Posttraumatic Stress Disorder (PTSD) NAD/Sirtuin Deficiency and SARM1-Mediated Synaptic Vulnerability: Evidence for Accelerated Brain Aging Subtypes Authors: Ngo Cheung Published: 2026 ABSTRACT Anxiety disorders and posttraumatic stress disorder (PTSD) often occur alongside signs of accelerated biological aging, yet the molecular pathways that connect genetic liability to synaptic and circuit-level dysfunction remain unclear. We performed a multi-gene-set transcriptome-wide association study using large-scale genome-wide association study (GWAS) summary statistics for anxiety disorders and PTSD, with a focus on curated aging-related pathways across brain regions involved in fear, reward, memory, and stress regulation.The analysis identified 185 significant enrichments, with strong cross-disorder concordance in senescence, telomere maintenance, and mitochondrial modules. PTSD showed a distinctive negative signature in nicotinamide adenine dinucleotide (NAD) metabolism and sirtuin-related pathways, most prominently driven by sirtuin 3 (SIRT3), together with evidence of sterile alpha and Toll/interleukin-1 receptor motif-containing 1 (SARM1)-linked axonal and synaptic vulnerability. Anxiety disorders showed stronger enrichment in mitochondrial intrinsic apoptosis and inflammatory-redox signaling, alongside glutamatergic and astrocytic plasticity changes. Both conditions shared senescence and deoxyribonucleic acid (DNA)-damage programs, with NIMA-related kinase 4 (NEK4) emerging as a common driver. These pathways converged on complement-mediated pruning and presynaptic remodeling, suggesting a mechanistic bridge from cellular aging to fear and reward circuit dysfunction.Together, the findings support an association-based model in which genetic liability for anxiety disorders and PTSD converges on stress-aging biology at the synapse, while diverging by disorder. This framework points to candidate biologically distinct subtypes and offers practical opportunities for biomarker development, patient stratification, and subtype-guided therapeutic strategies. These subtype labels are proposed biological hypotheses rather than validated clinical categories. Introduction Anxiety disorders and posttraumatic stress disorder (PTSD) are common and disabling conditions that affect large numbers of people worldwide. Their burden is not limited to fear, worry, avoidance, intrusive memories, or hyperarousal. Both conditions are also associated with medical comorbidities that are typically linked to aging, including cardiovascular disease, metabolic dysfunction, cognitive decline, and chronic inflammation [1-3]. A growing body of work also suggests that people with anxiety disorders or PTSD show markers of accelerated biological aging, including shorter leukocyte telomere length, epigenetic age acceleration, and higher allostatic load [4-7]. These findings suggest that the biology of these disorders may extend beyond acute fear and stress responses and involve deeper disturbances in cellular aging pathways. Psychiatric genetics has made important progress in identifying risk loci and broad biological signals for anxiety disorders and PTSD through genome-wide association studies (GWAS) [8-10]. Still, much of the field remains focused on individual loci or general inflammatory pathways. These approaches are useful, but they often leave a gap between genetic association and mechanism. In particular, they do not fully explain how genetic risk might affect synaptic function, circuit plasticity, and cellular aging programs in brain tissue. This gap matters because anxiety disorders and PTSD share clinical and neurobiological features, yet they also differ in trauma-related onset, symptom structure, and treatment response [11,12]. Transcriptome-wide association studies (TWAS) provide one way to bridge this gap. TWAS methods use expression quantitative trait loci (eQTLs) to estimate genetically predicted gene expression in disease-relevant tissues and then test whether that predicted expression is associated with disease risk [13-16]. When applied to brain regions involved in fear, reward, memory, and executive control--such as the caudate, nucleus accumbens, hippocampus, amygdala, frontal cortex, and anterior cingulate cortex--this approach can highlight functional pathways that may be altered by genetic liability. TWAS is especially useful for studying aging biology because major hallmarks of aging, including mitochondrial dysfunction, cellular senescence, altered nicotinamide adenine dinucleotide (NAD) metabolism, and inflammatory signaling, leave measurable signatures in gene-expression programs [17,18]. Here, “accelerated brain aging subtypes” refers to biologically defined profiles in which genetically predicted expression patterns resemble aging-related programs in brain tissue, such as NAD/sirtuin deficiency, senescence/telomere stress, mitochondrial-apoptotic vulnerability, or complement-pruning activation. The term is used as a mechanistic stratification hypothesis, not as a clinical diagnosis. We hypothesized that genetic liability for anxiety disorders and PTSD would converge on stress-aging biology. Specifically, we expected enrichment of pathways involving mitochondrial and NAD-related stress, deoxyribonucleic acid (DNA) damage, senescence, apoptosis, inflammatory signaling, and downstream effects on synaptic pruning and plasticity. At the same time, we expected the two disorders to diverge. PTSD risk was expected to show stronger signatures of NAD/sirtuin suppression and axonal or synaptic vulnerability, whereas anxiety risk was expected to emphasize mitochondrial apoptosis thresholds and inflammatory-glutamatergic plasticity. This model could help explain why both conditions show accelerated aging phenotypes while still producing different clinical patterns. It also links molecular findings to known circuit-level mechanisms involving fear generalization, extinction learning, reward processing, and threat regulation [19-22]. To test this model, we performed a multi-gene-set TWAS and enrichment analysis using summary statistics from two large European-ancestry GWAS: one for major anxiety disorders