Fecal Microbiome Associated with Both Colon Adenomas and Lifetime Colorectal Cancer Risk

Abstract

Background

Increasing data indicates the gut flora including bacteria and fungi combined with environmental factors are important in the pathogenesis of colorectal cancer (CRC). Understanding differences in the microbiome in patients with colon neoplasia will foster the development of biomarkers for early detection.

Aims

Determine the association of microbiome with presence of adenomas and predicted CRC risk.

Methods

In subjects referred for colonoscopy, the NCI CRC risk assessment tool was completed and stool for microbiome analysis as well as fecal immunochemical test (FIT) were collected. We calculated the microbiome alpha diversity using the Shannon index as well as individual bacterial and fungal species.

Results

Among 34 patients, we identified 10 with one or more adenomas. Only 2 patients were FIT positive. The median predicted lifetime CRC risk was 2.75% and the prevalence of adenoma was higher in the fourth quartile (P <0.001). The measured alpha diversity was somewhat higher in patients with adenomas (P = 0.07). We identified 4 bacterial species with an increased relative abundance among patients with adenomas [P <0.5]. Lifetime CRC risk was associated with 2 specific bacterial species, P. distasonis & E. hermannii [P = 0.05 & 0.09, respectively]. No associations were seen with fungal species and adenoma prevalence or lifetime CRC risk.

Conclusions

In addition to a strong correlation of predicted CRC risk and adenoma prevalence, we also found important differences in specific bacterial species and both adenoma prevalence and CRC risk. Larger trials are needed to potentially implement further data in the clinical setting.

Introduction

Increasing data indicates that the gut microbiome including bacteria (bacteriome) and fungi (mycobiome) are important in the pathogenesis of colorectal cancer (CRC). Approximately 50–60% of incident cases of CRC are estimated to be attributable to modifiable risk factors, including tobacco, obesity, physical inactivity, high consumption of red and processed meat, and low consumption of dietary fiber and whole grains [1]. Changes in the intestinal microbiota can combine with environmental factors to initiate and promote CRC and precursor colon adenomas [2]. Several previous studies have compared the microbiome in patients with colorectal carcinoma to healthy subjects. Patients with carcinoma are known to have both a less diverse bacteriome than controls [3] as well as expansions of species typically derived from the oral cavity [4]. The exact pathogenic mechanisms of bacteria in carcinogenesis are unclear and include chronic inflammation, dysfunction of immunity, increased levels of lipopolysaccharides, differences in bile acid composition, and leaky gut [1]. In addition to bacteria, distinct fungal community or mycobiome profiles have been identified in colorectal cancer patients though there are comparatively much fewer studies examining the role of the mycobiome compared to the bacteriome in this disease.

In addition to better understanding pathogenesis, analysis of the microbiome may augment current noninvasive colorectal screening approaches. Furthermore, understanding differences in the microbiome in patients with colon neoplasia compared to controls will foster the development of diagnostic/predictive biomarkers that may enable early detection of CRC and advanced adenomas to improve screening strategies [5–7].

We conducted the present study to extend these observations by enrolling a prospective cohort of subjects undergoing screening colonoscopy to examine the mycobiome and bacteriome in conjunction with data on established CRC risk factors. Our goals were to evaluate the diversity of the gut mycobiome and bacteriome in patients with colon adenomas and non-adenoma controls. In addition, we explored the value of mycobiome and bacteriome profiles compared to established CRC risk factors and fecal immunochemical testing (FIT) in predicting neoplasia.

Materials and Methods

Patients

The study was conducted at University Hospitals (UH) Cleveland Medical Center, an urban tertiary care center and three of its community-based satellite locations. The protocol was approved by the UH Institutional Review Board (study number 20200733) as well as the Case Comprehensive Cancer Center Protocol and Review Monitoring Board. Potential study patients were identified through the endoscopy scheduling system and medical records were reviewed for eligibility. Patients aged between 50–75 years, able to sign an informed consent, willing to collect a stool sample, and referred for colonoscopy for asymptomatic screening or surveillance because of a prior history of colon adenomas were eligible for enrolment. Patients with high-risk genetic conditions [familial adenomatous polyposis, hereditary non-polyposis CRC (Lynch Syndrome)], inflammatory bowel disease, hematochezia or melena within the past 30 days, or colonoscopy in the past 5 years were excluded.

