Abstract
Background/Aim: Gut microbiota plays an important role in colorectal cancer (CRC) and its composition in CRC patients can be influenced by ethnicity and tumour genomics. Herein, the aim was to study the possible associations of ethnicity and gene mutations with the gut microbiota in CRC patients. Materials and Methods: Bacterial composition in stool samples of 83 CRC patients and 60 controls from Iran and Finland was studied by 16S rRNA gene sequencing. The association of gut microbiota composition with CRC, host mutations in KRAS, NRAS and TP53, and ethnicity analysed. Results: Beta diversity analysis indicated significant differences between the Iranian and Finnish gut microbiota composition, in both controls and patients' groups. The Iranian controls had higher abundance of Prevotella and lower abundance of Bacteroides compared to the Finnish controls, while the Finnish patients had higher abundance of Clostridium compared to Iranian patients. Abundance of Ruminococcus was higher in patients compared to the controls. Higher abundances of Herbaspirillum, Catenibacterium and lower abundances of Barnesiella were associated with mutations in NRAS, TP53, and RAS respectively. Conclusion: A possible link of host gene mutations with gut bacterial composition is suggested.
The gut microbiota plays an integral role in the regulation of different gastrointestinal functions. There is increasing evidence that the gut microbiota not only contributes to the initiation and progression of colorectal cancer (CRC) but also influences the host response to cancer treatment, especially chemo- and immunotherapy (1). In addition to its regional effect in the gastrointestinal tract, the gut microbiota also affects distant organs via its role in shaping the immune system and host metabolism (2). Diet, lifestyle and ethnicity are among the main determinants of microbiota composition, and changes in diet associated with a modern lifestyle have been linked to microbiota-mediated cancer risk (3). The host genetic variation is also known to effect bacterial composition (4), although this area has not been investigated extensively. The host genetics is reported to strongly influence the abundance of gut Christensenellaceae (5), the bacteria associated with low body mass index. The Bifidobacterium abundance has been linked with a variation in the lactose tolerance gene LCT loci (6, 7). On the other hand, tumour genome is reported to be a factor modulating colonization of certain bacteria around colorectal tumour tissue (8, 9). A high abundance of Fusobacterium has been reported to be associated with CpG island methylator phenotype (CIMP), microsatellite instability (MSI) and mutations in CHD7/8 (10). However, the associations between tumour mutations and gut microbiota composition remain poorly understood.
The association of the gut microbiota composition with different disease states makes them useful as potential diagnostic markers of health and disease (11). Recent large-scale studies have however demonstrated that ethnicity (12) and geography (13) are strong independent factors explaining inter-individual differences in microbiota composition, even more so than either diet or metabolic disease. It is therefore important that associations of gut microbiota with serious diseases, like cancer, should be tested in different populations. We have earlier described an association of gut microbiota with different types of gastrointestinal neoplasms in Finnish patients (14). In the present study, the gut microbiota composition of Iranian CRC patients and controls were studied and compared to that seen among Finnish CRC patients and controls to find bacterial profile related to ethnicity and that related to CRC. We also compared bacterial composition in CRC patients with and without host gene mutations in KRAS, NRAS, HRAS and TP53 to examine the possible associations between gut microbiota and cancer gene mutations.
Characteristics of Iranian and Finnish colorectal cancer (CRC) patients and controls examined in this study.
Materials and Methods
Patients and sample collection. Informed consent was obtained from all Iranian and Finnish patients and controls before collection of stool samples. The ethical permission (351/13/03/02/2014) for the study was obtained from the Hospital District of Helsinki and Uusimaa (HUS) review board.
All of the Iranian samples were collected in Isfahan University-Alzahra Hospital (Isfahan, Iran) from CRC patients (IrCRC) and controls (IrC) living in the Isfahan province located in the central part of Iran. The stool samples were collected from 52 patients diagnosed with CRC before starting any cancer treatment and 47 control samples from individuals who had undergone colonoscopy due to some gastrointestinal symptoms but had normal findings on colonoscopy (Table I). Among the CRC patients, 11 had rectal cancer and 41 had colon cancer. Both the patients and controls were mainly Muslims, following Iranian dietary habits and did not consume either pork or alcohol. Sixteen percent of the patients and 9% of the controls were smokers. All samples were collected between July, 2016 and December 2017. Specimens were collected in the hospital and immediately stored at -80°C until DNA extraction.
