateCorrelation function to add a random effect for donor identity because, in some cases, samples from the same donor were processed in different batches. Additionally, since donors had between 2-3 samples, we used voomWithQualityWeights to adjust for differences in the number of samples per donor and number of cells aggregated/sample. We added RIN, PMI, batch, and age to the model as fixed effects. For RIN, PMI, batch, and age, values that were outside the sum of the third quartile and the product of 1.5 times the inter-quartile range from the median were winsorized to reduce the effect of extreme outliers. P-values were adjusted using Benjamini-Hochberg correction. We implemented the Bioconductor scRNA-seq pseudobulk pipeline to analyze DEGs between MDD and neurotypical controls in all clusters 34-38 . Differential chromatin accessibility analysis in MDD versus CTRL To determine cell-type specific differentially accessible regions (DARs) between subjects with MDD and neurotypical controls, the peaks matrix was aggregated and the pipeline for differential gene expression was applied. Additionally, a gene activity matrix was created from snATAC-seq using GeneActivity function in Signac. Chromatin accessibility at transcriptional and promoter regions was used to create the gene activity matrix. Then, the pipeline for predicted differentially active genes (DAGs) was applied to the aggregated gene activity matrices. Pre-ranked gene set enrichment analysis We performed pre-ranked Gene Set Enrichment Analysis (GSEA) 39 with FGSEA 40 using the product of log fold change and negative base 10 log of the Benjamini-Hochberg (BH) corrected p-value as the ranking method. We used the Gene Ontology (GO) annotation system to determine molecular function (MF, the molecular activities carried out by gene products), the cellular component (CC, the cellular location where gene products are active and perform their function), and biological process (BP, the processes and activities in which gene products are involved), with gene sets obtained from Molecular Signatures Database (MSigDB) 41 . We used Kyoto Encyclopedia of Genes and Genomes (KEGG) 42 to infer biological pathways related to DEGs, and used the Reactome pathway 43 to understand molecular mechanisms of biological reactions that are interconnected and result in pathways. The following parameters were used for the fgsea function: minSize=15, maxSize=400, nperm = 10000. Outputs with BH adjusted p-value less than 0.05 were considered significant. PsyGeNET analysis We ran enrichedP D from Psygenet2r 44 a R/Bioconductor package for the analysis of psychiatric disease genes, for leveraging PsyGeNET 45 a knowledge resource that collects and validates information of psychiatric diseases and their genes. We used database = ‘ALL’ to find psychiatric disorders for which DEGs from any cell type showed enrichment. We tested all DEGs from gene expression, all DEGs from gene activity, and the combination, and ran psygenetGene with database = ‘ALL’ and created geneAttrPlot for the evidence index. We then ran psygenetGene with DEGs from GN and ImGN (Fig. 4c) which were the cell types with most dysregulated genes across multiple modalities: transcript, chromatin accessibility and gene activity. Weighted gene co-expression network analysis (WGCNA) Weighted gene co-expression network analysis (WGCNA) was performed to identify co-expression modules using snRNA-seq data 22,46 . First, the aggregated expression for each sample (N = 52; 25 CTRL, 27 MDD) in the GN and ImGN cell types was calculated by summing the counts per gene across all nuclei. To account for known external sample traits, the counts were corrected for age and batch using limma 47 . In addition, lowly expressed genes with total counts of below 5 were removed. A soft thresholding power of 20 and 18, respectively, and a minimum module size of 30 genes, were used for network construction and module detection for GN. Each module was correlated with the phenotype (CTRL vs. MDD), and significance was determined using a p -value < 0.05. The online tool Metascape 48 was used to perform GO on the modules that showed a correlation with MDD. In Situ Hybridization For each donor, 3 serial sections of 16 m at 80 m intervals were used for Singleplex and Duplex RNAScope assays. For Singleplex RNAscope (ACDbio), the DCX probe (RNAscope Probe-Hs-DCX, Cat. No. 489551) was hybridized and amplified. Duplex RNAscope was used to investigate DCX co-localization with NEUROD1 , PROX1 , RBFOX3 , or TUBB3 mRNA in the human DG. Fresh frozen 16 m tissue sections were fixed in 4% paraformaldehyde for 15 minutes at room temperature (RT). Sections were washed in 1x PBS and then dehydrated in 50%, 70%, and twice in 100% EtOH for 5 minutes. RNAscope was performed according to manufactured instructions. The DCX probe (Cat. No. 489551-C2) and NEUROD1 probe (Cat. No. 437281), PROX1 probe (Cat. No. 530241), RBFOX3 probe (Cat. No. 415591), or TUBB3 probe (Cat No. 318901) were hybridized and amplified in C2 and C1 respectively. Slides were counterstained with 50% hematoxylin (Gil’s Hematoxylin No. 1) according to manufacturer’s instructions. The slides are then quickly dipped in xylene and mounted with Permount. Multiplex RNAscope was used to investigate the expression of GAD1 (Cat No 404031-C2), PROX1 (Cat No 530241-C3), and DCX (Cat No 489551) in the human DG, because single nuclei data showed co-expression of GAD1 in cell clusters expressing DCX . Sections were counterstained with 4’, 6-diamidino-2-phenylindole (DAPI). Gene expression analysis using qPCR RNA was isolated using RNeasy FFPE Kit (Qiagen, Germantown, MD) according to the manufacturer’s instructions. RNA quantity was assessed by evaluation of the A260/280 and A260/230 ratios using a Nanodrop spectrometer (Thermo Scientific 840274100). Superscript III enzyme (Invitrogen) was used to reverse transcribe 200 ng total RNA. Quantitative Real-Time PCR was performed on a StepOnePlus™ Real-Time PCR System using a TaqMan™ Gene Expression Master Mix (Applied Biosystems, Waltham, USA). A ready-made DCX TaqMan™ probe (Applied Biosystems, Waltham, USA) was used to amplify a 77bp region spanning an exon boundary within the coding sequence of the target gene. Each sample was assayed in triplicate and DCX target gene expression was normalized to the geometric mean of the reference gene beta-actin ( ACTB ). DCX gene expression was analyzed in PRISM using a Welch unpaired T-test. Double Immunofluorescence To examine expression of cell type specific markers at the protein level in human hippocampus, we performed double immunofluorescence (IF) for protein corresponding to RNAs of interest identified with the single nuclei and spatial transcriptomic analyses. Human hippocampus tissue sections (50 m thick) were washed and prepared as previously published 49 . The primary antibodies were incubated [(1:1000 guinea pig anti-DCX (Millipore), 1:5000 mouse anti-Glial Fibrillary Associated Protein (GFAP, Sigma-Aldrich), 1:500 rabbit anti-Ionized calcium binding adaptor molecule 1 (Iba1, Fujifilm), 1:50,000 mouse anti-Neurofilament (NF; Sigma-Aldrich); 1:2000 mouse anti-polysialylated neural cell adhesion marker (PSA-NCAM, Miltenyi Biotec); 1:500 rabbit anti-Calretinin (Swant Inc.), 1:500 rabbit anti-Calbindin (Swant Inc.), 1:500 rabbit anti-Neuronal Nuclear Marker (NeuN, Millipore)] in a 25% blocker overnight or for 5-days (Calretinin and Calbinin) at 4°C. The secondary antibodies were incubated overnight at 4°C with Jackson Alexa Fluor 488 goat anti-mouse or anti-rabbit (1:500) and Jackson Alexa Fluor 594 goat anti-guinea pig (1:500). The following day the tissue was washed for 30min (2x15’) in 0.05M PBS. Then the tissue was counterstained with 4’, 6-diamidino-2-phenylindole (DAPI) in PBS for 3min before being washed for 15min (3x5’) in 0.05M PBS. The tissue was then treated with 0.1% True Black (Biotium) in 70% ethanol in distilled water to quench autofluorescence. Sections were then washed for 1h (6x10’) in 0.05M PBS. The tissue was coverslipped using 90% glycerol and PBS. Immunohistochemistry Immunohistochemistry (IHC) for DCX was performed as previously described 49 Sections were incubated in guinea pig anti-doublecortin (1:30,000, Millipore) in 10% blocking solution overnight at 4°C on a shaker. Following 2-hour secondary incubation with anti-guinea pig biotin-conjugated antibody (Vector Laboratories) and 1-hour incubation in avidin-biotin-horse radish peroxidase system (Vector Laboratories), diluted 1:200, sections were washed in sodium acetate for 5 min, followed by a PBS wash for 20 min (2x10’) before being mounted on glass slides and desiccated. Sections were counterstained with Nuclear Red which involves dehydrating slides in ethanol, clarifying slides in xylenes and cover-slipping with Permount (Sigma-Aldrich). Sections were processed to identify mature granule neurons (anti-Neuronal Nuclear antigen [NeuN] mouse monoclonal antibody, 1:100,000; Chemicon, Temecula, CA), progenitor cells of Type II or III (anti-nestin mouse monoclonal antibody, 1:8000, Chemicon) and mitotic cells (anti-Ki-67 mouse monoclonal antibody, 1:200, Novocastra Clone-MM1, Newcastle Upon Tyne, UK). Immunohistochemistry and stereology were performed as previously described 50 . Confocal Microscopy Cell markers co-localization was visualized using an Olympus Fluoview FV1000 Confocal Laser Scanning System (Olympus, Waltham, MA) equipped with an Olympus IX81 Inverted Microscope, with electronic stage controller (Prior Scientific, Rockland, MA), camera interface, and Olympus Fluoview 1000 (v. 1.5) software, using the following objectives: UPLSAPO 10X and 20X air, with numerical aperture (NA) 0.40, 0.75; UPLFLN 40X oil with NA 1.30; and PLAFON 60X oil with NA 1.42 (Olympus). To quantify co-localization of cell markers (i.e., DCX/NeuN), image stacks of the whole DG were obtained using a confocal scanning microscope (Leica TCS SP8 2-Photon, Leica Microsystems Inc.), and then processed for Stereology using Stereoinvestigator software (MBF, Inc.). Fluorescence from the different fluorophores was detected in the following way: 1) 488-AffiniPure was excited at 488 nm and detected at 505-550 nm, 2) 594-AffiniPure was excited at 552 nm and detected at 600-650 nm, and 3) DAPI was excited at 405 nm and detected at 415-485 nm. For 488- and 552-nm excitation, the beam path included a TD 488/552/638 beamsplitter, while for 405 nm excitation a substrate beamsplitter was used. Hippocampal sections from each subject were scanned to produce 13 ± 2 μm image stacks, depending on the thickness of each tissue section. All z stacks were imaged with a dry Leica 20X objective (NA 0.70, working distance 0.5 mm), with a field of view of 553.6 × 553.6 μm, a pixel size of 0.54 × 0.54 μm, optical sectioning of 2.36 μm, and a z step of 1 μm. Positive and negative controls were run to ensure signal specificity and minimal background, creating an optimal signal to noise ratio. The same parameters were applied for every section of every subject. Image stacks were then imported in our MicroBrightfield system (MBF Bioscience, Williston, VT) to perform unbiased stereology. Cell quantification and statistical analysis We used an unbiased stereological approach (optical disector with fractionator) to estimate numbers of cells expressing DCX (DCX+), DCX/NeuN+ cells, DCX mRNA+ cells and DCX/TUBB3 mRNA+ cells. Olympus BX51-WI microscope was used with electronic stage controller LUDL MAC 6000 XYZ (BioVision Technologies, Exton, PA) and camera interface for Retiga 2000R CCD Scientific Camera (Q-Imaging, Surrey, BC, Canada). For anatomical alignment along the rostro-caudal axis of the DG, processed sections were anatomically matched with Nissl-stained sections at 1-mm interval using a stereoscope (Leica, Wild M3Z, Heerburg, Switzerland). We defined the anterior DG as the portion from the most rostral appearance of the DG to the start of the lateral geniculate. The mid DG was defined as the portion going from the start of the lateral geniculate to the disappearance of the lateral geniculate, and the posterior DG was the region from the end of the lateral geniculate to the end of the caudal DG. Outlines of the DG including SGZ and GCL were aligned from the most rostral to the most caudal hippocampal sections and used as boundaries for cell counting. To quantify cells, the first hippocampus section to be sampled was the one in which the DG first appeared. Subsequent sections were assayed every 2 mm thereafter in systematic fashion, until the DG disappeared, with an average of 8-10 sections per subject for DCX immunohistochemistry experiments. It should be noted that for DCX/NeuN and RNAscope experiments, an average of 3-5 sections of the anterior hippocampus were used per subject. Using the Stereo Investigator system, DG boundaries were defined at low magnification using the 4x objective. Cells were not counted in the upper and lower 3 μm in the z planes of the tissue sections (the guard zones). The estimated total number of cells was calculated as: N=∑Q − ×(1/hsf)×(1/asf)×(1/ssf) where Σ Q− = total number of cells counted, hsf (height sampling fraction) = h/t, where t is the mean section thickness, h is the height of the optical dissector , asf = area sampling fraction (area of the counting frame over the area of x,y step), and ssf = section sampling fraction (1/section interval). The cell counts were analyzed in PRISM using a Welch unpaired T-test. Proteomics We used Shotgun proteomics liquid chromatography with tandem mass spectrometry (LC/MS/MS), a strategy with broad applicability, based on digestion of proteins with proteolytic enzymes and analysis of the resulting peptides by mass spectrometry 51 . Fresh-frozen 4% paraformaldehyde post-fixed tissue was dissolved in 0.2% RapiGest SF in 50 mM ammonium bicarbonate, heated to 105 o C for 30 min, then to 80 o C for 2 h, digested with trypsin as previously described 52,53 , and then 50 fmol of yeast alcohol dehydrogenase were added as internal control. LC/MS/MS was run for 120 minutes, two chromatograms recorded for each biological replicate, in traveling-wave ion mobility spectrometry (TWIMS) MS E resolution mode 51,53,54 ; mass spectra were recorded every 0.6s; every 30s internal “lockmass” spectrum of Glu-1-Fibrinopeptide B was recorded (m/z 785.8426); 11,000 mass spectra were recorded per run; a total of 792,000 mass spectra were analyzed. Of the 1811 proteins detected, 680 were represented by a single peptide or with protein identification quality scores less than 250 and were not included in the analysis. The two groups (MDDs and CTRL) were compared against each other and differentially expressed proteins were identified with a PeptideTeller predicted error (false positive rate of 1%). P-values used in our report on these experimental results are based on the error function called xdev, which addresses false positives by using an advanced error model as described in previous papers 55,56 . The xdev-derived P-values were generated by the software Rosetta Elucidator Protein Expression Analysis System (V. 4.0.0.2.13, PerkinElmer, Boston, MA,). This software, developed by Merck & Co./Rosetta 57-59 extends the dynamic range of fold-change data, especially for low abundance proteins. This system matches accurate mass and retention time across all LC/MS/MS chromatograms in the experiment, enhancing the effectiveness of the mass spectrometry analyses. The xdev represent the difference of intensities between groups (pairwise comparisons) divided by the error of the difference, and the p-value of differential expression is calculated from an error function of a standard Gaussian distribution. Elucidator ran on Oracle Linux v. 6.5 server with 128 GB RAM and 32 Tb of RAID network attached storage. From within Elucidator, results were searched with Mascot server (Version 2.5.1, Matrix Science Ltd., London, UK). The search was against a SwissProt human canonical protein sequence database with isoforms (UniProtKB/Swiss-Prot Release 2017_01 containing 84,336 sequences; 48,572,808 residues, www.uniprot.org). Our quality control indices included peptide count above two for considering a protein a good candidate, and p-value < 0.05. Methods References 1 Rosoklija, G., Duma, A. & Dwork, A. J. Psychiatric brain collection in Macedonia: general lessons for scientific collaboration among countries of differing wealth. Pril (Makedon Akad Nauk Umet Odd Med Nauki) 34 , 95-98 (2013). 2 Kelly, T. M. & Mann, J. J. Validity of DSM-III-R diagnosis by psychological autopsy: a comparison with clinician ante-mortem diagnosis. Acta Psychiatr.Scand. 94 , 337-343 (1996). 3 Zhang, D. et al. Spatial epigenome-transcriptome co-profiling of mammalian tissues. Nature 616 , 113-122 (2023). https://doi.org:10.1038/s41586-023-05795-1 4 Deng, Y. et al. Spatial profiling of chromatin accessibility in mouse and human tissues. Nature 609 , 375-383 (2022). https://doi.org:10.1038/s41586-022-05094-1 5 Boldrini, M. et al. Resilience Is Associated With Larger Dentate Gyrus, While Suicide Decedents With Major Depressive Disorder Have Fewer Granule Neurons. Biol Psychiatry 85 , 850-862 (2019). https://doi.org:10.1016/j.biopsych.2018.12.022 6 Boldrini, M. et al. Human Hippocampal Neurogenesis Persists throughout Aging. Cell Stem Cell 22 , 589-599 e585 (2018). https://doi.org:10.1016/j.stem.2018.03.015 7 Besnard A, L. T., Levinson S, Chu D, Vicidomini C, Scobie KN, Dwork AJ, Arango V, Rosoklija GB, Mann JJ, Hen R, Leonardo ED, Boldrini M, Sahay A. Targeting Kruppel-like factor 9 (Klf9) in excitatory forebrain 1 neurons protects against chronic stress-induced dendritic spine enlargement and maladaptive fear responses. Cell Reports In Press (2018). 8 Maitra, M. et al. Extraction of nuclei from archived postmortem tissues for single-nucleus sequencing applications. Nature Protocols 16 , 2788-2801 (2021). 9 Hao, Y. et al. Integrated analysis of multimodal single-cell data. Cell 184 , 3573-3587.e3529 (2021). https://doi.org:https://doi.org/10.1016/j.cell.2021.04.048 10 Hao, Y. et al. Dictionary learning for integrative, multimodal and scalable single-cell analysis. Nature Biotechnology (2023). https://doi.org:10.1038/s41587-023-01767-y 11 Stuart, T., Srivastava, A., Madad, S., Lareau, C. A. & Satija, R. Single-cell chromatin state analysis with Signac. Nature Methods 18 , 1333-1341 (2021). https://doi.org:10.1038/s41592-021-01282-5 12 Lawrence, M. et al. Software for Computing and Annotating Genomic Ranges. PLOS Computational Biology 9 , e1003118 (2013). https://doi.org:10.1371/journal.pcbi.1003118 13 Korsunsky, I. et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nature methods 16 , 1289-1296 (2019). 14 Franjic, D. et al. Transcriptomic taxonomy and neurogenic trajectories of adult human, macaque, and pig hippocampal and entorhinal cells. Neuron 110 , 452-469.e414 (2022). https://doi.org:https://doi.org/10.1016/j.neuron.2021.10.036 15 Wang, W. et al. Transcriptome dynamics of hippocampal neurogenesis in macaques across the lifespan and aged humans. Cell Research 32 , 729-743 (2022). https://doi.org:10.1038/s41422-022-00678-y 16 Zhou, Y. et al. Molecular landscapes of human hippocampal immature neurons across lifespan. Nature 607 , 527-533 (2022). https://doi.org:10.1038/s41586-022-04912-w 17 Hochgerner, H., Zeisel, A., Lönnerberg, P. & Linnarsson, S. Conserved properties of dentate gyrus neurogenesis across postnatal development revealed by single-cell RNA sequencing. Nature Neuroscience 21 , 290-299 (2018). https://doi.org:10.1038/s41593-017-0056-2 18 Ayhan, F. et al. Resolving cellular and molecular diversity along the hippocampal anterior-to-posterior axis in humans. Neuron 109 , 2091-2105.e2096 (2021). https://doi.org:10.1016/j.neuron.2021.05.003 19 Habib, N. et al. Massively parallel single-nucleus RNA-seq with DroNc-seq. Nature methods 14 , 955-958 (2017). 20 Hao, Z.-Z. et al. Single-cell transcriptomics of adult macaque hippocampus reveals neural precursor cell populations. Nature Neuroscience 25 , 805-817 (2022). https://doi.org:10.1038/s41593-022-01073-x 21 van Bruggen, D. et al. Developmental landscape of human forebrain at a single-cell level identifies early waves of oligodendrogenesis. Developmental Cell 57 , 1421-1436.e1425 (2022). https://doi.org:https://doi.org/10.1016/j.devcel.2022.04.016 22 Maitra, M. et al. Cell type specific transcriptomic differences in depression show similar patterns between males and females but implicate distinct cell types and genes. Nature Communications 14 , 2912 (2023). https://doi.org:10.1038/s41467-023-38530-5 23 Blankvoort, S., Olsen, L. C. & Kentros, C. G. Single Cell Transcriptomic and Chromatin Profiles Suggest Layer Vb Is the Only Layer With Shared Excitatory Cell Types in the Medial and Lateral Entorhinal Cortex. Front Neural Circuits 15 , 806154 (2021). https://doi.org:10.3389/fncir.2021.806154 24 Di Bella, D. J. et al. Molecular logic of cellular diversification in the mouse cerebral cortex. Nature 595 , 554-559 (2021). https://doi.org:10.1038/s41586-021-03670-5 25 Morabito, S. et al. Single-nucleus chromatin accessibility and transcriptomic characterization of Alzheimer’s disease. Nature Genetics 53 , 1143-1155 (2021). https://doi.org:10.1038/s41588-021-00894-z 26 Grubman, A. et al. A single-cell atlas of entorhinal cortex from individuals with Alzheimer’s disease reveals cell-type-specific gene expression regulation. Nature Neuroscience 22 , 2087-2097 (2019). https://doi.org:10.1038/s41593-019-0539-4 27 La Manno, G. et al. RNA velocity of single cells. Nature 560 , 494-498 (2018). https://doi.org:10.1038/s41586-018-0414-6 28 Danecek, P. et al. Twelve years of SAMtools and BCFtools. Gigascience 10 (2021). https://doi.org:10.1093/gigascience/giab008 29 Bergen, V., Lange, M., Peidli, S., Wolf, F. A. & Theis, F. J. Generalizing RNA velocity to transient cell states through dynamical modeling. Nature Biotechnology 38 , 1408-1414 (2020). https://doi.org:10.1038/s41587-020-0591-3 30 Friedman, J. H., Hastie, T. & Tibshirani, R. Regularization Paths for Generalized Linear Models via Coordinate Descent. Journal of Statistical Software 33 , 1 - 22 (2010). https://doi.org:10.18637/jss.v033.i01 31 Setty, M. et al. Characterization of cell fate probabilities in single-cell data with Palantir. Nature Biotechnology 37 , 451-460 (2019). https://doi.org:10.1038/s41587-019-0068-4 32 Liu, R. et al. Why weight? Modelling sample and observational level variability improves power in RNA-seq analyses. Nucleic Acids Res 43 , e97 (2015). https://doi.org:10.1093/nar/gkv412 33 Phipson, B., Lee, S., Majewski, I. J., Alexander, W. S. & Smyth, G. K. ROBUST HYPERPARAMETER ESTIMATION PROTECTS AGAINST HYPERVARIABLE GENES AND IMPROVES POWER TO DETECT DIFFERENTIAL EXPRESSION. Ann Appl Stat 10 , 946-963 (2016). https://doi.org:10.1214/16-aoas920 34 Zhu, A., Ibrahim, J. G. & Love, M. I. Heavy-tailed prior distributions for sequence count data: removing the noise and preserving large differences. Bioinformatics 35 , 2084-2092 (2019). 