Viral and Bacterial Communities of Colorectal Cancer

19 Colorectal cancer is the second leading cause of cancer-related death in the United States and is a primary 20 cause of morbidity and mortality throughout the world. Colorectal cancer development has been linked to 21 differences in colonic bacterial community composition. Viruses are another important component of the 22 colonic microbial community, however they have yet to be studied in colorectal cancer despite their oncogenic 23 potential. We evaluated the colorectal cancer virome (virus community) in stool using a cohort of 90 human 24 subjects with either healthy, adenomatous (precancerous), or cancerous colons. We utilized 16S rRNA gene, 25 whole shotgun metagenomic, and purified virus metagenomic sequencing methods to compare the colorectal 26 cancer virome to the bacterial community. We identified no detectable difference in viral diversity (alpha or 27 beta) between healthy, adenomatous, or cancerous colonic samples, but more sophisticated random forest 28 models identified striking differences in the virome. The majority of the cancer-associated virome consisted 29 of temperate bacteriophages, suggesting that the community was indirectly linked to colorectal cancer by 30 modulating bacterial community structure and function. Our data suggest that the influential phages do 31 not exclusively infect influential bacteria, but rather act through the community as a whole. These results 32 provide foundational evidence that bacteriophage communities are associated with colorectal cancer and 33 likely impact cancer progression by altering the bacterial host communities. 34 2 peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/152868 doi: bioRxiv preprint first posted online Jun. 20, 2017; Significance Statement 35 Colorectal cancer is a leading cause of cancer-related death in the United States and worldwide. Its risk and 36 severity have been linked to colonic bacterial community composition. Although viruses have been linked 37 to other cancers and diseases, little is known about colorectal cancer virus communities. We addressed 38 this knowledge gap by identifying differences in colonic virus communities in the stool of colorectal cancer 39 patients and how they compared to bacterial community differences. The results suggested an indirect role 40 for the virome in impacting colorectal cancer by modulating their associated bacterial community. These 41 findings both support a biological role for viruses in colorectal cancer and provide a new understanding of 42 basic colorectal cancer etiology. 43 3 peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/152868 doi: bioRxiv preprint first posted online Jun. 20, 2017; Introduction 44 Due to their mutagenic abilities and propensity for functional manipulation, human viruses are strongly 45 associated with, and in many cases cause, cancer (1–4). Because bacteriophages (viruses that specifically 46 infect bacteria) are crucial for bacterial community stability and composition (5–7) and have been implicated 47 as oncogenic agents (8–11), bacteriophages have the potential to indirectly impact cancer. The gut virome 48 (the virus community of the gut) therefore has the potential to impact health and disease. Altered human 49 virome composition and diversity have been identified in diseases including periodontal disease (12), HIV 50 (13), cystic fibrosis (14), antibiotic exposure (15, 16), urinary tract infections (17), and inflammatory bowel 51 disease (18). The strong association of bacterial communities with colorectal cancer and the precedent for 52 the virome to impact other human diseases suggest that colorectal cancer may be associated with altered 53 virus communities. 54 Colorectal cancer is the second leading cause of cancer-related deaths in the United States (19). The 55 US National Cancer Institute estimates over 1.5 million Americans were diagnosed with colorectal cancer 56 in 2016 and over 500,000 Americans died from the disease (19). Growing evidence suggests that an 57 important component of colorectal cancer etiology may be perturbations in the colonic bacterial community 58 (8, 10, 11, 20, 21). Work in this area has led to a proposed disease model in which bacteria colonize the 59 colon, develop biofilms, promote inflammation, and enter an oncogenic synergy with the cancerous human 60 cells (22). This association also has allowed researchers to leverage bacterial community signatures as 61 biomarkers to provide accurate, noninvasive colorectal cancer detection from stool (8, 23, 24). While an 62 understanding of colorectal cancer bacterial communities has proven fruitful both for disease classification 63 and for identifying the underlying disease etiology, bacteria are only a subset of the colon microbiome. 