Stool samples were also collected from 31 Finnish CRC patients (HefeCRC) between April 2015 to May 2017, in Helsinki Uusimaa Hospital District: Surgical, Meilahti, and Jorvi Hospitals in Finland. Details of CRC patients and control group (HefeC) of Finnish origin have been described previously (14). The Finnish patients and controls followed mainly non-vegetarian diet, and on an average week consumed 10-20 standard units of alcohol. Nearly thirty percent of the Finnish patients and controls were smokers.
DNA extraction. From Iranian patients and controls, DNA was extracted from stool samples (180-220 mg) using the QIAamp DNA stool mini kit (Qiagen GmbH, Hidden, Germany) according to manufacturer's guidelines. DNA was quantified by Qubit 2.0 Fluorimeter (Thermo Fisher Scientific, Waltham, MA, USA) using the Qubit dsDNA BR Assay Kit. Extracted DNA was stored at −20°C. DNA extraction from Finnish stool samples was performed using PSP Spin Stool DNA Plus Kit (Stratec Molecular, Berlin, Germany), as previously described (14).
The results of 16S rRNA gene sequencing from 31 Finnish CRC patients and 13 healthy Finnish individuals published previously (14) were used to compare results with those of Iranian cohort. DNA samples of both Iranian and Finnish CRC patients and controls were checked for quality and quantity by Qubit 2.0 Fluorimeter and all the analysis were done in the same laboratory, by the same researchers in Helsinki, and following the same methodological procedures, as previously described (14).
Gene mutation analysis of host DNA from stool samples. Data regarding hotspot mutations in KRAS, NRAS, HRAS, and TP53 genes in DNA from stool samples of Finnish patients was collected from previously published results (15). Gene mutation information was available for 79/83 CRC patients; TP53 mutations were seen in 13, NRAS in 6, KRAS in 3 and no mutations in HRAS patients (all RAS mutations were found in 9 patients). DNA isolated from stool samples of both Finnish and Iranian CRC patients were studied for hotspot mutations in 22 cancer-related genes by targeted next generation sequencing using Ion AmpliSeq Colon and Lung Cancer panel v2 (Thermo Fisher Scientific, Waltham, MA, USA). DNA samples were sequenced on Ion Personal Genome Machine System (Thermo Fisher Scientific, Waltham, MA, USA) as previously described (15).
16S rRNA gene sequencing. Library preparation, template preparation and sequencing were performed on DNA isolated from Iranian CRC and controls as described earlier for Finnish CRC and control samples (14). For Finnish patients and controls, the results of 16S rRNA gene sequencing published earlier (14) were used for comparison with Iranian samples and for examining the association with gene mutations in DNA from stool samples.
Briefly, sequencing libraries were prepared from 3 ng of DNA, using Ion 16S Metagenomics kit (Thermo Fisher Scientific, Waltham, MA, USA) according to the vendor's instructions. Six hypervariable regions (Primer set V2, V4, V8 and Primer set V3, V6-7, V9) of 16S rRNA gene were amplified in two reactions/sample. After PCR, the samples were end-repaired, purified with Agencourt® AMPure® XP beads (Beckman Coulter, Brea, California, USA) and ligated to barcoded sequencing adapters according to the kit protocol. The libraries were quantified by the TapeStation (Agilent Technologies, Santa Clara, CA, USA) and samples were diluted to a 10 pM concentration.
The libraries were pooled and the template preparation was performed with either Ion OneTouch 2 system using the Ion PGM™ Hi-Q™ OT2 Kit (Thermo Fisher Scientific, Waltham, MA, USA) or Ion Chef system using the Ion PGM™ Hi-Q™ Chef Kit (Thermo Fisher Scientific, Waltham, MA, USA) following the kit protocols. The OneTouch 2 or the Ion Chef was used for emulsion PCR and the quality of resulting Ion Spheres were checked with Qubit 3.0 fluorometer (Thermo Fisher Scientific, Waltham, MA, USA). Sequencing was performed on the Ion PGM system using the Ion 318™ Chip (Thermo Fisher Scientific, Waltham, MA, USA) and Ion PGM Hi-Q Sequencing kit (Thermo Fisher Scientific, Waltham, MA, USA).
Data analysis. Operational taxonomic unit (OTU) tables were created for the 16S rRNA gene sequencing data of a total of 143 stool samples. The samples were classified according to nationality (Iranian/Finnish), cancer status (cancer/control), tumour location (colon/rectum), tumour stage (early/late), and gene mutation status of RAS (KRAS, NRAS, HRAS) and TP53. The between-sample normalization was performed by rarefaction of the read counts to an even depth and the rarefied read counts were converted into relative abundances using the phyloseq R package (16). A total of 163 unique genera were detected.
