Investigating oral microbiome dynamics in chronic kidney disease and post-transplantation in continuous culture

ABSTRACT

The oral microbiome is influenced by environmental factors in chronic kidney disease and following kidney transplantation affecting microbial composition, which may have implications for health and recovery. A major driver of oral microbiome perturbation is the accumulation of urea in saliva. We have modelled increased salivary urea concentrations associated with CKD and subsequent reductions that may occur post-transplantation. Oral microbiota were established in constant-depth film fermenters by inoculation with saliva. Duplicate validation runs were maintained with artificial saliva with baseline urea concentrations (0.205 mg/mL) for 21 days. Triplicate treatment runs were then done with baseline urea for 10 days (healthy phase) before urea was increased for 10 days to reflect CKD concentrations (0.92 mg/mL) (CKD phase). This was followed by reversion to baseline urea concentrations (post-transplant phase). Biofilms in primary validation runs reached dynamic stability within 5 days according to viable counting. DNA sequence data indicated minimal taxonomic variation over time and between low and high urea treatments despite background noise indicating changes in bacteria belonging to the family Gemellaceae and the genera TG5 and Leptotrichia. Significant differences in alpha and beta diversity occurred between low and high urea states but not following reversion to a low urea environment. Increased abundance of the TG5 was detected in late model phases, despite apparent count stability, and independent of changes in urea concentrations.

IMPORTANCE

This study investigates dynamic changes in the oral microbiome associated with changes in salivary urea concentration, an important factor in chronic kidney disease (CKD). The in vitro system modeled increased urea concentrations and subsequent reductions post-transplantation. The study provides insight into the oral microbial shifts during different simulated clinical phases. Understanding these dynamics is crucial for advancing our comprehension of CKD-associated oral microbiome variations and their potential impact on patient well-being and recovery.

INTRODUCTION

The oral microbiota changes compositionally following kidney transplantation (1). Studies have proposed that changes in the oral microbiota are linked to post-operative outcomes, including increased gingival overgrowth and Candida infection (2, 3). Using high-throughput sequencing techniques, two studies have reported increases in the abundance of opportunistic pathogens and decreased community diversity (4, 5). This is of particular potential concern, given the susceptibility of immunocompromised recipients to severe infections, which occur at an incidence of 25.5% within the first-year post-transplantation (6). Furthermore, the oral microbiome has been proposed as a reservoir for extra-oral colonization, which has implications beyond driving severe infections as this could drive inflammation (7) and which may increase the risk of allograft rejection.

Prior to, and during, kidney transplantation, the oral microbiome is subjected to several stressors, which could potentially drive observed changes in diversity and composition. These factors could be those associated with surgery and extended hospital stay. Broadly, chronic kidney disease (CKD), for which kidney transplantation is the gold-standard therapy, is associated with oral microbiome change. Differences in the oral microbiome of CKD patients include enrichment of Neisseria and depletion of Veillonella compared with those without CKD (8). A major CKD-associated factor which might play a role in driving microbial change is the availability of uremic toxins (9). As renal function reduces through the different stages of CKD, the body’s ability to excrete toxins and waste and metabolize vitamin D is also reduced (10–12). As uremic toxins begin to accumulate, they begin to interact with the immune system and potentially drive multiple immune system disturbances seen in uremic end-stage renal disease (ESRD) patients (13).

As kidney function declines, the metabolism of uremic toxins, urea and creatinine, is reduced, and their serum concentration increases (14, 15). This rise in serum concentrations is correlated with an increase in salivary concentrations (16), with CKD patients demonstrating higher salivary creatinine and urea levels than healthy controls (16). The effect of “above normal” levels of salivary urea is understudied in the literature. The resulting increase in localized environmental pH could have cascading effects on the network of microbiota in the oral cavity. Increased pH could have a differential effect on oral bacteria with different pH optima (1, 17–20). For instance, oral nitrate reduction to nitrite has been suggested to be key in systemic cardiovascular regulation (21–23) and is higher at a lower pH (24, 25). Ammonia production from arginine metabolism increases pH in a self-regulatory mechanism (26, 27). Therefore, the complex relationship between the oral microbiome, its metabolism, and host systemic health remains largely uncharacterized in CKD.