Measures

Materials for stool collection for bacteriome and mycobiome analysis and home FIT (OC-Sensor Diana, Polymedco), along with written instructions were mailed to the participants. Stool was collected from 90 days up to 7 days prior to the colonoscopy date. Both specimens were mailed back to study personnel using self-addressed packaging. Stool samples were processed for FIT through the UH laboratory using a cut off of 50 µg/gm stool as a positive test and stool microbiome was analyzed in the Integrated Microbiome core laboratory at UH/Case.

Stool Analysis for Microbiome and FIT

For microbiome analysis, on receipt, samples were immediately placed into previously prepared fast prep tubes (MP, Cat# 5076–200–34340) containing 500 μL of glass beads (Sigma-Aldrich G8772–100g) and 1 mL ASL^™ lysis buffer (Qiagen DNA Extraction Kit). In order to minimize batch effect, all collected samples were stored at −20 °C then processed and analyzed concurrently for bacterial and fungal composition. All samples were processed at one time by one technician to reduce technical error.

DNA was extracted using QIAamp Fast DNA Stool Mini kit (Qiagen GmpH, Hilden, Germany) according to manufacturer’s instructions. All swabs were transferred to tubes with 1 mL of InhibitEX lysis Buffer, incubated for 1 hour at 75 °C and shaken using Fastprep 96 two times for five minutes each at 1800 RPM. Equal amounts of 100 % ethanol and lysate were then mixed and passed through HiBind DNA Mini Columns (Omega Bio-tek, Georgia, USA). Resulting DNA pellets were eluted using 50 μL Molecular grade water.

Amplification of microbial 16S and 5.8S rRNA genes were performed using 16S −804 (5’-(TCC TAC GGG AGG CAG CAGT-3’) and 16S-515 (5’-GGA CTA CCA GGG TAT CTA ATC CTG- 3’), ITS1 (5’-(TCC GTA GGT GAA CCT GCG G- 3’), and ITS4 (5’-TCC TCC GCT TAT TGA TAT GC- 3’) primers, respectively. Q5 High-Fidelity Master Mix was used for PCR at a 1X concentration, along with a double volume of molecular grade water and 5 μL 100 mM of each primer. 1.5 μL of undiluted DNA was added to each 50 μL reaction. Thermocycling conditions consisted of an initial denaturation step (3 minutes at 98 °C), followed by 30 cycles of denaturation (10 seconds at 98 °C), annealing (10 seconds at 55 °C for the 16S primers and 20 seconds at 58 °C for the ITS primers), extension (10 seconds at 72 °C) then a final extension step of 3 minutes 72 °C. Ten μL of each PCR product was separated using gel electrophoresis on 1.5 % agarose gel (containing 7 μg/mL ethidium bromide).

