Integrative Analysis of Major Depressive Disorder and Ovarian Cancer: From Genetic Association to Single-Cell Mechanisms

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Summary:
Chen Liu et al. identified a potential link between major depressive disorder (MDD) and ovarian cancer through the CLSTN3 gene, highlighting the genetic variant rs3759416 as a possible contributing factor. The study is well designed and employs a robust multi-omics, integrative computational biology approach. It is both relevant and scientifically valuable. In particular, the integration of single-cell RNA sequencing with Mendelian randomization, the Harmony algorithm for batch correction, the scMetabolism package for comparing cellular metabolic activities, and additional tools for identifying differentially expressed genes provides a strong methodological framework. The identification of CLSTN3 as a potential molecular link between MDD and ovarian cancer offers novel insight into an emerging and important area of research. With minor revisions, I believe this manuscript would be suitable for publication. Addressing the following points outlined below will further enhance its clarity, rigor, and overall impact.

Major Comments
1.
The authors should provide a clear rationale for selecting ovarian cancer in relation to MDD. While the association between depression and cancer progression is increasingly recognized, it is unclear why ovarian cancer was specifically chosen over other cancer types (for example, prostate, lung, bladder, or cervical cancer). Including epidemiological, clinical, or biological justification in the Introduction would strengthen the study’s context and help readers better understand its focus and significance.
2.
In the Introduction, the authors state that “MDD is now recognized not merely as a psychiatric condition but as a systemic disease that predisposes individuals to coronary heart disease, diabetes, arthritis, and increasingly, cancer development.” However, the references provided primarily support the association with cancer and do not adequately substantiate the links with coronary heart disease, diabetes, and arthritis. The authors should include appropriate and up-to-date references for all listed conditions to ensure accuracy and completeness.
3.
The authors should clarify whether MDD similarly affects males and whether it is associated with cancer risk in men. If evidence exists, relevant information and references should be included in the Introduction to support the role of MDD in cancer development across both sexes.
4.
The Introduction would benefit from a broader discussion of how MDD may promote cancer progression through various biological mechanisms (e.g., immune dysregulation, inflammation, neuroendocrine pathways), not limited to ovarian cancer but extending to other solid tumors, supported by appropriate references.
5.
The manuscript states that depression increases activated cytotoxic (CD3⁺CD8⁺CD69⁺) and exhausted (CD3⁺Lag3⁺) T cells while reducing naïve and memory B cells; however, the supporting references are insufficient to substantiate this claim. Additional high-quality references should be included to support this statement.
6.
The authors used the GSE264489 dataset for ovarian cancer and non-cancer control samples, as well as the PRJCA032578 dataset. They should cite the original publications associated with these datasets, as indicated on the dataset repository pages.
7.
The role of the CLSTN3 gene identified in this study should be discussed in greater depth in the context of ovarian cancer and other cancer types. Supporting literature describing its potential oncogenic or biological role should be included.
8.
The authors did not demonstrate that CLSTN3 expression is elevated in ovarian cancer compared to normal controls. It is recommended to validate CLSTN3 expression using publicly available databases such as GEPIA, TIMER, and cBioPortal, and to include survival analyses from these platforms. This would strengthen the evidence supporting the oncogenic relevance of CLSTN3 and its potential role linking MDD to increased ovarian cancer risk.

Minor Comments
9.
The sample size is relatively small: treatment-naive ovarian cancer cases (n = 3), non-cancer controls (n = 6), MDD patients (n = 8), and healthy controls (n = 8). Although single-cell RNA sequencing enables analysis at the cellular level, increasing the number of patient samples (around 20 per group) would improve the robustness and generalizability of the findings.
10.
Several references throughout the manuscript appear to require verification or additional supporting citations. The authors are encouraged to carefully review and update the reference list to ensure that relevant and reliable sources support all statements.

Author Response

Dear reviewer,                                  

Thank you very much for giving us the opportunity to revise our manuscript. We are grateful to you for their thorough, constructive, and insightful comments, which have helped us improve the quality and rigor of our study considerably.

Below we provide a detailed, point-by-point response. All changes made to the manuscript are shown in Track Changes mode.

We hope you will find the revisions satisfactory, and that the manuscript can now be considered for publication. Please do not hesitate to contact us if anything further is needed.

 

Reviewer1#

Chen Liu et al. identified a potential link between major depressive disorder (MDD) and ovarian cancer through the CLSTN3 gene, highlighting the genetic variant rs3759416 as a possible contributing factor. The study is well designed and employs a robust multi-omics, integrative computational biology approach. It is both relevant and scientifically valuable. In particular, the integration of single-cell RNA sequencing with Mendelian randomization, the Harmony algorithm for batch correction, the scMetabolism package for comparing cellular metabolic activities, and additional tools for identifying differentially expressed genes provides a strong methodological framework. The identification of CLSTN3 as a potential molecular link between MDD and ovarian cancer offers novel insight into an emerging and important area of research. With minor revisions, I believe this manuscript would be suitable for publication. Addressing the following points outlined below will further enhance its clarity, rigor, and overall impact.

1.The authors should provide a clear rationale for selecting ovarian cancer in relation to MDD. While the association between depression and cancer progression is increasingly recognized, it is unclear why ovarian cancer was specifically chosen over other cancer types (for example, prostate, lung, bladder, or cervical cancer). Including epidemiological, clinical, or biological justification in the Introduction would strengthen the study’s context and help readers better understand its focus and significance.

