Association between gut microbiota and hepatocellular carcinoma and biliary tract cancer: A mendelian randomization study

Abstract

BACKGROUND

An increasing number of studies have begun to discuss the relationship between gut microbiota and diseases, yet there is currently a lack of corresponding articles describing the association between gut microbiota and hepatocellular carcinoma (HCC) and biliary tract cancer (BTC). This study aims to explore the relationship between them using Mendelian randomization (MR) analysis method.

AIM

To assess the relationship between gut microbiota and HCC and BTC.

METHODS

We obtained Genome-wide association study (GWAS) data for the gut microbiome from the intestinal microbiota genomic library (MiBioGen, https://mibiogen.gcc.rug.nl/). Additionally, we accessed data pertaining to HCC and BTC from the IEU open GWAS platform (https://gwas.mrcieu.ac.uk/). Our analysis employed fundamental instrumental variable analysis methods, including inverse-variance weighted, MR and Egger. To ensure the dependability of the results, we subjected the results to tests for multiple biases and heterogeneity.

RESULTS

During our investigation, we discovered 11 gut microbiota linked to an increased risk to BTC and HCC. The former included the genus Eubacterium hallii group (P = 0.017), Candidatus Soleaferrea (P = 0.034), Flavonifractor (P = 0.021), Lachnospiraceae FCS020 (P = 0.034), the order Victivallales (P = 0.018), and the class Lentisphaeria (P = 0.0.18). The latter included the genus Desulfovibrio (P = 0.042), Oscillibacter (P = 0.023), the family Coriobacteriaceae (P = 0.048), the order Coriobacteriales (P = 0.048), and the class Coriobacteriia (P = 0.048). Furthermore, in BTC, we observed 2 protective gut microbiota namely the genus Dorea (P = 0.041) and Lachnospiraceae ND3007 group (P = 0.045). All results showed no evidence of multiplicity or heterogeneity.

CONCLUSION

This study explores a causal link between gut microbiota and HCC and BTC. These insights may enhance the mechanistic knowledge of microbiota-related HCC and BTC pathways, potentially informing therapeutic strategies.

Key Words: Hepatocellular carcinoma; Biliary tract cancer; Gut microbiota; Mendelian randomization; Genetic variant

Core Tip: We apply the largest intestinal gut microbiota gene database to human disease genetic data, elucidating correlations between gut microbiota and diseases. This aids our understanding of how these bacteria influence or protect against diseases, thereby providing crucial insights into their role in such processes.

INTRODUCTION

Hepatobiliary diseases are major disease types among hepatocellular carcinoma (HCC) and biliary tract cancer (BTC). HCC stands as a predominant form of primary liver tumors, accounting for 75% of cases[1], and it is the fifth most common malignancy worldwide and the second leading cause of cancer-related deaths[2]. Etiologically, hepatitis B and C viruses remain the main contributors to HCC, driving nearly 70%-80% of case related to viral infection-induced cirrhosis, while alcoholic cirrhosis is another contributing factor[3,4]. In addition, non-alcoholic fatty liver disease[5], aflatoxin and autoimmune factors[6,7], and genetic factors are also associated with the occurrence of HCC. The incidence of BTC is also high, only second to HCC. It accounts for about 15% of primary liver tumors and takes form in various anatomical classifications, namely intrahepatic, hilar, and extrahepatic[8]. Risk factors for BTC include smoking, alcohol consumption, viral hepatitis infection, and liver fluke infection[9]. A common hallmark of these factors is the chronic inflammation within the bile duct epithelium and bile stasis[10].

Mendelian randomization (MR) is a widely employed epidemiological method used to evaluate the causal links between observed modifiable exposures or risk factors and clinical outcomes[11]. In this process, we need to identify an instrumental variable (IV) to examine the relationship between exposure and outcome, while also controlling for any confounding factors[12]. Single nucleotide polymorphism (SNP) has garnered increasing popularity as an IV in various contexts. This is because of the random assignment of its alleles during human gamete formation, thereby mitigating the influence of confounding factors[13]. In previous genome-wide association studies (GWAS), a large amount of genetic variation data has been collected from diverse individuals[14]. Therefore, eligible SNPs can be selected as IVs to explore the relationship between different exposures and outcome factors.