and one for PTSD. The analysis focused on curated aging-related gene sets. The goals were to identify enriched aging pathways and their main gene drivers in brain tissue, develop a mechanistic model linking these pathways to synaptic and circuit dysfunction, and derive translational implications for biomarkers, patient stratification, and therapeutic hypotheses. The primary objective was descriptive and comparative: to test whether curated aging-related pathways were enriched in anxiety disorders and PTSD and whether the two disorders showed shared or divergent profiles. Downstream mechanistic and translational interpretations were treated as hypotheses generated by these enrichment patterns. This work builds on several related areas. The hallmarks of the aging framework describe interconnected processes such as mitochondrial dysfunction, cellular senescence, and chronic inflammation that contribute to organismal decline [17,18]. Programmed axon degeneration, mediated in part by the NAD-consuming enzyme sterile alpha and Toll/interleukin-1 receptor motif-containing 1 (SARM1), provides a plausible link between metabolic stress and synaptic disconnection [23-25]. Complement-mediated synaptic pruning, first described as a developmental process, can become maladaptive in adult brains under inflammatory or stress-related conditions [26-28]. Finally, the neurocircuitry of fear conditioning and extinction, centered on the amygdala, hippocampus, and medial prefrontal cortex, provides a framework for understanding how molecular changes may translate into persistent threat memory, impaired safety learning, and avoidance behavior [20-22]. Materials and methods Genome-wide association study data sources We used publicly available summary statistics from two large-scale GWAS. The anxiety disorders analysis was based on a study by Strom et al. [10], which included 122,341 European-ancestry cases with major anxiety disorders and matched controls. The PTSD analysis used data from a study by Nievergelt et al. [9], comprising European-ancestry participants with PTSD diagnoses. Both datasets had undergone standard quality control, imputation, and association testing procedures as described in the original publications. TWAS and gene-set enrichment pipeline Genetically predicted gene expression was imputed using S-PrediXcan from the MetaXcan framework [13,14], with brain tissue-specific eQTL weights from Genotype-Tissue Expression Project version 8 (GTEx v8) [15]. We focused on the caudate basal ganglia, nucleus accumbens basal ganglia, hippocampus, amygdala, frontal cortex Brodmann area 9 (BA9), and anterior cingulate cortex Brodmann area 24 (BA24). These regions were selected because of their established roles in fear processing, reward, memory, and executive regulation. Multi-gene-set enrichment analyses were conducted using curated aging-related gene sets from Reactome, Kyoto Encyclopedia of Genes and Genomes (KEGG), Hallmark, Gene Ontology, and WikiPathways as represented in Molecular Signatures Database-derived collections [16]. The collections included cellular senescence, telomere maintenance, DNA damage response, intrinsic and extrinsic apoptosis, NAD biosynthesis and metabolism, sirtuin-related processes, mitochondrial function and oxidative phosphorylation, mitophagy, complement and innate immune activation, microglial activation, synaptic plasticity regulation, neurotransmitter release, glutamatergic and dopaminergic signaling, and insulin/glucagon-like peptide-1 (GLP-1)-related metabolic pathways. Gene identifiers were harmonized before gene-set assignment, duplicate genes within a set were removed, and only genes with available TWAS statistics in the relevant tissue were retained. The archived analysis output used the multi-gene-set mode and was generated on May 13, 2026. Statistical thresholds and filtering Enrichment was evaluated using Stouffer’s Z-score combination across genes within each set. We applied a nominal significance threshold of raw p < 0.05 and absolute Stouffer Z > 1.96. Additional filters included false discovery rate (FDR) correction where available, permutation-based p-values, Wilcoxon rank-sum tests, absolute Cohen’s d > 0.5, sign-concordance rate > 0.7, Kendall’s tau > 0.3, Pearson r > 0.3, Lin’s concordance correlation coefficient (CCC) > 0.5, and rank-biserial r > 0.3. Leave-one-out influence analysis was used to identify genes that disproportionately drove set-level signals. Results were retained only when they met at least two independent statistical criteria, such as a significant permutation p-value and the Stouffer Z threshold, to reduce false positives. This multi-criterion thresholding scheme was used because the analysis combined small, medium, and large gene sets, and no single test was expected to capture all robust pathway patterns. Stouffer Z and nominal p-values screened for pathway-level signal; permutation and Wilcoxon tests assessed robustness to gene-level distributional assumptions; effect-size and concordance filters emphasized consistent direction or meaningful magnitude; and FDR summarized multiple-testing control where available. Overlapping or correlated pathways were not pruned before testing because several aging pathways are biologically nested. Therefore, related pathway hits were interpreted as families of convergent signals rather than as statistically independent discoveries. No pathway family survived the most stringent family-wise FDR summary, so the results are presented as prioritized enrichment signals requiring replication rather than definitive family-wise discoveries. Results Overview of significant enrichments The multi-gene-set enrichment analysis identified 185 findings that met the pre-specified statistical criteria across the curated aging-related gene sets. Because no pathway family survived the most stringent family-wise FDR screen across the full analysis, the following results should be read as convergent, prioritized enrichment signals rather than final proof of pathway activation. Several core biological modules showed high concordance between anxiety disorders and PTSD (Table 1). Pearson correlations of gene-level