Response: Thank you for this important suggestion. We agree that a clear rationale for selecting ovarian cancer (OC) over other malignancies is essential for contextualizing our study. We have now added a dedicated paragraph in the introduction that outlines the epidemiological, clinical, and biological justification for focusing on OC in relation to major depressive disorder (MDD). The revised text highlights that: (1) OC patients exhibit a high prevalence of depression (~35%) and comorbid depression is associated with a 94% increased mortality risk. (2) OC 's poor prognosis and unique psychological burden distinguish it from other solid tumors; and (3) OC and MDD share convergent pathophysiological pathways involving HPA-axis dysregulation and pro-inflammatory cytokines (Pages 2-3, Lines 54-66). Thanks.

2.In the Introduction, the authors state that “MDD is now recognized not merely as a psychiatric condition but as a systemic disease that predisposes individuals to coronary heart disease, diabetes, arthritis, and increasingly, cancer development.” However, the references provided primarily support the association with cancer and do not adequately substantiate the links with coronary heart disease, diabetes, and arthritis. The authors should include appropriate and up-to-date references for all listed conditions to ensure accuracy and completeness.

Response: We sincerely thank the reviewer for this careful observation. We fully agree that the original references were insufficient to support the full spectrum of systemic comorbidities associated with MDD. The sentence in the Introduction (Page 2, Lines 43–45) and the reference list have been updated accordingly (References 5-7). We believe the statement is now accurately and comprehensively supported across all listed conditions, and we hope the reviewer finds this revision satisfactory. Thanks.

3.The authors should clarify whether MDD similarly affects males and whether it is associated with cancer risk in men. If evidence exists, relevant information and references should be included in the Introduction to support the role of MDD in cancer development across both sexes.

Response: Thank you for this important point. We have revised the Introduction to clarify that MDD affects both sexes, although women experience approximately twice the lifetime prevalence. We have added references showing that depression is also linked to increased cancer risk in men, including associations with prostate, colorectal, and lung cancer (Page 2, Lines 48–49). We further explain that while the MDD–cancer interplay extends to both sexes, this study focuses on ovarian cancer because it is a female-specific malignancy with particularly high rates of comorbid depression and unique stress-related biological mechanisms (HPA-axis dysregulation, peritoneal inflammation), offering distinct mechanistic insights. Thanks.

4.The Introduction would benefit from a broader discussion of how MDD may promote cancer progression through various biological mechanisms (e.g., immune dysregulation, inflammation, neuroendocrine pathways), not limited to ovarian cancer but extending to other solid tumors, supported by appropriate references.

Response: Thank you for this suggestion. We have expanded the Introduction to include a broader discussion of the biological mechanisms linking MDD to cancer progression across solid tumors, encompassing immune dysregulation, chronic inflammation, and neuroendocrine pathway alterations (e.g., HPA-axis activation and sympathetic nervous system signaling), with supporting references (Page 2, Lines 46–53). This provides the necessary mechanistic context before narrowing the focus to ovarian cancer. Thanks.

5. 
   The manuscript states that depression increases activated cytotoxic (CD3⁺CD8⁺CD69⁺) and exhausted (CD3⁺Lag3⁺) T cells while reducing naïve and memory B cells; however, the supporting references are insufficient to substantiate this claim. Additional high-quality references should be included to support this statement.

Response: Thank you for this comment. We have added high-quality references to substantiate these immunological findings. Specifically, we now cite Rachayon et al. (2024), who demonstrated that CD8⁺CD69⁺ T cells are expanded in acute-phase MDD and associated with pro-inflammatory responses, and that LAG3⁺ expression marks exhausted T cells in depressed patients (Reference 20). We also cite Ahmetspahic et al. (2018), who reported significantly reduced naïve B cell proportions in MDD patients (Reference 21). These references have been incorporated into the revised text to strengthen the evidence. Thanks.

6.The authors used the GSE264489 dataset for ovarian cancer and non-cancer control samples, as well as the PRJCA032578 dataset. They should cite the original publications associated with these datasets, as indicated on the dataset repository pages.

Response: We thank the reviewer for this important comment regarding data attribution. We fully agree that proper citation of the original publications associated with publicly available datasets is essential for scientific transparency and to acknowledge the contributions of the data generators. We have added citations to the original publications for both GSE264489 and PRJCA032578 in the Methods section (Page 4, Lines 88-90). and Reference list (Reference 23-24). These citations are now included alongside the dataset accession numbers to ensure proper attribution. Thanks.

7.The role of the CLSTN3 gene identified in this study should be discussed in greater depth in the context of ovarian cancer and other cancer types. Supporting literature describing its potential oncogenic or biological role should be included.

Response: Thank you for this constructive suggestion. We have expanded the Discussion to provide a comprehensive analysis of CLSTN3's biological and oncogenic roles. Specifically, we now describe:

(1) CLSTN3 as a cadherin superfamily member involved in cell adhesion and ECM remodeling; (2) evidence from colorectal cancer linking elevated CLSTN3 to tumor invasiveness and poor prognosis; (3) proteomic data showing CLSTN3 shedding from ovarian cancer cell lines, supporting its relevance to peritoneal dissemination; and (4) epigenetic regulation of CLSTN3 in hepatocellular carcinoma (Page 13, Lines 359-373). We also acknowledge that while CLSTN3's oncogenic role is supported in other malignancies, its specific function in ovarian cancer requires further experimental validation. Thanks.