64 Viruses are another important component of the colon microbial community that have yet to be studied in 65 the context of colorectal cancer. We evaluated disruptions in virus and bacterial community composition 66 in a human cohort whose stool was sampled at the three relevant stages of cancer development: healthy, 67 adenomatous, and cancerous. 68 Colorectal cancer progresses in a stepwise process that begins when healthy tissue develops into a 69 4 peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/152868 doi: bioRxiv preprint first posted online Jun. 20, 2017; precancerous polyp (i.e., adenoma) in the large intestine (25). If not removed, the adenoma may develop 70 into a cancerous lesion that can invade and metastasize, leading to severe illness and death. Progression to 71 cancer can be prevented when precancerous adenomas are detected and removed during routine screening 72 (26, 27). Survival for colorectal cancer patients may exceed 90% when the lesions are detected early and 73 removed (26). Thus, work that aims to facilitate early detection and prevention of progression beyond early 74 cancer stages has great potential to inform therapeutic development. 75 Here we address the knowledge gap of whether virus community composition is altered in colorectal cancer 76 and, if it is, how those differences might impact cancer progression and severity. We also aimed to evaluate 77 the virome’s potential for use as a diagnostic biomarker. The implications of this study are threefold. First, 78 this work supports a biological role for the virome in colorectal cancer development and suggests that more 79 than the bacterial members of the associated microbial communities are involved in the process. Second, we 80 present a supplementary, or even alternative, virus-based approach for classification modeling of colorectal 81 cancer using stool samples. Third, we provide initial support for the importance of studying the virome as a 82 component of the microbiome ecological network, especially in cancer. 83 Results 84 Cohort Design, Sample Collection, and Processing 85 Our study cohort consisted of 90 human subjects, 30 of whom had healthy colons, 30 of whom had 86 adenomas, and 30 of whom had carcinomas (Figure S1). Half of each stool sample was used to sequence 87 the bacterial communities using both 16S rRNA gene and shotgun sequencing techniques. The 16S 88 rRNA gene sequencing was performed for a previous study, and the sequences were re-analyzed using 89 contemporary methods (8). The other half of each stool sample was purified for virus like particles (VLPs) 90 before genomic DNA extraction and shotgun metagenomic sequencing. In the VLP purification, cells were 91 disrupted and extracellular DNA degraded (Figure S1) to allow the exclusive analysis of viral DNA within 92 virus capsids. In this manner, the extracellular virome of encapsulated viruses was targeted. 93 5 peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/152868 doi: bioRxiv preprint first posted online Jun. 20, 2017; Each extraction was performed with a blank buffer control to detect contaminants from reagents or other 94 unintentional sources. Only one of the nine controls contained detectable DNA at a minimal concentration of 95 0.011 ng/μl, thus providing evidence of the enrichment and purification of VLP genomic DNA over potential 96 contaminants (Figure S2 A). As was expected, these controls yielded few sequences and were almost 97 entirely removed while rarefying the datasets to a common number of sequences (Figure S2 B). The high 98 quality phage and bacterial sequences were assembled into highly covered contigs longer than 1 kb (Figure 99 S3). Because contigs represent genome fragments, we further clustered related bacterial contigs into 100 operational genomic units (OGUs) and viral contigs into operational viral units (OVUs) (Figure S3 S4) to 101 approximate organismal units. 102 Unaltered Virome Diversity in Colorectal Cancer 103 Microbiome and disease associations are often described as being of an altered diversity (i.e., “dysbiotic”). 104 We therefore first evaluated the influence of colorectal cancer on virome OVU diversity. We evaluated 105 differences in communities between disease states using the Shannon diversity, richness, and Bray-Curtis 106 metrics. We observed no significant alterations in either Shannon diversity or richness in the diseased states 107 as compared to the healthy state (Figure S5 C-D). There was no statistically significant clustering of the 108 disease groups (ANOSIM p-value = 0.4, Figure S5). Notably, there was a significant difference between 109 the few blank controls that remained after rarefying the data and the other study groups (ANOSIM p-value 110 < 0.001, Figure S6), further supporting the quality of the sample set. In summary, standard alpha and beta 111 diversity metrics were insufficient for capturing virus community differences between disease states (Figure 112 S5). This is consistent with what has been observed when the same metrics were applied to 16S rRNA 113 sequenced and metagenomic samples (8, 23, 24) and points to the need for alternate approaches to detect 114 the impact of colorectal cancer disease state on these communities. 