Z-scores between the two disorder profiles reached r = 0.60 for cellular senescence gene sets and r = 0.70 for telomere maintenance sets. Lin’s concordance correlation coefficients exceeded 0.69 in both modules. Mitochondrial and oxidative phosphorylation modules showed more moderate but still significant similarity, with correlations of about 0.45 to 0.50. Table 1Cross-Disorder Concordance in Core Aging-Related Modules (ptsd vs. anx_cc)Note: r = Pearson correlation coefficient; ρ = Spearman rank correlation coefficient; CCC = Lin's concordance correlation coefficient; τ = Kendall's tau; Sign = sign-concordance rate. All reported Pearson and Spearman values were significant at p < .05.PTSD = posttraumatic stress disorder; anx_cc = anxiety disorder case-control phenotype; HLA = human leukocyte antigen; GOBP = Gene Ontology Biological Process; REACTOME = Reactome pathway database; HALLMARK = Molecular Signatures Database Hallmark gene set; MOOTHA = Mootha mitochondrial gene set; n = number of genes.Module / Gene SetnrρCCCτSignHLA complex16.89.87.87.67.88GOBP regulation of cellular senescence35.76.63.73.47.74REACTOME telomere maintenance54.70.60.70.43.64GOBP cellular senescence75.60.54.57.39.65GOBP regulation of synaptic plasticity151.60.51.59.37.71HALLMARK apoptosis131.51.52.51.37.68MOOTHA mitochondria353.50.40.50.28.65HALLMARK oxidative phosphorylation149.45.40---GOBP intrinsic apoptotic signaling264.47.44.47.31.64 Cross-Disorder Concordance in Core Aging-Related Modules (ptsd vs. anx_cc) Note: r = Pearson correlation coefficient; ρ = Spearman rank correlation coefficient; CCC = Lin's concordance correlation coefficient; τ = Kendall's tau; Sign = sign-concordance rate. All reported Pearson and Spearman values were significant at p < .05. PTSD = posttraumatic stress disorder; anx_cc = anxiety disorder case-control phenotype; HLA = human leukocyte antigen; GOBP = Gene Ontology Biological Process; REACTOME = Reactome pathway database; HALLMARK = Molecular Signatures Database Hallmark gene set; MOOTHA = Mootha mitochondrial gene set; n = number of genes. Many of the strongest signals are localized to brain regions with established roles in emotion regulation and reward processing. Enrichment in the caudate and nucleus accumbens basal ganglia was especially consistent for apoptosis, senescence, and synaptic plasticity sets. Hippocampal and amygdala signals were prominent for DNA-damage response and inflammatory pathways. These regional patterns are consistent with the involvement of striatal and limbic circuits in both anxiety disorders and PTSD. PTSD-specific negative NAD/sirtuin/SIRT3 signature A clear divergence between PTSD and anxiety disorders emerged in NAD metabolism and sirtuin-related gene sets. PTSD showed consistent negative enrichment across several NAD biosynthetic and sirtuin-associated collections (Table 2). The WP NAD biosynthetic pathways "ALL" set yielded a Stouffer Z of -5.88, with permutation p = 0.015. Tissue-specific values were -2.96 in the anterior cingulate cortex, -2.95 in the amygdala, and -3.42 in the hippocampus. By contrast, anxiety disorders showed positive enrichment in overlapping nicotinate and nicotinamide metabolism sets, with an overall Z of +4.19. Table 2PTSD-Specific NAD/Sirtuin Signature and Leading SIRT3 DriverNote. Z = Stouffer's combined Z-score; p = permutation or Wilcoxon p-value, as reported; % = leave-one-out influence on pooled pathway Z. PTSD = posttraumatic stress disorder; anx_cc = anxiety disorder case-control phenotype; WP = WikiPathways; NAD = nicotinamide adenine dinucleotide; NAD⁺ = oxidized nicotinamide adenine dinucleotide; SIRT1 = sirtuin 1; SIRT3 = sirtuin 3; ACC = anterior cingulate cortex; BA24 = Brodmann area 24; KEGG = Kyoto Encyclopedia of Genes and Genomes; GOMF = Gene Ontology Molecular Function; NAPRT = nicotinate phosphoribosyltransferase; NMNAT2 = nicotinamide nucleotide adenylyltransferase 2; n = number of genes.Gene SetTraitTissuenZpWP NAD biosynthetic pathwaysptsdPooled19-5.88.015WP NAD biosynthetic pathwaysptsdHippocampus13-3.42.048WP NAD biosynthetic pathwaysptsdACC (BA24)17-2.96.020WP NAD biosynthetic pathwaysptsdAmygdala14-2.95.025WANG adipogenic genes repressed by SIRT1ptsdPooled18-4.98.041KEGG nicotinate and nicotinamide metabolismanx_ccPooled22+4.19.030Leading leave-one-out drivers (PTSD)GOMF NAD⁺ binding--SIRT3ptsd---7.5872.9%WP NAD metabolism/sirtuins/aging--SIRT3ptsd---7.58109.0%REACTOME nicotinate metabolism--NAPRTptsd---3.74-KEGG nicotinate/nicotinamide--NMNAT2ptsd--+2.62- PTSD-Specific NAD/Sirtuin Signature and Leading SIRT3 Driver Note. Z = Stouffer's combined Z-score; p = permutation or Wilcoxon p-value, as reported; % = leave-one-out influence on pooled pathway Z. PTSD = posttraumatic stress disorder; anx_cc = anxiety disorder case-control phenotype; WP = WikiPathways; NAD = nicotinamide adenine dinucleotide; NAD⁺ = oxidized nicotinamide adenine dinucleotide; SIRT1 = sirtuin 1; SIRT3 = sirtuin 3; ACC = anterior cingulate cortex; BA24 = Brodmann area 24; KEGG = Kyoto Encyclopedia of Genes and Genomes; GOMF = Gene Ontology Molecular Function; NAPRT = nicotinate phosphoribosyltransferase; NMNAT2 = nicotinamide nucleotide adenylyltransferase 2; n = number of genes. Sirtuin 3 (SIRT3), a mitochondrial deacetylase, was the most influential driver in these PTSD-associated sets. Across GOMF NAD⁺ binding, WP NAD biosynthetic pathways, and WP NAD metabolism/sirtuins/aging collections (SIRT3) showed Z = -7.58 and accounted for up to 72.9% of the pathway signal in leave-one-out analyses. Other notable drivers included nicotinamide nucleotide adenylyltransferase 2 (NMNAT2), with Z = +2.62 in PTSD; nicotinate phosphoribosyltransferase (NAPRT), with Z = -3.74 in PTSD compared with +2.30 in anxiety; and several poly(adenosine diphosphate-ribose) polymerase (PARP) family members in related NAD-binding collections. Differential testing confirmed that the WANG adipogenic genes repressed by SIRT1 ALL set were significantly more negative in PTSD than in anxiety, with a mean Z difference of -1.87, Cohen’s d = -0.74, and FDR-corrected p = 0.035. Overall, these findings are consistent with lower genetically predicted