Introduction

Oral diseases, including dental caries and periodontal disease, are seemingly non-life-threatening diseases that are economically costly and reduce quality of life for millions of people in the United States, especially among the elderly. Oral diseases and/or disorders can significantly impact a person’s overall health1. Research has shown that oral bacteria may contribute to increased risk of systemic disease and conditions, from cardiovascular disease to preterm birth2,3.

Oral biofilms such as dental plaque play important causative roles in the development of caries and periodontal diseases. Dental plaque is a complex and diverse microbial consortium consisting of hundreds of bacterial species embedded in polymer matrix and adhered tightly to tooth or oral mucosal surfaces4. Changes in the microenvironment can drive alterations in the composition and metabolic activity of the oral biofilm at a site, resulting in a local microbiome that can be deleterious to the host5. There is also ample evidence that less abundant pathogenic oral taxa can have large effects on host response such as inflammation, and thus promote dysbiotic states6. It is also apparent that a large number of taxa in the appropriate combinations may be dysbiotic and promote disease7. The net result is that induced change in levels of multiple taxa can result in disease at a site. One part of maintaining health is inhibiting these changes or reversing them.

Compared with their planktonic counterparts, the microbes in biofilms possess elevated virulence and are highly resistant to antimicrobial agents8. While mechanical plaque elimination with assorted devices remains the primary and most widely accepted means to maintain good oral health and to reduce plaque-mediated diseases, other options are used. Chemotherapeutic agents with a variety of direct antimicrobial mechanisms have been beneficial9,10. However, long-term use of broad-spectrum antibiotics can suppress resident oral microflora and permit overgrowth of pathogenic or opportunistic organisms, and anti-bacterials like chlorhexidine can act as irritants11. Agents that change the microenvironment can drive alterations in the composition and metabolic activity of the oral microflora resulting in one that may either be deleterious or be beneficial to the oral cavity of the host5. Therefore, disease can be prevented not only by targeting the putative pathogens directly by antimicrobial strategies, but also by interfering with the selection pressures responsible for their enrichment. There is a need for agents that use these means to promote health associated bacterial populations in the oral biofilm.

Plant-derived compounds offer an abundant source of antimicrobial agents that find wide acceptance by the public as “natural products.” In recent years, “phytomedicines” have gained significant popularity in the U.S. and various natural extracts have been incorporated into oral hygiene or therapeutic products. In addition to their demonstrated in vitro antimicrobial activity, many demonstrated potential in clinical utility12. The American cranberry (Vaccinium macrocarpum) is a rich source of polyphenolic bioactives, particularly the proanthocyanidins (PACs), which may contribute to human health. This natural product has been popular in the self-treatment of urinary tract infections13,14 and has the potential to aid in control of oral biofilm15,16,17. The PACs in cranberries have been shown to possess antimicrobial and anti-adhesion effects. In vitro studies have shown that cranberry extract and its PACs interfere with cariogenic Streptococcus mutansbiofilms by inhibiting glucosyltransferases and adherent glucan synthesis, thereby disrupting bacterial biofilms18,19. Cranberry components have also been found to inhibit periodontal pathogens in vitro20,21. Wu et al.. have reported that US-marketed cranberry juice drinks (CJC, containing 27% juice) caused rapid cellular aggregation and inhibited the ex vivogrowth and biofilm formation of oral bacteria17. When children’s plaque samples were cultured ex vivo and exposed to CJC, loosely attached biofilms were noted and were easily rinsed off surfaces. This showed that CJC’s inhibition extended from affecting single species bacteria to the complex plaque biofilm community. Pre-clinical testing of natural products such as those containing cranberry juice are an attempt to predict clinical efficacy.

