Integrated proteomic and metabolomic analyses implicate redox-metabolic pathways in PTSD-associated multisystem disease and accelerated aging Authors: Pei-Fen Kuan, Frank D Mann, Xiaohua Yang, Roman Kotov, Sean Clouston, Olga Ilkayeva, Michael Muehlbauer, Christopher B Newgard, Benjamin J Luft Published: 2026 ABSTRACT AbstractPosttraumatic stress disorder (PTSD) is associated with increased risk of chronic disease and premature aging, yet underlying molecular mechanisms remain unclear. We performed plasma proteomics (SomaScan; 9404 proteins) and targeted metabolomics (145 metabolites) in 393 World Trade Center responders (232 with PTSD, 161 trauma-exposed controls). A total of 114 proteins and seven metabolites were differentially expressed in PTSD. Top proteins included NCAN, BCAN, NCAM1, and GDF15. Top metabolites included serotonin, lactate, glutamic acid, and cystathionine. Integrative analyses showed coordinated proteomic-metabolomic alterations, with widespread correlations linking metabolites involved in redox and amino acid metabolism to synaptic and oxidative stress-related proteins. Gene ontology enrichment identified neuronal plasticity, immune activation, extracellular matrix remodeling, and oxidative stress. Proteomic organ aging analyses revealed accelerated aging in the pancreas, lung, and at the organismal level in PTSD. These results reveal a redox-metabolic mechanism through which PTSD may drive multisystem aging and elevate disease risk.Subject terms: Molecular neuroscience, Post-traumatic stress disorder Introduction Post-traumatic stress disorder (PTSD) is a chronic and often debilitating psychiatric condition that can arise after exposure to traumatic events. About 6.1% of U.S. adults are estimated to experience PTSD at some point in their lives1, highlighting its significance as a public health concern. PTSD affects more than mental health: individuals with the disorder also face higher risks of cardiovascular disease, metabolic problems, neurodegeneration, and premature death2–6. These observations suggest that trauma leaves lasting marks on the body, though we still do not fully understand the molecular processes that link psychological stress to long-term physical illness. Current treatments, including psychotherapy and medications, are only partially effective, as seen in populations like World Trade Center (WTC) responders, where high rates of PTSD persist nearly two decades after 9/117,8. Thus, it is critical to delineate the molecular processes underlying PTSD to develop therapies that can prevent physical morbidity in people with PTSD. For many years, biological studies of PTSD have focused on stress-responsive systems, especially hypothalamic-pituitary-adrenal axis signaling and immune activity9,10. While these pathways explain immediate stress reactions, they do not capture the long-lasting, multisystem effects or aging-related consequences of PTSD. In animal studies, chronic stress is shown to affect processes shared across the brain, blood vessels, immune system, and other tissues including redox balance, mitochondrial function, and metabolism11–13. Disruptions in these processes appear to contribute to aging-related disease and may link trauma exposure to downstream tissue dysfunction14. Yet few human studies have examined whether stress produces coordinated molecular changes across multiple organ systems in PTSD. High-dimensional molecular profiling offers a promising way to fill these gaps. Plasma proteomics allows for the direct measurement of proteins that control nearly all physiological processes, from intracellular to intercellular signaling15. Metabolomic profiling, which examines small molecules reflecting upstream gene and protein activity, provides a complementary view of biochemical states across the body16. By integrating proteomic and metabolomic data, we can uncover coordinated molecular networks that might be missed when each type of data is analyzed separately. This approach gives a more comprehensive, systems-level understanding of the biological changes associated with trauma17–21. Reproducibility in PTSD omics research has improved with larger datasets, but variability in sample types, processing methods, and assay platforms continues to present challenges22,23. Most PTSD proteomic studies, including those from our group, have used the Olink platform17–19. In our earlier study17, we examined 276 plasma proteins across three Olink targeted panels (Neurology, Neuro Exploratory, and Cardiovascular II) and identified six proteins, including NCAN and BCAN, associated with PTSD and mild cognitive impairment. A follow-up study in a larger cohort of 936 responders to WTC disaster, using the Olink Neurology panel confirmed several of these findings, highlighting neuronal and vascular pathways in PTSD18. Complementing these proteomic findings, our previous metabolomics study identified metabolites involved in fatty acid mobilization and oxidation, sphingolipid signaling, and amino acid–based neuroimmune pathways relevant to innate immunity, inflammation, and neuronal excitability20. Only one previous study used the SomaScan 1.3 K assay, which profiled 1305 proteins in 340 veterans and 180 active-duty soldiers24. Olink panels generally cover fewer proteins, focusing on specific pathways or diseases, except for Olink Explore HT, which measures roughly 5400 proteins. In contrast, SomaScan allows broader coverage: the v4 and 11 K v5 assays quantify over 5000 and 10,000 proteins, respectively. These assays also offer the unique ability to estimate organ-specific proteomic aging, including the heart, kidney, liver, and lung25. In studies of aging and disease, these organ-level aging estimates correlate with risks for cardiovascular, renal, and other chronic conditions, reflecting cumulative stress and systemic decline25,26. To date, organ-specific proteomic aging has not been studied in PTSD, despite evidence linking PTSD to cardiometabolic disease, kidney dysfunction, and accelerated systemic aging2–6. In this work, we expand previous proteomic studies of PTSD by using the SomaScan 11 K v5 platform, which quantifies over 10,000 protein analytes, providing unparalleled insight into a broad span of proteomic markers across multiple biological systems. This was complemented by targeted metabolomic profiling of 145 metabolites across diverse biochemical classes. WTC responders provide a uniquely informative cohort, having experienced a discrete episode of intense trauma and been followed longitudinally with detailed clinical phenotyping27. PTSD is highly prevalent in this population and strongly associated with chronic disease burden28, allowing examination of molecular correlates of trauma-related disease risk. In turn, integrating proteomic networks, metabolic signatures, and organ-specific aging estimates offers a unique opportunity to uncover multisystem molecular mechanisms connecting trauma to long-term organ dysfunction, thereby investigating both psychiatric and physical health vulnerabilities in the context of chronic PTSD. Results Characteristics of study groups The analytic cohort consisted of 393 participants, including 232 with PTSD and 161 trauma-exposed controls. Participants with PTSD were, on average, 3 years older and had higher body mass index (BMI), and greater WTC exposure severity compared to trauma-exposed controls (p < 0.01). Depressive symptom severity, measured by the PHQ-9, was also substantially higher among participants with PTSD (p < 0.001). Nonsteroidal anti-inflammatory drug (NSAID) use was modestly higher in the PTSD group (p = 0.046), while selective serotonin reuptake inhibitor (SSRI) use and medical comorbidities were substantially higher (p < 0.001). There were no significant differences between the PTSD and control groups with respect to gender, race, smoking status, or statin use (Table 1). To further contextualize the analytic sample relative to the broader WTC cohort, longitudinal PTSD symptom trajectory analyses showed that the subsample was enriched for individuals with higher baseline symptom burden and less favorable symptom trajectories over time, consistent with the targeted case-control sampling design (Supplementary Fig. 8). Table 1Clinical characteristics of the study sampleVariablePTSD n = 232Trauma-exposed control n = 161P-valueAge Mean (SD)56.6 (7.8)53.8 (7.2)2.0 × 10−4Gender N (%) Male208 (90)148 (92) Female24 (10)13 (8)0.560Race N (%) White203 (88)137 (85) Other29 (12)24 (15)0.591BMI Mean (SD)31.7 (5.9)29.8 (4.6)3.7 × 10−4PHQ score Mean (SD)13.3 (6.2)0.47 (1.0)<1 × 10−15Smoking N (%) Current26 (11)20 (12) Former77 (33)38 (24)0.107 Never127 (55)103 (64) Missing2 (1)0 (0)WTC exposure N (%) Low21 (9)30 (19)0.0017 Intermediate145 (63)100 (62) High40 (17)21 (13) Very high14 (6)1 (1) Missing12 (5)9 (5)NSAID use N (%) Yes48 (21)20 (12) No184 (79)141 (88)0.046SSRI use N (%) Yes68 (29)0 (0) No164 (71)161 (100)1.2 × 10−13Statin use N (%) Yes67 (29)38 (24) No165 (71)123 (76)0.295Medical comorbidity (any) N (%) Yes140 (60)53 (33) No92 (40)108 (67)1.6 × 10−7P-values were derived using two-sided t-tests for continuous variables (age, BMI, and PHQ score) and two-sided chi-squared tests for categorical variables (gender, race, smoking status, WTC exposure, NSAID use, SSRI use, statin use, and medical comorbidity). Missing categories were excluded from chi-square tests for smoking status and WTC exposure. Medical comorbidity was defined as the presence of COPD, cardiac disease, diabetes, hypertension, or stroke. Exact p-values are reported where possible. Clinical characteristics of the study sample P-values were derived using two-sided t-tests for continuous variables (age, BMI, and PHQ score) and two-sided chi-squared tests for categorical variables (gender, race, smoking status, WTC exposure, NSAID use, SSRI use, statin use, and medical comorbidity). Missing categories were excluded from chi-square tests for smoking status and WTC exposure. Medical comorbidity was defined as the presence of COPD, cardiac disease, diabetes, hypertension, or stroke. Exact p-values are reported where possible. PTSD is associated with coordinated proteomic and metabolomic alterations To characterize molecular alterations associated with PTSD, we performed differential analyses of plasma proteomics and targeted metabolomics. Figure 1A, B display the volcano plots of the proteomics and metabolomics data, respectively. For the proteomics data, there were approximately equal numbers of upregulated and downregulated proteins, whereas more metabolites were upregulated in PTSD. A total of 121 aptamer-based protein analytes, corresponding to 114 unique proteins, were differentially expressed between PTSD and controls at FDR < 0.05 from our primary discovery models. At the aptamer level, 73 analytes were upregulated and 48 were downregulated; when collapsed to unique proteins, 68 were upregulated and 46 were downregulated. In contrast, 7 metabolites were differentially expressed at FDR < 0.05 (6 upregulated, 1 downregulated). Heatmaps of the significant proteins and metabolites are presented in Fig. 1C, D. Differential proteomics and metabolomics analysis in PTSD. A Volcano plot of differential protein expression. B Volcano plot of differential metabolite expression. C Heatmap of the 121 aptamer-based protein analytes significantly associated with PTSD at FDR < 0.05. D Heatmap of the 7 metabolites significantly associated with PTSD at FDR < 0.05. Differential expression analyses were performed using two-sided linear regression models with HC3 robust standard errors, adjusting for age, gender, and race. Multiple testing correction was performed using the Benjamini-Hochberg false discovery rate procedure. Source data are provided as a Source Data file. The top ten protein analytes ranked by statistical significance were NCAM-120, NCAM1, GNPTG, NCAN, NPTXR, DNAJ89, MSTN, ANGPT2, GDF11 and GDF15, with GNPTG, DNAJ89, ANGPT2 and GDF15 showing upregulation in PTSD. The protein with the largest AUC was GDF15, with a value of 0.709, and was ranked tenth among differentially expressed protein analytes. The seven differentially expressed metabolites were lactate, serotonin, hydroxylysine, glutamic acid, cystathionine, proline, and sphingomyelin SM(d44:2). All except serotonin were upregulated in PTSD, highlighting alterations in energy and amino acid metabolism in affected individuals. Of note, glutamic acid and cystathionine, two metabolites linked to the system x_c⁻ transporter, were significantly upregulated in PTSD, while related metabolites in the same pathway (cysteine and cystine) showed similar upward trends (p = 0.06, FDR = 0.277 for cysteine and p = 0.02, FDR = 0.168 for cystine). The top metabolite, lactate is derived by glycolytic metabolism of glucose, which was also nominally upregulated in PTSD (p = 0.005, FDR = 0.07). Similarly, three metabolites produced by activation of lipolysis, namely glycerol (p = 0.03, FDR = 0.199), total ketones (p = 0.01, FDR = 0.154), and NEFA (p = 0.02, FDR = 0.168) were also nominally upregulated in PTSD. Together, these signatures of increased glycolytic metabolism of glucose and mobilization and oxidation of lipid stores, may suggest an increased catabolic state in individuals with PTSD. Sensitivity analyses confirmed the robustness of these findings (Supplementary Data 1). When additionally adjusted for BMI, smoking status, and WTC exposure, effect size estimates across all 9,404 protein analytes were highly concordant with the primary model (Spearman rs = 0.93), with complete directional agreement among the 121 protein analytes identified in the primary analysis. Although the number of FDR-significant analytes decreased under the more conservative model, 115 of the 121 protein analytes retained nominal significance (p < 0.05), indicating that the observed signal was largely preserved. To further address potential confounding by smoking or extreme WTC exposure, analyses were repeated in participants who were never or former smokers and had low or intermediate exposure (n = 263). Effect size estimates remained highly concordant with those from the full cohort (Spearman rs = 0.883), with 100% directional agreement among the 121 protein analytes (Supplementary Data 1). Additional sensitivity analyses, including nonlinear adjustment for age and BMI and models adjusting for NSAID, SSRI, and statin use, as well as medical comorbidities, yielded highly consistent results (Spearman rs > 0.92) with complete directional agreement among the 121 proteins (Supplementary Data 1). Restricting analyses to participants with minimal-to-mild depressive symptoms (PHQ < 10; n = 233) also produced highly concordant effect estimates (Spearman rs = 0.74), with 100% directional agreement among the 121 protein analytes. Detailed results of these analyses are provided in Supplementary Note 1 and Supplementary Data 1. Parallel sensitivity analyses for the metabolomics data yielded consistent results. Among the 7 metabolites identified in the primary analysis, most remained nominally significant after additional covariate adjustment and in restricted analyses, with complete directional concordance (Supplementary Data 1). In our secondary analysis of tryptophan (Trp)-related and branched chain amino acid (BCAA)-related metabolites, including the branched chain ketoacids (BCKA) and branched chain hydroxyacid (BCHA) metabolites, duplicated measurements of tryptophan, kynurenine, and serotonin across the primary and secondary metabolomics datasets showed moderate-to-high reproducibility (Supplementary Fig. 3). Five metabolites were identified at nominal significance (p < 0.05) (Supplementary Table 1): three Trp-related metabolites including indole-3-propionic acid (down), indole-3-lactic acid (up), and kynurenic acid (down), and two BCHA metabolites including 2-hydroxy-3-methylvaleric acid (HMVA) (up) and 2-hydroxyisovaleric acid (HIVA) (up). Among these, indole-3-propionic acid, HMVA and indole-3-lactic acid also remained significant at FDR < 0.05 when the FDR correction was restricted to the secondary metabolomics dataset. Additional analyses involving this dataset are provided in Supplementary Note 3 and Supplementary Figs. 4–7. Proteomic and metabolomic signatures show strong concordance with prior PTSD studies Our findings showed strong concordance with prior proteomic and metabolomic investigations. In our earlier targeted proteomic analysis using Olink panels covering 276 proteins17, eleven of the 121 differentially expressed protein analytes identified in the present work were measured, yet all (NCAN, MSTN, KYNU, BCAN, RGMA, THBS2, IL1RN, FGF21, ASGR1, and KITLG) but one (PIGR) demonstrated