Equal volumes of gel-extracted bacterial 16S rRNA and fungal ITS reaction products were pooled and cleaned using AMPure XP beads (Beckman Coulter, CA, USA) to remove excess unused primers. Using the Ion Plus fragment Library kit (ThermoFisher Scientific) pooled amplicons were exposed to end-repair enzyme for 20 minutes at room temperature. Following a second AMPure cleanup, ligation of Ion Torrent P1 and unique barcoded 'A' adaptors for each pooled sample was performed at 25 °C for 30 minutes. After AMPure removal of residual adaptors, 10 μL of each barcoded sample was pooled and mixed in one tube. Pooled library samples were then concentrated to one-fourth volume using Labconco Vacuum for 1 hour with heat. The concentrated pool was then size-selected for the anticipated base pair range of both 16S and ITS (300–800 bp) using Pippin Prep (Sage Bioscience). Amplification of the size-selected library using platinum PCR SuperMix High Fidelity (ThermoFisher Scientific) and the provided Ion Torrent Library amplification primer was done with conditions of 3 minutes denaturation at 95 °C, 7 cycles of denaturation at 95 °C for 15 seconds, annealing at 58 °C for 15 seconds, and extension at 70 °C for 90 seconds, and a final extension step of 5 minutes at 70 °C. Subsequent amplified library was then quantified using Ion Library TaqMan Quantification Kit (Applied Biosystems by Life Technologies) on StepOne qPCR instrument. A dilution of 300 pM was added into IonSphere templating reaction on the Ion Chef. Library sequencing was completed on Ion Torrent S5 sequencer (ThermoFisher Scientific).

Analysis of barcode-sorted samples was performed in a custom pipeline based on Greengenes V13_8 and Unite V7.2 databases designed for the taxonomic classification of 16SrRNA and ITS sequences, respectively. Qiime software was used for downstream data analysis.

Clinical Measures

The colonoscopy was performed per standard clinical care and the endoscopist was unaware of the findings of stool testing. The colonoscopy findings were reviewed and included the most advanced of no neoplasia, non-advanced adenoma (tubular adenoma(s) < 1 cm), advanced adenoma (adenoma > 1 cm, villous features or high-grade dysplasia), serrated adenoma and adenocarcinoma. All colonoscopies had adequate bowel preparation and the endoscopist was able to reach the cecum.

In addition to clinical data, all subjects completed the National Cancer Institute Colorectal Cancer Risk Assessment tool, a validated instrument that contains items about demographics, family history, lifestyle factors and personal medical history to estimate future absolute CRC risk (17).

Bioinformatic Analysis

Raw 16s and ITS data of the bacterial read counts were obtained and loaded to R version 4.0.3. Sample meta data were obtained and using R package microbiome 1.12.0 and phyloseq 1.34.0 a phyloseq object containing a read count matrix (species level identification), taxonomic table (kingdom to species) and metadata table was made. When dealing with both 16s and ITS data, each were initially handled separately and then combined after the read counts were normalized and transformed to relative abundance. Prior to creating the phyloseq object for the 16s and ITS data, the taxonomic table was cleaned to remove species and phyla that were annotated as “unidentified” or lacking identification at a species level annotated as an empty cell. This step allowed us to ensure only species that are identifiable are being utilized in the analysis. The main filtering step performed was on the total read counts per sample, where a cutoff of 500 read counts in the 16s data was used as a minimum requirement. Additionally, samples with missing meta data annotations were also removed (2 samples).

Composition bar graphs were generated on 16s and ITS relative abundance data separately after aggregating the data to the phyla level, and filtering on phyla prevalence of 25% amongst all samples. Ordinate analysis/principal component analysis was performed on bacteriome and mycobiome data separately, after aggregating data to the genus level as a mean to reduce complexity. The data were then filtered on prevalence of 25% amongst all samples. Ordinate analysis was performed using the “ordinate” function part of microbiome R package version 1.12.0 with method used as “MDS” (multidiemensional scaling) and distances using “Bray-curtis” method. Venn diagrams were generated using the 16s and ITS species level relative abundance. The core microbiome in patients with and without adenomas was determined using a prevalence of 50% within samples in each group. 16s and ITS relative abundance data were first filtered on species with a prevalence of 25% within all samples, and then the respective phyloseq objects were merged.

Wilcoxon-rank-sum nonparametric test was used to test for significantly different species between patients with and without adenomas, and the foldchange was calculated using mean relative abundance of a species within each group using the foldchange function from R package. Box plots were generated for all the significantly different species. Alpha diversity was calculated on all 16s samples using the Shannon index. The Wilcoxon-rank-sum-test was used to test for significant differences between the two groups. Spearman correlation was used to determine which of the core species microbiome (prevalence 25%) correlated with the lifetime CRC risk annotation through the NCI questionnaire.

