Introduction
Individuals exposed to adversity, such as trauma or stressful life events, have been reported to be at increased risk of psychiatric disorders and various subsequent somatic diseases (Ahmadi et al., Reference Ahmadi, Hajsadeghi, Mirshkarlo, Budoff, Yehuda and Ebrahimi2011; Sareen et al., Reference Sareen, Cox, Stein, Afifi, Fleet and Asmundson2007). However, although some individuals experience persistent or worsening psychological symptoms following trauma, others exhibit minimal symptoms or merely temporary distress that quickly resolves―a phenomenon known as psychological resilience (Galatzer-Levy, Huang, & Bonanno, Reference Galatzer-Levy, Huang and Bonanno2018; Kalisch et al., Reference Kalisch, Baker, Basten, Boks, Bonanno, Brummelman, Chmitorz, Fernàndez, Fiebach, Galatzer-Levy, Geuze, Groppa, Helmreich, Hendler, Hermans, Jovanovic, Kubiak, Lieb, Lutz and Kleim2017; Rutter, Reference Rutter2006). Because psychological resilience is conceptualized as a measure of an individual’s ability to effectively adapt to and recover from adversity or stress (Choi et al., Reference Choi, Chen, Ursano, Sun, Jain, Kessler, Koenen, Wang, Wynn, Campbell-Sills, Stein and Smoller2020; Vieira et al., Reference Vieira, Pedrotti Moreira, Mondin, Cardoso, Branco, Kapczinski, Jansen, Souza and da Silva2020), gaining a better understanding of its underlying biological mechanisms may help identify novel therapeutic targets for preventing psychological abnormalities following trauma.
Nevertheless, effectively measuring psychological resilience presents significant challenges. Prior research has primarily characterized resilience through self-rated questionnaires (e.g. the Connor-Davidson Resilience Scale) (Connor & Davidson, Reference Connor and Davidson2003), by assessing trait-like capacities (e.g. self-reported well-being) before the occurrence of adversity (Choi et al., Reference Choi, Stein, Dunn, Koenen and Smoller2019), or by the absence of psychiatric symptoms or disorders following adversity (Galatzer-Levy et al., Reference Galatzer-Levy, Huang and Bonanno2018; Niitsu et al., Reference Niitsu, Houfek, Barron, Stoltenberg, Kupzyk and Rice2017). However, these approaches are limited by their hypothetical nature (assessed without experience of real trauma), lack of objective validation, and failure to account for other relevant factors, such as psychological vulnerabilities, and the characteristics of adversity itself (Troy et al., Reference Troy, Willroth, Shallcross, Giuliani, Gross and Mauss2023), leading to inconsistent results from prior phenotypic studies. For instance, while several studies have supported the protective role of resilience in preventing psychiatric disorders (Castiglioni, Caldiroli, Manzoni, & Procaccia, Reference Castiglioni, Caldiroli, Manzoni and Procaccia2023; Navrady et al., Reference Navrady, Adams, Chan, Ritchie and McIntosh2018), the ability of self-rated resilience to predict psychiatric disorders after trauma has been shown to be somewhat constrained (Campbell-Sills et al., Reference Campbell-Sills, Kessler, Ursano, Sun, Taylor, Heeringa, Nock, Sampson, Jain and Stein2018).
Similarly, research on genetic determinants of psychological resilience has produced mixed findings. Twin studies have indicated that 25%–77% of variance in self-assessed resilience can be attributed to genetics (Amstadter, Myers, & Kendler, Reference Amstadter, Myers and Kendler2014; Waaktaar & Torgersen, Reference Waaktaar and Torgersen2012; Wolf et al., Reference Wolf, Miller, Sullivan, Amstadter, Mitchell, Goldberg and Magruder2018). Albeit, only a limited number of specific genetic markers have been identified based on studies with varying phenotype definitions, making them difficult to validate in external populations (Baselmans et al., Reference Baselmans, Jansen, Ip, van Dongen, Abdellaoui, van de Weijer, Bao, Smart, Kumari, Willemsen, Hottenga, Boomsma, de Geus, Nivard and Bartels2019; Feder, Nestler, & Charney, Reference Feder, Nestler and Charney2009; Maul et al., Reference Maul, Giegling, Fabbri, Corponi, Serretti and Rujescu2020; Mazzarotto et al., Reference Mazzarotto, Monteleone, Minelli, Mattevi, Cascino, Rocca, Rossi, Bertolino, Aguglia, Altamura, Amore, Bellomo, Bucci, Collantoni, Dell’Osso, Di Fabio, Fagiolini, Giuliani and Marchesi2024; Okbay et al., Reference Okbay, Baselmans, De Neve, Turley, Nivard, Fontana, Meddens, Linner, Rietveld, Derringer, Gratten, Lee, Liu, de Vlaming, Ahluwalia, Buchwald, Cavadino, Frazier-Wood, Furlotte and Cesarini2016; Qi & Chatterjee, Reference Qi and Chatterjee2018; Stein et al., Reference Stein, Choi, Jain, Campbell-Sills, Chen, Gelernter, He, Heeringa, Maihofer, Nievergelt, Nock, Ripke, Sun, Kessler, Smoller and Ursano2019). Recently, a residual approach has been employed to quantify continuous metrics of resilience in genetic studies, allowing for the consideration of multiple environmental factors (Dumitrescu et al., Reference Dumitrescu, Mahoney, Mukherjee, Lee, Bush, Engelman, Lu, Fardo, Trittschuh, Mez, Kaczorowski, Hernandez Saucedo, Widaman, Buckley, Properzi, Mormino, Yang, Harrison, Hedden and Hohman2020; Mukherjee et al., Reference Mukherjee, Kim, Gibbons, Nho, Risacher, Glymour, Habeck, Lee, Mormino, Ertekin-Taner, Montine, Decarli, Saykin and Crane2012; Ossenkoppele et al., Reference