To assess the causal pathway between physical activity, gut microbiota and cancer in European population, we used the GWAS summary data of physical activity, microbial traits and cancer, then conducted two-sample MR and two-step MR. A summary of the analysis pipeline is shown in Fig. 1. When studying the causal relationship between exercise and microbiota, after a weak IV with an F value less than 10 was removed, 59 SNPs were identified as IVs for physical activity, as shown in Supplementary Table 1. When studying the causal relationship between microbiota and cancer, we set the threshold for the P-value of SNPs to be 1E-05, after a weak IV with an F value less than 10 was removed, 5622 SNPs were identified as IVs for microbiota, as shown in Supplementary Table 2.
2.1. The causal association of physical activity with cancer risk
We found casual associations between physical activity and cancer. There are 11 meaningful results here, as shown in Supplementary Table 3. Then we conducted the MR Steiger directional test, and the results showed that only the steiger test values for Number of days/week of vigorous physical activity 10 + minutes (NVPA_10+) to ER- Breast cancer were greater than 0.05, The steiger test values for the remaining results were all less than 0.05, indicating that the hypothesis of a causal relationship from exercise to tumor is correct (Supplementary Table 4). In addition, there is a 95% confidence interval for the result OR that spans across 1, so this result is excluded.
Therefore, only 9 of the causal associations (Fig. 2) between physical activity and tumors are meaningful in the results. After conducting sensitivity analysis, we found the presence of horizontal pleiotropy in some of the results using the MR-PRESSO method, such as APAM were negative associated with Breast cancer (Oncoarray) (OR = 0.916, 95% confidence interval (CI): 0.877–0.957, PBH= 0.018), ER + Breast cancer (Oncoarray) (OR = 0.942, 95% confidence interval (CI): 0.899–0.988, PBH= 0.036). Apart from these results, none of the others have multiple effects, and the results are robust. For example, APAM were negative associated with Type of cancer: ICD10: C61 Malignant neoplasm of prostate (OR = 0.998, 95% confidence interval (CI): 0.997–0.999, PBH= 0.0008), Type of cancer: ICD10: D05.1 Intraductal carcinoma in situ (OR = 0.999, 95% CI: 0.998–1, PBH= 0.0009), Cancer of primary lymphoid hematopoietic (OR = 0.999, 95%CI: 0.998–0.999, PBH= 0.0024). Lucky, the association between APAM and breast cancer has been observed in more than one tumor dataset, as shown in Fig. 2. Furthermore, the results consistently indicate a negative correlation between the two. In addition to that, a similar phenomenon has also been observed between prostate cancer and APAM.
MR analyses suggested negative causal inferences for Number of days/week of moderate physical activity 10 + minutes (NMPA_10+) (OR = 0.373, 95%CI: 0.191–0.728, PBH= 0.023) with clear cell ovarian cancer.
The results of Wald ratio analyses demonstrated that NVPA_10 + associated with breast risk. We found that the causal relationship between the NVPA_10 + and ER + breast cancer (OR = 22.385, 95%CI: 9.138–54.835, PBH= 4.2E-11), as well as breast cancer (OR = 10.971, 95%CI: 5.095–23.623, PBH= 3.708E-09), is positive.
2.2. The causal association of physical activity, microbiota traits with cancer risk
In terms of microbiota, the physical activity has causal effects on gut microbiota (Supplementary Fig. 1, Supplementary Table 5). After conducting horizontal meta-regression analyses using the MR-PRESSO and MR-Egger methods, no evidence of horizontal pleiotropy was found in both methods. The results were robust and stable. Four physical activity features were found to be associated with 61 microbiota features in our results, you can refer to the Supplementary Table 5 for more details.
The results of IVW and Wald ratio analyses demonstrated that 17 microbiota features are associated with 8 cancer risk (PBH<0.05) (Supplementary Table 6). After conducting sensitivity analysis, it was found that a portion of the results exhibited horizontal pleiotropy using the MR-PRESSO method. Simultaneously, the MR-Egger method also detected some results with horizontal pleiotropy. Finally, we excluded the results that showed horizontal pleiotropy in both methods and in MR-Egger. The final result can be seen in the Supplementary Fig. 2. After conducting the directional test on MR Steiger, all significant results have been found to be in the correct direction (steiger test < 0.05, Supplementary Table 7).
