Abstract
Previous studies have shown significant sex-specific differences in major depressive disorder (MDD) in multiple biological parameters. Most studies focused on young and middle-aged adults, and there is a paucity of information about sex-specific biological differences in older adults with depression (aka, late-life depression (LLD)). To address this gap, this study aimed to evaluate sex-specific biological abnormalities in a large group of individuals with LLD using an untargeted proteomic analysis. We quantified 344 plasma proteins using a multiplex assay in 430 individuals with LLD and 140 healthy comparisons (HC) (age range between 60 to 85 years old for both groups). Sixty-six signaling proteins were differentially expressed in LLD (both sexes). Thirty-three proteins were uniquely associated with LLD in females, while six proteins were uniquely associated with LLD in males. The main biological processes affected by these proteins in females were related to immunoinflammatory control. In contrast, despite the smaller number of associated proteins, males showed dysregulations in a broader range of biological pathways, including immune regulation pathways, cell cycle control, and metabolic control. Sex has a significant impact on biomarker changes in LLD. Despite some overlap in differentially expressed biomarkers, males and females show different patterns of biomarkers changes, and males with LLD exhibit abnormalities in a larger set of biological processes compared to females. Our findings can provide novel targets for sex-specific interventions in LLD.
Keywords: late-life depression, sex differences, aging, biomarkers
Introduction
Major depressive disorder (MDD) is a leading cause of disability worldwide, and it is estimated that 170 million individuals are currently affected by depression worldwide (Herrman et al., 2022). A depressive episode is characterized by persistent feelings of sadness, and loss of interest in usual activities, among other symptoms that interfere with daily life activities (Kennedy, 2008). Many factors can significantly influence the manifestation and outcomes of MDD, including biological sex and gender expression (Schuch et al., 2014). Females have prevalence rates of MDD twice higher than males across the lifespan and usually report more severe symptoms (Riecher-Rossler, 2010). They also have higher rates of comorbid anxiety, while males experience higher rates of comorbid substance use (e.g., 48.1% vs. 24.5%) (Bekker and van Mens-Verhulst, 2007; Schuch et al., 2014). In addition, females have higher rates of suicide attempts than males (Bommersbach et al., 2022); however, suicide completion is four times higher in among the latter (https://www.cdc.gov/suicide/suicide-data-statistics.html, accessed on 1/26/2024). Biological sex may also influence treatment outcomes in MDD, with females having better treatment response to serotonin selective reuptake inhnibitors (SSRI), while men to tryciclic antidepressants (Kornstein et al., 2000). However, other studies showed no significant differential effect of sex on treatment response (reviewed in (Sramek et al., 2016).
These differences in incidence, clinical presentation, and treatment outcomes suggest that the mechanisms of MDD may differ based on biological sex. A recent work investigated sex-specific molecular signatures of MDD in post-mortem brain tissue and found 706 genes differentially expressed in men and 882 genes in women with MDD (Seney et al., 2018). Interestingly, 52 genes displayed transcriptional changes in opposite directions between males and females. Males had decreased expression of synapse-related genes and increased expression of oligodendrocytes, microglia, and inflammation-related genes. In contrast, females showed increased expression in synapse-related genes and decreased expression in the oligodendrocyte and microglia-related genes. Neuroimaging studies have also shown a differential effect of sex on subcortical brain region volume, e.g., a larger amygdala volume in females and reduced volume in males (Whittle et al., 2014). Another study indicated that depressed female adolescents had significantly higher myelin content in the left uncinate fasciculus and corpus callosum genu compared to adolescent males (Ho et al., 2021).
Most studies on the impact of biological sex on biomarkers differences in MDD included young and middle-aged adults. However, few studies have evaluated the impact of biological sex on biomarkers in older adults with MDD, also known as late-life depression (LLD). For example, a recent study found that three distinct single nucleotide polymorphisms of the CYP19A1 gene were associated with an increased risk of LLD only in females (Ancelin et al., 2020). No prior studies evaluated sex effects on protein expression related to multiple biological pathways in LLD. In this study, our primary aim was to characterise potential sex-specific protein expression patterns in LLD using an untargeted proteomic analysis. We leveraged a large study of older adults with LLD in whom we had available multiplex-based biomarker analyses. We also used weighted gene co-expression network analysis (WGCNA) to identify sex-specific protein expression modules and gene ontology analyses to identify the biological processes and pathways related to differentially expressed proteins in older females and males with LLD.
Methods
Study sample
This study used baseline cross-sectional data from participants in two NIH-funded studies: Incomplete Response in Late-Life Depression: Getting to Remission (IRL-GREY) (ClinicalTrials.gov Identifier: NCT00892047) and Mindfulness-Based Stress Reduction, Health Education and Exercise (MEDEX) (ClinicalTrials.gov Identifier: NCT02665481). There are differences between the IRL-GREY and MEDEX samples. For example, MEDEX participants were required to be off medications for diabetes or inflammatory conditions. Furthermore, these participants had to be considered medically healthy enough to exercise. In contrast, the IRL-GREY participants had no such exclusions.