Numerous models have been developed to assess the efficacy of natural antimicrobial formulations. The majority of bioactivity studies have been conducted in vitro against artificial oral biofilms made of a single or several species of bacteria. However, these models do not replicate the complexity of the human oral biofilm community, which consists of hundreds of bacterial species in the oral cavity, making results difficult to interpret. Efforts to mimic the oral environment have included the establishment of stable oral bacterial microcosms representative of the multitude of bacteria at relevant sites, such as subgingival and supragingival sites of the oral cavity; coating of artificial surfaces, such as hydroxyapatite or dental materials, with pooled cell free human saliva22,23; utilization of dynamic continuous growth systems with continual replenishment of culture media24,25,26,27; and many others. Research has shown that with the appropriate growth medium and the addition of human serum or specific carbohydrates, one can establish oral biofilm consortium that preserve most of the natural microbial taxa present at oral sites28,29,30. Formation of these biofilms seemed to mimic similar steps that occur with subgingival biofilm formation in vivo27. Using subgingival plaque samples pooled from healthy sites or periodontal disease sites, subsequent cultures retained the majority of taxa from the original sample sites over several days. These artificially formed biofilms can be sourced from oral mucosa surfaces, dental subgingival plaque, or saliva27,30,31,32,33,34,35. These oral microcosms are difficult to establish and can change with time but should better mimic the in vivo environment27,32,34,36. Freezing of donor oral microbial plaque samples has been used to facilitate the process, but it may affect the subsequent biofilm formation37,38. There is a need to develop superior models that can efficiently mimic the in vivo oral cavity microenvironment to characterize mechanisms and mode of actions of antimicrobial agents against oral microbial biofilms. The development of such models may aid in selection and design of clinical oral anti-bacterial trials to promote oral bacterial populations6,7 that as a network favor health.

In this study, the goal was to develop an ex vivo assay to evaluate the effects of antimicrobial agents, in this case, cranberry juice, on oral biofilm. This was done by measuring how this and other agents affect levels of both bacterial rRNA and rDNA. Optimization was performed by identifying reproducible bacterial targets with shed oral biofilm collected and tested on different days from each of the three respective individuals. This shed biofilm assay was used to determine oral bacteria targeted by cranberry juice. The antibiotics penicillin G and vancomycin were used as positive controls.

Results

This study takes advantage of the fact that a large fraction of the bacteria in saliva are in the form of aggregates in part derived from oral biofilm which are only there for short times8,39,40. Human saliva was used as the source for shed oral biofilm because it is a reservoir of microorganisms from all ecological niches in the human oral cavity with long-term stability. Unlike earlier work, no attempt was made to store or establish the isolates in culture28,29,30,36.

Reproducibility of salivary microbiome transcriptional activity when isolated and tested ex vivo over 1 week was examined. Saliva was isolated on midmorning on day 1, day 3 or 4, and day 7, after a 24-hour period without oral hygiene. Saliva volumes ranged from 15 to 26 mL and were stimulated by chewing 1 g of non-flavored gum base (Wrigley) for 5 min. Three subjects participated in the initial run. Shed biofilm samples were prepared then incubated anaerobically for 4 h at 37 °C in test medium. RNA was isolated and then subjected to cDNA synthesis and measurement of the 16S rRNA V4 region using next generation DNA sequencing (NGS). Analysis of alpha diversity revealed sample sets from each of the three subjects showed between-subject differences that were statistically significant. Chao1 diversity comparisons are shown in Fig. 1, as are those for Shannon diversity. Analysis of beta diversity, also shown in Fig. 1, revealed taxa profiles from a single individual grouped together and were again different from those of the two other subjects. Unsupervised clustering using the Ward method based on Euclidean Distances revealed three groups with only one sample misclassified to the wrong donor (Fig. 1).

Fig. 1
figure 1

Differences in 16S rRNA levels in ex vivo samples from three different subjects identified as C_, G_ and L_. Each point represents one ex vivo sample incubated for 4 h with test media but without an anti-bacterial challenge, done in duplicate. (A) Chao1 index, G_ vs. L_ p < 0.00014, C_ vs. L_ p < 0.00025. (B) Shannon index, G_ vs. L_ p < 0.14, C_ vs. L_ p < 0.0054. (C) Beta Diversity Bray Curtis distances, PERMANOVA G_ vs. L_ F-value 29.8916, R-squared = 0.74932, p < 0.002. PERMANOVA C_ vs. L_ F-value 36.9, R-squared = 0.787, p < 0.006. G_ vs. C_, F-value 4.24, R-squared = 0.29, p < 0.01. (D) Bar plot indicating relative level of 16S rRNA transcripts from the various taxa from different subjects’ ex vivo samples (E) Heat map of nonsupervised clustering of samples based on taxa 16 S rRNA levels. Samples clustered based on taxonomic profiles.

Full size image

These same shed biofilm ex vivo isolates, as prepared each day, were exposed individually to a number of agents, antibiotics penicillin G or vancomycin or 17% cranberry juice, over the 4-hour period. As will be described below, differences in taxa rRNA expression were detected. An examination of beta diversity (Bray Curtis distances) over the whole group of 45 samples, from the three individuals, tested in the presence of three different agents on three separate days, was done using adonis2 to determine the contribution to the diversity by two factors, treatment type and patient ID. Negative controls with no agent tested were done as technical replicates. For this test set, 28% of the beta diversity was contributed by donor identity and 30% by treatment type (control, 17% cranberry juice, penicillin G, or vancomycin).

