Oral Microbes are a Signature of Disease in the Gut

doi:10.21203/rs.3.rs-1631596/v1

Download PDF

Analysis

Oral Microbes are a Signature of Disease in the Gut

https://doi.org/10.21203/rs.3.rs-1631596/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this older preprint version

Read the latest preprint version →

Case-control studies have identified human gut microbiome taxa associated with diseases including diabetes, inflammatory bowel, and colorectal cancer. However, many taxa are significantly associated with different diseases, confounding the utility of taxa as disease-specific diagnostics and potentially indicating underlying similarities in physiologically distinct diseases. Using a meta-analysis of 4446 gut metagenomic samples linked to 55 diseases across 21 datasets, we identify taxa consistently associated with health or disease and validate these results on external datasets. We find that taxa associated with disease in the gut are either oral microbiome commensals or are members of the polyphyletic class Clostridiales, whereas healthy taxa are gut specialists. Prevalence of organisms in the oral cavity correlates with gut disease association, revealing a novel, conserved relationship between disease and the gut microbiome. Our work highlights a major limitation of traditional metagenomics analysis methods for identifying disease-specific markers, a key goal of many microbiome studies.

As the cost of genomic sequencing has steadily dropped over the past decades, the scope of sequencing efforts has increased drastically 1–3. Despite terabytes of sequence data on the human microbiome, the community of microorganisms, and their ecosystems in the human body, there remains a gap in understanding how the microbiome, particularly the gut microbiome, interacts with our host physiology 4. Deeper investigation into these links has revealed specific mechanisms by which microorganisms interact with our immune systems and intestinal epithelia 5, 6. There is substantial interest in using the microbiome as a diagnostic tool by identifying taxonomic and functional signatures which correspond to different phenotypes 7.

Microbiome disease-signature identification is difficult because of the high-dimensionality of the system and the large number of exogenous variables which can affect a case-control study 8. Batch effects, such as demographic and geographic differences between study populations, are thought to account for a significant portion of the variation in the microbiome and hence the differences in findings between similar studies 9, 10. To address these issues, researchers have begun aggregating studies together to extrapolate common signals for a given phenotype. For example, Thomas et al. and Wirbel et al. performed different meta-analyses on a collection of colorectal cancer microbiome studies and found a similar set of associated bacteria 11, 12. However, many of the taxa associated with specific diseases in these studies were also identified as disease associated in studies of other diseases. For example, the oral commensal Fusobacterium nucleatum has been associated with colorectal cancer 13, periodontitis 14, inflammatory bowel disease 15, and additional conditions 16.

Here, we hypothesize that many of the organisms identified as significant in case-control microbiome studies are actually general markers of health and disease. To test this hypothesis, we performed a meta-analysis by aggregating 4446 gut metagenomic samples linked to 55 diseases across 21 case-control studies datasets and ask how specific each taxon is for a given disease (Fig. 1).

We identify conserved, significant associations between individual taxa across different diseases and conversely identify taxa that are reproducibly associated with health in our case-control studies. We call these taxa ‘general’ biomarkers of health and disease. We validate our health-associated taxa with an external set of studies of healthy microbiomes across the world. Through further investigation of these general disease-associated taxa, we discover a novel, conserved association between oral microbiome taxa found in the gut and disease. We further observe that taxa associated generally with health tend to be gut specialists; the only outliers in this respect are members of the polyphyletic class Clostridiales which are both general disease markers and prevalent in the gut.

By performing a novel meta-analysis on thousands of whole community meta genomes generated by deep sequencing, our study provides novel biological insights into the link between human physiology in health and disease in the gut microbiome. In doing so, we also develop a new analysis framework to disentangle general signatures of disease and health from specific signatures of a disease.

A Core Set of Microorganisms are Biomarkers of Health or Disease Across a Variety of Settings.

We downloaded gut microbiome whole community metagenomic samples from 20 studies which included healthy controls and patients from studies of multiple diseases, including colorectal cancer (CRC), type 2 diabetes (T2D), type 1 diabetes (T1D), metabolic syndromes (IGT), inflammatory bowel disease (IBD), Behcet’s disease (BD), hypertension (HD), liver cirrhosis, and others (Table 1). We hypothesized that many of the differentially abundant microorganisms within these studies are general markers of disease in general, or hallmarks of ’dysbiosis’. To test this hypothesis, we computed and averaged association scores in the form of Mann-Whitney AUC statistics for each taxon in each study. We then generated a core set of taxa whose average association across all studies was statistically significant at a threshold of p < 0.01 (Fig. 2) and an expanded set of markers at p < 0.05 (Figure S1).