NAD/sirtuin-related pathway tone in PTSD, particularly in mitochondrial compartments, but they do not directly measure tissue NAD levels or sirtuin enzymatic activity. NAD-SARM1-synaptic/axon-vulnerability axis Integration across NAD metabolism, mitophagy, and pruning-related gene sets suggested a possible link between metabolic stress and axonal or synaptic vulnerability. SARM1, a key regulator of programmed axon degeneration, showed positive enrichment in both disorders within relevant mitophagy and pruning collections, with Z values around +3.10 to +3.15 (Table 3). This signal occurred alongside negative NAD/sirtuin findings in PTSD and positive mitochondrial and apoptosis signals in anxiety. Table 3Aging-to-Pruning/Plasticity Bridge: Complement, Microglia, Plasticity, and Release MachineryNote. Direction reflects the sign of Stouffer's Z-score: ↑ = positive enrichment; ↓ = negative enrichment. PTSD = posttraumatic stress disorder; anx_cc = anxiety disorder case-control phenotype; REACTOME = Reactome pathway database; GOBP = Gene Ontology Biological Process; ADULT_MICRO = adult microglial gene signature; C4A = complement C4A; C1QA = complement C1q A chain; C1QC = complement C1q C chain; TREM1 = triggering receptor expressed on myeloid cells 1; TREM2 = triggering receptor expressed on myeloid cells 2; SARM1 = sterile alpha and Toll/interleukin-1 receptor motif-containing 1; MAP1A = microtubule-associated protein 1A; STAU1 = staufen double-stranded ribonucleic acid binding protein 1; ARC = activity-regulated cytoskeleton-associated protein; DLG4 = discs large membrane-associated guanylate kinase scaffold protein 4; STX1A = syntaxin 1A; VAMP2 = vesicle-associated membrane protein 2; GRM3 = glutamate metabotropic receptor 3; GRIA1 = glutamate ionotropic receptor α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid type subunit 1.GeneModuleTraitZDirectionC4AREACTOME complementanx_cc+11.01↑C4AREACTOME complementptsd+12.33↑C1QAGOBP synapse pruningptsd-2.84↓C1QCGOBP synapse pruninganx_cc+2.47↑TREM1ADULT_MICROanx_cc+5.54↑TREM2GOBP synapse pruningptsd+3.72↑SARM1GOBP synapse pruninganx_cc / ptsd+3.10 / +3.15↑MAP1ARegulation of synaptic plasticityanx_cc / ptsd-7.33 / -10.51↓STAU1Regulation of synaptic plasticityanx_cc / ptsd-7.09 / -5.91↓ARCRegulation of synaptic plasticityanx_cc-4.64↓DLG4Regulation of synaptic plasticityptsd-3.96↓STX1AREACTOME neurotransmitter releaseptsd-7.01↓VAMP2REACTOME neurotransmitter releaseanx_cc / ptsd+6.98 / +3.57↑GRM3GOBP glutamatergicanx_cc+7.36↑GRIA1GOBP glutamatergicptsd+5.35↑ Aging-to-Pruning/Plasticity Bridge: Complement, Microglia, Plasticity, and Release Machinery Note. Direction reflects the sign of Stouffer's Z-score: ↑ = positive enrichment; ↓ = negative enrichment. PTSD = posttraumatic stress disorder; anx_cc = anxiety disorder case-control phenotype; REACTOME = Reactome pathway database; GOBP = Gene Ontology Biological Process; ADULT_MICRO = adult microglial gene signature; C4A = complement C4A; C1QA = complement C1q A chain; C1QC = complement C1q C chain; TREM1 = triggering receptor expressed on myeloid cells 1; TREM2 = triggering receptor expressed on myeloid cells 2; SARM1 = sterile alpha and Toll/interleukin-1 receptor motif-containing 1; MAP1A = microtubule-associated protein 1A; STAU1 = staufen double-stranded ribonucleic acid binding protein 1; ARC = activity-regulated cytoskeleton-associated protein; DLG4 = discs large membrane-associated guanylate kinase scaffold protein 4; STX1A = syntaxin 1A; VAMP2 = vesicle-associated membrane protein 2; GRM3 = glutamate metabotropic receptor 3; GRIA1 = glutamate ionotropic receptor α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid type subunit 1. Leave-one-out analyses indicated that SARM1 contributed substantially to mitophagy enrichment in both conditions. The overall pattern is compatible with a model in which NAD depletion or sirtuin suppression lowers the threshold for SARM1 activation, increasing susceptibility to axon or synapse disconnection. This remains an inferred pathway-level link and does not demonstrate SARM1 activation in patient brain tissue. This interpretation is consistent with studies showing that SARM1 activation promotes axon degeneration through NAD+ destruction and with broader work on programmed axon degeneration as a pathway linking injury and disease [23-25,29,30]. The axis was further supported by concurrent enrichment in mitochondrial dynamics genes, including mitofusin 2 (MFN2), optic atrophy 1 (OPA1), and ras homolog family member T1 (RHOT1), and synaptic vesicle release machinery, suggesting a coordinated effect on presynaptic integrity. Shared senescence/telomere/DNA-damage architecture Both disorders showed strong and highly concordant enrichment in senescence, telomere maintenance, and DNA-damage response gene sets (Table 4). Pearson correlations of Z-scores within these modules ranged from 0.60 to 0.70, with Lin’s concordance correlation coefficients between 0.69 and 0.73. The GOBP CELLULAR SENESCENCE ALL set yielded r = 0.599, p < 0.001, and rho = 0.537, p = 1 x 10^-6, between anxiety and PTSD profiles. Table 4Shared Senescence, Telomere, and DNA-Damage DriversNote. Δ = ΔStouffer, defined as the change in pathway-level Z-score with the gene removed; % = leave-one-out influence. PTSD = posttraumatic stress disorder; anx_cc = anxiety disorder case-control phenotype; GOBP = Gene Ontology Biological Process; REACTOME = Reactome pathway database; DNA = deoxyribonucleic acid; NEK4 = NIMA-related kinase 4; PPP6C = protein phosphatase 6 catalytic subunit; NBN = nibrin; ATM = ataxia telangiectasia mutated; POT1 = protection of telomeres 1; TERF1 = telomeric repeat binding factor 1; TERF2 = telomeric repeat binding factor 2; WRN = Werner syndrome RecQ-like helicase; RTEL1 = regulator of telomere elongation helicase 1.GeneTraitGene setZΔ%NEK4anx_ccGOBP cellular senescence+12.93+1.48100.5NEK4ptsdGOBP cellular senescence+7.8+0.8721.6NEK4anx_ccRegulation of cellular senescence+12.93+2.1597.8NEK4ptsdRegulation of cellular senescence+7.8+1.356PPP6Canx_ccREACTOME telomere+9.25+1.2341.5PPP6CptsdREACTOME telomere+11.46+1.5690.4NBNptsdREACTOME cellular senescence+7.43+0.74115.1ATMptsdREACTOME