Altered Vascular and Angiogenic Transcripts in the Neurogenic Niche in Neuroinflammatory-Type Schizophrenia

Abstract

Background and Hypothesis

Growing evidence indicates that vascular processes, including blood–brain barrier (BBB) function and angiogenesis, may be altered in schizophrenia and related to neuroinflammation. The subependymal zone (SEZ) neurogenic niche shows reduced neurogenesis markers in schizophrenia that are more pronounced with neuroinflammation. Since inflammatory neuropathology is related to increased diapedesis-related transcripts, we hypothesize that endothelial cells would be impacted in this neurogenic niche in schizophrenia.

Study Design

We measured the expression of four BBB-related mRNAs [claudin-5 (CLDN5), occludin (OCLN), platelet endothelial cell adhesion molecule-1 (PECAM1), and tight junction protein 1 (TJP1)] and five angiogenesis-related mRNAs [angiopoietin 1 (ANGPT1), angiopoietin 2 (ANGPT2), vascular endothelial growth factor A (VEGFA), vascular endothelial growth factor receptor 1 (VEGFR1), and TEK receptor tyrosine kinase (TEK)] via quantitative polymerase chain reaction, followed by semi-quantitative immunofluorescence for claudin-5 and collagen-IV protein. We estimated proportions of vascular cells by running cellular deconvolution on previous bulk RNA sequencing data.

Study Results

In high-inflammation schizophrenia, we found increased PECAM1 mRNA, CLDN5 mRNA and claudin-5 protein expression potentially relating to leukocyte trafficking and repair of endothelial tight junctions. The estimated proportion of vascular cells and VEGFA mRNA levels were increased, potentially indicating increased angiogenesis in high-inflammation schizophrenia. Higher VEGFA and CLDN5 mRNA levels were associated with higher expression of markers of immune cell transmigration but lower expression of immature neuron markers, suggesting that vascular dysfunction may impact neurogenesis.

Conclusions

These findings reveal that changes in the BBB and angiogenesis are more severe in high-inflammation schizophrenia and appear to be linked to reduced neurogenesis in the SEZ. This study underscores the importance of inflammation in shaping vascular and neurogenic pathology in schizophrenia, offering potential pathways for future therapeutic exploration.

Introduction

Changes to the vasculature exist in schizophrenia brain tissue, including molecular and functional changes leading to a leakier blood–brain barrier (BBB).^1,2 The human subependymal zone^3,4 (SEZ; also termed subventricular zone) is the brain’s largest neurogenic niche,⁵ where unique vasculature regulates all stages of neurogenesis, impacting neuronal reserves over the lifespan. Since we found blunted neurogenesis and neuroinflammation in schizophrenia,^6,7 we expected that molecular and cellular blood vessel pathology would be detectable in the SEZ. While relatively understudied, SEZ neurogenesis is highly relevant to schizophrenia, as the SEZ produces inhibitory interneurons, whose dysfunction is a hallmark of the disorder.⁸ Normally, the SEZ is highly vascularized, with many blood vessels lacking astrocytic end feet, increasing BBB permeability and peripheral communication.^9-11 In schizophrenia, the SEZ shows elevated macrophage density and transcriptomic signs of potential macrophage infiltration across the BBB.¹² Brain macrophages can originate from bone marrow and migrate between endothelial cells, requiring tight junction remodeling (eg, impacting claudin-5).¹³ However, the vascular changes in the schizophrenia SEZ remain untested, particularly whether BBB integrity molecules (structural) or angiogenesis (aberrant vessel growth) are altered.

About 40% of people with schizophrenia show elevated brain inflammation, based on elevated mRNA expression of key pro-inflammatory molecules [interleukin 6 (IL6), interleukin 1 beta (IL1B), C-X-C motif chemokine ligand 8 (CXCL8), interleukin 6 receptor (IL6R), interleukin 1 receptor type 1 (IL1R1), serpin family A member 3 (SERPINA3), and interleukin 6 signal transducer (IL6ST)].^7,14 In this high-inflammation subgroup, the SEZ has increased CD163+ macrophages, reduced microglia markers, and blunted neurogenesis markers [achaete-scute family bHLH transcription factor 1 (ASCL1), DLX6 antisense RNA 1 (DLX6-AS1), and doublecortin (DCX) mRNAs].^7,14 We previously found upregulated ICAM1 (intercellular adhesion molecule 1; that captures circulating white blood cells) in the SEZ,^7,14 cortex¹⁵ and midbrain¹⁶ of high-inflammation schizophrenia cases. Soluble or shed ICAM1 is also elevated in schizophrenia serum.^17,18 Further, RNA sequencing suggests inflammation in schizophrenia primarily affects the SEZ microenvironment including the extracellular matrix and the vasculature.¹⁹ Similar inflammatory subgroups appear in the midbrain,²⁰ where specific angiogenesis-related transcripts such as angiopoietins and SERPINE1 are increased.¹

Angiogenesis is the formation of new blood vessels by sprouting or remodeling. It is increased with inflammation²¹ and can perpetuate inflammation, as developing vessels have greater permeability due to insufficient tight junction proteins.^22,23 Angiogenesis also relates to neurogenesis through overlapping growth factor responsiveness/function.^24,25 Key regulators include vascular endothelial growth factor (VEGF), angiopoietins (ANGPT1, ANGPT2) and their respective receptors VEGFR1 (alias: FLT1) and TEK (alias: TIE2); which were a focus in this study. Angiopoietin-1 and angiopoietin-2 have opposing effects on the TIE2 receptor, with angiopoietin-1 being an agonist leading to vessel stabilization and maturation and angiopoietin-2 being an antagonist leading to vascular plasticity promoting angiogenesis.^26,27 VEGF and ANGPT2, produced by endothelial and immune cells,²⁸ are elevated in schizophrenia,^29,30 suggesting a link between angiogenesis and inflammation.³¹ We hypothesize that in the SEZ, people with schizophrenia and high inflammation would exhibit elevated angiogenesis-related mRNAs.