115 6 peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/152868 doi: bioRxiv preprint first posted online Jun. 20, 2017; Altered Virome Composition in Colorectal Cancer 116 As opposed to the diversity metrics discussed above, OTU-based relative abundance profiles generated 117 from 16S rRNA gene sequences are effective feature sets for classifying stool samples as originating from 118 individuals with healthy, adenomatous, or cancerous colons (8, 23). The exceptional performance of bacteria 119 in these classification models supports a role for bacteria in colorectal cancer. We built off of these findings by 120 evaluating the ability of virus community signatures to classify stool samples and compared their performance 121 to models built using bacterial community signatures. 122 To identify the altered virus communities associated with colorectal cancer, we built and tested random forest 123 models for classifying stool samples as belonging to individuals with either cancerous or healthy colons. We 124 confirmed that our bacterial 16S rRNA gene model replicated the performance of the original report which 125 used logit models instead of random forest models (Figure 1 A) (8). We then compared the bacterial OTU 126 model to a model built using OVU relative abundances. The viral model performed as well as the bacterial 127 model (corrected p-value = 0.4), with the viral and bacterial models achieving mean area under the curve 128 (AUC) values of 0.793 and 0.796, respectively (Figure 1 A B). To evaluate the ability of both bacterial and 129 viral biomarkers to classify samples, we built a combined model that used both bacterial and viral community 130 data. The combined model yielded a modest but statistically significant performance improvement beyond 131 the viral (corrected p-value = 0.002) and bacterial (corrected p-value = 0.002) models, yielding an AUC of 132 0.816 (Figure 1 A B). The combined features from the virus and bacterial communities improved our ability 133 to classify stool as belonging to individuals with cancerous colons. 134 To determine the advantage of viral metagenomic methods over bacterial metagenomic methods, we 135 compared the viral model to a model built using OGU relative abundance profiles from bacterial metagenomic 136 shotgun sequencing data. This model performed worse than the other models (mean AUC = 0.505) (Figure 137 1 A B). Because the coverage provided by the metagenomic sequencing was not as deep as the equivalent 138 16S rRNA gene sequencing, we attempted to compare the approaches at a common sequencing depth. 139 This investigation revealed that the bacterial 16S rRNA gene model was strongly driven by sparse and low 140 abundance OTUs (Figure S7). Removal of OTUs with a median abundance of zero resulted in the removal 141 7 peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/152868 doi: bioRxiv preprint first posted online Jun. 20, 2017; of six OTUs, and a loss of model performance down to what was observed in the metagenome-based model 142 (Figure S7 A). The majority of these OTUs had a relative abundance lower than 1% across the samples 143 (Figure S7 B). Although the features in the viral model also were of low abundance (Figure S9 F), the 144 coverage was sufficient for high model performance, likely because viral genomes are orders of magnitude 145 smaller than bacterial genomes. 146 The association between the bacterial and viral communities and colorectal cancer was driven by a few 147 important microbes. Fusobacterium was the primary driver of the bacterial association with colorectal cancer, 148 which is consistent with its previously described oncogenic potential (Figure 1 C)(22). The virome signature 149 also was driven by a few OVUs, suggesting a role for these viruses in tumorigenesis (Figure 1 D). The 150 identified viruses were bacteriophages, belonging to Siphoviridae, Myoviridae, and “unclassified” phage taxa. 151 Many of the important viruses were unidentifiable (denoted “unknown”). This is common in viromes across 152 habitats; studies have reported as much as 95% of virus sequences belonging to unknown genomic units 153 (14, 28–30). When the bacterial and viral community signatures were combined, both bacterial and viral 154 organisms drove the community association with cancer (Figure 1 E). 