Oral wash samples were collected in compliance with IRB and IBC approved protocol (IRBNET ID # 388951-2) and clinical consent was obtained from the volunteers. The study was carried out in the Microbial Biotechnology laboratory of the Charles E Schmidt College of Science at Florida Atlantic University, Davie, Florida.

Samples were collected from 5 healthy individuals above the age of 18 years. These individuals were antibiotic free; were not addicted to smoking for at least 3 months preceding the study; and adhered to their normal oral hygiene for 2 weeks before enrolment in the study. Samples were also collected from 5 smokers for comparison. The demographic information, periodontal parameters and smoking status for each group are summarized in Table 1. Oral wash samples were prepared by providing subjects with a 25 ml of bottled drinking water for oral gargle first thing in the morning. The sample is then collected in a 50 ml sterile Falcon tube that was provided by the lab. Samples were frozen at −20˚C. They were also asked to complete a questionnaire for reference.

a) Direct extraction of metagenomes from the oral wash

Samples were thawed by incubating at room temperature for 15 - 20 mins and 20 ml of the sample were transferred in 15 centrifuge tubes (1.5 ml) and centrifuged at 10,000 rpm for 10 mins in an aerosol-tight microcentrifuge (Eppendorf 5424R) to collect bacterial pellets. The pellets were used for direct extraction of the metagenomes. Total bacterial genomic DNA from oral wash samples was isolated using the Qiagen Dneasy Blood and Tissue DNA purification kit (Qiagen, Valencia, CA). In brief, the pellets were suspended in Phosphate Buffer Saline and washed twice by spinning at 15,000 rpm for 10 mins per wash. 200 µl of enzymatic lysis buffer was added to the final pellet and incubated at 37˚C for 1 - 2 hours. A 25 µl aliquot of a proteinase K (Qiagen, Valencia, CA) and 200 µl of AL buffer were added and incubated at 56˚C for 30mins. 200 µl of 96% - 100% ethanol was added to the sample and mixed by vortexing. The mixture was then be transferred in a QIAmp Mini Spin column with silica membrane for the adsorption of bacterial DNA Samples were washed with AW1 and AW2 buffer which were amended with alcohol as prescribed by manufacturer and the DNA was eluted using the 100 µl of the AE buffer.

b) Genomic DNA extraction from plated oral wash samples

A sub-set (1.5 ml) dissolved in Phosphate Buffered Saline (PBS) of each pellet from “a” above was plated (100 µl) on Brain Heart Infusion Agar and incubated aerobically for 48 hours at 30˚C. Two milliliters of sterile PBS was added to the agar culture plates to resuspend all visible growth into slurry. The community of cultivable bacteria was then transferred to microfuge tubes and centrifuged at 10,000 rpm for 10 mins to collect the bacterial pellets.

The quality and quantity of DNA samples were measured at 260 nm and 280 nm using a Nanodrop 2000C Spectrophotometer (Nanodrop products, DE, USA) respectively.

PCR was performed using the Mastercycler Nexus PCR Systems (Eppendorf North America, NY, USA). The bacterial 16S rDNA was amplified using the universal primer 1492 Reverse (5’GGTTACCTTGTTACGACTT- 3’) and the primer 27 Forward (5’-AGAGTTTGATCCTGGCTCAG-3’) that targets the V1-V9 region of the 16S rRNA gene. Reaction mixtures were incubated for 4 mins at 94˚C for denaturation, followed by 35 cycles consisting of 1 min at 94˚C, annealing for 30 s at 45˚C, and extension for 2 mins at 72˚C. PCR product (approximately 1500 bp) will be confirmed by electrophoresis in 1% agarose (w/v) get at 120 V for 30 mins. 10 µl of the amplified PCR product will be utilized for restriction digest.

PCR-amplified bacterial 16S rDNA was digested using 3 restriction enzymes: HaeIII, Alu1 and Sau3A (1U), ac-

cording to the manufacturer’s instructions (New England Biolabs, Beverly, Mass.) [29] . The DNA fragments were separated by gel electrophoresis on a 3% (w/v) low melting agarose gel run in 1X TAE (Tris Acetate- EDTA, pH-8) buffer. The RFLP ribopattern profile were digitally captured and recorded by means of the FOTO/ Analyst^® Investigator/FX System using the PC Image Acquistion Software (Fotodyne Incorporated, Hartland, WI).