The consequences of a more alkaline oral environment could go beyond the differences in microbial composition observed between CKD patients and healthy controls (8). A more basic environment could offer greater protection to enamel against acidification and demineralization (28, 29). The low prevalence of tooth decay in CKD is a long-documented phenomenon (30, 31). An investigation into the properties of the saliva and dental plaque of CKD patients found that plaque pH correlated directly with salivary urea nitrogen [often measured in blood or saliva, an indication of kidney function (32, 33)] concentration. Significantly, the same study also concluded that following a challenge with carbohydrate exposure, the minimum pH of CKD subjects did not reach cariogenic levels (while non-CKD subjects did) (31). Declining kidney function, therefore, might begin a process whereby increased salivary urea concentration drives microbial production of ammonia, which raises environmental pH, alters oral microbial profiles, and has consequences on oral health. Once this system has been established in the CKD host, the removal of chronic kidney disease by kidney transplantation is likely to cause further perturbation. It is unknown whether the return to normal salivary urea concentration causes the reversal of the above process, lowering pH and leaving the host more vulnerable to cariogenesis.

There are numerous concomitant factors which could influence the oral microbiome during kidney transplantation. Studying this effect in human transplant subjects is complicated by confounders, such as perioperative antibiotics, the introduction of immunosuppressive agents, and the effects that variation in oral pH (34, 35) and uremia (13) can have on the immune system. To study this effect in isolation, therefore, an in vitro modeling approach was used in the present study to understand how high salivary urea concentration can affect the composition and diversity of the CKD oral microbiome.

MATERIALS AND METHODS

Growth media

Bacterial isolates were grown on Wilkins Chalgren agar (WCA) (Scientific Laboratory Supplies, UK) and incubated overnight at 37°C under aerobic conditions. To grow liquid cultures, a single colony from a WCA plate was added to 10 mL of Wilkins Chalgren broth and grown aerobically overnight at 37°C with 200 rpm shaking.

Maintenance of oral microcosms

Oral microcosms were maintained in constant-depth film fermenters (CDFFs) as previously described (36–40). As per Ledder and Mistry (36), the temperature was maintained at 36°C by housing the CDFFs within perspex incubation chambers (Stuart Scientific, Redhill, Surrey, UK). The CDFF plugs were formed of polytetrafluoroethylene (PTFE). While hydroxyapatite has been commonly used in the past for plug surfaces in similar studies, the long-term growth of acidogenic plaques could risk dissolution (36); hence, PTFE was preferred. The PTFE plugs were set to a depth of 200 µm with a turntable rotor speed of 3 rpm. For artificial saliva medium, a semidefined McBain medium (pH 7) (41, 42) containing 2.5 g/L mucin, 2 g/L Bacto peptone, 2 g/L Trypticase peptone, 1 g/L yeast extract, 0.35 g/L NaCl, 0.2 g/L KCl, 0.2 g/L CaCl₂, 1 mg/mL hemin, and 2 mg/L vitamin K1 was added to the fermenter at 8 mL/h by a peristaltic pump (Minipuls 3; Gilson). Before inoculation, the PTFE plug surfaces and CDFF were conditioned for 24 hours with McBain medium at 36°C. The fermenters were inoculated with fresh saliva from a single healthy individual donor (age = 28 years) on two separate occasions (4.0 ± 0.5 mL/fermenter/inoculation) 10 hours apart. Sampling was performed every 48 hours; to avoid sampling immature plaques [per Ledder and Mistry (36)], pans were numbered and sampled sequentially. Samples were processed immediately for bacteriological analysis or stored at −80°C before extraction for high-throughput sequencing.

Validation of medium with the addition of urea

To validate the use of McBain medium with the addition of urea, two CDFFs were established with McBain medium with the addition of urea at the concentration found in the saliva of healthy controls by Lasisi et al. (16) (0.205 mg/mL). Previous studies have shown that dental biofilm communities within CDFFs reached dynamic steady states between 2 and 7 days (36, 41), as evidenced by viable count data. To validate the establishment of dynamic stable communities of oral biofilms in McBain medium with additional urea, communities of dental biofilms were sampled every 48 hours from two CDFFs for 21 days post-inoculation.

