Oral Health and the Altered Colonic Mucosa-associated Gut Microbiota

Abstract

Background

Systemic diseases have been associated with oral health and gut microbiota. We examined the association between oral health and the community composition and structure of the adherent colonic gut microbiota.

Methods

We obtained 197 snap-frozen colonic biopsies from 62 colonoscopy-confirmed polyp-free individuals. Microbial DNA was sequenced for the 16S rRNA V4 region using the Illumina MiSeq, and the sequences were assigned to the operational taxonomic unit based on SILVA. We used a questionnaire to ascertain tooth loss, gum disease, and lifestyle factors. We compared biodiversity and relative abundance of bacterial taxa based on the amount of tooth loss and the presence of gum disease. The multivariable negative binomial regression model for panel data was used to estimate the association between the bacterial count and oral health. False discovery rate-adjusted P value (q value) < .05 indicated statistical significance.

Results

More tooth loss and gum disease were associated with lower bacterial alpha diversity. The relative abundance of Faecalibacterium was lower (q values < .05) with more tooth loss. The association was significant after adjusting for age, ethnicity, obesity, smoking, alcohol use, hypertension, diabetes, and the colon segment. The relative abundance of Bacteroides was higher in those with gum disease.

Conclusions

Oral health was associated with alteration in the community composition and structure of the adherent gut bacteria in the colon. The reduced anti-inflammatory Faecalibacterium in participants with more tooth loss may indicate systemic inflammation. Future studies are warranted to confirm our findings and investigate the systemic role of Faecalibacterium.

Introduction

Poor oral health and periodontal inflammation have been associated with a number of chronic systemic diseases such as diabetes, cardiovascular disease, pulmonary disease, and colorectal/pancreatic cancers in observational epidemiologic research [1–4]. The underlying mechanisms by which poor oral health influences health outcomes are incompletely understood. Some suggest that periodontal bacteria can migrate to the distant organ to ignite inflammatory responses. For example, Streptococcus mutans has been found in human diseased cardiac valves [5]. Periodontal bacteria such as Porphyromonas gingivalis, Fusobacterium nucleatum, and Streptococcus gordonii have been shown to mediate inflammatory pathways. One interventional study showed these bacteria were associated with increased circulating dendritic cells and conversion of regulatory T-cells to Th17 cells [6]. Some suggest that poor oral health is an indicator of chronic systemic inflammation, which is common in many systemic diseases. Periodontitis has been associated with increased circulating interleukin-6 and C-reactive protein levels [7]. Treatment of periodontitis has been shown to decrease systemic inflammatory markers, reduce cholesterol levels, improve glycemic control in diabetics, and improve cognitive status in patients with Alzheimer’s disease [7–9]. There is overwhelming evidence that periodontitis, via the mediation of inflammatory pathways, is linked with systemic inflammation and diseases.

Accumulating research has linked gut microbiota changes, also known as dysbiosis, to multiple systemic diseases [10]. Gut bacteria can release endotoxins into the circulation where it induces the production of pro-inflammatory cytokines by enterocytes through toll-like receptor 4 [11, 12]. On the other hand, the gut bacteria such as Faecalibacterium prausnitzii is known to have a protective effect on systemic inflammation by increasing anti-inflammatory cytokines [13]. The depletion of anti-inflammatory bacteria and the overgrowth of opportunistic pathogenic bacteria can thus exacerbate systemic inflammation. However, the association between gut microbiota and oral health has not been well evaluated.

We examined the association between the community composition and structure of the colonic adherent microbiota and oral health using the amount of tooth loss and gum disease as the markers for oral health in a human population. We hypothesized that there would be a positive association between poor oral health and the relative abundance of pro-inflammatory microbiota or an inverse association between poor oral health and the relative abundance of anti-inflammatory microbiota in the colonic mucosa.

Materials and Methods

Study participants and data collection

Participants were prospectively and consecutively recruited from veterans undergoing colonoscopy at the endoscopy suite of the Michael E. DeBakey VA Medical center (MEDVAMC) in Houston between July 2013 and April 2017. The methods, including extensive eligibility criteria, were described previously [14].

We assessed the lifestyle and oral health of participants using an interviewer-administered questionnaire. We queried the number of adult teeth lost or removed due to gum disease or tooth decay (none, 1-5, 6 or more but not all, and all of the teeth), the use of full dentures, or history of implants or bridge, and whether participants had any history of gum disease (diagnosed gingivitis or periodontitis). If the participant answered “yes” to the last question, we defined this participant as “ever having gum disease”. In addition, a subset of 40 participants reported food intake in the past 12 months using the validated 2005 BLOCK food frequency questionnaire (FFQ). The average daily intake of food and nutrient was computed and energy-adjusted using the density method [15]. The 2005 healthy eating index (HEI) score, a measure of dietary quality specified by federal dietary guidance in 2005, was calculated [16].

Colonoscopy and Biopsy

The biopsy samples were acquired from the participants and stored using the same protocol as described previously [14]. Participants were advised to stop taking aspirin, anti-inflammatory drugs, blood thinners, iron, or vitamins with iron seven days before the procedure and stop diabetic medication one day before the procedure. Adequate endoscopic visualization was ensured. The endoscopists obtained biopsies from each colonic segment (cecum, ascending, transverse, descending, sigmoid colon, or rectum) when possible.

We enrolled 612 eligible study participants in the study. Of 174 study participants who were confirmed polyp-free, 134 consented to provide colonic mucosal biopsies, 69 were sent for microbiota profiling, and 63 responded to oral health questions. We, therefore, included 198 mucosa samples from 63 study participants in the present study. The FFQ was given to 46 participants, and 40 mailed back the completed FFQ.

All study participants provided informed consent form. The study protocol was approved by the Institutional Review Boards of Baylor College of Medicine (BCM) and MEDVAMC. The procedures followed were in accordance with the World Medical Association’s Declaration of Helsinki (1964, and its later amendments).

DNA extraction, Library construction, and 16S rRNA sequencing

The sequence analysis was performed at the Alkek Center for Metagenomics and Microbiome Research (CMMR) at BCM. Bacterial genomic DNA was extracted from the biopsies using the MO BIO PowerLyzer UltraClean Tissue & Cell DNA Isolation Kits (MO BIO Laboratories, Claridad, CA, a.k.a. Qiagen). All DNA samples were stored at −80°C until further analysis.

The 16S rRNA gene sequencing method was adapted from published methods [17–19]. The 16S rRNA variable region 4 (V4) was amplified by PCR using the barcoded Illumina adaptor-containing primers 515F and 806R and sequenced on the MiSeq platform (Illumina, San Diego, CA). The primers used for amplification contain adapters for the MiSeq sequencing and single-index barcodes so that the PCR products could be pooled and sequenced directly. The 2x250 bp paired-end protocol yields pair-end reads that overlap almost entirely [17].

Bioinformatics and Taxonomic Assignment

We used the CMMR bioinformatics pipeline for data analysis. The reads were merged using USEARCH v7.0.1090 [20]. A quality filter was applied to the merged reads, and those containing above 0.5% expected errors were discarded. The 16S rRNA gene sequences were clustered into the Operational Taxonomic Units (OTUs) at a similarity cutoff value of 97% using the UPARSE algorithm [21]. The OTUs were mapped to the SILVA (v128) to determine the taxonomies [22].

A rarefied OTU table was used for downstream analyses of alpha and beta diversity and phylogenic trends using the Agile Toolkit for Incisive Microbial Analyses (ATIMA) [23]. After using 1647 reads per sample as the rarefaction factor, two mucosal samples with poor sequencing results were eliminated. We were left with 197 mucosal samples from 62 participants for the final analysis.

Statistical Analysis

The bacterial alpha-diversity, beta-diversity, and the taxa abundance (mainly at the phylum and the genus level) were assessed based on tooth loss and gum disease. For the amount of tooth loss, participants were categorized into: none (n = 9), 1-5 (n = 27), 6 or more (n =19), and all teeth lost (n = 7). Participants were categorized into having gum disease (n = 13) or absence of gum diseases (n = 49). The study participants’ sociodemographic and clinical characteristics were compared based on tooth loss status using ANOVA for a continuous variable or Fisher’s exact test for a categorical variable. We performed the post hoc non-parametric trend test if the ANOVA test showed a significant difference in the distribution of the exposure variable across the amount of tooth loss. Alpha diversity of the bacteria was measured by observed OTU and the Shannon index and compared using the Kruskal-Wallis test based on the amount of tooth loss and the Mann-Whitney test based on gum disease. The observed OTU measures the number of different OTU per sample, i.e., “richness”. The relative abundance of the different OTUs making up the sample’s “richness” is defined as “evenness”. The Shannon index combines both richness and evenness of the OTUs. We used the multiple linear regression analysis for panel data (random effect) to examine the association between the Shannon index (the dependent variable) and oral health adjusting for age, ethnicity, BMI, smoking status, alcohol use, hypertension, and diabetes. PEMANOVA was used to evaluate the beta-diversity of the bacterial OTU (the dissimilarity of the bacterial community composition) based on the amount of tooth loss and gum disease using the weighted UniFrac as the distance matrix. The principal coordinate analysis (PCoA) plots were constructed to visualize the dissimilarity of the bacterial community composition [24]. The dissimilarity of the community composition based on tooth loss or gum diseases was also evaluated according to age groups (50-60, 60-<70, ≥70), ethnicity (non-Hispanic white, African American, and Hispanics), hypertension (yes vs. no), and diabetes (yes vs. no).

The relative abundances of bacterial taxa were compared between four groups using the Kruskal-Wallis test and two groups using the Mann-Whitney test. For those major bacteria (relative abundance ≥ 1%) that differed significantly by the amount of tooth loss and gum disease (q values < .05), the multivariable negative binomial regression analysis [25] was used to estimate the incidence rate ratio (IRR) and its 95% confidence interval (CI) of having the non-zero bacteria count in association with tooth loss and gum disease (two variables were simultaneously included in all the models as the ordinal variable), adjusting for age (continuous), race/ethnicity (non-Hispanic Caucasian, African American, and Hispanic), body mass index (BMI) (continuous), smoking status (never, former, and current), alcohol use (never, former, and current), diabetes (yes vs. no), hypertension (yes vs. no), biopsy segment, and HEI score. We conducted the complete case analysis when dietary data were not available. To account for multiple biopsies from some of the participants, we used the panel data analysis, assuming each individual was a panel (random effect). We also performed a sensitivity analysis by including 37 rectum samples from 37 participants to examine the robustness of our findings.

We used the STATA 16.0 (Stata Corp LLC, College Station, TX) and R program for data analysis. All tests were two-sided. In the general analyses, the P value < .05 indicated statistical significance. In the microbiota analysis, all P values were adjusted for multiple comparisons using the false discovery rate (FDR) algorithm [26]. FDR P-values (q value) < .05 indicated statistical significance.

Results

Our final analysis consisted of 62 participants, 51 to 75 years old, 95% were men, and 53% were non-Hispanic Caucasians. The proportion of patients with hypertension or diabetes was insignificantly higher in those with more tooth loss compared to less tooth loss. Participants who had more tooth loss were significantly older (P value for the non-parametric trend test < .0001) and were more likely to use dentures or implants. Those who wore full dentures did not report a history of gum diseases. Their tooth loss could be due to tooth decay or unknown reason. The proportion of current smokers was 42.9% in those who lost all their teeth vs. 0 in those with no missing tooth (P value = .21) (Table 1). We included 197 mucosal samples (35 cecum, 37 ascending, 27 transverse, 25 descending, 36 sigmoid colon, and 37 rectum samples). The composition of the gut microbiota did not differ by the colon segment (P value for beta-diversity = .99) (Supplemental Fig 1).

Table 1.

Basic characteristics of 62 participants based on the amount of tooth loss

Characteristics Mean (SD) or n (%)	Amount of Tooth Loss
	None (n =9)	1-5 (n=27)	6 or more (n=19)	All (n=7)	P value*
Age (years), mean (SD)	56.6 (6.0)	61.5 (6.8)	63.2 (5.0)	66.1 (4.8)	.01
Men, n (%)	9 (100)	25 (92.6)	18 (94.7)	7 (100)	.75
Ethnicity, n (%)					.50
Non-Hispanic Caucasian	5 (55.6)	15 (55.6)	11 (57.9)	2 (28.6)
African American	4 (44.4)	8 (29.6)	4 (21.1)	3 (42.9)
Hispanic	0 (0)	3 (11.1)	4 (21.1)	2 (28.6)
BMI (kg/m2)	33.7 (4.6)	32.0 (5.5)	33.1 (6.8)	32.0 (6.5)	.88
Hypertension (yes, n (%))	5 (55.6)	16 (59.3)	15 (78.9)	5 (71.4)	.48
Diabetes (yes, n (%))	2 (22.2)	10 (37.0)	10 (52.6)	4 (57.1)	.36
Smoking status (n (%))					.21
Never	5 (55.6)	10 (37.0)	7 (36.9)	0 (0)
Former	4 (44.4)	13 (48.2)	8 (42.1)	4 (57.1)
Current	0	4 (14.8)	4 (21.0)	3 (42.9)
Alcohol use (n (%))					.73
Never drinker	2 (22.2)	9 (33.3)	4 (22.2)	2 (28.6)
Former drinker	2 (22.2)	10 (37.0)	5 (27.8)	1 (14.3)
Current drinker	5 (55.6)	8 (29.6)	9 (50.0)	4 (57.1)
HEI score, mean (SD)	60.0 (6.5)	63.0 (9.7)	59.5 (9.2)	63.9 (11.2)	.68
Dietary quality (n (%))
HEI < 60	3 (33.3)	6 (22.2)	8 (42.1)	3 (42.9)	.48
HEI ≥ 60	6 (66.7)	21 (77.8)	11 (57.9)	4 (57.1)
Total fiber (g/1000 kcal/day)	6.47 (1.09)	7.85 (2.86)	7.78 (2.43)	8.59 (2.93)	.62
Saturated fat (g/1000 kcal/day)	13.3 (3.29)	12.7 (2.85)	13.8 (3.5)	13.0 (2.31)	.78
Wearing denture (n (%))
No	7 (77.8)	23 (85.2)	12 (63.1)	1 (14.3)	<.0001
Implant/partial denture	2 (22.2)	4 (14.8)	6 (31.6)	1 (14.3)
Full	0 (0)	0 (0)	1 (5.3)	5 (71.4)
Gum disease (n (%))					.30
No	8 (88.9)	21 (77.8)	13 (68.4)	7 (100)
Yes	1 (11.1)	6 (22.2)	6 (31.6)	0 (0)
Segments site (n (%))					.49
Cecum	5 (55.6)	17 (63)	9 (47.4)	4 (57.1)
Ascending	1 (11.1)	5 (18.5)	2 (10.5)	1 (14.3)
Transverse	0 (0)	0 (0)	2 (10.5)	1 (14.3)
Descending	0 (0)	0 (0)	1 (5.3)	1 (14.3)
Sigmoid	2 (22.2)	5 (18.5)	3 (15.8)	0 (0)
Rectum	1 (11.1)	0 (0)	2 (10.5)	0 (0)

BMI; Body mass index; HEI: Healthy eating index; SD: standard deviation.

^*P value for ANOVA test for a continuous variable (age, BMI, HEI score, total fiber, and saturated fat intake), and Fisher’s exact test for a categorical variable (sex, ethnicity, hypertension, diabetes, smoking status, alcohol use, dietary quality, wearing dentures, gum diseases, and segment site).

More tooth loss and the presence of gum disease were associated with lower alpha diversity (Fig. 1A and Fig. 1B). The Shannon index did not differ significantly based on the tooth loss category. However, the linear regression analysis adjusting for age, ethnicity, smoking status, alcohol, diabetes, and hypertension showed a significant association between the Shannon index and tooth loss (coefficient - 0.17, P value = .04). The beta diversity of the colonic microbiota also differed significantly by tooth loss and gum disease (P values < .001) (Fig. 2A and Fig. 2B). The dissimilarity of the community composition was seen regardless of age, ethnicity, hypertension, and diabetes (Supplemental Table 1). However, there was no distinct separation between the community compositions based on tooth loss or gum diseases.

Fig. 1.

A The difference in alpha diversity of the gut microbiota based on the amount of tooth loss was tested using the Kruskal-Wallis test. The q value was .04 for the observed OTU and was .15 for the Shannon index.

B The difference in alpha diversity of the gut microbiota based on the presence of gum diseases was tested using the Mann-Whitney test. The q values for the observed OTU and the Shannon index were < .05.

Fig. 2.

A Beta diversity of the gut microbiota based on the amount of tooth loss. Principal coordinate analysis of the bacterial OTU based on the weighted Unifrac distances matrix, encircled by 95% confidence interval ellipses. The proportion of variance explained by each principal component was denoted in the corresponding axis. P value was for the PERMANOVA test. Although P value = 0.001, there was no obvious separation of bacterial community composition based on the amount of tooth loss groups.

B Beta diversity of the gut microbiota based on the presence of gum diseases. Principal coordinate analysis of the bacterial OTU based on the weighted Unifrac distances matrix. The proportion of variance explained by each principal component was denoted in the corresponding axis label. P value was for the PERMANOVA test. Although P value = 0.002, there was no obvious separation of bacterial community composition based on the absence or presence of gum disease.

At the phylum level, more tooth loss was associated with a lower relative abundance of Firmicutes (q value = .004) and a higher relative abundance of Proteobacteria (q value = .027). Bacteroidetes was also more abundant in participants who lost some but not all teeth (q value = .002) (Fig. 3). Gum disease was not related to the relative abundance of bacterial phylum (data not shown).

Fig. 3.

Stacked bar chart of the relative abundance of the major phyla based on the amount of tooth loss. The Kruskal-Wallis test was used to test the difference. q values was .004, .006, and .03 for Firmicutes, Bacteroidetes, and Proteobacteria, respectively. The bacteria were ordered by q value. The q value for Firmicutes was the smallest.

Among 93 genera with relative abundance > 0.05%, 44 genera differed significantly by the amount of tooth loss (q values < .05). Fig 4 shows the major genera (relative abundance ≥ 1%) differed by tooth loss. Supplemental Fig 2 summarizes the less abundant (relative abundance between 0.05% and 1%) bacteria that differed significantly by the amount of tooth loss. A total of eight genera (relative abundance of ≥ 0.5%) differed significantly by the presence of gum disease (q value < .05) (Fig. 5).

Fig. 4.

Stacked bar chart of the relative abundance of the major bacterial genera (≥ 1% relative abundance) that differed significantly based on the amount of tooth loss (q values < .05, Kruskal-Wallis test). The bacteria were ordered by q value. The q value of the bacterium on the left side of the horizontal axis was the smallest.

Fig. 5.

Stacked bar chart of the relative abundance of the bacterial genera (relative abundance ≥ .05%) that differed significantly based on the presence of gum diseases (q < .05). The Mann-Whitney test was used to test the difference. The bacteria were ordered by q value. The q value of the bacterium on the left side of the horizontal axis was the smallest.

Using the multivariable negative binomial regression models for panel data, we found that the incidence rate of having a non-zero count of Faecalibacterium was lower with the higher amount of tooth loss (IRR: 0.62, 95% CI: 0.41-0.93) compared to no tooth loss. The incidence rate of having a non-zero count of Bacteroides, Escherichia, and LachnospiraceaeUnc94789 was insignificantly higher with a higher amount of tooth loss. The association between Bacteroides and gum disease was not statistically significant in the age-adjusted and multivariable model. Table 2 did not show the results of other major bacteria that were not associated with the amount of tooth loss or gum disease in the age-adjusted model.

Table 2.

Incidence rate ratio (IRR) and 95% confidence interval (CI) of having non-zero count of specific bacterial genera based on the amount of tooth loss and gum diseases

Taxa	IRR* (95% CI)	IRR** (95% CI)	IRR*** (95% CI)
	Amount of tooth loss
Faecalibacterium	0.66 (0.46-0.93)	0.62 (0.41-0.93)	0.54 (0.31-0.94)
Bacteroides	1.22 (1.04-1.42)	1.24 (1.02-1.49)	1.18 (0.87-1.60)
Escherichia/Shigella	1.44 (1.00-2.07)	1.23 (0.83-1.83)	0.93 (0.46-1.89)
	Gum disease
Bacteroides	1.29 (0.94-1.79)	1.14 (0.81-1.62)	1.50 (0.86-2.61)

CI: Confidence interval. IRR: incidence rate ratio. The significant associations were bolded.

^*In the negative binomial regression model for panel data, the rarefied bacterial count was the dependent variable. The amount of tooth loss and gum disease was included as the ordinal explanatory variables in the model simultaneously. The model was adjusted for age only. Compared to no missing tooth, the incidence rate of having non-zero Faecalibacterium count was reduced by 34% with one unit increase in the tooth loss category. The incidence rate of having non-zero Bacteroides count was increased by 22% with one unit increase in the tooth loss category.

^**The model was further adjusted for race (non-Hispanic white, African American, and Hispanic), body mass index, smoking status (never, former, and current), alcohol use (never, former and current), diabetes (yes vs. no), hypertension (yes vs. no), and biopsy segment. Compared to no missing tooth, the incidence rate of having non-zero Faecalibacterium count was reduced by 38% with one unit increase in the tooth loss category.

^***The model was further adjusted for the HEI score. Compared to no missing tooth, the incidence rate of having non-zero Faecalibacterium count was reduced by 46% with one unit increase in the tooth loss category.

The sensitivity analysis was performed based on the rectum mucosa (Supplemental Fig. 3). The observed OTU was the lowest among those who lost 6 or more teeth (q value = .09). The beta-diversity differed significantly by the amount of tooth loss (P value = .01). We observed a similar pattern of difference for Faecalibacterium, LachnospiraceaeUnc94789, and Escherichia in relation to the amount of tooth loss, as we observed in 197 biopsies. The relative abundance of Bacteroidetes was higher in those with 6 or more teeth lost (22.7%, 32.5%, 36.0%, and 21.3% for none, 1-5, 6 or more, and all teeth lost, respectively. P value = .27).

Discussion

In this cross-sectional study of polyp-free adults, more tooth loss was associated with lower alpha diversity of the gut bacteria, as well as the significant changes in the relative abundance of Firmicutes, Bacteroidetes, and Proteobacteria. More tooth loss was associated with a lower relative abundance of Faecalibacterium independent of age, ethnicity, smoking status, alcohol use, obesity, hypertension, diabetes, and diet. The analysis of the rectum samples confirmed the difference in Faecalibacterium based on the amount of tooth loss. The presence of gum disease was associated with the lower alpha diversity and the insignificant higher relative abundance of Bacteroides.

Lower alpha diversity was associated with more tooth loss and the presence of gum disease in our study. Similarly, Colombo et al. showed a lower alpha diversity in the fecal microbiota of individuals with chronic periodontitis [27]. Another study showed a lower alpha diversity of the oral microbiota of individuals with periodontal disease [28]. The composition of the gut microbiota also differed significantly by the amount of tooth loss and the presence of gum disease, similar to several studies on oral microbiota and periodontitis [29, 30]. In general, the findings on the biodiversity of gut microbiota in the colonic mucosa were in line with the previous studies using feces or subgingival samples.

We found that more tooth loss was significantly associated with the lower relative abundance of Faecalibacterium independent of multiple covariates. Faecalibacterium is one of the major genera in the Firmicutes phylum and the Ruminococcaceae family and can secrete short-chain fatty acids, including butyrate, which may have systemic anti-inflammatory properties [13, 31]. One in vitro study showed that peripheral blood mononuclear cell stimulation by F. prausnitzii led to reduced pro-inflammatory IL-12 and IFN-gamma production and increased anti-inflammatory IL-10 [13]. Another study showed that certain strains of Faecalibacterium could produce butyrate and promote the release of IL-10 [32]. Several animal studies have shown that Faecalibacterium could inhibit the NF-κB pathway and prevent colitis [33, 34]. The depletion of Faecalibacterium has been associated with inflammatory bowel disease and colorectal cancer [13, 35]. Our study suggested that poor oral health may be an indicator of the reduced anti-inflammatory status in the body mediated by Faecalibacterium or other bacteria in the colon.

Proteobacteria and Escherichia/Shigella were more abundant in those who lost more teeth. Our study is congruent with another study, which showed that periodontitis was associated with an increased abundance of Escherichia/Shigella in feces independent of diet [21]. Escherichia/Shigella has been associated with dysbiosis and several chronic diseases [36, 37]. Escherichia/Shigella is a known pro-inflammatory bacterium and can induce pro-inflammatory cytokines via the NF-κB pathway [38]. However, in our study, the association between Escherichia/Shigella and tooth loss was not significant in the multivariable negative binomial regression models.

The phylum Bacteroidetes and genus Bacteroides were more abundant in participants who lost some but not all their teeth. Higher Bacteroides was also associated with having gum disease. However, the difference was not significant in the multivariable model. Similarly, an oral microbiota study has shown an increase in these bacteria with periodontitis [39]. Because the 16S rRNA sequencing did not provide species and strain levels of information, the interpretation of our findings was not straightforward because different Bacteroides species exert different functions [40], and the function of the Bacteroides species is incompletely known. In our study, all participants who lost all their teeth wore dentures. Denture use has been shown to affect the community composition of the oral microbiota and has been associated with decreased HEI scores [41, 42]. However, the dietary quality of those who wore dentures did not differ significantly from those who lost none or some of the teeth in our study. Differences in other exposure factors may explain why Bacteroides did not further increase in those who lost all their teeth.

Several unclassified members of the Lachnospiraceae and Ruminococcaceae families differed by tooth loss and gum disease. These two bacterial families were mostly influenced by diet. We would expect that those with more tooth loss had an altered diet secondary to decreased mastication abilities. However, we did not observe a significant difference in dietary quality, total fiber intake, and saturated fat intake based on the amount of tooth loss. Nevertheless, adjusting for age attenuated the association between tooth loss and these bacteria, including Subdoligranulum and Lachnoclostridium. Future metagenomics studies with higher sequencing resolution should identify those unclassified bacteria and elucidate their functions in oral health and inflammation.

There are several possible mechanisms for the association between oral health and the colonic adherent microbiota. It has been shown that oral microbiota can travel via ingested food and drink to the gut and influence the gut microbiota [43]. However, because Faecalibacterium is an extremely oxygen-sensitive anaerobic bacterium that is found only in the gut [44], it is unlikely that oral bacteria would explain the observation on the lower Faecalibacterium abundance in those with more teeth lost. Therefore, we raised the other possibility, i.e., diet, smoking, and lifestyle factors not only can shape the oral microbiome, but also modulating the adherent gut microbiota [45–47]. The colonic microbiota could exert its influence via soluble inflammatory pathways [5–7]. The NF-κB pathway and the toll-like receptor signaling pathway in the host have been shown to be affected by Escherichia and Faecalibacterium [48, 49]. In summary, tooth loss may serve as an indicator of lifestyle-related chronic inflammation rooted in the colon, which was mediated by anti-inflammatory and possibly pro-inflammatory gut microbiota.

Our study had multiple strengths. We examined the adherent gut microbiota that may be more relevant to the immune functions of the host. We attempted to control for potential confounding factors in the multivariable analysis. Participants did not receive antibiotics in the past three months. However, there were several limitations. Firstly, the study participants were mostly men, which limited the generalizability of the study findings to the broader population. Secondly, we could not establish the causality between diet, tooth loss, and the gut microbiota in this cross-sectional study. We did not ascertain the history of tooth loss and diet history in early life. Participants may have lost their teeth at a young age due to lifestyle factors (such as smoking) that can promote chronic inflammation throughout their lifetime. Thirdly, because the diet information was available in 40 of 62 participants, the study power was reduced when the diet was considered a confounding factor. Lastly, we did not examine the oral microbiota, and thus could not compare the profile of the oral microbiota with the colonic mucosaln microbiota in the same individuals. Future studies should evaluate what oral microbiota profile can be a surrogate biomarker of “dysbiosis” of the gut microbiota.

In summary, our study showed that the community composition and structure of the colonic mucosa microbiota were associated with tooth loss, and to a lesser extent, gum disease. Poor oral health may be an indicator of inflammation rooted in the colon that partially attributable to unhealthy lifestyle factors. These lifestyle factors can influence both oral and gut microbiota. More research should be conducted to investigate the systemic role of the gut microbiota exert in response to lifestyle factors, medication use, or endogenous stress (such as biological aging), and their involvement in a myriad of diseases. Nevertheless, those individuals with poor oral health may benefit from adequate dental care, pre- or pro-biotic, or dietary intervention. Future study is needed to confirm our findings and elucidate the universal role of Faecalibacterium and other gut bacteria in chronic inflammatory diseases.