Genetic associations between resting-state brain functional networks and cancer risk: A bidirectional Mendelian randomization study

To the Editor: Resting-state functional magnetic resonance imaging (rsfMRI) enables the evaluation of spontaneous neural activity and has revealed intrinsic brain networks essential for emotion regulation, cognitive control, and systemic homeostasis.^[1] Increasing evidence suggests that brain functional connectivity may influence peripheral physiological processes, including immune surveillance and endocrine function—both of which are implicated in cancer development. Meanwhile, cancer remains a major public health challenge, with multifactorial etiologies and largely unexplored neurobiological determinants. Although previous observational studies have hinted at associations between brain function and cancer risk, their conclusions are often limited by confounding and reverse causality. Thus, the potential causal link between intrinsic brain network activity and cancer susceptibility remains unclear and requires rigorous investigation.

To evaluate the causal relationship between rsfMRI and cancer, we used a two-sample bidirectional Mendelian randomization (MR) analysis. We obtained genome-wide association study (GWAS) summary statistics of 191 rsfMRI phenotypes from the UK Biobank and summary data for 34 cancers from the Integrative Epidemiology Unit (IEU) Open GWAS project.^[2] Detailed information on rsfMRI phenotypes is available in the original publication and presented in Supplementary Table 1, http://links.lww.com/CM9/C700. The two-sample MR method necessitates that both independent samples be derived from the same population. Consequently, in the dataset retrieval process, we excluded studies that were duplicates, involved nonmalignant tumors, or focused on non-European populations. Instrumental variables (IVs) were selected based on genome-wide significance (P <5 × 10^-8), independence (r² <0.001), and F-statistics >10. We used inverse-variance weighted as the primary method, supplemented by MR-Egger, weighted median, and sensitivity analyses including Cochran’s Q, MR-Egger intercept, and leave-one-out tests. All original studies were conducted with appropriate ethical approval and the informed consent of all participants.

The research results are shown in Figure 1. Forward MR analysis revealed 14 rsfMRI traits with significant associations with cancer risk. Increased connectivity in the frontal and precentral networks was associated with a 23.4% decreased risk of breast cancer (odds ratio [OR] = 0.766, 95% confidence interval [CI]: 0.662–0.886, P_FDR = 0.034). Conversely, increased spontaneous neural activity in the occipital, frontal, and parietal regions was linked to a higher risk of liver and bile duct cancers (pheno 2: OR = 1.001, 95% CI: 1.001–1.002, P_FDR = 0.001; pheno 8: OR = 1.002, 95% CI: 1.001–1.003, P_FDR = 0.001; pheno 37: OR = 1.003, 95% CI: 1.001–1.004, P_FDR = 0.002) [Supplementary Tables 2 and 3, http://links.lww.com/CM9/C700]. Reverse MR analysis identified four rsfMRI phenotypes influenced by cancer. High-grade serous ovarian cancer was negatively associated with connectivity in the angular, parietal, and temporal regions (OR = 0.947, 95% CI: 0.920–0.975, P_FDR = 0.0497), whereas melanoma susceptibility was associated with diminished connectivity in default mode and salience networks, and increased precentral activity (pheno 1697: OR = 0.018, 95% CI: 0.002–0.165, P_FDR = 0.022; pheno 1269: OR = 0.010, 95% CI: 0.001–0.091, P_FDR = 0.007) [Supplementary Tables 4 and 5, http://links.lww.com/CM9/C700]. All results passed heterogeneity and pleiotropy tests, and sensitivity analyses confirmed their robustness [Supplementary Tables 6–9, http://links.lww.com/CM9/C700]. Furthermore, the results of the leave-one-out analysis were consistent, reinforcing the reliability of the study’s findings [Supplementary Figures 1 and 2, http://links.lww.com/CM9/C699].

Figure 1.

Research results. (A) Causality between breast cancer and brain functional networks in the forward MR. (B) Causality between liver and bile duct cancer and brain functional networks in the forward MR. (C) Causality between brain functional networks and high-grade serous ovarian cancer in the reverse MR. (D) Causality between brain functional networks and melanoma skin cancer in the reverse MR. fMRI: Functional magnetic resonance imaging; IV: Instrumental variable; IVW: Inverse variance weighted; MR: Mendelian randomization.

Our findings suggest that specific brain functional networks may play a role in modulating cancer risk. Notably, enhanced connectivity in the frontal and precentral regions was associated with a decreased risk of breast cancer, potentially reflecting the role of cognitive-emotional regulation in maintaining immune homeostasis.^[3] Increased spontaneous activity in the occipital, temporal, parietal, and cerebellar regions was positively associated with the risk of liver and bile duct cancers. These areas participate in visual integration and autonomic modulation, and their hyperactivation may reflect dysregulated neural–immune interactions or latent systemic stress signals.^[4,5] Reverse MR analysis revealed that genetic liability to melanoma and high-grade serous ovarian cancer was associated with altered brain functional connectivity. These associations raise the possibility that cancer-related systemic changes may influence intrinsic brain activity. The plausible mechanism is cancer-induced neuroinflammation: elevated cytokines (e.g., interleukin 6 [IL-6], tumor necrosis factor alpha [TNF-α], IL-1β) can cross the blood–brain barrier, activate microglia, and disrupt synaptic function. In addition, malignancy-related hormonal and metabolic alterations may affect neural circuits involved in mood and autonomic regulation.^[6] Although MR reflects genetic predisposition rather than direct clinical effects, these results suggest that cancer–brain interactions may begin earlier than clinically apparent.

Although some associations showed odds ratios close to 1.0, they remain biologically meaningful. These estimates reflect per-standard-deviation changes in complex brain traits and may indicate subtle but impactful shifts in immune or metabolic regulation. MR captures lifelong genetic predispositions, and even small effects can accumulate or interact with the environment to influence cancer risk.^[7] Thus, modest associations may still offer valuable insights into neural contributions to cancer susceptibility.

This study had several limitations. First, all GWAS data were derived from European populations, potentially limiting generalizability to other ethnic groups. Second, although we excluded weak instruments (F-statistic >10), bias from complex traits and residual confounding cannot be completely ruled out. Third, genome-wide instrument selection may introduce biological heterogeneity, especially for polygenic rsfMRI traits lacking functional annotation.^[8] Partial sample overlap between rsfMRI and cancer datasets may exist, though we minimized this through careful dataset selection and sensitivity analyses. Future studies using independent populations and biologically informed instruments are needed to confirm these findings.

Taken together, this study supports a previously underexplored dimension of cancer risk—the role of intrinsic brain network architecture. As brain-directed interventions (e.g., neuromodulation, stress reduction) gain traction in oncology, identifying high-risk neurofunctional profiles may enable personalized prevention strategies. Further longitudinal and experimental studies are warranted to unravel the mechanistic underpinnings linking brain function to cancer susceptibility.

Funding

This study was supported by the National Natural Science Foundation of China (Nos. 82372126 and 8207071577), Zhongda Hospital Affiliated to Southeast University, Jiangsu Province High-Level Hospital (Nos. 2023GSPKY11, GSP-LCYJFH01, and zdyyxy35), China Postdoctoral Science Foundation (No. 2024M750460), and Nanjing Postdoctoral Research Foundation Project (No. FTJ-bh-2).

Supplementary Material

cm9-139-2201-s001.pdf ^{(387.2KB, pdf)}

cm9-139-2201-s002.xls ^{(484KB, xls)}