Oral microcosm modeling of chronic kidney disease and post-transplant states

Following the establishment of a stable population size in oral microcosms with low urea in two “validation” models, three “treatment” models were established in CDFFs to challenge these environments with CKD (high urea) and post-transplant (low urea) states. Based on a study by Lasisi et al. (16), median salivary urea levels of CKD patients (92.00 mg/dL) were 4.5 times higher than healthy controls (20.50 mg/dL). For the three “treatment” models, CDFFs were conditioned with low-urea McBain medium (urea concentration 0.205 mg/mL) at 36°C for 24 hours before inoculation. CDFFs were sampled 24 hours after inoculation and then every 48 hours until day 31. After 10 days post-inoculation (hereafter referred to as the “healthy phase”), the flow of the low-urea McBain medium was stopped, and high-urea McBain medium (0.92 mg/mL) was commenced at the same flow rate (8 mL/h) for 10 days (hereafter referred to as the “CKD phase”). Following 10 days at the CKD phase, the flow of high-urea McBain medium (0.92 mg/mL) was stopped, and drip feeding with fresh low-urea McBain medium was commenced for 10 days (post-transplant phase).

Bacteriological analysis of CDFF samples

Five plugs were derived from each aseptic sampling (at each time point). Four plugs were chosen and split into pairs (at random). Pairs of sample plugs were each placed into separate universal containers (becoming replicates) and each vortexed for 1 min with 2-mm glass beads in 5 mL phosphate-buffered saline (PBS). Appropriate dilutions (0.2 mL) to target the production of plates containing 30–300 colonies were produced, and replicates were then serially diluted in (1.8 mL) PBS and plated in duplicate on selective and non-selective media. Where plates became contaminated or counts were below 30, plates with the highest available colony count were selected. For total aerobe counts, 0.2 mL of these dilutions was plated onto WCA and placed into an aerobic incubator at 37°C for 4 days. For total anaerobe counts, dilutions were plated onto WCA and placed into an anaerobic chamber (Don Whitley Scientific, Shipley, UK), with an atmosphere of 10% H₂, 10% CO₂, and 80% N₂ for 4 days. For total counts of Streptococcus species, dilutions were plated onto the streptococci-specialist (Mitis Salivarius) agar (Merck Life Science, UK). Colony counts were performed after 4 days.

High-throughput sequencing of CDFF samples

After extraction of sample plugs and vortexing with glass beads in PBS, 2 mL of the PBS medium was stored at −80°C for high-throughput sequencing. From the stored CDFF oral microcosm samples, 250 µL was extracted via the DNeasy Power Soil Pro Kit (Qiagen, Manchester, UK), according to the manufacturer’s instructions. Overall, 70 samples from the CDFFs were extracted; for each batch of extractions, a positive control (DNA from a culture isolate in 250 µL PBS) and an extraction negative control (250 µL PBS only) were included. For each batch of PCRs performed, a PCR-positive (a previously tested extraction of Lactobacillus plantarum) and PCR-negative [PCR-grade water (Qiagen, UK)] control were also included. Following extraction, the DNA was amplified by PCR using the 515F (5′-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGGTGYCAGCMGCCGCGGTAA-3′) and 806R (5′-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGGACTACNVGGGTWTCTAAT-3′) primers to amplify the variable region 4 (V4) of the 16S rRNA gene (43–45). Extracted DNA was added at a volume of 2.5 µL, followed by 5 µL of 515F primer, 5 µL of 806R primer, and 12.5 µL of KAPA HiFi HotStart ReadyMix (Roche, London, UK) with a final PCR reaction volume of 25 µL. Following a denaturing step at 95°C for 3 min, PCR was performed with 25 cycles of 95°C for 30 seconds, 55°C for 30 seconds, and 72°C for 30 seconds, followed by an elongation step of 72°C for 5 min. Amplification was confirmed with gel electrophoresis. After positive bands were confirmed (or absent, in the case of negative controls), samples were submitted to Deep Seq Next Generation Sequencing Facility at the University of Nottingham (Nottingham, UK) for sequencing via Illumina MiSeq.