BBB alterations have been implicated in the pathophysiology of schizophrenia,^15,32,33 but the SEZ BBB remains underexplored, despite the vasculature’s role neurogenesis.³⁴ We investigated key BBB components including mRNA expression of the adhesion molecule platelet endothelial cell adhesion molecule-1 (PECAM1) (which works with ICAM1 in immune cell diapedesis) and tight junction-related proteins claudin-5 (CLDN5), occludin (OCLN), and tight junction protein 1 (TJP1). Tight junction proteins encoded by CLDN5 and OCLN and their assembly component TJP1 (alias: ZO-1), tightly connect endothelial cells to regulate the diffusion of molecules.³⁵  CLDN5 lies within the 22q11 deletion syndrome (DiGeorge) locus, where 30% of deletion carriers (haploinsufficient for CLDN5) develop schizophrenia.² Suppression of CLDN5 in mice suggests a pathogenic role of altered tight junction proteins in schizophrenia-relevant behaviors and treatment response.² In schizophrenia brain tissue, there is increased CLDN5 mRNA expression in the dorsolateral prefrontal cortex (DLPFC),³⁶ the orbitofrontal cortex, and the occipital cortex,³⁷ but claudin-5 protein levels were reduced in the DLPFC³⁶ and hippocampus.³⁷ Further, schizophrenia patient-derived blood endothelial-like cells have an altered BBB signature, including higher TJP1 mRNA expression and changes to angiogenesis responsiveness.³⁸ Taken together, there may be fundamental changes to blood vessels in the brains of people with schizophrenia, and focusing on the SEZ will allow us to illuminate this pathology in relation to macrophage and neurogenic changes with implications for the perturbed development of interneurons.

Methods

Cohorts, Tissue Processing, and Gene Expression Quantification

Post-mortem human brain tissue was provided by the Stanley Medical Research Institute (Rockville, USA) and the NSW Brain Tissue Resource Centre (Sydney, Australia). The final cohort (n = 106) comprised 54 people with schizophrenia and 52 unaffected controls, which was matched for demographic factors except for brain pH, which is known to be lower in schizophrenia³⁹ (Table S1). The brain banks excluded cases with a history of central nervous system disease and/or cases with brain pathology; however, having an active peripheral inflammatory disease at the time of death was not part of the exclusion criteria, and information regarding the use of anti-inflammatory medication around the time of death was not available. The samples from these cohorts have previously been categorized into low- and high-inflammation subgroups based on a two-step recursive cluster analysis of their SEZ mRNA expression of IL6, IL1B, CXCL8, IL6R, IL1R1, SERPINA3, and IL6ST^7,14 (see Table S2 for inflammation subgroup demographics). The study was carried out in accordance with the Declaration of Helsinki and was approved by the University of New South Wales Human Research Ethics Committee (HREC 12435, HC 230253).

Methods for tissue processing, dissections, RNA extraction, complementary DNA (cDNA) synthesis, and gene expression measurements have been previously described.¹² Briefly, the SEZ was dissected along the lateral wall of the lateral ventricle from 60 μm coronal sections of fresh-frozen brain tissue from the anterior-third of the caudate nucleus at the level shown in photographs on pages 121-123 of the Atlas of the Human Brain.⁴⁰ Total RNA was extracted using TRIzol as per the manufacturer’s protocol (Thermo Fisher Scientific). cDNA was synthesized from 2 μg total RNA using the SuperScript® First-Strand Synthesis kit IV employing random hexamers (Thermo Fisher Scientific). Gene expression was measured with quantitative polymerase chain reactions (qPCRs) on the BioMark HD system (Fluidigm, South San Francisco, CA, USA) or ABI Prism 7900HT fast real-time PCR (Applied Biosystems, Foster City, CA, USA) using Taqman probes (Table S3). Negative controls, including no template or no reverse-transcriptase had no detectable expression for any of the mRNA targets studied.

Target Gene Selection and Their SEZ Cell Type Specificity

The choice of target genes for qPCR analysis was based on their known expression in endothelial cells, and/or their function being associated with the BBB and angiogenesis. However, we used total SEZ RNA without specifically isolating or enriching for the vasculature. Therefore, to confirm in which SEZ cell types the target genes are most highly expressed, we plotted the average expression of our chosen target genes across human SEZ cell type clusters from a single nuclei RNA sequencing (snRNAseq).⁴¹ In short, this data comprises 20 cell-type clusters that were identified in the human SEZ, including vascular cells (cluster 15). The expression of each target gene was calculated based on the normalized average expression of that gene from nuclei in each cell-type cluster. This was converted to a z-score for graphing/visualization due to the large variation in average expression levels across the different target genes. We calculated the fold-change difference between target gene expression in vasculature cells by dividing average target gene expression in the vasculature by the mean of their average expression in all other cell types. We evaluated statistical significance of the fold-changes in “vasculature” compared to other cell types using a permutation test for each target gene (10 000 permutations). The P-value was the proportion of permutations where the permuted fold change exceeded the observed fold change. A threshold of P < .05 was considered statistically significant.