To determine if, in response to penicillin G, general differences in taxa gene expression occur that are common over these three subjects’ shed biofilm ex vivo samples, we analyzed rRNA in samples from the three donors exposed to the antibiotic penicillin G at 5 µg/mL, the optimal pharmacologic plasma level of the drug41. This drug, which is known to target Gram-positive bacteria, might be expected to reduce relative levels of 16 S rRNA in Gram-positive bacteria. Shannon alpha diversity for richness and evenness of the distribution of taxa 16S rRNA was observed to be higher after penicillin G exposure, beta diversity was also significantly different than the mock treated samples (Fig. 2). Based on DeSeq2 analysis there were lower relative numbers of a large range of Gram-positive bacteria on the genus level. With penicillin G treatment, for the three subjects, 13 genera were at levels different than seen in the control (Table 1). Of the 8 Gram-positive genera that were differentially abundant, 7 were lower in relative levels with penicillin G exposure (Table 1). For the 5 Gram-negative taxa all showed higher relative levels versus vehicle-exposed shed biofilm samples. On the species level, differences were also noted against the vehicle-exposed control samples (Fig. 3).

Fig. 2
figure 2

The effect of different agents on the abundance of bacterial 16S rRNA transcripts in shed biofilm ex vivo samples. (A) Chao1 index showed no statistically significant differences. (B) Shannon index vancomycin (Van) vs. control (Con) p < 0.0000014, penicillin G (Pen) vs. control (Con), p < 0.000075, cranberry (Cra) vs. control (Con), p < 0.32. (C) Beta diversity based on Bray Curtis distances, 3 agents and the control. (D) Beta diversity 10 µg/mL vancomycin exposure vs. control, PERMANOVA F-value 11.5543, R-squared = 0.29969, p < 0.001. (E) Beta diversity 5 µg/mL penicillin G exposure vs. control, PERMANOVA F-value 11.5543, R-squared = 0.29969, p < 0.001. (F) Beta diversity 17% cranberry juice exposure vs. control, PERMANOVA F-value 2.42, R-squared = 0.082, p < 0.010 (F). (* indicates p < 0.05).

Full size image
Table 1 Genera of bacteria that encode 16S rRNA transcripts differentially abundant on exposure to these agents in the shed biofilm ex vivo ex vivo assay. Mean results of samples from 3 subjects collected and tested on 3 separate days. Gram positive genera are bolded.
Full size table
Fig. 3
figure 3

Differential abundance of 16S rRNA after incubation of shed biofilms from three subjects tested individually. (A) After exposure to vancomycin 16S rRNA assigned to these species detected by DeSeq2, FDR < 0.10, were differentially abundant compared to the control. (B) After exposure to penicillin G 16S rRNA assigned to these species detected by DeSeq2 FDR < 0.10 were differentially abundant compared to the control. (C) After exposure to 17% cranberry juice 16S rRNA assigned to these species detected by DeSeq2, FDR < 0.10, were differentially abundant compared to the control. The y axis shows the percentage of total reads assigned to each taxa for each differentially abundant 16S rRNA. All subjects’ samples collected and tested on three separate days. Control contains 0.4% DMSO vehicle as did anti-bacterial tests.

Full size image

Vancomycin works through a mechanism distinct from penicillin G to inhibit bacteria but also targets Gram-positive bacteria. Results were roughly similar to those for penicillin G-treated shed biofilms. With exposure to vancomycin at 10 µg/mL, the ideal pharmacologic serum concentration13,14, the bacteria 16S rRNAs from the three subjects were different than vehicle-treated samples. Shannon alpha diversity, for taxa richness and evenness of spread, was higher, p < 1.5 × 106, and Beta diversity was also different. As expected for an antibiotic that targets Gram-positive bacteria, rRNA levels from Gram-positive bacteria tended to be at lower levels versus that for the control. Sixteen genera showed differences in relative levels over the 4-hour incubation (Table 1). Seven were Gram-positive taxa and all 7 were lower relative to the other taxa. Nine were Gram-negative and all 9 showed higher relative levels versus the controls.