We identified 261 taxa at a p < .05 (S1) and 41 taxa p < .01 (Fig. 2a). Many health associated bacteria included common commensals such as Lachnospiraceae 17 and Ruminococcaceae families 18, 19, while many known pathogen-associated organisms such as Fusobacterium nucleatum 16, Streptococaceae 20–22family, and Solobacterium moorei 23were identified as consistently disease-associated. In particular, many of these taxa align with other studies. For example, out of 29 taxa identified as colorectal cancer-associated by Wirbel et al., 19 directly matched with disease-associated markers in our extended setTable 2 12.

To demonstrate the consistency of these markers across studies, we performed a two-sided t-test comparing control and experimental groups for each taxon in each study and found that our most significant health markers were health-associated in almost every study independently (and similarly for disease markers) (Fig. 2b, Figure S1b). This congruence indicates that our results were not biased by a handful of outlier studies but rather that significant taxa in each of these studies overlap despite involving completely different patient cohorts. Next, we calculated the percent of samples that contained each taxon. We found that healthy markers tend to be present in most samples while some disease markers, such as Anaerostipes 24, are rarely carried and others, such as Clostridium hathewayi 25, are found in most samples (Fig. 2d, Figure S1d). Finally, we plotted the relative abundance of the general health and disease markers in microbiomes carrying these taxa. Here, we observe a wide distribution in relative abundance across most taxa (Fig. 2d, Figure S1d).

Predominance of Health-Associated Taxa in Microbiomes

We next asked if the observation that health-associated markers are more abundant in gut metagenomes generalizes to other studies. We analyzed 15 additional datasets containing only healthy-labeled samples from Western populations and found that, in all cases, healthy taxa have a higher mean relative abundance than disease taxa (Fig. 3a). This observation extends to non- Western populations. We analyzed six datasets from geographies such as Tanzania, Fiji Islands, and mountainous village regions in Latin America, and all of them followed the same trend with the exception of Smith et al. 26–31 (Fig. 3b). We also observed that health taxa are more abundant than disease markers even in datasets for which we only had unhealthy patient samples. Taken together, our results suggest that healthy taxa are more predominant in general, regardless of underlying condition 32–35 (Fig. 3c).

Oral Microbes are Predominantly Associated With Dysbiosis

We next assessed potential biological or functional similarities in our disease-associated taxa. From a review of the literature, we observed that many of our disease-associated taxa are commonly found in oral microbiomes 36. To investigate this relationship in our data, we first quantified the mean abundance of each organism in the oral cavity of healthy individuals using 206 oral microbiome samples from three studies 26, 37, 38. We found that, on average, disease taxa are more abundant in the oral cavity than healthy taxa, whereas healthy taxa are more abundant in the gut (Fig. 4a). No such pattern emerged when performing the same analysis on skin and nasal body sites 39–42.

When we compare disease-associated markers with health-associated markers in gut studies, we find that health-associated markers have a higher mean relative abundance in the gut than disease-associated markers (Fig. 4b). We found a tendency for more abundant gut taxa to be associated with health regardless of whether the mean AUC was statistically significant (slope = 31.7, p = 1.26e-8). Most health-associated taxa are close to 0.5 in mean AUC until approximately 10^-1 mean abundance, at which point a fanning out occurs, where organisms above this threshold tend to be either strongly health- or disease-associated, although the trend leans towards higher-abundance organisms being more health-associated. (Fig. 4c)

By contrast, when we compare the mean relative abundance of bacteria in the oral cavity with their disease vs. health association, we find that the more abundant bacteria are in the oral cavity, the more likely they are to be disease-associated (Fig. 4c). We found that, of the 160 taxa that are found both in the oral cavity and are in the core gut group identified by our meta-analysis, 134 are associated with disease and only 26 with health. Furthermore, when we plotted the mean AUC score in gut studies vs. the log mean abundance of each oral taxon, we found that taxa with higher abundances in the oral cavity tend to be more disease-associated (regression slope = -50.4, p = 6.84e-22). This relationship also held true if we extended to all oral taxa; higher mean abundance was associated with lower mean AUC, even if that AUC score is not statistically significant.

Almost All Disease Markers are Oral Taxa or Class Clostridia

When analyzing taxonomic abundance data, we noticed that many taxa were completely absent from most samples, but when they were present, they would follow a consistent distribution (Fig. 2c). Hence, the mean relative abundance of a taxa may be biased downward for taxa that are infrequently carried. To better understand this phenomenon, we visualized our data based not on mean relative abundance but on frequency of carriage. We compared the percentage of samples those different markers are found in across different body sites and discovered a striking pattern: The majority of disease taxa are rarely found in the gut, but rather are common oral commensals (Fig. 5a, Figure S2). Furthermore, we found that all non-oral commensal disease markers were either from the class Clostridia or belonged to a higher taxonomic order such as

genus Peptostreptococcus (Fig. 5b-d).