Vascular Cell Proportion Estimates Using Bulk RNA Sequencing Data

We used CIBERSORTx⁴² to estimate the proportion of SEZ vascular cells and compared proportions across disease state and inflammation subgroups. Input data included transcripts per million expression values from our previous bulk RNAseq experiment in a subset of this cohort (n = 54) including schizophrenia and controls,¹² as well as inflammation subgroups of schizophrenia.¹⁹ We used data from the snRNAseq of the SEZ in healthy individuals⁴¹ as a custom cell type signature matrix. Marker genes for each of 20 initial clusters were identified using adjusted P-value (<.05) and log2 fold change (>0.25), then selecting the top 150 genes per cell cluster. Duplicated genes were removed, yielding 1957 unique marker genes. Average expression per cluster was extracted from a full transcriptomic expression matrix. Clusters with overlapping cell type identity and transcriptional profiles (eg, MSPNs, Astro_NSC) were merged based on principal component analysis (PCA) proximity and naming, by averaging gene expression values across merged clusters. The job parameters in CIBERSORTx were the “impute cell fractions” job type, B-mode batch correction, quantile normalization disabled, and the “relative” run mode.

Immunofluorescence, Scanning Microscopy, and Quantification

A double-label immunofluorescence for claudin-5 and collagen-IV was performed on fresh frozen human SEZ tissue (14 μm sections) from schizophrenia and control groups (n = 32 per group) from the Stanley Medical Research Institute. Scientists were blinded to the subject diagnoses for all stages of the experiment, image acquisition, and quantification. The slides were thawed at room temperature and fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS; pH 7.4) for 10 min. Slides were then blocked with 10% normal blocking serum (NBS; Millipore, Temecula, CA, USA) for 60 min at room temperature. After draining the NBS, the primary antibodies specific to claudin-5 (1:500 dilution, hosted in mouse; Invitrogen 35-2500) and collagen-IV (1:1000 dilution, hosted in rabbit; Abcam ab6568) were applied to the slides and left to incubate at 4 °C overnight. Slides with no primary antibodies were run as controls. Slides then underwent 3 × 5-min washes in PBS, and the fluorescent secondary antibodies were applied to the slides for 60 min in the dark (both 1:500 dilution; Thermofisher, Alexa Fluor 555, A-31570 hosted in mouse for claudin-5; Thermofisher, Alexa Fluor 647, A-31573 hosted in rabbit for collagen-IV). After the slides were washed in PBS, slides were washed in 5 mM cupric sulfate (Sigma-Aldrich) and 50 mM ammonium acetate (Sigma-Aldrich) for 30 min to minimize autofluorescence. The slides were then washed in PBS and counterstained with 1:1000 DAPI (Sigma-Aldrich). The slides were coverslipped with anti-fade mounting media (Citifluor AF1, ProSciTech, Kirwan, AUS) and stored at 4 °C.

Images were captured on the Olympus VS200 E26 scanning microscope with the 20x objective. The channels used were DAPI, Cy3 for claudin-5 (555 nm) and Cy5 for collagen-IV (647 nm). Due to some blood vessels or tissue being folded, z-stacking was utilized to increase clarity and image quality of blood vessels.

QuPath v0.4.0 was used for image analysis and quantification. The SEZ was outlined in the DAPI channel and was defined as the region of tissue with sparse cells and then densely packed nuclei between the monolayer of ependymal cells along the lateral wall of the lateral ventricle and the caudate nucleus (capture area on average ~ 2.4 mm² ± 1.2 standard deviation). Once the entire SEZ on each slide was outlined, the area within that region that also contained claudin-5 and/or collagen-IV was also determined in their fluorescent channels. The threshold classification tool was used to set a threshold for staining significantly above background, and this was used to determine the percentage of stained area in the entire SEZ immunopositive for claudin-5 (pixel level of 550) and for collagen-IV (pixel level of 350, minimum pixel size 30 μm²). The no primary antibody control slides were used to determine threshold cut-offs. The percentage of claudin-5 per blood vessel was quantified by measuring the overlapping area of claudin-5 and collagen-IV fluorescence as a percentage of total collagen-IV area.

Statistical Analyses

IBM SPSS Statistics Version 28 (IBM, Armonk, NY, USA) was used to conduct statistical analysis, and GraphPad Prism Version 7 (GraphPad Software, La Jolla, CA, USA) was used to plot the data. Target gene expression (transcript of interest) was normalized to the geometric mean of three housekeeper transcripts (UBC, GAPDH, and TBP), and group outliers were removed if the levels were more than two standard deviations from the mean of that gene (1-5 per diagnostic group). The geometric mean of housekeepers did not differ according to group. Normality was checked with the Shapiro–Wilk analysis and data that did not pass the normal distribution test was log transformed (CLDN5, PECAM1, VEGFA and ANGPT2 mRNAs). Pearson’s product–moment and Spearman’s rank correlations were conducted to assess the relationship between all target genes and demographic, post-mortem, and clinical factors across all groups. Correlations between target genes and the clinical factors “age of disease onset,” “duration of illness,” and “lifetime antipsychotic dose” are detailed in Table S4. Correlations between target genes and the potential covariates including age, brain pH, RNA integrity number (RIN), or post-mortem interval (PMI) are detailed in Table S5. Significant correlations with covariates were statistically controlled for with ANCOVA or Quade’s rank analysis of covariance, except for brain pH because it is reduced in schizophrenia³⁹ and associated with inflammation.⁴³ Otherwise, independent samples t-tests or Mann–Whitney U tests were used to detect group differences in diagnoses, and ANOVA (with Least Significant Differences post hoc test) or Welch’s ANOVA were used to detect group differences in inflammation subgroups. Pearson’s product–moment (R), Spearman’s rank (rho), partial (r), or semi-partial (sr) correlation analysis were used to assess the relationship between target genes and markers for immune cells or neurogenesis. An alpha level of 0.05 was considered statistically significant.