155 Shifted Phage Influence Between Cancer Progression Stages 156 Because previous work has identified shifts in which bacteria were most important at different stages of 157 colorectal cancer (8, 20, 22), we explored whether shifts in the relative influence of specific phages could be 158 detected between healthy, adenomatous, and cancerous colons. We evaluated community shifts between 159 the two disease stage transitions (healthy to adenomatous and adenomatous to cancerous) by building 160 random forest models to compare only the diagnosis groups around the transitions. While bacterial OTU 161 models performed equally well for all disease class comparisons, the virome model performances differed 162 (Figure S8 A-B). Like bacteria (Figure S8 F-H), different virome members were important between the healthy 163 to adenomatous and adenomatous to cancerous stages (Figure S8 C-E). 164 After evaluating our ability to classify samples between two disease states, we performed a three-class 165 random forest model including all disease states. The 16S rRNA gene model yielded a mean AUC of 0.771 166 8 peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/152868 doi: bioRxiv preprint first posted online Jun. 20, 2017; and outperformed the viral community model, which yielded a mean AUC of 0.699 (p-value < 0.001, Figure 167 S9 A-C). The microbes important for the healthy versus cancer and healthy versus adenoma models were 168 also important for the three-class model (Figure S9 D-E). The most important bacterium in the two and three 169 class models was the same Fusobacterium (OTU 4) (Figure 1 C, Figure S9 D). The viruses most important 170 to the three-class model were identified as bacteriophages (Figure 1 D, Figure S9 E), but not all important 171 OVUs were of increased abundance in the diseased state (Figure S9 F). 172 Bacteriophage Dominance in Colorectal Cancer Virome 173 Differences in the colorectal cancer virome could have been driven directly by eukaryotic viruses or indirectly 174 by bacteriophages acting through their bacterial hosts. To better understand the types of viruses that were 175 important for colorectal cancer, we identified the virome OVUs as being similar to either eukaryotic viruses 176 or bacteriophages. The most important viruses to the classification model were identified as bacteriophages 177 (Figure S9). Overall, we were able to identify 78.8% of the OVUs as known viruses, and 93.8% of those 178 viral OVUs aligned to bacteriophage reference genomes. It is important to note that this could have been 179 influenced by our methodological biases against enveloped viruses (more common of eukaryotic viruses than 180 bacteriophage), due to chloroform and DNase treatment for purification. 181 We evaluated whether the phages in the community were primarily lytic (i.e. obligately lyse their hosts after 182 replication) or temperate (i.e. able to integrate into their host’s genome to form a lysogen, and subsequently 183 transition to a lytic mode). We accomplished this by identifying three markers for temperate phages in the 184 OVU representative sequences: 1) presence of phage integrase genes, 2) presence of known prophage 185 genes, according the the ACLAME (A CLAssification of Mobile genetic Elements) database, and 3) nucleotide 186 similarity to regions of bacterial genomes (29, 31, 32). We found that the majority of the phages were 187 temperate and that the overall fraction of temperate phages remained consistent throughout the healthy, 188 adenomatous, and cancerous stages (Figure S10 E). These findings were consistent with previous reports 189 suggesting the gut virome is primarily composed of temperate phages (13, 18, 31, 33). 190 9 peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/152868 doi: bioRxiv preprint first posted online Jun. 20, 2017; Community Context of Influential Phages 191 Because the link between colorectal cancer and the virome was driven by bacteriophages, we hypothesized 192 that the influential phages were primarily predators of the influential bacteria, and thus influenced their relative 193 abundance through predation. If this hypothesis were true, we would expect a correlation between the relative 194 abundances of influential bacteria and phages. Instead, we observed a strikingly low correlation between 195 bacterial and phage relative abundances (Figure 2 A,C). Overall, there was an absence of correlation 196 between the most influential OVUs and bacterial OTUs (Figure 2 B). This evidence supported our null 197 hypothesis that the influential phages were not primarily predators of influential bacteria. 