Based on the results above, multiple microbial features representing the same microorganism are causally related to the same tumor outcome, such as OTU97_105 (Ruminococcaceae) prevalence (GWAS_ID = HkgyA9, OR = 0.955, 95%CI: 0.936–0.974, PBH= 0.03), OTU99_120 (Ruminococcaceae) prevalence (GWAS_ID = XJMkxn, OR = 0.961, 95%CI: 0.941–0.982, PBH= 0.03), although these two microorganisms belong to different OTUs, they both represent F_Ruminococcaceae(Supplementary Table 8), and both have a negative causal relationship with ER + Breast cancer (GWAS_ID = ieu-a-1127). The direction of the causal relationship is consistent.
Another scenario observed here is that the same microbial feature has a causal relationship with the same type of tumor from different datasets, for instance OTU97_73 (Prevotella) prevalence (GWAS_ID = nx1iLW), represent Prevotella (Supplementary Table 8), have a negative causal relationship with Endometrial cancer (GWAS_ID = ebi-a-GCST006464, OR = 1.07, 95%CI: 1.034–1.108, PBH= 0.014), Endometrial cancer (endometrioid histology) (GWAS_ID = ebi-a-GCST006465, OR = 1.08, 95%CI: 1.037–1.125, PBH= 0.034). Both of these tumors belong to endometrial cancer, and the causal relationship between them and Prevotella is positive and consistent in direction.
2.3. Validate data analyses
In the validate microbiota data, when studying the causal relationship between physical activity and microbiota, some of the findings overlap with the German dataset, thereby validating the results obtained from analyzing the Finnish dataset, such as Bacteroides stercoris abundance (Supplementary Table 9), have a negative causal relationship with Breast cancer (OR = 0.0033, 95%CI: 0.003–0.004, PBH= 0). Phascolarctobacterium sp003150755(OR = 33.577, 95%CI: 22.124–50.958, PBH=3.18E-61), Prevotella sp have a positive causal relationship with Breast cancer (Supplementary Table 9). When studying the causal relationship between microbiota and cancers, there also have some findings overlap, for example, NMPA_10 + have a negative causal relationship with Alistipes abundance (OR = 0.712, 95%CI: 0.558–0.907, PBH= 0.0059).
2.4. Sensitivity analyses
We future carried out Cochran’s Q statistics test to evaluate the MR results. In the research of all results, many results exhibit heterogeneity, so these results were obtained using the Inverse variance weighted (multiplicative random effects) method. If you would like to see more detailed information, please refer to Supplementary Table 10.
Pleiotropy test using MR-Egger and MR-PRESSO showed some results of MR study are not robust (Supplementary Table 11). In the MR study on the relationship between microorganisms and tumors, there are some results that show pleiotropic (P-value < 0.05). Please refer to the details for more information in Supplementary Table 11.
For reverse MR, MR-Egger regression and MR-PROSSSO showed that there was no horizontal pleiotropy between instrumental variables and outcome. In addition, the results of Cochrane Q statistics showed no significant heterogeneity. Detailed information in the reverse MR analyses is shown in Supplementary Table 12.
2.5. Mediation Analyses
Given the limited evidence suggesting that physical activity have a causal effect on microbiota, and microbiota have a causal effect on microbial cancer risk. We conducted mediation analyses (two-step MR) to investigate whether the effect of physical activity on cancer was mediated by microbial traits.
The mediation analyses shown that the effect of Moderate to vigorous physical activity levels was partially mediated by OTU97_73 (Prevotella) prevalence for Endometrial cancer. The same conclusion was also obtained in another dataset of endometrial cancer. And we found a significant mediation effect for OTU99_84 (Prevotella) prevalence in the causal effect of the effect of Moderate to vigorous physical activity levels on Breast cancer. However, both OTU97_73 (Prevotella) prevalence and OTU99_84 (Prevotella) represents Prevotella (Supplementary Table 8), moreover, the tag Sequences of these two OTUs are consistent (Supplementary Table 8), and their identity as Prevotella stercorea can be found in the BLAST(https://blast.ncbi.nlm.nih.gov/). Also, we found the mediation analyses shown that the effect of Moderate to vigorous physical activity levels was partially mediated by OTU99_84 (Prevotella) prevalence for Breast cancer. In addition, we also found a causal relationship between Prevotella and Breast cancer in Finish microbial datasets[18], and there is a positive effect (mediate(beta) > 0) between them (Supplementary Table 9). However, there is no direct interaction between Moderate to vigorous physical activity levels and Endometrial cancer, Breast cancer. Therefore, there is only an indirect interaction mediated by Prevotella between Moderate to vigorous physical activity levels and these two types of tumors (Fig. 3, Supplementary Table 13).