cellular senescence+6.3+0.6397.4POT1ptsdREACTOME telomere+7.00+0.9454.9TERF1ptsdREACTOME telomere+4.72+0.6336.8TERF2ptsdREACTOME cellular senescence-3.61-0.3656.6WRNptsdREACTOME telomere+3.5+0.4727RTEL1ptsdREACTOME telomere+3.44+0.4626.5 Shared Senescence, Telomere, and DNA-Damage Drivers Note. Δ = ΔStouffer, defined as the change in pathway-level Z-score with the gene removed; % = leave-one-out influence. PTSD = posttraumatic stress disorder; anx_cc = anxiety disorder case-control phenotype; GOBP = Gene Ontology Biological Process; REACTOME = Reactome pathway database; DNA = deoxyribonucleic acid; NEK4 = NIMA-related kinase 4; PPP6C = protein phosphatase 6 catalytic subunit; NBN = nibrin; ATM = ataxia telangiectasia mutated; POT1 = protection of telomeres 1; TERF1 = telomeric repeat binding factor 1; TERF2 = telomeric repeat binding factor 2; WRN = Werner syndrome RecQ-like helicase; RTEL1 = regulator of telomere elongation helicase 1. Prominent driver genes included ataxia telangiectasia mutated (ATM), with Z = +6.30 in PTSD senescence and apoptosis sets; nibrin (NBN), with Z = +7.43; protection of telomeres 1 (POT1), with Z = +7.00 in PTSD telomere and senescence sets; telomeric repeat binding factor 1 (TERF1), with Z = +4.72; telomeric repeat binding factor 2 (TERF2), with Z = -3.61; protein phosphatase 6 catalytic subunit (PPP6C), with Z = +9.25 in anxiety and +11.46 in PTSD; Werner syndrome RecQ-like helicase (WRN), with Z = +3.50; and regulator of telomere elongation helicase 1 (RTEL1), with Z = +3.44. Other influential genes included cyclin-dependent kinase inhibitor 1B (CDKN1B), cyclin-dependent kinase inhibitor 2A (CDKN2A), lamin B1 (LMNB1), enhancer of zeste homolog 2 (EZH2), and suppressor of zeste 12 (SUZ12). These findings indicate that genetic risk for both conditions intersects with DNA-damage checkpoint and telomere-protection programs, consistent with broader evidence that anxiety disorders and PTSD are associated with accelerated biological aging phenotypes [4,6,7,17,18,31]. NEK4 as a novel shared senescence/checkpoint driver NEK4 emerged as one of the strongest and most consistent gene-level signals across senescence-related sets. In the GOBP CELLULAR SENESCENCE ALL collection, NEK4 showed Z = +12.93 in anxiety and Z = +7.80 in PTSD. It represented the largest single-gene contribution to these enrichments, with 100.5% leave-one-out influence in anxiety and 21.6% in PTSD. NEK4 was also the top driver in the regulation-of-cellular-senescence set for both disorders. NEK4 is a kinase implicated in DNA-damage response and cell-cycle checkpoint regulation. Its dominant influence in both anxiety and PTSD senescence modules suggests a shared genetic mechanism that may sensitize cells to stress-induced checkpoint activation or senescence-like states in brain tissue. Because this inference is based on genetically predicted expression, functional studies are needed before NEK4 can be considered a confirmed causal driver of stress-related brain aging. Anxiety-enriched mitochondrial intrinsic apoptosis Anxiety disorders showed particularly strong enrichment in intrinsic apoptosis pathways. The REACTOME INTRINSIC PATHWAY FOR APOPTOSIS ALL set produced an overall Stouffer Z of +6.43, with permutation p = 0.0056 (Table 5). Tissue-specific values were +4.94 in the nucleus accumbens, +5.33 in the caudate, and +3.48 in the hippocampus. KEGG APOPTOSIS ALL and HALLMARK APOPTOSIS ALL showed parallel positive enrichment, with overall Z values between +3.63 and +3.88. Table 5Anxiety-Enriched Intrinsic Apoptosis and Key BCL-2 / Caspase DriversNote. For gene-level rows, the % column denotes leave-one-out influence on pathway-level Stouffer's Z-score. PTSD = posttraumatic stress disorder; anx_cc = anxiety disorder case-control phenotype; REACTOME = Reactome pathway database; KEGG = Kyoto Encyclopedia of Genes and Genomes; GOCC = Gene Ontology Cellular Component; BCL-2 = B-cell lymphoma 2; Nuc. Accumbens = nucleus accumbens; BAK1 = BCL2 antagonist/killer 1; BIK = BCL2 interacting killer; BCL2L1 = BCL2-like 1; CASP10 = caspase 10; CASP7 = caspase 7; GPX1 = glutathione peroxidase 1; n = number of genes; Z = Stouffer's combined Z-score; p = p-value.Gene set / driverTraitTissuenZpREACTOME intrinsic pathway for apoptosisanx_ccPooled47+6.43.006REACTOME intrinsic pathway for apoptosisanx_ccCaudate38+5.33.001REACTOME intrinsic pathway for apoptosisanx_ccNuc. Accumbens35+4.94< .001REACTOME intrinsic pathway for apoptosisanx_ccHippocampus27+3.48.01HALLMARK apoptosisanx_ccPooled136+3.88.036KEGG apoptosisanx_ccPooled69+3.63-Gene-level drivers BAK1--GOCC BCL-2 family complexanx_cc---3.9298.90%BIK--GOCC BCL-2 family complexanx_cc--+2.9765.40%BCL2L1--GOCC BCL-2 family complexanx_cc--+2.8261.90%CASP10--KEGG apoptosisanx_cc--+6.420.60%CASP7--KEGG apoptosisptsd--+4.8437.30%GPX1--HALLMARK apoptosisptsd--+10.7831.80% Anxiety-Enriched Intrinsic Apoptosis and Key BCL-2 / Caspase Drivers Note. For gene-level rows, the % column denotes leave-one-out influence on pathway-level Stouffer's Z-score. PTSD = posttraumatic stress disorder; anx_cc = anxiety disorder case-control phenotype; REACTOME = Reactome pathway database; KEGG = Kyoto Encyclopedia of Genes and Genomes; GOCC = Gene Ontology Cellular Component; BCL-2 = B-cell lymphoma 2; Nuc. Accumbens = nucleus accumbens; BAK1 = BCL2 antagonist/killer 1; BIK = BCL2 interacting killer; BCL2L1 = BCL2-like 1; CASP10 = caspase 10; CASP7 = caspase 7; GPX1 = glutathione peroxidase 1; n = number of genes; Z = Stouffer's combined Z-score; p = p-value. Key driver genes within B-cell lymphoma 2 (BCL-2) family and executioner caspase modules included BCL2 antagonist/killer 1 (BAK1), with Z = -3.92 and 98.9% leave-one-out influence; BCL2 interacting killer (BIK), with Z = +2.97; BCL2-like 1 (BCL2L1), with Z = +2.82; BCL2-associated agonist of cell death (BAD), with Z = +2.67; B-cell lymphoma 2 (BCL2), with Z = -1.18; caspase 7 (CASP7), with Z = +4.84 in PTSD but also influence in anxiety; and caspase 10 (CASP10), with Z = +6.40. Glutathione peroxidase 1 (GPX1), an antioxidant enzyme, was a major contributor in PTSD apoptosis sets, with Z = +10.78. Taken