35 McCarthy, D. J., Campbell, K. R., Lun, A. T. L. & Wills, Q. F. Scater: pre-processing, quality control, normalization and visualization of single-cell RNA-seq data in R. Bioinformatics 33 , 1179-1186 (2017). https://doi.org:10.1093/bioinformatics/btw777 36 Amezquita, R. A. et al. Orchestrating single-cell analysis with Bioconductor. Nature Methods 17 , 137-145 (2020). https://doi.org:10.1038/s41592-019-0654-x 37 Love, M. I., Anders, S. & Huber, W. Analyzing RNA-seq data with DESeq2. Bioconductor 2 , 1-63 (2017). 38 Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biology 15 , 550 (2014). https://doi.org:10.1186/s13059-014-0550-8 39 Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A 102 , 15545-15550 (2005). https://doi.org:10.1073/pnas.0506580102 40 Korotkevich, G. et al. Fast gene set enrichment analysis. bioRxiv , 060012 (2021). https://doi.org:10.1101/060012 41 Liberzon, A. et al. The Molecular Signatures Database (MSigDB) hallmark gene set collection. Cell Syst 1 , 417-425 (2015). https://doi.org:10.1016/j.cels.2015.12.004 42 Kanehisa, M., Furumichi, M., Tanabe, M., Sato, Y. & Morishima, K. KEGG: new perspectives on genomes, pathways, diseases and drugs. Nucleic Acids Res 45 , D353-d361 (2017). https://doi.org:10.1093/nar/gkw1092 43 Milacic, M. et al. The Reactome Pathway Knowledgebase 2024. Nucleic Acids Research 52 , D672-D678 (2023). https://doi.org:10.1093/nar/gkad1025 44 Gutiérrez-Sacristán, A., Hernández-Ferrer, C., González, J. R. & Furlong, L. I. psygenet2r: a R/Bioconductor package for the analysis of psychiatric disease genes. Bioinformatics 33 , 4004-4006 (2017). https://doi.org:10.1093/bioinformatics/btx506 45 Gutiérrez-Sacristán, A. et al. PsyGeNET: a knowledge platform on psychiatric disorders and their genes. Bioinformatics 31 , 3075-3077 (2015). https://doi.org:10.1093/bioinformatics/btv301 46 Zhang, B. & Horvath, S. A general framework for weighted gene co-expression network analysis. Stat Appl Genet Mol Biol 4 , Article17 (2005). https://doi.org:10.2202/1544-6115.1128 47 Ritchie, M. E. et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 43 , e47 (2015). https://doi.org:10.1093/nar/gkv007 48 Zhou, Y. et al. Metascape provides a biologist-oriented resource for the analysis of systems-level datasets. Nat Commun 10 , 1523 (2019). https://doi.org:10.1038/s41467-019-09234-6 49 Boldrini, M. et al. Human hippocampal neurogenesis persists throughout aging. Cell stem cell 22 , 589-599. e585 (2018). 50 Boldrini, M. et al. Benzodiazepines and the potential trophic effect of antidepressants on dentate gyrus cells in mood disorders. Int J Neuropsychopharmacol 17 , 1923-1933 (2014). https://doi.org:10.1017/S1461145714000844 51 Brown, L. M. Quantitative shotgun proteomics with data-independent acquisition and traveling wave ion mobility spectrometry: a versatile tool in the life sciences. Adv Exp Med Biol 806 , 79-91 (2014). https://doi.org:10.1007/978-3-319-06068-2_4 52 Shimada, K. et al. Global survey of cell death mechanisms reveals metabolic regulation of ferroptosis. Nat Chem Biol 12 , 497-503 (2016). 53 Yang, Wan S. et al. Regulation of Ferroptotic Cancer Cell Death by GPX4. Cell 156 , 317-331 (2014). 54 Oswald, E. S., Brown, L. M., Bulinski, J. C. & Hung, C. T. Label-free protein profiling of adipose-derived human stem cells under hyperosmotic treatment. J Proteome Res 10 , 3050-3059 (2011). 55 Dai, H., Meyer, M., Stepaniants, S., Ziman, M. & Stoughton, R. Use of hybridization kinetics for differentiating specific from non-specific binding to oligonucleotide microarrays. Nucleic Acids Res 30 , e86 (2002). https://doi.org:10.1093/nar/gnf085 56 Weng, L. et al. Rosetta error model for gene expression analysis. Bioinformatics 22 , 1111-1121 (2006). 57 Paweletz, C. P. et al. Application of an End-to-End Biomarker Discovery Platform to Identify Target Engagement Markers in Cerebrospinal Fluid by High Resolution Differential Mass Spectrometry. Journal of Proteome Research 9 , 1392-1401 (2010). 58 Hendrickson, R. C. et al. High Resolution Discovery Proteomics Reveals Candidate Disease Progression Markers of Alzheimer's Disease in Human Cerebrospinal Fluid. PloS one 10 (2015). https://doi.org:e0135365 10.1371/journal.pone.0135365 59 McAvoy, T. et al. Quantification of tau in cerebrospinal fluid by immunoaffinity enrichment and tandem mass spectrometry. Clin. Chem. 60 , 683-689 (2014). Additional Declarations The authors declare potential competing interests as follows: HG and her family own stocks in Illumina, Inc.; JJM receives royalties from Columbia University for the Columbia Pathways App. and from the Research Foundation of Mental Hygiene for commercial use of the C-SSRS. Supplementary Files BoldriniHippocampusMultiomeExtendedDataFigures.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8593764","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":574606591,"identity":"77b43ba7-0ad2-4530-a2fe-5df9a0aebfa6","order_by":0,"name":"Madeline B. Mariani","email":"","orcid":"","institution":"Columbia University - NYSPI/RFMH","correspondingAuthor":false,"prefix":"","firstName":"Madeline","middleName":"B.","lastName":"Mariani","suffix":""},{"id":574606592,"identity":"2b684bae-96c7-4cab-a870-3068a8ce63db","order_by":1,"name":"Cheick A. Sissoko","email":"","orcid":"","institution":"Columbia University - NYSPI/RFMH","correspondingAuthor":false,"prefix":"","firstName":"Cheick","middleName":"A.","lastName":"Sissoko","suffix":""},{"id":574606593,"identity":"c1542c36-38bb-4bad-ae7a-abe8f2499199","order_by":2,"name":"Alexandra M Wamalwa","email":"","orcid":"","institution":"Columbia University - NYSPI/RFMH","correspondingAuthor":false,"prefix":"","firstName":"Alexandra","middleName":"M","lastName":"Wamalwa","suffix":""},{"id":574606594,"identity":"cba24613-8632-4673-8d55-18222b00c5e4","order_by":3,"name":"Anthony Ramnauth","email":"","orcid":"","institution":"Columbia University - NYSPI/RFMH","correspondingAuthor":false,"prefix":"","firstName":"Anthony","middleName":"","lastName":"Ramnauth","suffix":""},{"id":574606595,"identity":"2d2c75be-b553-4754-82d8-ebb9bc85067f","order_by":4,"name":"Yang Xiao","email":"","orcid":"","institution":"Columbia University - NYSPI/RFMH","correspondingAuthor":false,"prefix":"","firstName":"Yang","middleName":"","lastName":"Xiao","suffix":""},{"id":574606596,"identity":"5fbaeb9f-a45d-4cf6-b1ff-2ce229994f1a","order_by":5,"name":"Alexandria N. Tartt","email":"","orcid":"","institution":"Columbia University - NYSPI/RFMH","correspondingAuthor":false,"prefix":"","firstName":"Alexandria","middleName":"N.","lastName":"Tartt","suffix":""},{"id":574606597,"identity":"0797b7d0-c6c4-4663-9a71-a9ea149155c1","order_by":6,"name":"Camille Fulmore","email":"","orcid":"","institution":"Columbia University - NYSPI/RFMH","correspondingAuthor":false,"prefix":"","firstName":"Camille","middleName":"","lastName":"Fulmore","suffix":""},{"id":574606598,"identity":"755ce668-6ef5-4275-a608-1835cd8bcda1","order_by":7,"name":"Ying Liu","email":"","orcid":"","institution":"Columbia University - NYSPI/RFMH","correspondingAuthor":false,"prefix":"","firstName":"Ying","middleName":"","lastName":"Liu","suffix":""},{"id":574606599,"identity":"e89bde26-e89b-4030-a3e0-8c48a68cf47b","order_by":8,"name":"Wenpin Hou","email":"","orcid":"","institution":"Columbia University - NYSPI/RFMH","correspondingAuthor":false,"prefix":"","firstName":"Wenpin","middleName":"","lastName":"Hou","suffix":""},{"id":574606600,"identity":"18aa0b21-daef-4b0b-89ec-42961e70b56d","order_by":9,"name":"Yung-yu Huang","email":"","orcid":"","institution":"Columbia University - NYSPI/RFMH","correspondingAuthor":false,"prefix":"","firstName":"Yung-yu","middleName":"","lastName":"Huang","suffix":""},{"id":574606601,"identity":"12574b7a-3407-4f5d-a5ef-22b83424ebb3","order_by":10,"name":"Andrew J. Dwork","email":"","orcid":"","institution":"Columbia University - NYSPI/RFMH","correspondingAuthor":false,"prefix":"","firstName":"Andrew","middleName":"J.","lastName":"Dwork","suffix":""},{"id":574606602,"identity":"d9a22c99-667c-4a8a-a400-83bec5fdee63","order_by":11,"name":"Gorazd B. Rosoklija","email":"","orcid":"","institution":"Columbia University - NYSPI/RFMH","correspondingAuthor":false,"prefix":"","firstName":"Gorazd","middleName":"B.","lastName":"Rosoklija","suffix":""},{"id":574606603,"identity":"ef8f50b8-9f54-4189-a699-94411ace9bdc","order_by":12,"name":"Shujuan T. McDonald","email":"","orcid":"","institution":"Columbia University - NYSPI/RFMH","correspondingAuthor":false,"prefix":"","firstName":"Shujuan","middleName":"T.","lastName":"McDonald","suffix":""},{"id":574606604,"identity":"9b951921-f243-4551-a93f-d67d946fdd88","order_by":13,"name":"Victoria Arango","email":"","orcid":"","institution":"Columbia University - NYSPI/RFMH","correspondingAuthor":false,"prefix":"","firstName":"Victoria","middleName":"","lastName":"Arango","suffix":""},{"id":574606605,"identity":"ba2979fb-6927-4246-8839-12beaaa9052f","order_by":14,"name":"J. John Mann","email":"","orcid":"","institution":"Columbia University - NYSPI/RFMH","correspondingAuthor":false,"prefix":"","firstName":"J.","middleName":"John","lastName":"Mann","suffix":""},{"id":574606606,"identity":"8d2d03a2-85db-446d-a6c7-e43f8b5e6b8c","order_by":15,"name":"René Hen","email":"","orcid":"","institution":"Columbia University - NYSPI/RFMH","correspondingAuthor":false,"prefix":"","firstName":"René","middleName":"","lastName":"Hen","suffix":""},{"id":574606607,"identity":"f6080ca4-ba3b-476c-ba2a-a5d4ca3ef242","order_by":16,"name":"Lewis M. Brown","email":"","orcid":"","institution":"Columbia University","correspondingAuthor":false,"prefix":"","firstName":"Lewis","middleName":"M.","lastName":"Brown","suffix":""},{"id":574606608,"identity":"fbb52685-8ee8-4e3e-939c-5fbe3851b223","order_by":17,"name":"Hanga Galfalvy","email":"","orcid":"","institution":"Columbia University - NYSPI/RFMH","correspondingAuthor":false,"prefix":"","firstName":"Hanga","middleName":"","lastName":"Galfalvy","suffix":""},{"id":574606590,"identity":"b77a98bc-ad28-410a-967d-c70d9c6706ae","order_by":18,"name":"Maura B. Dupont","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAz0lEQVRIiWNgGAWjYDACCeaDDz5UMDDwMTCwMSSARA4Q1MKWbDjjDEg98Vp4zKR526BaGIjRoju7x0By5jw7OTaJ7LQHD2oY5PhuJODXYnbnWIHBx23JxmwSudsNEo4xGEsS1HIjeUPizG0HEtskcrdJJLAxJG4grCXB4DDvHJiWfwz1RGhJMWzmbYBqSWxjSDAgwi/JjDOOAf3C83a7QWKfhOHMMw8IaLndfPzHhxo7OX723G0Pf3yzkec7TsAWdCBBmvJRMApGwSgYBdgBAKBqSfx9PoUlAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0003-3574-7641","institution":"Columbia University - NYSPI/RFMH","correspondingAuthor":true,"prefix":"","firstName":"Maura","middleName":"B.","lastName":"Dupont","suffix":""}],"badges":[],"createdAt":"2026-01-13 15:30:54","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":false,"conflictsOfInterestStatement":true,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-8593764/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8593764/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":100426318,"identity":"8064ee70-1c46-4273-b5e5-4ed04f73d35f","added_by":"auto","created_at":"2026-01-16 14:19:19","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":212451,"visible":true,"origin":"","legend":"","description":"","filename":"HippocampusMultiomeManuscriptMethodsResearchSquare.docx","url":"https://assets-eu.researchsquare.com/files/rs-8593764/v1/81423bc6074d3b666c39fcd0.docx"},{"id":100425824,"identity":"9a443f35-189b-41be-a3cf-eeb59622719a","added_by":"auto","created_at":"2026-01-16 14:18:57","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":342,"visible":true,"origin":"","legend":"","description":"","filename":"rs8593764.json","url":"https://assets-eu.researchsquare.com/files/rs-8593764/v1/382283935d41809b27c23b3d.json"},{"id":100426029,"identity":"1b3beea9-e148-40d9-9d15-488bee3761f0","added_by":"auto","created_at":"2026-01-16 14:19:10","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":161880,"visible":true,"origin":"","legend":"","description":"","filename":"rs85937641enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8593764/v1/bcb1c5f22fe0be4c3c2570a4.xml"},{"id":100426300,"identity":"f06628da-d284-4979-ae52-c32e6e0debfa","added_by":"auto","created_at":"2026-01-16 14:19:18","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":157766,"visible":true,"origin":"","legend":"","description":"","filename":"rs85937641structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8593764/v1/2b67fee5b1390ddcc4b5c76a.xml"},{"id":100426140,"identity":"d2cda569-48bc-4472-adee-23bb7550527b","added_by":"auto","created_at":"2026-01-16 14:19:12","extension":"html","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":188956,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8593764/v1/55a194c8e8fed4021fa471cc.html"},{"id":100425837,"identity":"2ececed4-80a6-436d-b395-4b805f0c0619","added_by":"auto","created_at":"2026-01-16 14:19:00","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":2189621,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDefining hippocampal cell types characterized by gene expression and accessible chromatin sequencing. a.\u003c/strong\u003e Schematic of study design. \u003cstrong\u003eb-d.\u003c/strong\u003e UMAP plots using batch-, sample-, and donor-corrected Harmony components. \u003cstrong\u003eb. \u003c/strong\u003eUMAP visualization of 349,847 cells derived from gene expression data and colored by cell types. \u003cstrong\u003ec.\u003c/strong\u003eUMAP visualization of 349,847 cells derived from accessible chromatin data and colored by cell types. \u003cstrong\u003ed.\u003c/strong\u003e UMAP visualization of 349,847 cells derived from integration of two modalities using WNN analysis colored by annotations from unsupervised clustering.\u003cstrong\u003e e.\u003c/strong\u003e Pearson’s correlation of natural log1p of the average gene expression and average predicted gene activity calculated from accessible chromatin in each cell type. \u003cstrong\u003ef-g.\u003c/strong\u003e Dotplot depicting the expression of marker genes (\u003cem\u003ePDFGRA\u003c/em\u003e – oligodendrocyte progenitor cells (OPC); \u003cem\u003eMBP\u003c/em\u003e – oligodendrocytes (Olig); \u003cem\u003ePDGFRB\u003c/em\u003e, \u003cem\u003ePECAM1\u003c/em\u003e, \u003cem\u003eCEMIP\u003c/em\u003e– vasculature cells (Vasc); \u003cem\u003eAPBB1IP\u003c/em\u003e, \u003cem\u003eP2RY12\u003c/em\u003e, \u003cem\u003eMRC1\u003c/em\u003e – immune cells (Immune); \u003cem\u003eCFAP54\u003c/em\u003e – ependymal cells (Epe); \u003cem\u003eTTR\u003c/em\u003e, \u003cem\u003eHTR2C\u003c/em\u003e – epithelial cells of choroid plexus (Ch.Pl); \u003cem\u003eAQP4\u003c/em\u003e, \u003cem\u003eSCL1A2\u003c/em\u003e, \u003cem\u003eGFAP\u003c/em\u003e – astrocytes (Astro); \u003cem\u003eRELN\u003c/em\u003e – Cajal Retzius cells (CR); \u003cem\u003eRBFOX3\u003c/em\u003e, \u003cem\u003eSYT1\u003c/em\u003e, \u003cem\u003eSV2B\u003c/em\u003e – neurons; \u003cem\u003eSLC17A7\u003c/em\u003e, \u003cem\u003eCAMK2A\u003c/em\u003e – excitatory neurons; \u003cem\u003ePROX1\u003c/em\u003e, \u003cem\u003eCALB1\u003c/em\u003e – granule neurons (GN); \u003cem\u003eEPHA5\u003c/em\u003e, \u003cem\u003eNEUROD6\u003c/em\u003e \u0026nbsp;– \u0026nbsp;Cornu Ammonis excitatory neurons (CA.ExN); \u003cem\u003eTSHZ2\u003c/em\u003e – \u0026nbsp;other excitatory neurons (O.ExN); \u003cem\u003eGAD1\u003c/em\u003e, \u003cem\u003eGAD2\u003c/em\u003e, \u003cem\u003eDCX\u003c/em\u003e, \u003cem\u003eSCL32A1\u003c/em\u003e – inhibitory neurons (InN). (For detailed dotplot with 31 clusters and specific marker genes see Extended Data Fig. 4c). \u003cstrong\u003ef.\u003c/strong\u003e Dotplot depicting expression of marker RNA transcripts. \u003cstrong\u003eg.\u003c/strong\u003e Dotplot depicting predicted gene activity based on chromatin accessibility. \u003cstrong\u003eh-i.\u003c/strong\u003e Anchor-based integration of single-nuclei gene expression data with \u003cstrong\u003e(h)\u003c/strong\u003e whole-transcriptome spatial gene expression and \u003cstrong\u003e(i) \u003c/strong\u003esingle-cell in-situ hybridization with 266 probes.\u003c/p\u003e","description":"","filename":"Fig1.png","url":"https://assets-eu.researchsquare.com/files/rs-8593764/v1/fdb8aba17864cdf320260776.png"},{"id":100546281,"identity":"0cb08fd1-0566-410e-942f-199e07c3f3eb","added_by":"auto","created_at":"2026-01-19 08:04:50","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":3376983,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eVisium spatial transcriptomics and Xenium in-situ hybridization, anchor-based integration of single nuclei data, and RNA Velocity. a. \u003c/strong\u003eExample of Visium capture area with manual annotation of hippocampal subfields. \u003cstrong\u003eb.\u003c/strong\u003eSubfield annotations projected onto Visium section at a spot level. \u003cstrong\u003ec.\u003c/strong\u003e Visium unsupervised clustering annotation identifies 13 distinct clusters on the tissue section.\u003cstrong\u003e d. \u003c/strong\u003eUMAP visualization of 39,230 spots from 12 spatial gene expression samples. UMAP is derived from anchor-based integration. \u0026nbsp;\u003cstrong\u003ec-d. \u003c/strong\u003eCluster color-coding is maintained. \u003cstrong\u003ee.\u003c/strong\u003e Anchor-based integration of whole transcriptome spatial gene expression and single-nuclei (sn) gene expression provides a probabilistic transfer of annotations from reference (sn) to query (spatial) to annotate 55um spatial spots. Color shows probability score. From top to bottom: \u003cem\u003eCornu Ammonis\u003c/em\u003e (CA) excitatory neurons (CA.ExN) have high probability score on CA regions, other excitatory neuron (O.ExN) have low probability score on CA regions, vasculature (Vasc) is diffuse across section. \u003cstrong\u003ef\u003c/strong\u003e. Anchor-based integration of snRNA-seq data (reference) with single-cell in-situ hybridization using 266 probes over 12x24 mm capture area (query) The larger capture area allows the spatial differentiation of CA.ExN in CA regions from O.ExN in subiculum. X and Y-axes represent microns. \u003cstrong\u003eg-j.\u003c/strong\u003e UMAP projection of subset of Visium unsupervised clusters gcl, sgz.pl, and sgz.ml \u003cstrong\u003eg.\u003c/strong\u003e with stream of velocity vectors from sgz.ml to sgz.pl to gcl estimated using RNA Velocity dynamical modeling. \u003cstrong\u003eh. \u003c/strong\u003ePartition-based graph abstraction (PAGA) with velocity-directed edges, a much simpler abstracted graph, in which edge weight represents confidence in the presence of connections, providing an easily interpretable overview of lineage trajectory. \u003cstrong\u003ei.\u003c/strong\u003e UMAP projection colored by latent time which represents the internal time of the spot and is based only on its transcriptional dynamics. \u003cstrong\u003ej.\u003c/strong\u003e UMAP colored by pseudotime which is calculated from the velocity graph. The root spot, or ‘zero’ is implicitly inferred (i.e. not used as an input parameter).\u003cstrong\u003e k.\u003c/strong\u003eHeatmap of gene expression (yellow = high, dark blue = low) by latent time, showing spots ordered by latent time on x-axis and genes on the y-axis.\u003c/p\u003e","description":"","filename":"Fig2.png","url":"https://assets-eu.researchsquare.com/files/rs-8593764/v1/de34f605ac3c0c3bf5ae778d.png"},{"id":100426034,"identity":"74f9bb4a-9ba5-4ef6-bdf5-240ffd33fd57","added_by":"auto","created_at":"2026-01-16 14:19:10","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":1722870,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eachine learning approach and Palantir trajectory inference to investigate neurogenic lineage | . a. \u003c/strong\u003eSchematic representation of gene expression in prototypical cells used to train and test the LASSO regression. \u003cstrong\u003eb.\u003c/strong\u003e UMAP plot of TypeI, TypeII, TypeIII, ImGN, and GN cells using batch-, sample-, and donor-corrected Harmony principal components, colored by cell type. \u003cstrong\u003ec.\u003c/strong\u003e Palantir-inferred pseudotime showing TypeI as youngest, TypeII and TypeIII following, and ImGN and GN as oldest. \u003cstrong\u003ed.\u003c/strong\u003e Violin plots of gene expression in cell types identified as TypeI, TypeII, TypeIII, and ImGN by LASSO regression. CA.ExN and O.ExN are merged to ExN. \u0026nbsp;\u003cstrong\u003ee. \u003c/strong\u003eGene expression trends for the neurogenic lineage using Palantir Generalized Additive Models. \u0026nbsp;Color represents the z-scored scale gene trends of imputed expression. \u003cstrong\u003ef.\u003c/strong\u003e Coverage plots showing accessible chromatin peaks in each cell type with violin plot to showing RNA expression. From left to right: \u003cem\u003ePAX6\u003c/em\u003e, \u003cem\u003eSOX2\u003c/em\u003e, and \u003cem\u003ePROX1\u003c/em\u003e. CA.ExN and O.ExN are merged. \u003cstrong\u003eg.\u003c/strong\u003e Anchor based integration of snRNA-seq data (reference) with Xenium single-cell in-situ hybridization (query). ImGN in red are in the subgranular zone (SGZ) and granule cell layer (GCL). X and Y-axes represent microns. \u003cstrong\u003eh.\u003c/strong\u003e Two probe-based protocols (RNAScope and Xenium) used to validate co-expression of \u003cem\u003ePROX1\u003c/em\u003e (red), \u003cem\u003eGAD1\u003c/em\u003e (green), and \u003cem\u003eDCX\u003c/em\u003e (yellow) in cells located in the SGZ and GCL of the dentate gyrus.\u003c/p\u003e","description":"","filename":"Fig3.png","url":"https://assets-eu.researchsquare.com/files/rs-8593764/v1/51068f9c4ad0d593f82c0d0d.png"},{"id":100426078,"identity":"70efd02b-3369-4901-9e64-204e3086d0db","added_by":"auto","created_at":"2026-01-16 14:19:11","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1208613,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCell type-specific differential gene expression, chromatin accessibility, and predicted gene activity in MDD versus CTRL. \u0026nbsp;a.\u003c/strong\u003e Distribution of differentially expressed genes in clusters (left) and cell types (right). Upregulated genes on the right of the dashed line and downregulated gene on the left. Clusters Olig, OPC, Epe, and Ch.Pl were not aggregated, so they are not shown again in the cell type graph. Clusters or cell types with no differential gene expression in MDD are not shown. \u003cstrong\u003eb.\u003c/strong\u003e Dotplot showing log(cpm)-adjusted counts for top 15 differentially expressed genes in MDD ImGN (left) and GN (right). \u003cstrong\u003ec.\u003c/strong\u003e Distribution of differentially accessible chromatin peaks in MDD in clusters (left) and cell types (right). Regions more accessible are on the right of the dashed line and less accessible on the left. \u003cstrong\u003ed.\u003c/strong\u003e Dotplot showing log(cpm)-adjusted counts for top 30 differentially expressed accessible chromatin peaks in MDD GN. \u003cstrong\u003ee.