Late-Life Depression (LLD) participants:
430 older adults (60 years and older) diagnosed with a current major depressive episode (MDE) who were participants in the IRL-GREY study were included in this analysis. Details about the study methods and main results have been previously published (Lenze et al., 2015). In brief, the participants of this trial were recruited from 3 sites (St Louis, MO; Pittsburgh, PA; and Toronto, Canada). The diagnosis of MDE was made based on the Structured Clinical Interview for DSM-IV (SCID-IV) (First et al., 2002), based on the DSM-IV diagnostic criteria for MDE (2000). Depressive episode recurrence, duration, history of previous suicide attempts, comorbid anxiety disorders, and the age of onset of major depressive disorder were also collected via the SCID-IV. The severity of depressive symptoms was measured by the Montgomery-Asberg Depression Rating Scale (MADRS) (Montgomery and Asberg, 1979). Medical comorbidity burden was quantified with the Cumulative Illness Rating Scale – Geriatrics (CIRS-G) (Miller et al., 1992). The CIRS-G is a 14-item instrument that measures the number and severity of physical health problems (13 organ systems, 0–4 score for each system, total scores ranging from 0 to 56). Higher scores are indicative of a greater medical comorbidity burden.
Healthy Comparison (HC) group participants:
140 older adults (60 years and older) without evidence of a current major depressive episode or major neurocognitive impairment were included as a healthy comparison group (HC) in this study. They were recruited as part of the MEDEX study, a randomized clinical trial to evaluate the effects of mindfulness and multimodal exercise to improve cognitive performance. MEDEX study participants underwent a similar clinical assessment as the IRL-GREY study participants, namely the evaluation of current and past psychiatric history using the MINI 7.0(Sheehan et al., 1998). However, the MEDEX study did not use the same cognitive battery as IRL-GREY. Details on descriptive characteristics of the study participants were previously published by Wetherell et al. (Wetherell et al., 2020).
Biomarker analyses
Laboratory analysis:
Blood samples were collected with EDTA tubes by antecubital venous puncture. Plasma samples were separated, aliquoted, and stored at − 80 °C until laboratory analyses.
Plasma samples were analyzed using LUMINEX 100/200 (Luminex Corp., Austin, TX) using customized multiplex immunoassays from R&D/Biotechne (Minneapolis, MN) and according to the manufacturer’s instructions. All laboratory analyses were carried out in 2 different analysis batches with HC and LLD samples randomly distributed across them. The intra-assay coefficient of variance was below 10% for all biomarkers and the inter-assay coefficient of variance ranged from 3% to 22%.
A total of 342 different analytes were included in the customized panel (Supplementary table 1). This multiplex assay aimed to provide comprehensive coverage of biomarkers related to different biological processes and pathways relevant to age-related disorders, including cancer, cardiovascular diseases, and metabolic disorders, as well as psychiatric and neurodegenerative disorders.
Statistical Methods
Proteomics data preprocessing
The preprocessing procedures include a normalization to improve the comparability between samples by reducing the experimental variances and the imputation of missing values. The normalization method is the Cycloess implemented by the R package NormalyzerDE (Willforss et al., 2019), where Cycloess estimates and removes the bias function of log expression ratio versus log expression abundance (which should be independent). Then the missing values are imputed using multiple imputation R package mice (m = 5) and the method “predictive mean matching” (method = ‘pmm’) (van Buuren and Groothuis-Oudshoorn, 2011). We chose the method based on the performance benchmark of four imputation methods: K-nearest-neighbor, predictive mean matching, bayesian linear regression, and random forest imputations by randomly deleting non-missing observations and then comparing the imputation error (supplementary Figure S1).
Differential expression (DE) analysis
We fitted random intercept models with covariate selection to study the effect of LLD, self-reported sex assigned at birth, and their interaction for the expression of biomarkers. Covariates included in the model were age, CIRS-G scores, and BMI, where two at most were selected for each biomarker based on the Bayesian information criterion (BIC). We also included batch and study sites as random intercepts to adjust for the batch effect and site differences.
The data analyses were repeated for each imputed dataset (m = 5) and pooled for final estimation. We calculated the p-values of the main effects (sex, group (HC vs. LLD), and interaction) by permuting the sample data against their covariates, eliminating the bias introduced by the additional covariate selection procedure. To adjust for multiple testing across the protein panel, we computed Benjamini-Hochberg (BH) adjusted p-values (p.BH) and set the false discovery rate (FDR) at 5%. We performed gene set enrichment analysis (GSEA) with the R package fgsea (Korotkevich et al., 2021) and MSigDB 3.0 (38) ontology gene sets for a threshold-free comparison of DE-enriched pathways between the sexes.
Identification of sex-specific DE proteins
To identify DE proteins in both sexes, in females and males only, we calculated the combined p-values of both sexes using the adaptively weighted Fisher’s method (AW-Fisher) (Huo et al., 2020). The AW-Fisher method takes p-values of the LLD effect from both sexes and computes combined p-values with weights indicating which sex has a significant DE. For example, an AW-Fisher p-value less than 0.05 with a weight (1, 0) indicates a significant DE in females but not in males, whereas a weight (1, 1) indicates a significant DE in both females and males. This procedure serves the same purpose as drawing a Venn diagram but is more statistically rigorous. We adjusted the AW-Fisher p-values to AW-Fisher q-values with Benjamini-Hochberg at FDR < 5% to account for multiple testing.
In this study, genes with AW-Fisher q-value less than 0.05 are considered DE in either sex, then the DE status in each sex is decided by the AW-Fisher weight. Finally, genes that are DE in either sex are plotted in volcano plots.