Alpha diversity and observed richness of the gut microbiota at the genus level was analysed with the microbiome (17) and vegan R packages (18). The Shannon index was used to estimate the bacterial diversity, with community richness being quantified by the number of unique taxa observed. Kruskal-Wallis test was applied to test for the significance of group-level differences and correction for multiple testing was performed for each group with the Benjamini-Hochberg FDR method (19).
Unsupervised principal coordinates analysis (PCoA), based on Euclidean distance between Hellinger-transformed abundance profiles (20) were performed with the phyloseq R package (16). Only the genera that were detected in at least 20% of all samples were included in the analysis. Community level differences between the groups were tested with PERMANOVA for CLR-transformed (centered log-ratio transformed) abundances (21) with the R package compositions (22) to remove compositionality bias. ANCOM was applied to assess the significance of the differences in the abundance of the individual genera (23).
Association of age and sex with taxa abundances was examined by Spearman correlation and Wilcoxon test respectively, both performed for CLR-transformed data in order to remove compositionality bias. Since no significant association of age or sex with taxa abundances was observed, we did not include these as covariates in our analysis.
Results
The characteristics of the CRC patients and controls groups included in the study are described in Table I.
Bacterial alpha-diversity. The most significant difference in gut microbiota alpha-diversity (Shannon index) was observed between the Iranian and the Finnish controls (p=0.004). Finnish controls had a higher bacterial richness than their Iranian counterparts (Figure 1A). However, no difference in alpha-diversity was observed between the Iranian and the Finnish CRC patients (p=1.0) (Figure1B); between the Iranian CRC patients and controls (p=0.51); and between Finnish CRC patients and controls (p=0.10). All p-values have been corrected for multiple testing as described in the Materials and Methods.
Bacterial beta-diversity. The beta diversity of gut microbiota composition (Figure 2; Table II) indicated that the Iranian controls were significantly different from the Finnish controls (p=0.006), and the Iranian CRC patients were significantly different from the Finnish CRC patients (p=0.006), indicating a strong ethnic influence on the gut microbiota composition. The CRC patients and controls were not significantly different for either nationality (p>0.1).
Bacterial genera with significant differences in relative abundances in different group comparisons. Significant differences were observed between the Iranian and Finnish controls also in the relative abundance of many individual genera (Table III). The most notable difference was the high relative abundance of Prevotella in the Iranian controls compared to the Finnish controls (Figure 3A). Other significant genera [Bacteroides, Roseburia, (Ruminococcus), Eubacterium, Faecalibacterium and Lactococcus] had higher relative abundances in Finnish control samples compared to the Iranian controls. On the other hand, a comparison of Iranian and Finnish cancer patients showed significant differences in the relative abundance of only Clostridium; that had a higher abundance in the Finnish CRC patients as compared to Iranian patients (Figure 3B).
The CRC patients had a higher relative abundance of Ruminococcus compared to healthy individuals, when samples were not classified according to ethnicity. Higher relative abundance of Ruminococcus and Bifidobacterium was also seen in Iranian CRC patients compared to Iranian controls. On the other hand, lower relative abundance of Bifidobacterium, Phascolarctobacterium and Lachnoclostridium was seen in Finnish CRC patients compared to the Finnish controls (Table III).
Association of host gene mutations and gut microbiota. CRC patients with RAS (KRAS/NRAS) gene mutations in the DNA isolated from stool samples had lower abundance of Barnesiella in comparison to those patients without RAS gene mutation (Figure 3C; Table III). The relative abundance of Herbaspirillum was higher in NRAS mutated (Figure 3D); of Catenibacterium higher (Figure 3) and Blautia lower in TP53 mutated patients compared to the patients with no mutation in the corresponding genes (Table III).
Comparison of alpha diversity between the groups. (A) Iranian (IrC) and Finnish (HefeC) controls, and (B) Iranian (IrCRC) and Finnish (HefeCRC) colorectal cancer (CRC) patients.
Fusobacterium among CRC and controls. We observed presence of Fusobacterium in 18.6% of the samples from CRC patients, which was significantly (p=0.01; Fisher's exact test) higher compared to 3.3% among controls.
Discussion
Composition of gut bacteria is affected by both environmental and genetic factors. Diet, lifestyle, ethnicity, age, sex and host genotype are among the main determinants of microbiota composition and the changes in gut bacterial composition are linked to microbiota-mediated cancer risk (24-27). We aimed to study gut microbiota and host gene mutations in stool samples from CRC patients and controls from Iran and Finland, in order to see the associations of ethnicity, CRC and cancer gene mutations on the gut microbiota.