\u003c/strong\u003eDistribution of differentially predicted gene activity in MDD calculated from accessible peaks counts in clusters (left) and cell types (right) with increased predicted gene activity in MDD on the right of the dashed line and decreased on the right. \u003cstrong\u003ef. \u003c/strong\u003eDotplot showing log(cpm)-adjusted counts for ImGN and GN differentially predicted gene activity in MDD calculated from accessible peaks counts.\u003c/p\u003e","description":"","filename":"Fig4.png","url":"https://assets-eu.researchsquare.com/files/rs-8593764/v1/9165f5c65e121254e41da4a9.png"},{"id":100425876,"identity":"1cb84e8f-73d5-46ce-aaf6-72e5c0601066","added_by":"auto","created_at":"2026-01-16 14:19:04","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1208613,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePathway dysregulation in MDD. a. \u003c/strong\u003ePsyGeNET database reported gene-disease association network plot for all DEGs from snRNA-seq data in our sample and genes previously linked to psychiatric disorders. \u003cstrong\u003eb.\u003c/strong\u003e Bar graph of gene-disease association showing number of genes associated with disease (blue) and number of genes uniquely associated with each disease (yellow). \u003cstrong\u003ec.\u003c/strong\u003e Gene-disease association heatmap for GN DEGs (note that THEMIS is also a DEG in ImGN). Evidence index of 1 (blue) indicates that all literature support association. Evidence index between 0 and 1 (grey) indicate partial support for association.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ed-g. \u003c/strong\u003eGene set enrichment analysis for ImGN and GN. Genes are ranked using log(fold change)-log10 (Benjamini-Hochberg adjusted p-value). Terms in blue are enriched for genes upregulated in MDD. Terms in red are enriched for genes downregulated in MDD. \u003cstrong\u003e\u0026nbsp;d.\u003c/strong\u003e KEGG and REACTOME pathway dysregulation in ImGN\u003cstrong\u003e. e.\u003c/strong\u003e Hallmark pathway terms dysregulated in ImGN. \u003cstrong\u003ef. \u003c/strong\u003eKEGG and Reactome pathway dysregulation in GN. \u003cstrong\u003eg.\u003c/strong\u003e Hallmark pathway terms dysregulated in GN\u003cstrong\u003e h.\u003c/strong\u003e\u0026nbsp;Weighted gene co-expression network analysis (WGCNA) heatmap showing module eigengenes (ME) correlation and associated\u0026nbsp;\u003cem\u003ep\u003c/em\u003e-value (in parentheses) for ImGN, with MDD-CTRL status and covariates (age, RIN, PMI). \u003cstrong\u003ei. \u003c/strong\u003eThe turquoise module was positively associated with MDD in ImGN and was enriched for lipid modification and metabolism and the blue module was enriched for RNA splicing, RNA metabolism, translation and response to stress, and protein folding. \u003cstrong\u003ej.\u003c/strong\u003e The blue module negatively associated with MDD in ImGN was enriched metabolism, and lipid modifications. \u0026nbsp;\u003cstrong\u003ek.\u003c/strong\u003e Heatmap showing the correlation and associated\u0026nbsp;\u003cem\u003ep\u003c/em\u003e-value (in parentheses) of GN WGCNA module eigengenes with MDD-CTRL status and covariates (age, pH, PMI). \u003cstrong\u003el. \u003c/strong\u003eThe turquoise module positively associated with MDD in GN was enriched for cell morphogenesis and projection development, and VEGF pathways. \u003cstrong\u003em. \u003c/strong\u003eThe blue module negatively associated with MDD in GN was enriched for lipid and metabolic processes.\u003c/p\u003e","description":"","filename":"Fig5.png","url":"https://assets-eu.researchsquare.com/files/rs-8593764/v1/3a56fa2af822db166069afde.png"},{"id":100553626,"identity":"bd34fc66-66c2-4637-a2e1-cdb842e3e19f","added_by":"auto","created_at":"2026-01-19 08:38:12","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":11493448,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8593764/v1/64c30fde-9ffb-44ad-980d-cbdda7160618.pdf"},{"id":100426150,"identity":"a78edeac-508f-40f8-a645-adab3d61c2b6","added_by":"auto","created_at":"2026-01-16 14:19:13","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":21436407,"visible":true,"origin":"","legend":"","description":"","filename":"BoldriniHippocampusMultiomeExtendedDataFigures.docx","url":"https://assets-eu.researchsquare.com/files/rs-8593764/v1/09e7ff6606a6349bb4f07cea.docx"}],"financialInterests":"The authors declare potential competing interests as follows: HG and her family own stocks in Illumina, Inc.; JJM receives royalties from Columbia University for the Columbia Pathways App. and from the Research Foundation of Mental Hygiene for commercial use of the C-SSRS.","formattedTitle":"Human hippocampus niche and circuit neurotypical and major depression profiles","fulltext":[{"header":"Main","content":"\u003cp\u003eMajor depressive disorder (MDD) is recurrent and episodic, often treatment resistant\u003csup\u003e2\u003c/sup\u003e, a major cause of disability worldwide, and the most frequent diagnosis in those who die by suicide\u003csup\u003e1\u003c/sup\u003e. Human brain postmortem studies found smaller hippocampus dentate gyrus (DG) and fewer granule neurons (GNs)\u003csup\u003e3,13\u003c/sup\u003e in MDD, which may reflect impaired neuroplasticity resulting in MDD symptoms like impaired emotional dysregulation and memory\u003csup\u003e2\u003c/sup\u003e. Moreover, hippocampus neuronal and neuropil loss were correlated with MDD severity and illness duration\u003csup\u003e3,13\u003c/sup\u003e. \u003cem\u003eIn vivo\u003c/em\u003e brain imaging confirmed smaller hippocampal volume in MDD compared to controls\u003csup\u003e14\u003c/sup\u003e. It remains unknown whether smaller hippocampal volume and diminished neuroplasticity in MDD are the result of increased apoptosis, decreased AHN, or other cellular and molecular changes.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eDeficits in AHN have been linked to depressive-like behavior in rodents\u003csup\u003e15\u003c/sup\u003e. Conversely, adult born GNs support stress recovery and emotional regulation\u003csup\u003e16\u003c/sup\u003e, enhance hippocampal plasticity, and circuit rewiring\u003csup\u003e17\u003c/sup\u003e. Transcriptomics studies agreed that rodent markers are inadequate to identify progenitor cells and their trajectory in humans and non-human primates\u003csup\u003e5,7,8\u003c/sup\u003e. In humans, early studies showed proliferating cells becoming\u0026nbsp;GNs\u003csup\u003e18\u003c/sup\u003e and estimated neuron birth date based on \u003csup\u003e14\u003c/sup\u003eC decay indicated substantial AHN in people who had been subjected to radioactive exposure\u003csup\u003e19\u003c/sup\u003e. Nevertheless, human AHN remains debated\u003csup\u003e20,21\u003c/sup\u003e, as immunofluorescence (IF), immunohistochemistry (IHC)\u003csup\u003e22-24\u003c/sup\u003e, \u003cem\u003ein situ\u003c/em\u003e hybridization (ISH)\u003csup\u003e23,25\u003c/sup\u003e, and transcriptome sequencing studies\u003csup\u003e5-8\u003c/sup\u003e led to opposite conclusions, likely driven by tissue processing methods, subjects\u0026rsquo; medical history, mode of death, and psychotropic drugs exposure\u003csup\u003e26\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo resolve this debate, provide new insights into the molecular characteristics of human hippocampus cell types, and discover the molecular basis of cellular disruption in MDD, we assessed 106 subjects and for the first time profiled almost 350,000 cells in untreated MDD subjects (MDDs) and neurotypical controls (CTRL) with multi-omic single-nucleus RNA sequencing (snRNA-seq) and assay for transposase-accessible chromatin with sequencing (snATAC-seq), spatial RNA-seq, and extensive molecular marker validation (Fig. 1a). Here we present results from the largest and most deeply sequenced single-nucleus multiome and spatial transcriptome resource dataset of neurotypical and MDD human hippocampus cell types to date\u003csup\u003e27\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSingle-nucleus multiome and spatial transcriptome profiling of adult human hippocampus\u0026nbsp;\u003c/strong\u003eWe simultaneously generated snRNA-seq and snATAC-seq data on 349,847\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003enuclei from deeply clinically phenotyped sudden death subjects with high-quality brain tissue (52 samples from 18 subjects, 9 male MDDs and 9 male CTRL, with 53% of nuclei from MDDs; Fig. 1a-d, Supplementary Tables 1-3; Methods). Number of samples per donor, cells profiled per donor, RNA integrity number, post-mortem interval, and age did not differ between MDDs and CTRL (Extended Data Fig. 1a-e). The medium number of unique molecular indices, chromatin regions, and genes per donor did not differ between MDDs and CTRL (Extended Data Fig. 1f-i). We used Harmony\u003csup\u003e28\u003c/sup\u003e to correct for sample, donor, and batch in RNA and ATAC data separately (Extended Data Fig. 2). \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo integrate snRNA-seq and snATAC-seq data, we used weighted nearest neighbor (WNN) analysis\u003csup\u003e29\u003c/sup\u003e (Fig. 1b-d), an unsupervised method in Seurat (v5) to obtain a unified representation of snRNA-seq and snATAC-seq data. We identified 31 unsupervised clusters, representing 12 broader cell types (Fig. 1b-c), that were not driven by batch or donor effects (Fig. 1b-d; Extended Data Fig. 3a-d). In each cluster and cell type, average GEX and predicted GEX inferred by accessible chromatin (or gene activity) data were strongly correlated with each other (Fig. 1e; Extended Data Fig. 4e) and used to annotate cell types based on their top-expressed genes (Fig. 1f-g; Extended Data Fig. 4c-d; Methods; Supplementary Table 4). Clusters included oligodendrocyte progenitor cells (OPC), oligodendrocytes (Olig), endothelial cells (Endo), pericytes (Peri), vascular leptomeningeal cells (VLMC), microglia (Micro), macrophages (Macro), T-cells (Tcell), choroid plexus cells (Ch.Pl), subventricular zone (SVZ) ependymal cells (Epe), Cajal Retzius cells (CR), three astrocyte (Astro1-3), nine inhibitory neuron (InN.SST, InN.LHX6, InN.PVALB, InN.VIP, InN.LAMP5, InN.SLC17A8, InN.PENK, InN.TUBB3, InN.PROX1), six non-granule excitatory neuron (ExN1-6), and two GN (GN1, GN2) clusters (Fig. 1d, Extended Data Fig. 4a-d). These 31 cluster annotations aligned with those from recent human hippocampus snRNA-seq studies that did not include ATAC sequencing\u003csup\u003e6-8\u003c/sup\u003e (Extended Data Fig. 5; Supplementary Table 5). The 12 broader cell types included: OPC, Olig, Ch.Pl, Epe, and CR (identical to the clusters), and merged clusters of vasculature cells (Vasc; including Endo, Peri, and VLMC), immune system cells (Immune; including Micro, Macro, and Tcell), astrocytes (Astro; including all three astrocyte clusters), granule neurons (GN, including both GN clusters), inhibitory neurons (InN, including all nine InN clusters), CA excitatory neurons (CA.ExN, including ExN1, ExN 3) and other excitatory neurons (O.ExN, including ExN2, ExN4-6 that did not map on CA regions, Fig. 2e-f; Supplementary Table 6).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSpatial GEX is concordant with single-nucleus cell type annotations\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo investigate GEX in whole tissue, including cytoplasm and neuronal processes, and provide anatomical mapping of GEX onto hippocampal subfields, we used Visium (10X Genomics) on intact hippocampus proper tissue sections (6 subjects, 2 samples each; 3 male MDDs and 3 male CTRL; Supplementary Table 7). First, we used histology (Fig. 2a) to guide hippocampus manual anatomical annotation into granule cell layer (GCL), subgranular zone (SGZ), molecular layer (ML), CA1-4, and white matter (WM, Fig. 2b). Then we performed unsupervised clustering of Visium spots (55x55 \u0026micro;m in size) which provided a more detailed spot-type annotation than manual annotation of hippocampal subfields and delivered 13 clusters that expressed genes concordant with expected cell types based on spot location in the respective subfields (Fig. 2c-d; Supplementary Table 8). Unsupervised spot clusters included (Methods): 2 hilus clusters (hilus1, hilus2), CA3 stratum pyramidales (ca3), stratum oriens-stratum lucidum (so-slu), so-slu of the CA3 (ca3-so-slu), stratum radiatum (sr), stratum oriens-stratum lacunosum (so-sla), stratum moleculare (ml), stratum granulosum (gcl), SGZ-polymorphic layer (sgz-pl), stratum moleculare (sgz-ml).\u003c/p\u003e\n\u003cp\u003eWe further used Visium for validation of single-nucleus cell type annotations, performing anchor-based integration of sn-multiome-based cell types (reference) with whole-transcriptome spatial GEX (query), and found that Olig and Astro cell types mapped in WM, GN in GCL (Fig. 1h), and CA.ExN in CA regions, while O.ExN had low prediction score in CA regions, and vasculature was spread across the tissue (Fig. 2e). Different ExN clusters mapped on specific subfields, with ExN1 mapping on CA1, ExN3 on CA3, while Ch.Pl and Epe correctly mapped on choroid plexus and SVZ (Extended Data Fig. 4f).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWe then used Xenium multiplex ISH (10X Genomics, 266 probes, 10.45x22.45 mm capture area) for additional anchor-based integration of single-nuclei cell types. While CA.ExN mapped on CA1-4, O.ExN mapped on the subiculum (Fig. 2e-f), which was not possible to verify with Visium because the smaller capture area could not include the subiculum. Xenium confirmed the expected location of Astro, Olig, GN (Fig. 1i), and Vasc (Fig. 2f).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSpatial RNA Velocity identifies a cell trajectory from the SGZ to the GCL\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo infer cell trajectories based on proportions of spliced and unspliced RNAs, we applied RNA Velocity analysis to spatial RNA-seq rather than to snRNA-seq data because the method relies on both cytoplasmic and nuclear RNAs to infer transcription and degradation rates\u003csup\u003e11\u003c/sup\u003e. Since we hypothesized that, if a neurogenic trajectory exists, it would be in the DG, to infer RNA Velocity, we selected spots located in the SGZ, GCL, and ML (Fig. 2g; Extended Data Fig. 6; Methods). The partition-based graph abstraction (PAGA) analysis identified a lineage from spatial cluster sgz.ml to cluster sgz.pl and then to gcl (Fig. 2h), indicating that sgz.ml could have the most immature cells with cell bodies in the SGZ and apical dendrites in ML, likely radial-glia-like quiescent neural stem cells\u003csup\u003e30\u003c/sup\u003e. The next cluster on the trajectory was sgz.pl likely containing intermediate progenitors located in SGZ. Ultimately the gcl spatial cluster would comprise GN at different dynamic functional states. Latent time (Fig. 2i) and pseudotime calculated from RNA Velocity (Fig. 2j) agreed with time zero corresponding to cluster sgz.ml and increasing time latency going towards clusters sgz.ml and then gcl. GEX-by-latent-time analysis showed high expression of \u003cem\u003ePCNA\u003c/em\u003e, \u003cem\u003eHESX1\u003c/em\u003e, \u003cem\u003eMKI67\u003c/em\u003e, and \u003cem\u003eENOX1\u003c/em\u003e at the earliest stage (sgz.ml spatial cluster), \u003cem\u003eDCX\u003c/em\u003e, \u003cem\u003eCALB2\u003c/em\u003e, \u003cem\u003eSOX11\u003c/em\u003e, \u003cem\u003eGAD1\u003c/em\u003e, \u003cem\u003eGAD2\u003c/em\u003e, and \u003cem\u003eSTMN2\u003c/em\u003e at the intermediate stage (sgz.pl spatial cluster), and \u003cem\u003eCAMK2A\u003c/em\u003e, \u003cem\u003eNEUROG3\u003c/em\u003e, \u003cem\u003eCAMK2A\u003c/em\u003e, \u003cem\u003eSNAP25\u003c/em\u003e, \u003cem\u003ePPFIA2\u003c/em\u003e, and \u003cem\u003eRBFOX3\u003c/em\u003e at the final stage of maturation (gcl spatial cluster, Fig. 2k).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSupervised machine learning characterization of neurogenic cell trajectory\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo determine the presence of a human hippocampus neurogenic niche at single-nucleus resolution, we used a supervised machine learning approach. Supervised models have previously been used to distinguish transcriptionally ambiguous cell subtypes\u003csup\u003e9,10\u003c/sup\u003e, including those of the neurogenesis trajectory\u003csup\u003e8\u003c/sup\u003e. We scored all 349,847 high quality nuclei using normalized counts of genes known to be expressed in mammalian hippocampus quiescent neural stem cells (Type I), amplifying neural progenitors (Type II), intermediate neural progenitors (Type III), and immature GN\u0026nbsp;(ImGN;\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eFig. 3a) as per rodent and human literature\u003csup\u003e5,8,20,30,31\u003c/sup\u003e. A percentage of high confidence cells were selected (Methods). High confidence cells with high expression of required marker genes and low expression of genes typically not expressed in that cell type\u003csup\u003e5,7,8,20,30\u003c/sup\u003e were used to train the models: for Type I we used \u003cem\u003eGFAP\u003c/em\u003e, \u003cem\u003eSOX2\u003c/em\u003e, \u003cem\u003ePAX6\u003c/em\u003e, \u003cem\u003eNES\u003c/em\u003e, \u003cem\u003eFABP7\u003c/em\u003e, \u003cem\u003eHES5\u003c/em\u003e, and \u003cem\u003eETNPPL\u003c/em\u003e; Type II had \u003cem\u003eASCL1\u003c/em\u003e, \u003cem\u003eFOXO3\u003c/em\u003e, \u003cem\u003eNR2E1\u003c/em\u003e, \u003cem\u003eEOMES\u003c/em\u003e, \u003cem\u003eNEUROG2\u003c/em\u003e, \u003cem\u003eNES\u003c/em\u003e, \u003cem\u003eMCM2\u003c/em\u003e, \u003cem\u003ePCNA\u003c/em\u003e, \u003cem\u003eMKI67\u003c/em\u003e, \u003cem\u003eTOP2A\u003c/em\u003e, and not \u003cem\u003eMBP\u003c/em\u003e or \u003cem\u003eMOBP\u003c/em\u003e, to distinguish them from oligodendrocyte progenitors; for Type III we used \u003cem\u003eMYT1L\u003c/em\u003e, \u003cem\u003ePROX1\u003c/em\u003e, \u003cem\u003eDCX\u003c/em\u003e, \u003cem\u003eCALB2\u003c/em\u003e, \u003cem\u003eRELN\u003c/em\u003e, \u003cem\u003eST8SI2\u003c/em\u003e, \u003cem\u003eST8SIA4\u003c/em\u003e, and not mature neuron marker \u003cem\u003eRBFOX3\u003c/em\u003e, or inhibitory neuron marker \u003cem\u003eVIP\u003c/em\u003e; for ImGN,\u0026nbsp;\u003cem\u003eRBFOX3\u003c/em\u003e, \u003cem\u003ePROX1\u003c/em\u003e, \u003cem\u003eNEUROD1\u003c/em\u003e, \u003cem\u003eDCX\u003c/em\u003e, \u003cem\u003eBHLHE22\u003c/em\u003e, \u003cem\u003eCOL25A1\u003c/em\u003e, and \u003cem\u003ePOSTN\u003c/em\u003e. Cells with 85% prediction score or higher were annotated as TypeI (4,413 cells, 7.47% of GN) TypeII (1,680 cells, 2.83% of GN), TypeIII (4,423 cells, 7.49% of GN) and ImGN (5,611 cells, 9.50% of GN) and clustered separately (Fig. 3b). The LASSO approach extracted these cells from the original clusters (Supplementary Table 9). TypeI derived from Astro, as radial-glia-like progenitors are most like astrocytes\u003csup\u003e30\u003c/sup\u003e. TypeII derived mostly from OPC, Astro, and Endo, as those clusters contain proliferating cells. TypeIII derived from InN and CR clusters. CR cells have a prominent role in brain development\u003csup\u003e32\u003c/sup\u003e and express CALB2 and TUBB3 and not SYT1 (Extended Data Fig. 4c-d), not typical of adult excitatory CR. InN clusters expressed immature neuron markers: \u003cem\u003eCALB2\u003c/em\u003e (InN.VIP, InN.PROX1), \u003cem\u003eTUBB3\u003c/em\u003e (InN.TUBB3), \u003cem\u003eRELN\u003c/em\u003e (InN.LAMP5), \u003cem\u003eDCX\u0026nbsp;\u003c/em\u003e(all InN subtypes), and \u003cem\u003ePROX1\u003c/em\u003e (InN.PROX1, InN.VIP, InN.LAMP5; Fig. 1f-g, Extended Data Fig. 4c,d). Moreover, \u003cem\u003eGAD1\u003c/em\u003e and \u003cem\u003eGAD2\u003c/em\u003e, classically used as InN markers, have transcript variants expressed during brain development until neurons fully differentiate\u003csup\u003e33\u003c/sup\u003e. Using Xenium and RNAscope (ACDbio), we examined the expression of \u003cem\u003eGAD1\u003c/em\u003e in the DG, finding \u003cem\u003eGAD1\u003c/em\u003e co-expressed with \u003cem\u003ePROX1\u003c/em\u003e and \u003cem\u003eDCX\u003c/em\u003e in the GCL (Fig. 3h). Finally, ImGN were pulled mostly from GN, in line with a recent publication\u003csup\u003e8\u003c/sup\u003e.\u0026nbsp;To validate spatial location of ImGN, we used anchor-based integration of single-nuclei GEX (reference) and Xenium ISH (query) and found that ImGN are in SGZ and GCL (Fig. 3g).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe maturity stages of the machine learning-identified cell types were reflected by gene ontology terms showing TypeI enriched for cell fate commitment, TypeII for stem cell differentiation and neural precursor cell proliferation, TypeIII for neuron migration and axon development, and ImGN for cell growth and learning or memory (Supplementary Table 10) .\u003c/p\u003e\n\u003cp\u003eIn a larger sample (Supplementary Table 1), using RNAscope, we validated expression of DCX in human SGZ and GCL as previously shown\u003csup\u003e21\u003c/sup\u003e, and demonstrated DCX co-localizing with NEUROD1, PROX1, TUBB3 and RBFOX3 (Extended Data. Fig. 7a-g). Using IHC, we confirmed expression of Nestin and Ki67 proteins in human SGZ, DCX protein in human SGZ and GCL, and NeuN protein in GCL, as previously reported\u003csup\u003e18\u003c/sup\u003e. Using double IF, we replicated co-localization of DCX with PSA-NCAM\u003csup\u003e18\u003c/sup\u003e, and showed DCX co-labeling with neurofilament, calretinin, calbindin, and NeuN, and no DCX co-localization with glial markers GFAP and IBA1 (Extended Data Fig. 7-8).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSingle-nucleus pseudotime trajectory inference supports a human neurogenic lineage\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo test if machine learning-identified neurogenic cell types TypeI, TypeII, TypeIII, and ImGN belonged to a differentiation trajectory, we used Palantir\u003csup\u003e12\u003c/sup\u003e to model differentiation capturing the continuity in cell states and cell fate determination (Fig. 3c-e). We included TypeI, TypeII, TypeIII, ImGN, and the cells remaining in the GN cell type after applying the LASSO model (Fig. 3b). Palantir assigned pseudotime in concordance with our hypothesis and ordered cell types from TypeI, TypeII, TypeIII to ImGN and GN (Fig. 3c). GEX along the neurogenic lineage included \u003cem\u003ePAX6\u003c/em\u003e, \u003cem\u003eETNPPL\u003c/em\u003e, and \u003cem\u003eSOX2\u003c/em\u003e in TypeI quiescent progenitors; \u003cem\u003eMKI67\u003c/em\u003e, \u003cem\u003eTOP2A\u003c/em\u003e, \u003cem\u003ePCNA\u003c/em\u003e in TypeII proliferating progenitors; \u003cem\u003eDCX\u003c/em\u003e, \u003cem\u003eGAD1\u003c/em\u003e, \u003cem\u003eGAD2\u003c/em\u003e, \u003cem\u003eEOMES\u003c/em\u003e, \u003cem\u003eTUBB3\u003c/em\u003e, \u003cem\u003ePROX1\u003c/em\u003e, \u003cem\u003eSTMN1\u003c/em\u003e in TypeIII neuroblasts and ImGN; and finally, \u003cem\u003eSCL17A7\u003c/em\u003e, \u003cem\u003eCALB1\u003c/em\u003e, \u003cem\u003eCAMK2A\u003c/em\u003e, \u003cem\u003eSYT1\u003c/em\u003e, and \u003cem\u003eRBFOX3\u003c/em\u003e in mature GN, also corresponding to open chromatin at those genes (Fig. 3d-f).