A small number of plasmid colonies of PCR products were transformed into chemically competent E. coli cells using the Gene Hunter PCR-Trap vector and pCR 4-TOPO high copy number vectors (Invitrogen, Carlsbad, CA, USA) according to manufacturer’s manual. 500 µls of each transformation were plated on LB plates containing 50 ug/ml tetracycline and were grown overnight. Cells from 10 colonies picked from selective LB agar plates were used to create a library and were dispatched for sequencing. Sequencing reactions were performed on an ABI GeneAmp^® 9700 thermal cycler (Applied Biosystems) using v3.1 Big Dye kit according to the manufacturer’s standard protocol of 25 cycles, followed by purification using Performa^® gel filtration short plates (Edgebio, Gaithersburg, MD, USA). The sequencing runs were performed using POP7 polymer on 36-cm capillaries in an ABI 3730 DNA sequencer (Applied Biosystems) and were analyzed by ABI Sequencing Analysis software v5.1 and KB Basecaller (both from Applied Biosystems) capable of sequence quality value (QV) prediction on each base call. Sequences were analyzed by BLAST on the RDB II database and the HOMD. Diversity indices were also calculated for PCR-RFLP analysis: the Shannon-Wiener index (H’) was calculated from H’ = −ΣP_ilnP_i) where P_i is the proportion of the phylotypes.

An image of the RFLP gel was captured by Fotodyne Imager as mentioned above. The banding patterns were examined to identify the commonalities and differences (which is presumed to represent the pattern for the biomarker strain) among the 2 different study groups. This was achieved by using the Gel Compare II (Applied Maths, Kortrijk, Belgium) software to analyze the 16s rDNA fingerprint. This software searches for discriminative bands between selected groups of patterns and also searches for unique and common bands within selections. To obtain an objective comparison and normalization of the bands on different slab gels, a molecular weight marker was included in each gel at least two times. Digitized images were normalized and combined [31] . A similarity matrix was created by using the Dice similarity coefficient [32] [33] . For the same slab gels similarity matrix was created using the Pearson correlation coefficient using the densitometry curve created by the fingerprints. Similarities were displayed graphically as a dendrogram. For band comparison, a band position tolerance value of 0.8% was allowed to compensate for misalignment of homologous bands due to technical imperfections. The unweighted pair group method using average linkages (UPGMA) [31] was used to cluster the patterns.

DNA isolates from direct oral wash and cultured samples of ten subjects (5 healthy and 5 smokers) 3 months prior and post yielded 40 samples in total. All 40 samples were digested with the selected endonucleases for restriction analysis. The three enzymes generated distinctive PCR-RFLP finger prints in cultured and metagenomic subsets. The ribopattern for the cultured and metagenomic subset were also diverse. The analysis and comparison of restriction profiles obtained with the digestion of the amplified 16SrRNA gene sequences allowed the identification of a set of appropriate frequent-cutter endonucleases. Restriction enzyme HaeIII and Sau3AI was found to give the clearest and most discriminatory profiles for PCR-RFLP patterns compared to AluI.

PCR-RFLP ripo pattern using HaeIII of initial samples for Healthy non-smokers and smoker yielded 3 well resolved stable bands of 700 bp, 500 bp and 350 bp in both cultured and metagenomic subsets (Figure 1(a) & Figure 1(b)). A fourth band (1500 bp) of lighter intensity was seen in three smokers (Smoker #3, Smoker #4 and #5) and 2 non-smokers (Healthy #1 and #2) in the cultured subset. In the same subset four smokers (Smoker #1, 2, 4 and 5) also displayed a unique 300 bp band which was absent in the other subjects (Figure 1(b)). The restriction patterns of the amplified 16S rDNA were used for cluster analysis by UPGMA. Analysis of the cluster revealed that both healthy non-smoker and smoker groups were tightly grouped together in both cultured and metagenomic sets at a linkage level of (S_D) 70%. However, it aligned one healthy subject (Healthy #3) with the smokers in the cultured subgroup (Figure 2(a)).

Post 3 months analysis showed the presence of three similar ubiquitous bands as was observed during the initial data analysis, affirming the existence of a core group in all the samples. This outcome suggested that this core group is well established and does not change over time. The 1500 bp fragment which was more prominent in smokers was intensified in the healthy subjects, Healthy #1 and #2. Clustering data of the cultured subset

showed that the smokers were clustered together but the healthy groups were scattered with Healthy #1, #2 and #3 falling out of the cluster. In the metagenomic subset the healthy samples showed similarities with the smokers (Healthy #1 with Smoker #2, Healthy #4 with Smoker #1 and Healthy #3 with Smoker #5) (Figure 2(b)).