 

8.The authors did not demonstrate that CLSTN3 expression is elevated in ovarian cancer compared to normal controls. It is recommended to validate CLSTN3 expression using publicly available databases such as GEPIA, TIMER, and cBioPortal, and to include survival analyses from these platforms. This would strengthen the evidence supporting the oncogenic relevance of CLSTN3 and its potential role linking MDD to increased ovarian cancer risk.
Response: We sincerely appreciate the reviewer's constructive suggestion regarding the validation of CLSTN3 expression in ovarian cancer (OC) using publicly available databases. Thank you for this valuable suggestion.

To address this, we have incorporated additional analyses using publicly available databases. As shown in Figure 8A–D, CLSTN3 expression is significantly upregulated in ovarian cancer tissues compared with normal controls, and elevated CLSTN3 expression is associated with poor prognosis (Page 11, Lines 306-313). These findings strengthen the evidence supporting the oncogenic relevance of CLSTN3 in ovarian cancer. Thanks.

Minor Comments
9. The sample size is relatively small: treatment-naive ovarian cancer cases (n = 3), non-cancer controls (n = 6), MDD patients (n = 8), and healthy controls (n = 8). Although single-cell RNA sequencing enables analysis at the cellular level, increasing the number of patient samples (around 20 per group) would improve the robustness and generalizability of the findings.

Response: We sincerely appreciate the reviewer's concern regarding the sample size in our single-cell RNA sequencing (scRNA-seq) analysis. We acknowledge this limitation. The modest sample size reflects the inherent challenges of scRNA-seq in clinical oncology, including strict quality control requirements for fresh surgical specimens, high costs, and the difficulty of recruiting patients with comorbid MDD and OC. However, scRNA-seq captures transcriptomic profiles at single-cell resolution, with each tissue sample yielding thousands of cells, providing substantial statistical power for cell-type-specific analyses.

We have also added an explicit discussion of this limitation and stated that future multi-center studies with larger cohorts are warranted (Pages 13-14, Lines 383-403). Thanks.

10. 
    Several references throughout the manuscript appear to require verification or additional supporting citations. The authors are encouraged to carefully review and update the reference list to ensure that relevant and reliable sources support all statements.

Response: Thank you for this important suggestion. We have conducted a comprehensive review of the entire reference list in the revised manuscript. Specifically, we verified all original citations against the published articles, replaced outdated or misquoted references, and supplemented with recent high-quality studies. All in-text citations have been cross-checked with the reference list for consistency. We believe these revisions have improved the scholarly rigor of the manuscript. Thanks.

Reviewer 2 Report

Comments and Suggestions for Authors

1. The scRNA‑seq dataset includes only 3 treatment‑naïve ovarian cancer patients, 6 non‑cancer controls, 8 MDD patients, and 8 healthy controls. For a disease as heterogeneous as ovarian cancer, n=3 is grossly insufficient to draw any meaningful conclusions about immune cell populations or gene expression changes. The results are almost certainly underpowered and subject to high variability.
2. The MR analysis uses publicly available GWAS summary statistics with large sample sizes, which is acceptable for the genetic component. However, the initial gene selection (554 DEGs from CD8_EM T cells) is based on the tiny scRNA‑seq cohort, meaning that the entire downstream MR pipeline is built on an unstable foundation. False positives due to small sample size are highly likely.
3.The authors identify 554 DEGs in CD8_EM T cells, then test 554 genes for association with ovarian cancer risk using MR. No correction for multiple comparisons (e.g., Bonferroni, FDR) is reported. With 554 tests, the probability of finding at least one “significant” association by chance alone is extremely high. The nominal p‑values presented (e.g., CLSTN3 OR 1.21, 95% CI 1.03–1.43, p=0.019) would not survive correction for 554 tests. The claim that CLSTN3 is a “significant risk factor” is therefore unsupported.
4. MR can suggest a causal relationship if the instrumental variables are valid, but it does not prove biological mechanism. The authors conclude that “CLSTN3 is a promising candidate gene that may influence malignant progression” and even refer to it as a “therapeutic target.” MR alone cannot establish a target for drug development; it only provides genetic evidence of association. The effect size (OR 1.21) is modest and, even if real, would have limited clinical relevance.
5. The colocalization analysis (PPH4 = 99.99%) is presented as strong evidence, but colocalization of an eQTL with a GWAS signal does not prove that the gene is the causal effector; it only suggests that the same variant influences both expression and disease risk. Many colocalized genes are not drug targets.
6. The title and abstract claim CLSTN3 as a “therapeutic target,” yet the manuscript contains no experimental validation of CLSTN3 function in either depression or ovarian cancer. There are no in vitro or in vivo experiments (e.g., knockdown or overexpression in OC cell lines, animal models, or immune cell assays). The “potential physiological function” section (Figure 6) is purely descriptive and does not test any hypothesis.
7. The authors show that CLSTN3 is expressed in CD8_EM T cells and that its expression correlates with certain metabolic pathways, but these are correlative observations. There is no evidence that CLSTN3 directly affects T‑cell function, tumor growth, or depression-related biology.
8. The study begins with the premise that depression increases ovarian cancer risk, citing epidemiological literature. However, the analysis does not actually test whether depression (or any depression‑related phenotype) causally affects CLSTN3 expression or OC risk. The MR analysis uses eQTL data from CD8_EM T cells (which are not specific to depression) and GWAS data for ovarian cancer. There is no MR analysis of depression as an exposure. Therefore, the connection between MDD and CLSTN3 is entirely based on the observation that CD8_EM T cells are expanded in both conditions (from the tiny scRNA‑seq dataset). This is a weak associative link, not a causal chain.
9. The phrase “neuro‑immuno‑oncology” is grandiose relative to the evidence. The manuscript does not investigate neural mechanisms, nor does it provide any neurobiological data.
10. The integration of datasets from different sources (GSE264489 for OC, PRJCA032578 for MDD) without clear demonstration of batch correction adequacy is concerning. The authors used Harmony, but no metrics are provided to show that biological variation is preserved and batch effects removed. Given the small number of samples per group, over‑correction is a risk.
11. The definition of CD8_EM T cells is based on marker genes, but no validation (e.g., flow cytometry or independent dataset) is provided. The claim that this population is “consistently expanded” in both MDD and OC is based on proportional changes in a handful of samples; with n=3 for OC, this is not convincing.
12. The raw scRNA‑seq data accessions are given, but the code for the full analysis pipeline is not provided. Key parameters for clustering, dimensionality reduction, and differential expression are not sufficiently detailed to allow replication.
13. The MR analysis uses GWAS summary statistics from the GWAS Catalog, but the exact identifiers (e.g., GCT90436511) are not standard; the authors should provide the original study citations and ensure that the GWAS data are publicly accessible with clear documentation.
14. The conclusion states: “This finding provides new insights for future intervention strategies and emphasizes the potential impact of mental health on cancer outcomes.” No intervention strategy is proposed or tested. The manuscript does not provide any evidence that targeting CLSTN3 would improve cancer outcomes or mental health. Such statements are misleading.