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eMoreover, Palantir lineage cell GEX is concordant with progressive GEX in Visium spots ordered by latent time based on RNA Velocity (Fig. 2k), and they both suggest a phase in which \u003cem\u003eGAD1\u003c/em\u003e and \u003cem\u003eGAD2\u003c/em\u003e are expressed in TypeIII. In fact, all InN expressed \u003cem\u003eDCX\u003c/em\u003e; InN.LAMP5 expressed \u003cem\u003eRELN\u003c/em\u003e; InN.VIP and InN.PROX1 expressed \u003cem\u003eCALB2\u003c/em\u003e; InN.TUBB3 expressed \u003cem\u003eTUBB3\u003c/em\u003e; and InN.PROX1, InN.VIP, and InN.LAMP5 expressed \u003cem\u003ePROX1\u003c/em\u003e (Fig. 1f; Extended Data Fig. 4c).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCell-type specific gene expression and chromatin accessibility are altered in MDD\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo understand molecular underpinnings of hippocampus anatomical and functional disruption in MDD\u003csup\u003e2,3,13,14\u003c/sup\u003e, we analyzed cell-type specific differentially expressed genes (DEGs), differentially accessible chromatin regions (DARs), and predicted differentially active genes (DAGs) both in the 31 clusters and the 16 cell types (OPC, Olig, Ch.Pl, Epe, CR, Vasc, Immune, Astro, TypeI, TypeII, TypeIII, ImGN, GN, InN, CA.ExN, O.ExN, Supplementary Tables 11-13).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe clusters with the highest number of DEGs were GN2, GN1, ExN2, Epe, OPC, and Oligo, (Fig. 4a-b). GN1 and GN2 had 75 and 43 DEGs, contributing to 13.2% and 7.6% of all DEGs respectively. Epe and ExN2 had 182 DEGs, both contributing to 32% of DEGs. Regarding DARs, ExN1 had the most with 497 hits (87.5% of all), Oligo and GN1 had 39 and 36, contributing to 6.9% and 6.3% of DARs respectively (Fig. 4c-d). As for predicted DAGs, 97.6% were identified in OPCs, and GN1, GN2, and Oligo had 15, 13, and 13 respectively (Fig. 4e-f).\u003c/p\u003e\n\u003cp\u003eWhen analyzing cell types, GN had 101 DEGs, O.ExN \u0026nbsp;had 38, ImGN had 20, TypeI and TypeIII each had one DEG: \u003cem\u003eLINC02315\u003c/em\u003e (upregulated) and \u003cem\u003eZNF257\u003c/em\u003e (upregulated) respectively (Fig. 4a-b). DARs were the highest in Immune (14227 DARs), followed by GN (41 DARs; Fig. 4c-d). DAGs were 60 in GN, 41 in O.ExN, 2 in ImGN, and one in TypeI and TypeIII (\u003cem\u003eACOT1,\u003c/em\u003e upregulated in MDDs, Fig. 4e-f). Taken together, GN had the most consistently dysregulated genes across modalities.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePsyGeNET Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo explore which of the identified dysregulated genes had previously been associated with MDD, we used PsyGeNET text mining database\u003csup\u003e34\u003c/sup\u003e. In all clusters, we found 43 genes with DEGs and DAGs overlapping with psychiatric disease-associated genes in PsyGeNET, and significant for depressive disorders (p=0.033017458) with 15 overlapping genes (\u003cem\u003eADCY8\u003c/em\u003e, \u003cem\u003eATF3\u003c/em\u003e, \u003cem\u003eDISC1\u003c/em\u003e, \u003cem\u003eEMP1\u003c/em\u003e, \u003cem\u003eENPEP\u003c/em\u003e, \u003cem\u003eFOLH1\u003c/em\u003e, \u003cem\u003eFOS\u003c/em\u003e, \u003cem\u003eGRIN2A\u003c/em\u003e, \u003cem\u003eGRK5\u003c/em\u003e, \u003cem\u003eIL18\u003c/em\u003e, \u003cem\u003eIL19\u003c/em\u003e, \u003cem\u003eIL6R\u003c/em\u003e, \u003cem\u003eSOD2\u003c/em\u003e, \u003cem\u003eSYNDIG1\u003c/em\u003e), and for alcohol use disorder (pl=0.009390476), with 83 overlapping genes (Fig. 5a-b, Supplementary Table 14).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWhen analyzing cell types, GN had 7 DEGs previously linked to psychiatric disorders, including depression, schizophrenia, and mood disorders (\u003cem\u003eADCY8\u003c/em\u003e, \u003cem\u003eCCND2\u003c/em\u003e, \u003cem\u003eDNAH8\u003c/em\u003e, \u003cem\u003eGABBR1\u003c/em\u003e, \u003cem\u003eHLA-DQB1\u003c/em\u003e, \u003cem\u003ePRKCB\u003c/em\u003e, \u003cem\u003eTHEMIS\u003c/em\u003e; Fig. 5c), without any significant disease-specific associations.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eBiological pathways of cell-type specific DEGs in MDD\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo infer pathways associated with cell-type specific multi-omics changes in MDD, we performed pre-ranked gene set enrichment analysis (GSEA, Supplementary Table 15). We focused on GN that had the most dysregulated genes across modalities, and ImGN to understand why GN are fewer in MDD\u003csup\u003e3,4,13\u003c/sup\u003e. In ImGN, pathways related to interferon signaling, cellular stress, Golgi, MTORC1, sex hormone response and apoptosis were enriched, while lipid homeostasis and glycosylation were negatively regulated in MDD (Fig. 5d-e). In GN, pathways related to PTEN, MTORC1, apoptosis, sex hormone and immune response, and thyroid, Parkinson, Huntington and Alzheimer\u0026rsquo;s diseases were enriched, and lipid homeostasis was negatively regulated in MDD (Fig. 5f-g). Notably, in both ImGN and GN, apoptosis was enriched in MDD with top genes \u003cem\u003eCCND2\u003c/em\u003e, \u003cem\u003eJUN\u003c/em\u003e, \u003cem\u003eMADD\u003c/em\u003e, \u003cem\u003eBCAP31\u003c/em\u003e, \u003cem\u003eDNAJA1\u003c/em\u003e, \u003cem\u003eNEFH\u003c/em\u003e, \u003cem\u003ePEA15\u003c/em\u003e, \u003cem\u003eDAP\u003c/em\u003e, \u003cem\u003eENO2\u003c/em\u003e, \u003cem\u003eBMP2\u003c/em\u003e (Supplementary Table 15).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eWeighted gene co-expression\u0026nbsp;networks dysregulation in MDD\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWhile differential GEX analysis provides information on single-gene expression differences, MDD is highly polygenic\u003csup\u003e35\u003c/sup\u003e and disease state may be driven by gene network changes rather than single-gene alterations\u003csup\u003e36\u003c/sup\u003e. Therefore,\u0026nbsp;we used pseudo-bulk GEX in GN and ImGN to perform weighted-gene co-expression network analysis (WGCNA)\u003csup\u003e37\u003c/sup\u003e and determine clusters of highly correlated genes that have an association with MDD. \u0026nbsp;Each cluster was represented by a color-coded module eigengene (ME), the first principal component of the module representing the overall expression level of the module.\u003c/p\u003e\n\u003cp\u003eIn ImGN, six out of seven modules had correlations with MDD (Fig. 5h-j; Extended Data Fig. 9a-d). Four ME were positively associated and two were negatively associated with MDD. MEblue, enriched for RNA metabolism, splicing, translation and response to cellular stress, was positively correlated with MDD (correlation 0.99, p=1\u0026times;\u0026thinsp;10\u003csup\u003e\u0026minus;40\u003c/sup\u003e), in line with\u0026nbsp;GSEA findings described above. The MEturquoise module, enriched for lipid modification and metabolism, was negatively correlated with MDD (correlation -0.99, p=2\u0026times;\u0026thinsp;10\u003csup\u003e\u0026minus;47\u003c/sup\u003e) underscoring GSEA findings of lipid homeostasis being negatively regulated in MDD ImGN and GN (Fig. 5e-g).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn GN, 8 modules out of 9 had correlations with MDD (Fig. 5k-m; Extended Data Fig. 9e-h). Three ME were positively associated and five were negatively associated with MDD. The MEturquoise module, enriched for lipid modification and metabolism, was positively correlated with MDD (correlation 0.99, p=2\u0026thinsp;\u0026times;\u0026thinsp;10\u003csup\u003e\u0026minus;43\u003c/sup\u003e) and showed significant overlap (p=2\u0026thinsp;\u0026times;\u0026thinsp;10\u003csup\u003e\u0026minus;04\u003c/sup\u003e) with upregulated DEGs in GN, and with\u0026nbsp;GSEA findings\u0026nbsp;(Fig. 5e-g). The MEblue module, enriched for development and cellular morphogenesis, was negatively correlated with MDD (correlation -0.99, p=1\u0026times;\u0026thinsp;10\u003csup\u003e\u0026minus;50\u003c/sup\u003e) and showed significant overlap (\u003cem\u003ep\u003c/em\u003e=7\u0026thinsp;\u0026times;\u0026thinsp;10\u003csup\u003e\u0026minus;06\u003c/sup\u003e) with downregulated DEGs in GN, and with GSEA findings (Fig. 5e-g).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eQuantification of target marker RNA and protein expression in MDD\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn separate and enlarged samples, using unbiased stereology, we quantified number of cells expressing neurogenic lineage RNAs and proteins labeled by multiplex ISH and IF, and single IHC, and used quantitative real-time polymerase chain reaction (qPCR) to quantify RNA in bulk tissue (Methods;\u0026nbsp;Supplementary Table 16, Extended Data Fig. 7-8).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eRNAscope DCX+/TUBB3+ cells were fewer in MDDs (N=11) \u003cem\u003evs.\u003c/em\u003e CTRL (N=11) in GCL (t=2.294; p=0.04) but not SGZ (Extended Data Fig. 7n). DCX RNA measured by qPCR was lower in MDDs (N=5) \u003cem\u003evs.\u003c/em\u003e CTRL (N=5) in anterior (t=3.371, p=0.0145) but not posterior DG (Extended Data Fig. 7o).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eSingle-IHC Ki67+ cells were fewer in MDDs (N=17) \u003cem\u003evs.\u003c/em\u003e CTRL (N=23) in anterior (t=2.780; p=0.0084) but not mid or posterior DG (Extended Data Fig 8p). MDDs had fewer Nestin+ cells in anterior (34 MDDs, 38\u003cem\u003e\u0026nbsp;\u003c/em\u003eCTRL, t=4.360; p\u0026lt;0.0001) and posterior DG (31 MDD, 28 CTRL, t=3.654; pval=0.0008) but not in mid DG (Extended Data Fig. 7q). Nissl+ glial cells were not different between MDDs and CTRL in any region (Extended Data Fig. 7r). Single-IHC DCX+ cells were fewer in MDDs (N=22) \u003cem\u003evs.\u003c/em\u003e CTRL (N=36) in anterior (t=3.547; p=0.0008), but not in mid or posterior DG (Extended Data Fig. 7s). MDD had fewer single-IHC NeuN+ GNs in anterior (35 MDDs, 32 CTRL, t=2.951; p=0.004) and mid DG (28 MDDs, 26 CTRL, t=2.410; p=0.01), but not in posterior DG (Extended Data Fig. 7t).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIF-stained DCX+/NeuN- cells were fewer in MDDs (N=8) \u003cem\u003evs.\u003c/em\u003e CTRL (N=8) in SGZ (welch-corrected t [t]= 2.318; p=0.04) but not GCL, and DCX+/NeuN+ cells did not differ between groups (Extended Data Fig. 8l-o).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eProteomics\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe further used high resolution shotgun proteomics liquid chromatography with tandem mass spectrometry (LC/MS/MS)\u003csup\u003e38\u003c/sup\u003e,\u0026nbsp;to quantify expression of proteins corresponding to DEGs in MDDs \u003cem\u003evs.\u003c/em\u003e CTRL as per snRNA-seq data (Methods;\u0026nbsp;Extended Data Fig. 9i-m; Supplementary Table 17).\u003c/p\u003e\n\u003cp\u003eAfter running proteomics as we described\u003csup\u003e38\u003c/sup\u003e, Elucidator (Methods) detected 80,504 peptides across 48 runs and identified 1811 proteins with protein score at a 4% false discovery rate (Extended Data Fig. 9i-m; Supplementary Table 17). The 1131 proteins with two or more peptides and protein scores greater than 250 were used for the analyses.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWe found 308 differentially expressed proteins between MDDs and CTRL, of which four corresponded to detected DEGs: CPNE8, HSDL2, NCH1, and RAC1 (Extended Data Fig. 9n).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eHSDL2, a key factor of fatty acid regulation in lipid metabolism\u003csup\u003e39\u003c/sup\u003e, was upregulated at both the RNA level in GN and the protein level in bulk tissue.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eRAC1, important in neurodevelopmental function including neurogenesis, synaptogenesis, and plasticity and expressed in migrating cortical neurons and adult neurons\u003csup\u003e40\u003c/sup\u003e, was lower at the protein level in MDDs \u003cem\u003evs.\u003c/em\u003e CTRL. RAC1 is negatively regulated by ARHGAP15\u003csup\u003e41\u003c/sup\u003e, whose genomic sequence overlaps with long non-coding RNA \u003cem\u003eAC079793.2\u003c/em\u003e, also downregulated in ImGN in MDD.\u003c/p\u003e\n\u003cp\u003eConversely, CPNE8 regulating cell adhesion\u003csup\u003e42\u003c/sup\u003e showed downregulated RNA expression in MDD GN and upregulated protein expression in bulk MDD tissue.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eSimilarly, NCEH, a protector from \u0026alpha;-synuclein neurotoxicity\u003csup\u003e43\u003c/sup\u003e, was downregulated at the RNA level in Epe but the protein was upregulated in bulk hippocampus tissue in MDD.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eEmploying the first single-nuclei multiome and spatial transcriptome profiling of human anterior hippocampus cells, combined with shotgun proteomics in anterior hippocampus tissue, we compared sudden death neurotypical adult subjects and unmedicated patients with MDD.\u003c/p\u003e\n\u003cp\u003eImplementing a machine learning approach for rare cells identification, we detected quiescent (Type I), amplifying (Type II), and intermediate (Type III) progenitor cells, and replicated the identification of immature neurons\u003csup\u003e8\u003c/sup\u003e. Our integrated analysis identified distinct clusters representing all hippocampal cell types, in agreement with other publications\u003csup\u003e6\u0026ndash;8\u003c/sup\u003e, identified a neurogenic lineage from early progenitors to immature granule neurons, and defined cell type-specific gene expression and chromatin accessibility dysregulation in MDD. Importantly, the annotation of these cell types, including those of the neurogenesis trajectory, was validated by spatial GEX profiling using Visium, RNAScope, and Xenium, and by protein marker expression identification by double IF and IHC, providing robust evidence of cell identity and relative cellular and spatial distribution within the hippocampus.\u003c/p\u003e\n\u003cp\u003eCell trajectory inference using both single-nuclei and spatial transcriptome sequencing, supported the neurogenic cell-lineage trajectory, and gene expression corresponding to maturational stages described in mice\u003csup\u003e5,7\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eFindings suggest transcriptional and epigenetic similarities of astroglia-like (TypeI) progenitors with astrocytes, amplifying progenitors (TypeII) with other dividing cells (OPC, Astro and Endo).\u003c/p\u003e\n\u003cp\u003eThese data do not demonstrate when cells of the neurogenic trajectory were generated but the expression of proliferation markers (\u003cem\u003ePCNA\u003c/em\u003e, \u003cem\u003eMKI67\u003c/em\u003e, \u003cem\u003eTOP2A\u003c/em\u003e; Fig.\u0026nbsp;2k, 3e) in TypeII suggests the presence of mitosis.\u003c/p\u003e\n\u003cp\u003eTypeIII origination from GAD\u0026thinsp;+\u0026thinsp;neurons initially classified as InN, is largely due to DCX expression in GAD\u0026thinsp;+\u0026thinsp;cells, and warrants further investigation on transitional GAD expression in maturing cells.\u003c/p\u003e\n\u003cp\u003eWe demonstrated for the first time \u003cem\u003eGAD1\u003c/em\u003e expression in TypeIII, colocalizing with \u003cem\u003eDCX\u003c/em\u003e, \u003cem\u003eTUBB3\u003c/em\u003e and \u003cem\u003ePROX1\u003c/em\u003e, and \u003cem\u003eDCX\u003c/em\u003e mapping onto SGZ and GCL, in line with findings that GAD1 alternative transcripts are expressed in human hippocampus during neuronal development\u003csup\u003e33,44\u003c/sup\u003e. Additionally, GABAergic basket cells could be generated in human DG as demonstrated in adult rats\u003csup\u003e45\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eImGN coming from the GN cluster has been previously observed. Notably the number of ImGN is similar to what previously estimated (close to 10% of all GNs), suggesting cells might remain in this immature status for a long time, which has profound implications for the functional significance of AHN in the human brain.\u003c/p\u003e\n\u003cp\u003eThe presence of immaturity markers suggests a regenerative potential of the human SGZ. It remains unknown how prolonged the immature state is and how long it takes to make a human neuron. Different sequencing technologies, including epigenome analysis and deeper sequencing, may explain our finding the full neurogenic lineage when others have not\u003csup\u003e6\u0026ndash;8\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eThis validity of our trajectory reconstruction is strengthened by RNA Velocity and pseudotime analysis inferring of cell trajectory, and its identified molecular signatures that characterize neural stem cell activation, proliferation, differentiation, and ultimately, neuronal maturation, matching those found in the literature\u003csup\u003e20,26,30\u003c/sup\u003e. These findings provide a valuable resource for future studies investigating the regulation of AHN and its potential modulation for therapeutic purposes.\u003c/p\u003e\n\u003cp\u003eIn MDD, elevated inflammation, response to cellular stress and apoptosis pathways in ImGN and GN suggests the molecular basis of smaller \u003cem\u003ein vivo\u003c/em\u003e hippocampal volume\u003csup\u003e14\u003c/sup\u003e and postmortem findings of fewer GNs\u003csup\u003e4,13\u003c/sup\u003e and agree with finding in MDD PFC\u003csup\u003e27\u003c/sup\u003e and patient blood\u003csup\u003e46\u0026ndash;48\u003c/sup\u003e. Overlap between DEGs and PsyGeNET MDD and alcohol use disorder genes, may explain the comorbidity between the two.\u003c/p\u003e\n\u003cp\u003eUpregulation of \u003cem\u003eLINC02315\u003c/em\u003e in TypeI and OPC in MDD could prevent proliferation or transition to the amplifying stage because this long-intergenic non-coding RNA is protective in lung squamous cell carcinoma by blocking proliferation\u003csup\u003e49\u003c/sup\u003e. In TypeIII, InN and Olig, upregulation of \u003cem\u003eZNF257\u003c/em\u003e which has over 1,000 binding sites in gene promoters and is believed to regulate transcription in the developing brain\u003csup\u003e50\u003c/sup\u003e, may contribute to increased nucleic acid binding, transcription and Vitamin D receptor pathway activity in MDD affecting developmental processes. Dysregulation of MTORC1 in ImGN and GN supports altered cell growth and metabolism\u003csup\u003e2\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eUsing both proteomics and transcriptomics, we find genes dysregulated at both levels. Upregulated HSDL2, a key factor of fatty acid regulation in lipid metabolism\u003csup\u003e39\u003c/sup\u003e, at the RNA level in GN and protein level in bulk tissue suggests lipid metabolism dysregulation in MDD. Downregulation of both long non-coding RNA \u003cem\u003eAC079793.2\u003c/em\u003e in ImGN, overlapping in genomic sequence with \u003cem\u003eARHGAP15\u003c/em\u003e, and possibly negatively regulating it, and its target protein RAC1 in bulk tissue, suggest reduced neurogenesis, synaptogenesis, and plasticity\u003csup\u003e40\u003c/sup\u003e. The fact that \u003cem\u003eCPNE8\u003c/em\u003e, regulating cell adhesion\u003csup\u003e42\u003c/sup\u003e, and NCEH, protecting from \u0026alpha;-synuclein neurotoxicity\u003csup\u003e43\u003c/sup\u003e, have downregulated RNA expression in GN and Epe respectively, while the proteins are upregulated in bulk tissue, may be due to tissue-specific transcriptomic differences involving DG and Epe, or compensatory post-transcriptional mechanisms.\u003c/p\u003e\n\u003cp\u003eAlthough this is the first study combining single-nucleus multi-omic (RNA and ATAC) profiling, with spatial transcriptomic and bulk proteomics in human hippocampus, and comparing sudded death neurotypical controls with unmedicated depressed subjects that had MDD, limitations include small sample, not including women, possible clinical variability, and few subjects having received benzodiazepines prior to death \u0026ndash; although we excluded comorbidity and patients on antidepressants or antipsychotics. Results should be replicated in larger samples that include women and low testing effects of clinical and sociodemographic variables, because in animal models stress greatly affects neurogenesis.\u003c/p\u003e\n\u003cp\u003eOverall, this comprehensive single-nuclei multi-omic, spatial transcriptomic, and proteomic profiling of brain tissue supports the presence of a neurogenic niche and suggests molecular underpinnings of anterior hippocampus atrophy in MDD. Dysregulated markers will drive future mechanistic studies investigating their cellular, molecular, and behavioral effects, and will lead blood and brain imaging biomarker discovery and drug development for MDD and hippocampus trophic support.\u003c/p\u003e\n\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003cbr\u003e\u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003eData availability\u003c/p\u003e\n\u003cp\u003eThe single cell and spatial genomics data will be available at the GEX Omnibus database upon publication. Specimen information and sequencing statistics are described in Supplementary Tables 1-3 and 7. Sources of the published sequencing datasets used in this study are described in Supplementary Table 5.\u003c/p\u003e\n\u003cp\u003eRaw mass spectroscopy data are available at the MassIVE repository entry under dataset identifier MSV000094862 (reviewer password: reviewer052424).\u003c/p\u003e\n\u003cp\u003eThe final datasets for this study, which will be archived in our servers, will include demographic data, psychiatric assessment data, psychological ratings, and postmortem studies data. \u0026nbsp;Because the data we collect are sensitive in nature, we carefully ensure that appropriate privacy safeguards are in place. All data will be stripped of identifiers prior to release for sharing.\u003c/p\u003e\n\u003cp\u003eIn addition, we will provide relevant protocols and published genetic and phenotypic data upon request. Material transfers will be made with no more restrictive terms than in the Simple Letter Agreement (SLA) or the Uniform Biological Materials Transfer Agreement (UBMTA) and without reach through requirements.\u003c/p\u003e\n\u003cp\u003eShould any intellectual property arise which requires a patent, we will ensure that the technology (materials and data) remains widely available to the research community in accordance with the NIH Principles and Guidelines.\u003c/p\u003e\n\u003cp\u003eCode availability\u003c/p\u003e\n\u003cp\u003eThe computational code used in this study is available at GitHub or upon request.