consistent effect directions, with NCAN and BCAN reaching statistical significance in both datasets (Table 2). When comparing proteins previously associated with comorbid PTSD and mild cognitive impairment in our earlier study17, the ten corresponding aptamers in the present dataset also showed a high degree of directional agreement: eight demonstrated effects in the same direction (Supplementary Data 2). Table 2Subset of differentially expressed (DE) aptamer-based protein analytes identified in the present study that were replicated in other studies, namely WTC Study 1 (Kuan et al.17), PSYCHENCODE2-UKBB19, and Muhie et al.24AptNameTargetGene SymbolESWTC Study1 KuanPSYENCODE2 UKBBMilitaryseq.4498.62NCAM-120NCAM1DownDown ***Yesseq.20161.41NCAM1NCAM1DownDown ***Yesseq.15573.110CSPG3NCANDownDown ***seq.27221.34Myostatin-PropeptideMSTNDownDownDown ***seq.4374.45MIC-1GDF15UpUp ***seq.34255.32ATS8ADAMTS8DownDown ***seq.16818.200CDCP1CDCP1UpUp ***seq.2677.1ERBB1EGFRDownYesseq.6556.5ENPP5ENPP5DownDown ***seq.4559.64KYNUKYNUUpUpUp ***seq.5635.66TREM2TREM2UpUp ***seq.30695.4PON3PON3DownDown ***seq.3461.58PGCBBCANDownDown ***Down ***seq.2201.17EndostatinCOL18A1UpUp ***seq.31918.65EST2CES2UpUp ***seq.3216.2PIGRPIGRUpDownUp ***seq.5483.1RGMARGMADownDownDown ***seq.7182.1carboxylesterase, liverCES1UpUp ***seq.3339.33TSP2THBS2UpUpseq.11196.31Collagen alpha-3(VI):BPTI/Kunitz inhibitorCOL6A3UpUp ***seq.5307.12Coagulation Factor IXabF9UpUp ***seq.9282.12CRIS2CRISP2DownDown ***seq.2974.61contactin-1CNTN1DownDown ***Yesseq.11851.21TREM2TREM2UpUp ***seq.10970.3NAR3ART3DownDown ***seq.16300.4TREM2TREM2UpUp ***seq.5353.89IL-1RaIL1RNUpUpUp ***seq.29534.9FGF21FGF21UpUpUp ***seq.4568.17SLIK5SLITRK5DownYesseq.20590.13NPYNPYDownUp ***seq.5452.71ASGR1ASGR1UpUpUp ***seq.19241.31RBP-IIIRBP5UpUp ***seq.15386.7FABPAFABP4UpUp ***seq.16307.22UNC5H4UNC5DDownYesseq.15487.164carboxylesterase, liverCES1UpUp ***seq.3446.7IL-18 RaIL18R1UpUp ***seq.9377.25SCFKITLGDownDownFor WTC Study 1 (Kuan et al.17) and PSYCHENCODE2-UKBB19, we report the effect size (ES) direction and whether the protein was significant at FDR < 0.05 in the original study (***). Subset of differentially expressed (DE) aptamer-based protein analytes identified in the present study that were replicated in other studies, namely WTC Study 1 (Kuan et al.17), PSYCHENCODE2-UKBB19, and Muhie et al.24 For WTC Study 1 (Kuan et al.17) and PSYCHENCODE2-UKBB19, we report the effect size (ES) direction and whether the protein was significant at FDR < 0.05 in the original study (***). Our second prior study18, which analyzed 92 neurologically relevant proteins in a larger cohort of 936 WTC responders, identified ten proteins associated with PTSD diagnosis or symptom severity (PCL total scores). These proteins corresponded to sixteen aptamers in the current analysis, and most showed consistent effect directions, including replicated associations for NCAN and BCAN, i.e., two proteins repeatedly emerging across studies (Supplementary Data 2). Results were comparable in subset analyses by excluding all participants who overlapped with either prior study (n = 91 removed) (Supplementary Data 2). We observed a strong convergence when comparing our findings with the recent PSYCHENCODE2-UKBB PTSD proteomic study19, which assessed 2941 plasma analytes using Olink Explore 3072 and identified 299 PTSD-associated proteins. Of the 339 corresponding aptamers (some proteins have multiple aptamers) in our dataset, 244 (72%) showed consistent effect directions and 98 achieved nominal significance (p < 0.05), demonstrating broad reproducibility across platforms (Supplementary Data 2). Notably, 32 of the 121 differentially expressed protein analytes in our study overlapped directly with the PSYCHENCODE2-UKBB set, including NCAM1, BCAN, and GDF15 (Table 2). Among these overlapping analytes, 31 of 32 showed consistent effect directions. Finally, six of the 121 differentially expressed protein analytes (NCAM-120, NCAM1, EGFR, CNTN1, SLITRK5, and UNC5D) overlapped with the 36 PTSD-relevant proteins identified by Muhie et al.24 through network-based analyses incorporating genetic associations (Table 2), providing further evidence that the molecular signals observed here converge with those reported in a different population. For metabolomics, our findings also aligned with our prior Metabolon-based study20, which identified hexosylceramide HCER(26:1) as upregulated in PTSD. In the present study, this metabolite also showed higher levels in PTSD, though it was not statistically significant (nominal p = 0.221). Moreover, all seven significant metabolites identified in the present study showed consistent effect directions compared to our previous results20. Serotonin, in particular, was nominally significant (p < 0.05) in our earlier study. PTSD-associated molecular alterations implicate neuronal, immune, and extracellular matrix pathways Gene Ontology (GO) enrichment analysis of proteomics data identified 25 significantly enriched categories at FDR < 0.1 (Table 3). Twelve were directly related to neuronal processes, including neuron projection regeneration, glial cell differentiation, gliogenesis, axon regeneration, response to axon injury, regulation of nervous system development, regulation of axon regeneration, regulation of neuron projection regeneration, negative regulation of neuron projection regeneration, axon development, regulation of neurogenesis, and glial cell development. Additional categories included immune function, oxidative stress and inflammatory pathways, extracellular matrix (ECM) structure, developmental and growth regulation, and nutrient metabolism. No KEGG pathways reached statistical significance at FDR < 0.1. Table 3Gene Ontology (GO) categories significantly enriched at FDR < 0.1OntologyIDDescriptionMajor themeCCGO:0030312external encapsulating structureStructural/Extracellular MatrixCCGO:0031012extracellular matrixStructural/Extracellular MatrixCCGO:0062023collagen-containing extracellular matrixStructural/Extracellular MatrixBPGO:0099560synaptic membrane adhesionStructural/Extracellular MatrixBPGO:0097106postsynaptic density organizationStructural/Extracellular MatrixBPGO:0031102neuron projection regenerationNeuronal/Nervous SystemBPGO:0031103axon regenerationNeuronal/Nervous SystemBPGO:0048678response to axon injuryNeuronal/Nervous SystemBPGO:0048679regulation of axon regenerationNeuronal/Nervous SystemBPGO:0070570regulation of neuron projection regenerationNeuronal/Nervous SystemBPGO:0070571negative regulation of neuron projection regenerationNeuronal/Nervous SystemBPGO:0061564axon developmentNeuronal/Nervous SystemBPGO:0051960regulation of nervous system developmentNeuronal/Nervous SystemBPGO:0050767regulation of neurogenesisNeuronal/Nervous SystemBPGO:0010001glial cell differentiationNeuronal/Nervous SystemBPGO:0042063gliogenesisNeuronal/Nervous SystemBPGO:0021782glial cell developmentNeuronal/Nervous SystemBPGO:0031099regenerationDevelopmental/Growth RegulationBPGO:0042554superoxide anion generationImmune/Oxidative Stress/InflammationBPGO:0009410response to xenobiotic stimulusImmune/Oxidative Stress/InflammationBPGO:0071466cellular response to xenobiotic stimulusImmune/Oxidative Stress/InflammationBPGO:0009749response to glucoseMetabolic/Nutrient ResponseBPGO:0009746response to hexoseMetabolic/Nutrient ResponseBPGO:0034284response to monosaccharideMetabolic/Nutrient ResponseCCGO:0005788endoplasmic reticulum lumenCellular Components (Intracellular) Gene Ontology (GO) categories significantly enriched at FDR < 0.1 Results from Independent Component Analysis (ICA) identified top modules associated with PTSD that were enriched in pathways related to neuronal development, axon guidance, and synaptic organization; ECM structure and cell adhesion; growth factor and receptor-mediated signaling; immune and inflammatory processes; and developmental and morphogenetic programs related to tissue remodeling (Supplementary Note 2 and Supplementary Data 5). Integrated proteomic-metabolomic analyses reveal cystine-glutamate and redox dysregulation in PTSD Integrative analysis of proteomic and metabolomic data revealed coordinated molecular alterations in PTSD. Across the seven significant metabolites, 120 protein analytes were significantly correlated with at least one of the metabolites (Fig. 2A). The number of