When shed biofilm ex vivo samples from the same three subjects prepared on the same three days were exposed to a 17% mixture of cranberry juice, a large number of bacterial taxa were observed to show levels distinct from those in the vehicle-exposed control. Average alpha diversity of samples exposed to cranberry mixture versus the control vehicle-exposed isolates were similar. Beta diversity was substantially different among the two groups as shown in Fig. 2 at p < 0.002. DESeq2 was used to identify specific taxa differentially abundant for the taxa specific 16S rRNAs on average in shed biofilms samples treated with 17% cranberry versus those of the negative controls (Fig. 3; Table 1). These included on the genus level: Peptostreptococcus, Veillonella, Campylobacter, Aggregatibacter, Haemophilus, Fusobacteria, etc. and also a large number identified on the species level (Fig. 3; Table 1).

Shed biofilm ex vivo samples from a fourth subject were prepared on three separate days. They were similarly challenged with vancomycin, cranberry, or vehicle exposed in triplicate to validate some of the above findings independently (Supplemental Table 1). 16S rRNA amplicon sequencing of the RNA independently revealed substantial overlap with the findings from the earlier three subjects. Five out of 16 genera differentially represented with vancomycin treatment, including Peptostreptococcus, Alloprevotella, Granulicatella, Haemophilus, and Leptotrichia, showed similar differences in the earlier experiment (Table 1, Supplemental Table 1). For cranberry exposure, 7 out of 16 genera were similarly differentially abundant, including Peptostreptococcus, Veillonella, Campylobacter, Aggregatibacter Haemophilus, Rothia, and Alloprevotella (Table 1, Supplemental Table 1).

To address the question of if changes in rRNA expression in shed biofilm ex vivo samples treated with vancomycin, penicillin G, or cranberry were reflected in differences in relative 16S rDNA levels, both DNA and rRNA were isolated from samples exposed to the three agents and the control. Samples were from the 4 subjects, with an N of 21 for each agent versus the controls. Table 2 reveals the taxa differentially abundant on the level of 16S rRNA after exposure to each agent. This analysis is of a smaller group of samples and uses an abundance measure distinct from than that used in support of Table 1; Fig. 342, so some divergences from the results displayed there were expected. Table 3 reveals the taxa that were differentially abundant in the form of 16S rDNA (the 16S rRNA gene). Overall, more taxa showed differentially abundant 16S rRNA transcripts than 16S rDNAs after exposure to each agent versus the control (Tables 1 and 2).

Table 2 MaAsLin2 analysis revealed taxa differentially abundant on the RNA level in the shed biofilm ex vivo ex vivo assay with the various agents in a subset of test samples from 4 subjects. 16S rRNA levels were compared to those from vehicle exposed test samples.
Full size table
Table 3 MaAsLin2 analysis reveals taxa differentially abundant on the DNA level in the shed biofilm ex vivo assay of various agents in a subset of test samples from 4 subjects.
Full size table

An important question was whether levels of total 16S rDNA and 16S rRNA transcripts differed with the three different treatments versus the control. Analysis of total relative yield of 16S rDNA in each sample based on PCR revealed no differences in vancomycin-treated samples versus controls, which were at 0.91x the level of the control at p < 0.85; lower levels in penicillin G-treated samples. which showed a 0.59x the level of control samples, at p < 0.022; and for cranberry-treated samples 0.14x versus the control, at p < 0.022 (Fig. 4). When the same analysis was done at the 16S rRNA (RNA) level all treatments resulted in lower levels of 16S rRNA transcripts with exposure to vancomycin, 048x p < 0.039, penicillin G, 0.22x, p < 0.0072, and with cranberry, 0.13x, p < 0.012.

Fig. 4
figure 4

Differential abundance of 16S rRNA product and 16S rDNA ((the 16S rRNA gene) after 4 h incubation of shed biofilms with the 3 agents or vehicle. Total 16S rDNA for each incubation is normalized to the relative level of that molecule in the vehicle-treated control. Total 16S rRNA is normalized to the total 16S rRNA in the control. *Indicates differential abundance versus control based on Student t-test (p < 0.05).