Author Response

Dear reviewer,

Thank you very much for giving us the opportunity to revise our manuscript. We are grateful to you for their thorough, constructive, and insightful comments, which have helped us improve the quality and rigor of our study considerably.

Below we provide a detailed, point-by-point response. All changes made to the manuscript are shown in Track Changes mode.

We hope you will find the revisions satisfactory, and that the manuscript can now be considered for publication. Please do not hesitate to contact us if anything further is needed.

 

 

Reviewer2#

1. The scRNA‑seq dataset includes only 3 treatment‑naïve ovarian cancer patients, 6 non‑cancer controls, 8 MDD patients, and 8 healthy controls. For a disease as heterogeneous as ovarian cancer, n=3 is grossly insufficient to draw any meaningful conclusions about immune cell populations or gene expression changes. The results are almost certainly underpowered and subject to high variability.

Response: Thank you for this critical assessment. We fully acknowledge that the scRNA-seq discovery cohort is small and insufficient to capture the full inter-patient heterogeneity of ovarian cancer. We have explicitly stated this limitation in the revised Discussion (Pages 13-14, Lines 383-403), emphasizing that all scRNA-seq findings should be interpreted as exploratory.

To mitigate this, we have: (1) validated key CLSTN3 findings in an independent bulk RNA-seq dataset with substantially larger sample sizes; (2) performed orthogonal flow cytometry validation in an independent cohort (5 OC vs. 5 healthy controls); and (3) prioritized signatures showing cross-disease consistency. We agree that larger, multi-center scRNA-seq cohorts are warranted for definitive conclusions. Thanks.

2. 
   The MR analysis uses publicly available GWAS summary statistics with large sample sizes, which is acceptable for the genetic component. However, the initial gene selection (554 DEGs from CD8_EM T cells) is based on the tiny scRNA‑seq cohort, meaning that the entire downstream MR pipeline is built on an unstable foundation. False positives due to small sample size are highly likely.

Response: Thank you for this critical concern. We acknowledge that the scRNA-seq discovery cohort is small. However, we wish to clarify that the MR pipeline is structurally separated from the discovery phase: the 554 DEGs served solely to nominate candidate genes, while the MR instruments were extracted from independent, large-scale eQTL databases (eQTLGen, n = 31,684) with stringent filtering (genome-wide significance, F-statistic > 10, LD clumping r² < 0.001).

Importantly, CLSTN3 replicated as one of only four genes showing consistent significant associations across two independent ovarian cancer GWAS datasets, and reverse MR excluded reverse causation. We have added an explicit statement in the Discussion acknowledging this limitation (Pages 13-14, Lines 383-403). Thanks.

3.The authors identify 554 DEGs in CD8_EM T cells, then test 554 genes for association with ovarian cancer risk using MR. No correction for multiple comparisons (e.g., Bonferroni, FDR) is reported. With 554 tests, the probability of finding at least one “significant” association by chance alone is extremely high. The nominal p‑values presented (e.g., CLSTN3 OR 1.21, 95% CI 1.S03–1.43, p=0.019) would not survive correction for 554 tests. The claim that CLSTN3 is a “significant risk factor” is therefore unsupported.