Results

Our study cohort included 35 patients who were referred for screening or surveillance colonoscopy. We collected fecal samples and clinical data from 34 subjects, including 24 women and 10 men. The median age was 58 years and 91% of the subjects were white (Table 1).

Table 1.

Demographic Characteristics of the Cohort

	Adenoma (n = 10; %= 29%)	Non-adenoma (n = 23; %= 67%)
Age (years)
(50–59)	5 (50%)	14 (60%)
(60 75)	5 (50%)	9 (39%)
Race
Black	2 (20%)	0
White	S (80%)	22 (95%)
Other	0	1 (4%)
Gender
Male	5 (50%)	5 (21%)
Female	5 (50%)	18 (78%)

All 34 patients underwent their scheduled colonoscopy. Colonoscopy detected neoplastic polyps in 11 (33%) subjects; one of the polyps was not retrieved. All polyps were benign; with the largest polyp measuring 13 mm. Of those, 2 subjects were found to have advanced adenomas and 2 subjects were observed to have sessile serrated adenomas; the latter were included in the adenoma group in the analysis. Only 2 patients had a positive FIT, one of whom had an advanced adenoma and the other had a normal examination. On subsequent medical record review, none of the patients were later diagnosed with CRC or advanced adenomas or had a follow up colonoscopy.

Based on the NCI risk calculator the median predicted lifetime CRC risk was 2.75% with an interquartile range of 2.3%. When patients were divided into quartiles based on their predicted lifetime CRC risk, the prevalence of adenomas was 1.5%, 37.5%, 1.5% and 62.5% in the lowest through highest risk quartile, respectively (P < 0.001). The prevalence of advanced adenoma was 1.5% in both the 2^nd and 4^th quartiles whereas none were found in 1^st and 3^rd quartiles.

On stool microbiome analysis, we compared the composition of the bacterial and fungal species at the phyla level between patients with and without adenomas. There were no clear clustering patterns between the two groups and on principal component analysis for bacteriome and mycobiome; there was no clear clustering of samples based on their adenoma status (figures 1a and b). The observed Shannon alpha diversity was significantly higher in patients with adenomas compared to those without (P = 0.01) while the measured one was close to significance (P = 0.07) (figures 2a and b). On further analysis, ten unique species were found in the adenoma group compared to none in the non-adenoma and nineteen species were common between the two groups.

Fig. 1.

Clustering of bacteriome (panel A) and mycobiome (panel B) by adenoma status

Fig. 2.

Observed (Panel A) and Measured (Panel B) Shannon Alpha Diversity

We identified four bacterial species, including Veillonella parvula, Blautia obeum, Shigella boydii and Lysinibacillus boronitolerans, which showed an increased relative abundance among patients with adenomas (mean = 0.0005, 0.0007, 0.0007 and 0.0002; P = 0.04, 0.01, 0.04 and 0.04, respectively) compared to others. One bacterial species, Bacteroides uniformis and one fungal species, Metarhizium anisopliae have an increased relative abundance among the non-adenoma group (mean = 0.02 and zero; P = 0.02 and 0.04, respectively) compared to the adenoma group.

In addition, we found six bacterial species whose relative abundance has significantly correlates with the lifetime CRC risk. A positive correlation was found in two specific bacterial species, including Parabacteroides distasonis (Figure 3a, Spearman’s rho 0.3; P = 0.05) and Escherichia hermannii (Figure 3b, Spearman’s rho 0.3; P = 0.09), whereas a negative correlation was observed in four other bacterial species; Lactobacillus zeae, Eubacterium dolichum, Ifidobacterium adolescentis and Eggerthella lenta (Spearman’s rho 0.3, 0.3, 0.4 and 0.4; P = 0.09, 0.07, 0.01 and 0.005, respectively) (Figure 4a, b, c &d). Overall, no associations were seen with fungal species and adenoma prevalence or lifetime CRC risk.