\u003c/p\u003e\n\u003cp\u003eAcknowledgements\u003c/p\u003e\n\u003cp\u003eWe thank the families for donating the brain tissue used in this study. This work is supported by NIH grants AI164769, AG076949, MH133561, AG080790. We thank Stefano Marenco and the Human Brain Collection Core at the National Institute of Mental Health, and Mark Underwood at the New York State Psychiatric Institute for providing some of the postmortem human brain tissue samples. We thank Erin Bush and Peter Sims and the Columbia Single Cell Analysis Core and JP Sulzberger Columbia Genome Center for their services. We thank Jeremy Worley and the Columbia Genome Engineering Core for Xenium (10X Genomics) experimentation and analysis services. We thank Vinod Gupta and the Columbia Center for Computational Biology and Bioinformatics for services in data storage and high-performance computing infrastructure usage. We thank Tao Su and the Molecular Pathology Shared Resource at Columbia for RNA extraction and quantification services. This study used the Confocal and Specialized Microscopy Shared Resource of the Herbert Irving Comprehensive Cancer Center at Columbia University, funded in part through the NIH/NCI Cancer Center Support Grant P30CA013696. Some schematic illustrations were created or modified using BioRender.com.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eContributions\u003c/p\u003e\n\u003cp\u003eMBM and CAS performed sequencing experiments, single-nuclei and spatial library generation, and bioinformatics. AMW performed WGCNA analysis. AR worked on spatial data analysis and data interpretation. YX advised on bioinformatics. ANT, CF, and AMW completed IHC, IF and ISH experiments. YL designed the machine learning approach. \u0026nbsp;WH advised on spatial transcriptomics approaches. YYH worked on nuclei isolation. MBD, GR, JJM, VA and AD provided the human postmortem specimens and clinical characterization of the subjects. STMD and LB worked on proteomics. RH helped with the interpretation of the neurogenesis findings. HG designed the biostatistics approach. MBD secured funding for this project, guided experimental design and approaches, worked on result interpretation and reporting strategy. MBM, and MBD wrote the first draft of the manuscript. All authors reviewed the manuscript.\u003c/p\u003e\n\u003cp\u003eCompeting Interests\u003c/p\u003e\n\u003cp\u003eHG and her family own stocks in Illumina, Inc.; JJM receives royalties from Columbia University for the Columbia Pathways App. and from the Research Foundation of Mental Hygiene for commercial use of the C-SSRS.\u003c/p\u003e\n\u003cp\u003eCorresponding author\u003c/p\u003e\n\u003cp\u003eCorrespondence to [email protected]\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cbr\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eJames, S. L.\u003cem\u003e et al.\u003c/em\u003e Global, regional, and national incidence, prevalence, and years lived with disability for 354 diseases and injuries for 195 countries and territories, 1990\u0026ndash;2017: a systematic analysis for the Global Burden of Disease Study 2017. \u003cem\u003eThe Lancet\u003c/em\u003e \u003cstrong\u003e392\u003c/strong\u003e, 1789-1858 (2018). \u003c/li\u003e\n\u003cli\u003eTartt, A. N., Mariani, M. B., Hen, R., Mann, J. J. \u0026amp; Boldrini, M. Dysregulation of adult hippocampal neuroplasticity in major depression: pathogenesis and therapeutic implications. \u003cem\u003eMol Psychiatry\u003c/em\u003e (2022). https://doi.org:10.1038/s41380-022-01520-y\u003c/li\u003e\n\u003cli\u003eCobb, J. A.\u003cem\u003e et al.\u003c/em\u003e Hippocampal volume and total cell numbers in major depressive disorder. \u003cem\u003eJ Psychiatr Res\u003c/em\u003e \u003cstrong\u003e47\u003c/strong\u003e, 299-306 (2013). https://doi.org:10.1016/j.jpsychires.2012.10.020\u003c/li\u003e\n\u003cli\u003eBoldrini, M.\u003cem\u003e et al.\u003c/em\u003e Resilience Is Associated With Larger Dentate Gyrus, While Suicide Decedents With Major Depressive Disorder Have Fewer Granule Neurons. \u003cem\u003eBiological Psychiatry\u003c/em\u003e \u003cstrong\u003e85\u003c/strong\u003e, 850-862 (2019). https://doi.org:https://doi.org/10.1016/j.biopsych.2018.12.022\u003c/li\u003e\n\u003cli\u003eWang, W.\u003cem\u003e et al.\u003c/em\u003e Transcriptome dynamics of hippocampal neurogenesis in macaques across the lifespan and aged humans. \u003cem\u003eCell Research\u003c/em\u003e \u003cstrong\u003e32\u003c/strong\u003e, 729-743 (2022). https://doi.org:10.1038/s41422-022-00678-y\u003c/li\u003e\n\u003cli\u003eAyhan, F.\u003cem\u003e et al.\u003c/em\u003e Resolving cellular and molecular diversity along the hippocampal anterior-to-posterior axis in humans. \u003cem\u003eNeuron\u003c/em\u003e \u003cstrong\u003e109\u003c/strong\u003e, 2091-2105.e2096 (2021). https://doi.org:10.1016/j.neuron.2021.05.003\u003c/li\u003e\n\u003cli\u003eFranjic, D.\u003cem\u003e et al.\u003c/em\u003e Transcriptomic taxonomy and neurogenic trajectories of adult human, macaque, and pig hippocampal and entorhinal cells. \u003cem\u003eNeuron\u003c/em\u003e \u003cstrong\u003e110\u003c/strong\u003e, 452-469. e414 (2022). \u003c/li\u003e\n\u003cli\u003eZhou, Y.\u003cem\u003e et al.\u003c/em\u003e Molecular landscapes of human hippocampal immature neurons across lifespan. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e607\u003c/strong\u003e, 527-533 (2022). https://doi.org:10.1038/s41586-022-04912-w\u003c/li\u003e\n\u003cli\u003eHabib, N.\u003cem\u003e et al.\u003c/em\u003e Massively parallel single-nucleus RNA-seq with DroNc-seq. \u003cem\u003eNature methods\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 955-958 (2017). \u003c/li\u003e\n\u003cli\u003eShekhar, K.\u003cem\u003e et al.\u003c/em\u003e Comprehensive Classification of Retinal Bipolar Neurons by Single-Cell Transcriptomics. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e166\u003c/strong\u003e, 1308-1323 e1330 (2016). https://doi.org:10.1016/j.cell.2016.07.054\u003c/li\u003e\n\u003cli\u003eLa Manno, G.\u003cem\u003e et al.\u003c/em\u003e RNA velocity of single cells. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e560\u003c/strong\u003e, 494-498 (2018). https://doi.org:10.1038/s41586-018-0414-6\u003c/li\u003e\n\u003cli\u003eSetty, M.\u003cem\u003e et al.\u003c/em\u003e Characterization of cell fate probabilities in single-cell data with Palantir. \u003cem\u003eNat Biotechnol\u003c/em\u003e \u003cstrong\u003e37\u003c/strong\u003e, 451-460 (2019). https://doi.org:10.1038/s41587-019-0068-4\u003c/li\u003e\n\u003cli\u003eBoldrini, M.\u003cem\u003e et al.\u003c/em\u003e Hippocampal Granule Neuron Number and Dentate Gyrus Volume in Antidepressant-Treated and Untreated Major Depression. \u003cem\u003eNeuropsychopharmacology\u003c/em\u003e \u003cstrong\u003e38\u003c/strong\u003e, 1068-1077 (2013). https://doi.org:10.1038/npp.2013.5\u003c/li\u003e\n\u003cli\u003eNolan, M.\u003cem\u003e et al.\u003c/em\u003e Hippocampal and Amygdalar Volume Changes in Major Depressive Disorder: A Targeted Review and Focus on Stress. \u003cem\u003eChronic Stress\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, 2470547020944553 (2020). https://doi.org:10.1177/2470547020944553\u003c/li\u003e\n\u003cli\u003eSurget, A.\u003cem\u003e et al.\u003c/em\u003e Antidepressants recruit new neurons to improve stress response regulation. \u003cem\u003eMol.Psychiatry\u003c/em\u003e (2011). \u003c/li\u003e\n\u003cli\u003eAnacker, C.\u003cem\u003e et al.\u003c/em\u003e Hippocampal neurogenesis confers stress resilience by inhibiting the ventral dentate gyrus. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e559\u003c/strong\u003e (2018). https://doi.org:https://doi.org/10.1038/s41586-018-0262-4\u003c/li\u003e\n\u003cli\u003eTemprana, S. G.\u003cem\u003e et al.\u003c/em\u003e Delayed coupling to feedback inhibition during a critical period for the integration of adult-born granule cells. \u003cem\u003eNeuron\u003c/em\u003e \u003cstrong\u003e85\u003c/strong\u003e, 116-130 (2015). https://doi.org:10.1016/j.neuron.2014.11.023\u003c/li\u003e\n\u003cli\u003eEriksson, P. S.\u003cem\u003e et al.\u003c/em\u003e Neurogenesis in the adult human hippocampus. \u003cem\u003eNat.Med.\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, 1313-1317 (1998). \u003c/li\u003e\n\u003cli\u003eSpalding, K. L.\u003cem\u003e et al.\u003c/em\u003e Dynamics of hippocampal neurogenesis in adult humans. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e153\u003c/strong\u003e, 1219-1227 (2013). https://doi.org:10.1016/j.cell.2013.05.002\u003c/li\u003e\n\u003cli\u003eGage, F. H. Adult neurogenesis in mammals. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e364\u003c/strong\u003e, 827-828 (2019). \u003c/li\u003e\n\u003cli\u003eTosoni, G.\u003cem\u003e et al.\u003c/em\u003e Mapping human adult hippocampal neurogenesis with single-cell transcriptomics: Reconciling controversy or fueling the debate? \u003cem\u003eNeuron\u003c/em\u003e \u003cstrong\u003e111\u003c/strong\u003e, 1714-1731. e1713 (2023). \u003c/li\u003e\n\u003cli\u003eBoldrini, M.\u003cem\u003e et al.\u003c/em\u003e Human hippocampal neurogenesis persists throughout aging. \u003cem\u003eCell stem cell\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 589-599. e585 (2018). \u003c/li\u003e\n\u003cli\u003eSorrells, S. F.\u003cem\u003e et al.\u003c/em\u003e Positive controls in adults and children support that very few, if any, new neurons are born in the adult human hippocampus. \u003cem\u003eJournal of Neuroscience\u003c/em\u003e \u003cstrong\u003e41\u003c/strong\u003e, 2554-2565 (2021). \u003c/li\u003e\n\u003cli\u003eMoreno-Jim\u0026eacute;nez, E. P., Terreros-Roncal, J., Flor-Garc\u0026iacute;a, M., R\u0026aacute;bano, A. \u0026amp; Llorens-Mart\u0026iacute;n, M. Evidences for adult hippocampal neurogenesis in humans. \u003cem\u003eJournal of Neuroscience\u003c/em\u003e \u003cstrong\u003e41\u003c/strong\u003e, 2541-2553 (2021). \u003c/li\u003e\n\u003cli\u003eTartt, A. N.\u003cem\u003e et al.\u003c/em\u003e Considerations for Assessing the Extent of Hippocampal Neurogenesis in the Adult and Aging Human Brain. \u003cem\u003eCell Stem Cell\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 782-783 (2018). https://doi.org:10.1016/j.stem.2018.10.025\u003c/li\u003e\n\u003cli\u003eKempermann, G.\u003cem\u003e et al.\u003c/em\u003e Human adult neurogenesis: evidence and remaining questions. \u003cem\u003eCell stem cell\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 25-30 (2018). \u003c/li\u003e\n\u003cli\u003eMaitra, M.\u003cem\u003e et al.\u003c/em\u003e Cell type specific transcriptomic differences in depression show similar patterns between males and females but implicate distinct cell types and genes. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 2912 (2023). https://doi.org:10.1038/s41467-023-38530-5\u003c/li\u003e\n\u003cli\u003eKorsunsky, I.\u003cem\u003e et al.\u003c/em\u003e Fast, sensitive and accurate integration of single-cell data with Harmony. \u003cem\u003eNature methods\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 1289-1296 (2019). \u003c/li\u003e\n\u003cli\u003eHao, Y.\u003cem\u003e et al.\u003c/em\u003e Integrated analysis of multimodal single-cell data. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e184\u003c/strong\u003e, 3573-3587.e3529 (2021). https://doi.org:https://doi.org/10.1016/j.cell.2021.04.048\u003c/li\u003e\n\u003cli\u003eMing, G.-l. \u0026amp; Song, H. Adult neurogenesis in the mammalian brain: significant answers and significant questions. \u003cem\u003eNeuron\u003c/em\u003e \u003cstrong\u003e70\u003c/strong\u003e, 687-702 (2011). \u003c/li\u003e\n\u003cli\u003eAbbott, L. C. \u0026amp; Nigussie, F. Adult neurogenesis in the mammalian dentate gyrus. \u003cem\u003eAnatomia, Histologia, Embryologia\u003c/em\u003e \u003cstrong\u003e49\u003c/strong\u003e, 3-16 (2020). https://doi.org:https://doi.org/10.1111/ahe.12496\u003c/li\u003e\n\u003cli\u003eGil-Sanz, C.\u003cem\u003e et al.\u003c/em\u003e Cajal-Retzius cells instruct neuronal migration by coincidence signaling between secreted and contact-dependent guidance cues. \u003cem\u003eNeuron\u003c/em\u003e \u003cstrong\u003e79\u003c/strong\u003e, 461-477 (2013). https://doi.org:10.1016/j.neuron.2013.06.040\u003c/li\u003e\n\u003cli\u003eTao, R.\u003cem\u003e et al.\u003c/em\u003e GAD1 alternative transcripts and DNA methylation in human prefrontal cortex and hippocampus in brain development, schizophrenia. \u003cem\u003eMol Psychiatry\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 1496-1505 (2018). https://doi.org:10.1038/mp.2017.105\u003c/li\u003e\n\u003cli\u003eGuti\u0026eacute;rrez-Sacrist\u0026aacute;n, A.\u003cem\u003e et al.\u003c/em\u003e Text mining and expert curation to develop a database on psychiatric diseases and their genes. \u003cem\u003eDatabase (Oxford)\u003c/em\u003e \u003cstrong\u003e2017\u003c/strong\u003e (2017). https://doi.org:10.1093/database/bax043\u003c/li\u003e\n\u003cli\u003eKendall, K. M.\u003cem\u003e et al.\u003c/em\u003e The genetic basis of major depression. \u003cem\u003ePsychol Med\u003c/em\u003e \u003cstrong\u003e51\u003c/strong\u003e, 2217-2230 (2021). https://doi.org:10.1017/s0033291721000441\u003c/li\u003e\n\u003cli\u003ePerez, J. M.\u003cem\u003e et al.\u003c/em\u003e Hippocampal subfield transcriptome analysis in schizophrenia psychosis. \u003cem\u003eMolecular Psychiatry\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 2577-2589 (2021). https://doi.org:10.1038/s41380-020-0696-6\u003c/li\u003e\n\u003cli\u003eLi, J.\u003cem\u003e et al.\u003c/em\u003e Application of Weighted Gene Co-expression Network Analysis for Data from Paired Design. \u003cem\u003eScientific reports\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 622 (2018). https://doi.org:10.1038/s41598-017-18705-z\u003c/li\u003e\n\u003cli\u003eWerth, E. G., Rajbhandari, P., Stockwell, B. R. \u0026amp; Brown, L. M. Time Course of Changes in Sorafenib-Treated Hepatocellular Carcinoma Cells Suggests Involvement of Phospho-Regulated Signaling in Ferroptosis Induction. \u003cem\u003eProteomics\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, e2000006 (2020). https://doi.org:10.1002/pmic.202000006\u003c/li\u003e\n\u003cli\u003eYang, Y.\u003cem\u003e et al.\u003c/em\u003e Lipid metabolism regulator human hydroxysteroid dehydrogenase-like 2 (HSDL2) modulates cervical cancer cell proliferation and metastasis. \u003cem\u003eJ Cell Mol Med\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 4846-4859 (2021). https://doi.org:10.1111/jcmm.16461\u003c/li\u003e\n\u003cli\u003eLiaci, C.\u003cem\u003e et al.\u003c/em\u003e Loss of ARHGAP15 affects the directional control of migrating interneurons in the embryonic cortex and increases susceptibility to epilepsy. \u003cem\u003eFront Cell Dev Biol\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 875468 (2022). https://doi.org:10.3389/fcell.2022.875468\u003c/li\u003e\n\u003cli\u003eZamboni, V.\u003cem\u003e et al.\u003c/em\u003e Disruption of ArhGAP15 results in hyperactive Rac1, affects the architecture and function of hippocampal inhibitory neurons and causes cognitive deficits. \u003cem\u003eScientific Reports\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 34877 (2016). https://doi.org:10.1038/srep34877\u003c/li\u003e\n\u003cli\u003eZhang, P.\u003cem\u003e et al.\u003c/em\u003e CPNE8 Promotes Gastric Cancer Metastasis by Modulating Focal Adhesion Pathway and Tumor Microenvironment. \u003cem\u003eInt J Biol Sci\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 4932-4949 (2022). https://doi.org:10.7150/ijbs.76425\u003c/li\u003e\n\u003cli\u003eZhang, S., Glukhova, S. A., Caldwell, K. A. \u0026amp; Caldwell, G. A. NCEH-1 modulates cholesterol metabolism and protects against \u0026alpha;-synuclein toxicity in a C. elegans model of Parkinson\u0026apos;s disease. \u003cem\u003eHum Mol Genet\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 3823-3836 (2017). https://doi.org:10.1093/hmg/ddx269\u003c/li\u003e\n\u003cli\u003eRamnauth, A. D.\u003cem\u003e et al.\u003c/em\u003e Spatially-resolved transcriptomics of human dentate gyrus across postnatal lifespan reveals heterogeneity in markers for proliferation, extracellular matrix, and neuroinflammation. \u003cem\u003ebioRxiv\u003c/em\u003e (2023). https://doi.org:10.1101/2023.11.20.567883\u003c/li\u003e\n\u003cli\u003eLiu, S.\u003cem\u003e et al.\u003c/em\u003e Generation of functional inhibitory neurons in the adult rat hippocampus. \u003cem\u003eJournal of Neuroscience\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 732-736 (2003). \u003c/li\u003e\n\u003cli\u003eFelger, J. C.\u003cem\u003e et al.\u003c/em\u003e What does plasma CRP tell us about peripheral and central inflammation in depression? \u003cem\u003eMolecular Psychiatry\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 1301-1311 (2020). https://doi.org:10.1038/s41380-018-0096-3\u003c/li\u003e\n\u003cli\u003eEnache, D., Pariante, C. M. \u0026amp; Mondelli, V. Markers of central inflammation in major depressive disorder: A systematic review and meta-analysis of studies examining cerebrospinal fluid, positron emission tomography and post-mortem brain tissue. \u003cem\u003eBrain, Behavior, and Immunity\u003c/em\u003e \u003cstrong\u003e81\u003c/strong\u003e, 24-40 (2019). https://doi.org:https://doi.org/10.1016/j.bbi.2019.06.015\u003c/li\u003e\n\u003cli\u003eZhan, Y.\u003cem\u003e et al.\u003c/em\u003e Alterations of multiple peripheral inflammatory cytokine levels after repeated ketamine infusions in major depressive disorder. \u003cem\u003eTranslational Psychiatry\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 246 (2020). https://doi.org:10.1038/s41398-020-00933-z\u003c/li\u003e\n\u003cli\u003eLiu, J., Yao, Y., Hu, Z., Zhou, H. \u0026amp; Zhong, M. Transcriptional profiling of long‐intergenic noncoding RNAs in lung squamous cell carcinoma and its value in diagnosis and prognosis. \u003cem\u003eMolecular genetics \u0026amp; genomic medicine\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, e994 (2019). \u003c/li\u003e\n\u003cli\u003eFarmiloe, G., Lodewijk, G. A., Robben, S. F., van Bree, E. J. \u0026amp; Jacobs, F. M. J. Widespread correlation of KRAB zinc finger protein binding with brain-developmental gene expression patterns. \u003cem\u003ePhilos Trans R Soc Lond B Biol Sci\u003c/em\u003e \u003cstrong\u003e375\u003c/strong\u003e, 20190333 (2020). https://doi.org:10.1098/rstb.2019.0333 \u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003eBrain Collection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eBrain tissue was obtained from the Brain Collection of the New York State Psychiatric Institute (NYSPI) at Columbia University, which includes brain samples from the Republic of Macedonia\u003csup\u003e1\u003c/sup\u003e, and the Human Brain Collection Core (HBCC) of the National Institute of Mental Health (NIMH). Brain tissue collection was conducted with IRB approval and informed consent obtained from next of kin who agreed to donate the brains and participate in psychological autopsy interviews\u003csup\u003e2\u003c/sup\u003e which provide diagnoses according to the Diagnostic Statistical Manual of Mental Disorders (DSM) through structured proxy-based interviews comparable in reliability to the Structured Clinical Interview for DSM diagnoses (SCID) in living patients, and has been extensively used in postmortem studies\u003csup\u003e3-7\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSubjects\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe included 106 subjects with sudden death, no neuropathology including microvascular disease, short agonal state, no resuscitation with prolonged (\u0026gt;10 min) hypoxia, no alcohol or drug use disorder, or chronic illness that may affect CNS function (e.g. epilepsy, renal failure, metastatic malignancy, HIV, Supplementary Table 1).\u003c/p\u003e\n\u003cp\u003eAll 51 subjects with Major Depressive Disorder (MDD) included in this study had a diagnosis of MDD, a major depressive episode within 4 months of death, were not in remission at death, had no history of treatment during the three months before death, and had negative toxicology for psychoactive drugs, except for 15 subjects with positive toxicology for benzodiazepines that could have been given at the time of death in the emergency room, or previously prescribed.\u003c/p\u003e\n\u003cp\u003eAll 55 controls included in this study had no neurological or psychiatric diagnoses, no pathological or traumatic condition affecting the brain, clear toxicology reports, and had no history of psychiatric treatments or suicide attempts.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eBrain tissue processing\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eRight hemispheres were sliced at 2-cm intervals, flash frozen and stored at -80\u0026ordm;C at the time of death. Routine neuropathological examination and toxicology exams were performed. If the subject was 45 years old or older, hippocampal formation and neocortex are stained for senile plaques and neurofibrillary tangles.\u003c/p\u003e\n\u003cp\u003eThe whole hippocampal formation was dissected from two or three consecutive 2 cm-thick frozen coronal blocks. Frozen adjacent sections of 50 m thickness were collected in tubes for snRNA/ATAC-seq, and two adjacent sections of 14 m were mounted on barcoded slides for Visium spatial transcriptomics (10X Genomics) per donor. Frozen tissue sections (16 m) were also used for qPCR, RNAscope\u0026reg; Duplex, Multiplex and Hiplex in situ hybridization (ACDbio), and for Western Blots (WB). Tissue fixed in 4% paraformaldehyde (length of tissue fixation was 5\u0026rsquo; per each mm of block thickness) was used for immunohistochemistry (IHC) and immunofluorescence (IF). Supplementary Table 1 describes experiments performed in each sample.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eNuclei isolation\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe adapted methods from Maitra et al. 2021\u003csup\u003e8\u003c/sup\u003e for isolating high-quality nuclei from human frozen brain tissue. Using a glass douce tissue homogenizer (Wheaton, 357542), approximately 90 mg of tissue were homogenized in 5mL lysis buffer (10mM tris-HCl pH 7.4, 10mM NaCl, 3mM MgCl\u003csub\u003e2\u0026nbsp;\u003c/sub\u003e6H\u003csub\u003e2\u003c/sub\u003eO, 0.05% NP-40) and incubated on ice for 5 minutes. Then, 5mL nuclei wash buffer (NWB: 5% BSA, 0.25% glycerol, 0.001% protector RNase inhibitor (MilliporeSigma, cat. no. 333539901) in 1x PBS) were added to sample to quench the lysis. Homogenate was filtered through a 40\u003cem\u003e\u0026nbsp;\u003c/em\u003e m Corning cell strainer (Millipore Sigma, cat. no. CLS431750). Filtrate was centrifuged three times at 500g for 5 minutes at 4\u0026deg;C. After each incubation, pellet was resuspended in NWB. Final resuspension was ultracentrifuged at 10,000g for 30 minutes at 4\u0026deg;C over iodixanol cushion (Millipore Sigma, cat. no. D1556-250). Nuclei pellet was resuspended in less than 100 L nuclei wash buffer (10X Genomics). Nuclei concentration and cell viability were estimated using trypan blue and DAPI in Countess II Hemocytometer (Applied Biosystems). Nuclei concentrations varied between 2-9 10\u003csup\u003e6\u003c/sup\u003e nuclei/mL.