Response: Thank you for this critical methodological comment. We agree that multiple comparison correction is essential. Since 24 genes of 554 DEGs significantly linked to ovarian cancer susceptibility, we have now applied Benjamini-Hochberg FDR and Bonferroni correction to 24 MR analyses (Page 9, Lines 253-255). no genes survived correction (FDR q < 0.05: 0 genes; Bonferroni-corrected p < 0.05: 0 genes), including CLSTN3 (nominal p = 0.024, FDR q = 0.492) (Table1). Therefore, we have removed all claims that CLSTN3 is a "significant risk factor" based on MR alone. Instead, we now describe CLSTN3 as a candidate molecule nominated by convergent evidence:

(1) nominal genetic association; (2) strong colocalization with the OC GWAS signal (PPH4 = 99.99%); (3) specific expression in CD8_EM T cells; and (4) functional validation showing siRNA-mediated knockdown significantly inhibits OC cell proliferation in vitro. We have added an explicit statement in the Discussion acknowledging that the MR evidence is hypothesis-generating and that the decision to pursue CLSTN3 was based on the integration of these multiple lines of evidence, not on the nominal p-value alone. "We also note that similar MR studies in this field have historically relied on nominal p-values for gene prioritization (Reference 27); however, we fully agree with the reviewer that rigorous multiple comparison correction should be reported, and we have now done so." Thanks.

4. 
   MR can suggest a causal relationship if the instrumental variables are valid, but it does not prove biological mechanism. The authors conclude that “CLSTN3 is a promising candidate gene that may influence malignant progression” and even refer to it as a “therapeutic target.” MR alone cannot establish a target for drug development; it only provides genetic evidence of association. The effect size (OR 1.21) is modest and, even if real, would have limited clinical relevance.

Response: Thank you for this incisive critique. We fully agree that MR provides genetic evidence of association, not biological mechanism or drug target validation. We have made the following revisions:

(1) We no longer refer to CLSTN3 as a "therapeutic target" and now describe it as a "potential candidate molecule" throughout the manuscript. (2) To move beyond association, we have added in vitro loss-of-function experiments showing that siRNA-mediated CLSTN3 knockdown significantly inhibits ovarian cancer cell proliferation (CCK-8 assay, revised Figure 9) (Page 11, Lines 317-319). (3) We have explicitly distinguished "genetic candidate" from "validated drug target" in the Discussion, noting that target tractability, safety profiling, and in vivo efficacy are beyond the scope of this study (Pages 13-14, Lines 383-403). Thanks.

5. 
   The colocalization analysis (PPH4 = 99.99%) is presented as strong evidence, but colocalization of an eQTL with a GWAS signal does not prove that the gene is the causal effector; it only suggests that the same variant influences both expression and disease risk. Many colocalized genes are not drug targets.

Response: Thank you for this important clarification. We fully agree that colocalization indicates a shared genetic signal rather than proving CLSTN3 is the causal effector or a validated drug target. We have revised the manuscript to accurately interpret the colocalization evidence: we now state that rs3759416 influences both CLSTN3 expression and ovarian cancer risk, but we explicitly caution that this does not establish CLSTN3 as the causal mediator. Accordingly, we have removed all language describing CLSTN3 as a "therapeutic target" and now refer to it as a "candidate molecule" or "potential shared immune factor."

We have also added a limitation statement clarifying that colocalization alone is insufficient to infer therapeutic relevance, and that functional validation in vivo and vitro is required to determine whether CLSTN3 is indeed the effector gene underlying the observed association (Pages 13-14, Lines 383-403). Thanks.

 

6. The title and abstract claim CLSTN3 as a “therapeutic target,” yet the manuscript contains no experimental validation of CLSTN3 function in either depression or ovarian cancer. There are no in vitro or in vivo experiments (e.g., knockdown or overexpression in OC cell lines, animal models, or immune cell assays). The “potential physiological function” section (Figure 6) is purely descriptive and does not test any hypothesis.

Response: We thank the reviewer for this important comment. We have now performed in vitro functional validation experiments to test the role of CLSTN3 in ovarian cancer (OC) cell proliferation. Briefly, we knocked down CLSTN3 expression in HOC7 OC cell lines using siRNA and assessed cell proliferation by CCK-8 assay. As shown in [Figure 9], CLSTN3 knockdown significantly inhibited OC cell proliferation compared to the negative control group (p < 0.05). These results provide experimental evidence supporting a functional role of CLSTN3 in promoting OC cell growth, consistent with our MR findings that higher CLSTN3 expression is associated with increased ovarian cancer risk.

Accordingly, we have revised the title to " Integrative Analysis of Major Depressive Disorder and Ovarian Cancer: From Genetic Association to Single-Cell Mechanisms" and modified the abstract and discussion to reflect the preliminary nature of these findings. We have also added a paragraph in the Discussion acknowledging that in vivo validation and mechanistic studies are still needed (Pages 13-14, Lines 383-403). Thanks.

7. 
   The authors show that CLSTN3 is expressed in CD8_EM T cells and that its expression correlates with certain metabolic pathways, but these are correlative observations. There is no evidence that CLSTN3 directly affects T‑cell function, tumor growth, or depression-related biology.