The gut microbiota refers to the total number of bacteria that live in the human digestive tract, surpassing 100 trillion in number. This includes various types of microorganisms that together make up a genome that is 150 times more expansive than the human genome[15]. These microorganisms have multiple effects on the human body. For example, studies on the relationship between gut microbiota and alcoholism have found that their impact not only on alcohol, but also on the manner in which alcohol and its byproducts interact with the body[16]. Moreover, research has shown that gut microbiota also promotes the development of non-alcoholic liver disease[17]. The influence of microbiota on malignant tumors has also gained increasing attention. Alterations in gut bacteria composition may reshape the development of HCC[18]. In the patients of HCC, their intestinal microbiome's diversity and quantity fluctuate-showing a probable decrease in Bifidobacteria while Enterococcus and Escherichia increase[19,20]. These changes offer a deeper understanding of the causal relationship between gut microbiota and HCC and BTC. However, these findings are often affected by confounding factors, such as age, gender, and weight. Therefore, this study utilized MR to investigate the causal association between gut microbiota and HCC and BTC objectively.

MATERIALS AND METHODS

We employed a two-sample MR analysis to investigate the relationship between gut microbiota and HCC and BTC. In MR analysis, the following three assumptions must be met to establish the connection between the exposure and outcome: (1) The selection of SNPs significantly associated with the gut microbiota as IVs; (2) the use of IVs is independent of other confounding factors, such as alcohol consumption and age; and (3) IVs only affect the outcome through the association between exposure and outcome, devoid of any other causal pathways (Figure 1).

Figure 1 Assumptions of Mendelian randomization and the processing. GWAS: Genome-wide association study; IVs: Instrumental variables; SNP: Single nucleotide polymorphism; MR: Mendelian randomization; IVW: Inverse-variance weighted; HCC: Hepatocellular carcinoma; BTC: Biliary tract cancer; R2: R-squared; KB: Kilobase.

Statistical resource

The data on the human gut microbiome was obtained from MiBioGen, encompassing 18340 research participants from 11 different countries and 24 research cohorts. The survey results covered 211 genera, 131 families, 35 orders, 16 classes, and 9 phyla, underscoring a highly diverse microbial ecosystem.

The HCC data (bbj-a-158) and BTC data (bbj-a-92) were obtained from the IEU open GWAS (https://gwas.mrcieu.ac.uk/) database. The HCC study included 1866 cases and 195745 controls, while the BTC study incorporated 339 cases and 195745 controls.

Selection of IVs

In different taxonomic classifications, encompassing phylum, order, family, and genus, we first screened for SNPs that are significantly associated with gut microbiota using a threshold of P < 5 × 10^-8. However, the yield of available SNPs was limited, potentially not meet the research requirements. Therefore, we adjusted the threshold to P < 1 × 10^-5 and performed a second round of screening to obtain SNPs for IV. To ensure result accuracy and minimize bias, we referred to related studies and conducted linkage disequilibrium screening based on the parameters of R-squared (R2) < 0.01 and kilobase (KB) = 500. Given the presence of the second hypothesis, we extended our focus to identifying SNPs related to confounding factors. To achieve this, we used the PhenoScannerV2 database to retrieve, replace, or delete SNPs. Furthermore, to avoid the influence of weak IVs on results, we used the F-statistic to test for IV strength and set F < 10 as the threshold indicative of weak IV.

MR analysis

MR analysis was used to investigate the causal relationship between gut microbiota and HCC and BTC. We utilized three common MR methods: Inverse variance weighted testing (IVW), weighted median method, and MR-Egger regression. IVW was used as the primary diagnostic benchmark, while the other testing methods provided supplementary references. We considered P < 0.05 to be statistically significant.

Heterogeneity

To ensure the reliability of our results, we conducted multiple sensitivity and specificity analyses, including IVW's Cochran's Q statistic, MR-Egger regression test, and MR-Pleiotropy Residual Sum and Outlier (PRESSO), in order to evaluate the horizontal pleiotropic effects. P < 0.05 indicated the presence of heterogeneity (Figure 1).