Results

Expression of PECAM1 and TJP1 mRNAs, but not CLDN5 or OCLN mRNAs, Were Significantly Altered in Schizophrenia Compared to Controls

We first compared the mRNA expression levels of adhesion molecule PECAM1 and tight-junction proteins (CLDN5, OCLN, TJP1) between schizophrenia and controls. PECAM1 was upregulated by 50% in schizophrenia compared to controls [ANCOVA (RIN) F (2, 82) = 15.72, P < .001; Figure 1A]. Expression of CLDN5 (P = .09) and OCLN (P = .062) did not significantly differ between diagnostic groups (Figure 1B and C). TJP1 mRNA was downregulated by 8% in schizophrenia compared to controls [t (100) = 2.0, P = .047; Figure 1D].

Figure 1.

mRNA expression of blood–brain barrier-associated genes in the SEZ in schizophrenia and control brains (A) Adhesion molecule PECAM1 was significantly increased in schizophrenia compared to controls. (B and C) Tight junction proteins CLDN5 and OCLN mRNA were not significantly different. (D) Tight junction assembly factor TJP1 (alias ZO-1) was significantly reduced in schizophrenia. (E) snRNAseq data from the SEZ of healthy individuals,⁴¹ indicates PECAM1, CLDN5, and OCLN are primarily expressed in SEZ vasculature cells, whereas TJP1 has higher mean expression in oligodendrocytes. In A-D, data are shown as a percentage of control mean ± standard error of the mean. In E, data are graphed as z-scores representing the average expression of the gene per cell-type cluster, where clusters were determined by hierarchical clustering of transcriptomic profiles, and labeled on the basis of expression quantitation of cell-type characteristic markers. ^*P < .05; ^***P < .001; Astro/NSC, astrocyte/neural stem cells; MSPNs, medium spiny neurons; SST, somatostatin; OPCs, oligodendrocyte progenitor cells; Oligos, oligodendrocytes; CLDN5, claudin 5; OCLN, occludin; PECAM1, platelet endothelial cell adhesion molecule-1; SEZ, subependymal zone; TJP1, tight junction protein 1; snRNAseq, single nuclei RNA sequencing; SCZ, schizophrenia

To determine in which human SEZ cell types these genes are normally expressed, we plotted their average expression across 20 different cell types from a previously published snRNAseq analysis of the post-mortem SEZ and adjacent caudate tissue from eight healthy adults.⁴¹  PECAM1, CLDN5, and OCLN mRNAs were primarily expressed in vasculature cells (Figure 1E; fold-change in vasculature of 25.28, 10.26, and 4.33 respectively; all P < .0001, permutation test). However, TJP1 was expressed across a range of SEZ cell types including highest expression levels in oligodendrocytes (fold-change in vasculature = 0.85; P = .447, permutation test).

CLDN5 and PECAM1 Were Significantly Altered in HI-SCZ

To determine the role of inflammatory state on the BBB in schizophrenia, we analyzed the BBB gene expression between low-inflammation controls (LI-CTRL), low-inflammation schizophrenia (LI-SCZ), and high-inflammation schizophrenia (HI-SCZ). The high inflammation control group was too small (n = 6) for robust statistical analyses, but these cases are plotted on the graphs as gray points for reference. PECAM1 was significantly increased in a stepwise manner from LI-CTRL to LI-SCZ to HI-SCZ, with HI-SCZ having double the LI-CTRL level of PECAM1 [Welch’s ANOVA, F (2, 75) = 12.91, P < .001; all significant post hocs P ≤ .015; Figure 2A]. CLDN5 was significantly increased in HI-SCZ compared to LI-CTRL and LI-SCZ by ~100% [F (2,90) = 16.85, P < .001; all significant post hocs P < .001; this remains significant when controlling for age as a covariate (P < .001) Figure 2B]. OCLN and TJP1 did not significantly differ between inflammatory subgroups (both F < 1.65, P > .19; Figure 2C and D).

Figure 2.

mRNA expression of blood–brain barrier-associated genes in the SEZ in inflammation subgroups of schizophrenia and controls (A) Adhesion molecule PECAM1 mRNA was significantly increased, in a stepwise manner, from LI-CTRLs to LI-SCZ to HI-SCZ. (B) CLDN5 was significantly increased specifically in HI-SCZ compared to LI-SCZ and LI-CTRLS. (C and D) OCLN and TJP1 did not significantly differ based on inflammatory subgroup. Data are shown as a percentage of LI-CTRL mean ± standard error of the mean. High-inflammation controls are represented in gray but were not included in statistical analyses of group-wise comparisons due to low representation of this subgroup (n = 6). ^*P < .05; ^**P < .01; ^****P < .0001; HI, high inflammation; LI, low inflammation; SCZ, schizophrenia; CLDN5, claudin 5; OCLN, occludin; PECAM1, platelet endothelial cell adhesion molecule-1; TJP1, tight junction protein 1; LI-CTRL, low inflammation controls; LI-SCZ, low-inflammation schizophrenia; HI-SCZ, high-inflammation schizophrenia

Claudin-5 Protein is Localized to Blood Vessels and is Increased in HI-SCZ

Consistent with our snRNAseq data, we found that claudin-5 protein was localized exclusively to blood vessels, which was evident by overlap with collagen-IV (Figure 3A). There was large variation in the claudin-5 coverage of blood vessels with both small and large vessels displaying minimal (Figure 3B and E), moderate (Figure 3C and F), or high (Figure 3D and G) coverage of claudin-5. In rare instances, blood vessels were seemingly devoid of claudin-5 (Figure 3H). Within the human SEZ, a higher percentage of claudin-5 stained area was associated with increased CLDN5 mRNA expression (rho = 0.286, P = .028; n = 59 overlapping cases).

Figure 3.