Response: Thank you for this important suggestion. We fully agree that correlation does not imply causation and have revised the manuscript to carefully distinguish associative evidence from functional validation.

(1) Regarding tumor growth, we have performed siRNA-mediated CLSTN3 knockdown in ovarian cancer cells, which significantly inhibited cell proliferation in vitro (CCK-8 assay, revised Figure 9), providing experimental evidence beyond pure correlation. (2) Regarding T cell function, we acknowledge that we have not directly manipulated CLSTN3 in CD8_EM T cells; these findings are now explicitly labeled as hypothesis-generating in the revised text. (3) Regarding depression, we clarify that the CLSTN3–depression link is based on genetic association and comorbidity patterns, not direct biological validation, and we have removed any language implying a direct molecular effect on depression pathophysiology. All correlative observations are now framed with appropriate caution throughout the manuscript (Pages 13-14, Lines 383-403). Thanks.

8. 
   The study begins with the premise that depression increases ovarian cancer risk, citing epidemiological literature. However, the analysis does not actually test whether depression (or any depression‑related phenotype) causally affects CLSTN3 expression or OC risk. The MR analysis uses eQTL data from CD8_EM T cells (which are not specific to depression) and GWAS data for ovarian cancer. There is no MR analysis of depression as an exposure. Therefore, the connection between MDD and CLSTN3 is entirely based on the observation that CD8_EM T cells are expanded in both conditions (from the tiny scRNA‑seq dataset). This is a weak associative link, not a causal chain.

Response: Thank you for this critical comment. We acknowledge that our original manuscript lacked a formal MR analysis with depression as the exposure. We initially attempted two-sample MR using MDD GWAS summary statistics; however, we encountered a methodological concern regarding the large number of available SNPs (>200), which raised uncertainty about the appropriateness of the iterative MR approach under these conditions.

To robustly explore the depression–ovarian cancer association, we instead analyzed data from the UK Biobank, a large-scale prospective cohort, which confirmed that depression is significantly associated with increased ovarian cancer risk, exhibiting a clear dose–response relationship (Figure 1) (Pages 7-8, Lines 196-207). We have revised the manuscript to clearly distinguish between this epidemiological association and the hypothesis-generating role of CLSTN3 identified through scRNA-seq and eQTL-MR. We explicitly state that we have not demonstrated a direct causal effect of depression on CLSTN3 expression, and that CLSTN3's role as a mechanistic link between MDD and OC remains hypothetical and requires further validation. Thanks.

9. 
   The phrase “neuro‑immuno‑oncology” is grandiose relative to the evidence. The manuscript does not investigate neural mechanisms, nor does it provide any neurobiological data.

Response: Thank you for this important comment. We fully agree that the term was inappropriate given the scope of our data. We have revised the title to: "Integrative Analysis of Major Depressive Disorder and Ovarian Cancer: From Genetic Association to Single-Cell Mechanisms."

We have also removed all instances of "neuro-immuno-oncology" from the Abstract, Introduction, and Discussion, replacing them with precise descriptions such as "comorbidity between depression and cancer" and "shared immune dysregulation." The revised title now accurately reflects our study's focus on causal inference and immune characterization. Thanks.

10. 
    The integration of datasets from different sources (GSE264489 for OC, PRJCA032578 for MDD) without clear demonstration of batch correction adequacy is concerning. The authors used Harmony, but no metrics are provided to show that biological variation is preserved, and batch effects removed. Given the small number of samples per group, over‑correction is a risk.

Response: We thank the reviewer for this important concern. We now provide the key metric demonstrating batch correction adequacy: LISI (batch) improved from 1.63 (pre-correction) to 1.99 (post-correction), where a value of 2 indicates perfect batch mixing(Page 4, Lines 106-112).

This confirms effective batch integration without over-correction, while biological variation is preserved as evidenced by distinct cell-type clusters in UMAP and retained specificity of canonical markers. All analysis scripts could be deposited at the journal's data repository. Thanks.

11. 
    The definition of CD8_EM T cells is based on marker genes, but no validation (e.g., flow cytometry or independent dataset) is provided. The claim that this population is “consistently expanded” in both MDD and OC is based on proportional changes in a handful of samples; with n=3 for OC, this is not convincing.

Response: Thank you for this excellent suggestion. We have performed flow cytometric analysis of PBMCs from 5 healthy controls and 5 ovarian cancer patients. As shown in revised Figure 8E-8F, CLSTN3 protein levels were significantly elevated in activated CD8⁺ T cells from OC patients compared to controls (Page 11, Lines 313-316).

We wish to note a technical clarification: Due to the limited availability of reliable CD45RA/CCR7 antibodies for our panel, CD8⁺ effector memory T cells were functionally identified as CD3⁺CD8⁺TNF-α⁺ cells following PMA/ionomycin stimulation, rather than by the canonical surface markers. While this approach captures activated/effector CD8⁺ T cells with cytokine-producing capacity rather than the strictly defined CD45RA⁻CCR7⁻ subset, this population reflects the functional effector state relevant to our study (Page 6, Lines 152-161). We have revised the Methods and Results to explicitly state this limitation and acknowledge that future validation using a comprehensive memory-marker panel (CD45RA/CCR7) is warranted (Pages 13-14, Lines 383-403). Thanks.
12. The raw scRNA‑seq data accessions are given, but the code for the full analysis pipeline is not provided. Key parameters for clustering, dimensionality reduction, and differential expression are not sufficiently detailed to allow replication.