Identification of sex-specific protein modules
We used weighted gene co-expression network analysis (WGCNA) to build protein modules where module memberships have highly correlated expression. For each sex, we performed hierarchical clustering to proteins based on their absolute correlation with each other, and the Dynamic Tree Cut method (Langfelder et al., 2008) assigned proteins with high absolute correlation into the same module. We named the protein modules after representative GO pathways enriched by the module proteins. We used Fisher exact tests to assess the enriched GO pathways based on MSigDB 3.0 (Liberzon et al., 2011).
We then calculated the module connectivity of identified protein modules in HC and LLD groups in each sex separately and performed module differential connectivity (MDC) analyses. The module connectivity is the sum of network degrees, which measures how many connections are present in the network. The MDC analysis calculates the connectivity ratio of identified protein modules between HC and LLD groups (the LLD group was the reference group). An MDC >1 indicates that the module has higher connectivity (gain-of-connectivity) in HC compared to LLD, while an MDC <1 indicates that the module has lower connectivity (loss-of-connectivity) in HC. The statistical significance of the MDC result was then estimated using permutation (n = 1000). We generated the null distribution of MDC by permuting both samples and the proteins: by permuting the samples, we generated the null distribution of protein connectivity; by permuting the proteins, we generated random protein modules.
We also conducted module-trait association analyses for identified modules and clinical traits of interest (age, CIRS-G, systolic blood pressure, diastolic blood pressure, BMI, and MADRS (LLD individuals only). For each module, we computed the first principal component (PC) of member proteins as the module eigengene (ME), and then the module-trait association is calculated as the pearson correlation between ME and selected traits. The p-values of correlation tests between multiple modules and clinical traits are also corrected with Benjamini-Hochberg procedure.
Results
Demographic and clinical characteristics of participants
The characteristics of the sample are shown in Table 1. There were statistically significant differences in the sex distribution, age, medical comorbidity burden, and BMI between LLD and HC. We found a statistically significant difference in the distribution of males and females between LLD and HC groups (p=0.036). However, such a difference would not impact our analysis since we control for sex in the differential expression analysis and stratify data by sex in gene module detection. Since CIRS-G scores and BMI were significantly higher in the LLD group, these variables were included as covariates in all analyses to control for their potential confounding effects in the biomarker levels.
Table 1.
Demographic and clinical characteristics between LLD patients and HC.
| Healthy Comparison (n=140) | Late-Life Depression (n=430) | p-value | |
|---|---|---|---|
| Sex (Males (%)) | 36 (25.7) | 154 (35.8) | 0.036 |
| Race (%) | 0.546 | ||
| White | 119 (85.0) | 380 (88.4) | |
| Black | 17 (12.1) | 42 (9.8) | |
| Others | 4 (2.9) | 8 (1.9) | |
| Age (mean (SD.)) | 71.29 (4.80) | 68.89 (7.09) | <0.001 |
| Study Site (%) | <0.001 | ||
| Pittsburgh | 0 (0.0) | 177 (41.2) | |
| Toronto | 0 (0.0) | 121 (28.1) | |
| San Diego | 44 (31.4) | 0 (0.0) | |
| St. Louis | 96 (68.6) | 132 (30.7) | |
| CIRS-G (mean (SD.)) | 6.47 (2.92) | 9.94 (4.47) | <0.001 |
| Systolic BP (mean (SD)) | 135.35 (17.20) | 132.02 (19.02) | 0.067 |
| Diastolic BP (mean (SD)) | 75.58 (11.06) | 76.26 (11.56) | 0.539 |
| BMI (mean (SD.)) | 27.76 (4.77) | 29.76 (6.82) | 0.001 |
Others include American Indian, Asian Pacific and more than one race
CIRS-G: Cumulative Illness Rating Scale – Geriatric; BP: blood pressure; BMI: body mass index.
Biomarkers’ data preprocessing
Of the 344 biomarkers assayed for this study, 54 had missing data due to random laboratory variation and falling below the lower limit of quantitation (LLOQ) for each biomarker. The missingness rate varied from less than 1% to 36%, with a mode value of 24.6%. We imputed all the missing data using the multiple imputation R package “mice” with predictive mean matching (pmm) and imputation times of 5.
Differential Expression Analysis
Overall, we identified 102 DE proteins between LLD and HC, among which 33 were female-specific, 6 male-specific, and 63 were shared by both sexes (Figure 1A). The top 15 biomarkers with the most significant DE in females and males are listed in Table 2 and Table 3, respectively. Additional information about DE proteins in males and females are available in the Supplementary table 2. The volcano plots (Figure 1B) show the three types of DE proteins in both sexes: the shared DE proteins (plotted in red, e.g., CXCL14 and S100A9) had similar effect sizes and p-values; the sex-specific DE proteins showed a noticeable drop in effect sizes and increased in p-values in the plot for the other sex, e.g., KLK3 in the male plot has a large negative effect while the effect size is close to zero in the female plot. The supplementary material shows the coefficients and p-values for each DE protein category.
Figure 1 -.
Differential expressed (DE) proteins. A. Number of DE proteins in each sex by a cutoff of AW-Fisher q-value 0.05. B. Volcano plots for proteins that are DE in either females or males. The effect sizes and p-values for the LLD effect are plotted separately for each sex. C. Gene set enrichment analysis for females and males separately. The signs indicate the directions of the enrichment score (ES) from GSEA. The signs are the same for females and males in presented pathways.