Sau 3AI yielded a more complex fingerprint. Sau3AI restriction produced 4 well-defined stable fragments (1000 bp, 380 bp, 180 bp and 150 bp) in both the subsets (Figure 3(a) & Figure 3(b)). The fingerprint of the cultured subtype showed 2 fragmented bands (800 bp and 700 bp) which were variable and becomes less intense/disappears in smokers (Figure 3(b)). This suggested the alteration of well-established core microbiome with changing oral health. Hence, a shift in oral microbial community was captured. These bands are critical and could serve as potential signatures of compromised oral health. Clustering pattern of the initial metagenomic isolates showed a tight clustering of smokers (Figure 4(a)). The healthy samples were widely distributed some showing similarity with smokers. Mention worthy were samples Healthy #1, #3 and #4 in both cultured and metagenomic subset (Figure 4(a)). There was no noticeable change in the ribopatterns of samples collected after 3 months (Figure 4(b)). Clustering of the fingerprints exhibited the similar relatedness of Healthy #1, #3 and #4 with smokers among the metagenomic subgroup (r > 0.60). Both groups demonstrated noticeable tight clustering unlike initial pre 3 months analysis.

Alu I produced 5 fragments (1200 bp, 900 bp, 580 bp, 300 bp, 260 bp) present in both cultured plate wash and metagenomic extracts (Figure 5(a) & Figure 5(b)). A fragment of 450 bp length was distict in smokers in the cultured plate wash category (Figure 5(b)). The cluster analysis based on densitomety curve grouped healthy non-smokers and smokers into closely clustered groups within the culture subset of the initial sample. The metagenome isolates however showed similar alignment of the Healthy #3 and #4 with the smokers (Figure 6(a)). There were no distinguishable differences in the data analysis post three months (Figure 6(b)).

HaeIII ribopattern for the initial sampling showed that both healthy and smokers were tightly clustered. Noteworthy, initial cluster analysis of the fingerprints produced by Sau3AI captured the shift in the health status of Healthy #1, #3 and #4 at an early stage. This shift was later captured by HaeIII ribotyping in the post 3 month’s analysis as mentioned earlier. This observation prompted a careful analysis of the questionnaire. It showed that Healthy #1 was a smoker for 15 years. It was evident that even though the subject quit there was a signature that persists and thereby aligning him with the smokers. It was also found that Healthy #4 was exposed to passive cigarette smoke and hence harbours a signature tax on that led to the clustering with smokers. It is noteworthy that the subjects self-identified their categories but the protocol developed in this study was able to determine the appropriate groups. This outcome correlates with the finding that every individual has a variable set of microbiome which is affected by the host’s environment and lifestyle. It was however uncertain why Healthy #3 showed a similarity with smokers.

When subjected to PCR with primer pair F27/R1492 both the metagenomic and cultured plate wash isolates

yieldeda 1300 bp product. Sequencing data for the amplified PCR in Healthy non-smokers and smoker showed that the microbial diversity was significantly high in smokers than in non-smokers (p = 0.04) (Figure 7).

Comparing healthy with smokers we observed a shift in the composition of oral microbiota composing a well-defined transition from gram-positive to gram-negative dominated community in smokers. Staphylococcus and gram-negative genera like Neisseria, Thiomargarita, Granulicatella, and Synechococcus were more preva-

lent in smokers. Interestingly, Healthy #1 and Healthy #3 showed presence of the genera Pyrobaculum and Neisseria asserting our previous finding of a shift in the oral microbial community and a resemblance with smokers. These strains were not seen among other healthy samples. It is noteworthy that Streptococcus mitis which is being associated with various pancreatic diseases, including pancreatic cancer, are less abundant in healthy subjects and predominant in smokers. Xanthomonas, Streptococcus and phylum Candidatus, were also well distributed in both the groups and are probable members of the core microbiota (Figure 8). Streptococcus sanguinis were abundant in both groups. Streptococcus infantis, Streptococcusoralis and Streptococcus perosis were dominant in smokers. Interestingly, Streptococcus pneumonae was abundant in healthy non-smokers however studies have shown increased pneumoccal adherence in smokers than in non-smokers. Streptococcus oralis, are recognized as important etiological agents of dental caries and sub-acute bacterial endocarditis and septicemia [34] .