Response: Thank you for this important comment. All analysis scripts could been deposited at the journal's data repository, including all scripts for scRNA-seq preprocessing, integration, clustering, and MR analysis, along with a README documenting software versions and execution instructions.

We have also expanded the Methods section to specify key parameters: Seurat v4.3.0 with SCTransform normalization, Harmony integration (theta = 2), PCA (dims = 1:30), UMAP (dims = 1:20, min.dist = 0.3), FindNeighbors (dims = 1:30), FindClusters (resolution = 0.5) (Page 4, Lines 102-106). Thanks.

13. 
    The MR analysis uses GWAS summary statistics from the GWAS Catalog, but the exact identifiers (e.g., GCT90436511) are not standard; the authors should provide the original study citations and ensure that the GWAS data are publicly accessible with clear documentation.

Response: Thank you for this important comment; we apologize for the typographical errors in the identifiers and have now corrected "GCT90436511" to the standard GWAS Catalog accession GCST90436511 while providing complete provenances for all datasets in our updated Methods and Supplementary Table X. The primary GWAS summary statistics for ovarian cancer were derived from two large-scale cohorts: GCST90011821 from the pan-cancer study by Rashkin S, et al. (2020) [PMID: 32887889], which identified genetic risk variants in 1,259 cases and 410,350 controls, and GCST90436511 from Zhou W, et al. (2018) [PMID: 30104761], which utilized the SAIGE method to account for case-control imbalance in 2,103 cases of malignant neoplasm of the ovary.

We have added a Data Availability statement confirming that all summary statistics are publicly accessible via the GWAS Catalog database, with detailed documentation on download dates, allele harmonization, and genome builds provided to ensure full reproducibility(Pages 27-28, Lines 609-613). Thanks.

14. 
    The conclusion states: “This finding provides new insights for future intervention strategies and emphasizes the potential impact of mental health on cancer outcomes.” No intervention strategy is proposed or tested. The manuscript does not provide any evidence that targeting CLSTN3 would improve cancer outcomes or mental health. Such statements are misleading.

Response: Thank you for this important criticism. We fully agree that our original conclusion overstated the translational implications (Page 2, Lines 29-32). We have thoroughly revised the Conclusion and Discussion to eliminate any misleading claims. Specifically, we have removed statements implying that our findings constitute validated intervention strategies or proven therapeutic efficacy.

The revised Conclusion now frames our findings as strictly hypothesis-generating, describing CLSTN3 as a candidate molecule rather than a validated therapeutic target. We explicitly state that our in vitro proliferation assays demonstrate a cell-intrinsic effect but do not address therapeutic efficacy, drug delivery, or clinical outcomes. We have also added a dedicated limitations statement clarifying that this study does not propose or test any specific intervention, and that extensive preclinical validation is required before any therapeutic relevance can be claimed (Pages 13-14, Lines 383-403). We believe these revisions accurately reflect the scope and limitations of our work. Thanks.

Reviewer 3 Report

Comments and Suggestions for Authors

This manuscript presents a compelling and methodologically sophisticated investigation into the biological link between Major Depressive Disorder (MDD) and ovarian cancer (OC). The integration of single-cell RNA sequencing data from peripheral blood with Mendelian Randomization (MR) is a notable strength, offering a novel, data-driven approach to a complex epidemiological observation. The identification of CLSTN3 in CD8_EM T cells as a potential causal risk factor for OC and MDD is an intriguing finding. The manuscript is generally well-written. However, several aspects require clarification and further statistical validation to strengthen the causal claim and biological interpretation. 

1. The choice of CD8_EM T cells as the target population is critical to the study's narrative. To better support this choice, please provide a more detailed explanation in the results or methods section. This should include the comparative data that demonstrated the unique relevance of CD8_EM T cells (e.g., showing they were the most expanded or dysregulated population common to both conditions) and the biological reasoning behind focusing on them.
2. Clearly state the limitationissue of small sample size in the "Discussion" section of the manuscript. Adjust the wording of all conclusions from a definitive tone to a hypothesis-generating tone (e.g., "Our exploratory analysis suggests...", "This raises the possibility that..."). Emphasize the need for future validation in larger-scale studi
3. To provide direct experimental validation of your compelling genetic findings, it would be crucial to examine CLSTN3 expression at the protein level specifically within CD8⁺ T cells. I recommend performingflow cytometric analysis on peripheral blood mononuclear cell (PBMC) samples to quantify CLSTN3 protein levels in CD8_EM T cells. Ideally, this analysis should compare samples from individuals with MDD or ovarian cancer cases against healthy controls to strengthen the link between the psychological state and the molecular alteration in the immune cell population of interest.
4. Please provide additional information on the potential for drug development targeting CLSTN3.

Author Response

Dear reviewer,

Thank you very much for giving us the opportunity to revise our manuscript. We are grateful to you for their thorough, constructive, and insightful comments, which have helped us improve the quality and rigor of our study considerably.

Below we provide a detailed, point-by-point response. All changes made to the manuscript are shown in Track Changes mode.