High-throughput sequence data processing

Paired-end sequence data were imported into QIIME2 version 2022.2 (46) via the Casava 1.8 paired-end demultiplexed fastq format. From demultiplexed sequences, before denoising, there were 4,118,143 paired forward and reverse reads (median per sample: 49,148; range: 65–77,216). Since DADA2 is a denoiser stated for use with Illumina-generated data (47), this denoiser was preferred to the Deblur option (48). The amplicon sequence variants (ASVs) generated by DADA2 were aligned using mafft (49), and a phylogeny was constructed using FastTree2 (50). After denoising (including internal quality control and filtering of phiX reads and chimeric sequences), 3,350,577 reads remained with 549 ASVs identified. After rarefaction (subsampling without replacement) to an even depth, alpha and beta diversity metrics were estimated by q2 diversity (46). A rarefaction depth of 29,919 was chosen for diversity analyses [validation run (V) V1: 11, V2: 12; treatment run (T) T1: 16, T2: 14, T3: 15]. Alpha diversity was measured via Shannon’s diversity index (51), Pielou’s evenness (52), Faith’s phylogenetic diversity [PD; (53)], and the observed ASVs (54, 55). Significant differences in alpha diversity metrics between groups were tested using unpaired t-test or one-way analysis of variance (ANOVA) tests (and, where significant, with Tukey’s multiple comparison test) within GraphPad Prism 9 (GraphPad Software, California, US) and plotted using the same software. For beta diversity analysis, four metrics applied to each data set before analysis were Jaccard (56), Bray-Curtis (57), unweighted UniFrac, and weighted UniFrac (58, 59). The clustering of beta diversity metrics between these groups was tested via permutational multivariate ANOVA (PERMANOVA; 999 permutations) within QIIME2 (60), and the principal coordinates analysis (PCoA) plots were generated in Rstudio (61) using the qiime2R (62), phyloseq (63), and tidyverse (64) packages. Taxonomy was assigned using the q2-feature-classifier (65) classify-sklearn Naïve Bayes taxonomic classifier against the Greengenes 13_8 97% OTU reference sequences (66). Stacked taxonomic bar plots were produced using the ggplot2 package (67) in Rstudio (61). Following the recommendations of Nearing et al. (68), three differential abundance testers were selected to look for differentially abundant taxa. Before differential abundance testing, the sample from day 27 of treatment run 3 (T3) was removed due to low ASV count (27) and the presence of only ASVs belonging to V. dispar at 100% relative abundance. Differential abundance testing was performed within QIIME2 via ANCOM (69) and ALDEx2 (70) using the q2-aldex2 plugin (71, 72). The third differential abundance tester was the DESeq2 R package (73), which was chosen due to its high sensitivity in small data sets (74).

RESULTS

Validations runs

To test the effects of urea on the oral microbiota, we used microcosms, seeded from saliva, cultured for a period of weeks, with regular sampling, in CDFFs. We started with runs to validate the use of CDFFs with McBain medium and urea at the level of saliva from healthy individuals before moving on to treatment runs where urea levels were manipulated. Compared to the saliva donor, validation runs did not differ by alpha diversity (P = 0.13) but did by beta diversity (weighted UniFrac) (P = 0.012) (data not shown).

Dynamic stability in validation runs by total colony counts

Validation CDFF models were run in duplicate for 21 days with no change in conditions to determine model stability. Dynamic stability of oral microcosm biofilms was achieved in both validation (V1 and V2) runs around 5 days after inoculation. This is evidenced by colony counts (Fig. 1), which showed the maintenance of stable colony counts for total aerobes, anaerobes, and streptococci from day 5 until day 21 after inoculation. The validation runs established oral microcosm communities with similar mean log₁₀ colony-forming units per milliliter (log CFU/mL) of total aerobes.

Fig 1.

Population dynamics in oral biofilm microcosms drip fed McBain medium with the addition of urea at the median concentration found in the saliva of healthy controls (0.205 mg/mL). Open and closed symbols differentiate duplicates (open symbol, validation V1; closed symbol, validation V2). Data are means ± standard deviation.

Dynamic stability in treatment runs by total colony counts

Colony counts indicated that oral microcosm communities were quickly established in dynamic stability (at a stable population size), and this was maintained for 31 days following inoculation (Fig. 2). For treatment run 1 (T1), the mean log CFU colony count (per milliliter) of total aerobes was 6.93 (range: 5.23–7.87).

Fig 2.