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLibrary preparation and sequencing\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eApproximately 16,000 nuclei for each sample were loaded into the 10X Chromium Controller immediately after quantification. Gene expression (GEX) and Assay for Transposase assessable chromatin (ATAC) libraries were prepared according to Chromium Next GEM Single Cell Multiome ATAC + GEX (CG000338) User Guide. We generated 2-4 libraries per sample. The libraries were sequenced on an Illumina NovaSeq 6000 using v4 chemistry to an average sequencing depth of 80,611 reads per cell for RNA libraries and 37,119 reads per cell for ATAC libraries (Supplementary Tables 3 and 7).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePre-processing and filtering multiome data\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA count matrix and a fragment file were generated by 10X CellRanger (v2.0.0) pipeline with the chemistry parameter set to Single Cell Multiome ATAC + GEX v1 (Supplementary Table 3)\u003cem\u003e.\u003c/em\u003e The reads were aligned to the GRCh38 human genome. RNA and ATAC data for each sample were preprocessed in parallelly using Seurat\u003csup\u003e9,10\u003c/sup\u003e (V5) and Signac\u003csup\u003e11\u003c/sup\u003e (V1)R packages. The count matrix was filtered for cells that have greater than 500 total number of unique transcripts and 200 genes expressed detected to remove dead or dying cells, less than 10% mitochondrial DNA, less than 150,000 UMIs, and less than 20,000 genes expressed to remove multiplets, an artifact that occurs when two or more cells are encapsulated in the same GEM (Gel beads-in-EMulsion) and are tagged with the same barcode (Extended Data Fig. 3e). Multiplets were then removed by considering canonical markers of expected cell types. Clusters that expressed multiple markers of more than one cell type will be removed. Remaining cells were pre-processed again using the standard Seurat workflow. For all samples, peaks BED (Browser Extensible Data) files output from CellRanger were converted to GenomicRanges\u003csup\u003e12\u003c/sup\u003e, an R/Bioconductor package for manipulating genomic intervals and combined to create a unified peak set representing all samples in the dataset. Peaks longer than 10,000 and shorter than 20 bases were removed. Barcodes for cells that were previously identified to be multiplets, dying, or quality were removed. Cells with transcription start site (TSS) enrichment score less than 1, nucleosome signal greater than 3, ATAC read counts less than 200 and greater than 100,000 were removed (Extended Data Fig. 3e). Cells that passed quality control (QC) in both modalities were retained for further analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMulti-modal integration\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn RNA and ATAC modalities, 52 samples were merged. Genes and peaks expressed in less than 0.01% of cells that passed QC were removed. For chromatin accessibility data, we followed the Signac workflow for merging multiple objects. Peaks on nonstandard chromosomes and in blacklist regions from the hg38 genome were removed. Merged gene expression data were processed as follows: to account for different number of counts per gene in each cell, the gene counts for each cell were divided by the sum of counts for that cell, multiplied by 10,000 and natural log transformed using Seurat function NormalizeData with default parameters. The intersection of the top 2,000 variable genes and a list of 93 genes involved in neurogenesis and neural maturation (sum: 2,065 genes) was passed to Seurat function ScaleData for linear transformation, to center mean expression across cells for each gene to zero, and scale expression of each gene, so that highly expressed genes were not given more weight in downstream analysis. Principal component analysis (PCA) was run (RunPCA) with features set to these 2,065 genes, to prevent overfitting. Otherwise, default parameters were used. Fifty PCA cell embeddings were passed to Seurat\u0026rsquo;s wrapper for Harmony\u003csup\u003e13\u003c/sup\u003e with \u003cem\u003egroup.by.vars\u003c/em\u003e set to sample, donor, and batch. The Harmony-corrected PCA embedding was used for UMAP (Uniform Manifold Approximation and Projection) and Louvain clustering (Extended Data Fig. 2). Similarly, accessible chromatin data are processed using Signac workflow. We used latent semantic indexing (LSI) first by term frequency-inverse document frequency normalization (TF-IDF) and then singular value decomposition (SVD) dimension reduction with Signac functions \u003cem\u003eRunTFIDF\u003c/em\u003e and \u003cem\u003eRunSVD\u003c/em\u003e. TF-IDF accounts for differences in sequencing depth between cells and gives higher values to rarer peaks, so they are not overlooked in downstream analysis. All features in the TF-IDF matrix are passed to SVD. A total of 50 LSI components, excluding the first, were passed to \u003cem\u003eRunHarmony\u003c/em\u003e with \u003cem\u003egroup.by.vars\u003c/em\u003e parameter set to sample, donor, and batch. The Harmony-corrected PCA and LSI reductions were passed to Seurat \u003cem\u003eFindMultiModalNeighbors\u003c/em\u003e function for Weighted Nearest Neighbor (WNN) analysis\u003csup\u003e9\u003c/sup\u003e. The weighted nearest neighbor and weighted shared nearest neighbor graphs were passed to UMAP and Smart Local Moving (SLM) clustering algorithm according to Seurat WNN vignette. Unsupervised clustering implemented with \u003cem\u003eFindClusters\u003c/em\u003e function in Seurat with resolution set to 1.6 yielded 113 clusters. Clusters with less than 20 cells and clusters driven by one donor were merged with the nearest cluster using \u003cem\u003eBuildClusterTree\u003c/em\u003e with slot set to \u0026lsquo;scale.data\u0026rsquo; resulting in 31 clusters (Fig. 1d, Extended Data Fig. 3). We then created a gene activity matrix inferred from single nuclei ATAC sequencing (snATAC-seq) data with the \u003cem\u003eGeneActivity\u003c/em\u003e function in Signac which quantifies ATAC-seq counts in the 2 kb-upstream region and gene body.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCell cluster identification\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe 31 cell clusters were annotated using canonical gene expression in accordance with published single-cell transcriptomics studies of mammalian hippocampus\u003csup\u003e14-18\u003c/sup\u003e, markers identified by a Wilcoxon rank-sum test implemented by FindAllMarkers function in Seurat (parameter setting RNA: only.pos = TRUE, min.pct = 0.2, min.diff.pct = 0.1, logFC_threshold = 1; parameter setting ATAC: only.pos = T, min.pct = 0.05, test.use = \u0026lsquo;LR\u0026rsquo;, latent.var = \u0026lsquo;nCount_ATAC\u0026rsquo;), and prediction scores from integration with spatial data implemented with TransferData function in Seurat. The following canonical cell types were annotated using the respective recognized markers (ordered as in Fig. 1f-g): oligodendrocyte progenitor cell (OPC): \u003cem\u003ePDGFRA\u003c/em\u003e\u003csup\u003e14,16,19-21\u003c/sup\u003e, \u003cem\u003eOLIG1\u003c/em\u003e\u003csup\u003e21\u003c/sup\u003e; oligodendrocyte (olig) \u003cem\u003eMBP\u003c/em\u003e\u003csup\u003e19,22\u003c/sup\u003e; vasculature cells including endothelial cells (Endo): \u003cem\u003eFLT1\u003c/em\u003e, \u003cem\u003ePECAM1\u003c/em\u003e (protein name: CD31),\u003cem\u003e\u0026nbsp;CLDN5\u003c/em\u003e, \u003cem\u003eNES\u003c/em\u003e; vascular leptomeningeal cells (VLMC): \u003cem\u003eCEMIP\u003c/em\u003e, \u003cem\u003eCOL1A2\u003c/em\u003e; pericyte (Peri): \u003cem\u003ePDGFRB\u003c/em\u003e, \u003cem\u003eABCC9\u003c/em\u003e; immune cells including microglia (Micro) \u003cem\u003eAPBB1IP\u003c/em\u003e, \u003cem\u003ePTPRC\u003c/em\u003e\u003csup\u003e14,20\u003c/sup\u003e,\u003cem\u003e\u0026nbsp;P2RY12\u003c/em\u003e\u003csup\u003e14\u003c/sup\u003e, macrophage (Macro): \u003cem\u003eCD163\u003c/em\u003e (alias is macrophage associated protein); T-cell: \u003cem\u003eCD247\u003c/em\u003e (alias is T-cell surface glycoprotein); ependymal cells of the subventricular zone (Epe): \u003cem\u003eCFAP54\u003c/em\u003e, \u003cem\u003eDNAH9\u003c/em\u003e, \u003cem\u003eFOXJ1\u003c/em\u003e; choroid plexus cells (Ch.Pl): \u003cem\u003eHTR2C\u003c/em\u003e, \u003cem\u003eCLIC6\u003c/em\u003e, \u003cem\u003eTTR\u003c/em\u003e; astrocytes (Astro, 3 clusters): \u003cem\u003eAQP4\u003c/em\u003e, \u003cem\u003eSLC1A2\u003c/em\u003e, \u003cem\u003eALDH1L1\u003c/em\u003e, \u003cem\u003eGLUL\u003c/em\u003e, \u003cem\u003eGFAP\u003c/em\u003e, with subtypes Astro1: \u003cem\u003eGPC5\u003c/em\u003e, \u003cem\u003eWIF1\u003c/em\u003e, GFAP; Astro2: \u003cem\u003eCD44\u003c/em\u003e, \u003cem\u003eTNC\u003c/em\u003e, \u003cem\u003eGFAP\u003c/em\u003e; Astro3: \u003cem\u003eTNC\u003c/em\u003e, \u003cem\u003eGFAP\u003c/em\u003e; Cajal Retzius cells (CR): \u003cem\u003eRELN\u003c/em\u003e\u003csup\u003e14,16\u003c/sup\u003e; granule neurons (GN, 2 clusters): \u003cem\u003eRBFOX3\u003c/em\u003e, \u003cem\u003eSYT1\u003c/em\u003e, \u003cem\u003eMYT1L\u003c/em\u003e, \u003cem\u003eSLC17A7\u003c/em\u003e, \u003cem\u003eCAMK2A\u003c/em\u003e, \u003cem\u003ePROX1\u003c/em\u003e, \u003cem\u003eCALB1\u003c/em\u003e, \u003cem\u003eTLL1\u003c/em\u003e, \u003cem\u003ePPFIA2\u003c/em\u003e, \u003cem\u003eSEMA5A\u003c/em\u003e, \u003cem\u003eBDNF\u003c/em\u003e, with subtypes \u003cem\u003ePOSTN\u003c/em\u003e\u003csup\u003e-\u0026nbsp;\u003c/sup\u003e(GN1) and \u003cem\u003ePOSTN\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e (GN2); excitatory neurons (ExN, 6 clusters): with \u003cem\u003eRBFOX3\u003c/em\u003e, \u003cem\u003eMYT1L\u003c/em\u003e, \u003cem\u003eCAMK2A\u003c/em\u003e, \u003cem\u003eSLC17A7\u003c/em\u003e, \u003cem\u003eSV2B\u003c/em\u003e expressed by all ExN and specific subtype markers for ExN1: \u003cem\u003eCADPS2\u003c/em\u003e, \u003cem\u003eNEUROD6\u003c/em\u003e; ExN2: \u003cem\u003eSATB2\u003c/em\u003e, \u003cem\u003eSLC22A10\u003c/em\u003e; ExN3: \u003cem\u003eNEUROD6\u003c/em\u003e, \u003cem\u003eTRHDE\u003c/em\u003e, \u003cem\u003eCPNE4\u003c/em\u003e, \u003cem\u003eHTR7\u003c/em\u003e; ExN4: \u003cem\u003eSATB2\u003c/em\u003e, \u003cem\u003eSLC22A10\u003c/em\u003e, \u003cem\u003eHS3ST2\u003c/em\u003e, \u003cem\u003eSYNPO2\u003c/em\u003e; ExN5: \u003cem\u003eHS3ST2\u003c/em\u003e, \u003cem\u003eNTNG1\u003c/em\u003e; ExN6: \u003cem\u003eHS3ST2\u003c/em\u003e, \u003cem\u003eCHRM2\u003c/em\u003e, \u003cem\u003ePCP4\u003c/em\u003e; inhibitory neurons (InN, 9 clusters): all expressing \u003cem\u003eRBFOX3\u003c/em\u003e, \u003cem\u003eMYT1L\u003c/em\u003e, \u003cem\u003eGAD1\u003c/em\u003e, \u003cem\u003eGAD2\u003c/em\u003e, \u003cem\u003eSLC32A1\u003c/em\u003e, \u003cem\u003eDCX\u003c/em\u003e, with InN subtypes expressing: \u003cem\u003eVIP\u003c/em\u003e, \u003cem\u003eSST\u003c/em\u003e, \u003cem\u003ePVALB\u003c/em\u003e, \u003cem\u003eLHX6\u003c/em\u003e, \u003cem\u003eLAMP5\u003c/em\u003e, \u003cem\u003eSLC17A8\u003c/em\u003e, \u003cem\u003eTUBB3\u003c/em\u003e, \u003cem\u003ePROX1\u003c/em\u003e, \u003cem\u003ePENK\u003c/em\u003e.\u003c/p\u003e\n\u003cp\u003eOut of the 31 cell clusters (Extended Data Fig. 4), 12 cell types were identified by merging all astrocyte clusters, all inhibitory neuron clusters, all immune clusters, all vasculature clusters, the two GN clusters, ExN1 and ExN3 merged to CA.ExN (\u003cem\u003eEPHA5\u003c/em\u003e\u003csup\u003e16\u003c/sup\u003e, \u003cem\u003eNEUROD6\u003c/em\u003e\u003csup\u003e19\u003c/sup\u003e), and ExN2 and ExN4-6 merged to Other ExN (O.ExN; expressing \u003cem\u003eTSHZ2\u003c/em\u003e\u003csup\u003e23,24\u003c/sup\u003e).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSingle nuclei integration with other datasets\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe integrated our database with publicly available databases to achieve a better characterization of the molecular landscapes defining cell types and states in the human hippocampus in neurotypical controls and MDD. We downloaded the count matrix of datasets generated by recently published studies focusing on the hippocampus\u003csup\u003e14,16,25,26\u003c/sup\u003e. Independently generated datasets were preprocessed using our standard pipeline using Seurat\u003csup\u003e10\u003c/sup\u003e and Harmony\u003csup\u003e13\u003c/sup\u003e in R. Then, following a published approach\u003csup\u003e14\u003c/sup\u003e, for the top 2000 variably expressed genes, Pearson\u0026rsquo;s R was calculated for the average expression across cell types (Extended Data Fig. 5).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTissue and library preparation for Visium spatial gene expression\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTwo adjacent sections per donor were sectioned at 14 \u0026micro;m at -20C and mounted within capture areas on Visium slides. Slides were stored at -80C, then fixed, stained, and imaged according to 10X User Guide CG000160 Rev C. Spatial gene expression libraries were prepared according to 10X User Guide CG000239 Rev F. The libraries were sequenced on an Illumina NovaSeq 6000 using v4 chemistry to an average sequencing depth of about 67,000 reads/spot and 1965 median genes/spot (Supplementary Table 7). SpaceRanger (v1.3.0) with chemistry set to Spatial 3\u0026rsquo; v1 used to generate count matrices.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTissue preparation and loading for Xenium in-situ hybridization\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFresh frozen tissue was sectioned at 10\u0026micro;m onto 10.45mm x 22.45mm sampling area on 10X Genomics Xenium Slide according to 10X Demonstrated Protocol CG000579 Rev C. Tissue was fixed and permeabilized according to 10X Demonstrated Protocol CG000581 Rev C. Probe hybridization, ligation, and amplification were carried out according to 10X User Guide CG000582 Rev C. The output from the Xenium analyzer was loaded into Xenium Explorer 1.3 or RStudio for integration with single-nuclei gene expression data.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSpatial gene expression data pre-processing, integration, and clustering\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eVisium (10X Genomics) count matrices and spatial data were loaded into Seurat using the Load10X_Spatial command with the default parameters. Feature counts for each spot were divided by total counts for that spot and multiplied by scale factor (10,000) and natural log transformed using function NormalizeData. Features that were outliers to mean variability were found with local polynomial regression of log(variance) and log(mean) which gives the observed mean and expected variance of feature expression implemented using the FindVariableFeatures function. Then, data were scaled and centered using ScaleData function. Spatial data from replicates were merged using the merge function. Dimensionality reduction and clustering were calculated using RunPCA, FindNeighbors, FindClusters, and RunUMAP with dimensions parameter set to 1:30.\u003c/p\u003e\n\u003cp\u003eSpatial data were integrated with multiome data using FindTransferAnchors and TransferData functions in Seurat\u003csup\u003e10\u003c/sup\u003e (Fig. 1h, Fig. 2e).\u003c/p\u003e\n\u003cp\u003eHippocampalsubfields, including granule cell layer (GCL), subgranular zone (SGZ), molecular layer (ML), \u003cem\u003eCornu Ammonis\u003c/em\u003e (CA), white matter (WM), vasculature (vasc), chorioid plexus (chor.plex), and subventricular zone (SVZ) were manually annotated in Loupe Browser within the Visium fiducial frame (Fig. 2b). Folds in the tissue were labeled as such and excluded from downstream analysis. Spots with less than 500 unique molecular identifiers were excluded from downstream analysis. Unsupervised clustering of merged spatial gene expression data was driven largely by sample and donor identify, so spatial gene expression data were integrated using canonical correlation analysis (CCA) in Seurat (Extended Data Fig. 6a-b).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSpatial cluster identification\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eUsing \u003cem\u003eFindClusters\u003c/em\u003e in Seurat, 13 clusters were identified at 0.5 resolution: two hilus clusters, hilus1 comprising mostly excitatory neurons, and hilus2 showing markers of inhibitory neurons, mossy cell markers, serotonin receptor 2c (\u003cem\u003eHRT2C\u003c/em\u003e) and tryptophan hydroxylases 2 (\u003cem\u003eTPH2\u003c/em\u003e); CA3 stratum pyramidales (ca3) comprising excitatory neurons; stratum oriens-stratum lucidum (so-slu) representing WM; so-slu of the CA3 (ca3-so-slu) expressing parvalbumin and markers involved in cell-cell adhesion, extracellular matric, methylation, calcium signaling, and trophic factors; stratum radiatum (sr) comprised of axonal markers; stratum oriens-stratum lacunosum (so-sla) made up of dendrites; stratum moleculare (ml) comprised of apical dendrite-associate markers and \u003cem\u003eHTR2B\u003c/em\u003e; stratum granulosum (gcl) made up mostly of GNs; SGZ-polymorphic layer (sgz-pl) expressing markers involved in neurogenesis, development and differentiation, \u003cem\u003eTPH1\u003c/em\u003e and \u003cem\u003eTPH2\u003c/em\u003e; scattered spots in SGZ and stratum moleculare (sgz-ml) expressing stem cell, differentiation and neurogenesis markers and \u003cem\u003eHTR3A\u003c/em\u003e and \u003cem\u003eHTR1D\u003c/em\u003e; vasculature (vasc) expressing endothelial and collagen markers, and \u003cem\u003eHTR2B\u003c/em\u003e and \u003cem\u003eTPH1\u003c/em\u003e; and choroid plexus (ch.pl) expressing cilia and flagella associated markers, \u003cem\u003eHTR2C\u003c/em\u003e, and astrocyte markers (Supplementary Table 8).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIntegration of spatial and single-nucleus gene expression datasets\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe used Seurat v3 anchor-based integration to transfer annotations from single nucleus data to spatial Visium data. Single-nucleus annotations were used as the reference dataset and both whole transcriptome Visium spatial gene expression and Xenium in-situ hybridization were used as query data. Spots were assigned probability scores based on transcriptional congruence with cell types from single nucleus data (Fig. 1h-i, Fig. 2e-f, Extended Data Fig. 4f).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRNA Velocity inference on Visium spatial transcriptomic data\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo understand the kinetics of cell state in human hippocampus, we computed RNA velocity using velocyto\u003csup\u003e27\u003c/sup\u003e\u003cu\u003e\u0026nbsp;\u003c/u\u003ebased on the relative abundance of nascent (unspliced) mRNAs and mature (spliced) mRNAs on Visium spots (55x55 m). Samtools\u003csup\u003e28\u003c/sup\u003e (v1.9) was used to sort bamfiles and \u003cem\u003evelocyto run\u003c/em\u003e\u003csup\u003e27\u003c/sup\u003e was used to run velocity analysis and output a loom file. Pseudotime lineage analysis was performed by RNA velocity (scVelo Python package\u003csup\u003e29\u003c/sup\u003e , Fig. 2g-k, Extended Data Fig 6). The proportion of spliced/unspliced RNAs did not differ between unsupervised spot clusters, and we did not have power to test differences between MDD and CTRL.\u003c/p\u003e\n\u003cp\u003eWe used unsupervised clusters located in SGZ and GCL (Supplementary Table 8) that expressed proliferation, differentiation, maturation and neurogenesis markers among their top 100, to subset spots hypothesized to be part of neurogenesis trajectory (Extended Data Fig. 6c-h). We selected spatial clusters sgz.ml, sgz.pl, and gcl (Supplementary Table 8) and recalculated UMAP embedding using the top 5000 variably expressed genes and additional genes known to be involved in neurogenesis (total of 5074 genes) and the first 15 PCs for FindNeighbors and RunUMAP. Seurat object was translated to AnnData using SeuratDisk functions SaveH5Seurat and Convert with \u0026lsquo;dest\u0026rsquo; set to \u0026lsquo;h5ad.\u0026rsquo; The \u0026lsquo;data\u0026rsquo; layer from the Seurat object was used for gene expression calculations in scvelo. The AnnData object was merged with loom files output of velocyto run on the command line. The dynamical model was run with the default parameters.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIdentification of neurogenic niche stem cell progeny using machine learning on single nuclei transcriptome data\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe apply a supervised learning approach to identify neurogenic cell types, as recently published\u003csup\u003e16\u003c/sup\u003e. Prototypes for each cell type were discovered by first retaining the sparse counts matrix for the top 4000 variable genes and an additional 54 genes involved in neurogenesis and neuronal maturation. The Seurat function \u003cem\u003eNormalizeData\u003c/em\u003e was rerun on this matrix with the default parameters. For each gene, normalized counts that were above the sum of the third quartile and the product of 1.5 times the inter-quartile range were brought down to this value to reduce the effect of extreme outliers. Prototype scores were assigned to cells for identification of the following cell types: TypeI progenitors (quiescent neural stem cells), TypeII progenitors (intermediate amplifying neural progenitors), TypeIII (differentiating neurons), and ImGN (immature granule neurons, which are maturing and integrating into the DG circuit) using normalized count data. The distribution of summed normalized counts of selected genes expected to be expressed at each maturation stage was used to determine a cutoff for defining prototypical cells, considering that, based on previous studies in rodents and primates, progenitors are expected to be a small proportion of the total number of adult human dentate gyrus cells (\u0026lt;1% of all cells). We summed normalized counts of the following stage-specific genes: for TypeI, we summed normalized counts for \u003cem\u003eGFAP\u003c/em\u003e, \u003cem\u003eSOX2\u003c/em\u003e, \u003cem\u003ePAX6\u003c/em\u003e, \u003cem\u003eNES\u003c/em\u003e, \u003cem\u003eFABP7\u003c/em\u003e, \u003cem\u003eHES5\u003c/em\u003e, and \u003cem\u003eETNPPL\u003c/em\u003e; for TypeII, normalized counts for \u003cem\u003eASCL1\u003c/em\u003e, \u003cem\u003eFOXO3\u003c/em\u003e, \u003cem\u003eNR2E1\u003c/em\u003e, \u003cem\u003eEOMES\u003c/em\u003e, \u003cem\u003eNEUROG2\u003c/em\u003e, \u003cem\u003eNEUROD1\u003c/em\u003e, \u003cem\u003eNES\u003c/em\u003e, \u003cem\u003eMCM2\u003c/em\u003e, \u003cem\u003ePCNA\u003c/em\u003e, \u003cem\u003eMKI67\u003c/em\u003e, and \u003cem\u003eTOP2A\u003c/em\u003e were summed, and because oligodendrocyte progenitors can express markers in common to other progenitors, normalized counts for \u003cem\u003eMBP\u003c/em\u003e and \u003cem\u003eMOBP\u003c/em\u003e were subtracted so that cells that could be on the oligodendrocyte differentiation trajectory were less likely to be determined to be TypeII prototypes; for TypeIII, we summed normalized counts for \u003cem\u003eMYT1L\u003c/em\u003e, \u003cem\u003ePROX1\u003c/em\u003e, \u003cem\u003eDCX\u003c/em\u003e, \u003cem\u003eCALB2\u003c/em\u003e, \u003cem\u003eRELN\u003c/em\u003e, \u003cem\u003eST8SIA2\u003c/em\u003e, and \u003cem\u003eST8SIA4\u003c/em\u003e and subtracted normalized counts for \u003cem\u003eRBFOX3\u003c/em\u003e and \u003cem\u003eVIP\u003c/em\u003e to decrease the likelihood that TypeIII prototypes were at a later stage of maturation or inhibitory neurons; for ImGN prototypes, normalized counts of \u003cem\u003eRBFOX3\u003c/em\u003e, \u003cem\u003ePROX1\u003c/em\u003e, \u003cem\u003eNEUROD1\u003c/em\u003e, \u003cem\u003eDCX\u003c/em\u003e, \u003cem\u003eBHLHE22\u003c/em\u003e, \u003cem\u003eCOL25A1\u003c/em\u003e, and \u003cem\u003ePOSTN\u003c/em\u003e were summed. Cells with high prototypes scores were identified as prototypical, or high confidence cells.