Full size image

An analysis (N = 26) of total 16S rDNA levels in shed biofilm samples at experiment start and then after the 4-hour incubation with 17% cranberry extract revealed a decrease in rDNA levels to 0.43x, p < 0.0203. Shed biofilm samples incubated with penicillin G in contrast showed minimal change to 1.6x, p < 0.12 and those incubated with vancomycin trended as an increase at to 2.4x, p < 0.0658 (Table 4, right). A similar analysis of total 16S rRNA levels (N = 25) was done under identical assay conditions comparing shed biofilm samples put in test media than stopped immediately by adding RNA preservative. These were compared to samples incubated for 4 h with test media and 17% cranberry juice, vancomycin or penicillin G prior to termination (Table 4, left). No significant changes in 16S rRNA levels were observed.

Table 4 Comparison of relative level of 16S rRNA before and after 4 h exposure to agents. Comparison of relative level of 16S rDNA before and after 4 h exposure to the agents.
Full size table

Levels of bulk RNA and DNA recovered from the assayed samples revealed somewhat similar results to those with 16S rRNA and rDNA for 17% cranberry, vancomycin, and penicillin G compared to the vehicle treated control. Exposure to 17% cranberry juice treatment over the 4 h resulted in a decrease in bulk DNA versus the baseline (Base) shed saliva biofilm at the start of the assay (Supplemental Fig. 1).

Discussion

Changes in rRNA with the agents, some are true decreases

Differences in levels of rRNA in response to agents known to affect bacteria were observed in this ex vivo assay of shed oral biofilm. Like oral biofilms, salivary aggregates are thought to share the properties of relative antibiotic resistance, resistance to immune regulation, and protection from chemicals8. Relative higher and lower levels of specific 16 S rRNAs from various taxa was observed after exposure to each of the three agents versus the control. It was expected that changes indicative of loss of ribosomal RNA might be minimal during the 4-hour period, as rRNA can have turnover rates over days even after cell death, in the majority of conditions though certainly not all43,44,45. On the other hand, rapid increases in rRNA levels versus control are known to occur in responsive taxa as they increase in activity and protein synthesis levels go up46,47,48,49. For the penicillin G- and vancomycin-treated samples, one could argue lower levels of 16S rRNA levels versus the control for some species were in fact a lack of an increase. At this juncture it is not known if taxa that are lower than the control on the 16S rRNA level (Table 1; Fig. 3A, B,C) are showing true decreases or are merely bacteria that fail to be stimulated in the presence of an inhibitory agent. If the latter is the case, the assay would be restricted to testing the bacteria in the population that maintain the ability to remain active, or even begin the proliferation process, in the ex vivo state. It is reassuring that a large variety of taxa show differential levels in the presence of each of the three agents (Fig. 3A, B,C) suggesting many species are being queried.

Reproducibility of results

It was unclear how oral bacteria isolated in bulk in their native communities would respond to the same insult but on different days or times of day. In vivo studies on people suggest oral bacteria taxa DNA levels do not change greatly over time2,50, but one might think that bacteria rRNA levels may show more variability30,51. How cranberry, penicillin G, or vancomycin affects oral bacteria rRNA in vivo is not known. In addition, variability in characteristics of the ex vivo samples from day to day may contribute to lack of reproducibility. This could be due to differences in nutrients, host factors, or cells that are trapped in the matrix of the ex vivo material and not washed away during the sample preparation. However, rRNA analysis demonstrated reproducibility for three samples collected over a week from an individual and incubated as negative controls in test media for 4 h (Fig. 1). Also, the high level of agreement after exposure to each of the tested agents among samples from different subjects was not expected given subject-specific differences in negative control 16S rRNA taxa profiles (Fig. 1). Taxa rRNA abundance differences in response to tested agents were quite similar among different donors (Fig. 2C-F). A fourth person’s saliva biofilm sample tested on three different days showed substantial overlap in the taxa that were more abundant depending on the agent tested compared to control (Table 1, Supplemental Table 1)52,53. This hints at the robustness of the approach of using shed biofilms exposed to various agents to detect common bacterial targets in different people’s mouths.

Selectivity of controls

The antibiotics used as controls in this study, penicillin G and vancomycin, both target cell wall synthesis though they have distinct mechanisms54,55. They target Gram-positive bacteria with the thick cell walls unprotected by a lipid membrane. While vancomycin is a glycopeptide that binds directly to cell wall peptidoglycan to prevent crosslinking and leading to a faulty cell wall, penicillin G is a beta lactam molecule that binds to and inhibits the enzyme that makes the cell wall cross links. Only rarely can either of these agents cross the bacterial outer membrane of some taxa, reducing the effectiveness of these drugs on Gram-negative bacteria. The experiments with the shed biofilm samples showed both drugs selectively produced lower levels of Gram-positive bacteria when taxa analysis was restricted to the genus level with one exception (Table 1). When abundance at the species level was examined, some Gram-negative bacteria were seen to be at lower levels which may or may not be direct or indirect effects of these two antibiotics. This might occur if a Gram-negative and Gram-positive species have a mutualistic relationship. Though notably some Gram-negative taxa that readily uptake these drugs may indeed be directly sensitive54,55.