198 Given these findings, we hypothesized that the most influential phages were acting by infecting a wide range 199 of bacteria in the overall community, instead of just the influential bacteria. In other words, we hypothesized 200 that the influential bacteriophages were community hubs (central members) within the bacteria and phage 201 interactive network. We investigated the potential host ranges of all phage OVUs using a previously 202 developed random forest model that relies on sequence features to predict which phages infected which 203 bacteria in the community (Figure 3 A) (34). The predicted interactions were then used to identify phage 204 community hubs. We calculated the alpha centrality (measure of importance in the ecological network) 205 of each phage OVU’s connection to the rest of the network. The phages with high centrality values were 206 defined as community hubs. Next, the centrality of each OVU was compared to its importance in the 207 colorectal cancer classification model. Phage OVU centrality was significantly and positively correlated 208 with importance to the disease model (p-value = 0.02, R = 0.14), suggesting that phages important in 209 driving colorectal cancer also were more likely to be community hubs (Figure 3 B). Together these findings 210 supported our hypothesis that influential phages were hubs within their microbial communities and had 211 broad host ranges. 212 10 peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/152868 doi: bioRxiv preprint first posted online Jun. 20, 2017; Discussion 213 Because of their propensity for mutagenesis and capacity for modulating their host functionality, many viruses 214 are oncogenic (1–4). Some bacteria also have oncogenic properties, suggesting that bacteriophages may 215 play an indirect role in promoting carcinogenesis by influencing bacterial community composition and 216 dynamics (8–10). Despite their carcinogenic potential and the strong association between bacteria and 217 colorectal cancer, a mechanistic link between virus colorectal communities and colorectal cancer has yet to 218 be evaluated. Here we show that, like colonic bacterial communities, the colon virome was altered in patients 219 with colorectal cancer relative to those with healthy colons. Our findings support a working hypothesis for 220 oncogenesis by phage-modulated bacterial community composition. 221 Here, we have begun to delineate the role the colonic virome plays in colorectal cancer (Figure 4 A). We 222 found that basic diversity metrics of alpha diversity (richness and Shannon diversity) and beta diversity 223 (Bray-Curtis dissimilarity) were insufficient for identifying virome community differences between healthy 224 and cancerous states. By implementing a more sophisticated machine learning approach (random forest 225 classification), we detected strong associations between the colon virus community composition and 226 colorectal cancer. The colorectal cancer virome was composed primarily of bacteriophages. These phage 227 communities were not exclusively predators of the most influential bacteria, as demonstrated by the lack 228 of correlation between the abundances of the bacterial and phage populations. Instead, we identified 229 influential phages as being community hubs, suggesting phages influence cancer by altering the greater 230 bacterial community instead of directly modulating the influential bacteria. Our previous work has shown that 231 modifying colon bacterial communities alters colorectal cancer progression and tumor burden in mice (10, 232 20). This provides a precedent for phage indirectly influencing colorectal cancer progression by altering the 233 bacterial community composition. Overall, our data support a model in which the bacteriophage community 234 modulates the bacterial community, and through those interactions indirectly influences the bacteria driving 235 colorectal cancer progression (Figure 4 A). Although our evidence suggested phages indirectly influenced 236 colorectal cancer development, we were not able to rule out the role of phages directly interacting with the 237 human host (35, 36). 238 11 peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/152868 doi: bioRxiv preprint first posted online Jun. 20, 2017; In addition to modeling the potential connections between virus communities, bacteria communities, and 239 colorectal cancer, we also used our data and existing knowledge of phage biology to develop a working 240 hypothesis for the mechanisms by which this may occur. This was done by incorporating our findings into the 241 current model for colorectal cancer development (Figure 4 B) (22). We hypothesize that the process began 242 with broadly infectious phages in the colon lysing and thereby disrupting the existing bacterial communities. 