\u003c/p\u003e\n\u003cp\u003eWe applied the LASSO regression (implemented with \u003cem\u003eglmnet\u003c/em\u003e in R\u003csup\u003e30\u003c/sup\u003e) for each cell type individually. For each cell type, the prototypes and a randomly sampled quantity of cells required to reach 50,000 were used for training and testing the model. The trained model uses a list of positively and negatively ranked coefficients to rank genes in their ability to predict cell types. This method uses many genes to characterize these cell types (for TypeI, TypeII, TypeIII, and ImGN it used 372, 410, 555, and 422 genes respectively), instead of just a few arbitrarily chosen genes. Few cells (approximately 150 cells) were assigned two identities. For these cells, we manually assigned the identity with the highest predicted probability from \u003cem\u003epredict.glm\u003c/em\u003e with type set to \u0026lsquo;response\u0026rsquo; and the tuning parameter s was chosen by 10-fold cross validation by cv.glmnet. For final counts of cells in each LASSO identified cell type, see Supplementary Table 9.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCell trajectory inference based on single nucleus transcriptomic data\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn the absence of cytoplasmic mRNA and thus degradation rate in the single-nuclei gene expression data, we used Palantir\u003csup\u003e31\u003c/sup\u003e to characterize the trajectory of cells identified as TypeI, TypeII, TypeIII, ImGN, and GN from LASSO regression. TypeI, TypeII, TypeIII, ImGN, and GN cell types were subset, and re-preprocessed using Seurat and Harmony as described in Multi-modal integration. The first 5 Harmony batch, sample, and donor corrected PCs were used for FindNeighbors and RunUMAP (Fig. 3b). Seurat object was converted to AnnData using SeuratDisk functions SaveH5Seurat and Convert with \u0026lsquo;dest\u0026rsquo; set to \u0026lsquo;h5ad.\u0026rsquo; Palantir function core.run_palantir was used to calculate psueodtime ordering of each cell (Fig. 3c), terminal state probabilities, and entropy (a quantitive measure of the differentiation potential of each cell). Generalized Additive Models\u003csup\u003e31\u003c/sup\u003e were used to determine the gene expression trends along the neurogenesis lineage using the function presults.compute_gene_trends (Fig. 3e).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDifferential gene expression analysis in MDD \u003cem\u003eversus\u003c/em\u003e CTRL\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo determine cell-type specific differentially expressed genes (DEGs) between subjects with MDD and neurotypical controls, count, peak, and gene activity matrices were aggregated on the sample and cell type level. Data were normalized using center log ratio transformation implemented using the clr function in the compositions R package. Then, we used the differential expression pipeline from R package limma\u003csup\u003e32,33\u003c/sup\u003e. Samples from the same donor were not aggregated. Instead, we used limma\u0026rsquo;s DuplicateCorrelation function to add a random effect for donor identity because, in some cases, samples from the same donor were processed in different batches. Additionally, since donors had between 2-3 samples, we used voomWithQualityWeights to adjust for differences in the number of samples per donor and number of cells aggregated/sample. We added RIN, PMI, batch, and age to the model as fixed effects. For RIN, PMI, batch, and age, values that were outside the sum of the third quartile and the product of 1.5 times the inter-quartile range from the median were winsorized to reduce the effect of extreme outliers. P-values were adjusted using Benjamini-Hochberg correction. We implemented the Bioconductor scRNA-seq pseudobulk pipeline to analyze DEGs between MDD and neurotypical controls in all clusters\u003csup\u003e34-38\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDifferential chromatin accessibility analysis in MDD \u003cem\u003eversus\u003c/em\u003e CTRL\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo determine cell-type specific differentially accessible regions (DARs) between subjects with MDD and neurotypical controls, the peaks matrix was aggregated and the pipeline for differential gene expression was applied. Additionally, a gene activity matrix was created from snATAC-seq using GeneActivity function in Signac. Chromatin accessibility at transcriptional and promoter regions was used to create the gene activity matrix. Then, the pipeline for predicted differentially active genes (DAGs) was applied to the aggregated gene activity matrices.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePre-ranked gene set enrichment analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe performed pre-ranked Gene Set Enrichment Analysis (GSEA)\u003csup\u003e39\u003c/sup\u003e with FGSEA\u003csup\u003e40\u003c/sup\u003e using the product of log fold change and negative base 10 log of the Benjamini-Hochberg (BH) corrected p-value as the ranking method. We used the Gene Ontology (GO) annotation system to determine molecular function (MF, the molecular activities carried out by gene products), the cellular component (CC, the cellular location where gene products are active and perform their function), and biological process (BP, the processes and activities in which gene products are involved), with gene sets obtained from Molecular Signatures Database (MSigDB)\u003csup\u003e41\u003c/sup\u003e. We used Kyoto Encyclopedia of Genes and Genomes (KEGG)\u003csup\u003e42\u003c/sup\u003e to infer biological pathways related to DEGs, and used the Reactome pathway\u003csup\u003e43\u003c/sup\u003e to understand molecular mechanisms of biological reactions that are interconnected and result in pathways. The following parameters were used for the \u003cem\u003efgsea\u003c/em\u003e function: minSize=15, maxSize=400, nperm = 10000. Outputs with BH adjusted p-value less than 0.05 were considered significant.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePsyGeNET analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe ran \u003cem\u003eenrichedP\u003c/em\u003eD from Psygenet2r\u003csup\u003e44\u003c/sup\u003e a R/Bioconductor package for the analysis of psychiatric disease genes, for leveraging PsyGeNET\u003csup\u003e45\u003c/sup\u003e a knowledge resource that collects and validates information of psychiatric diseases and their genes. We used database = \u0026lsquo;ALL\u0026rsquo; to find psychiatric disorders for which DEGs from any cell type showed enrichment. We tested all DEGs from gene expression, all DEGs from gene activity, and the combination, and ran \u003cem\u003epsygenetGene\u003c/em\u003e with database = \u0026lsquo;ALL\u0026rsquo; and created \u003cem\u003egeneAttrPlot\u003c/em\u003e for the evidence index. We then ran \u003cem\u003epsygenetGene\u003c/em\u003e with DEGs from GN and ImGN (Fig. 4c) which were the cell types with most dysregulated genes across multiple modalities: transcript, chromatin accessibility and gene activity.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eWeighted gene co-expression network analysis (WGCNA)\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWeighted gene co-expression network analysis (WGCNA) was performed to identify co-expression modules using snRNA-seq data\u003csup\u003e22,46\u003c/sup\u003e. First, the aggregated expression for each sample (N = 52; 25 CTRL, 27 MDD) in the GN and ImGN cell types was calculated by summing the counts per gene across all nuclei. To account for known external sample traits, the counts were corrected for age and batch using limma\u003csup\u003e47\u003c/sup\u003e. In addition, lowly expressed genes with total counts of below 5 were removed. A soft thresholding power of 20 and 18, respectively, and a minimum module size of 30 genes, were used for network construction and module detection for GN. Each module was correlated with the phenotype (CTRL \u003cem\u003evs.\u003c/em\u003e MDD), and significance was determined using a \u003cem\u003ep\u003c/em\u003e-value \u0026lt; 0.05. The online tool Metascape\u003csup\u003e48\u003c/sup\u003e was used to perform GO on the modules that showed a correlation with MDD.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIn Situ Hybridization\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor each donor, 3 serial sections of 16 m at 80 m intervals were used for Singleplex and Duplex RNAScope assays. For Singleplex RNAscope (ACDbio), the \u003cem\u003eDCX\u003c/em\u003e probe (RNAscope Probe-Hs-DCX, Cat. No. 489551) was hybridized and amplified. Duplex RNAscope was used to investigate \u003cem\u003eDCX\u003c/em\u003e co-localization with \u003cem\u003eNEUROD1\u003c/em\u003e, \u003cem\u003ePROX1\u003c/em\u003e, \u003cem\u003eRBFOX3\u003c/em\u003e, or \u003cem\u003eTUBB3\u003c/em\u003e mRNA in the human DG. Fresh frozen 16 m tissue sections were fixed in 4% paraformaldehyde for 15 minutes at room temperature (RT). Sections were washed in 1x PBS and then dehydrated in 50%, 70%, and twice in 100% EtOH for 5 minutes. RNAscope was performed according to manufactured instructions. The \u003cem\u003eDCX\u003c/em\u003e probe (Cat. No. 489551-C2) and \u003cem\u003eNEUROD1\u003c/em\u003e probe (Cat. No. 437281), \u003cem\u003ePROX1\u003c/em\u003e probe (Cat. No. 530241), \u003cem\u003eRBFOX3\u003c/em\u003e probe (Cat. No. 415591), or \u003cem\u003eTUBB3\u003c/em\u003e probe (Cat No. 318901) were hybridized and amplified in C2 and C1 respectively. Slides were counterstained with 50% hematoxylin (Gil\u0026rsquo;s Hematoxylin No. 1) according to manufacturer\u0026rsquo;s instructions. The slides are then quickly dipped in xylene and mounted with Permount.\u003c/p\u003e\n\u003cp\u003eMultiplex RNAscope was used to investigate the expression of \u003cem\u003eGAD1\u003c/em\u003e (Cat No 404031-C2), \u003cem\u003ePROX1\u003c/em\u003e (Cat No 530241-C3), and \u003cem\u003eDCX\u003c/em\u003e (Cat No 489551) in the human DG, because single nuclei data showed co-expression of \u003cem\u003eGAD1\u003c/em\u003e in cell clusters expressing \u003cem\u003eDCX\u003c/em\u003e. Sections were counterstained with 4\u0026rsquo;, 6-diamidino-2-phenylindole (DAPI).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGene expression analysis using qPCR\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eRNA was isolated using RNeasy FFPE Kit (Qiagen, Germantown, MD) according to the manufacturer\u0026rsquo;s instructions. RNA quantity was assessed by evaluation of the A260/280 and A260/230 ratios using a Nanodrop spectrometer (Thermo Scientific 840274100). Superscript III enzyme (Invitrogen) was used to reverse transcribe 200 ng total RNA. Quantitative Real-Time PCR was performed on a StepOnePlus\u0026trade; Real-Time PCR System using a TaqMan\u0026trade; Gene Expression Master Mix (Applied Biosystems, Waltham, USA). A ready-made \u003cem\u003eDCX\u003c/em\u003e TaqMan\u0026trade; probe (Applied Biosystems, Waltham, USA) was used to amplify a 77bp region spanning an exon boundary within the coding sequence of the target gene. Each sample was assayed in triplicate and \u003cem\u003eDCX\u003c/em\u003e target gene expression was normalized to the geometric mean of the reference gene beta-actin (\u003cem\u003eACTB\u003c/em\u003e). \u003cem\u003eDCX\u003c/em\u003e gene expression was analyzed in PRISM using a Welch unpaired T-test.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDouble Immunofluorescence\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo examine expression of cell type specific markers at the protein level in human hippocampus, we performed double immunofluorescence (IF) for protein corresponding to RNAs of interest identified with the single nuclei and spatial transcriptomic analyses.\u003c/p\u003e\n\u003cp\u003eHuman hippocampus tissue sections (50 m thick) were washed and prepared as previously published\u003csup\u003e49\u003c/sup\u003e. The primary antibodies were incubated [(1:1000 guinea pig anti-DCX (Millipore), 1:5000 mouse anti-Glial Fibrillary Associated Protein (GFAP, Sigma-Aldrich), 1:500 rabbit anti-Ionized calcium binding adaptor molecule 1 (Iba1, Fujifilm), 1:50,000 mouse anti-Neurofilament (NF; Sigma-Aldrich); 1:2000 mouse anti-polysialylated neural cell adhesion marker (PSA-NCAM, Miltenyi Biotec); 1:500 rabbit anti-Calretinin (Swant Inc.), 1:500 rabbit anti-Calbindin (Swant Inc.), 1:500 rabbit anti-Neuronal Nuclear Marker (NeuN, Millipore)] in a 25% blocker overnight or for 5-days (Calretinin and Calbinin) at 4\u0026deg;C. The secondary antibodies were incubated overnight at 4\u0026deg;C with Jackson Alexa Fluor 488 goat anti-mouse or anti-rabbit (1:500) and Jackson Alexa Fluor 594 goat anti-guinea pig (1:500). The following day the tissue was washed for 30min (2x15\u0026rsquo;) in 0.05M PBS. Then the tissue was counterstained with 4\u0026rsquo;, 6-diamidino-2-phenylindole (DAPI) in PBS for 3min before being washed for 15min (3x5\u0026rsquo;) in 0.05M PBS. The tissue was then treated with 0.1% True Black (Biotium) in 70% ethanol in distilled water to quench autofluorescence. Sections were then washed for 1h (6x10\u0026rsquo;) in 0.05M PBS. The tissue was coverslipped using 90% glycerol and PBS.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eImmunohistochemistry\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eImmunohistochemistry (IHC) for DCX was performed as previously described\u003csup\u003e49\u003c/sup\u003e Sections were incubated in guinea pig anti-doublecortin (1:30,000, Millipore) in 10% blocking solution overnight at 4\u0026deg;C on a shaker. Following 2-hour secondary incubation with anti-guinea pig biotin-conjugated antibody (Vector Laboratories) and 1-hour incubation in avidin-biotin-horse radish peroxidase system (Vector Laboratories), diluted 1:200, sections were washed in sodium acetate for 5 min, followed by a PBS wash for 20 min (2x10\u0026rsquo;) before being mounted on glass slides and desiccated. Sections were counterstained with Nuclear Red which involves dehydrating slides in ethanol, clarifying slides in xylenes and cover-slipping with Permount (Sigma-Aldrich).\u003c/p\u003e\n\u003cp\u003eSections were processed to identify mature granule neurons (anti-Neuronal Nuclear antigen [NeuN] mouse monoclonal antibody, 1:100,000; Chemicon, Temecula, CA), progenitor cells of Type II or III (anti-nestin mouse monoclonal antibody, 1:8000, Chemicon) and mitotic cells (anti-Ki-67 mouse monoclonal antibody, 1:200, Novocastra Clone-MM1, Newcastle Upon Tyne, UK). Immunohistochemistry and stereology were performed as previously described\u003csup\u003e50\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConfocal Microscopy\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCell markers co-localization was visualized using an Olympus Fluoview FV1000 Confocal Laser Scanning System (Olympus, Waltham, MA) equipped with an Olympus IX81 Inverted Microscope, with electronic stage controller (Prior Scientific, Rockland, MA), camera interface, and Olympus Fluoview 1000 (v. 1.5) software, using the following objectives: UPLSAPO 10X and 20X air, with numerical aperture (NA) 0.40, 0.75; UPLFLN 40X oil with NA 1.30; and PLAFON 60X oil with NA 1.42 (Olympus).\u003c/p\u003e\n\u003cp\u003eTo quantify co-localization of cell markers (i.e., DCX/NeuN), image stacks of the whole DG were obtained using a confocal scanning microscope (Leica TCS SP8 2-Photon, Leica Microsystems Inc.), and then processed for Stereology using Stereoinvestigator software (MBF, Inc.). Fluorescence from the different fluorophores was detected in the following way: 1) 488-AffiniPure was excited at 488 nm and detected at 505-550 nm, 2) 594-AffiniPure was excited at 552 nm and detected at 600-650 nm, and 3) DAPI was excited at 405 nm and detected at 415-485 nm. For 488- and 552-nm excitation, the beam path included a TD 488/552/638 beamsplitter, while for 405 nm excitation a substrate beamsplitter was used.\u003c/p\u003e\n\u003cp\u003eHippocampal sections from each subject were scanned to produce 13 \u0026plusmn; 2 \u0026mu;m image stacks, depending on the thickness of each tissue section. All z stacks were imaged with a dry Leica 20X objective (NA 0.70, working distance 0.5 mm), with a field of view of 553.6 \u0026times; 553.6 \u0026mu;m, a pixel size of 0.54 \u0026times; 0.54 \u0026mu;m, optical sectioning of 2.36 \u0026mu;m, and a z step of 1 \u0026mu;m. Positive and negative controls were run to ensure signal specificity and minimal background, creating an optimal signal to noise ratio. The same parameters were applied for every section of every subject. Image stacks were then imported in our MicroBrightfield system (MBF Bioscience, Williston, VT) to perform unbiased stereology.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCell quantification and statistical analysis\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe used an unbiased stereological approach (optical \u003cem\u003edisector\u003c/em\u003e with fractionator) to estimate numbers of cells expressing DCX (DCX+), DCX/NeuN+ cells, DCX mRNA+ cells and DCX/TUBB3 mRNA+ cells. Olympus BX51-WI microscope was used with electronic stage controller LUDL MAC 6000 XYZ (BioVision Technologies, Exton, PA) and camera interface for Retiga 2000R CCD Scientific Camera (Q-Imaging, Surrey, BC, Canada).\u003c/p\u003e\n\u003cp\u003eFor anatomical alignment along the rostro-caudal axis of the DG, processed sections were anatomically matched with Nissl-stained sections at 1-mm interval using a stereoscope (Leica, Wild M3Z, Heerburg, Switzerland). We defined the anterior DG as the portion from the most rostral appearance of the DG to the start of the lateral geniculate. The mid DG was defined as the portion going from the start of the lateral geniculate to the disappearance of the lateral geniculate, and the posterior DG was the region from the end of the lateral geniculate to the end of the caudal DG. Outlines of the DG including SGZ and GCL were aligned from the most rostral to the most caudal hippocampal sections and used as boundaries for cell counting. To quantify cells, the first hippocampus section to be sampled was the one in which the DG first appeared. Subsequent sections were assayed every 2 mm thereafter in systematic fashion, until the DG disappeared, with an average of 8-10 sections per subject for DCX immunohistochemistry experiments. It should be noted that for DCX/NeuN and RNAscope experiments, an average of 3-5 sections of the anterior hippocampus were used per subject. Using the Stereo Investigator system, DG boundaries were defined at low magnification using the 4x objective. Cells were not counted in the upper and lower 3 \u0026mu;m in the z planes of the tissue sections (the guard zones). The estimated total number of cells was calculated as:\u003c/p\u003e\n\u003cp\u003eN=\u0026sum;Q\u003csup\u003e\u0026minus;\u003c/sup\u003e\u0026times;(1/hsf)\u0026times;(1/asf)\u0026times;(1/ssf)\u003c/p\u003e\n\u003cp\u003ewhere \u003cem\u003e\u0026Sigma;\u003c/em\u003e\u003cem\u003e\u003csup\u003eQ\u0026minus;\u003c/sup\u003e\u003c/em\u003e = total number of cells counted, hsf (height sampling fraction) = h/t, where \u003cem\u003et\u003c/em\u003e is the mean section thickness, \u003cem\u003eh\u003c/em\u003e is the height of the optical \u003cem\u003edissector\u003c/em\u003e, \u003cem\u003easf\u003c/em\u003e = area sampling fraction (area of the counting frame over the area of x,y step), and \u003cem\u003essf\u003c/em\u003e = section sampling fraction (1/section interval).\u003c/p\u003e\n\u003cp\u003eThe cell counts were analyzed in PRISM using a Welch unpaired T-test.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eProteomics\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe used Shotgun proteomics liquid chromatography with tandem mass spectrometry (LC/MS/MS), a strategy with broad applicability, based on digestion of proteins with proteolytic enzymes and analysis of the resulting peptides by mass spectrometry\u003csup\u003e51\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eFresh-frozen 4% paraformaldehyde post-fixed tissue was dissolved in 0.2% RapiGest SF in 50 mM ammonium bicarbonate, heated to 105 \u003csup\u003eo\u003c/sup\u003eC for 30 min, then to 80 \u003csup\u003eo\u003c/sup\u003eC for 2 h, digested with trypsin as previously described\u003csup\u003e52,53\u003c/sup\u003e, and then 50 fmol of yeast alcohol dehydrogenase were added as internal control.\u003c/p\u003e\n\u003cp\u003eLC/MS/MS was run for 120 minutes, two chromatograms recorded for each biological replicate, in traveling-wave ion mobility spectrometry (TWIMS) MS\u003csup\u003eE\u003c/sup\u003e resolution mode \u003csup\u003e51,53,54\u003c/sup\u003e; mass spectra were recorded every 0.6s; every 30s internal \u0026ldquo;lockmass\u0026rdquo; spectrum of Glu-1-Fibrinopeptide B was recorded (m/z 785.8426); 11,000 mass spectra were recorded per run; a total of 792,000 mass spectra were analyzed. Of the 1811 proteins detected, 680 were represented by a single peptide or with protein identification quality scores less than 250 and were not included in the analysis.