correlated protein analytes varied by metabolite: 104 with lactate, 96 with cystathionine, 108 with glutamic acid, 108 with hydroxylysine, 71 with proline, and 36 with serotonin. No proteins were significantly correlated with sphingomyelin SM(d44:2) at the FDR < 0.05 threshold, indicating that it did not show strong coordinated associations in this dataset. Among these 120 protein analytes, eighteen protein analytes (NCAM-120, NCAM1, GDF15, HTRA1, RP9, MORC3, ACSM4, THBS2, PXDN, CNTN1, GBE1, MPST, GOLM1, NBPF1, NCF2, NCKAP5, TNFRSF10D, and FABP4) were correlated with six of the seven metabolites, suggesting they may act as hub proteins driving multi-omic dysregulation in PTSD. Notably, GDF15, a key mediator of cachexia and catabolic metabolism was broadly correlated across metabolites, consistent with a central role in systemic metabolic and stress-response dysregulation in PTSD. Integrative proteomics-metabolomics analysis in PTSD. A Heatmap summarizing significant protein-metabolite correlations (FDR < 0.05, denoted by *), identified from differential expression analysis. Rows represent aptamer-based protein analytes and columns represent metabolites. B–D Representative examples of protein-metabolite pairs showing nominal interaction effects (p < 0.01) by PTSD status. Scatterplots illustrate differential slopes between participants with PTSD and trauma-exposed controls: B NCAN versus cystathionine. C NOX4 versus cystathionine. D MXRA8 versus hydroxylysine. Protein-metabolite correlations in (A) were evaluated using two-sided Spearman rank correlation tests with Benjamini-Hochberg false discovery rate correction applied within metabolite. Interaction effects in (B–D) were evaluated using two-sided linear regression models including protein analyte × PTSD interaction terms, adjusting for age, gender, and race. Statistical inference was based on HC3 robust standard errors. Shaded bands represent 95% confidence intervals around fitted linear regression lines. Source data are provided as a Source Data file. At FDR < 0.05, no protein-metabolite interaction terms remained statistically significant by PTSD status. At a nominal threshold of p < 0.01, six protein-metabolite pairs showed evidence of interaction, including cystathionine versus NCAN, cystathionine versus NOX4, cystathionine versus ENPP5, cystathionine versus CACNA2D3, hydroxylysine versus MXRA8, and serotonin versus MORC3 (Fig. 2). We therefore performed exploratory stratified analyses by PTSD status for these six pairs. In five pairs (cystathionine versus NCAN, cystathionine versus NOX4, cystathionine versus ENPP5, cystathionine versus CACNA2D3, and hydroxylysine versus MXRA8), the metabolite-protein association was stronger and statistically significant among participants with PTSD, while associations were weak or non-significant in trauma-exposed controls. In contrast, the serotonin versus MORC3 association appeared stronger in controls than in participants with PTSD. Stratified results are provided in Supplementary Data 3. Although these findings suggest potential PTSD-related differences in selected metabolite-protein relationships, interaction terms did not survive multiple testing correction and should therefore be interpreted as exploratory. Supplementary Fig. 2 presents pairwise (protein-metabolite) plots of metabolites associated with the system x_c⁻ transporter (glutamic acid, cystathionine, cysteine, cystine) and NOX4, showing generally positive correlations. Beyond the cystathionine-NOX4 pair, the cysteine-cystine and cystathionine-glutamic acid pairs also exhibited stronger correlations in the PTSD group compared with controls, highlighting altered regulation of system x_c⁻-related metabolic pathways in PTSD. PTSD is associated with accelerated proteomic aging across multiple organ systems We next used our proteomics data set to examine whether PTSD was associated with accelerated biological aging across multiple organ systems. The Spearman rank correlations between predicted age measures and chronological age ranged from 0.222 to 0.843 (Table 4), where the global aging measures showed strong associations with chronological age (Conventional: Spearman rs = 0.843; Organismal: Spearman rs = 0.826; CognitionOrganismal: Spearman rs = 0.721). Five of the 25 age measures were significantly associated with PTSD at FDR < 0.05: Pancreas, CognitionLung, Organismal, CognitionOrganismal, and Conventional (Fig. 3). In all cases, PTSD participants exhibited accelerated predicted biological aging compared to trauma-exposed controls. These findings suggest that molecular signatures of PTSD extend beyond brain-related processes to encompass systemic alterations in metabolic and organ-specific aging. Table 4Spearman rank correlations between predicted organ age measures and chronological ageOrganrsConventional0.843Organismal0.826CognitionOrganismal0.721Immune0.561Liver0.46Brain0.401Pancreas0.395Intestine0.388Muscle0.381CognitionLiver0.367CognitionIntestine0.365CognitionHeart0.355Heart0.355CognitionPancreas0.353CognitionBrain0.336Artery0.313CognitionImmune0.311CognitionMuscle0.31CognitionArtery0.309Lung0.302Kidney0.298Adipose0.283CognitionAdipose0.27CognitionKidney0.268CognitionLung0.222 Spearman rank correlations between predicted organ age measures and chronological age Associations between chronological age and predicted proteomic age measures, and comparison of age acceleration by PTSD status. Scatterplots comparing chronological age with predicted age are shown in the top row, and boxplots comparing organ aging between individuals with PTSD and trauma-exposed controls are shown in the bottom row for organ systems significantly associated with PTSD. A Pancreas. B CognitionLung. C Organismal. D CognitionOrganismal. (E) Conventional. Each point represents an independent participant sample from the WTC responder cohort (n = 391 biologically independent plasma samples with SomaScan data; PTSD = 231, trauma-exposed controls = 160). No technical replicates were used in the statistical analyses. In the boxplots, the center line represents the median, box bounds represent the 25th and 75th percentiles, and whiskers extend to the smallest and largest values within 1.5 × IQR from the hinges. Individual participant values are overlaid as jittered points. Source data are provided as a Source Data file. Enrichment analyses of the full ranked proteomic results from the PTSD differential expression analysis revealed significant enrichment for brain-specific proteins. Among the 121 differentially expressed protein analytes, seven proteins (BCAN, OMG, SEZ6L, CNTN1, NPTXR, NCAN, and IGLON5) overlapped with brain-specific proteins identified by [12] (Supplementary Data 4). Discussion In this study, we used large-scale plasma proteomics and targeted metabolomics to study the molecular changes associated with PTSD in WTC responders, nearly 18 years after the 9/11 attacks. Using primary discovery models adjusted for age, gender, and race, together with robust inference with heteroskedasticity-consistent (HC3) standard errors, we identified 121 aptamer-based protein analytes (114 unique proteins) and 7 metabolites that were differentially expressed in individuals with PTSD. These molecular signatures converge on pathways related to neuronal structure, immune activation, ECM remodeling, metabolic regulation, and redox balance. Beyond replicating previously reported proteomic and metabolomic changes, our findings provide important insights into system x_c⁻ metabolite dysregulation and reveal coordinated multi-omic patterns linking sulfur amino acid metabolism, oxidative stress, and tissue remodeling. In addition, we observed evidence of accelerated biological aging across multiple organ-enriched plasma proteomic signatures. Together, these findings indicate that PTSD is accompanied by coordinated, multisystem molecular dysregulation that extends well beyond the central nervous system and provides insight into the long-term biological embedding of traumatic stress. In our proteomic analyses, we observed a mix of proteins that were upregulated or downregulated in PTSD. Among these were NCAN, BCAN, NCAM1, and GDF15, i.e., proteins that have been previously linked to psychiatric disorders, neurodegeneration, and stress-related biology17–19. Interestingly, GDF15 stood out as having the strongest discriminative