RESULTS

Selection of IVs

We obtained SNPs from a total of 211 microbial communities and subsequently removed those in linkage disequilibrium (R2 < 0.01, KB = 500). This process resulted in 128, 235, 294, 303, and 1761 SNPs at the phylum, class, order, family, and genus levels, respectively, with a significance level of P < 1 × 10^-5. Our selected IV are listed in Table 1, and their F-statistics all exceeded 10, indicating their instrumental strength satisfying the required criteria. For further details on the IVs, please refer to Supplementary Table 1.

Table 1 Causal associations between genetically predicted 13 microbial taxa and hepatocellular carcinoma and biliary tract cancer.

Outcome	Exposure	Method	nSNP	β	SE	OR (95%CI)	P value
Biliary tract cancer
	Genus Eubacterium hallii group	IVW	11	0.838	0.351	2.31 (1.16-4.61)	0.017
	Genus Candidatus Soleaferrea	IVW	9	0.802	0.377	2.23 (1.06-4.67)	0.034
	Genus Dorea	IVW	7	-1.232	0.603	0.29 (0.09-0.95)	0.041
	Genus Flavonifractor	IVW	5	1.619	0.702	5.05 (1.27-19.9)	0.021
	Genus Lachnospiraceae FCS020 group	IVW	11	0.710	0.335	2.03 (1.06-3.92)	0.034
	Genus Lachnospiraceae ND3007 group	IVW	3	-2.430	1.214	0.09 (0.01-0.95)	0.045
	Order Victivallales	IVW	6	0.791	0.334	2.21 (1.15-4.25)	0.018
	Class Lentisphaeria	IVW	6	0.791	0.334	2.21 (1.15-4.25)	0.018
Hepatocellular carcinoma
	Genus Desulfovibrio	IVW	9	0.324	0.159	1.38 (1.01-1.89)	0.042
	Genus Oscillibacter	IVW	9	0.345	0.151	1.41 (1.05-1.90)	0.023
	Family Coriobacteriaceae	IVW	15	0.377	0.190	1.46 (1.00-2.12)	0.048
	Order Coriobacteriales	IVW	15	0.377	0.190	1.46 (1.00-2.12)	0.048
	Class Coriobacteriia	IVW	15	0.377	0.190	1.46 (1.00-2.12)	0.048

IVW: Inverse-variance weighted; OR: Odds ratio; Nsnp: Number of Single nucleotide polymorphism; SE: Standard error.

The impact of gut microbiota on HCC and BTC

Our primary results were obtained through the application of IVW method in MR analysis. And all the results were showed in Supplementary Table 2. There were 8 relevant microbial communities identified in BTC. After excluding one unknown microbial community, 5 microbial communities were found to associate with HCC (Figure 2).

Figure 2 Forest plot of the associations between genetically determined 13 gut microbial genera with the risks of hepatocellular carcinoma and biliary tract cancer. OR: Odds ratio; Nsnp: Number of single nucleotide polymorphism.

BTC: We observed a protective effect in patients with BTC for the genus Dorea [P = 0.041, odds ratio (OR) = 0.29, 95%CI: 0.09-0.95] and Lachnospiraceae ND3007group (P = 0.045, OR = 0.09, 95%CI: 0.01-0.95). However, it appears that the genus Eubacterium hallii group (P = 0.017, OR = 2.31, 95%CI: 1.16-4.61), CandidatusSoleaferrea (P = 0.034, OR = 2.23, 95%CI: 1.06-4.67), Flavonifractor (P = 0.021, OR = 5.05, 95%CI: 1.27-19.9), Lachnospiraceae FCS020 group (P = 0.034, OR = 2.03, 95%CI: 1.06-3.92), the order Victivallales (P = 0.018, OR = 2.21, 95%CI: 1.15-4.25), and the class Lentisphaeria (P = 0.018, OR = 2.21, 95%CI: 1.15-4.25) are associated with adverse outcomes in patients with BTC.

HCC: In HCC, we observed an unfavorable influence associated with the genus Desulfovibrio (P = 0.042, OR = 1.38, 95%CI: 1.01-1.89), Oscillibacter (P = 0.023, OR = 1.41, 95%CI: 1.05-1.90), the family Coriobacteriaceae (P = 0.048, OR = 1.46, 95%CI: 1.00-2.12), the order Coriobacteriales (P = 0.048, OR = 1.46, 95%CI: 1.00-2.12), and the class Coriobacteriia (P = 0.048, OR = 1.46, 95%CI: 1.00-2.12). However, we did not identify any bacterial group having a protective role.