together, these results are consistent with altered mitochondrial apoptosis-threshold regulation in anxiety disorders, particularly in striatal and hippocampal circuits. Aging-to-pruning/plasticity bridge A bridge between aging biology and synaptic remodeling was evident across complement, microglial, structural plasticity, and presynaptic release gene sets. The complement set showed very large positive enrichment, with complement C4A (C4A) displaying Z = +11.01 in anxiety and Z = +12.33 in PTSD. Complement C1q A chain (C1QA) and complement C1q C chain (C1QC) were also influential, with C1QA showing Z = -2.84 in PTSD and C1QC showing Z = +2.47 in anxiety. Microglial state markers suggested a possible disorder-level divergence. Triggering receptor expressed on myeloid cells 2 (TREM2) was positive in PTSD pruning contexts, with Z = +3.72, whereas triggering receptor expressed on myeloid cells 1 (TREM1) was strongly positive in anxiety, with Z = +5.54. These findings are notable given the role of TREM-family signaling in microglial biology and neurodegenerative disease contexts [32,33]. Structural plasticity genes showed consistent negative signals, including microtubule-associated protein 1A (MAP1A) with Z = -7.33 in anxiety and -10.51 in PTSD, staufen double-stranded ribonucleic acid binding protein 1 (STAU1) with Z = -7.09 and -5.91, activity-regulated cytoskeleton-associated protein (ARC) with Z = -4.64 in anxiety, and discs large membrane-associated guanylate kinase scaffold protein 4 (DLG4) with Z = -3.96 in PTSD. Presynaptic vesicle machinery was also altered. Syntaxin 1A (STX1A) reached Z = -7.01 in PTSD, while vesicle-associated membrane protein 2 (VAMP2) showed Z = +6.98 in anxiety and +3.57 in PTSD. Glutamatergic genes included glutamate metabotropic receptor 3 (GRM3), with Z = +7.36 in anxiety, and glutamate ionotropic receptor alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid type subunit 1 (GRIA1), with Z = +5.35 in PTSD. These patterns suggest that aging-related stress may converge on both microglial pruning and activity-dependent synaptic remodeling, but with disorder-specific emphasis. Supporting modules Additional modules provided further context. Inflammatory and senescence-associated secretory phenotype-like sets, including HALLMARK INFLAMMATORY RESPONSE and HALLMARK TNFA SIGNALING VIA NFKB, showed positive enrichment, especially in anxiety, with Z values up to +4.69. Nitric oxide synthase 2 (NOS2) was a recurrent driver, with Z = +8.24 in anxiety and +7.71 in PTSD. Insulin secretion and GLP-1-related sets showed mixed but significant signals, including strong negative enrichment in PTSD for regulation of insulin secretion, with an overall Z of -6.20. Circadian clock genes showed modest positive enrichment in PTSD amygdala and nucleus accumbens, with Z values between +2.34 and +3.05. An oligodendrocyte/myelin signature was also detected, with ADULT OLIGO ALL showing Z = +5.30 in PTSD, driven largely by glycerate kinase (GLYCTK) and pleckstrin homology domain-containing B1 (PLEKHB1). Cross-disorder contrasts and gene-level drivers Differential testing identified several sets with statistically significant differences in magnitude between disorders (Table 6). The WANG ADIPOGENIC GENES REPRESSED BY SIRT1 ALL set was more negative in PTSD, with a mean Z difference of -1.869, Mann-Whitney U p = 0.035, and paired Cohen’s d = -0.74. HALLMARK TNFA SIGNALING VIA NFKB and REACTOME CYTOKINE SIGNALING IN IMMUNE SYSTEM showed greater effect sizes in anxiety. Table 6Cross-Disorder Differential Contrasts and Supporting ModulesNote. ΔZ = mean Stouffer Z-score difference, calculated as PTSD - anx_cc; dpaired = paired Cohen's d; MWU = Mann–Whitney U test; Paired W = paired Wilcoxon test; Perm = sign-flip permutation test. Dashes indicate metrics that were not computed or did not reach the reporting threshold. PTSD = posttraumatic stress disorder; anx_cc = anxiety disorder case-control phenotype; FDR = false discovery rate; SIRT1 = sirtuin 1; TNFA = tumor necrosis factor alpha; NFKB = nuclear factor kappa B; KEGG = Kyoto Encyclopedia of Genes and Genomes; GOBP = Gene Ontology Biological Process; ADULT_OLIGO = adult oligodendrocyte gene signature; GLYCTK = glycerate kinase; PLEKHB1 = pleckstrin homology domain-containing B1.Gene setΔZdpairedMWU p (FDR)Paired W pPerm pWANG adipogenic genes repressed by SIRT1 (ptsd < anx_cc)-1.87-0.74.035.008.002HALLMARK TNFA signaling via NFKB (anx_cc > ptsd)-0.80-0.51.117.044.033REACTOME cytokine signaling in immune system (anx_cc > ptsd)-0.75-0.51---KEGG nicotinate and nicotinamide metabolism (anx_cc > ptsd)-1.12-0.49--.032GOBP cellular senescence (ptsd > anx_cc)+0.63+0.27-.024.026Supporting pathway-level signals HALLMARK inflammatory response -- anx_cc (pooled)---.052.048 HALLMARK TNFA signaling via NFKB -- anx_cc (pooled, Z = +4.62)---.044.047 REACTOME regulation of insulin secretion -- ptsd (pooled, Z = -6.20)---.079.012 REACTOME circadian clock -- ptsd amygdala (Z = +3.05)---.031- ADULT_OLIGO -- ptsd (pooled, Z = +5.30; GLYCTK, PLEKHB1 drivers)---.156.03 Cross-Disorder Differential Contrasts and Supporting Modules Note. ΔZ = mean Stouffer Z-score difference, calculated as PTSD - anx_cc; dpaired = paired Cohen's d; MWU = Mann–Whitney U test; Paired W = paired Wilcoxon test; Perm = sign-flip permutation test. Dashes indicate metrics that were not computed or did not reach the reporting threshold. PTSD = posttraumatic stress disorder; anx_cc = anxiety disorder case-control phenotype; FDR = false discovery rate; SIRT1 = sirtuin 1; TNFA = tumor necrosis factor alpha; NFKB = nuclear factor kappa B; KEGG = Kyoto Encyclopedia of Genes and Genomes; GOBP = Gene Ontology Biological Process; ADULT_OLIGO = adult oligodendrocyte gene signature; GLYCTK = glycerate kinase; PLEKHB1 = pleckstrin homology domain-containing B1. Across all significant sets, leave-one-out analyses identified 1,113 influential gene-disease entries. The most recurrent high-impact drivers were SIRT3 in PTSD NAD sets, NEK4 in senescence sets for both disorders, C4A in complement sets, STX1A and VAMP2 in presynaptic release sets, GRM3 in anxiety glutamatergic sets, and BAK1 and BIK in