Primarily, a strong influence of ethnicity on the gut bacterial composition was noted in beta-diversity analysis. Significant differences were seen between the Iranian and Finnish controls (p=0.006) and between the Iranian and Finnish CRC patients (p=0.006), while the differences were not significant between patients and controls from the same ethnic group (Table II; Figure 2). Thus, ethnicity seems to have a more significant effect on the overall community composition than the CRC status. The differences in alpha diversity were however significant only between healthy controls of Iranian and Finnish origin and non-significant between CRC patients of these ethnic groups (Figure 1). Finnish controls had higher bacterial diversity and richness (Figure 1A) than Iranian controls. Similar higher gut microbiota alpha-diversity among individuals of north European origin (Dutch) compared to individuals of Middle-Eastern and African ethnic groups living in the same city (Amsterdam) has been reported earlier (11). A higher bacterial alpha diversity has also been associated with a healthier metabolic state (28).
In the present study, the most striking difference observed between the controls was the high abundance of Prevotella in Iranian controls, whereas the levels of these genera were very low in Finnish controls (Figure 3A, Table III). A marked lower abundance of Prevotella has been similarly reported in Dutch individuals as compared to the high abundance found in Moroccans, Turks, and Ghanaians (12). It therefore seems that individuals of north European descent have a low Provetella abundance, while individuals from Middle-Eastern and other neighbouring countries harbour high amounts of these bacteria. Previous studies have described presence of three enterotypes driven by Provetella, Bacteroides and Ruminococcus (29). The higher abundance of Bacteroides among Finnish controls and Provetella among Iranian controls could be indicative of higher prevalence of Bacteroides –driven enterotype among Finnish controls and Provetella –driven enterotype among Iranian controls. Moreover, the enterotype driven by Provetella is associated with non-western or fibre-rich diet and enterotype driven by Bacteroides is associated with animal protein and saturated fat diet (30).
Group-level comparisons based on community composition. Analysed for beta diversity differences among groups by PERMANOVA with 999 permutations and Benjamini-Hochberg FDR correction for multiple testing. The associations with FDR<0.25 (adjusted p-value) are shown.
Although many genera had differential abundance among the controls of Iranian and Finnish origin, only Clostridium showed significant difference in abundance among the CRC patients of these two populations, with lower abundance among Iranian CRC patients. The differential abundances of the genera between Iranian and Finnish groups might have been influenced by diet. The major differences in the dietary habits between the two ethnic groups are the absence of pork meat and alcohol intake among Iranian patients and controls. Higher abundance of Clostridium (seen in Finnish CRC patients) and higher abundance of Bacteroides (seen in Finnish controls) has been previously associated with alcohol consumption (31).
Gut bacterial dysbiosis is commonly seen in patients with CRC. Although in literature there are discrepancies regarding the bacterial profiles associated with CRC, higher abundance of Fusobacterium is one of the most frequent observations associated with CRC status (32, 33). In line with previous studies, we observed significant differences (p=0.01) in the presence of Fusobacterium among CRC patients and controls. Fusobacterium has been found to be strongly associated with CRC (32) and involved in metastasis of CRC (33).
In order to identify which bacteria were associated with CRC, we compared the bacterial composition of all CRC patients with all controls, irrespective of ethnicity. Only, Ruminococcus, had a significantly higher abundance in cancer patients compared to the controls, as well as in Iranian CRC compared to Iranian controls. According to the literature, Ruminococcus is able to break down resistant starches in the gut and supply nutrients to the host and also provides growth substrates for other bacteria (35). Differential abundance of Ruminococcus has been previously reported in CRC (34) and a recent study revealed higher abundance of Ruminococcus in patients at stage-1 of CRC compared to healthy individuals (36). Similarly, in our study, all Iranian CRC patients and the majority of the Finnish patients had early stage of cancer (Table I). Therefore, the increased abundance of Ruminococcus might be associated with early stages of CRC.
Genera which were significantly differentially abundant in the group-wise comparisons (False discovery rate, FDR<0.25). The Log10 fold change estimates are based on centered log-ratio (CLR)-transformed abundances between the group means.
The Finnish CRC had lower abundances of Bifidobacterium, Phascolarctobacterium and Lachnoclostridium than their healthy counterparts. Lower levels of both Phascolarctobacterium and Lachnospiraceae NK4A136 have been related to psychosocial stress in children (37). Bifidobacterium and Phascolarctobacterium confer beneficial health properties, and moreover, the presence of Phascolarctobacterium has been associated with a positive mood in the host (38). The higher abundances of these bacteria in Finnish controls might be linked to a healthier state.