Ossenkoppele, Lyoo, Jester-Broms, Sudre, Cho, Ryu, Choi, Smith, Strandberg, Palmqvist, Kramer, Boxer, Gorno-Tempini, Miller, la Joie, Rabinovici and Hansson2020). In this framework, resilience is quantified as the discrepancy between observed and predicted psychological symptoms, where predictions are derived from models incorporating relevant risk factors. Individuals exhibiting better-than-expected symptoms are interpreted as having higher resilience. This approach yields a psychometrically grounded indicator that is more heritable and more closely aligned with underlying pathophysiological processes than conventional measures. However, most of these studies relied on small sample sizes (Mukherjee et al., Reference Mukherjee, Kim, Gibbons, Nho, Risacher, Glymour, Habeck, Lee, Mormino, Ertekin-Taner, Montine, Decarli, Saykin and Crane2012; Ossenkoppele et al., Reference Ossenkoppele, Lyoo, Jester-Broms, Sudre, Cho, Ryu, Choi, Smith, Strandberg, Palmqvist, Kramer, Boxer, Gorno-Tempini, Miller, la Joie, Rabinovici and Hansson2020), cross-sectional data (Mukherjee et al., Reference Mukherjee, Kim, Gibbons, Nho, Risacher, Glymour, Habeck, Lee, Mormino, Ertekin-Taner, Montine, Decarli, Saykin and Crane2012), or data from a specific population or ancestry (Dumitrescu et al., Reference Dumitrescu, Mahoney, Mukherjee, Lee, Bush, Engelman, Lu, Fardo, Trittschuh, Mez, Kaczorowski, Hernandez Saucedo, Widaman, Buckley, Properzi, Mormino, Yang, Harrison, Hedden and Hohman2020; Mukherjee et al., Reference Mukherjee, Kim, Gibbons, Nho, Risacher, Glymour, Habeck, Lee, Mormino, Ertekin-Taner, Montine, Decarli, Saykin and Crane2012; Ossenkoppele et al., Reference Ossenkoppele, Lyoo, Jester-Broms, Sudre, Cho, Ryu, Choi, Smith, Strandberg, Palmqvist, Kramer, Boxer, Gorno-Tempini, Miller, la Joie, Rabinovici and Hansson2020), leading to findings that require replication.
To this end, leveraging data from the China Surgery and Anesthesia Cohort (CSAC), which provides comprehensive information on sociodemographic, environmental, trauma-related features, along with psychological status that was measured repeatedly at baseline and follow-ups, and individual-level genotyping data for surgical patients, we aimed to identify participants exhibiting post-trauma psychological resilience using the residual approach (i.e. a better-than-expected depressive symptom) and uncover its associated genetic components. With data from an independent cohort (i.e. the China Severe Trauma Cohort [CSTC]) for further genotype stratification and additive model analyses, findings from this research could potentially contribute to the early identification of individuals likely to exhibit psychological resilience or non-resilience to adversity, offer new insights into the biological mechanisms underlying resilience, and ultimately enhance risk assessment and management for traumatized individuals.
Methods
This study was structured into three parts (Figure 1). In the first part, we conducted phenotypic analyses utilizing longitudinal data from an exploration cohort (n = 12,946) constructed based on the CSAC dataset. We established models to predict depressive symptoms at 1, 6, and 12 months post-surgery, where the ranked model residuals were used to identify individuals displaying psychological resilience. In the second part, our focus was narrowed to participants with genotyping data available after performing standard GWAS quality control in the CSAC (n = 7,299), where we performed GWAS analyses to identify candidate genetic signals linked to psychological resilience. In the third part, we sought to examine the effects of the identified genetic variants on psychological trajectories using data from an independent trauma cohort that recruited 2,285 patients who experienced severe physical trauma within 14 days.
The study flowchart. Note: PHQ-9, ‘9-item Patient Health Questionnaire’.
Figure 1. Long description
The flowchart is divided into two main columns. The left column, titled Selection of study populations, shows the participant flow. The right column, titled The analytic steps, aligns with the stages of the study.
1. Top Section: China Surgery and Anesthesia Cohort. 14,046 participants recruited between July 15, 2020, and December 24, 2023, who underwent noncardiac surgery.
* An arrow points down to an exclusion list: 14 withdrew, 162 received more than one surgery, 283 with a history of psychiatric disorder, and 641 without any data on P H Q-9 during follow-up.
* This leads to 12,946 participants: 2,328 with psychological resilience and 10,618 without psychological resilience.
* Corresponding Analytic Step: Part 1. Phenotypic analyses, including identifying psychological resilience using the residual approach and descriptive analyses.
2. Middle Section: An arrow points down from the 12,946 participants to another exclusion list: 6,653 without genotype data and 4 did not keep through quality control.
* This leads to 7,299 participants with qualified genetic data: 1,213 with psychological resilience and 6,086 without psychological resilience.
* Corresponding Analytic Step: Part 2. Genetic analyses, including Genome-wide association study analyses of psychological resilience.
3. Bottom Section: China Severe Trauma Cohort. 2,285 participants experienced trauma within 14 days and recruited between March 1, 2020, and January 12, 2024.