apoptosis sets. These gene-level results reinforce the pathway findings and identify specific candidates for follow-up functional studies. Discussion This TWAS-based enrichment analysis suggests that genetic liability for anxiety disorders and PTSD intersects with core programs of cellular aging. Across 185 significant findings, we observed enrichment in pathways related to mitochondrial function, NAD metabolism, senescence, telomere maintenance, apoptosis, inflammation, and synaptic remodeling. Some modules showed substantial concordance between the two disorders, while others differed in direction or magnitude. These patterns may help explain why anxiety disorders and PTSD share features of accelerated biological aging but differ in clinical expression and neurocircuit involvement. The findings should be interpreted as genetically informed pathway associations, not as direct evidence of altered transcript, protein, metabolite, or cellular pathway activity in affected individuals. A unifying model emerges from the data. Genetic risk appears to converge on a stress-aging cascade in which mitochondrial and NAD-related perturbations contribute to DNA damage and senescence checkpoint activation. These changes may then engage apoptosis and inflammatory signaling, including SASP-like responses. Downstream, these upstream cellular stress programs appear to feed into complement- and microglia-mediated synaptic pruning, structural plasticity changes, and presynaptic remodeling. This model is consistent with the broader hallmarks of the aging framework and with evidence that stress-related disorders involve bioenergetic, inflammatory, and plasticity-related disturbances [5,17,18,34,35]. Within this shared framework, PTSD showed a route marked by SIRT3 suppression and broader reduction in NAD/sirtuin tone. This may lower the threshold for SARM1-mediated axonal or synaptic disconnection. The pattern is supported by negative enrichment in NAD biosynthetic and sirtuin-related gene sets, as well as positive SARM1 signals across mitophagy and pruning collections. Work on SARM1 has shown that its activation can trigger local axon degeneration through NAD+ destruction, and reviews of programmed axon degeneration have emphasized its relevance to injury and disease [23-25,29,30]. However, the present data do not establish that SIRT3 suppression causes NAD depletion or that SARM1 is enzymatically activated in PTSD; this axis should be viewed as a testable mechanistic hypothesis. In anxiety disorders, the dominant pattern involved altered mitochondrial apoptosis thresholds, especially in striatal and hippocampal regions, together with inflammatory-redox activation involving NOS2 and tumor necrosis factor (TNF) signaling and glutamatergic/astrocytic plasticity changes involving GRM3. The most clearly shared component was the senescence, telomere, and DNA-damage module, where NEK4, ATM, NBN, POT1, TERF1, TERF2, and PPP6C emerged as recurrent drivers, and cross-disorder correlations exceeded r = 0.60. These molecular findings map onto known neurocircuit abnormalities. Fear generalization and impaired safety learning have been linked to dysregulation in amygdala-hippocampal and amygdala-prefrontal circuits [19,20]. The pruning and plasticity signals found here, particularly those involving C4A, TREM2, TREM1, ARC, DLG4, and presynaptic vesicle machinery, could contribute to excessive or poorly timed synapse elimination in these circuits. Complement and microglial pathways are well established in developmental synapse elimination, and their reactivation or dysregulation in adult brains may contribute to disease-related circuit remodeling [26-28]. The presynaptic remodeling findings involving STX1A, VAMP2, and synaptotagmin 1 (SYT1) are also relevant because precise neurotransmitter release timing is important for extinction learning and fear memory updating [21]. The regional results also fit with known clinical patterns. Striatal involvement, especially in the caudate and nucleus accumbens, is consistent with reward-processing changes and avoidance biases in both disorders. These may be expressed differently across diagnoses, with avoidance more prominent in anxiety disorders and intrusive memory, hyperarousal, and trauma-related reactivity more characteristic of PTSD [11,12]. The mitochondrial and NAD-related signals also align with reports of bioenergetic dysfunction and oxidative stress in stress-related conditions [35,36]. Energetic insufficiency at synapses may make them more vulnerable to functional disconnection or elimination, particularly under chronic stress. From a translational standpoint, the findings support the development of biologically informed subtypes rather than treating anxiety disorders and PTSD as single, uniform categories. At this stage, these subtype labels are intended as research constructs for biomarker testing and trial enrichment, not as validated diagnostic categories. A mitochondrial-resilience PTSD subtype could be defined by low NAD/sirtuin tone, elevated oxidative stress markers, and evidence of axonal or synaptic injury, such as neurofilament light chain. An axon-vulnerability PTSD subtype might be further characterized by combined NAD/SARM1 signatures and diffusion magnetic resonance imaging (MRI) or connectivity abnormalities. For anxiety disorders, a mitochondrial-apoptotic subtype could be marked by elevated apoptosis-threshold gene signals together with redox and inflammatory markers. An inflammatory-redox subtype might be identified by prominent NOS2/TNF and SASP-like profiles. A broader accelerated-aging stress disorder subtype, spanning both diagnoses, could be defined by epigenetic age acceleration, telomere length, and senescence-associated inflammatory markers. Such stratification would allow future trials to enrich for biologically relevant subgroups and better match patients to existing or emerging interventions. Several biomarker panels follow directly from the top modules. A PTSD mitochondrial-resilience panel could combine NAD metabolomics, including oxidized/reduced nicotinamide adenine dinucleotide (NAD+/NADH) ratio, nicotinamide (NAM), and nicotinamide