We next grouped our taxa based on their frequency of carriage and disease score. As Fig. 5c shows, there are no core taxa that are infrequently (< 20% of samples) carried in the gut and also health associated. These observations hold true in the extended set of health and disease markers with the exception of two broad categories (family Bacillaceae and Genus Siphoviridae, a phage taxa) and Clostridium sp-L2-50 (Figure S3). Additionally, there are very few core markers commonly carried in the gut and disease associated. Taxa that fit into this category in the core belong to Clostridia, and in the extended group are mostly higher order categories (for example Class Gammaproteobacteria and Order Enterobacteriales). By contrast, there are no core health associated taxa commonly found in oral microbiomes (Fig. 5d). This also applies to the extended set again with the exception of a few higher order category taxa (Figure S4). The remaining disease-associated taxa are infrequently found in any body site (Fig. 5b). These are relatively rare organisms like Holdemania sp AP2 (2.37% carriage) and Atopobium sp ICM58 (.08% carriage), while the rarest health-associated taxa is Bacteroides intestinalis (27.08% carriage) (Fig. 2d). There are a few exceptions to this pattern in the extended set, such as the disease marker Lachnospiraceae bacterium 3-1-57FAA-CT1 (29.68% carriage) (Figs. 6 and 10). All other core healthy markers are carried by more than half of stool samples and only Clostridia disease markers are ever found at that frequency in the gut. These findings serve as a validation, reinforcing the consistency of the identified taxa as truly general markers of health and disease.

Our work demonstrates that microbiome biomarkers identified in case-control studies may not be purely specific to the phenotypes under investigation, which changes the interpretation of these studies’ findings. Our results thus raise an important question for microbiome clinical research: is a taxon identified in a case-control study associated with a specific disease or with general stress pathways that are common across many diseases? This question adds to the list of challenges associated with such studies which includes batch effects 43, sequence annotation errors 44, and sampling strategy 45.

Specificity is key to designing robust diagnostic tools, developing interventional therapies, and understanding mechanistic links between microbes and human pathophysiology. For example, an assay designed to diagnose CRC using bacterial readouts can only be effective if it has high sensitivity and is specific enough to not raise false-positives when performed on samples from patients with a different disease, such as IBD 46, 47. Ideally, such an assay should be compared against patients with conditions that could have similar signatures to IBD to truly validate its effectiveness⁴⁸. Historically, microbiome case-control studies are set up to compare a healthy control group against an experimental cohort with some phenotype of interest. Here we show that without also comparing the experimental cohort against ‘similar’ groups (eg. other disease groups), one cannot ascertain specificity. This limitation curtails the effectiveness of follow-up analyses which attempt to elucidate the mechanisms of specific genes, pathways, or taxa identified by a case-control study in the disease phenotype. Our approach identifies general microbiome/disease associations that are robust across multiple disease studies rather than specific to the pathophysiology of an individual disease. From prior studies, mechanisms of microbiome/disease interactions could include short chain fatty acid metabolism, regulatory T cell activity, cytokines interactions, and overall response to bacterial antigens 49. Hence, there is an urgent, unmet need to identify and differentiate general- vs specific- taxa, genes, and pathways associated with disease 50.

Our findings also identified a striking pattern in that almost all disease markers are either common oral taxa or belong to the class Clostridia. This could suggest that, in dysbiosis, bacteria which do not normally inhabit the gut may be able to migrate in from the oral cavity 13, 51. Indeed, there is evidence that oral infections such as gingivitis can influence gut disorders such as inflammatory bowel disease 52. One may ask why this does not happen normally; bacteria entering the stomach should usually be destroyed during digestion due to the low pH 53. However, there is evidence that there is always some level of transmission of bacteria between oral to gut environments 54. A disease process could increase this transmission rate, for example by increasing transit time of food and making it easier for oral bacteria to enter the gut 55, 56. We note that the reproducibility of our findings across studies suggests that these taxa are not simply transiting, in which case we would not expect to see the same oral taxa repeatedly, but that they are also taking up residence in the gut. Hence, increased intestinal stress may create an environment that is more conducive to invasion by oral taxa 57. In the ecology literature, there is also evidence that ecosystem invasibility can increase in situations where turnover rises or the overall system becomes less stable 58. Regardless, such an interpretation implies that these markers are not causal but rather are a consequence of intestinal stress.

The Clostridia disease markers are an exception to the oral-gut axis observation. Bacteria from class Clostridia are rarely found in the oral cavity and are in some cases gut commensals. These microbes are difficult to classify; the class is polyphyletic and many taxa have been renamed since this dataset was first downloaded in March 2021 59. However, many Clostridia are spore- forming and, building on the invasibility hypothesis, environmental stress may make it easier for their spores to grow 60–62. Hence, sporulation and altered transit time may represent two parallel routes of entry for microbes into a perturbed gut environment that is more vulnerable to invasion.