Claudin-5 and collagen-IV protein localization and quantification in the human SEZ (A) Typical blood vessel showing claudin-5 (yellow) overlapping with blood vessel marker collagen-IV (red) and endothelial cell nuclei (dapi/blue), separated channels are shown sequentially. (B-G) Both small and large blood vessels had varied claudin-5 coverage ranging from minimal, moderate and high. (H) some blood vessels were devoid of claudin-5 and may represent angiogenesis. (I) Percentage area of blood vessels in the SEZ with detectable claudin-5 protein was significantly increased HI-SCZ. (J and K) Collagen-IV and the ratio of claudin-5/collagen-IV did not significantly differ across groups. (I-K) Data are shown as a percentage of LI-CTRL mean ± standard error of the mean. High-inflammation controls are represented in gray but were not included in statistical analyses of group-wise comparisons due to low representation of this subgroup (n = 6). ^**P < .01; HI, high inflammation; LI, low inflammation; SCZ, schizophrenia; SEZ, subependymal zone; HI-SCZ, high-inflammation schizophrenia; LI-CTRL, low-inflammation controls

The percentage of claudin-5 and collagen-IV stained area did not significantly differ between controls and schizophrenia. However, the percentage of claudin-5 positive SEZ area was increased by about 50% in HI-SCZ compared to LI-SCZ and LI-CTRL [F (2, 48), P = .010; all significant post hocs P ≤ .007; Figure 3I]. In contrast, the percentage of collagen-IV stained area did not significantly differ between inflammatory subgroups [F (2, 49) = 3.69, P = .135; Figure 3J]. The percentage of claudin-5 per blood vessel area did not significantly differ across diagnosis [t(58) = 0.44, P = .66] or inflammation subgroups [F (2, 47) = 0.11, P = .89; Figure 3K]. No clinical variables correlated with the percentage of claudin-5 or collagen-IV stained area (Table S6).

ANGPT1 and ANGPT2 mRNAs Were Significantly Downregulated in Schizophrenia Compared to Controls whereas VEGF mRNA was not Changed

We next determined whether the mRNA expression of proteins regulating angiogenesis was altered in schizophrenia. ANGPT1 and ANGPT2 mRNAs were significantly lower in schizophrenia compared to controls [F (2, 82) = 6.93, P = .004; t (97) = 2.8, P = .006, respectively; Figure 4A and B]. In contrast, the mRNAs encoding another angiogenic growth factor VEGFA, its receptor VEGFR1, and the receptor of angiopoietins, TEK, were not significantly different across diagnostic groups (all P > .014; Figure 4C, D, and  E). The estimated proportion of SEZ vascular cells (using cellular deconvolution of bulk RNAseq data) was increased in schizophrenia (mean = 9.12%) compared to controls [mean = 6.64%; ANCOVA (age, RIN) F (1, 50) = 16.67, P < .001; Figure 4F].

Figure 4.

mRNA expression of angiogenesis-associated markers in the SEZ in schizophrenia and controls (A and B) Although ANGPT1 and ANGPT2 have divergent functions, their mRNA was significantly reduced in schizophrenia compared to controls. (C) VEGFA and (D) its receptor VEGFR1 mRNA did not significantly differ by diagnosis. (E) Angiopoietin receptor TEK did not significantly differ by diagnosis. (F) The estimated proportion of vascular cells using cellular deconvolution of bulk RNAseq suggests a higher proportion of vascular cells in schizophrenia compared to controls. (G) snRNAseq data from the SEZ of healthy individuals,⁴¹ indicates that ANGPT1 is primarily expressed in ependymal and Astro/NSC, ANGPT2 had its highest mRNA expression in immature neurons, VEGFA was expressed in proliferative cell types like Astro/NSCs and OPCs, and VEGFR1 was specific to vasculature. TEK was not detected in that snRNAseq study. For A-E, data are shown as a percentage of control mean ± standard error of the mean. In F data are graphed as z-scores representing the average expression of the gene per cell-type cluster, where clusters were determined by hierarchical clustering of transcriptomic profiles, and labeled on the basis of expression quantitation of cell-type characteristic markers. ^**P < .01; SCZ, schizophrenia; Astro/NSC, astrocyte/neural stem cells; MSPNs, medium spiny neurons; SST, somatostatin; OPCs, oligodendrocyte progenitor cells; Oligos, oligodendrocytes; SEZ, subependymal zone; ANGPT1, angiopoietin-1; ANGPT2, angiopoietin-2; snRNAseq, single nuclei RNA sequencing

Interestingly, by snRNAseq, we found that ANGPT1 mRNA was most highly expressed in SEZ cell types associated with ependymal and astrocytes/neural stem cells, ANGPT2 mRNA was most highly expressed in immature neurons, whereas VEGFA mRNA was expressed in a range of immature cell types including stem cells, neuroblasts, and oligodendrocyte progenitor cells (fold-change in vasculature of 1.19, 0.30, and 0.24, respectively; all P > .334, permutation test). In contrast, VEGFR1 was quite specifically found in cells of the vasculature (Figure 4G; fold change in vasculature = 7.68; P < .0001, permutation test). TEK mRNA was not identified in the previous snRNAseq, likely due to low sequencing depth.

ANGPT1, ANGPT2, and VEGFA mRNA Expression was Changed between Inflammatory Subgroups

ANGPT1 was reduced in LI-SCZ by ~30% compared to LI-CTRL, [F (2, 74) = 4.10, P = .021; post hoc P = .012; Figure 5A]. ANGPT2 was reduced in LI-SCZ compared to LI-CTRL and HI-SCZ [F (2, 90) = 9.24, P < .001; all significant post hocs P ≤ .03; Figure 5B]. In contrast, VEGFA mRNA was increased by ~50% in HI-SCZ compared to both LI-CTRL and LI-SCZ [F (2, 90) = 9.24, P < .001; all significant post hocs P < .001; Figure 5C]. VEGFR1 and TEK were not significantly different across inflammatory subgroups (both P > .372; Figure 5D and E). The estimated proportion of SEZ vascular cells was increased in HI-SCZ (mean = 11.05%) compared to LI-CTRL (mean = 6.64%) and LI-SCZ [mean = 7.04%; ANCOVA (age, RIN) F (2, 49) = 26.54, P < .0001; both significant post hocs P ≤ .001; Figure 5F].