* Corresponding Analytic Step: Part 3. Analyses of independent cohort, focusing on the effect of identified genetic variants on post-trauma psychological trajectories.
Study design
Cohort study for psychological resilience ascertainment
Cohort construction
The CSAC is an ongoing prospective cohort study recruiting patients (aged 40–65 years) who underwent elective surgery and general anesthesia in 4 medical centers in China (3 hospitals in Sichuan and 1 in Hebei Province) since July 2020. Information on sociodemographic and environmental factors was collected at baseline (i.e. 1 day before the surgery) by trained data collectors. Participants were also invited to donate a blood sample. We also obtained surgery and anesthesia-related conditions through linkage with the anesthesia information management system (AIMS) and electronic medical record (EMR) system. Psychological status and physical well-being were mainly assessed using self-rated scales or questionnaires at baseline and at 1, 6, and 12 months post-surgery. Details of this cohort have been described elsewhere (Yang et al., Reference Yang, Chen, Chen, He, Wang, Qu, Yang, Tang, Zeng, Deng, Liu, Huang, Li, du, Liu, Li and Song2024).
In the present study, we included 14,046 participants who underwent noncardiac surgery between July 15, 2020, and December 24, 2023. We then excluded participants who withdrew (n = 14), had undergone more than one surgery (n = 162), had a history of psychiatric disorder (n = 283), and without any data on 9-item Patient Health Questionnaire (PHQ-9) during follow-ups (n = 641), leaving 12,946 participants for the phenotypic analyses (Supplementary eTable 1).
Ascertainment of psychological resilience
In brief, we applied the residual approach to ascertain individuals with psychological resilience according to the presence of better-than-expected depressive symptoms post-surgery. Particularly, as we had repeated measurements of depressive symptoms (i.e. by PHQ-9 scores measured at 1, 6, and 12 months post-surgery), we first established and validated a linear mixed-effects model to predict individuals’ PHQ-9 score at each post-surgery point (Bates, Mächler, Bolker, & Walker, Reference Bates, Mächler, Bolker and Walker2015). To ensure that residuals reflect genuine resilience rather than confounding, we included a comprehensive set of 23 variables (sociodemographic and environmental factors, surgery- and anesthesia-related factors, as well as baseline PHQ-9 score) to adjust for confounders (Supplementary eAppendix 1; Supplementary eTables 2–3). Then, we ranked residuals (i.e. actual − predicted PHQ-9 score) derived from the fitted model and defined ‘better-than-expected’, ‘comparable’, and ‘worse-than-expected’ statuses by lowest 25th, 26–74th, and top 25th percentiles of residuals, respectively, at each follow-up point. Individuals with ‘better-than-expected’ status at ≥1 follow-up point and without ‘worse-than-expected’ status at any follow-up point were considered to have psychological resilience, while the others were considered as controls (i.e., not resilient).
Case–control study for identifying genetic determinants
Among the 12,946 CSAC participants, 7,299 had qualified genotyping data (including 6,683,986 single-nucleotide polymorphisms [SNPs] after quality control and imputation; Supplementary eAppendix 2). According to the ascertainment approach mentioned above, we included 1,213 cases (having psychological resilience) and 6,086 controls (not psychological resilience) in the genetic analyses.
Independent cohort for examining the effects of identified genetic variants on post-trauma psychological trajectories
We included an independent trauma cohort (i.e. the CSTC) to study the effect of identified genetic variants on post-trauma psychological trajectories. In brief, the CSTC is a prospective cohort study that recruited 4,122 patients (aged 12–80 years) who were admitted to the Trauma Center of West China Hospital for severe physical trauma experienced within 14 days between March 2020–January 12, 2024. Data on sociodemographic and environmental factors, as well as trauma-related conditions (e.g. time passed since injury, injury cause), were collected at recruitment. Psychological status (i.e. depressive symptoms, anxiety symptoms, and sleep disturbance) and physical well-being were assessed at baseline and at 1, 3, 6, and 12 months after recruitment. After exclusion of participants who withdrew (n = 47), had a history of psychiatric disorder (n = 170), had no PHQ-9 data during follow-ups (n = 195), and had no qualified genotyping data (n = 1,425; Supplementary eAppendix 2), a total of 2,285 participants were included in this analysis.
This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines (von Elm et al., Reference von Elm, Altman, Egger, Pocock, Gøtzsche and Vandenbroucke2007).
Statistical analyses
Descriptive analyses of the CSAC cohort
To aid interpretation of the identified resilience phenotype, as well as to test if the residual approach can validly capture resilient individuals, we first plotted the changes of multiple psychological and physical conditions post-surgery among CSAC participants with different statuses of psychological resilience. Studied conditions included post-surgery depressive symptoms (using PHQ-9 scores or PHQ-9 ≥ 5 for the presence [yes/no]), anxiety symptoms (using scores of the 7-item Generalized Anxiety Disorder [GAD-7] or GAD-7 ≥ 5, Zhang et al., Reference Zhang, Wang, Zeng, Zhao, Zhang, Zhai, Meng, Wang and Liu2021), sleep disturbance (using scores of the 19-item Pittsburgh Sleep Quality Index [PSQI], or PSQI > 10, Ho & Fong, Reference Ho and Fong2014), physical recovery (on a self-rated scale of 0–10, recoding of the data [10-score] was performed to make sure higher values representing poorer recovery condition) and life satisfaction (answers ‘very satisfied’, ‘satisfied’, ‘average’, ‘dissatisfied’ or ‘very dissatisfied’ to the question: ‘are you basically satisfied with your current life?’).