Our study is limited by the number of datasets available for each disease. Hence, we focused on identifying and understanding the general markers. We also note that our observed effect sizes for individual microbes are very small. This, along with the acknowledgement that our results may only be associative rather than causal, suggests that individual markers are poor candidates for probiotics meant to foster a healthy microbiome or for diagnostic assays 63, 64. Instead, the combination of many markers should be utilized in order to account for diversity between individuals and we posit that functional data, in combination with taxonomic data, will also assist in separating true signal from noise.

As interest in the microbiome as a research target grows, so too have concerns emerged about the reproducibility and interpretability of study findings 65. Every step in a study has various risks, from the sample collection protocols to the nucleotide extraction protocols, to the sequencing technologies, primer kits, and base-calling software, to the tools used to annotate metagenomes, and even the statistical analyses performed at the end. Work from other labs has highlighted risks due to experimental batch effects and possible solutions for controlling for these issues 9, 10, but to our knowledge, there has been very little discourse about controlling for the specificity of results 66. Our work in this paper highlights both the dangers of ignoring this specificity and the observation that many diseases linked to the microbiome display common signatures. There is a wealth of data now available that can be used to indirectly control for these issues; we expect that future case-control studies will not only take steps to control for experimental noise, but also to compare their results against other (unrelated) studies to quantify which findings are specific and which are nonspecific to their investigation.

All figures were generated originally using scripts written in Python and assembled using Adobe Illustrator.

Data Curation

The data sets for this study were sourced from ‘CuratedMetagenomicData‘ (downloaded Mar 2021) 59, an R package providing uniform metagenomic annotations for all studies included via ‘MetaPhlAn2‘ 67 and ‘HUManN2‘ 68 along with standardized metadata.

To prepare this data for analysis, we wrote two Python packages (available on PyPi). The first, ‘curated_metagenomic_data_extractor‘, downloads all available annotations and metadata. The metadata is re-uploaded into a (PostgreSQL) database, and the annotations files are uploaded

to a Google Cloud Storage bucket. The purpose of this is to make it easier to query samples and studies based on specific criteria (eg. get all samples for a body site, with a specific disease label, etc.).

The second package, called ‘curated_metagenomic_data_loader‘, creates datasets for statistical analysis. Here, we write SQL queries to select samples based on their metadata, and the package downloads the annotations for those samples from the storage bucket and consolidates all of this information into a TensorFrame, which is a data structure we created for easily working with this type of data 69 .

Dataset Selection

We chose datasets which were available in ‘curatedMetagenomicData’ at the time of analysis which contained stool microbiome samples and involved a healthy control group and an unhealthy (labeled disease phenotype) group. These are described in Table 1. Not all studies were classic case-control studies; for example, Vatanen et al. was a longitudinal study following newborn infants across different Eastern European countries 70. However, every study used had samples labeled with a disease phenotype or as control/healthy.

Meta-Analysis Technique

At a high level, we identified putative health and disease markers by assigning an association score to each taxon for each dataset. We then looked for taxa that had consistently high (healthy) or low (disease) association scores across all studies.

For each study and taxa, we performed a Mann-Whitney U test comparing the distribution of relative abundances of that taxa between experimental and control groups 71. This is a non-parametric test of the null-hypothesis that a random sample chosen from the experimental group is no likely to be greater or lesser than a random sample chosen from the control group. The U- statistic from this test can be converted into an estimate of the probability that a random sample from the control group is greater than a random sample from the experimental group, and this is called the AUC (area under the curve) 72. The reason for this nomenclature is that in receiver operator characteristic (ROC) analysis for binary classification, the area under a receiver-operator characteristic curve (commonly denoted as AUC) is equivalent to the probability that a classifier assigns a greater score to a positive label than a negative label 73. Hence, the U-statistic AUC can be interpreted as a binary classification: if the AUC score is greater than 0.5, the taxa is associated with healthy, and if the score is less than 0.5, it is associated with disease, and 0.5 means no association. This association is quantitative.

To assign a final score to a taxon, we computed its mean AUC-score across all studies. We then performed a (two sided) T-test of the null hypothesis that this mean score is 0.5. We used a cutoff of p < 0.01 to generate the core set of taxa that are shown in Fig. 1, and a cutoff of p <0.05 to generate an expanded set of markers.