Figure 5.

mRNA expression of angiogenesis-associated markers in the SEZ in inflammation subgroups of schizophrenia and controls (A) ANGPT1 mRNA was reduced in LI-SCZ compared to LI-CTRLs. (B) ANGPT2 mRNA was reduced in LI-SCZ compared to LI-CTRLs and LI-SCZ. (C) VEGFA mRNA was increased in HI-SCZ compared to low inflammation groups. (D and E) VEGFR1 and TEK were not significantly changed. (F) There is a higher estimated proportion of vascular cells in HI-SCZ compared to LI-SCZ and LI-controls. Data are shown as a percentage of LI-CTRL mean ± standard error of the mean. High-inflammation controls are represented in gray but were not included in statistical analyses of group-wise comparisons due to low representation of this subgroup (n = 6). ^*P < .05; ^***P < .001; bold type (in F) = P ≤ .05; HI, high inflammation; LI, low inflammation; NSC, neural stem cell; rho, Spearman’s correlation coefficient; SCZ, schizophrenia; sr, semi-partial correlation coefficient; ANGPT1, angiopoietin-1; LI-SCZ, low-inflammation schizophrenia; LI-CTRL, low-inflammation controls

CLDN5 and VEGFA Correlate with Infiltrating Macrophage and Neurogenesis Markers

We next determined if the significant changes in vascular related mRNAs (CLDN5 and VEGFA, which were both increased in HI-SCZ) were associated with markers for immune cells and/or their adhesion (PECAM1, ICAM1, CD163) and/or with markers for different stages of neurogenesis [GFAPD (stem cells), ASCL1 (neuronal progenitor), DLX6-AS1 (immature neuron), DCX (immature neuron)], which we previously measured in this cohort.⁷ Higher transcript levels of both CLDN5 and VEGFA were associated with higher transcript levels of PECAM1 and CD163 mRNAs, with CLDN5 additionally correlating positively with ICAM1 mRNA (Table S7). Interestingly, higher CLDN5 mRNA was associated with higher quiescent neural stem cell marker GFAPD but with lower expression of the immature neuron markers DLX6-AS1 and DCX mRNA (Table S7). Higher VEGFA mRNA levels were associated with lower DLX6-AS1 (Table S7).

Overall, the correlations between CLDN5 mRNA and cell adhesion, macrophage, and neurogenesis mRNA markers were more numerous and of a greater strength as compared to those with VEGF mRNA.

Correlations with Clinical Features

None of the BBB or angiogenesis related genes or proteins of interest in this study significantly correlated with lifetime antipsychotic dose (all r ≤ 0.236, P ≥ .096) or age of schizophrenia onset (all r ≤ 0.183, P ≥ .096; Tables S4 and S6). PECAM1 and ANGPT2 significantly correlated with duration of illness (both r ≥ 0.298, P ≤ .035), but no other genes or proteins of interest significantly correlated.

Discussion

This study identifies significant molecular alterations to the vasculature and in angiogenesis-related transcripts that impact the vasculature in the SEZ of individuals with schizophrenia, particularly the subgroup with elevated inflammation. We find that these vascular changes are linked to indices of macrophage attraction and suppression of neurogenesis. We confirmed that CLDN5, PECAM1, OCLN and VEGFR1 are predominantly expressed by vasculature cells, whereas angiogenesis regulating factors (ANGPT1, ANGPT2, VEGFA) are expressed across a range of SEZ neurogenic cells including immature neurons, neural stem cells, glia and ependymal cells. Using cellular deconvolution, we found the estimated proportion of vascular cells in the SEZ to be significantly higher in HI-SCZ. Our findings raise the interesting question as to whether vascular changes may precede or follow putative activation/accumulation of macrophages in the neurogenic niche. These findings support the hypothesis that vascular dysfunction and inflammation may lead to an altered microenvironment that impairs potential routes for brain plasticity in schizophrenia.

Angiogenesis May Relate to Immune Cell Transmigration in HI-SCZ

The increased estimated proportion of vascular cells, together with elevated VEGFA expression from non-vascular cells in HI-SCZ, may signify increased SEZ angiogenesis, which in turn can enhance BBB permeability and promote immune cell diapedesis.^21,44 This cyclical relationship is supported here by the positive correlations between VEGFA, adhesion molecule PECAM1, and macrophage marker CD163, and is consistent with our previous SEZ bulk RNA sequencing study showing VEGFA and VEGFR1 upregulation in high inflammation schizophrenia alongside associations between another adhesion molecule, ICAM1, and CD163.¹⁹

We found elevated PECAM1 and CLDN5 in neuroinflammatory schizophrenia (HI-SCZ), which may relate to the higher proportion of vascular cells in this group. However, other vascular-enriched mRNAs, such as OCLN and VEGFRA, were not changed in HI-SCZ, suggesting there could be other reasons for the PECAM1 and CLDN5 elevation and that there may be more PECAM1 and CLDN5 mRNA per endothelial cell. PECAM1 also plays a positive role in angiogenesis by supporting endothelial cell motility,⁴⁵ so higher PECAM1 could contribute to the increased proportion of SEZ endothelial cells found here in the schizophrenia SEZ.

