Introduction

Childhood trauma is a well-known environmental stressor associated with mental illness, conferring a threefold increase in risk of both affective and psychotic disorders1,2,3,4. However, the biological mechanisms linking trauma to psychopathology are poorly understood. Increasing the knowledge of the long-term biological correlates of early life trauma in severe mental disorders is a pivotal step in the development of new and personalized treatment targets.

Studies indicate that childhood trauma may lead to long-lasting effects by interfering with developmental trajectories of various biological systems5. However, the factors determining why only some individuals develop psychopathology following childhood trauma experiences remain largely unknown. One plausible explanation is that the distinct trajectory from early adverse events to clinical end-points is largely shaped by epigenetic or other modifications regulating the expression of genes6. Given that individuals differ in their genetic make-up, they will also differ in their transcriptional response to environmental influences6. In addition to increasing the risk of affective and psychotic disorders, a history of childhood trauma is associated with an earlier age at onset and an increased number of episodes after illness onset in bipolar disorder (BD) and schizophrenia (SZ)1,7,8, also after adjusting for current mood at the time of the assessment9. Earlier onset is more often observed in patients with a severe illness trajectory, and these patients are more likely to have higher inherited risk than those with later onset10. Cumulative clinical episodes, especially within the first three years of the illness, are associated with increased risk of treatment resistance11.

It is therefore plausible that expression levels of genes associated with both psychiatric disease and exposure to early life trauma could be determinants of clinical outcomes.

One gene of relevance is ANK3 (Ankyrin 3), which encodes a scaffolding protein localized to the axon and is required for the proper organization of ion channels12. Changes in ANK3 expression have been persistently linked to experiences of early life adverse events as well as psychopathology, including BD5. Variants of ANK3 have also consistently been identified as top hits in all three BD genome-wide association studies (GWAS) from the Psychiatric Genomics Consortium (PGC)13,14,15. Interestingly, ANK3 may be associated with SZ and post-traumatic stress disorder (PTSD)16,17. Moreover, ANK3 expression is associated with anhedonia and stress processing18, both of which are important features of traumatized patients including those with BD and SZ8,19,20.

BD and SZ are often viewed as part of the same continuum with shared neural, genetic and psychological mechanisms21,22, and both are associated with altered ANK3 expression16,18,23,25,,24. Animal and human studies have also demonstrated that early life stress may alter the methylation status of ANK35, with behavioral correlates, further pointing to ANK3 expression as a candidate moderator of early life stress effects in psychiatric diseases. A recent animal study also showed that repeated social defeat induces a persistent upregulation of ANK3 mRNA in the pituitary gland25, strengthening a link between ANK3 expression and early adversities.

The main objective of the current study was to investigate how ANK3 expression affects the association between childhood trauma and clinical manifestations such as age at onset and number of episodes of BD and SZ. Our overall hypothesis is that ANK3 expression will moderate the association between childhood adverse events and clinical trait characteristics of BD and SZ, in particular affective symptoms, with and without adjusting for current depressive symptoms. Specifically, we hypothesize that patients with high ANK3 mRNA levels and a history of childhood trauma experiences will have the earliest age at onset and the highest number of illness episodes.

Methods

Participants

Data were obtained from the ongoing Thematically Organized Psychosis (TOP) study at the Norwegian Center for Mental Disorders Research (NORMENT). Participants were enrolled between 2007 and 2021 from psychiatric units within the major hospitals in Oslo, Norway. For the current study, we included a total sample of n = 465 consisting of 291 patients with SZ spectrum disorder (schizophrenia, n = 164; schizophreniform, n = 24; schizoaffective, n = 46; other psychosis n = 57), and 174 patients with BD (bipolar 1, n = 135; bipolar II, n = 16; and not otherwise specified (NOS), n = 23), all recruited from the same catchment area. Exclusion criteria for patients were: age outside the range of 18–65 years, intellectual disability (IQ under 70), organic or substance-induced affective disorder or psychosis, or medical conditions interfering with brain function including neurological disorders, autoimmune diseases or cancer. Informed consent was obtained from all participants. The protocol for the study was approved by the South Eastern Norway Regional Committee for Medical Research Genetic Ethics (REK; 2009/2485/REK Sør-Øst). The Regional Committee for Medical Research Ethics and the Norwegian Data Inspectorate approved all aspects of the study. All methods were performed in accordance with the relevant guidelines and regulations.