Deduplication

For the purpose of simplifying visualization, we consolidated taxa that were directly related and had the same association scores. This occurs when the taxonomic annotations for a sample give the same relative abundance to a taxon and its parent. For example, s Clostridium_symbiosum and s Clostridium_symbiosum|t Clostridium_symbiosum_unclassified had the same score (0.418). In such instances, we show only the parent taxon in figures that mention marker names. This was done for Fig. 2, Fig. 5, Figure S1, and all supplemental figures. Figures 3 and 4 utilized the expanded marker list with no deduplication because including duplicate markers does not change the visualization or interpretation of the figure as duplicate markers sit on top of each other.

Comparison to Other Meta-Analyses

Our approach can be thought of as a more permissive form of the method used by LefSe algorithm to detect significant markers across studies 74. Briefly, LefSe considers a marker to be significant by requiring that the marker be statistically significantly different between each combination of experimental and control groups across each study being analyzed. LefSe thus performs a Mann-Whitney U test for each pair of experiment / control groups and demands that every one of them be significant. While this approach is effective for small numbers of datasets, the number of tests performed this way (n*(n-1)/2) results in a very strict criterion when considering many studies. To loosen this requirement, we consider those features that are statistically significant in most of the studies considered and on average. To quantify features that are statistically significant in most studies, we compute a Mann-Whitney U test for each study and convert the resulting statistic to an AUC score, where a value greater than 0.5 indicates a positive association, less than 0.5 indicates a negative association, and the closer to 0.5, the weaker the association. To accept a feature, we then require that the mean AUC across all studies for that feature be significantly different from 0.5 (using a two-sided T-test, p < 0.01 was used to define ‘core’ healthy and unhealthy marker taxa and p < 0.05 was used to define an expanded set of marker taxa).

Our approach thus essentially requires that a feature be consistently (but not always) associated with the same side of a study (either positively enriched for the healthy cohort or for the experimental group). While less strict, this method is more flexible, and reflective of the noise in microbiome studies, enabling us to incorporate more studies and account for natural variation that may cause a feature to not always be differentially abundant between study groups.

Body Site Comparisons

To generate the results in Figs. 3a, 3b, and 4, we estimated how often each taxon was present in different body sites. To do this, we utilized all of the available datasets in curatedMetagenomicData. For each body site, we took all available samples from that body site and computed the mean relative abundance across all such samples to estimate the mean frequency of a taxon in a site. Organisms with mean relative abundance greater than zero were included. This approach enabled us to statistically characterize oral-associated taxa in our dataset.

Funding

SK was supported by a National Institutes of Health (NIH) training grant 2T32GM007288-45 (Medical Scientist Training Program) at Albert Einstein College of Medicine and an NIH T32 fellowship on Geographic Medicine and Emerging Infectious Diseases (2T32AI070117-13). LK is supported in part by NIH R01HL069438, a Department of Defense Peer Reviewed Cancer Research Career Development Award (CA171019), the Elsa U Pardee Foundation, and an Irma T. Hirschl Career Scientist award. Computing support for the project was funded by the Google Cloud Research Credits program (GCP19980904) and an Nvidia GPU research grant.

Authors’ contributions

SK and LK conceived of the study. SK performed all the of the data processing, statistical analysis, and figure generation. SK and LK co-wrote the text of the paper.

Competing interests

The authors have no competing interests to disclose.