One of the major differences found in schizophrenia brain is a higher CD163+ macrophage density,⁷ so the correlation between CLDN5 mRNA and diapedesis markers (PECAM1, ICAM1, CD163) supports that tight junction remodeling occurs alongside leukocyte migration. In fact, leukocyte transmigration across the BBB predominantly takes place through endothelial cell junctions (≥98%), triggering claudin-5 reorganization, with PECAM-1 itself being critical for transendothelial migration.¹³ Together supporting the hypothesis of vascular alterations in concert with macrophage transmigration in high inflammation schizophrenia.

While Lizano, Pong, Santarriaga, Bannai, Karmacharya⁴⁶ reported reduced CLDN5 mRNA in schizophrenia patient-derived brain microvascular endothelial cells, our findings indicate increased CLDN5 in the SEZ, specifically in HI-SCZ. These apparently discrepant findings could indicate that lower CLDN5 may lead to more permeable tight junctions and a leakier BBB earlier in the course of schizophrenia causing increased inflammation. However, when inflammation is present long term, which could be the case in our cohort, this may lead to more CLDN5 mRNA with higher accumulation of claudin-5 protein. Thus, our findings may reflect an adaptive response to environmental stressors, such as chronic inflammation, rather than a direct genetic effect. Our findings, in the SEZ, contrast with research from other brain regions where the BBB may be more intact. For example, in the hippocampus there is reduced claudin-5 protein³⁷ and in the prefrontal cortex there is reduced claudin-5 protein but elevated claudin-5 mRNA in schizophrenia.³⁶ In the SEZ, we found both claudin-5 protein and mRNA levels to be increased and positively correlated, suggesting that this region may undergo different BBB remodeling compared to other areas, which is not surprising considering the SEZ vasculature is already unique to support neurogenesis.⁴⁷ Notably, claudin-5 elevations appear most pronounced in schizophrenia cases with elevated inflammation, both here and in previous work in the DLPFC where a “high-inflammation” subtype of schizophrenia was determined from bulk transcriptomics,⁴⁸ highlighting the importance of stratifying by inflammatory status as this may help to resolve inconsistencies in the literature in future studies. These findings seem unique to SEZ inflammation in schizophrenia, considering other inflammation-related neurological disorders, such as multiple sclerosis, stroke and major depression, are associated with reduced claudin-5.³⁵ Although we found reduced TJP1 mRNA in schizophrenia, suggesting possible tight junction weakening, TJP1 is expressed across multiple SEZ cell types including oligodendrocytes, so its functional implications for the BBB should be interpreted cautiously. Together, these results could have various interpretations, including that claudin-5 upregulation is due to increased blood vessels and angiogenesis, or that it may represent an active, inflammation-driven attempt to reinforce the BBB during increased immune cell transmigration in HI-SCZ, or potentially both.

Vasculature Changes May Impact Neurogenesis in HI-SCZ

Considering VEGFA expression is highest in cells with the capacity to proliferate in the SEZ, such as NSCs, neuroblasts and oligodendrocyte progenitor cells, but its receptor is specific to endothelial cells, this could suggest dividing cells may signal for increased angiogenesis, especially in states of high inflammation. This is plausible given the reliance of cell genesis on trophic support from blood vessels.⁴⁹ In our study, VEGFA is also associated with reduced immature neuron marker expression, suggesting that the benefits of enhanced vascular support are negated by a putative deleterious consequences of increased angiogenesis, which could result in elevated inflammation and immune cell infiltration. This aligns with our previous SEZ bulk RNA sequencing study which showed correlative associations between increased angiogenesis markers with reduced neuroblast markers, neuronatin, and fibroblast growth factor receptor 3.¹⁹ These changes are regionally specific, because in the midbrain there were no significant changes in VEGFA mRNA in schizophrenia, yet other angiogenesis factors like SERPINE1 were elevated.¹ Additionally, it seems that VEGFA mRNA is decreased in the DLPFC,⁵⁰ but VEGF protein is increased in serum in schizophrenia.²⁹

While in the midbrain angiopoietin-2 is expressed on GFAP+ astrocytes¹ in the SEZ it is most highly expressed in immature neurons, suggesting regional specificity and a propensity for neuroplastic neurons to promote angiogenesis in the neurogenic niche. ANGPT1 and ANGPT2 mRNAs were both decreased in schizophrenia and more specifically in low inflammation schizophrenia. Considering angiopoietin-1 reduces permeability and stabilizes blood vessels whereas angiopoietin-2 destabilizes blood vessels to promote angiogenesis,^26,27 the functional implications would be that the reduced expression of both ANGPT1 and ANGPT2 could balance each other out and/or lead to reduced plasticity of blood vessels.

We also found increased CLDN5 was associated with reduced immature neuron markers. There is no clear direct relationship between CLDN5 and neurogenesis in the literature, although CLDN5 is reduced in proliferative glioblastoma environments, suggesting a potential link to cell proliferation.⁵¹ Others studies show that a leakier SEZ BBB inhibits neuronal differentiation in favor of astrogenesis,⁵² which would suggest that a higher amount of claudin-5 would promote neurogenesis, contrary to our findings. Yet, if upregulated CLDN5 is related to higher immune cell infiltration, the correlation between higher CLDN5 and reduced immature neuron markers may reflect the detrimental role of elevated inflammation and macrophages on neurogenesis.^53-55 Given the unique SEZ vasculature supporting neurogenesis, and changes in the vasculature, particularly in the neuroinflammatory schizophrenia subgroup^1,19 with reduced neurogenesis markers,^7,14 further research is needed to elucidate whether there is a causal role of the vasculature pathology in reducing neurogenesis, or whether neurogenesis and vasculature changes are both downstream of inflammation.