243 This shift led to novel niche space that enabled opportunistic bacteria (such as Fusobacterium nucleatum) 244 to colonize. Once the initial influential founder bacteria established themselves in the epithelium, secondary 245 opportunistic bacteria were able to adhere to the founders, colonize, and begin establishing a biofilm. Phages 246 may have played a role in biofilm dispersal and growth by lysing bacteria within the biofilm, a process 247 important for effective biofilm growth (37). The oncogenic bacteria may then have been able to transform 248 the epithelial cells and disrupt tight junctions to infiltrate the epithelium, thereby initiating an inflammatory 249 immune response. As the adenomatous polyps developed and progressed towards carcinogenesis, we 250 observed a shift in the phages and bacteria whose relative abundances were most influential. As the bacteria 251 entered their oncogenic synergy with the epithelium, we conjecture that the phages continued mediating 252 biofilm dispersal. This process would thereby support the colonized oncogenic bacteria by lysing competing 253 cells and releasing nutrients to other bacteria in the form of cellular lysates. In addition to highlighting the 254 likely mechanisms by which the colorectal cancer virome is interacting with the bacterial communities, this 255 outline will guide future research investigations of the role the virome plays colorectal cancer. 256 A notable finding was the poor performance observed using bacterial metagenomic methods compared 257 to the performance of models using viral metagenomes or 16S rRNA gene sequences. We believe this 258 observation speaks to the importance of sequencing coverage in microbial community studies and the 259 advantage of the high coverage of 16S rRNA gene sequencing relative to the lower per OTU coverage 260 possible using whole metagenomic shotgun sequencing. To demonstrate this concept, consider that six 261 bacterial OTUs drove the performance of the 16S rRNA gene classification model and these OTUs were all 262 sparsely present and lowly abundant. Filtration of OTUs with a median relative abundance of zero resulted 263 in the removal of the six important OTUs and reduced model performance to being nearly random like 264 the bacterial metagenomic model. The bacterial metagenomic OGUs represented only the most abundant 265 12 peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/152868 doi: bioRxiv preprint first posted online Jun. 20, 2017; taxa, which was not informative for this application. There has been some success in using shotgun 266 metagenomic approaches for stool colorectal cancer classification (24), but this previous approach relied on 267 lowly abundant signatures and did not utilize OGU clustering, as done here. In that former case, the models 268 only performed as well as the 16S rRNA gene model (24). Thus, the targeted 16S rRNA gene sequencing 269 approach, which represented only a fraction of the bacterial metagenomic sequencing depth, was more 270 effective for detecting colorectal cancer in stool samples. Despite a loss of enthusiasm for 16S rRNA gene 271 sequencing in favor of shotgun metagenomic techniques, 16S rRNA gene sequencing is still a superior 272 methodological approach for some important applications. 273 In addition to the diagnostic ramifications for understanding the colorectal cancer microbiome, our findings 274 suggest that viruses, while understudied and currently under-appreciated in the human microbiome, are 275 likely to be an important contributor to human disease. Viral community dynamics have the potential to 276 provide an abundance of information to supplement those of bacterial communities. Evidence has suggested 277 that the virome is a crucial component to the microbiome and that bacteriophages are important players. 278 Bacteriophage and bacterial communities cannot maintain stability and co-evolution without one another (6, 279 38). Not only is the human virome an important element to consider in human health and disease (12–18), 280 but our findings support that it is likely to have a significant impact on cancer etiology and progression. 281 Materials & Methods 282 This study was approved by the University of Michigan Institutional Review Board and all subjects provided 283 informed consent. A detailed description of protocols used for sample collection, processing, and analysis 284 is provided in SI Materials and Methods. All study sequences are available on the NCBI Sequence Read 285 Archive under the BioProject ID PRJNA389927. All associated source code is available at the following 286 GitHub repository: 287 https://github.com/SchlossLab/Hannigan_CRCVirome_PNAS_2017. 288 13 peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/152868 doi: bioRxiv preprint first posted online Jun. 20, 2017; Acknowledgments 289 The authors thank the Schloss lab members for their underlying contributions, and the Great Lakes-New 290 England Early Detection Research Network for providing the fecal samples that were used in this study. GD 291 Hannigan was supported in part by the Molecular Mechanisms in Microbial Pathogenesis Training Program 292 (T32 AI007528). PD Schloss was supported by funding from the National Institutes of Health (P30DK034933). 293 MT Ruffin was supported by funding from the National Institutes of Health (5U01CA86400). 294 14 peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/152868 doi: bioRxiv preprint first posted online Jun. 20, 2017;