Effects of agents on 16S rDNA

The focus of this study was on rRNA as it was thought it would more rapidly show changes in levels due to stimulation or inhibition compared to those shown by DNA. However, over 4 h it was certainly possible some bacteria could be induced to proliferate and others could be induced to die, which could result in an increase or decrease in their DNA56. Indeed even over 4 h, differences in total 16S rDNA were detectable over the course of the assay with cranberry and penicillin G exposure compared to the vehicle-treated control (Fig. 4). 16S rDNA levels of vancomycin-treated cells were not different than those of controls; penicillin G treatment showed lower levels of bacteria DNA overall, at 0.59x p < 0.022, while cranberry exposure resulted in 16S rDNA (gene) level at 0.13x p < 0.0005. Given the much lower level of total rRNA and rDNA with exposure to 17% cranberry versus the vehicle-treated shed biofilm, we sought to determine if the deficit was an actual decrease from prior to the incubation or just a block to bacteria DNA production. Indeed, comparisons of before and after 4-hour exposure to 17% cranberry juice showed a statistically significant decrease in 16S rDNA levels to 43%, p < 0.020 compared to the total level of 16S rDNA at assay start (Table 4). When effects on bulk DNA were measured, 17% cranberry treatment induced reductions in DNA levels over the 4-hour incubation (Supp. Figure 1). It was surprising that cell DNA loss would be so rapid based on earlier publications56,57,58. The results highlight the destructive capabilities of the 17% cranberry59.

Effects of agents on 16S rDNA vs. 16S rRNA transcripts

It is well known that in various microbiomes, RNA and DNA levels assigned to specific taxa often do not agree as DNA levels are proportional to the number of bacteria (dead or alive) and RNA is more indicative of the state of the bacteria cell46,47,48,49,60. The rationale of the current study was that the ex vivo assay challenge with various agents would induce changes in the RNA faster than those in DNA, even for rRNA. This would alleviate the need to wait many hours as is routinely done to test bactericidal agents specificity monitored on the DNA level. Still, it was possible that the assay of 16S rRNA showed differences merely reflective of cell proliferation or bacteria loss induced by an agent. In that case, we might expect RNA differences to mimic those that occur with DNA. Tables 2 and 3, which allow a head-on comparison of RNA and DNA differences in the same rRNA genes or rRNA transcripts in the same ex vivo tested samples subjected to vancomycin, penicillin G, or cranberry treatment, revealed minimal overlap. For this experiment, only samples available both as RNA and DNA were included, and to allow batch correction the MaAsLin2 test was used to compare levels of 16S rRNA (Tables 2 and 3). These findings are not directly comparable to those in Fig. 3, which is for a larger number of samples. For this particular study, more differences in specific taxa were observed on the RNA level than DNA level with the three agents tested (Tables 2 and 3). The lack of sensitivity of the DNA study makes it difficult to compare 16S rRNA and rDNA profiles to see how related they are. More work will need to be done to determine if detection in perturbation in rRNA levels is truly more sensitive than DNA detection as a way to determine taxa inhibited by tested agents. Also addressing the latter question, the reduction in total 16S rRNA seen with exposure to all three agents versus vehicle-exposed bacteria seems to indicate a higher level of sensitivity of the 16S rRNA assay than that of the assay for 16S rDNA to agent perturbations. This is some support for the hypothesis that rRNA measurement is superior to measuring rDNA to determine targets of anti-bacterial agents at early timepoints.