All findings are depicted through scatter plots (refer to Supplementary Figure 1) and forest plots (see Supplementary Figure 2) for visual representation.

Heterogeneity

In addition, we employed Cochran's Q test, MR-Egger regression test and MR-PRESSO analysis to assess our data. The results from the Cochran's Q test revealed no significant heterogeneity (P > 0.05). Furthermore, the MR-Egger test did not indicate any evidence of pleiotropy (P > 0.05). After precise MR-PRESSO analysis, we found no detectable outliers, indicating the absence of horizontal pleiotropy (all P > 0.05) (Table 2). When using leave-one-out analysis, we identified biased genetic prediction in certain SNPs (Supplementary Figure 3).

Table 2 The heterogeneity results from the Cochran’s Q test, Egger-intercept analysis and PRESSO-analysis.

Outcome	Exposure	IVW-Q	Q-P value	Egger-Q	Q-P value	Egger-intercept	SE	P value
Biliary tract cancer
	Genus Eubacterium hallii group	7.393	0.687	4.848	0.847	0.100	0.062	0.145
	Genus Candidatus Soleaferrea	9.726	0.284	5.012	0.658	0.795	0.366	0.066
	Genus Dorea	0.971	0.986	0.962	0.965	-0.010	0.107	0.928
	Genus Flavonifractor	4.673	0.322	2.396	0.494	-0.576	0.381	0.228
	Genus Lachnospiraceae FCS020 group	10.554	0.393	10.46	0.313	-0.016	0.060	0.791
	Genus Lachnospiraceae ND3007 group	3.375	0.184	1.100	0.294	1.473	1.024	0.386
	Order Victivallales	3.460	0.629	3.425	0.489	-0.034	0.182	0.859
	Class Lentisphaeria	3.460	0.629	3.425	0.489	-0.034	0.182	0.859
Hepatocellular carcinoma
	Genus Desulfovibrio	7.263	0.508	6.627	0.468	0.038	0.048	0.451
	Genus Oscillibacter	3.114	0.926	2.956	0.888	0.024	0.062	0.703
	Family Coriobacteriaceae	16.794	0.267	16.76	0.210	0.007	0.051	0.881
	Order Coriobacteriales	16.794	0.267	16.76	0.210	0.007	0.051	0.881
	Class Coriobacteriia	16.794	0.267	16.76	0.210	0.007	0.051	0.881

IVW: Inverse-variance weighted; OR: Odds ratio; SE: Standard error.

DISCUSSION

Our human bodies coexist with countless microbes, abundant on our skin and in cavities linked to the outside world. These invaluable partners play significant roles in various physiological processes of us humans[21]. And more and more studies have underscored the critical role of gut microbiota in human health. Functioning with other organisms, gut microbiota can be influenced by host-related factors, exerting both beneficial and detrimental impacts on their host’s well-being[16]. These effects are reflected in various areas of investigation, indicating that gut microbiota could potentially contribute to neurodegenerative diseases like Alzheimer's and Parkinson's, metabolic disorders such as obesity and diabetes, and cardiovascular diseases such as hypertension[22]. The anatomical connection bridging the gut and biliary tract establishes a communication pathway with the liver, suggesting that a causal relationship may exist between gut microbiota and HCC and BTC. Notably, Milosevic et al[23] have described the close relationship between various liver diseases and gut microbiota in detail. These collective research findings affirm the widespread recognition of the gut microbiota’s significance. Gut microbiota is closely related to human health and plays a crucial role in the development of various diseases. Therefore, a deeper understanding of the composition and function of gut microbiota could offer insights into the prevention and management of numerous diseases, providing more effective protection for human health.

There is a lack of research focusing on the interplay between gut microbiota and HCC and BTC. However, with the use of MR in large-scale GWAS data, potential causal relationships could be explored between gut microbiota and HCC and BTC. This approach could potentially illuminate the underlying mechanisms, thereby advancing progress in preventing and treating HCC and BTC.