Genetic analyses of the case–control study to identify associated genetic determinants
Restricting to CSAC participants with qualified genetic data, we performed GWAS analyses to identify genetic variants associated with psychological resilience, where estimates were obtained from logistic regression models adjusted for age, sex, batch, and the top five principal components, using PLINK version 2.00 (Purcell et al., Reference Purcell, Neale, Todd-Brown, Thomas, Ferreira, Bender, Maller, Sklar, de Bakker, Daly and Sham2007). These analyses were repeated after additionally adjusting for baseline GAD-7 and PSQI scores and polygenic risk scores (PRS) for depression (Supplementary eTable 1), given that pre-surgery anxiety, sleep disturbance, and individuals’ genetic predisposition to depression could also contribute to the post-surgery depressive symptoms. In addition, to test the robustness of results to the selected thresholds for caseness (i.e. lowest 25th percentiles of residuals used for defining ‘better-than-expected’ status), we repeated the GWAS analyses using alternative thresholds (the lowest 20th, or the lowest 30th percentiles). We also performed subgroup-specific GWAS analyses stratified by site of surgery, to examine whether the identified SNPs remained associated with the resilience phenotype across these different trauma contexts.
SNPs were considered genome-wide significant if their P values were < 5 × 10−8. The significant SNPs, together with surrounding genomic loci identified based on linkage disequilibrium (LD) structure (i.e. r 2 ≥ 0.6), were further gene-mapped and functionally annotated on the web-based platform FUMA (http://fuma.ctglab.nl/) (Watanabe, Taskesen, van Bochoven, & Posthuma, Reference Watanabe, Taskesen, Van Bochoven and Posthuma2017). Then, mapped genes were then included in the gene set enrichment analysis to identify potential biological pathways using the web-based tool Metascape (Zhou et al., Reference Zhou, Zhou, Pache, Chang, Khodabakhshi, Tanaseichuk, Benner and Chanda2019).
Analyses of the CSTC to test the effect of identified genetic variants on post-trauma psychological trajectories
Based on the independent cohort of CSTC, we examined the associations between the lead SNP (rs11766511) and post-trauma psychological trajectories (i.e. the changes of post-trauma depressive symptoms, anxiety symptoms, and sleep disturbance) and physical conditions. We performed genotype-stratified analyses treating rs11766511 as a categorical variable (AA, GA, GG), with AA genotype as the reference. We also conducted additive genetic model analyses, coding the number of effect alleles (G) as 0, 1, or 2, to estimate additive effects. For scale scores as continuous variables, we used Beta with 95%CI derived from linear mixed-effect models for association assessment, while for the presence of post-trauma psychological symptoms (using the same measuring tools and cut-off criteria as the CSAC) as binary variables, the associations were measured by ORs with 95%CI derived from generalized linear mixed-effect models. All models were adjusted for age, sex, batch, and the top five genetic principal components (Supplementary eTable 4).
All data analyses were completed using R (Version 4.2.1), PLINK (Version 2.0), and GCTB (Version 2.0). A two-tailed test with P < 0.05 was considered statistically significant.
Results
Post-surgery psychological trajectories and characteristics among individuals with different statuses of psychological resilience in the exploration cohort
Of 12,946 eligible participants of CSAC, the mean age (standard deviation) was 52.2 (7.1) years, with a female dominance (7,569 [58.5%]) (Supplementary eTable 5). According to the ascertainment approach stated above, we identified 2,328 (18.0%) patients with psychological resilience. Briefly, when examining the changes in multiple psychological and physical conditions over the post-surgery period based on psychological resilience status, we consistently observed better performance among individuals with psychological resilience, compared to those without (Figure 2; Supplementary eFigure 1). For example, resilient participants had lower GAD-7 scores (e.g. 0.22 vs. 0.76 at 1 month post-surgery, and 0.17 vs. 0.72 at 12 months post-surgery) and smaller proportions of anxiety symptoms (e.g. 0.7% vs. 5.2% at 1 month post-surgery, and 0.6% vs. 5.0% at 12 months post-surgery) at all follow-up time points, compared to non-resilient participants (Figure 2c and d). Likewise, physical recovery was slightly better among resilient patients, compared to non-resilient counterparts (e.g. mean score = 2.69 vs. 2.78 at 1 month post-surgery, and 1.49 vs. 1.60 at 12 months post-surgery; Supplementary eFigure 1A).
Changes of psychological status during the post-surgery period among 12,946 participants in the China Surgery and Anesthesia Cohort, by psychological resilience status. PHQ-9 scores and depressive symptoms are presented as definitional outcomes used to derive the residual-based resilience phenotype. Anxiety symptoms and sleep disturbance are presented to aid interpretation of the identified phenotype and to examine whether resilient participants also showed more favorable psychological profiles after surgery. Note: PHQ-9, ‘9-item Patient Health Questionnaire’; GAD-7, ‘7-item Generalized Anxiety Disorder’; PSQI, ‘19-item Pittsburgh Sleep Quality Index’.
Figure 2. Long description
A multi-panel figure with six charts organized into three rows. A legend at the top indicates red for non-resilient participants and blue for resilient participants.
Row 1: Depressive symptoms.
* Panel A: Bar chart of P H Q 9 mean scores. Non-resilient scores are 2.16 at 1m, 1.78 at 6m, and 1.84 at 12m. Resilient scores are significantly lower at 0.57, 0.40, and 0.49 respectively.