Limitations of this work are that head-to-head comparison of RNA and DNA levels from the same samples can be less accurate depending on methods of bacterial lysis and nucleic acid purification used. More work remains to be done to establish how quickly changes in rRNA can be detected in the assay. Ideally, direct comparisons between exposure to a potential bactericidal agent in vivo and exposure of that agent to ex vivo shed biofilms would reveal how representative the assay is of what occurs in vivo. Recently completed in vivo studies in adults and children have demonstrated that rinsing orally with 100% pure cranberry juice inhibited subsequent dental plaque regrowth and metabolic activity including acid production (Christine Wu, manuscript in preparation). This supports the general idea that the ex vivo assay to test cranberry juice would reveal detectable effects on specific taxa. One human in vivo study on cranberry extract effects did not report specific oral taxa differences in the cranberry extract-exposed group in a study of 11 days61. Two in vitro earlier studies showed a reduction in Streptococcus mutans in some oral compartments after 4 or 6 weeks of rinsing with a cranberry extract62,63. This agrees with the reduction of Streptococcus genus in the current work (Fig. 3C). S. mutans were at or near undetectable in all donor samples except one, so effects on that taxa were not adequately tested. No reduction in S. mutans was observed but levels in the caries-free donors that provided samples were low. Another study showed that Actinomyces naeslundii and to a lesser degree S. mutans were inhibited in planktonic culture by cranberry, but not Lactobacillus paracresis15. Sanchez et al. examined effects of short-term exposure of 100% cranberry juice on planktonic and biofilm viability of 6 oral bacteria and were able to show marked inhibitory effects against Veillonella parvula and A. naeslundii16. In the ex vivo assay described here, the species V. parvula and A. naeslundii were queried. Only V. parvula was at lower levels in the cranberry-exposed shed biofilm samples (Fig. 3C). Usage of full-length 16S rRNA amplicon sequencing to identify bacteria would help determine the level of agreement between earlier studies using standard in vitro assay systems and our ex vivo test. It is also not clear how closely the system used here mimics the in vivo situation where people can rinse with 100% cranberry juice. In this study, the ex vivo assay used 17% juice over a 4-hour incubation period which may not produce effects on bacteria similar to what would occur in vivo. More work needs to be done to determine how closely shed biofilm tested anaerobically responds to insults the way it does in vivo.

Conclusion and speculation

The shed biofilm assay was shown to be reproducible in that oral bacteria isolated on different days showed similar 16 S rRNA profiles after the assay (Fig. 1). It also revealed both positive control agents, penicillin G and vancomycin, caused there to be lower steady state levels of almost exclusively Gram-positive genera 16 S rRNA (Fig. 3A, B). This was consistent with their known inhibitory effects on Gram-positives54,55. Diluted cranberry extract in contrast caused reduced 16 S rRNA levels in a range of bacterial taxa versus the control. Comparison of pre- and post-incubation with diluted cranberry extract indicated reduced 16 S rDNA levels suggesting active killing was occurring by that agent. This was a surprising result in that bacterial DNA is thought to be quite stable even in dead cells56,57,58. Streptococcus was reduced after cranberry oral exposure to human in in vivo studies, and it was also reduced in the shed biofilm assay described here62,63. In sum, the shed oral biofilm assay described in this work may allow reproducible assessment of oral taxa susceptibility for an individual.

Materials and methods

Materials: penicillin G was prepared at 12.5 mg/mL in water and vancomycin was dissolved in DMSO at 2.5 mg/mL; final concentrations were 5 µg/mL and 10 µg/mL, respectively. Pure unsweetened cranberry juice (Ocean Spray) was used at 16.7% final concentration.

Ethics approval was by the Institutional Review Board I at the University of Illinois Chicago (IRB protocol #2016-0696). Verbal and written informed consent were obtained by all participants prior to saliva collection. All experiments were performed in accordance with relevant guidelines and regulations.

Sample collection, DNA/RNA extraction, and sequencing

Participants in overall good health without overt systematic disease and not taking antibiotics over the previous three months, were directed to refrain from oral hygiene for 24 h prior to supplying stimulated saliva accumulated from chewing 1 g of non-flavored gum base (Wrigley) for 5 min. Saliva was put on ice, centrifuged at 5500 x G for 5 min at 4 ˚C, then washed with transport medium65, resuspended in PBS and adjusted to OD600nm= 1.0, then centrifuged and resuspended in same volume of modified SHI medium34,36 supplemented with 10% FBS and 0.1% glucose, without sheep blood added. Salivary bacterial suspension (400 µL) was then added into 2 mL polypropylene centrifuge tubes containing test agent (in total 100 µL volume) and incubated anaerobically at 37 ˚C for 4 h in an anaerobic chamber (Forma) filled with anaerobic gas (5% CO2, 10% H2, 85% N2). Water was used as a control and all test samples had end concentrations of 0.40% DMSO including controls with no agent.