Table 2.
Top 15 DE of LLD in Females (HC as the reference group)
| Protein name | Effect Size | corrected.p | FDR |
|---|---|---|---|
| CXCL-14 | −1.61131148 | 1.84E-05 | 0.006330046 |
| S100A9 | 1.695555142 | 3.68E-05 | 0.006330046 |
| Tissue Factor | −0.92618374 | 5.52E-05 | 0.006330046 |
| NSE | −1.07438629 | 7.36E-05 | 0.006330046 |
| Midkine | −0.71406161 | 9.20E-05 | 0.006330046 |
| TGF-α | 0.712774801 | 0.000110408 | 0.006330046 |
| ALDH1A1 | −0.94101965 | 0.000128809 | 0.006330046 |
| Park-7 | −0.59674537 | 0.00014721 | 0.006330046 |
| CCL-28 | −0.59227911 | 0.000165612 | 0.006330046 |
| MPO | 0.910149546 | 0.000184013 | 0.006330046 |
| CD40-L | 0.691039408 | 0.000202414 | 0.006330046 |
| CXCL-9 | −0.30076528 | 0.000220816 | 0.006330046 |
| HB-EGF | −0.40483448 | 0.000239217 | 0.006330046 |
| XCL-1 | −0.36565214 | 0.000257618 | 0.006330046 |
| TNFSF-14 | 0.594331983 | 0.000276019 | 0.006330046 |
Table 3.
Top 15 DE of LLD in Males (HC as the reference group)
| Protein name | Effect Size | corrected.p | FDR |
|---|---|---|---|
| CXCL-14 | −1.5794986 | 1.84E-05 | 0.00633005 |
| S100A9 | 1.6472944 | 3.68E-05 | 0.00633005 |
| Tissue Factor | −0.8154824 | 5.52E-05 | 0.00633005 |
| CCL-28 | −0.6882839 | 7.36E-05 | 0.00633005 |
| NSE | −0.8563452 | 9.20E-05 | 0.00633005 |
| TGF-α | 0.60553874 | 0.00011041 | 0.00633005 |
| Midkine | −0.5663875 | 0.00012881 | 0.00633005 |
| CD40-L | 0.62032037 | 0.00018401 | 0.00791256 |
| Park-7 | −0.4391619 | 0.00029442 | 0.01055008 |
| ALDH1A1 | −0.6757815 | 0.00033122 | 0.01055008 |
| IL6-R | −0.2561894 | 0.00034962 | 0.01055008 |
| Uromodulin | −0.4410915 | 0.00036803 | 0.01055008 |
| MPO | 0.71937915 | 0.00042323 | 0.01085151 |
| HB-EGF | −0.4021217 | 0.00044163 | 0.01085151 |
| XCL-1 | −0.3709321 | 0.00047843 | 0.01097208 |
The gene set enrichment analysis (GSEA) showed that DE proteins from both sexes were enriched in many biological processes. The threshold-free GSEA identified a higher number of significantly enriched pathways in males (n=43) than in females (n=24). Figure 1C plots the enriched pathways with a p-value < 0.01 in either sex. Female-specific pathways were mainly related to immunoinflammatory control, while the male-specific pathways covered a broader range of biological processes, including immunoinflammatory control, cell cycle and cell fate control, vascular processes, and metabolic control.
Sex-specific protein modules and enrichment analysis
We next used WGCNA to identify protein modules in each sex within which the proteins are highly correlated. We identified nine protein modules in females and sixteen protein modules in males (Figure 2A). The module differential connectivity (MDC) analysis identified two and three differentially connected modules in females and males, respectively (Figure 2B). In females, we found a significant gain-of-connectivity (FDR < 0.00001) in the module related to “Cellular adhesion” and a loss-of-connectivity (FDR < 0.00001) in the proteins from the module related to “Alpha-synuclein pathology”. In males, we found a significant gain-of-connectivity in modules related to “T cell-mediated cytotoxicity” (FDR = 0.008) and “T cell migration” (FDR = 0.008) and significant loss-of-connectivity in modules related to “Gene expression control” (FDR < 0.00001). Then we conducted a module-trait association analysis to evaluate how each protein module was associated with demographic variables and clinical characteristics (Figure 2C). In males with LLD, MADRS scores showed a weak but significant association with the “Apoptosis regulation” protein module. We found age was positively associated with “Cellular Adhesion” with a moderate correlation but negatively associated with “Regulation of neuronal apoptosis process” and “T cell migration”. The CIRS-G was also positively associated with “Cellular Adhesion” and “Apoptosis regulation” but negatively associated with “Metabolic control”. Finally, we found a strong negative association between “Metabolic control” and BMI. Interestingly, the association between protein modules and clinical traits was weaker in females than in males. In females, MADRS scores showed a weak association, albeit statistically significant, with the “extracellular vesicle synthesis” protein module.
Figure 2.
Weighted gene co-expression network analysis (WGCNA) results. A. Protein dendrogram with modules identified by hierarchical clustering. Proteins mapped to the same color are assigned to the same module. B. The module differential connectivity (MDC) result for each module in the two sexes. The color on the outer circle maps the module colors in A. And the inner circle indicates if the module gains or loses connectivity by FDR<0.05. NS stands for not significant. C. The Module-trait correlation heatmap. The warmer color stands for a higher positive correlation and the colder color stands for a higher negative correlation. The number on top of each cell is the estimated correlation and the number in the parenthesis is the BH-adjusted p-value for the correlation test.