* Panel B: Line graph of depressive symptom percentages. Non-resilient group starts at 14.0 percent, dips to 11.6 percent at 6m, and ends at 12.4 percent. Resilient group remains near 0.0 percent across all time points.
Row 2: Anxiety symptoms.
* Panel C: Bar chart of G A D 7 mean scores. Non-resilient scores remain stable around 0.72 to 0.78. Resilient scores are lower, ranging from 0.17 to 0.22.
* Panel D: Line graph of anxiety symptom percentages. Non-resilient group fluctuates between 5.0 percent and 5.9 percent. Resilient group stays low between 0.6 percent and 0.7 percent.
Row 3: Sleep disturbance.
* Panel E: Bar chart of P S Q I mean scores. Non-resilient scores are 6.44 at 1m, 5.79 at 6m, and 5.91 at 12m. Resilient scores are 5.24, 4.43, and 4.25 respectively.
* Panel F: Line graph of sleep disturbance percentages. Non-resilient group starts at 15.7 percent and ends at 13.6 percent. Resilient group shows a steady decline from 4.8 percent at 1m to 1.4 percent at 12m.
Regarding basic characteristics, we found individuals with and without psychological resilience were largely comparable with regard to most sociodemographic, environmental, and surgery- and anesthesia-related factors, while resilient patients were more likely to be female (66.4% vs. 56.7%) and to have a higher likelihood of experiencing preoperative pain (11.6% vs. 8.3%) or receiving an intraoperative blood transfusion (6.4% vs. 2.6%) compared to non-resilient patients (Supplementary eTable 6).
Genetic determinants for psychological resilience
The GWAS analyses revealed one lead SNP significantly associated with psychological resilience (rs11766511, with the G allele as the effect allele; OR = 1.477 [95%CI 1.285–1.698, P = 3.082 × 10−8), which is located on chromosome 7p22.2 (Figure 3). The lead SNP together with 17 highly correlated SNPs based on LD structure (i.e. r 2 ≥ 0.6; Table 1) were mapped to one gene sidekick cell adhesion molecule 1 (SDK1, gene ID 221935), with reported traits including well-being indicators (e.g. positive affect and life satisfaction), and psychological and neurological diseases (e.g. insomnia and depression, Alzheimer’s disease, autism spectrum disorder) (Supplementary eTables 7 and 8). The most related biological process of SDK1 was response to cocaine (Gene Ontology [GO]: 0048148, GO: 0042220).
Results of genome-wide association study for identified psychological resilience based on the China Surgery and Anesthesia Cohort. Note: GO, ‘Gene Ontology’. (a) Manhattan plot, the x-axis is chromosomal position, and the y-axis is the significance on a −log10 scale, the upper dashed line marks the threshold for genome-wide significance (P < 5 × 10−8); (b) quantile–quantile plot, indicating no noticeable systematic biases; (c) regional plot, one significant single-nucleotide polymorphisms (rs11766511) located on chromosome 7p22.2, together with surrounding genomic loci that identified based on linkage disequilibrium structure (i.e. r 2 ≥ 0.6), were further gene-mapped to one gene sidekick cell adhesion molecule 1 (SDK1); (d) the short description for identical genes were annotated using the Metascape tool.
Figure 3. Long description
The figure consists of four panels labeled A through D.
Panel A is a Manhattan plot showing the association of genetic variants across chromosomes 1 to 22 on the x-axis against negative log 10 P-values on the y-axis. A horizontal dashed red line indicates the genome-wide significance threshold. A single peak on chromosome 7 is labeled r s 1 1 7 6 6 5 1 1, which surpasses the significance threshold.
Panel B is a Q Q plot comparing expected negative log 10 P-values on the x-axis to observed negative log 10 P-values on the y-axis. The data points follow the diagonal identity line closely until the tail, where they deviate upward, indicating significant associations.
Panel C is a regional association plot focused on chromosome 7. The x-axis shows the genomic position from 4,140,000 to 4,190,000. The y-axis shows negative log 10 P-values. The lead S N P r s 1 1 7 6 6 5 1 1 is highlighted in purple. Other S N P s are colored by their linkage disequilibrium r-squared values relative to the lead S N P, ranging from red (high) to blue (low). Below the plot, a gene track shows the S D K 1 gene locus.
Panel D is a table titled Mapped gene. It contains four rows:
* Gene I D: 221935.
* Gene Symbol: S D K 1.
* Description: Sidekick cell adhesion molecule 1.
* Biological Process (G O): Response to cocaine (G O 0048148, G O 0042220).
The candidate single-nucleotide polymorphisms (SNPs) associated with psychological resilience
Table 1. Long description
The table contains 11 columns: S N P s, Chromosome, B P, Mapped gene, Non-effect allele, Effect allele, M A F, r super 2, Function, O R 95% C I, and P value. All 18 S N P s are located on Chromosome 7 and mapped to the S D K 1 gene.
Key data points include:
* rs11766511: B P 4173613, Non-effect A, Effect G, M A F 0.103, r super 2 1.000, intronic, O R 1.477, P value 3.082 times 10 super minus 8.
* rs13224870: B P 4172006, Non-effect C, Effect T, M A F 0.099, r super 2 0.939, exonic synonymous S N V exon 28, O R 1.471, P value 7.145 times 10 super minus 8.
* rs13224986: B P 4172109, Non-effect C, Effect T, M A F 0.099, r super 2 0.939, intronic, O R 1.469, P value 7.970 times 10 super minus 8.