Post incubation with various agents under anaerobic conditions; 2 volumes RNA Protect Cell (Qiagen) was added to each tube to stop reactions and preserve nucleic acid. Samples were placed on ice for 30 min. Samples were then centrifuged for 10 min at 5,500x G, and supernatant removed. Pellets were dissolved in guanidium thiocyanate solution (GTC, 4 M guanidinium thiocyanate, 25 mM sodium citrate, pH 7.0), 0.5% (w/v) N-laurosylsarcosine (Sarkosyl), and 0.1 M 2-mercaptoethanol66, or RLT from Qiagen, followed by homogenization with 0.5 and 0.1 mm glass-zirconium beads in the Mini-Bead Beater 3x at 1 min each. Acid phenol extraction was followed by silica-based purification using Zymo Research RNA Clean and Concentrator-5. RNA samples were all exposed to Turbo DNase at 1 unit/mL for 30 min at 37 ˚C prior to repurification. cDNA synthesis was carried out using random sequence hexamers and Superscript III (Invitrogen, Thermo Fisher Scientific) using 1% of each sample harvested from the original bacterial pellets, then subjected to amplicon sequencing similar to DNA samples as described below.

DNA was extracted from a 150 µL aliquot of lysed bacteria in GTC at −80 ˚C. which was left over from mechanical lysis done for RNA extraction. After addition of 2 volumes Bashing Buffer, the Quick-DNA Fungal/Bacteria Miniprep Kit (Zymo Research) was used to purify the DNA. The V4 variable region of bacterial 16 S rRNA genes was amplified using the primer set CS1_515F: ACACTGACGACATGGTTCTACAGTGTGYCAGCMGCCGCGGTAA.

CS2_806R: TACGGTAGCAGAGACTTGGTCTCCGGACTACNVGGGTWTCTAAT followed at the Rush University Genomics and Microbiome Core by a second PCR amplification when sample specific barcodes were added followed by cleanup and sequencing as described earlier67,68. Sequencing was performed on the Illumina MiniSeq2 at 150 cycles (Illumina, Inc, San Diego, CA, USA). Negative controls were samples that started with H2O instead of saliva DNA. Additional controls were technical replicates of several assays.

For bulk RNA and DNA measurements either Nanodrop was used to measure A260 of nucleic acids, or a Qubit was used with Biotium AccuGreen™ High Sensitivity dsDNA and AccuBlue® Broad Range RNA Quantitation Kits. For measurement of bulk 16 S rRNA and rDNA a PCR-based method was used {Nadkarni, 2002 #130}.

Microbial community analysis

For taxa assignment and measurement, forward sequence reads from the FASTQ files were analyzed using the software package QIIME2 (v2023.5)69. Sequences were all at average quality score of 25 without trimming and were used as a size of 151. DADA2-plugin in QIIME2 was used to denoise the sequence and generate feature data and feature tables for the dataset of DNA sequences70. Taxonomy assignment was done by classify-consensus-blast function using the Blast + consensus taxonomy classifier to determine 98% match identity of the query sequences to the Human Oral Microbiome Database (v15.22)71,72. On average there were 52,815 reads per sample with the minimum 17,243 for cDNA amplicon sequencing and average 67,417 reads per samples with the minimum 38,208 reads for amplicon DNA sequencing.

Taxa that had fewer than 2 reads in 80% of samples were eliminated. Alpha and beta diversity were determined on the feature level using MicrobiomeAnalyst71. For the initial data set of shed biofilm samples from three subjects with agent exposure on three separate days, the stored RNA samples were converted to cDNA and then sequenced all at once to avoid batch effects. This allowed the use of DESeq2 to determine differentially abundant 16 S rRNAs73.

Adonis2, which can partition distance matrices over sources of variation, was used to determine the contribution of sample donor and treatment exposure of the samples to the derived beta diversity as Bray Curtis Distance, among all samples, from the first three subjects74. Unlike Adonis, order of variable entry for testing does not affect results.

Statistical analysis was performed using MaAsLin2 software for multivariable analysis of the sample data. The linear model method (LM) was used for the data analysis, with Total Sum Scaling normalization, log transformation, minimum abundance set to 4, minimum prevalence at 20%, and the FDR (q-value) less than 0.142. Categorical variables were summarized using frequencies and percentages. Continuous variables were presented as mean, and standard deviation.

To assess the differences in RNA and DNA levels, the Student t-test was conducted when data followed a normal distribution75. Otherwise, when noted, the Mann Whitney test was used due to non-normal distribution of some data.