ability. This protein has been associated with mitochondrial stress, systemic inflammation, cachexia, and cellular senescence29,30, suggesting that it could be a meaningful marker of PTSD-related biological changes. Furthermore, our result is consistent with earlier Olink-based studies from our group17,18 and largely overlaps with the large-scale PSYCHENCODE2-UKBB list of PTSD-associated proteins19. Moreover, some of these proteins also appeared in co-expression network analyses that mapped to genetic variants linked to PCL scores in military personnel24. Taken together, these results point toward a set of core molecular signals that may contribute to PTSD risk across different populations. Our metabolomic analysis highlighted alterations in energy metabolism and excitatory neurotransmission. Specifically, elevated lactate and glutamic acid indicate shifts in glycolytic flux and excitatory signaling31, while elevation of cystathionine in concert with glutamic acid suggests dysregulation of the system x_c⁻ antiporter, which regulates glutamate-cystine exchange and oxidative stress regulation32. This finding is supported by similar upward trends in two other system x_c- metabolites, cysteine and cystine, with cystine reaching nominal significance and cysteine showing a comparable trend. Upregulation of lactate, glycerol, ketones, and NEFA indicates an increased catabolic state and altered systemic energy utilization in PTSD, consistent with prior findings of disrupted energy metabolism in PTSD20,33. These findings may relate to the increase in GDF15, a circulating peptide that acts centrally to cause reductions in food intake and peripherally to enhance lipolysis and fatty acid oxidation34. Serotonin was reduced among PTSD cases in the present study, consistent with decades of evidence for serotonergic dysfunction in trauma-related disorders, although the direction of regulation has been mixed across studies35,36, partly due to differences in biospecimen types. The downregulation pattern observed here may contribute to specific symptoms of PTSD, including hyperarousal, irritability, and sleep disturbance35, and is also consistent with the clinical efficacy of SSRIs for treatment of PTSD37. We note that global upregulation of glutamatergic and oxidative stress signaling supports excitotoxicity and redox disruption models of chronic PTSD38,39, suggesting that serotonergic, metabolic, and immune signaling dysregulation may collectively drive both psychiatric illness and systemic biological alterations. Secondary analyses of tryptophan (Trp)-, BCAA-related metabolites suggested PTSD-related dysregulation of BCAA, tryptophan, and kynurenine pathways, further reinforcing alterations in redox balance, mitochondrial function, and stress-related pathways. Our GO analysis revealed processes involved in neuronal development and repair, including axon regeneration, synapse organization, and glial differentiation. These results are in line with evidence that maladaptive plasticity and abnormal remodeling of neural circuits are important features of PTSD40. We also observed enrichment of immune, inflammatory, and ECM pathways, reinforcing the systemic nature of PTSD biology and the interplay between neuroinflammation, tissue remodeling, and peripheral immune dysregulation10. Notably, despite the chronicity of PTSD in our participants, our findings show enrichment of pathways related to neuronal regeneration, suggesting the presence of molecular repair processes long after the traumatic event, and in some cases decades later. Findings from ICA reinforce this view, indicating that PTSD involves coordinated disruptions across immune, structural, and neuronal systems, which aligns with the patterns seen in our differential expression and pathway analyses. In integrative proteome-metabolome analyses, we observed broad coordinated associations between the significant metabolites and many protein analytes, consistent with multi-omic coupling of metabolic and proteomic systems. However, protein-metabolite interaction terms did not remain statistically significant after multiple testing correction. At a nominal threshold, a small number of pairs showed suggestive evidence of differential metabolite-protein slopes by PTSD status, including cystathionine with synaptic, ECM and oxidative stress-related proteins (e.g., NCAN and NOX4), as well as hydroxylysine with MXRA8. Exploratory stratified analyses suggested that several of these metabolite-protein associations were stronger among participants with PTSD. Cystathionine is an important metabolite in transsulfuration that links methionine metabolism with glutathione synthesis and cellular redox stabilization32. The stronger association with NCAN (which regulates the composition of the ECM and synaptic stability41) suggest that sulfur amino acid metabolism may regulate neuronal structural remodeling associated with PTSD. In addition, hydroxylysine, a collagen-derived amino acid associated with ECM turnover42, showed a nominal interaction with MXRA8, a matrix-remodeling-associated protein43, further supporting a potential link between metabolite signals of ECM remodeling and proteomic markers of tissue structural regulation. This observation is consistent with our enrichment analyses, which highlighted ECM organization and tissue remodeling as prominent biological themes in the PTSD-associated proteomic signatures. We also observed a robust correlation with NOX4, which generates reactive oxygen species and mediates signaling related to oxidative stress44, thus offering a potential mechanistic link between metabolic dysregulation and oxidative/inflammatory pathways. Furthermore, the elevation of GDF15 observed in PTSD is consistent with the upregulation of the stress response downstream of redox and mitochondrial dysfunction, offering a potential systemic correlate of the metabolic and oxidative stress signatures that we identified through our proteome-metabolome analysis. Because these interaction effects did not survive FDR correction, they should be interpreted as exploratory and hypothesis-generating. Nonetheless, they provide biologically coherent candidates linking sulfur amino acid metabolism to oxidative stress and neuronal or ECM remodeling in chronic PTSD. Supplementary Fig. 2 further supports these observations, showing that metabolites associated with the system x_c⁻ transporter (glutamic acid, cystathionine, cysteine, cystine) exhibit generally positive correlations with each other, with stronger pairwise relationships among PTSD cases compared to controls. Notably, cystathionine-glutamic acid and cysteine-cystine correlations were heightened in PTSD, reinforcing dysregulation of glutamate-cystine exchange and redox balance32. Given the role of system x_c⁻ in regulation of cellular redox status via synthesis of glutathione32, these findings could indicate alterations in responses to oxidative stress in PTSD. Altogether, these findings suggest a coordinated shift in metabolic protein networks that couple sulfur amino acids, excitatory neurotransmission, and oxidative and inflammatory states with protein markers for synaptic and ECM remodeling. An important finding of our study is that PTSD is associated with accelerated proteomic aging across multiple organ-enriched plasma signatures, including the pancreas and lung, as well as at the organismal level. In particular, accelerated pancreatic aging suggests that long-term trauma exposure may be linked to earlier declines in pancreatic function, although our analysis at this stage does not discriminate the endocrine from the exocrine pancreas. The combination of dysregulation of glutamate-cystine exchange and redox balance, and possible alterations in pancreatic hormone secretion could contribute to the observed increase in the risk of metabolic diseases such as type 2 diabetes in subjects with PTSD5. We also found that both conventional and cognition-informed proteomic aging models were accelerated in PTSD, reinforcing PTSD as a systemic condition that accelerates aging. These findings extend prior work linking PTSD to telomere shortening45, DNA methylation and transcriptomic age acceleration3,46, and heightened risk of age-related disease2,4–6. Importantly, complementary enrichment analyses revealed that PTSD-associated proteomic alterations