Clinical assessment

Participants were diagnosed by trained psychologists and medical doctors using the Structured Clinical Interview for DSM-IV Axis I disorders (SCID-I), chapters A-E26. Age at onset of first SCID-verified episode and number of episodes were assessed using SCID-I and medical records. Clinicians involved in the diagnostic ratings received regular clinical supervision from senior researchers and professors, both individually and in groups. A good inter-rater reliability for diagnostic assessments at the TOP study was indicated, with an overall kappa score between 0.92 and 0.99 across assessment teams27. Regular medication was presented as Daily Defined Dosage (DDD). In patients with a SZ spectrum diagnosis, we assessed depressive symptoms using the Calgary Depression Scale for Schizophrenia (CDSS)28, and in BD by the Inventory for Depressive Symptomatology (IDS)29. Information on alcohol use was obtain by the Alcohol use disorders identification test (AUDIT; http://whqlibdoc.who.int/hq/2001/who_msd_msb_01.6a.pdf. Information on smoking was defined as daily smoking yes, no. Childhood trauma was measured using the retrospective Childhood Trauma Questionnaire (CTQ)30. Childhood traumatic events were assessed as a continuous variable ranging from 25 to 125, with a higher score indicating greater severity30. Subtypes of sexual abuse, emotional abuse, physical abuse, emotional neglect and physical neglect were also investigated (scores ranging from a minimum score of 5 to a maximum score of 25 for each subtype). Moderate to severe cutoff scores from the CTQ manual are presented in Supplementary Material Table S1. A CTQ total score is defined as CTQ composite score throughout the study. Although other studies of childhood trauma have been published in overlapping datasets7, the association between ANK3 expression and childhood trauma experiences within this study population has not yet been investigated.

RNA microarray analysis and quality control

Blood sampling was done in the morning after one night fasting (median time of blood sampling was 9 AM). Quantification of mRNA levels was carried out using microarray-based gene expression analysis as described elsewhere31. Briefly, blood samples were collected in Tempus Blood RNA Tubes (Life Technologies Corporation). Total RNA was extracted with the TEMPUS 12-Port RNA Isolation Kit (Applied Biosystems) and ABI PRISM 6100 Nucleic Acid PrepStation (Applied Biosystems) according to manufacturer’s protocol. Gene expression analyses were performed with Illumina HumanHT-12 v4 Expression BeadChip (Illumina, Inc.). Multidimensional scaling and hierarchical clustering were used for regular quality control, including sample quality measurements and removal of outliers, as well as removal of multiple batch effects (RNA extraction batch, RNA extraction method, DNase treatment batch, cRNA labelling batch, and chip hybridization).

Statistical analyses

Data were analyzed with IBM Statistics SPSS v27. Independent sample t-tests were performed for comparisons of demographic variables. The statistical modeling tool Process v.3.532 was applied to investigate interaction effects between ANK3 mRNA levels (moderator) and childhood trauma (independent variable) on clinical variables analyzed one at a time (as the dependent variable). The following clinical variables were investigated: current depressive symptoms, number of episodes (psychotic, elevated mood or depressive mood) and age at first episode. To control the Type I error rate we only investigated conditional effects if the interaction analysis was statistical significant. P-values presented were further adjusted for False Discovery Rate (FDR) and considered statistically significant if receiving a score of 1. All p-values in the result section are FDR-adjusted. Since a recent meta-analysis linked childhood abuse and childhood neglect to different symptoms in adults with a psychotic disorder1, the main analyses included different subtypes of trauma from the CTQ (physical abuse, sexual abuse, emotional abuse, physical neglect, and emotional neglect), as well as a CTQ total score. All variables in Process were entered as continuous variables. If the moderation analyses have a p value of < 0.05, Process calculates conditional effect of the predictor at the value of high, intermediate or low levels of the moderator, where high ANK3 group = 1 SD > mean levels, intermediate ANK3 group = mean ANK3 levels, and low ANK3 group = 1 SD < mean levels.

In the present study, interaction analyses and the conditional effects of the predictor at the different values of the moderator (high, intermediate, low) were run in Process with 5 000 bootstrap samples. As Process includes bootstrapping, it is a well-suited method for analyzing variables that are not normally distributed (including age at onset, and depressive symptoms from the CDSS and IDS). All statistical analyses involving ANK3 expression measurements were based on the batch-adjusted log2-transformed data. Assumptions for regression analyses were checked and found satisfying. The main analyses were adjusted for age, sex, diagnosis, time of blood sampling, and medication use represented as defined daily dose (DDD) in accordance with the World Health Organization (WHO) guidelines (https://www.whocc.no/ddd/). We adjusted for duration of illness in the analyses of number of episodes. As current mood may affect retrospective information of childhood trauma events33, analyses were also conducted with and without adjustment for current depressive symptoms. As smoking status has been linked to changes in methylation and gene expression34, sensitivity analyses was also conducted adjusting for daily smoking (yes, no). A sample of 465 participants will have 80% power of detecting a medium effect, at a significance criterion of α = 0.05.