We hope you will find the revisions satisfactory, and that the manuscript can now be considered for publication. Please do not hesitate to contact us if anything further is needed.

 

Reviewer3#

This manuscript presents a compelling and methodologically sophisticated investigation into the biological link between Major Depressive Disorder (MDD) and ovarian cancer (OC). The integration of single-cell RNA sequencing data from peripheral blood with Mendelian Randomization (MR) is a notable strength, offering a novel, data-driven approach to a complex epidemiological observation. The identification of CLSTN3 in CD8_EM T cells as a potential causal risk factor for OC and MDD is an intriguing finding. The manuscript is generally well-written. However, several aspects require clarification and further statistical validation to strengthen the causal claim and biological interpretation.

1. The choice of CD8_EM T cells as the target population is critical to the study's narrative. To better support this choice, please provide a more detailed explanation in the results or methods section. This should include the comparative data that demonstrated the unique relevance of CD8_EM T cells (e.g., showing they were the most expanded or dysregulated population common to both conditions) and the biological reasoning behind focusing on them.

Response: Thank you for this important suggestion. We have revised both the Results sections to provide a detailed justification for selecting CD8_EM T cells. Specifically, we now highlight that comparative scRNA-seq analysis identified CD8_EM T cells as the most prominently expanded and functionally dysregulated immune subset common to both MDD and OC patients.

Biologically, CD8_EM T cells are critical mediators of anti-tumor immunity, and their persistent activation in chronic inflammatory states such as MDD can lead to exhaustion and impaired tumor surveillance. We hypothesized that MDD-driven expansion of dysfunctional CD8_EM T cells may compromise effective immune responses against ovarian cancer. These points are now clearly stated in the revised manuscript (Page 8, Lines 218-228). Thanks.

 

2. Clearly state the limitationissue of small sample size in the "Discussion" section of the manuscript. Adjust the wording of all conclusions from a definitive tone to a hypothesis-generating tone (e.g., "Our exploratory analysis suggests...", "This raises the possibility that..."). Emphasize the need for future validation in larger-scale study.

Response: We sincerely thank the reviewer for this critical suggestion regarding the tone of our conclusions and the need to explicitly address sample size limitations. We fully agree that given the relatively small sample size in our single-cell and clinical analyses, a more cautious, hypothesis-generating tone is both scientifically appropriate and necessary. We have added a dedicated paragraph in the Discussion explicitly acknowledging the small sample size limitation.

We have also systematically revised the wording throughout the manuscript—from the Abstract to the Figure Legends—replacing definitive statements with cautious, hypothesis-generating language. Furthermore, we have expanded the Future Directions section to emphasize that our findings require validation in larger, multi-center cohorts and in vivo models. We believe these revisions appropriately frame our study as hypothesis-generating (Pages 13-14, Lines 383-403). Thanks.

 

3. To provide direct experimental validation of your compelling genetic findings, it would be crucial to examine CLSTN3 expression at the protein level specifically within CD8⁺ T cells. I recommend performingflow cytometric analysis on peripheral blood mononuclear cell (PBMC) samples to quantify CLSTN3 protein levels in CD8_EM T cells. Ideally, this analysis should compare samples from individuals with MDD or ovarian cancer cases against healthy controls to strengthen the link between the psychological state and the molecular alteration in the immune cell population of interest.

Response: Thank you for this excellent suggestion. We have performed flow cytometric analysis of PBMCs from 5 healthy controls and 5 ovarian cancer patients. As shown in revised Figure 8E-8F, CLSTN3 protein levels were significantly elevated in activated CD8⁺ T cells from OC patients compared to controls (Page 11, Lines 313-316).

We wish to note a technical clarification: Due to the limited availability of reliable CD45RA/CCR7 antibodies for our panel, CD8⁺ effector memory T cells were functionally identified as CD3⁺CD8⁺TNF-α⁺ cells following PMA/ionomycin stimulation, rather than by the canonical surface markers. While this approach captures activated/effector CD8⁺ T cells with cytokine-producing capacity rather than the strictly defined CD45RA⁻CCR7⁻ subset, this population reflects the functional effector state relevant to our study (Page 6, Lines 152-161). We have revised the Methods and Results to explicitly state this limitation and acknowledge that future validation using a comprehensive memory-marker panel (CD45RA/CCR7) is warranted (Pages 13-14, Lines 383-403). Thanks.

 

4. Please provide additional information on the potential for drug development targeting CLSTN3.

Response: Thank you for this insightful suggestion. We have added a dedicated paragraph to the Discussion. We note that no CLSTN3-specific therapeutics have yet entered clinical development, but our findings suggest several potential strategies:

(1) disease-context-dependent targeting, given divergent CLSTN3 expression patterns in ovarian cancer versus MDD; (2) pharmacological intervention directed at CLSTN3-mediated signaling or lysosomal trafficking in T cells; and (3) RNA-targeted approaches (e.g., siRNA, ASOs) for cell-type-specific modulation. We also acknowledge key challenges, including neuronal expression (CNS side-effect risk), delivery specificity for CD8_EM T cells, and potential functional redundancy with CLSTN1/CLSTN2. Overall, we propose CLSTN3 as a candidate biomarker with long-term therapeutic relevance rather than an immediately druggable target and have outlined priority next steps accordingly (Page 13, Lines 374-382). Thanks.