Gu, W., Miller, S. & Chiu, C. Y. Clinical Metagenomic Next-Generation Sequencing for Pathogen Detection. Annu Rev Pathol 14, 319–338 (2019).
Malla, M. A. et al. Exploring the Human Microbiome: The Potential Future Role of NextGeneration Sequencing in Disease Diagnosis and Treatment. Front Immunol 9, 2868 (2018).
Levy, S. E. & Boone, B. E. Next-Generation Sequencing Strategies. Cold Spring Harb Perspect Med 9 (2019).
Hatzenpichler, R., Krukenberg, V., Spietz, R. L. & Jay, Z. J. Next-generation physiology approaches to study microbiome function at single cell level. Nat Rev Microbiol 18, 241–256 (2020).
Lambring, C. B. et al. Impact of the Microbiome on the Immune System. Crit Rev Immunol 39, 313–328 (2019).
Shi, N., Li, N., Duan, X. & Niu, H. Interaction between the gut microbiome and mucosal immune system. Mil Med Res 4, 14 (2017).
Schlaberg, R. Microbiome Diagnostics. Clin Chem 66, 68–76 (2020).
Tsilimigras, M. C. & Fodor, A. A. Compositional data analysis of the microbiome: fundamentals, tools, and challenges. Ann Epidemiol 26, 330–5 (2016).
Wang, Y. & LêCao, K. A. Managing batch effects in microbiome data. Brief Bioinform 21, 1954–1970 (2020).
Gibbons, S. M., Duvallet, C. & Alm, E. J. Correcting for batch effects in case-control microbiome studies. PLoS Comput Biol 14, e1006102 (2018).
Thomas, A. M. et al. Metagenomic analysis of colorectal cancer datasets identifies crosscohort microbial diagnostic signatures and a link with choline degradation. Nat Med 25, 667–678 (2019).
Wirbel, J. et al. Meta-analysis of fecal metagenomes reveals global microbial signatures that are specific for colorectal cancer. Nat Med 25, 679–689 (2019).
Park, S. Y. et al. Oral-Gut Microbiome Axis in Gastrointestinal Disease and Cancer. Cancers (Basel) 13 (2021).
More Than Just a Periodontal Pathogen –the Research Progress on Fusobacterium nucleatum. Frontiers in Cellular and Infection Microbiology 12 (2022). URL 10.3389/fcimb.2022. 815318.
Huh, J. W. & Roh, T. Y. Opportunistic detection of Fusobacterium nucleatum as a marker for the early gut microbial dysbiosis. BMC Microbiol 20, 208–208 (2020).
Han, Y. W. Fusobacterium nucleatum: a commensal-turned pathogen. Curr Opin Microbiol 23, 141–147 (2015).
Vacca, M. et al. The Controversial Role of Human Gut Lachnospiraceae. Microorganisms 8, 573–573 (2020).
Lee, G., You, H. J. & Bajaj, J. S. Distinct signatures of gut microbiome and metabolites associated with significant fibrosis in non-obese NAFLD. Nat Commun 11, 4982–4982 (2020).
Gu, X. et al. Gut Ruminococcaceae levels at baseline correlate with risk of antibiotic- associated diarrhea. (2021).
Yu, D. et al. Fecal Streptococcus Alteration Is Associated with Gastric Cancer Occurrence and Liver Metastasis. mBio 12, e02994–21 (2021). URL 10.1128/mBio.02994-21.
Vitetta, L., Llewellyn, H. & Oldfield, D. (2019).
Garcia-Mantrana, I., Selma-Royo, M., Alcantara, C. & Collado, M. C. (2018).
Barrak, I., Stájer, A., Gajdács, M. & Urbán, E. Small, but smelly: the importance of Solobacterium moorei in halitosis and other human infections. Heliyon 6 (2020).
Identifying Gut Microbiota Associated With Colorectal Cancer Using a Zero-Inflated Lognormal Model. Frontiers in Microbiology 10 (2019). URL 10.3389/fmicb.2019.00826.
Hernández-Juárez, L. E. et al. Analyses of publicly available Hungatella hathewayi genomes revealed genetic distances indicating they belong to more than one species. Virulence 12 (1950).
Brito, I. L. et al. Mobile genes in the human microbiome are structured from global to individual scales. Nature 535, 435–439 (2016).
Dhakan, D. B. et al. The unique composition of Indian gut microbiome, gene catalogue, and associated fecal metabolome deciphered using multi-omics approaches. Gigascience 8 (2019).
Liu, W. et al. Unique Features of Ethnic Mongolian Gut Microbiome revealed by metagenomic analysis. Sci Rep 6, 34826 (2016).
Pehrsson, E. C. et al. Interconnected microbiomes and resistomes in low-income human habitats. Nature 533, 212–6 (2016).
Rampelli, S. et al. Metagenome Sequencing of the Hadza Hunter-Gatherer Gut Microbiota. Curr Biol 25, 1682–93 (2015).
Smits, S. A. et al. Seasonal cycling in the gut microbiome of the Hadza hunter-gatherers of Tanzania. Science 357, 802–806 (2017).
Wen, C. et al. Quantitative metagenomics reveals unique gut microbiome biomarkers in ankylosing spondylitis. Genome Biol 18, 142 (2017).
Gopalakrishnan, V., Helmink, B. A., Spencer, C. N., Reuben, A. & Wargo, J. A. The Influence of the Gut Microbiome on Cancer, Immunity, and Cancer Immunotherapy. Cancer Cell 33, 570–580 (2018).