Results

Sample description

A total of 291 patients with SZ and 174 patients with BD were included (Table 1). The age at inclusion was 30.5 ± 10.2 (mean ± SD) years. The male-to-female ratio was 1.4 (56% males and 44% females). The sample consisted of 84.3% Europeans. The mean CTQ composite score was 43.84 ± 15.38 (range 25–117), with emotional neglect being the most frequently reported trauma reported by 29.3% of the patients. Adjusted for age, sex, and time of day of blood sampling, a trend was found for higher ANK3 mRNA in BD compared to SZ (5.54 ± 0.38 and 5.47 ± 0.39, respectively, t = 1.90, P = 0.06). The mean age at onset in the total sample was 22.05 ± 9.04 and the average number of total episodes (including depressive, hypomanic/manic and psychotic episodes) were 5.47 ± 5.42. One hundred and fifty-four (74%) of BD patients had at least one psychotic episode. One hundred and thirty-six (47%) of the SZ spectrum group had at least one depressive episode, and 29 (10%) had at least one manic or hypomanic episode. No significant difference in current depressive symptoms or medication (DDD) was observed between the two patient groups, but patients with BD had more depressive and manic/hypomanic episodes compared to SZ. No association was observed between subtypes of medication (antipsychotic, mood, or antidepressants DDD) and ANK3 levels (p > 0.1, data not shown). Smoking status was associated with having lower ANK3 levels (current smokers, ANK3 mean ± SD = 5.46 ± 0.38 compared to 5.54 ± 0.40 in non-smokers, t = − 2.36, p = 0.009, CI =  = − 0.16–0.01).

Table 1 Clinical characteristics and sample overview.
Full size table

Patients with BD were more likely to be married and working or fulltime student compared to patients with SZ, while no difference was observed for daily tobacco use or alcohol use from the Audit.

Patients with higher ANK3 mRNA also showed a trend of more manic/hypomanic episodes and earlier age at onset of mania/hypomania (see Supplementary Material Tables S2–S3).

ANK3 expression, childhood trauma and clinical characteristics

Within the total sample, ANK3 mRNA levels moderated the association between CTQ composite score and current depressive symptoms (ßsexual abuse × ANK30.01, p < 0.05, CI =  < 0.01–0.003; Fig. 1 r2 = 0.01).

Figure 1
figure 1

ANK3 mRNA, childhood trauma composite score and depressive symptoms. Process, moderation analysis. ANK3 mRNA levels moderated the association between CTQ composite score and current depressive symptoms (ßCTQ composite score × ANK3 = .01, p < .05. Depressive symptoms were assessed by the z-scores of Calgary Depression Scale for Schizophrenia (CDSS) and The Inventory for Depressive Symptomatology (IDS) for bipolar disorder.

Full size image

ANK3 mRNA moderated the effects of childhood sexual abuse and emotional neglect on number of manic/hypomanic episodes (ßsexual abuse × ANK3 = 0.43, P = 0.003, CI = 0.14–0.71; r2 = 0.01 and ßsexual abuse × ANK3 = 0.24, P = 0.0009, CI = 0.06–0.042; r2 = 0.01 respectively; Table 2). Patients with high ANK3 mRNA and with the most severe levels of sexual abuse or emotional neglect had more manic/ hypomanic episodes (ß = 0.27; p = 0.0005, CI = 0.12–0.42; and ß = 0.21; p < 0.001 CI = 0.11–0.31; respectively, Fig. 2a, b). No significant associations were observed for depressive or psychotic episodes (p > 0.1).

Table 2 ANK3 mRNA moderates the association between subtypes of trauma and clinical traits.
Full size table
Figure 2
figure 2

(a–b) ANK3 mRNA moderates the relationship between sexual abuse, emotional neglect, and elevated mood episodes. Process, moderation analysis. Sexual abuse: ß = .43, p = .003. Emotional neglect: ß = .24, p = .0009. Adjusted for age, sex, medication, Daily Defined Dose, diagnosis (schizophrenia spectrum/bipolar disorders), duration of illness, and time of blood sample. High ANK3 group = 1 SD > mean levels. Intermediate ANK3 group = mean ANK3 levels, and Low ANK3 group = 1 SD < mean levels.

Full size image

ANK3 mRNA levels moderated the association between childhood sexual abuse and age at onset of mania/hypomania (ßsexual abuse × ANK3 = − 1.23, p = 0.006, CI = − 2.10 to − 0.36; r2 = 0.02). Patients with high ANK3 mRNA levels and severe scores of sexual traumas had the earliest age at onset of mania/hypomania (ß = − 0.64; P = 0.01, CI = − 1.14 to − 0.15; see Fig. 3). No statistically significant relationship between ANK3 expression, childhood traumas and age at first psychotic or depressive episode (P > 0.05).