Population dynamics in oral biofilm microcosms drip fed McBain medium with the addition of urea at the median concentration found in the saliva of healthy controls (0.205 mg/mL) for 10 days, followed by urea at the median concentration found in the saliva of chronic kidney disease patients (0.92 mg/mL), followed by a further 11 days at a median concentration of healthy controls. Open, closed, and cross symbols differentiate triplicates (closed symbol, treatment run T1; open symbol, treatment run T2; cross symbol, treatment run T3). Data are means ± standard deviation (n = 3 model runs).

Alpha diversity of validation runs in early (days 1–10) and late (days 11–21) phase

The difference in alpha diversity between the first 10 days post-inoculation and the last 11 days post-inoculation in validation runs was measured using four metrics (Fig. 3). Comparing these groups via t-test, the “late” phase of the CDFF run was found to have higher alpha diversity via all four metrics.

Fig 3.

The alpha diversity of 68 samples taken from 2 CDFF validation runs (V1 and V2), comparing the early (days 1–10) and late (days 11–21) phases via four metrics (Shannon’s diversity index, observed ASVs, Faith’s phylogenetic diversity, and Pielou’s evenness). The significance of statistical (t-test) comparisons between groups is indicated above brackets.

Beta diversity of validation runs in early (days 1–10) and late (days 11–12) phase

Beta diversity was calculated using principal coordinates analysis with four metrics Jaccard, Bray-Curtis, unweighted UniFrac, and weighted UniFrac. Plotting these data (Fig. 4) revealed differential clustering patterns in all four metrics between samples taken from the early and late phases. This was confirmed statistically by permutational multivariate ANOVA, which found significant differences between early and late phases using all four metrics (P = 0.001). In all four cases, the separation between early and late is largely along the first principal component and that the late samples have a much narrower distribution on that axis than the early samples, suggesting a degree of stabilization over time.

Fig 4.

Beta diversity analysis of samples from validation CDFF runs (V1, circle symbols; V2, triangle symbols) measured via four metrics (Jaccard, Bray-Curtis, unweighted UniFrac, and weighted UniFrac) for early (days 1–10) and late (11–21) phases, shown in red and blue, respectively. Significance, as detected via PERMANOVA testing, is indicated in the bottom right of each.

Relative abundance of genera for validation runs (V1 and V2)

The two validation runs developed established microbiome profiles with Streptococcus, Peptostreptococcus, Prevotella, and Neisseria among the genera above 3% abundance in phases in both runs (Fig. 5). The 10 genera with the highest relative abundances (of the 2 runs combined) are shown for each phase in Table 1 and include Neisseria, Streptococcus, Peptostreptococcus, Fusobacterium, Selenomonas, Parvimonas, Prevotella, TG5, Leptotrichia, and Oribacterium. In the late phase, a gradual increase in TG5 is apparent in both runs beginning at days 13 (V1) and 15 (V2) (respectively). A decrease in the relative abundance of Neisseria and Streptococcus is also apparent in the late phase.

Fig 5.

(Top) Microbiome profiles of oral microcosm samples collected from CDFF validation runs (V1 and V2). The stacked bar chart shows the mean proportion of genera (relative abundance) in each sample from each validation run. (Bottom) Mean relative abundance of genera from each validation run in early (days 1–10) and late (days 11–12) phase. Each genus is represented by a different color (legend below). Genera found at relative abundance below 3% are grouped together (gray).

TABLE 1.

The 10 defined genera with the highest combined (model runs) relative abundance^a

Genera	Days 1–10 (%)	Days 11–21 (%)
Neisseria	26.49	11.98
Streptococcus	15.12	3.09
Peptostreptococcus	8.69	8.23
Fusobacterium	7.27	8.00
Selenomonas	2.53	7.50
Parvimonas	3.69	5.55
Prevotella	3.37	5.12
TG5	0.18	7.23
Leptotrichia	2.92	1.53
Oribacterium	1.83	2.61

^a Relative abundance of each genus in the early (days 1–10) and late (days 11–21) phases is shown.

Treatment runs

To test the effects of kidney transplantation on the oral microbiota, CDFF runs were in three consecutive phases: healthy (low urea), CKD (high urea), and post-transplant (low urea), with each phase lasting 10 days.