Discussion
In this study, we explored the impact of biological sex in peripheral protein expression in LLD. First, we showed that males and females show significant sex-specific differences in protein expression in LLD. Females showed a larger set of differentially expressed proteins than males, but the males showed a greater diversity of enriched biological processes and pathways. Furthermore, we identified nine protein modules of highly correlated proteins in females compared to the 16 protein modules in males. Females had a gain of connectivity in the module related to cellular adhesion and a loss of connectivity in modules related to alpha-synuclein pathology. Males had a gain of connectivity in modules related to T cell-mediated cytotoxicity and a loss of connectivity in the module related to gene expression control. Our findings indicate that biological sex has a distinct effect on the pattern of protein expression profiles in LLD.
Gene ontology analyses of differentially expressed proteins showed that both females and males had significant abnormalities in multiple biological processes related to immune-inflammatory control. This finding is in line with a rich literature suggesting that abnormalities in immune-inflammatory control is common in LLD and can be a therapeutic target for this condition (Diniz et al., 2016; Drevets et al., 2022; Janssen et al., 2021). However, males with LLD showed broader abnormalities in biological process that are relevant to the development of depression, including the hormonal signaling, metabolic control, neuronal apoptosis and cell fate. For example, we uncovered the association between lower levels of kallikrein 3 (aka., prostate specific antigen, PSA) and LLD in males. No previous study have provided a link between PSA levels and LLD or major depression across the lifespan in males and plausible mechanistic links between PSA and LLD are unclear. However, our finding reinforces the importance of biological sex differences in LLD since PSA is expressed at much higher levels in males than females. Moreover, the broader impact of DE proteins in biological processes observed in males can help to explain the higher risk of adverse outcomes (e.g., mortality) observed in males (Diniz et al., 2014). For example, since many of these biological processes are features commonly associated with accelerated biological aging (López-Otín et al., 2013), such abnormalities can provide a hypothetical mechanistic link for why males with LLD have higher risk of age-related outcomes, including mortality. Interestingly, recent studies showed that males with LLD have a higher senescence burden, a hallmark of biological aging, than females (Diniz et al., 2022; Seitz-Holland et al., 2023).
Our findings are in line with previous studies in the literature but add novel information about sex-specific differences in biological abnormalities in LLD. A previous investigation of sex-dependent biomarkers of MDD analyzed 171 serum molecules in adults (Ramsey et al., 2016). They found 28 biomarkers that were sex-dependent, with males presenting with elevated levels of proteins such as CRP, TFF3, cystatin-C, fetuin-A, and sTNFR2, which are implicated in the function of T cells, monocytes, and macrophages. These findings are in concordance with our findings that males with LLD have a gain of connectivity in protein modules relating to T cell migration and T cell-mediated cytotoxicity. Other studies found a significant sex-dependent effect on genes related to neuronal development, immune function, and vascular functions (Blokland et al., 2022; Kang et al., 2020). Altogether, these findings suggest that biological sex significantly impacts immune function and inflammatory pathways in LLD. These results are also consistent with findings that sex significantly impacts immune response in other psychiatric disorders, including schizophrenia and bipolar disorder (Blokland et al., 2022).
The WGCNA can highlight fundamental differences in the patterns of protein co-expression and how they are related to different clinical characteristics (“traits”) in any given condition(Horvath and Dong, 2008). Specifically, these analyses can evaluate how different protein groups can underlie specific clinical characteristics or manifestations of different diseases. Our WGCNA analyses revealed that males and females had different patterns of protein co-expression modules. In males, the association between age, CIRS-G scores, and BMI with different protein modules (e.g., “cellular adhesion’”, T cell migration”, “apoptosis regulation”, and “metabolic control”) suggest that males may be more susceptible to the biological effects of inflammation and pro-aging biological processes. Interestingly, neither males nor females showed a moderate to strong correlation between any protein modules and MADRS scores, a measure of the severity of depressive symptoms. These results may indicate that depressive symptoms have a non-specific effect on different protein expression profiles captured in the differential expression analyses in females. Finally, our findings from WGCNA reinforce that males and females have different patterns of biological abnormalities which have sex-specific associations with distinct clinical traits in LLD.
Our results need to be interpreted in light of the study’s limitations. First, our results are limited by the relatively low racial and ethnic diversity of the sample. Further work is needed to extend these findings to non-white ethnic and other racial groups. Second, the HC and depressed participants originated from two different cohort studies, and as a consequence, our results might be due to methodological differences and sample inclusion/exclusion criteria differences between studies. To address this limitation, we randomly allocated the samples (IRL-GREY and MEDEX studies) in the two different experimental batches. We also included the laboratory batch as a random variable in the statistical models to minimize the risk of bias in the laboratory and statistical analyses. We did not include neurocognitive performance in our models, although both samples had the same cognitive exclusion criterion (i.e., dementia). It is widely acknowledged that cognitive impairment has significant impact on protein expression patterns in LLD (Diniz et al., 2015). Finally, we used a multiplex assay that does not cover the whole proteome and there may be other sex-dependent differences in protein expression that we could not observe because of the limited number of proteins we measured compared to the whole plasma proteome.