mononucleotide (NMN), with SIRT3-informed polygenic or predicted-expression scores, redox markers such as reduced/oxidized glutathione (GSH/GSSG) and 8-hydroxy-2′-deoxyguanosine (8-OHdG), and synaptic injury indicators such as neurofilament light. An aging panel applicable to both disorders could include GrimAge or DunedinPACE epigenetic clocks, leukocyte telomere length, p16/cyclin-dependent kinase inhibitor 2A (p16/CDKN2A) expression, and SASP cytokines. Pruning and plasticity signatures could be assessed through complement components such as C4A, microglial state markers such as the TREM2/TREM1 ratio, and synaptic proteins measured in neuron-derived extracellular vesicles. These panels could be tested against symptom clusters, treatment resistance, cognitive impairment, and longitudinal outcomes in existing cohorts. The same model suggests several therapeutic hypotheses. Biomarker-guided aerobic exercise and sleep stabilization are immediately testable, low-risk interventions that engage mitochondrial biogenesis, NAD metabolism, and plasticity processes. Metabolic optimization, including GLP-1-related pathways where relevant, may be useful in subgroups showing insulin secretion or glucose-homeostasis signals. NAD-precursor or sirtuin-modulating strategies could be evaluated in enriched PTSD mitochondrial-resilience subgroups, while SARM1-pathway modulators remain a longer-term possibility pending further validation. Plasticity-focused interventions could include optimized sleep timing around exposure sessions, exercise augmentation of trauma-focused therapy, and circadian-aligned treatment scheduling in patients with high pruning or plasticity gene signatures. Glutamate-modulating or anti-inflammatory adjuncts may be relevant for GRM3-high or NOS2-high anxiety subgroups. The study has several strengths. Brain-region-specific TWAS weights allowed more tissue-relevant interpretation than locus-level GWAS alone. The multi-gene-set design made it possible to detect coordinated pathway signals rather than isolated gene effects. Cross-disorder comparisons identified both shared and disorder-specific biology, and leave-one-out analyses provided gene-level resolution. Integrating aging modules with synaptic plasticity and pruning sets also provided a mechanistic bridge that has often been missing from prior genetic studies of anxiety disorders and PTSD. Several limitations should also be noted. TWAS estimates genetically predicted expression and does not directly measure transcript or protein levels in patient tissue, so causal inference remains provisional. Some of the strongest pathway signals were strongly influenced by individual genes, including SIRT3, NEK4, and C4A. This indicates that pathway results can be sensitive to dominant drivers and highlights the need for colocalization, replication, and functional validation. Complement and human leukocyte antigen (HLA)-related findings require caution because of extended linkage disequilibrium and possible immune-cell composition effects. Major histocompatibility complex (MHC)-exclusion analyses were performed, but they cannot fully remove this concern. The analysis also depends on curated gene-set definitions, which vary by database and often contain overlapping genes; therefore, related pathway hits should not be interpreted as statistically independent biological processes. Both GWAS inputs were restricted to European-ancestry participants, which improves ancestry consistency but limits generalizability to other populations. No pathway family survived the most stringent family-wise FDR summary, so the reported findings should be considered hypothesis-generating and prioritized for replication. Finally, the cross-sectional nature of the GWAS data prevents definitive conclusions about temporal order or causality in the proposed cascade. Future work should prioritize functional validation in human-induced pluripotent stem cell-derived neurons, astrocytes, and microglia. Targeted perturbation of SIRT3, SARM1, NEK4, and key pruning genes under glucocorticoid or inflammatory stress would be especially informative. Clinical cohort studies should combine the proposed biomarker panels with longitudinal symptom tracking, extinction-learning tasks, and multimodal neuroimaging, including diffusion MRI, resting-state connectivity, and fear-conditioning functional magnetic resonance imaging (fMRI). Subtype-enriched randomized trials comparing standard care with adjunctive mitochondrial support or plasticity-enhancing interventions would be a logical step toward precision treatment. In sum, the data support an association-based model in which genetic risk for anxiety disorders and PTSD converges on aging biology at the synapse while diverging in the molecular route taken. This framework moves beyond broad ideas of stress or inflammation and toward mechanistically grounded candidate subtypes that may improve risk stratification, biomarker development, and targeted intervention strategies. Conclusions This TWAS enrichment study suggests that genetic liability for anxiety disorders and PTSD intersects with core programs of cellular aging in brain tissue. The findings support a hypothesis-generating stress-aging-synapse cascade that is shared across disorders but also contains disorder-specific branches. PTSD risk showed a pronounced signature of NAD/sirtuin suppression and SARM1-linked synaptic vulnerability, whereas anxiety risk emphasized mitochondrial apoptosis thresholds and inflammatory-glutamatergic plasticity changes. Both disorders shared senescence, telomere, and DNA-damage programs. These findings support the hypothesis of biologically distinct subtypes, including a mitochondrial-resilience PTSD subtype and a plasticity/pruning profile that may help predict exposure-therapy response. They also point to biomarker panels and adjunctive treatment strategies that could be tested in future studies. By connecting molecular aging mechanisms to circuit-level dysfunction, this work provides a foundation for moving from descriptive genetic associations toward more precise approaches to risk assessment and treatment in anxiety disorders and PTSD.