Loomba, R. et al. Gut Microbiome-Based Metagenomic Signature for Non-invasive Detection of Advanced Fibrosis in Human Nonalcoholic Fatty Liver Disease. Cell Metab 25 (2017).
Matson, V. et al. The commensal microbiome is associated with anti-PD-1 efficacy in metastatic melanoma patients. Science 359, 104–108 (2018).
Willis, J. R. & Gabaldón, T. The Human Oral Microbiome in Health and Disease: From Sequences to. Ecosystems. Microorganisms 8, 308–308 (2020).
Ferretti, P. et al. Mother-to-Infant Microbial Transmission from Different Body Sites Shapes the Developing Infant Gut Microbiome. Cell Host Microbe 24 (2018).
Olm, M. R. et al. Identical bacterial populations colonize premature infant gut, skin, and oral microbiomes and exhibit different in situ growth rates. Genome Res 27, 601–612 (2017).
Tett, A. et al. Unexplored diversity and strain-level structure of the skin microbiome associated with psoriasis. NPJ Biofilms Microbiomes 3, 14 (2017).
Chng, K., Tay, A. & Li, C. Whole metagenome profiling reveals skin microbiome-dependent susceptibility to atopic dermatitis flare. Nat Microbiol 1, 16106–16106 (2016).
Oh, J. et al. Biogeography and individuality shape function in the human skin metagenome. Nature 514, 59–64 (2014).
Structure, function and diversity of the healthy human microbiome. Nature 486, 207–14 (2012).
Tierney, B. T. et al. Systematically assessing microbiome-disease associations identifies drivers of inconsistency in metagenomic research. PLoS Biol 20, e3001556 (2022).
Sun, Z., Huang, S. & Zhang, M. Challenges in benchmarking metagenomic profilers. Nat Methods 18, 618–626 (2021).
Current Sampling Methods for Gut Microbiota: A Call for More Precise Devices. Frontiers in Cellular and Infection Microbiology 10 (2020). URL 10.3389/fcimb.2020.00151.
Vogtmann, E. et al. Colorectal Cancer and the Human Gut Microbiome: Reproducibility with Whole-Genome Shotgun Sequencing. PLoS One 11, e0155362 (2016).
Zeller, G. et al. Potential of fecal microbiota for early-stage detection of colorectal cancer. Mol Syst Biol 10, 766 (2014).
Olesen, S., Alm, E. Dysbiosis is not an answer. Nat Microbiol 1, 16228 (2016). https://doi.org/10.1038/nmicrobiol.2016.228
Zheng, D., Liwinski, T. & Elinav, E. Interaction between microbiota and immunity in health and disease. Cell Res 30, 492–506 (2020).
Minot, S. S. & Willis, A. D. Clustering co-abundant genes identifies components of the gut microbiome that are reproducibly associated with colorectal cancer and inflammatory bowel disease. Microbiome 7, 110 (2019).
Olsen, I. & Yamazaki, K. Can oral bacteria affect the microbiome of the gut? J Oral Microbiol 11, 1586422 (2019).
Byrd, K. M. & Gulati, A. S. The Gum-Gut Axis in Inflammatory Bowel Diseases: A Hypothesis-Driven Review of Associations and Advances. Front Immunol 12, 620124 (2021).
Tennant, S. M. et al. Influence of gastric acid on susceptibility to infection with ingested bacterial pathogens. Infect Immun 76, 639–45 (2008).
Schmidt, T. S. et al. Extensive transmission of microbes along the gastrointestinal tract. {eLife} 8 (2019). URL 10.7554/elife.42693.
Kojima, M. et al. Bowel movement frequency and risk of colorectal cancer in a large cohort study of Japanese men and women. Br J Cancer 90, 1397–401 (2004).
Haase, A. M. et al. Regional gastrointestinal transit times in severe ulcerative colitis. Neurogastroenterol Motil 28, 217–24 (2016).
Albright, M. B. N. et al. Solutions in microbiome engineering: prioritizing barriers to organism establishment. ISME J 16, 331–338 (2022).
Hui, C., Richardson, D. M. & Landi, P. Defining invasiveness and invasibility in ecological networks. Biol Invasions 18, 971–983 (2016).
Pasolli, E., Schiffer, L. & Manghi, P. Accessible, curated metagenomic data through ExperimentHub. Nat Methods 14, 1023–1024 (2017).
Clostridia: Sporeforming Anaerobic Bacilli - Medical Microbiology - NCBI Bookshelf.
Al-Hinai, M. A., Jones, S. W. & Papoutsakis, E. T. The Clostridium sporulation programs: diversity and preservation of endospore differentiation. Microbiol Mol Biol Rev 79, 19–37 (2015).
Mallozzi, M., Viswanathan, V. K. & Vedantam, G. Spore-forming Bacilli and Clostridia in human disease. Future Microbiol 5, 1109–23 (2010).
Cunningham, M. et al. Shaping the Future of Probiotics and Prebiotics. Trends in Microbiology 29, 667–685 (2021). URL https://doi.org/10.1016/j.tim.2021.01.003.