In conclusion, our results show that individuals with LLD have significant sex-specific biological abnormalities. We provide new evidence of sex-dependent effects of protein clusters on different clinical variables and characteristics of the depressive episode. Our results also point to the highly heterogeneous nature of the biological abnormalities in LLD and the importance of taking into consideration sex-specific differences when developing and testing interventions for LLD.
Supplementary Material
Highlights.
Sex has a significant impact on the manifestation and outcomes of MDD in the lifespan.
Older adults with MDD have sex-specific biomarker signatures.
Males have more extensive abnormalities in biological processes than females.
Funding sources:
This work was funded by NIH grants R01MH118311 (Diniz & Tseng), R01 MH083660 (Reynolds, Lenze, Mulsant), and R01AG049369 (Lenze & Wetherell).
Footnotes
Conflict of interests: Dr. Lenze is a consultant for Merck, Prodeo, Boehringer-Ingelheim, Pritikin ICR, and IngenioRxDr. Lenze has a patent application for sigma-1 agonists in COVID-19 treatment. He also receives research support from PCORI, the COVID Early Treatment Fund, Emergent Venture FastGrants, and MagStim. Dr. Wu is consultant for Genzyme, Novartis, Roche, and The Department of Justice. Dr. Karp reports receipt of honorarium from Otsuka for preparation and presentation of a webinar (disease-state, not product-focused) and from NightWare for scientific advising and equity from Aifred Health for scientific advising. Dr. Mulsant holds and receives support from the Labatt Family Chair in Biology of Depression in Late-Life Adults at the University of Toronto. He also receives compensation from the Centre for Addiction and Mental Health (CAMH), Toronto, Ontario. He currently receives research support from Brain Canada, the Canadian Institutes of Health Research, the CAMH Foundation, the US National Institute of Health (NIH), Capital Solution Design LLC (software used in a study founded by CAMH Foundation), and HAPPYneuron (software used in a study founded by Brain Canada). Within the past three years, he has received research support from the Patient-Centered Outcomes Research Institute (PCORI) and he has been an unpaid consultant to Myriad Neuroscience. Dr. Blumberger receives research support from CIHR, N.I.H., Brain Canada and the Temerty Family through the CAMH Foundation and the Campbell Research Institute. He received research support and in-kind equipment support for an investigator-initiated study from Brainsway Ltd. and he is the site principal investigator for one sponsor-initiated study for Brainsway Ltd. He also receives in-kind equipment support from Magventure for investigator-initiated studies. He received medication supplies for an investigator-initiated trial from Indivior. The other co-authors do not have a conflict of interest related to this study.
Ethical statement
We, the authors, confirm that this manuscript was written to meet the highest standards of ethics in publishing scientific content for peer review publications.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
References
- 2000. Diagnostic and statistical manual of mental disorders : DSM-IV-TR. Fourth edition, text revision. Washington, DC: : American Psychiatric Association, [2000] ©2000. [Google Scholar]
- Ancelin ML, Norton J, Canonico M, Scarabin PY, Ritchie K, Ryan J, 2020. Aromatase (CYP19A1) gene variants, sex steroid levels, and late-life depression. Depress Anxiety 37 (2), 146–155. [DOI] [PubMed] [Google Scholar]
- Bekker MH, van Mens-Verhulst J, 2007. Anxiety disorders: sex differences in prevalence, degree, and background, but gender-neutral treatment. Gend Med 4 Suppl B, S178–193. [DOI] [PubMed] [Google Scholar]
- Blokland GAM, Grove J, Chen CY, Cotsapas C, Tobet S, Handa R, St Clair D, Lencz T, Mowry BJ, Periyasamy S, Cairns MJ, Tooney PA, Wu JQ, Kelly B, Kirov G, Sullivan PF, Corvin A, Riley BP, Esko T, Milani L, Jönsson EG, Palotie A, Ehrenreich H, Begemann M, Steixner-Kumar A, Sham PC, Iwata N, Weinberger DR, Gejman PV, Sanders AR, Buxbaum JD, Rujescu D, Giegling I, Konte B, Hartmann AM, Bramon E, Murray RM, Pato MT, Lee J, Melle I, Molden E, Ophoff RA, McQuillin A, Bass NJ, Adolfsson R, Malhotra AK, Martin NG, Fullerton JM, Mitchell PB, Schofield PR, Forstner AJ, Degenhardt F, Schaupp S, Comes AL, Kogevinas M, Guzman-Parra J, Reif A, Streit F, Sirignano L, Cichon S, Grigoroiu-Serbanescu M, Hauser J, Lissowska J, Mayoral F, Müller-Myhsok B, Świątkowska B, Schulze TG, Nöthen MM, Rietschel M, Kelsoe J, Leboyer M, Jamain S, Etain B, Bellivier F, Vincent JB, Alda M, O’Donovan C, Cervantes P, Biernacka JM, Frye M, McElroy SL, Scott LJ, Stahl EA, Landén M, Hamshere ML, Smeland OB, Djurovic S, Vaaler AE, Andreassen OA, Baune BT, Air T, Preisig M, Uher R, Levinson DF, Weissman MM, Potash JB, Shi J, Knowles JA, Perlis RH, Lucae S, Boomsma DI, Penninx B, Hottenga JJ, de Geus EJC, Willemsen G, Milaneschi Y, Tiemeier H, Grabe HJ, Teumer A, Van der Auwera S, Völker U, Hamilton SP, Magnusson PKE, Viktorin A, Mehta D, Mullins N, Adams MJ, Breen G, McIntosh AM, Lewis CM, Hougaard DM, Nordentoft M, Mors O, Mortensen PB, Werge T, Als TD, Børglum AD, Petryshen TL, Smoller JW, Goldstein JM, 2022. Sex-Dependent Shared and Nonshared Genetic Architecture Across Mood and Psychotic Disorders. Biol Psychiatry 91 (1), 102–117. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Bommersbach TJ, Rosenheck RA, Petrakis IL, Rhee TG, 2022. Why are women more likely to attempt suicide than men? Analysis of lifetime suicide attempts among US adults in a nationally representative sample. J Affect Disord 311, 157–164. [DOI] [PubMed] [Google Scholar]
- Diniz BS, Lin CW, Sibille E, Tseng G, Lotrich F, Aizenstein HJ, Reynolds CF, Butters MA, 2016. Circulating biosignatures of late-life depression (LLD): Towards a comprehensive, data-driven approach to understanding LLD pathophysiology. J Psychiatr Res 82, 1–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Diniz BS, Mulsant BH, Reynolds CF 3rd, Blumberger DM, Karp JF, Butters MA, Mendes-Silva AP, Vieira EL, Tseng G, Lenze EJ, 2022. Association of Molecular Senescence Markers in Late-Life Depression With Clinical Characteristics and Treatment Outcome. JAMA Netw Open 5 (6), e2219678. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Diniz BS, Reynolds CF 3rd, Butters MA, Dew MA, Firmo JO, Lima-Costa MF, Castro-Costa E, 2014. The effect of gender, age, and symptom severity in late-life depression on the risk of all-cause mortality: the Bambui Cohort Study of Aging. Depress Anxiety 31 (9), 787–795. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Diniz BS, Sibille E, Ding Y, Tseng G, Aizenstein HJ, Lotrich F, Becker JT, Lopez OL, Lotze MT, Klunk WE, Reynolds CF, Butters MA, 2015. Plasma biosignature and brain pathology related to persistent cognitive impairment in late-life depression. Mol Psychiatry 20 (5), 594–601. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Drevets WC, Wittenberg GM, Bullmore ET, Manji HK, 2022. Immune targets for therapeutic development in depression: towards precision medicine. Nat Rev Drug Discov 21 (3), 224–244. [DOI] [PMC free article] [PubMed] [Google Scholar]
- First M, Spitzer RL, Gibbon M, Williams JBW, 2002. Structured Clinical Interview for DSM-IV-TR Axis I Disorders, Research Version, Patient Edition. (SCID-I/P). Biometrics Research, New York State Psychiatric Institute, New York. [Google Scholar]
- Herrman H, Patel V, Kieling C, Berk M, Buchweitz C, Cuijpers P, Furukawa TA, Kessler RC, Kohrt BA, Maj M, McGorry P, Reynolds CF 3rd, Weissman MM, Chibanda D, Dowrick C, Howard LM, Hoven CW, Knapp M, Mayberg HS, Penninx B, Xiao S, Trivedi M, Uher R, Vijayakumar L, Wolpert M, 2022. Time for united action on depression: a Lancet-World Psychiatric Association Commission. Lancet 399 (10328), 957–1022. [DOI] [PubMed] [Google Scholar]
- Ho TC, Sisk LM, Kulla A, Teresi GI, Hansen MM, Wu H, Gotlib IH, 2021. Sex differences in myelin content of white matter tracts in adolescents with depression. Neuropsychopharmacology 46 (13), 2295–2303. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Horvath S, Dong J, 2008. Geometric interpretation of gene coexpression network analysis. PLoS Comput Biol 4 (8), e1000117. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Huo Z, Tang S, Park Y, Tseng G, 2020. P-value evaluation, variability index and biomarker categorization for adaptively weighted Fisher’s meta-analysis method in omics applications. Bioinformatics 36 (2), 524–532. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Janssen EPCJ, Köhler S, Geraets AFJ, Stehouwer CDA, Schaper NC, Sep SJS, Henry RMA, Kallen C.J.H.v.d., Schalkwijk CG, Koster A, Verhey FR, Schram MT, 2021. Low-grade inflammation and endothelial dysfunction predict four-year risk and course of depressive symptoms: The Maastricht study. Brain, Behavior, and Immunity 97, 61–67. [DOI] [PubMed] [Google Scholar]
- Kang HJ, Park Y, Yoo KH, Kim KT, Kim ES, Kim JW, Kim SW, Shin IS, Yoon JS, Kim JH, Kim JM, 2020. Sex differences in the genetic architecture of depression. Sci Rep 10 (1), 9927. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kennedy SH, 2008. Core symptoms of major depressive disorder: relevance to diagnosis and treatment. Dialogues Clin Neurosci 10 (3), 271–277. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kornstein SG, Schatzberg AF, Thase ME, Yonkers KA, McCullough JP, Keitner GI, Gelenberg AJ, Davis SM, Harrison WM, Keller MB, 2000. Gender differences in treatment response to sertraline versus imipramine in chronic depression. Am J Psychiatry 157 (9), 1445–1452. [DOI] [PubMed] [Google Scholar]