\u003c/p\u003e\n\u003cp\u003eThe two groups (MDDs and CTRL) were compared against each other and differentially expressed proteins were identified with a PeptideTeller predicted error (false positive rate of 1%). P-values used in our report on these experimental results are based on the error function called xdev, which addresses false positives by using an advanced error model as described in previous papers\u003csup\u003e55,56\u003c/sup\u003e. The xdev-derived P-values were generated by the software Rosetta Elucidator Protein Expression Analysis System (V. 4.0.0.2.13, PerkinElmer, Boston, MA,). This software, developed by Merck \u0026amp; Co./Rosetta\u003csup\u003e57-59\u003c/sup\u003e extends the dynamic range of fold-change data, especially for low abundance proteins. This system matches accurate mass and retention time across all LC/MS/MS chromatograms in the experiment, enhancing the effectiveness of the mass spectrometry analyses. The xdev represent the difference of intensities between groups (pairwise comparisons) divided by the error of the difference, and the p-value of differential expression is calculated from an error function of a standard Gaussian distribution. Elucidator ran on Oracle Linux v. 6.5 server with 128 GB RAM and 32 Tb of RAID network attached storage. From within Elucidator, results were searched with Mascot server (Version 2.5.1, Matrix Science Ltd., London, UK). The search was against a SwissProt human canonical protein sequence database with isoforms (UniProtKB/Swiss-Prot Release 2017_01 containing 84,336 sequences; 48,572,808 residues, www.uniprot.org). Our quality control indices included peptide count above two for considering a protein a good candidate, and p-value \u0026lt; 0.05.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods References\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e1 Rosoklija, G., Duma, A. \u0026amp; Dwork, A. J. Psychiatric brain collection in Macedonia: general lessons for scientific collaboration among countries of differing wealth. \u003cem\u003ePril (Makedon Akad Nauk Umet Odd Med Nauki)\u003c/em\u003e \u003cstrong\u003e34\u003c/strong\u003e, 95-98 (2013).\u003c/p\u003e\n\u003cp\u003e2 Kelly, T. M. \u0026amp; Mann, J. J. Validity of DSM-III-R diagnosis by psychological autopsy: a comparison with clinician ante-mortem diagnosis. \u003cem\u003eActa Psychiatr.Scand.\u003c/em\u003e \u003cstrong\u003e94\u003c/strong\u003e, 337-343 (1996).\u003c/p\u003e\n\u003cp\u003e3 Zhang, D.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Spatial epigenome-transcriptome co-profiling of mammalian tissues. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e616\u003c/strong\u003e, 113-122 (2023). https://doi.org:10.1038/s41586-023-05795-1\u003c/p\u003e\n\u003cp\u003e4 Deng, Y.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Spatial profiling of chromatin accessibility in mouse and human tissues. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e609\u003c/strong\u003e, 375-383 (2022). https://doi.org:10.1038/s41586-022-05094-1\u003c/p\u003e\n\u003cp\u003e5 Boldrini, M.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Resilience Is Associated With Larger Dentate Gyrus, While Suicide Decedents With Major Depressive Disorder Have Fewer Granule Neurons. \u003cem\u003eBiol Psychiatry\u003c/em\u003e \u003cstrong\u003e85\u003c/strong\u003e, 850-862 (2019). https://doi.org:10.1016/j.biopsych.2018.12.022\u003c/p\u003e\n\u003cp\u003e6 Boldrini, M.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Human Hippocampal Neurogenesis Persists throughout Aging. \u003cem\u003eCell Stem Cell\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 589-599 e585 (2018). https://doi.org:10.1016/j.stem.2018.03.015\u003c/p\u003e\n\u003cp\u003e7 Besnard A, L. T., Levinson S, Chu D, Vicidomini C, Scobie KN, Dwork AJ, Arango V, Rosoklija GB, Mann JJ, Hen R, Leonardo ED, Boldrini M, Sahay A. Targeting Kruppel-like factor 9 (Klf9) in excitatory forebrain 1 neurons protects against chronic stress-induced dendritic spine enlargement and maladaptive fear responses. \u003cem\u003eCell Reports\u003c/em\u003e \u003cstrong\u003eIn Press\u003c/strong\u003e (2018).\u003c/p\u003e\n\u003cp\u003e8 Maitra, M.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Extraction of nuclei from archived postmortem tissues for single-nucleus sequencing applications. \u003cem\u003eNature Protocols\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 2788-2801 (2021).\u003c/p\u003e\n\u003cp\u003e9 Hao, Y.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Integrated analysis of multimodal single-cell data. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e184\u003c/strong\u003e, 3573-3587.e3529 (2021). https://doi.org:https://doi.org/10.1016/j.cell.2021.04.048\u003c/p\u003e\n\u003cp\u003e10 Hao, Y.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Dictionary learning for integrative, multimodal and scalable single-cell analysis. \u003cem\u003eNature Biotechnology\u003c/em\u003e (2023). https://doi.org:10.1038/s41587-023-01767-y\u003c/p\u003e\n\u003cp\u003e11 Stuart, T., Srivastava, A., Madad, S., Lareau, C. A. \u0026amp; Satija, R. Single-cell chromatin state analysis with Signac. \u003cem\u003eNature Methods\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 1333-1341 (2021). https://doi.org:10.1038/s41592-021-01282-5\u003c/p\u003e\n\u003cp\u003e12 Lawrence, M.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Software for Computing and Annotating Genomic Ranges. \u003cem\u003ePLOS Computational Biology\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, e1003118 (2013). https://doi.org:10.1371/journal.pcbi.1003118\u003c/p\u003e\n\u003cp\u003e13 Korsunsky, I.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Fast, sensitive and accurate integration of single-cell data with Harmony. \u003cem\u003eNature methods\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 1289-1296 (2019).\u003c/p\u003e\n\u003cp\u003e14 Franjic, D.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Transcriptomic taxonomy and neurogenic trajectories of adult human, macaque, and pig hippocampal and entorhinal cells. \u003cem\u003eNeuron\u003c/em\u003e \u003cstrong\u003e110\u003c/strong\u003e, 452-469.e414 (2022). https://doi.org:https://doi.org/10.1016/j.neuron.2021.10.036\u003c/p\u003e\n\u003cp\u003e15 Wang, W.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Transcriptome dynamics of hippocampal neurogenesis in macaques across the lifespan and aged humans. \u003cem\u003eCell Research\u003c/em\u003e \u003cstrong\u003e32\u003c/strong\u003e, 729-743 (2022). https://doi.org:10.1038/s41422-022-00678-y\u003c/p\u003e\n\u003cp\u003e16 Zhou, Y.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Molecular landscapes of human hippocampal immature neurons across lifespan. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e607\u003c/strong\u003e, 527-533 (2022). https://doi.org:10.1038/s41586-022-04912-w\u003c/p\u003e\n\u003cp\u003e17 Hochgerner, H., Zeisel, A., L\u0026ouml;nnerberg, P. \u0026amp; Linnarsson, S. Conserved properties of dentate gyrus neurogenesis across postnatal development revealed by single-cell RNA sequencing. \u003cem\u003eNature Neuroscience\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 290-299 (2018). https://doi.org:10.1038/s41593-017-0056-2\u003c/p\u003e\n\u003cp\u003e18 Ayhan, F.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Resolving cellular and molecular diversity along the hippocampal anterior-to-posterior axis in humans. \u003cem\u003eNeuron\u003c/em\u003e \u003cstrong\u003e109\u003c/strong\u003e, 2091-2105.e2096 (2021). https://doi.org:10.1016/j.neuron.2021.05.003\u003c/p\u003e\n\u003cp\u003e19 Habib, N.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Massively parallel single-nucleus RNA-seq with DroNc-seq. \u003cem\u003eNature methods\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 955-958 (2017).\u003c/p\u003e\n\u003cp\u003e20 Hao, Z.-Z.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Single-cell transcriptomics of adult macaque hippocampus reveals neural precursor cell populations. \u003cem\u003eNature Neuroscience\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 805-817 (2022). https://doi.org:10.1038/s41593-022-01073-x\u003c/p\u003e\n\u003cp\u003e21 van Bruggen, D.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Developmental landscape of human forebrain at a single-cell level identifies early waves of oligodendrogenesis. \u003cem\u003eDevelopmental Cell\u003c/em\u003e \u003cstrong\u003e57\u003c/strong\u003e, 1421-1436.e1425 (2022). https://doi.org:https://doi.org/10.1016/j.devcel.2022.04.016\u003c/p\u003e\n\u003cp\u003e22 Maitra, M.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Cell type specific transcriptomic differences in depression show similar patterns between males and females but implicate distinct cell types and genes. \u003cem\u003eNature Communications\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 2912 (2023). https://doi.org:10.1038/s41467-023-38530-5\u003c/p\u003e\n\u003cp\u003e23 Blankvoort, S., Olsen, L. C. \u0026amp; Kentros, C. G. Single Cell Transcriptomic and Chromatin Profiles Suggest Layer Vb Is the Only Layer With Shared Excitatory Cell Types in the Medial and Lateral Entorhinal Cortex. \u003cem\u003eFront Neural Circuits\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 806154 (2021). https://doi.org:10.3389/fncir.2021.806154\u003c/p\u003e\n\u003cp\u003e24 Di Bella, D. J.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Molecular logic of cellular diversification in the mouse cerebral cortex. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e595\u003c/strong\u003e, 554-559 (2021). https://doi.org:10.1038/s41586-021-03670-5\u003c/p\u003e\n\u003cp\u003e25 Morabito, S.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Single-nucleus chromatin accessibility and transcriptomic characterization of Alzheimer\u0026rsquo;s disease. \u003cem\u003eNature Genetics\u003c/em\u003e \u003cstrong\u003e53\u003c/strong\u003e, 1143-1155 (2021). https://doi.org:10.1038/s41588-021-00894-z\u003c/p\u003e\n\u003cp\u003e26 Grubman, A.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e A single-cell atlas of entorhinal cortex from individuals with Alzheimer\u0026rsquo;s disease reveals cell-type-specific gene expression regulation. \u003cem\u003eNature Neuroscience\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 2087-2097 (2019). https://doi.org:10.1038/s41593-019-0539-4\u003c/p\u003e\n\u003cp\u003e27 La Manno, G.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e RNA velocity of single cells. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e560\u003c/strong\u003e, 494-498 (2018). https://doi.org:10.1038/s41586-018-0414-6\u003c/p\u003e\n\u003cp\u003e28 Danecek, P.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Twelve years of SAMtools and BCFtools. \u003cem\u003eGigascience\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e (2021). https://doi.org:10.1093/gigascience/giab008\u003c/p\u003e\n\u003cp\u003e29 Bergen, V., Lange, M., Peidli, S., Wolf, F. A. \u0026amp; Theis, F. J. Generalizing RNA velocity to transient cell states through dynamical modeling. \u003cem\u003eNature Biotechnology\u003c/em\u003e \u003cstrong\u003e38\u003c/strong\u003e, 1408-1414 (2020). https://doi.org:10.1038/s41587-020-0591-3\u003c/p\u003e\n\u003cp\u003e30 Friedman, J. H., Hastie, T. \u0026amp; Tibshirani, R. Regularization Paths for Generalized Linear Models via Coordinate Descent. \u003cem\u003eJournal of Statistical Software\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 1 - 22 (2010). https://doi.org:10.18637/jss.v033.i01\u003c/p\u003e\n\u003cp\u003e31 Setty, M.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Characterization of cell fate probabilities in single-cell data with Palantir. \u003cem\u003eNature Biotechnology\u003c/em\u003e \u003cstrong\u003e37\u003c/strong\u003e, 451-460 (2019). https://doi.org:10.1038/s41587-019-0068-4\u003c/p\u003e\n\u003cp\u003e32 Liu, R.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Why weight? Modelling sample and observational level variability improves power in RNA-seq analyses. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e43\u003c/strong\u003e, e97 (2015). https://doi.org:10.1093/nar/gkv412\u003c/p\u003e\n\u003cp\u003e33 Phipson, B., Lee, S., Majewski, I. J., Alexander, W. S. \u0026amp; Smyth, G. K. ROBUST HYPERPARAMETER ESTIMATION PROTECTS AGAINST HYPERVARIABLE GENES AND IMPROVES POWER TO DETECT DIFFERENTIAL EXPRESSION. \u003cem\u003eAnn Appl Stat\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 946-963 (2016). https://doi.org:10.1214/16-aoas920\u003c/p\u003e\n\u003cp\u003e34 Zhu, A., Ibrahim, J. G. \u0026amp; Love, M. I. Heavy-tailed prior distributions for sequence count data: removing the noise and preserving large differences. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e35\u003c/strong\u003e, 2084-2092 (2019).\u003c/p\u003e\n\u003cp\u003e35 McCarthy, D. J., Campbell, K. R., Lun, A. T. L. \u0026amp; Wills, Q. F. Scater: pre-processing, quality control, normalization and visualization of single-cell RNA-seq data in R. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 1179-1186 (2017). https://doi.org:10.1093/bioinformatics/btw777\u003c/p\u003e\n\u003cp\u003e36 Amezquita, R. A.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Orchestrating single-cell analysis with Bioconductor. \u003cem\u003eNature Methods\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, 137-145 (2020). https://doi.org:10.1038/s41592-019-0654-x\u003c/p\u003e\n\u003cp\u003e37 Love, M. I., Anders, S. \u0026amp; Huber, W. Analyzing RNA-seq data with DESeq2. \u003cem\u003eBioconductor\u003c/em\u003e \u003cstrong\u003e2\u003c/strong\u003e, 1-63 (2017).\u003c/p\u003e\n\u003cp\u003e38 Love, M. I., Huber, W. \u0026amp; Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. \u003cem\u003eGenome Biology\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 550 (2014). https://doi.org:10.1186/s13059-014-0550-8\u003c/p\u003e\n\u003cp\u003e39 Subramanian, A.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. \u003cem\u003eProc Natl Acad Sci U S A\u003c/em\u003e \u003cstrong\u003e102\u003c/strong\u003e, 15545-15550 (2005). https://doi.org:10.1073/pnas.0506580102\u003c/p\u003e\n\u003cp\u003e40 Korotkevich, G.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Fast gene set enrichment analysis. \u003cem\u003ebioRxiv\u003c/em\u003e, 060012 (2021). https://doi.org:10.1101/060012\u003c/p\u003e\n\u003cp\u003e41 Liberzon, A.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e The Molecular Signatures Database (MSigDB) hallmark gene set collection. \u003cem\u003eCell Syst\u003c/em\u003e \u003cstrong\u003e1\u003c/strong\u003e, 417-425 (2015). https://doi.org:10.1016/j.cels.2015.12.004\u003c/p\u003e\n\u003cp\u003e42 Kanehisa, M., Furumichi, M., Tanabe, M., Sato, Y. \u0026amp; Morishima, K. KEGG: new perspectives on genomes, pathways, diseases and drugs. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e45\u003c/strong\u003e, D353-d361 (2017). https://doi.org:10.1093/nar/gkw1092\u003c/p\u003e\n\u003cp\u003e43 Milacic, M.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e The Reactome Pathway Knowledgebase 2024. \u003cem\u003eNucleic Acids Research\u003c/em\u003e \u003cstrong\u003e52\u003c/strong\u003e, D672-D678 (2023). https://doi.org:10.1093/nar/gkad1025\u003c/p\u003e\n\u003cp\u003e44 Guti\u0026eacute;rrez-Sacrist\u0026aacute;n, A., Hern\u0026aacute;ndez-Ferrer, C., Gonz\u0026aacute;lez, J. R. \u0026amp; Furlong, L. I. psygenet2r: a R/Bioconductor package for the analysis of psychiatric disease genes. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 4004-4006 (2017). https://doi.org:10.1093/bioinformatics/btx506\u003c/p\u003e\n\u003cp\u003e45 Guti\u0026eacute;rrez-Sacrist\u0026aacute;n, A.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e PsyGeNET: a knowledge platform on psychiatric disorders and their genes. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e31\u003c/strong\u003e, 3075-3077 (2015). https://doi.org:10.1093/bioinformatics/btv301\u003c/p\u003e\n\u003cp\u003e46 Zhang, B. \u0026amp; Horvath, S. A general framework for weighted gene co-expression network analysis. \u003cem\u003eStat Appl Genet Mol Biol\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, Article17 (2005). https://doi.org:10.2202/1544-6115.1128\u003c/p\u003e\n\u003cp\u003e47 Ritchie, M. E.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e limma powers differential expression analyses for RNA-sequencing and microarray studies. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e43\u003c/strong\u003e, e47 (2015). https://doi.org:10.1093/nar/gkv007\u003c/p\u003e\n\u003cp\u003e48 Zhou, Y.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Metascape provides a biologist-oriented resource for the analysis of systems-level datasets. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 1523 (2019). https://doi.org:10.1038/s41467-019-09234-6\u003c/p\u003e\n\u003cp\u003e49 Boldrini, M.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Human hippocampal neurogenesis persists throughout aging. \u003cem\u003eCell stem cell\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 589-599. e585 (2018).\u003c/p\u003e\n\u003cp\u003e50 Boldrini, M.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Benzodiazepines and the potential trophic effect of antidepressants on dentate gyrus cells in mood disorders. \u003cem\u003eInt J Neuropsychopharmacol\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, 1923-1933 (2014). https://doi.org:10.1017/S1461145714000844\u003c/p\u003e\n\u003cp\u003e51 Brown, L. M. Quantitative shotgun proteomics with data-independent acquisition and traveling wave ion mobility spectrometry: a versatile tool in the life sciences. \u003cem\u003eAdv Exp Med Biol\u003c/em\u003e \u003cstrong\u003e806\u003c/strong\u003e, 79-91 (2014). https://doi.org:10.1007/978-3-319-06068-2_4\u003c/p\u003e\n\u003cp\u003e52 Shimada, K.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Global survey of cell death mechanisms reveals metabolic regulation of ferroptosis. \u003cem\u003eNat Chem Biol\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 497-503 (2016).\u003c/p\u003e\n\u003cp\u003e53 Yang, Wan S.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Regulation of Ferroptotic Cancer Cell Death by GPX4. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e156\u003c/strong\u003e, 317-331 (2014).\u003c/p\u003e\n\u003cp\u003e54 Oswald, E. S., Brown, L. M., Bulinski, J. C. \u0026amp; Hung, C. T. Label-free protein profiling of adipose-derived human stem cells under hyperosmotic treatment. \u003cem\u003eJ Proteome Res\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 3050-3059 (2011).\u003c/p\u003e\n\u003cp\u003e55 Dai, H., Meyer, M., Stepaniants, S., Ziman, M. \u0026amp; Stoughton, R. Use of hybridization kinetics for differentiating specific from non-specific binding to oligonucleotide microarrays. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, e86 (2002). https://doi.org:10.1093/nar/gnf085\u003c/p\u003e\n\u003cp\u003e56 Weng, L.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Rosetta error model for gene expression analysis. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 1111-1121 (2006).\u003c/p\u003e\n\u003cp\u003e57 Paweletz, C. P.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Application of an End-to-End Biomarker Discovery Platform to Identify Target Engagement Markers in Cerebrospinal Fluid by High Resolution Differential Mass Spectrometry. \u003cem\u003eJournal of Proteome Research\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 1392-1401 (2010).\u003c/p\u003e\n\u003cp\u003e58 Hendrickson, R. C.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e High Resolution Discovery Proteomics Reveals Candidate Disease Progression Markers of Alzheimer\u0026apos;s Disease in Human Cerebrospinal Fluid. \u003cem\u003ePloS one\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e (2015). https://doi.org:e0135365\u003c/p\u003e\n\u003cp\u003e10.1371/journal.pone.0135365\u003c/p\u003e\n\u003cp\u003e59 McAvoy, T.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Quantification of tau in cerebrospinal fluid by immunoaffinity enrichment and tandem mass spectrometry. \u003cem\u003eClin. Chem.\u003c/em\u003e \u003cstrong\u003e60\u003c/strong\u003e, 683-689 (2014).\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[{"identity":"a1dd4dea-8a6b-4676-91b6-e3d524c2f747","identifier":"10.13039/100000002","name":"National Institutes of Health","awardNumber":"AI164769, AG076949, MH133561, AG080790","order_by":0}],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Columbia University","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-8593764/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8593764/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eHippocampal neuroplasticity regulates memory and emotional responses, which are altered in major depression, a leading cause of disability worldwide\u003csup\u003e1\u003c/sup\u003e. Molecular underpinnings of reduced hippocampal neuroplasticity\u003csup\u003e2\u003c/sup\u003e, fewer neurons, and smaller volume\u003csup\u003e3,4\u003c/sup\u003e in MDD are unknown and may involve blunted adult hippocampal neurogenesis (AHN), a debated phenomenon in human brain\u003csup\u003e5\u0026ndash;8\u003c/sup\u003e. Dynamic changes in gene expression (GEX) are mediated, in part, by cis-regulatory elements (CREs), in response to developmental cues, environmental signals, and cellular contexts. However, the accessibility of CREs and their relation to GEX in specific hippocampus cell types in MDD is unknown. To this end, we simultaneously profiled chromatin accessibility and GEX in 349,847 human hippocampal nuclei, used machine learning\u003csup\u003e9,10\u003c/sup\u003e, RNA Velocity\u003csup\u003e11\u003c/sup\u003e and pseudotime trajectory inference\u003csup\u003e12\u003c/sup\u003e to investigate the presence of progenitor cells and their progeny, and spatial transcriptomics in intact hippocampus sections to sequence cytoplasm and neuronal processes providing anatomical transcriptome mapping. We identified Type I, II and III neural progenitors and immature granule neurons. We find hyperactivated inflammation and apoptosis in immature and mature granule neurons, possibly affecting AHN and cell survival in MDD. Proteomics and transcriptomics convergent dysregulations in MDD, point to lipid metabolism, neurogenesis, synaptogenesis, cell adhesion, and plasticity. Identified molecular markers may drive biomarker and drug discovery for MDD and hippocampus resilience.\u003c/p\u003e","manuscriptTitle":"Human hippocampus niche and circuit neurotypical and major depression profiles","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-16 14:09:10","doi":"10.21203/rs.3.rs-8593764/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"62cd4148-1829-4021-97ec-b40c4d47cfed","owner":[],"postedDate":"January 16th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":61143765,"name":"Cellular \u0026 Molecular Neuroscience"}],"tags":[],"updatedAt":"2026-01-22T21:35:09+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-16 14:09:10","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8593764","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8593764","identity":"rs-8593764","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}