Fig. 3.

Correlation with the lifetime CRC risk in Parabacteroides distasonis (Panel A) and Escherichia hermannii (Panel B)

Fig. 4.

Correlation with the lifetime CRC risk in Lactobacillus zeae (Panel A), Eubacterium dolichum (Panel B), Ifidobacterium adolescentis (Panel C) and Eggerthella lenta (Panel D)

Discussion

It is known that changes in the intestinal microbiome can combine with environmental factors to initiate and promote CRC and precursor colon adenomas. Understanding differences in the microbiome in patients with colon neoplasia compared to controls may foster the development of biomarkers that may enable early detection of CRC and advanced adenomas to improve screening strategies. In this prospective study of patients referred for outpatient colonoscopy, we identified several bacterial species that differed between patients with and without colon adenomas. More significantly, we found six species that correlated with an individual’s lifetime CRC risk which if validated in additional studies, could also inform work in CRC pathogenesis.

Previous investigations have identified differences in the bacteriome of patients with CRC. In particular, increased levels of Fusobacterium nucleatum DNA and RNA sequences have been observed in tumor specimens, especially in right-sided cancers with microsatellite instability [8–10]. Others have reported an association with enterotoxigenic Bacteroides fragilis [11], E. coli [12] and Streptococcus bovis [13], with increases in Malasseziomycetes and decreases in Saccaromycetes and Pneumocystidomycetes [14]. Another study compared the bacteriome in patients with serrated adenomas and controls and only found a modest difference in taxa and diversity [18]. In addition, the feasibility of microbiome analysis as a screening tool has been evaluated. In a study of 490 patients, including 120 with cancer and 198 with adenomas, analysis of 16S rRNA gene sequencing was able to detect 92% of cancers and 45% of adenomas with a specificity of 90% and outperformed FIT [5]. In another study that included 104 patients with cancer, 103 with advanced adenomas and 102 controls, relative abundance of Fusobacterium nucleatum improved the prediction of FIT alone in the detection of both cancer and advanced adenomas [6].

Whereas numerous studies have examined the bacterial community in CRC, very few have examined the mycobiome [14,16]. In fact, in preliminary studies of patients with colon adenomas and controls, conducted by our team, there was a greater diversity of the mycobiome than microbiome. However, in the present study we did not identify any differences in the fungal flora among individuals with adenoma, which may relate to the relatively small sample size.

Information obtained from this study may guide the development of a microbial biomarker for an accurate detection of advanced colorectal lesions or early detection of CRC. By performing biomarker discovery within a screening population, the generalizability of the findings to future screening cohorts is likely to be significant. The validation of a microbiome-based biomarker will require many steps before enabling full implantation. These could then be combined by means of a biomarker panel for the development of a rapid test, which has the potential of being integrated into national CRC screening programs. The cost-effectiveness of adding a microbial biomarker to the FIT test should be carefully evaluated before implementation.

We recognize several limitations of the study. It was performed on a small sample in a single center, which could affect the generalizability of the results. Gender and ethnicity characteristics were not balanced, as most patients were white and female. Although all patients were referred for routine colonoscopy for screening or adenoma surveillance, there may have been a bias toward a high risk sample, with colonoscopy selected over non-invasive methods. We also assessed CRC risk using a predictive model in lieu of patients followed longitudinally for development of CRC. We have studied stool microbiome, but we did not examine the microbiome at the mucosal level to assess for any difference. The predominance of specific microbiome over others in patients with adenoma was not explained in terms of metabolome.

In summary, we have found differences in the fecal microbiome that correlated not only with the presence of adenomas but a patient’s lifetime CRC risk. Larger trials are needed to potentially implement further data in the clinical setting.

Abbreviations

CRC

colorectal cancer

FIT

fecal immunochemical test

NCI

National Cancer Institute

MDS

multidimensional scaling

DNA

deoxyribonucleic acid