Comparison of community composition between groups based on beta diversity. Redundancy analysis (RDA) showing community variations in the different group comparisons: (A) Iranian controls (IrC) and Finnish controls (HefeC), (B) The Iranian colorectal cancer patients (IrCRC) and the Finnish colorectal cancer patients (HefeCRC), (C) The Iranian controls (IrC) and the Iranian colorectal cancer patients (IrCRC), (D) The Finnish controls (HefeC) and the Finnish colorectal cancer patients (HefeCRC), and (E) All controls (AllC) and all colorectal cancer (AllCRC) patients.
Bacterial genera showing greatest differences in abundances (centered log-transformed; CLR) for various group comparisons. (A) Iranian controls (IrC) and Finnish controls (HefeC), (B) Iranian colorectal cancer patients (IrCRC) and Finnish colorectal cancer patients (HefeCRC), (C) colorectal cancer patients with and without the RAS mutations, (D) colorectal cancer patients with and without the NRAS mutations, and (E) colorectal cancer patients with and without the TP53 mutations.
The most significant association of host gene mutations with gut bacterial composition was a lower abundance of Barnesiella in patients with RAS mutations and higher abundance of Herbaspirillum in NRAS mutated patients compared to patients without these gene mutations. TP53 mutated patients showed higher abundance of Catenibacterium and lower abundance of Blautia compared to patients without mutations in TP53 gene. Barnesiella has been reported to protect against the colonization of the highly antibiotic – resistant Enterococcus faecium bacterium by restricting the growth of these pathogens (39). Importantly, Barnesiella has been associated with an anti-cancer immune response; these bacteria have been found to modulate the efficacy of the anti-cancer immunomodulatory agent, cyclophosphamide (40), by enhancing the Tc1 (cytotoxic T lymphocytes) and Th1 (T helper cell) response and restoring the function of intra-tumoral interferon gamma (IFN-γ)-producing T cells. Gene expression analysis from The Cancer Genome Atlas and the KFSYSCC (Koo Foundation Sun Yat-Sen Cancer Centre) data sets have revealed a suppressed Th1-/cytotoxicity in KRAS mutant colorectal cancer patients (41). The absence of Barnesiella in RAS mutated patients in our study, might also be associated with compromised tumour immune surveillance in these patients which would promote the development of the tumour.
Higher urinary abundance of Herbaspirillum has been linked to high risk of recurrence of bladder cancer (42). The abundance of Catenibacterium, seen in TP53 mutated patients, has been associated with animal fat diet (43) and diet rich in red meat, while food containing trans-fatty acids has been reported to be associated with TP53 mutations in colon cancer (44).
Since age and gender are important factors affecting gut microbiota composition, a limitation of our study is that the Finnish controls were of relatively younger age than the other groups (Table I). However, no significant association of age and sex with the observed abundances was found in our data. Moreover, the small number of patients with RAS and TP53 mutations in our study did not allow drawing any firm conclusions. A more in-depth study on larger number of patients with RAS and TP53 mutations is warranted to confirm the associations of bacterial genera with RAS and TP53 mutations found in this study.
We found that ethnicity is more significantly associated with gut microbiota composition than is the cancer status. Therefore, our results suggest that bacterial markers associated with cancer status should be further tested in patients from different ethnic groups, in order to assess their usefulness as cancer marker. Further studies on the association of bacterial genera with RAS and TP53 gene mutations could elucidate the role of gut microbiota in the growth of mutated cells and moreover, the impact of cellular pathways that are dysregulated in mutated cells on the abundance of specific bacteria. This knowledge could be useful in identifying new therapeutic approaches.
Acknowledgements
The Authors wish to thank Dr Ewen MacDonald for his linguistic correction of the manuscript. The study was supported by Sigrid Jusélius Foundation grant to SK. LL was supported by Academy of Finland (decision 295741). Rasoul Salehi greatly acknowledges financial support provided to him by the Iran National Scientific Foundation (INSF) through grant number 90002549.
Footnotes
Authors' Contributions
SK, VS, LL, PP and AK contributed to the study design. VS, SK, FS, OY, AK, TK, MT and RS contributed to sample collection and analysis. Bioinformatics and statistical data analysis was coordinated by LL. VS, LL, HR and SK contributed to manuscript preparation. All Authors approved the final version.
Conflicts of Interest
The Authors declare no conflicts of interest.
- Received January 15, 2020.
- Revision received February 4, 2020.
- Accepted February 11, 2020.
- Copyright© 2020, International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved
