Evaluation of Buccal Cell Samples for Studies of Oral Microbiota

Recent studies have revealed associations between the features of human microbial communities (human microbiota) and various diseases, including autoimmune disorders, diabetes, obesity, colon cancer, and even psychiatric conditions (1–7). Large prospective, population-based studies involving thousands of participants are needed to confirm those results and systematically study the role of the human microbiota in health. Profiles of the microbiota can be generated with next-generation DNA sequencing methods, but very few large long-term population-based cohorts have collected biological specimens for the purpose of microbiota research. Buccal cells, which have been obtained from mouthwash in many cohorts for the study of human genomic DNA, might be employed to study the oral microbiota. The purpose of the current study was to evaluate whether buccal cell specimens, collected by the protocol used in a prospective cancer cohort (http://prevention.cancer.gov/major-programs/prostate-lung-colorectal), could be used for oral microbiota research by comparing buccal cell specimens with eight other oral sample types from the same subjects.

Forty-three subjects were recruited at Eastman Institute of Oral Health, University of Rochester (Rochester, NY). All the subjects signed informed consent and filled out questionnaires. Results are based on 41 subjects after quality control–determined exclusions. Individuals with antibiotic usage or professional dental cleaning within the last 3 months or diagnosed with severe periodontal disease or cancer were excluded. The study was approved by Institutional Review Boards of the NCI (Rockville, MD) and University of Rochester (Rochester, NY).

We collected 8 oral samples, including 2 dental plaque samples (supra- and subgingival plaque), raw saliva, and swabs from 5 soft tissue sites (keratinized gingiva, hard palate, buccal mucosa, palatine tonsil, and tongue dorsum) by following the protocols of the Human Microbiome Project (HMP; ref. 8; http://hmpdacc.org/doc/HMP_MOP_Version12_0_072910.pdf; see Supplementary Fig. S1 for detailed sampling locations). In addition, buccal cells from mouthwash were also collected following the protocol used in the Prostate, Lung, Colorectal and Ovarian (PLCO) cancer cohort (Supplementary Methods).

DNA was extracted and purified as described previously (9). Briefly, samples were thawed on ice and then vortexed for 5 minutes to resuspend the cells. A 0.5-mL aliquot was transferred to a sterile 2.0-mL tube for cell lysis chemically and physically (bead-beating). The resulting crude lysate was processed using QIAamp DNA Mini Kit (Qiagen) according to the manufacturer's guidance. The samples were eluted with 2 × 200 μL of AE buffer (10 mmol/L Tris-Cl, 0.5 mmol/L EDTA; pH 9.0) into separate tubes. The DNA concentrations in the samples were measured using the Quant-iT PicoGreen dsDNA Assay Kit from Molecular Probes (Invitrogen).

The V3–V4 regions of the 16S rRNA gene were amplified and sequenced on an Illumina MiSeq instrument using the 300 paired-end protocol at the Institute for Genome Sciences, Genomic Resource Center, University of Maryland School of Medicine (Baltimore, MD; ref. 10). The sequence reads were processed by removing the low-quality and short-sequence reads and samples with <1,000 reads (11) and then clustered into species-level operational taxonomy units (OTU; species-level OTUs or observed species) at 97% identity in the Quantitative Insights Into Microbial Ecology (QIIME 1.8.0) pipeline (12). The sequence data have been submitted to NCBI (BioProject accession number PRJNA316469).

Alpha diversity measures of number of observed species (species-level OTUs), and Shannon index (adjusted for species relative abundance; ref. 13) were estimated by averaging more than 20 rarefied tables (1,000 sequence reads/sample). Beta diversity, which measures the pairwise difference among samples, was estimated as the UniFrac distance, unweighted and also weighted by the relative abundance of each species (14). To rule out the batch effects, 19 random samples were duplicated within and between batches. No differences within and between batches were found in alpha and beta diversity measures (Supplementary Fig. S2).

Differences in alpha diversity between buccal cells and each of the other sample types were compared by Wilcoxon signed-rank tests. To examine differences in beta diversity, we compared the UniFrac distance within buccal cell samples (within-buccal cell) with the UniFrac distance within each of the other sample types (within-other), and the distance between buccal cells and each of the other sample types (between-groups) by permutation tests (1,000 Monte Carlo repetitions).

To construct the dendogram for similarity and difference among the oral sample types, Bray–Curtis distance matrices were calculated from the species-level profiles by sample type and then clustered by the complete linkage clustering method. Bootstrap statistics with 1,000 bootstrap samples were calculated to estimate the proportion of samples in which each cluster was identified. A heatmap is prepared to visualize the genus-level profiles and their difference by sample types. To better show difference by samples types in each genus, z-score [(x − μ)/SD] was calculated for each genus, where x is a particular sample's relative abundance, μ and SD are mean and standard deviation of relative abundance of all samples.

The 41 participants ranged in age from 25 to 66 years; they included 20 African Americans and 21 Caucasians, 22 men, 19 women, 22 current smokers, and 19 never smokers. As reported elsewhere (submitted), essentially no differences in oral microbiome metrics were found by age, race, sex, or smoking status. All 41 participants had successful sequencing (with >1,000 reads per sample) for buccal cell, buccal mucosa, and palatine tonsil samples, whereas samples from the other six oral sites were successfully sequenced in as few as 10 participants for subgingival plaque and up to 37 participants for hard palate (Fig. 1).

Across the nine oral sample types, buccal cells had the highest alpha diversity by both observed species (unadjusted for relative abundance) and Shannon index (adjusted for relative abundance), followed by supra- and subgingival plaque, saliva and, at the low end, the five oral soft tissue swabs (Fig. 1). Of the five soft tissue swab types, buccal mucosa had the highest alpha diversity, followed by tongue dorsum, hard palate, palatine tonsils, and keratinized gingiva (Fig. 1). Buccal cells differed from each of the eight other sample types (P < 0.05) according to Wilcoxon signed-rank tests with Bonferroni correction for multiple comparisons (Supplementary Table S1).

The microbiota composition within buccal cell samples tended to be more similar across participants, as suggested by relatively smaller within-sample weighted and unweighted UniFrac distances (Fig. 2, white boxes). Specifically, for all sites except keratinized gingiva, palatine tonsils, saliva, and hard palate, within-buccal-cell samples unweighted distance was smaller (P < 0.05) than each of the within-other sample type unweighted distances (Fig. 2A, white vs. light gray boxes, Supplementary Table S2). Likewise for weighted distances (Fig. 2B), within-buccal cell samples distance (white boxes) was smaller than each of the within-other sample type (light gray boxes) weighted distances except tongue dorsum and keratinized gingiva.

Compared with UniFrac distances within buccal cell samples, UniFrac distances between buccal cells and each of the other oral sample types were larger (P < 0.05; Supplementary Table S2) as indicated in Fig. 2 (white vs. dark gray boxes), implying site-specific differences in microbiota composition.

According to the cladogram based on species-level profiles (Fig. 3, top), the oral sample types formed at least two clusters with bootstrap support >70%. One cluster included two dental plaques and tongue dorsum. Another included all the other sample types except keratinized gingiva swab. The cladogram suggested that buccal cell samples most closely resembled saliva, then hard palate and buccal mucosa, distinct from dental plaques and tongue dorsum. As shown in the heatmap (Fig. 3) for visualization of the average genus-level profile by sample type, the cluster of buccal cell, saliva, hard palate, and buccal mucosa sample had the genus-level profile different from other sample types. In particular, it featured with relative less abundance of Streptococcus.

This study found that buccal cell specimens, which were collected in mouthwash following the PLCO cohort protocol, had significantly more observed species, higher alpha diversity, and microbial communities varied less across healthy participants (beta diversity) compared with saliva and with swabs/scrapings from seven other oral sites. Thus, microbiota from buccal cell samples is diverse and relatively stable across healthy subjects. Differences were found in microbial profiles between buccal cell and other oral communities. Some differences were large, as between buccal cells and dental plaques or tongue dorsum, indicating that the microbial profile of buccal cell cannot represent the profiles of all other oral samples. Buccal cell microbiota resembles microbiota from saliva more closely than from other sites. Taken together, our study indicates that the mouthwash-based buccal specimens that have been collected and are being collected in prospective cohort studies have a diverse microbial profile that lends itself to studies of associations with cancer and other diseases.

Mouthwash collection of buccal cells for human DNA has been employed for more than a decade (15, 16). Thus, there are repositories of buccal cells that can be used to study associations of microbiota with disease. This may prove fruitful, because cancer associations have been established with bacteria in other oral sites. For example, periodontal disease is associated with several aerodigestive malignancies and two periodontal pathogens, Fusobacterium nucleatum and Porphyromonas gingivalis, that have been associated with colorectal cancer in cross-sectional studies (6, 17–19). In addition, risk of incident pancreas cancer was increased 2-fold among participants in the European Prospective Investigation into Cancer and Nutrition study who had elevated serum antibodies against Porphyromonas gingivalis (20).

The distal gut harbors some 90% of the human microbiota, and methods for collecting and stabilizing feces for microbiome research have been described (21, 22) and are being used for a wide range of clinical research (23). Promising studies with fecal specimens are only beginning, although it will require some time to achieve adequate prospective follow-up for disease outcome. In the meantime, hypotheses involving cancer and other conditions in relation to the oral microbiome could be developed from retrospective cohort studies with previously collected buccal cells. One could also conduct cross-sectional studies to determine whether variations in buccal cell microbiota are associated with variations in gut microbiota.

A logical extension of the current study is to characterize buccal cell samples that were previously collected during the course of a prospective cohort study. This will permit measurements of microbiota in relation to time of storage, collections and storage characteristics, and diverse covariates. Larger sample sizes and additional repeat studies can help better characterize random variation. We explicitly followed the protocol used by the PLCO cohort, so the available results will generalize to this large representative study. A limitation of our work is that usable sequence data were not obtained from all participants for 6 of the oral sample types. Nonetheless, investigation of the primary objective, the performance of buccal cell specimens for microbiome research, was effectively addressed, because sequencing was successful from buccal cells for all participants.

In summary, microbial DNA was successfully sequenced from buccal cell samples derived from mouthwash. Moreover, buccal cells from healthy patients in a dental clinic had microbial profiles with higher alpha diversity and less variation across participants (beta diversity) compared with other types of oral specimens. Thus, stored buccal cell samples from previous cohorts are a promising resource for studies of oral microbiota to find microbial associations with disease, although additional studies are needed to establish stability in storage. Sampling from specific oral sites may be required when dysbiosis is postulated to act locally, as may be the case for periodontitis or hairy leukoplakia (24).

No potential conflicts of interest were disclosed.

Conception and design: G. Yu, J.J. Goedert, Y. Ren, N.E. Caporaso

Development of methodology: G. Yu, S. Phillips, M.H. Gail, J. Ravel, N.E. Caporaso

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G. Yu, S. Phillips, M. Humphrys, J. Ravel, Y. Ren, N.E. Caporaso

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G. Yu, M.H. Gail, J. Ravel, N.E. Caporaso

Writing, review, and/or revision of the manuscript: G. Yu, M.H. Gail, J.J. Goedert, Y. Ren, N.E. Caporaso

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G. Yu, S. Phillips, J. Ravel, Y. Ren, N.E. Caporaso

Study supervision: G. Yu, Y. Ren, N.E. Caporaso

This work was supported by intramural research funding at Division of Cancer Epidemiology and Genetics, NCI at the NIH.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.