* rs13242365: B P 4175025, Non-effect G, Effect T, M A F 0.108, r super 2 0.894, intronic, O R 1.404, P value 1.112 times 10 super minus 6.
* rs1077472: B P 4175323, Non-effect G, Effect A, M A F 0.108, r super 2 0.894, intronic, O R 1.401, P value 1.336 times 10 super minus 6.
* rs12701426: B P 4176315, Non-effect G, Effect A, M A F 0.119, r super 2 0.719, intronic, O R 1.463, P value 1.752 times 10 super minus 7.
* rs12701427: B P 4176472, Non-effect G, Effect A, M A F 0.123, r super 2 0.629, intronic, O R 1.443, P value 1.546 times 10 super minus 7.
All remaining S N P s (rs1880856, rs588123, rs654384, rs612476, rs72569241, rs651245, rs13223924, rs12701405, rs60440642, rs8180800, and rs4723396) are intronic with r super 2 values ranging from 0.692 to 0.796 and P values between 10 super minus 4 and 10 super minus 7.
Note: SNP, single-nucleotide polymorphism; BP, genomic position in the human genome assembly GRCh37 (hg19); MAF, minor allele frequency based on 1000G; OR, odds ratio; CI, confidence interval; SDK1, sidekick cell adhesion molecule 1.
a ORs were derived from a logistic regression in PLINK version 2.00, adjusted for age, sex, genotyping batch, and the first five principal components.
The results of GWAS analyses additionally adjusted for baseline GAD-7 scores, PSQI scores, and PRS for depression corroborated the main finding by pinpointing the same SNP as the top signal (i.e. for rs11766511, fully-adjusted model: OR = 1.486, P = 2.952 × 10−8; Supplementary eTable 9). In the sensitivity analysis where the modified threshold was applied for defining caseness, no genome-wide significant variants were identified at P < 5 × 10−8 but the identified variant (i.e. rs11766511) in the main analysis was considered suggestively significant (the lowest 20th percentile: OR = 1.316, P = 5.869 × 10−7; the lowest 30th percentile: OR = 1.253, P = 0.001; Supplementary eTable 10). In stratified analyses by site of surgery (Supplementary eTable 11), all candidate SNPs showed consistent effect directions across the four subgroups, and several identified variants in the main analysis demonstrated nominally significant associations in multiple subgroups (P < 0.05).
The effect of rs11766511 on post-trauma psychological trajectories in the independent trauma cohort
When analyzing the changes in post-traumatic psychological status based on rs11766511 genotypes, we observed better performance among individuals with GG genotype, compared to AA genotype and GA genotype (Supplementary eTable 12, Figure 4a–f). Further genotype-stratified analyses suggested that compared with AA carriers, GG carriers had lower PHQ-9 scores (Beta = −0.94 [−1.76−−0.12]) and PSQI scores (Beta = −1.02 [−1.94 to −0.10]), as well as lower odds of sleep disturbance (OR = 0.12 [0.02–0.68]). However, the GA genotype was not associated with post-trauma psychological trajectories (Figure 4g). In addition, patients with rs11766511 GG genotype tended to have better self-reported physical recovery condition (Supplementary eFigure 2). Under the additive genetic model, the G allele showed a directionally consistent but non-significant protective pattern for post-trauma psychological symptoms and physical recovery (Supplementary eTable 13).
The effect of rs11766511 genotypes on post-trauma psychological trajectories among 2,285 participants in the China Severe Trauma Cohort. Note: PHQ-9, ‘9-item Patient Health Questionnaire’; GAD-7, ‘7-item Generalized Anxiety Disorder’; PSQI, ‘19-item Pittsburgh Sleep Quality Index’; OR, ‘odds ratios’; CI, ‘confidence interval’. aBeta coefficients and 95% CIs were estimated using linear mixed-effects models; odds ratio (ORs) and 95% CIs were estimated using generalized linear mixed-effects models. Models were adjusted for age, sex, batch, and the top five principal components.
Figure 4. Long description
The figure is organized into four horizontal sections.
1. Depressive symptoms: Panel A is a bar chart of P H Q-9 mean scores at 1, 3, 6, and 12 months. A A genotype (red) shows the highest scores, followed by G A (green) and G G (blue). Panel B is a line graph showing the percentage of participants with symptoms, where A A starts at 17.1% and ends at 13.7%, while G G starts at 3.0% and rises to 10.3%.
2. Anxiety symptoms: Panel C shows G A D-7 mean scores. A A and G A genotypes maintain higher scores (above 1.2) compared to G G (below 0.9). Panel D shows percentages, with A A decreasing from 13.9% to 9.1% and G G remaining lowest, ending at 3.4%.
3. Sleep disturbance: Panel E shows P S Q I mean scores. All genotypes show a downward trend over 12 months, with A A consistently highest. Panel F shows percentages where A A drops from 14.61% to 8.75%, and G G fluctuates between 3.13% and 3.45%.
4. The effect of rs 1 1 7 6 6 5 1 1 genotypes on post-trauma psychological trajectories: Panel G is a forest plot table. It lists Beta values and Odds Ratios (O R) with 95% Confidence Intervals (C I). For all three symptom categories, the G G genotype shows a significant negative association compared to the A A reference, with Beta values of -0.94 for depression, -0.76 for anxiety, and -1.02 for sleep disturbance. The O R for G G genotype in sleep disturbance is 0.12 (0.02 to 0.68), indicating a protective effect.