Alpha diversity of treatment runs in healthy, CKD, and post-transplant phases

The difference in alpha diversity between the three phases of feeding with (low or high urea) artificial saliva “treatment” runs was measured using four metrics (Fig. 6). Comparing these groups via one-way ANOVA, no differences were found between phases using Shannon’s diversity index (P = 0.4641), observed ASVs (P = 0.0711), or Pielou’s evenness (P = 0.7551). A one-way ANOVA, performed on Faith’s phylogenetic diversity metric, indicated there was a statistically significant difference between at least two groups (P = 0.0039). Tukey’s multiple comparisons test found that the mean Faith’s PD alpha diversity for samples taken from the healthy phase (6.62 ± 1.45) was significantly lower than samples from the CKD (7.92 ± 1.23) phase (P adj. = 0.0219) and the post-transplant phase (8.076 ± 0.87), P adj. = 0.0052. No significant difference was found between the mean alpha diversity of the CKD phase (7.922) and the post-transplant phase using the Faith’s PD metric, P adj. = 0.9394.

Fig 6.

The alpha diversity of samples taken from the CDFF “treatment” runs (T1, T2, and T3), comparing the “healthy” (days 1–10), CKD (days 11–20), and “post-transplant” (days 21–31) phases via four metrics (Shannon’s diversity index, observed ASVs, Faith’s phylogenetic diversity, and Pielou’s evenness). The significance of statistical comparisons between groups is indicated above brackets (where P < 0.05).

Beta diversity of treatment runs in healthy, CKD, and post-transplant phases

Beta diversity was measured using the unweighted UniFrac, weighted UniFrac, Jaccard, and Bray-Curtis distance metrics. Following principal coordinates analysis, plotting (Fig. 7) showed clear distinct clustering based upon treatment run (shapes) as well as clustering based upon treatment phase (colors). Different clustering patterns based upon the treatment phase were confirmed statistically via permutational multivariate ANOVA tests (999 permutations), which found an overall significant difference between phases using every metric tested (unweighted UniFrac P = 0.003, weighted UniFrac P = 0.017, Jaccard P = 0.003, and Bray-Curtis P = 0.009).

Fig 7.

Beta diversity analysis of samples from treatment CDFF runs (T1, circle symbols; T2, triangle symbols; T3, square symbols) measured via four metrics (Jaccard, Bray-Curtis, unweighted UniFrac, and weighted UniFrac) for three phases of artificial saliva with urea treatment (healthy, CKD, and post-transplant) in red, green, and blue (respectively). Significance (P) relative to overall PERMANOVA testing between phases is indicated for each model.

Pairwise post hoc statistical comparisons found differences in beta diversity between healthy and CKD phases using three out of four metrics (unweighted UniFrac P adj. = 0.036, Jaccard P adj. = 0.033, and Bray-Curtis P adj. = 0.050). Comparing the healthy phase with the post-transplant phase, post hoc analysis revealed differences using all four metrics (unweighted UniFrac P adj. = 0.006, weighted UniFrac P adj. = 0.015, Jaccard P adj. = 0.006, and Bray-Curtis P adj. = 0.003). Comparisons of the CKD and post-transplant phases showed no significant differences using any beta diversity metric.

Relative abundance of genera in treatment runs (T1, T2, and T3)

The three treatment runs established microbiome profiles with Streptococcus, Neisseria, Veillonella, and Haemophilus among the genera above 3% abundance in each run (Fig. 8). The 10 genera with the highest (sum) relative abundance are shown for each phase in Table 2 and include Parvimonas, Sphingobacterium, Fusobacterium, Neisseria, TG5, Serratia, Streptococcus, Selenomonas, Prevotella, and Peptostreptococcus. In treatment runs T1 and T2, gradual increases in TG5 are apparent from days 17 and 13 (respectively) (Fig. 8), and this is reflected in the overall abundance in each phase shown in Table 2. In treatment run T3, a gradual increase in Sphingobacterium is apparent (Fig. 8) and reflected in Table 2.

Fig 8.