- Korotkevich G, Sukhov V, Budin N, Shpak B, Artyomov MN, Sergushichev A, 2021. Fast gene set enrichment analysis. BioRxiv, 060012. [Google Scholar]
- Langfelder P, Zhang B, Horvath S, 2008. Defining clusters from a hierarchical cluster tree: the Dynamic Tree Cut package for R. Bioinformatics 24 (5), 719–720. [DOI] [PubMed] [Google Scholar]
- Lenze EJ, Mulsant BH, Blumberger DM, Karp JF, Newcomer JW, Anderson SJ, Dew MA, Butters MA, Stack JA, Begley AE, Reynolds CF 3rd, 2015. Efficacy, safety, and tolerability of augmentation pharmacotherapy with aripiprazole for treatment-resistant depression in late life: a randomised, double-blind, placebo-controlled trial. Lancet 386 (10011), 2404–2412. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Liberzon A, Subramanian A, Pinchback R, Thorvaldsdóttir H, Tamayo P, Mesirov JP, 2011. Molecular signatures database (MSigDB) 3.0. Bioinformatics 27 (12), 1739–1740. [DOI] [PMC free article] [PubMed] [Google Scholar]
- López-Otín C, Blasco MA, Partridge L, Serrano M, Kroemer G, 2013. The hallmarks of aging. Cell 153 (6), 1194–1217. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Miller MD, Paradis CF, Houck PR, Mazumdar S, Stack JA, Rifai AH, Mulsant B, Reynolds CF 3rd, 1992. Rating chronic medical illness burden in geropsychiatric practice and research: application of the Cumulative Illness Rating Scale. Psychiatry Res 41 (3), 237–248. [DOI] [PubMed] [Google Scholar]
- Montgomery SA, Asberg M, 1979. A new depression scale designed to be sensitive to change. Br J Psychiatry 134, 382–389. [DOI] [PubMed] [Google Scholar]
- Ramsey JM, Cooper JD, Bot M, Guest PC, Lamers F, Weickert CS, Penninx BW, Bahn S, 2016. Sex Differences in Serum Markers of Major Depressive Disorder in the Netherlands Study of Depression and Anxiety (NESDA). PLoS One 11 (5), e0156624. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Riecher-Rossler A, 2010. Prospects for the classification of mental disorders in women. Eur Psychiatry 25 (4), 189–196. [DOI] [PubMed] [Google Scholar]
- Schuch JJ, Roest AM, Nolen WA, Penninx BW, de Jonge P, 2014. Gender differences in major depressive disorder: results from the Netherlands study of depression and anxiety. J Affect Disord 156, 156–163. [DOI] [PubMed] [Google Scholar]
- Seitz-Holland J, Mulsant BH Iii, C.F.R., Blumberger DM, Karp JF, Butters MA, Mendes-Silva AP, Vieira EL, Tseng G, Lenze EJ, Diniz BS, 2023. Major depression, physical health and molecular senescence markers abnormalities. Nature Mental Health 1 (3), 200–209. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Seney ML, Huo Z, Cahill K, French L, Puralewski R, Zhang J, Logan RW, Tseng G, Lewis DA, Sibille E, 2018. Opposite Molecular Signatures of Depression in Men and Women. Biol Psychiatry 84 (1), 18–27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Sheehan DV, Lecrubier Y, Sheehan KH, Amorim P, Janavs J, Weiller E, Hergueta T, Baker R, Dunbar GC, 1998. The Mini-International Neuropsychiatric Interview (M.I.N.I.): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10. J Clin Psychiatry 59 Suppl 20, 22–33;quiz 34–57. [PubMed] [Google Scholar]
- Sramek JJ, Murphy MF, Cutler NR, 2016. Sex differences in the psychopharmacological treatment of depression. Dialogues Clin Neurosci 18 (4), 447–457. [DOI] [PMC free article] [PubMed] [Google Scholar]
- van Buuren S, Groothuis-Oudshoorn K, 2011. mice: Multivariate Imputation by Chained Equations in R. Journal of Statistical Software 45 (3), 1–67. [Google Scholar]
- Wetherell JL, Ripperger HS, Voegtle M, Ances BM, Balota D, Bower ES, Depp C, Eyler L, Foster ER, Head D, Hershey T, Hickman S, Kamantigue N, Klein S, Miller JP, Yingling MD, Nichols J, Nicol GE, Patterson BW, Rodebaugh TL, Shimony JS, Snyder A, Stephens M, Tate S, Uhrich ML, Wing D, Wu GF, Lenze EJ, 2020. Mindfulness, Education, and Exercise for age-related cognitive decline: Study protocol, pilot study results, and description of the baseline sample. Clin Trials 17 (5), 581–594. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Whittle S, Lichter R, Dennison M, Vijayakumar N, Schwartz O, Byrne ML, Simmons JG, Yücel M, Pantelis C, McGorry P, Allen NB, 2014. Structural brain development and depression onset during adolescence: a prospective longitudinal study. Am J Psychiatry 171 (5), 564–571. [DOI] [PubMed] [Google Scholar]
- Willforss J, Chawade A, Levander F, 2019. NormalyzerDE: Online Tool for Improved Normalization of Omics Expression Data and High-Sensitivity Differential Expression Analysis. J Proteome Res 18 (2), 732–740. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.