are not randomly distributed but are preferentially enriched among brain-specific proteins. The overlap of multiple differentially expressed proteins with brain-specific markers, of which many are involved in synaptic structure and neuronal connectivity, provides convergent evidence that central nervous system-relevant molecular perturbations are detectable in plasma long after trauma exposure. Together, the organ-level insights provided by proteomic aging models25 reveal that PTSD may affect different physiological systems in distinct ways, highlighting the potential of organ-specific biomarkers to predict long-term health outcomes in this high-risk population. At the same time, we emphasize that these plasma-based “organ age” measures should not be interpreted as reflecting proteins exclusively expressed in a single organ or as direct measurements of tissue-specific aging. Rather, they represent plasma-derived aging signatures constructed from protein sets enriched for organ-predominant expression (as defined using transcriptomic reference atlases) and validated in prior work against organ-relevant outcomes25. Thus, the observed associations support the interpretation that PTSD relates to systemic aging biology captured by organ-enriched plasma proteomic patterns, while acknowledging the inherent limitation that plasma signals cannot uniquely localize biological processes to specific tissues. Our findings were robust across multiple sensitivity analyses in both proteomic and metabolomic data. Effect estimates remained consistent after adjusting for BMI, smoking status, and WTC exposure, as well as in analyses restricted to never or former smokers with lower exposure. Additional adjustment for medical comorbidities and medication use had minimal impact on the results. Restricting analyses to participants with minimal-to-mild depressive symptoms yielded similar patterns, indicating that the observed molecular signatures are not driven solely by depressive symptoms. Overall, these findings suggest that the identified signatures reflect the broader biology of chronic PTSD in a real-world, comorbid population rather than “pure” PTSD in isolation. One of the key strengths of this study is the integration of large-scale proteomic and metabolomic data within a well-characterized cohort, along with replication against prior datasets and the innovative use of validated organ-enriched aging signatures. Together, these approaches provide a broad and detailed view of the molecular changes linked to PTSD. Nonetheless, our findings must be considered in the context of several limitations. While the WTC cohort offers a unique chance to study trauma-related biology in people with a shared exposure history, the analytic sample represents an enriched case-control design drawn from a larger parent cohort and may not be representative of cohort-wide distributions; replication in independent PTSD populations will be necessary to confirm that our findings generalize more broadly. Another limitation is that plasma-based proteomics and metabolomics may not fully capture events in the brain or other organs; complementary studies using cerebrospinal fluid, neuroimaging, or postmortem brain and other tissue sampling could provide deeper insight. Despite extensive sensitivity analyses, residual confounding by unmeasured clinical, environmental, and treatment-related factors could still have influenced our results. Finally, because this analysis was cross-sectional, we cannot determine causality or directionality of the associations observed. Given that biospecimens were collected many years after the initial trauma exposure, the molecular signatures identified here likely reflect long-term biological embedding of chronic PTSD rather than acute post-trauma responses. This interpretation is further supported by longitudinal PTSD symptom trajectory analyses8, which indicate that the analytic subsample was enriched for individuals with persistently elevated symptom burden and less favorable symptom trajectories over time, consistent with chronic rather than transient PTSD. Longitudinal research will be crucial to test whether these biomarkers can predict disease progression, treatment response, or longer-term health outcomes. In conclusion, we present one of the most comprehensive proteomic and metabolomic profiles of PTSD to date. Our analyses uncover reproducible protein and metabolite signatures, crosstalk between molecular pathways, and point to accelerated biological aging across multiple organs. Together, these findings link PTSD to neuronal, immune, metabolic, and aging-related processes, emphasizing its systemic nature and informing future biomarker research. Methods Participants and PTSD assessment Participants were recruited from the Stony Brook World Trade Center (WTC) Health Program27. The study was approved by the Stony Brook University Institutional Review Board, and all participants provided written informed consent. Blood samples were collected in 2019, nearly 18 years after the 9/11 attacks. Inclusion criteria required sufficient English language proficiency to complete the study protocol. Both male and female participants were eligible. Gender information was obtained through participant self-report and was included as a covariate in all primary statistical analyses. Participants did not receive financial compensation for participation in this study. Responders with a history of head injuries, brain tumors, or cancers were ineligible for the study. Probable PTSD was assessed using the 17-item Posttraumatic Stress Disorder Checklist for a specific stressor (PCL-S)47, a validated self-report measure based on DSM-IV criteria. Participants rated items on a 5-point Likert scale (1 = “not at all” to 5 = “extremely”), with excellent internal consistency (Cronbach’s α = 0.96). A PCL total score >43 was used to indicate probable PTSD. The trauma-exposed control group was defined as asymptomatic (PCL total score <23) and underwent an additional medical record review to exclude individuals with a clinical history of other psychiatric disorders, including major depressive disorder. Participants meeting criteria for probable PTSD were not excluded on the basis of psychiatric comorbidities. The analytic sample was assembled using an enriched case-control design from the larger WTC responder cohort, resulting in overrepresentation of individuals at the extremes of PTSD symptom burden. Accordingly, the case-control composition does not reflect the prevalence of PTSD in the broader WTC cohort. Depressive symptom severity was assessed using the Patient Health Questionnaire-9 (PHQ-9)48, a validated 9-item self-report measure of depressive symptoms over the past 2 weeks. Items are rated on a 4-point Likert scale (0 = “not at all” to 3 = “nearly every day”), yielding a total score ranging from 0 to 27, with higher scores indicating greater depressive symptom burden. The analytic cohort included 393 participants (232 PTSD, 161 trauma-exposed controls), with a mean age of 55.5 years (SD = 7.7), 87% White, and 41% current or former smokers. Additional covariates included BMI and exposure49. Exposure was defined using a composite measure incorporating duration of work at the WTC site, exposure to the dust cloud from the collapse of the buildings, and work on the debris pile. Participants were categorized into four ordered exposure groups (low, intermediate, high, and very high), reflecting increasing exposure intensity49. Physician-diagnosed medical conditions were abstracted from electronic medical records within the WTC Health Program. A composite medical comorbidity variable was defined as the presence of chronic obstructive pulmonary disease, cardiac disease, diabetes mellitus, hypertension, or stroke (any vs none). Current medication use at the time of blood draw was assessed using structured clinical intake forms, incorporating both program records and participant self-report. NSAID use included both prescription and over-the-counter medications. SSRIs use and statin use were similarly defined using combined program and self-reported data. All medication variables were coded as binary indicators (yes/no) reflecting use at the study visit. To contextualize the analytic sample relative to the broader WTC cohort, longitudinal PTSD symptom trajectories derived