Figure 3
figure 3

ANK3 mRNA moderates the relationship between sexual abuse and earlier age at onset of mania/hypomania. Process, moderation analysis. ß = − 1.23, p = .006. Adjusted for age, sex, medication, Daily Defined Dose, diagnosis (schizophrenia spectrum/bipolar disorders) and time of blood sample. High ANK3 group = 1 SD > mean levels. Intermediate ANK3 group = mean ANK3 levels, and Low ANK3 group = 1 SD < mean levels, AAO = age at onset.

Full size image

Clinical traits adjusting for current depressive symptoms

Sensitivity analyses adjusting for current depressive symptoms showed that ANK3 expression moderated the relationship between specific types of childhood traumatic events (sexual abuse and neglect) and number of mania/hypomanic episodes and corresponding age at onset (see Table 3) also after taking into consideration current depressive symptoms. Findings above remained unchanged when smoking status was added into the model (data not shown).

Table 3 ANK3 mRNA moderates the association between subtypes of trauma and clinical traits adjusted for current depressive symptoms.
Full size table

Discussion

We found that ANK3 mRNA levels moderated the relationship between specific types of childhood traumas and affective episodes in severe mental disorders. Within the total sample, Patients with both high ANK3 expression and with the most severe childhood sexual abuse and emotional neglect had more manic/hypomanic episodes and an earlier age at onset of the first episode. No relationship was observed between number of psychotic or depressive episodes and ANK3 mRNA levels.

A recent meta-analysis of childhood trauma in psychosis empathized that biological mechanisms linking trauma to clinical presentations are lacking (Alameda et al.1). Our study suggests an interplay between ANK3 mRNA levels and childhood trauma and affective episodes. Affective symptoms are the most consistent clinical dimension associated with childhood trauma experiences1. As previously discussed, childhood trauma is a well-known environmental stressor associated with mental illness, conferring a threefold increase in risk of both affective and psychotic disorders1,2,3,4. As previously reported in partly overlapping samples7,35, our current study showed that patients with childhood trauma experiences had more severe clinical features, including more mood episodes and an earlier age at onset of affective episodes (see Supplementary Material Tables S1, S2). However, the factors determining why only some individuals develop psychopathology following childhood trauma experiences remain largely unknown. One plausible explanation is that the expression of genes that shape the trajectory from early adverse events to clinical end-points6. Large GWAS studies in severe mental disorders demonstrate an altered ANK3 expression in this group compared to healthy individuals14,15, and our study suggests an interplay between ANK3 mRNA levels and childhood trauma exposure which increases vulnerability to develop affective symptoms, as is also supported by previous studies showing a direct link between early stress and ANK3 levels5,25.

Exposure to stress or early adversities both after and in the perinatal period has been linked to changes in the methylation status of the ANK3 gene5, which is supported by RNA changes in animals exposed to stress25. Furthermore, in humans an interaction has been observed between a polymorphism affecting ANK3 and obstetric complications impacting connectivity during a working memory task often found to be impaired in mental disorders5. As discussed by Luoni and colleagues5, ANK3 may be an important mediator between early life stress and psychopathology by changing neuronal circuits important for psychiatric illness. Since ANK3 expression can be modified by pharmacological interventions and by long-term lithium use, ANK3 could be seen as a target for drug intervention in individuals with stress-related disorders with neurodevelopmental origin, but more research is needed in this field.

While our study has several strengths, such as a well characterized clinical sample with detailed information on gene expression and clinical characteristics, it also has some limitations. First, history of childhood trauma was measured retrospectively by asking adults about experiences in their childhood. A recent meta-analysis showed low overlap between prospective and retrospective data of childhood trauma33, and current mood at the time of assessment may contribute to over or under-reporting of retrospective experiences36. As discussed by33 retrospective data may have higher sensitivity in detecting actual cases that otherwise could go unnoticed, but reliance solely on retrospective assessment methods might also lead to misclassification of exposed groups33,37,38. Therefore, improving the reliability of retrospective data, such including sensitivity analyses adjusting for current depressive symptoms level, is warranted which we have done in this study. Second, ANK3 expression was measured in whole blood, which may not wholly recapitulate expression patterns in brain. However, previous studies have demonstrated a significant up-regulation of ANK3 in peripheral blood from BD patients23,39, which is consistent with the predicted direction of effect of the protective ANK3 variant associated with BD23. This suggests that peripheral expression of ANK3 may be an acceptable proxy for brain expression. The role of physical health as well as information from healthy controls should be further addressed in future studies. Our study specifically focused on ANK3 mRNA expression as the gene has been found to be associated with both BD and schizophrenia in large GWAS14,16, and studies have also shown that ANK3 is influenced by early trauma experiences5,18. Future studies should also investigate in larger samples a broader specter of gene expression and its alteration by childhood trauma in severe mental disorders. In summary, our study points to ANK3 mRNA expression as a biological marker influencing the relationship between childhood trauma experiences and more severe affective clinical traits in severe mental disorders.