Discussion
Utilizing longitudinal data on depressive symptoms assessed before and after surgery in a large cohort of surgical patients, we applied a residual-based approach to identify individuals who exhibited better-than-expected psychological outcomes –s defined as psychological resilience. Our GWAS analyses revealed a significant genetic locus, rs11766511, mapped to the SDK1 gene on chromosome 7p22.2, which was associated with resilience. In an independent cohort of 2,285 traumatic patients, the GG genotype of rs11766511 showed a protective effect against multiple post-trauma psychological symptoms, compared to the AA genotype; However, an additive genetic model showed a directionally consistent but non-significant protective pattern of the G allele, requiring confirmation in larger samples.
Prior studies struggled to reach a consensus on the definition of psychological resilience, resulting in inconsistent findings regarding its genetic basis (Baselmans et al., Reference Baselmans, Jansen, Ip, van Dongen, Abdellaoui, van de Weijer, Bao, Smart, Kumari, Willemsen, Hottenga, Boomsma, de Geus, Nivard and Bartels2019; Feder et al., Reference Feder, Nestler and Charney2009; Maul et al., Reference Maul, Giegling, Fabbri, Corponi, Serretti and Rujescu2020; Mazzarotto et al., Reference Mazzarotto, Monteleone, Minelli, Mattevi, Cascino, Rocca, Rossi, Bertolino, Aguglia, Altamura, Amore, Bellomo, Bucci, Collantoni, Dell’Osso, Di Fabio, Fagiolini, Giuliani and Marchesi2024; Okbay et al., Reference Okbay, Baselmans, De Neve, Turley, Nivard, Fontana, Meddens, Linner, Rietveld, Derringer, Gratten, Lee, Liu, de Vlaming, Ahluwalia, Buchwald, Cavadino, Frazier-Wood, Furlotte and Cesarini2016; Qi & Chatterjee, Reference Qi and Chatterjee2018; Stein et al., Reference Stein, Choi, Jain, Campbell-Sills, Chen, Gelernter, He, Heeringa, Maihofer, Nievergelt, Nock, Ripke, Sun, Kessler, Smoller and Ursano2019). For instance, a PRS derived from the resilience GWAS was not significantly associated with resilience assessed by the absence of psychiatric symptoms or disorders (Stein et al., Reference Stein, Choi, Jain, Campbell-Sills, Chen, Gelernter, He, Heeringa, Maihofer, Nievergelt, Nock, Ripke, Sun, Kessler, Smoller and Ursano2019). In addition to the use of data from a large, well-characterized surgical patient cohort, our study is distinctive in its use of a residual-based method for ascertaining resilience, allowing for the integration of diverse factors (sociodemographic and environmental influences, adversity characteristics) into the analysis. By focusing on individuals who exhibited better-than-expected outcomes, we were able to identify a more objective and quantifiable measure of resilience, while ensuring that the identified genetic associations are robust. Indeed, this notion gains support from further analyses, where the reliability of identified candidate genetic signals for resilience was further demonstrated in an independent cohort of patients who experienced physical trauma. Such a phenomenon underscores that resilience to varied traumatic experiences may have shared biological underpinnings.
As a candidate locus requiring further replication, the mapped gene SDK1 may influence neural signaling pathways during the stress response and contribute to the regulation of mood and behavior in the face of adversity, which was in line with previous research. For instance, animal studies support that SDK1 could be involved in the response to cocaine (Scobie et al., Reference Scobie, Damez-Werno, Sun, Shao, Gancarz, Panganiban, Dias, Koo, Caiafa, Kaufman, Neve, Dietz, Shen and Nestler2014) and chronic social defeat stress (Bagot et al., Reference Bagot, Cates, Purushothaman, Lorsch, Walker, Wang and Nestler2016; Hultman et al., Reference Hultman, Ulrich, Sachs, Blount, Carlson, Ndubuizu, Bagot, Parise, Vu, Gallagher, Wang, Silva, Deisseroth, Mague, Caron, Nestler, Carin and Dzirasa2018) through multiple neural signaling pathways (e.g. upregulating poly polymerase-1 binding, regulating synaptic connections, and as a transcript regulated in the nucleus accumbens). Besides, results of human GWAS studies indicated that identified genetic locus (mapped to SDK1) was associated with multiple well-being indicators (e.g. positive affect, life satisfaction) (Baselmans et al., Reference Baselmans, Jansen, Ip, van Dongen, Abdellaoui, van de Weijer, Bao, Smart, Kumari, Willemsen, Hottenga, Boomsma, de Geus, Nivard and Bartels2019; Okbay et al., Reference Okbay, Baselmans, De Neve, Turley, Nivard, Fontana, Meddens, Linner, Rietveld, Derringer, Gratten, Lee, Liu, de Vlaming, Ahluwalia, Buchwald, Cavadino, Frazier-Wood, Furlotte and Cesarini2016; Qi & Chatterjee, Reference Qi and Chatterjee2018), as well as psychoneurological diseases (e.g. depression [Maul et al., Reference Maul, Giegling, Fabbri, Corponi, Serretti and Rujescu2020; Okbay et al., Reference Okbay, Baselmans, De Neve, Turley, Nivard, Fontana, Meddens, Linner, Rietveld, Derringer, Gratten, Lee, Liu, de Vlaming, Ahluwalia, Buchwald, Cavadino, Frazier-Wood, Furlotte and Cesarini2016] and Alzheimer’s disease [Adewuyi et al., Reference Adewuyi, O’Brien, Nyholt, Porter and Laws2022]). The methylation level of SDK1 (measured in the dorsolateral prefrontal cortex tissue, blood, or sperm) has been linked to stress responses and depressive symptoms (Rastad et al., Reference Rastad, Barjaste, Lanjanian, Moeini, Kiani and Masoudi-Nejad2023; Roberts et al., Reference Roberts, Gladish, Gatev, Jones, Chen, MacIsaac, Tworoger, Austin, Tanrikut, Chavarro, Baccarelli and Kobor2018; Snijders et al., Reference Snijders, Maihofer, Ratanatharathorn, Baker, Boks, Geuze, Jain, Kessler, Pishva, Risbrough, Stein, Ursano, Vermetten, Vinkers, Smith, Uddin, Rutten and Nievergelt2020; Yang et al., Reference Yang, Xu, Bierer, Flory, Gautam, Bader, Lehrner, Makotkine, Desarnaud, Miller, Jett, Hammamieh and Yehuda2021), further supporting its role in shaping resilience.