from repeated PCL assessments were compared between the analytic subsample and the full cohort (Supplementary Fig. 8). Blood sampling Participants were instructed to fast for at least 8 h and to refrain from NSAID use, aspirin, and vigorous exercise on the morning of blood collection. Blood draws were conducted under standardized clinical procedures during routine WTC Health Program visits. Whole blood was collected into K2-EDTA tubes (BD Vacutainer, Franklin Lakes, NJ), gently inverted 8–10 times, and centrifuged at 2000 × g for 15 min at 4 °C. Plasma was aliquoted into polyethylene tubes and stored at −80 °C until analysis. Proteomics profiling Plasma proteins were quantified using the SomaScan 11 K v5 assay, a high-throughput multiplex platform based on modified single-stranded DNA aptamers, which provide high specificity and sensitivity. The assay measures more than 11,000 aptamer-based protein analytes, comprising over 10,000 aptamers representing more than 9600 unique human proteins; some proteins are represented by multiple aptamers. These target analytes span a broad range of biological pathways, including neurology, cardiology, inflammation, and oncology. Additional details are available elsewhere (https://somalogic.com/). Proteomics data preprocessing Protein abundances were quantified as relative fluorescence units (RFU). Normalization included the following steps: (i) hybridization normalization to reduce readout bias, (ii) intraplate median normalization to correct for within-plate variation, (iii) plate scaling for between-plate variation, (iv) calibration to minimize run-to-run variability, and (v) Adaptive Normalization by Maximum Likelihood (ANML) to adjust for inter-sample variability. Quality control (QC) was performed by comparing replicate signals against reference standards. RFU values were log10-transformed and standardized. Samples failing Somalogic-provided row-level QC (RowCheck flag; n = 2) and non-human aptamers (k = 295) were removed. Analytes were further filtered based on assay QC performance (QC CV < 0.1 and ≤2 plates with calibration QC ratios outside 0.8–1.2). The final dataset included 391 participants (231 with PTSD, 160 controls) and 9404 aptamer-based protein analytes targeting 8514 unique human proteins. To evaluate potential batch effects, we examined the distribution of PTSD cases and controls across assay plates and assessed sample clustering using principal component analysis (PCA) of normalized proteomic data. There was no significant association between plate and PTSD status (chi-square test p = 0.28), indicating comparable distribution of cases and controls across plates. Visualization of the first two principal components demonstrated no clustering by assay plate or PTSD status (Supplementary Fig. 1). We also examined inter-individual variability across all log10-transformed proteomic analytes prior to differential analyses. Variability was assessed using the standard deviation (SD) and median absolute deviation (MAD) of log10-transformed, unscaled data. The distribution of these measures across 9,404 aptamers indicated substantial variability, with no evidence of a substantial subset of near-invariant features (SD: median = 0.114, IQR = 0.087-0.156; range = 0.009–1.085; MAD: median = 0.067, IQR = 0.046–0.106; range = 0.008–0.730). Accordingly, no additional variability-based filtering was applied. Metabolomics profiling Metabolomic profiling was performed using five custom panels targeting 145 metabolites, including 5 conventional metabolites, 44 amino acids and biogenic amines, 21 ceramides, 30 sphingomyelins, and 45 acylcarnitines. These panels were chosen to capture direct and indirect pathways including fatty acid mobilization and oxidation, sphingolipid signaling, and amino acid-based neuroimmune pathways relevant to innate immunity, inflammation, and neuronal excitability identified from our previous work20. Acylcarnitines were quantified by flow injection tandem mass spectrometry (MS/MS) as previously described50, using a Waters Xevo TQD mass spectrometer coupled with an Acquity™ UPLC system. Amino acids and biogenic amines were measured by LC-MS/MS following an adapted method from51. Briefly, 10 µL of plasma was spiked with stable isotope-labeled standards, deproteinized with methanol, and derivatized with AccQTag reagent at 55 °C. Separation was performed on a Waters Acquity UPLC HSS T3 column using a formic acid-acetonitrile gradient, and analytes detected in positive ion mode via multiple reaction monitoring on a Xevo TQ-XS system. Ceramides and sphingomyelins were extracted as described previously52 and analyzed by MS in positive ionization mode. Conventional metabolites, including glucose, lactate, glycerol, non-esterified fatty acids (NEFA), and total ketone bodies, were quantified using a Beckman-Coulter DxC 600 clinical analyzer, as previously described53. Metabolomics data were available for 356 participants (213 with PTSD, 143 controls). For acylcarnitines, limited plasma availability yielded data for 333 participants (196 with PTSD, 137 controls). All metabolite values were log10-transformed. Inter-individual variability was assessed using the SD and MAD of log10-transformed metabolite levels. The distributions indicated adequate dispersion across participants (SD: median = 0.155, IQR = 0.118–0.257; range = 0.066–2.569; MAD: median = 0.139, IQR = 0.108–0.181; range = 0.065–1.053), with no evidence of near-invariant metabolites. Secondary metabolomics profiling included 10 additional tryptophan pathway metabolites, 3 branched-chain hydroxy acid (BCHA) metabolites, and 3 branched-chain keto acid (BCKA) metabolites in 276 participants (165 with PTSD, 111 controls). Tryptophan pathway metabolites were analyzed on a Waters Xevo TQ-XS triple quadrupole mass spectrometer (Milford, MA) coupled to a Waters Acquity UPLC system using a HSS T3 column (1.8 µm, 2.1 × 150 mm), as described previously54. BCHA metabolites were quantified by LC-MS/MS using a Waters Xevo TQ-XS system coupled to a Waters Acquity UPLC, as reported previously55. BCKA metabolites were analyzed by LC-MS/MS as previously described56, using isotopically labeled internal standards KIC-d3, KIV-5C13 (Cambridge Isotope Laboratories), and KMV-d8 (Toronto Research Chemicals). Differential proteomics and metabolomics analysis For primary discovery models, separate linear regression analyses were performed for each aptamer-based protein analyte and metabolite, using log10-transformed that were subsequently standardized (z-scored) analyte levels as the dependent variable and PTSD status as the independent variable, adjusting for age, gender, and race. To account for potential heteroskedasticity and influential observations, regression models were re-estimated using heteroskedasticity-consistent (HC3) standard errors. Robust Wald tests were implemented using the sandwich (v3.1.1) and lmtest (v0.9.40) packages in R. A false discovery rate (FDR) < 0.05 was used to identify significant proteins and metabolites. The area under the receiver operating characteristic curve (AUC) was also computed to assess discriminative power of each analyte. To evaluate potential confounding by environmental and lifestyle factors, primary models were additionally adjusted for BMI, smoking status, and WTC exposure. WTC exposure was modeled as an ordinal variable (low, intermediate, high, very high) to reflect increasing exposure intensity. Additional analyses were performed in a restricted subset of participants who were never or former smokers and had low or intermediate exposure levels. Sensitivity analyses further evaluated the robustness of findings to alternative model specifications. To assess potential nonlinear confounding, models incorporating quadratic terms (Age² and BMI²) and natural spline functions (df = 3) for age and BMI were examined. Additional models adjusted for medical comorbidities and medication use (NSAIDs, SSRIs, and statins), including analyses excluding NSAID users. Given the high comorbidity between PTSD and depressive symptoms, analyses were also repeated after restricting participants to those with minimal-to-mild depressive symptoms (PHQ < 10). Parallel sensitivity analyses were performed for the metabolomics data. Results were compared with findings from our prior proteomics studies17,18, our metabolomics