(Top) Microbiome profiles of oral microcosm samples collected from CDFF treatment runs (T1, T2, and T3). The stacked bar chart shows the mean proportion of genera (relative abundance) in each sample from each treatment run through healthy (days 1–10), CKD (days 11–20, between dashed lines), and post-transplant (days 21–31) phases. (Bottom) Mean relative abundance of genera from each treatment run in healthy (days 1–10), CKD (days 11–20), and post-transplant (days 21–31) phases. Each genus is represented by a different color (legend below). Genera found at relative abundance below 3% are grouped together (gray). Prevotella in square brackets denotes recommended, but not verified, taxonomies in the Greengenes database.

TABLE 2.

The 10 defined genera with the highest combined relative abundance in treatment runs^a

Genera	Healthy (%)	CKD (%)	Post-transplant (%)
Parvimonas	5.45	8.44	11.09
Sphingobacterium	3.63	9.86	10.70
Fusobacterium	7.43	5.81	6.98
Neisseria	15.10	1.32	1.61
TG5	0.18	4.03	9.91
Serratia	3.07	7.27	3.13
Streptococcus	5.53	2.28	3.73
Selenomonas	2.80	5.61	3.01
Prevotella	3.78	3.01	2.94
Peptostreptococcus	2.66	3.59	3.47

^a Relative abundance of each genus in the healthy (days 1–10), CDK (days 11–20), and post-transplant (days 21–31) phases is shown.

DISCUSSION

Kidney failure progresses through CKD toward ESRD. In doing so, kidney function shows a continual decline and a reduced capacity to metabolize uremic toxins, such as creatinine and urea (14). Ultimately, renal replacement therapy in the form of dialysis or transplantation is required to preserve life. The inability to process uremic toxins causes their accumulation in the serum and saliva (14). In the saliva, median urea concentrations of CKD patients can rise up to 4.5 times higher than healthy controls (16). The oral microbiome is detectably different following kidney transplantation (1). Evidence from microbiome studies has suggested an increase in opportunistic pathogens following transplantation (2–5). Yet, these changes in oral bacterial communities have not been well studied. To investigate this, the present study used in vitro models of microcosm communities to study the response of oral bacteria to CKD and renal transplant.

The two (validation) runs, which were drip-fed McBain medium with the urea concentration of the saliva in healthy individuals (16) for 21 days, reached dynamic stability (measured by the congruence of viable counts) within 5 days, similar to previous studies (39, 40, 75, 76). Few examples exist that have employed the (highly sensitive) high-throughput sequencing method in CDFFs (75). The present study suggests that using this more sensitive technique, measurable changes might still occur (in the absence of a change in experimental conditions) beyond stable population sizes according to viable counts. Alpha and beta diversity showed significantly different patterns in early and late phases of validation runs, and trends, such as increasing relative abundance of genus TG5 and decreasing abundance of Neisseria, were apparent until the runs were terminated.

Three treatment runs were also performed to model the successive urea concentrations that the oral microbiome of a kidney transplant recipient might be exposed to. The healthy phase differed from both the CKD and post-transplant phases in terms of (Faith’s) alpha diversity and [(both): unweighted UniFrac, Jaccard, and Bray; (post-transplant only): weighted UniFrac] beta diversity metrics. Taxa found in (consensus) differential abundance between these phases included V. dispar, which has previously been found to outperform other oral species in CDFFs where urea has been removed after establishment (77) and TG5, which detectably increased in abundance in the absence of urea concentration change (Validations runs), and might represent ongoing ecological succession occurring independent of urea concentration change. Limited differences were observed in the comparison of the CKD and post-transplant model phases. No difference was observed in alpha or beta diversity, and only one consensus ASV was detected in differential abundance.

The limited differences observed in the oral microcosms during the treatment phases could reflect the fact that other factors drive the changes observed in patient studies or that the observation period for these models (11 days “post-transplant”) was too short. Future studies using CDFFs to model the effect of the onset and resolution of CKD on the oral microbiome could include replicate CDFFs inoculated with the same communities and apply variations in urea concentration in parallel. Such studies can establish more consistent baseline microbial population dynamics, including underlying growth rates and microbial interactions, facilitating a better understanding of oral microbiome changes that occur after transplantation (5).

ETHICS APPROVAL

Samples were collected following approval from the University of Manchester Research Ethics Committee (Reference: 2019-5929-9344).

DATA AVAILABILITY

The data presented in the study are deposited in the NCBI SRA repository, accession number PRJNA1141113.