The main strengths of our study include the use of two large prospective cohorts in China for examining the association for the candidate locus across different types of physical stressors/traumatic events, which enhances the generalizability of our findings across different age groups and types of physical stressors/traumatic events. Additionally, the use of a residual-based approach to model psychological trajectory from pre- to post-trauma period addresses some of the limitations of previous studies, which often relied on self-reported resilience measures that may be influenced by subjective bias, confounding factors, or fail to capture the complexity of resilience as it manifests in real-world contexts.
Several limitations should be acknowledged. First, the PHQ-9 and GAD-7 may not fully capture the complexity of psychological symptoms; more detailed instruments could refine resilience phenotyping (Li et al., Reference Li, Li, Ding, Liu, Guo and Liu2024). Second, resilience was defined based on depressive symptom trajectories, assuming a shared liability with depression, and may therefore reflect resilience to depression rather than broader mental health domains. Third, resilience was assessed in the context of illness- and surgery-related stress and may therefore reflect resilience to physical trauma, limiting generalizability to other stressors such as early-life or chronic psychosocial adversity. Fourth, the genetic findings from CSTC should be interpreted cautiously. The G allele showed a directionally consistent but non-significant protective pattern under the additive genetic model, while the nominal genotype-stratified signal was driven by the small number of GG homozygotes. Thus, rs11766511 should be considered a candidate locus requiring further confirmation. Last, the relatively modest effect size of the identified genetic locus suggests that resilience is likely influenced by multiple genetic factors, each contributing a small effect. Larger, more diverse studies are needed to uncover additional loci and to better understand the polygenic nature of resilience.
Conclusions
In conclusion, this study identified rs11766511 and the mapped gene SDK1 as a candidate genetic signal of psychological resilience, providing preliminary clues for understanding the biological mechanisms that may help maintain mental health in the face of adversity. These findings should be interpreted cautiously and warrant confirmation in larger independent cohorts, as well as further biological investigation.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S0033291726105790.
Abbreviations
- AIMS
-
anesthesia information management system
- CSAC
-
China Surgery and Anesthesia Cohort
- CSTC
-
China Severe Trauma Cohort
- EMR
-
electronic medical record
- GAD-7
-
7-item Generalized Anxiety Disorder
- GWAS
-
genome-wide association study
- PHQ-9
-
9-item Patient Health Questionnaire
- PRS
-
polygenic risk scores
- PSQI
-
19-item Pittsburgh Sleep Quality Index
- SDK1
-
sidekick cell adhesion molecule 1
- SNPs
-
single-nucleotide polymorphisms
Data availability statement
Currently, data from CSAC and CSTC are not available to the public due to data protection and privacy policies. However, we welcome clinicians and researchers to contact us to initiate collaborative projects or join CSAC and CSTC as new recruitment centers. Please visit our website at http://biomedbdc.wchscu.cn/JoylabErasePM/hx-portals/#/index for more details, and contact corresponding author, HS (songhuan@wchscu.cn), for collaboration requests.
Acknowledgments
We thank all participants for providing their information. We also thank the team members and colleagues involved in the West China Surgery and Anesthesia Cohort project and the China Severe Trauma Cohort project for their support.
Author contribution
Acquisition of data: H.Y., W.C., J.H., Q.L., H.S.; Data and project management: H.W., W.C., Y.Z., Q.L., H.S.; Data cleaning and analysis: H.W., Y.Z., L.L., J.L., Z.C.; Drafting of the manuscript: H.W. and H.S.; Interpreted the data: H.W., F.F., J.S., J.H., Q.L., H.S.; Study concept and design: H.W., J.S., H.S. Approved the final manuscript as submitted and agree to be accountable for all aspects of the work: all authors.
Funding statement
This work was supported by the National Natural Science Foundation of China (No. 82402390 to HW, No. 82471535 to HS), the Natural Science Foundation of Sichuan Province (No. 2025ZNSFSC1757 to HW), the ‘Qimingxing’ Research Fund for Young Talents (No. HXQMX0175 for HW), the Sichuan Science and Technology Program (No. 2023ZYD0168), and the China Postdoctoral Science Foundation (No. 2023M742466 to WC).
Competing interests
The authors declare no competing interests.
Ethical standards
All participants enrolled in these two cohorts signed informed consent forms prior to any data or sample collection. The conduct of the CSAC and CSTC was approved by the Ethics Committee of West China Hospital, Sichuan University (approval numbers: 2020.469 and 2020.243).
Open access