In our study, a total of 13 microbial groups with causal relationship to HCC and BTC were identified. Among them, six bacterial groups in BTC and five in HCC were associated with higher disease risks. Eubacterium, a core genus within the human gut microbiota, can influence the host through various pathways[24], including cholesterol conversion and butyrate production, which can lower the risk of inflammatory bowel diseases[25]. However, Eubacterium also appears to be involved in the onset of certain diseases, such as obesity[26]. Recent studies, including our own, have found that the abundance of Eubacterium hallii group is significantly higher in patients with BTC compared to healthy individuals[27]. The class Lentisphaeria, initially discovered in the ocean in 2004[28], harbors the order Victivallales within its ranks. We found an increased abundance of this group in multiple diseases, including autoimmune hepatitis and multiple sclerosis[29,30], suggesting a potential association with disease susceptibility. Endotoxin, discovered in 1907, has been of great interest due to its association with liver damage[31]. Subsequent research has demonstrated that it can increase inflammatory factors and cause liver tissue damage[32]. Moreover, endotoxin mediates the generation of reactive nitrogen and oxygen species, leading to cell death and liver damage[33]. It has been shown that the genus Desulfovibrio may be involved in increasing the risk of HCC through the production of endotoxin. Despite Oscillibacter, Candidatus, Soleaferrea, Flavonifractor, the Coriobacteriaceae family, and other microbiota's being linked to adverse outcomes, further study of the mechanisms is required.

Additionally, our investigation led us to identify only two protective bacterial groups, namely genus Dorea and genus Lachnospiraceae ND3007 group, in BTC. Previous studies have shown a positive correlation between the increase in genus Dorea and the varying levels of secondary bile acid called Deoxycholic acid (DCA), indicating a connection[34,35]. Interestingly, DCA exhibit impressive effects on cell metabolism, influencing glycolysis and oxidative phosphorylation process, making it a potent ally against cancer[36,37]. Additionally, the genus Lachnospiraceae ND3007 group, as a member of the family Lachnospiraceae , exerts influence through the generation of fatty acids and the promotion of primary bile acid metabolism[38,39]. This dynamic interaction impacts both the gut microbiome and the host. However, given the broad spectrum of Lachnospiraceae family, its impact on the host is not entirely same. Similar observations have been made by Vacca et al[38], which showed that different strains from Lachnospiraceae family affected host differently in different organs or individuals. This highlights the substantial heterogeneity inherent in the effects of microbial communities, both across distinct species and even within the same species. Therefore, a more refined microbial classification would be more beneficial for future research. We did not find any bacterial groups that exhibited simultaneous influence on both factors, indicating the heterogeneity between these two structures, despite their anatomical proximity. This underscores the diverse actions enacted by gut microbiome bacterial groups within a single individual based on their interaction with different system organs. In several studies, we have found the changes in inflammatory markers such as Neutrophil-to-Lymphocyte Ratio and platelet lymphocyte ratio could lead to poor cancer prognosis. Importantly, the bacterial groups identified as influential in this study could potentially induce the adverse inflammatory reactions that lead to poor prognosis. This finding offers us a new perspective in optimizing cancer treatment strategies.

Our study still has several limitations. Although we employed IVW's Cochran's Q statistic, MR-Egger regression test, and MR-PRESSO to reduce the impact of multiple effects, the selection of GWAS data from different regions and ethnicities may still introduce potential biases into the results. Moreover, due to the large number of SNPs, we had to adjust the screening P value from 5 × 10^-8 to 1 × 10^-5. This adjustment increased the number of SNPs in the study, but may also affected the result precision.

In summary, although our MR analysis has yielded significant insights, further experiments and verification are still needed. Our study offers a foundational reference, introducing novel avenues to inspire more extensive research. We look forward to future investigations revealing the impact of gut microbiota on HCC and BTC.

CONCLUSION

With the assistance of public databases and MR analysis method, we successfully identified causative relationships between specific gut microbiota and HCC and BTC. This finding could potentially impact in the early screening and treatment of HCC and BTC by earmarking intestinal microbiota as viable targets or valuable biomarkers. Our results also provide guidance and reference for future studies centered around the relationship between gut microbiota and HCC and BTC.

ACKNOWLEDGEMENTS

The authors express their gratitude to the participants and investigators of the IEU open GWAS project. The authors also appreciate the MiBioGen consortium for releasing the gut microbiota GWAS summary statistics.