study20, the PSYCHENCODE2 PTSD study utilizing UKBB plasma proteomics19 and a study of military personnel24. Among the 391 participants with proteomics data in the current study, 88 overlapped with our first prior study17 and 91 overlapped with our second prior study18. To evaluate the independence of the present findings, we conducted sensitivity analyses excluding all participants who overlapped with either prior study (n = 91 removed). Differential metabolomics analyses were also performed on the secondary metabolomics dataset, which included tryptophan-, BCHA-, and BCKA-related metabolites. Due to the smaller sample size (n = 276), this dataset was not included in primary integrative analyses. Pathway analysis Pathway enrichment was performed on differentially expressed proteins using GO terms57 and KEGG canonical pathways58. Integrative proteomics-metabolomics analysis We performed an integrative analysis by examining the correlations and potential synergy (i.e., interaction effects) between differentially expressed protein analytes and metabolites to further characterize coordinated molecular alterations. Specifically, Spearman rank correlation analyses and linear regression models were applied to test whether protein-metabolite relationships differed between individuals with PTSD and trauma-exposed controls. Potential moderation effects were evaluated using product interaction terms in the regression models (protein analyte × case status), adjusting for age, gender, and race. For protein-metabolite pairs demonstrating evidence of interaction at a nominal threshold (p < 0.01), exploratory stratified analyses were conducted within PTSD and control groups separately to estimate group-specific slopes. As in the differential expression analyses, statistical inference was based on heteroskedasticity-consistent (HC3) standard errors and robust Wald tests. Statistically significant protein-metabolite interactions were defined at a metabolite-wise FDR < 0.05. ICA was performed to identify modules of co-expressed protein analytes and metabolites59. Each module was then refined by retaining proteins with significant contribution scores (FDR < 0.05). Associations between ICA modules and PTSD were examined using linear regression models adjusted for age, gender, and race, with statistical significance defined at FDR < 0.05. Pathway enrichment analysis was conducted for significant modules using GO terms57. Organ aging Organ-specific proteomic ages were generated using a validated plasma-based organ age calculator for 11 organs (adipose, artery, brain, heart, immune system, intestine, kidney, liver, lung, muscle, and pancreas)25. Two sets of models were used: (i) organ-specific models trained on corresponding proteins and (ii) “CognitionX” models trained on proteins identified by the FIBA algorithm applied to CDRGLOB (global clinical dementia rating), incorporating both chronological age and cognitive decline. Additionally, three global measures were estimated: Conventional (all proteins), Organismal (organ non-specific proteins), and CognitionOrganismal (CDRGLOB-FIBA trained). Altogether, 25 age estimates were derived. Spearman rank correlations were computed between each predicted organ age and chronological age to evaluate age-tracking performance. For each estimate, an “age gap” was calculated as the difference between predicted and expected age relative to same-aged peers from the models’ training cohort. Associations between age gaps and PTSD were tested using linear regression, adjusted for chronological age, gender, and race, with statistical significance defined at FDR < 0.05 based on heteroskedasticity-consistent (HC3) standard errors and robust Wald tests. We also assessed overlap between our differentially expressed protein analytes and organ-specific protein sets used to train the organ aging models in ref. 25, and performed enrichment analyses using the full ranked list of protein analytes ranked by the product of −log p-value and effect size direction for each organ, with statistical significance defined at FDR < 0.05. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Supplementary Data 5 Reporting Summary Transparent Peer Review file Source data Source Data Acknowledgements We gratefully acknowledge the support of the first responders and rescue/recovery workers through their participation in this study. We also thank the staff of the Stony Brook World Trade Center Health Program and the World Trade Center Health Program Data Monitoring Center for ongoing support. The findings and conclusions presented in this article are those of the authors and do not represent the official position of NIOSH, the CDC, NIH or the U.S. Public Health Service. During manuscript preparation, the authors used a generative AI tool for language refinement and consultation of general background information during the literature review. All scientific interpretations, data analyses, and conclusions were developed solely by the research team, who take full responsibility for the content. Author contributions The authors confirm contribution to the paper as follows: Study conception, design, and funding acquisition: P.F.K., B.J.L. Plasma sample retrieval and coordination of the proteomic and metabolomic assays: X.Y. Metabolomics assay generation: O.I., M.M. Data analysis, primary interpretation of results, and drafting of the original manuscript: P.F.K. Proteomic and metabolomic interpretation: C.B.N. Critical revision of the manuscript for important intellectual content: F.D.M., X.Y., R.K., S.C., O.I., M.M., C.B.N., B.J.L. All authors reviewed the results and approved the final manuscript. Peer review Peer review information Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available. Funding The current study was supported in part by CDC/NIOSH, NIH/NIA and NIH/NIDDK awards: CDC-75D30122c15522 (PI: B.J.L.), U01 OH012466 (PI: P.F.K.), R21 AG081480 (PI: P.F.K.), U01 OH012257 (MPI: P.F.K., S.C.), P30 DK124723 (PI: C.B.N.) and the SUNY Research Foundation. Data availability Summary statistics from association analyses of all analyzed proteins and metabolites have been deposited in the Open Science Framework (https://osf.io/z5hse). Individual-level clinical and omics data cannot be publicly shared in order to protect the privacy and confidentiality of World Trade Center responders and due to ethical restrictions governing human subjects research. De-identified individual-level data may be made available to qualified researchers through the Stony Brook World Trade Center Health and Wellness Program by contacting Dr. Benjamin Luft (Benjamin.Luft@stonybrookmedicine.edu). Access is subject to Institutional Review Board approval, applicable data use agreements, and institutional policies governing human subjects research. Source data are provided with this paper. Code availability Analyses were conducted in R (v4.4.2). The analysis code has been deposited in the Open Science Framework (https://osf.io/z5hse). Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Pei-Fen Kuan, Email: PeiFen.Kuan@stonybrook.edu. Benjamin J. Luft, Email: Benjamin.Luft@stonybrookmedicine.edu Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-74757-8. References Associated Data Supplementary Materials Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Supplementary Data 5 Reporting Summary Transparent Peer Review file Source Data Data Availability Statement Summary statistics from association analyses of all analyzed proteins and metabolites have been deposited in the Open Science Framework (https://osf.io/z5hse). Individual-level clinical and omics data cannot be publicly shared in order to protect the privacy and confidentiality of World Trade Center responders and due to ethical restrictions governing human subjects research. De-identified individual-level data may be made available to qualified researchers through the Stony Brook World Trade Center Health and Wellness Program by contacting Dr. Benjamin Luft (Benjamin.Luft@stonybrookmedicine.edu). Access is subject to Institutional Review Board approval, applicable data use agreements, and institutional policies governing human subjects research. Source data are provided with this paper. Analyses were conducted in R (v4.4.2). The analysis code has been deposited in the Open Science Framework (https://osf.io/z5hse).