Limitations and Future Directions

The use herein of post-mortem human brain tissue offers a unique opportunity to directly examine molecular and cellular changes in the human neurogenic niche of individuals with schizophrenia, yielding insights not achievable in animal models or in vitro systems. Our well-powered cohort revealed subgroup-specific alterations, such as those linked to high inflammation, often obscured in heterogeneous schizophrenia populations, underscoring the value of post-mortem research in uncovering region-specific pathologies in complex disorders. Nonetheless, post-mortem studies have limitations. In our cohort, there were too few controls with high inflammation (~10%) to enable statistical comparison. Thus, the inflammation related changes may not be unique to schizophrenia, but they are important to elucidate due to ~40% of schizophrenia cases having elevated inflammation. Another potential confound is the use antipsychotic treatments. Although prior studies suggest antipsychotics can increase claudin-5 expression,² we found no significant correlation between CLDN5 mRNA or protein levels and lifetime antipsychotic dose, indicating that CLDN5 upregulation in HI-SCZ likely reflects an adaptive response to BBB disruption and inflammation rather than being solely a medication effect. However, we do not have data on medication dose at the time of death, which may exhibit a stronger relationship with gene expression if there is an impact of medication on CLDN5.

Lower, more acidic, pH is consistently reported in schizophrenia post-mortem brain tissue³⁹ and in our cohort brain pH is lowest in the high inflammation group. This is not surprising as brain acidity is often associated with inflammation in ischemia, Alzheimer’s, and aging,⁵⁶ as well as being associated with mitochondrial impairment.⁵⁷ Immune cells have acid chemosensory receptors which differentially impact innate and adaptive immune cells; for example, a more acidic environment triggers release of pro-inflammatory cytokines from monocytes/macrophages,^58,59 but suppresses T-cell and natural killer cell activity.⁶⁰ This response to lower pH aligns with our findings of more pro-inflammatory macrophages in schizophrenia, which relates to lower pH.^7,14 In brain tissue, a major acid sensing receptor, GPR4, is specifically expressed in endothelial cells⁶¹ and rodent studies suggest that angiogenesis is enhanced in response to moderately acidic pH.⁶² While the relationships between brain acidosis, inflammation, and angiogenesis are still being elucidated,⁶³ our findings of higher CLDN5, PECAM1, VEGFR1, TEK, and estimated proportion of vascular cells correlating with brain pH suggest that angiogenesis could be upregulated as a restorative process in the more acidic brains in high-inflammation schizophrenia, as is the case in cerebral ischemia.⁶⁴

With gene expression analysis by qPCR from homogenate tissue, we cannot distinguish between increased expression per cell or increased abundance of cells that express the target gene. Therefore, future research should employ single-nucleus RNA sequencing in disease states in the SEZ to resolve cell type-specific transcriptional changes and examine the cellular communication pathways that could drive changes in neurogenesis, angiogenesis, and inflammation in the SEZ in schizophrenia. While we find changes in tight junction and angiogenesis related genes, future studies are needed to directly assess BBB integrity, for example assessing the leakiness of the vasculature by measuring plasmalemma vesicle-associated protein and/or by determining how serum proteins (albumin, immunoglobulins or fibrinogen) may be distributed within brain tissue. Longitudinal studies in humans using blood biomarkers indicative of brain events and animal models that can control variables and isolate their direct impact on distinct brain cells are also needed to explore the temporal relationships between vascular changes, inflammation, and neurogenesis, alongside determining the impact of clinical factors like antipsychotic treatments and systemic inflammation.

Conclusion

This study provides evidence of significant vascular and BBB-related changes in the SEZ of individuals with schizophrenia, particularly within a subgroup of patients with a high-inflammation biotype. These findings support the hypothesis that inflammation-driven vascular changes and elevated angiogenesis play a critical role in SEZ-specific neuropathology, likely influencing neurogenesis and immune cell infiltration. The observed upregulation of claudin-5 in high inflammation schizophrenia may reflect either more blood vessels/endothelial cells, or the tight junction remodeling that occurs during immune cell transmigration. It is unlikely to be a direct effect of antipsychotic medication, emphasizing the dynamic interplay between inflammation, angiogenesis, and BBB integrity.

These results highlight the importance of subgroup-specific and region-focused analyses in understanding schizophrenia's complexity, as alterations in the SEZ differ from those observed in other brain regions and when looking at schizophrenia as a single group. While our findings provide valuable insights, they also raise important questions about the causal relationships between inflammation, vascular dysfunction, and neurogenesis. By improving our understanding of SEZ-specific pathology, this research paves the way for targeted therapeutic strategies aimed at mitigating inflammation, preserving BBB integrity, and supporting neurogenesis in schizophrenia.

Author Contributions

H.F.N.: formal analysis, supervision, visualization, methodology, writing—original draft, writing—review & editing; J.L.: formal analysis, methodology, visualization; J.M.F., funding acquisition, supervision, writing—review & editing; M.J.W.: resources, methodology, writing—review & editing; and C.S.W.: conceptualization, funding acquisition, supervision, writing—review & editing

Funding

This work was funded by the NIH RO1 grant “Neuroinflammation and Neurogenesis in Schizophrenia” (PI Dr. Cynthia Weickert; 1R01MH126108; USA), the National Health and Medical Research Council (NHMRC) (Investigator Grant to CSW (APP# 2009237; Australia), and the NSW Health Schizophrenia Research Grant (CI CSW #RG220489; Australia). J.M.F. was supported by NHMRC Medical Research Futures Fund (GNT1200428).

Conflicts of Interest

The Authors have declared that there are no conflicts of interest in relation to the subject of this study.

Data Availability

The data from this study are available from the corresponding author upon request.