Damhorst, M. W. G. L., Adelman, Michael, H., Woodworth, C. S. & Kraft. Current Capabilities of Gut Microbiome-Based Diagnostics and the Promise of Clinical Application. Pages S270-S275 223 (2021).
Schloss, P. D. Identifying and Overcoming Threats to Reproducibility, Replicability, Robustness, and Generalizability in Microbiome Research. (2018).
Wirbel, J., Zych, K. & Essex, M. Microbiome meta-analysis and cross-disease comparison enabled by the SIAMCAT machine learning toolbox. Genome Biol 22, 93–93 (2021).
Truong, D., Franzosa, E. & Tickle, T. MetaPhlAn2 for enhanced metagenomic taxonomic profiling. Nat Methods 12, 902–903 (2015).
Franzosa, E. A. et al. Species-level functional profiling of metagenomes and metatranscriptomes. Nature methods 15, 962–968 (2018).
S, K. & L, K. Fireworks: Reproducible Machine Learning and Preprocessing with PyTorch. The Journal of Open Source Software (2019).
Vatanen, T. et al. Variation in Microbiome LPS Immunogenicity Contributes to Autoimmunity in Humans. Cell 165, 842–53 (2016).
Mann, H. B. & Whitney, D. R. On a Test of Whether one of Two Random Variables is Stochastically Larger than the Other. The Annals of Mathematical Statistics 18, 50–60 (1947).
Horton, B. AUC Meets the Wilcoxon-Mann-Whitney U-Statistic (2017). URL https://blog. revolutionanalytics.com/2017/03/auc-meets-u-stat.html.
Drury, M. The Probabilistic Interpretation of AUC (2017). URL http://madrury.github.io/ jekyll/update/statistics/2017/06/21/auc-proof.html.
Segata, N., Izard, J. & Waldron, L. Metagenomic biomarker discovery and explanation. Genome Biol 12, 60–60 (2011).
David, L. A. et al. Diet rapidly and reproducibly alters the human gut microbiome. Nature 505, 559–63 (2014).
Feng, Q. et al. Gut microbiome development along the colorectal adenoma-carcinoma sequence. Nat Commun 6, 6528 (2015).
Hannigan, G. D., Duhaime, M. B., 4th Ruffin, M., Koumpouras, C. C. & Schloss, P. D. Diagnostic Potential and Interactive Dynamics of the Colorectal Cancer Virome. mBio 9 (2018).
Jie, Z. et al. The gut microbiome in atherosclerotic cardiovascular disease. Nat Commun 8, 845 (2017).
Karlsson, F. H. et al. Gut metagenome in European women with normal, impaired and diabetic glucose control. Nature 498, 99–103 (2013).
Kieser, S. et al. Bangladeshi children with acute diarrhoea show faecal microbiomes with increased Streptococcus abundance, irrespective of diarrhoea aetiology. Environ Microbiol 20, 2256–2269 (2018).
Kostic, A. D. et al. The dynamics of the human infant gut microbiome in development and in progression toward type 1 diabetes. Cell Host Microbe 17, 260–73 (2015).
Li, J. et al. Gut microbiota dysbiosis contributes to the development of hypertension. Microbiome 5, 14 (2017).
Li, S. S. et al. Durable coexistence of donor and recipient strains after fecal microbiota transplantation. Science 352, 586–9 (2016).
Nielsen, H. B. et al. Identification and assembly of genomes and genetic elements in complex metagenomic samples without using reference genomes. Nat Biotechnol 32, 822–8 (2014).
Qin, J. et al. A metagenome-wide association study of gut microbiota in type 2 diabetes. Nature 490, 55–60 (2012).
Qin, N. et al. Alterations of the human gut microbiome in liver cirrhosis. Nature 513, 59–64 (2014).
Vincent, C. et al. Bloom and bust: intestinal microbiota dynamics in response to hospital exposures and Clostridium difficile colonization or infection. Microbiome 4, 12 (2016).
Ye, Z. et al. A metagenomic study of the gut microbiome in Behcet’s disease. Microbiome 6, 135 (2018).
Yu, J. et al. Metagenomic analysis of faecal microbiome as a tool towards targeted non- invasive biomarkers for colorectal cancer. Gut 66, 70–78 (2017).
Milanese, A., Mende, D. R. & Paoli, L. Microbial abundance, activity and population genomic profiling with mOTUs2. Nat Commun 10, 1014–1014 (2019).

Figures 6 and 10 are not available with this version.

Tables 1-2 are available in the Supplementary Files section.

There is NO Competing Interest.

table1.pdf
Table 1
table2.pdf
Table 2
figures1v1.pdf
Figure s1
figures2v1.pdf
Figure s2
figures3v1.pdf
Figure s3
figures4v1.pdf
Figure s4
figures5v1.pdf
Figure s5

Download PDF

Version 1

posted

You are reading this older preprint version

Read the latest preprint version →

Oral Microbes are a Signature of Disease in the Gut

Status:

Version 1

Abstract

Figures

1 Introduction

2 Results

3 Discussion

4 Methods

Declarations

References

Figures 6 and 10

Tables 1-2

Additional Declarations

Supplementary Files

Status:

Version 1