Exploration of severe early childhood caries microbiota through a novel developed nutrient enriched microbiological medium, high through-put 16S rRNA sequencing and culturomics

Subjects and sample collection

Dental plaque samples were collected from eleven patients with S-ECC. Three boys and six girls, aged from 2.53 to 5.33 years old, were diagnosed S-ECC with decayed, missing, and filled tooth surface (dmfs) scores were 14.11  ± 3.59 (Table S1). Participants were required to avoid eating, drinking and brushing their teeth at 2 h before dental plaque sampling. Ethical approval of the study was granted by Peking University School of Stomatology (PKUSSIRB-201839140), and written informed consent was obtained from all study participants’ guardians.

Eleven plaque samples were collected with a sterile curette from the labial/buccal smooth surfaces or the decayed cavity of the affected teeth. The dental plaque was transferred to a 1.5 mL sterile centrifuge tube full of pre-reduced medium (0.15 g sodium mercaptoacetate, 0.9 mL fresh 1% calcium chloride, and 100 mL distilled water, filtered for removing bacteria) and taken to the laboratory and processed for culture within 3 min. Samples were dispersed by vortexing for 30 s and diluted tenfold consecutively from 10⁻¹ to 10⁻⁸.

Culture media

All samples were grown in duplicate at 37 °C under anaerobic conditions (80% N₂, 10% H₂, and 10% CO₂) and normoxic conditions (air with 5% CO₂). To mimic the nutritional environment of the dental plaque, a novel culture medium was designed. The basal liquid medium consisted of brain-heart infusion broth (OXOID, Basingstoke, UK) 24.5 g/L, agar (Solarbio, Beijing, China) 15 g/L, dextrose (Solarbio) 5 g/L, corn starch (Solarbio) 1.5 g/L, casein acid hydrolysate (Macklin, Shanghai, China) 17.5 g/L, hemin (Solarbio) 5 mg/L, VitK (Solarbio) 1 mg/L, VitB12 (Macklin) 100 µg/L, L-glutamine (Macklin) 0.1 g/L, adenine (Macklin) 10 mg/L, guanine hydrochloride (Macklin) 300 µg/L, p-aminobenzoic acid (Macklin) 130 µg/L, nicotinamide adenine dinucleotide (Macklin) 2.5 mg/L, thiamine pyrophosphate (Macklin) 1 mg/L, ferric nitrate (Macklin) 200 µg/L, cysteine hydrochloride (Solarbio) 0.259 g/L, L-cysteine (Solarbio) 11 mg/L, and sheep blood (Pingrui Biotechnology company) 5%. SHI medium was prepared as described before (Tian et al., 2010).

Dental plaque culture

Samples were enriched aerobically and anaerobically using two liquid media SHI and NEMM designed by this study and subcultured in solid media after optimal dilution, whose single colonies numbers range from 100 to 500. The cultures were divided into four groups: (1) aNEMM (samples grown in NEMM anaerobically); (2) aSHI (samples grown in SHI medium anaerobically); (3) nNEMM (samples grown in NEMM under normoxic condition); and (4) nSHI (samples grown in SHI medium under normoxic condition). Consecutive tenfold dilutions of the plaque samples in two media were prepared by PBS and used to inoculate NEMM and SHI-agar plates in triplicate. Plates were incubated and photographed at 0, 1, 3, 6, 9, 15, 21 and 28 days. Meanwhile, 500 µL suspension cultures at these different time points were collected for 16S rRNA gene sequencing analysis and bacterial isolation. For clinical isolates, suspension cultures were gradient diluted from 10⁻¹ to 10⁻⁸, and then the diluted bacterial solutions were plated on the NEMM- and SHI-agar plates. The clinical isolates were picked up from the plates according to morphology and then identified by sequencing the full length of the 16S rRNA gene as described below.

High through-put 16S rRNA gene sequencing

Nine clinical dental plaque samples and their cultures in NEMM and SHI media under anaerobic and normoxic conditions at eight indicated time points (9 clinical samples + 144 SHI cultures +144 NEMM cultures) were respectively subjected to high through-put 16S rRNA gene sequencing at Biomarker Technologies (Biomarker Technologies, Beijing, China) where the total DNA was isolated, amplified and sequenced according to their standard procedures. In brief, microbial DNA was extracted from the bacterial suspensions according to the manufacturer’s protocols. DNA concentration was assessed by a Nanodrop (Thermo Scientific) and quality was determined by agarose gel electrophoresis. PCR was used to amplify the variable regions 4 and 5 (V4-V5) of bacterial 16S rRNA gene by using 515F (ACTCCTACGGGAGGCAGCA) and 907R (GGACTACHVGGGTWTCTAAT, H = A/C/T, V = A/C/G, W = A/T) PCR primers. The PCR procedure was set as previously published research (Li et al., 2017). Purified amplicons were pooled in equimolar solutions and paired-end sequenced (2 × 300) on an Illumina MiSeq platform. After passing data quality check, the original high-throughput sequencing data were converted into FASTQ format files for bioinformatics analysis. The high-quality tag data were obtained after filtering and compared with sequences in the SILVA database (http://www.arb-silva.de).

Bioinformatics analysis

The bioinformatics analysis was conducted using QIIME2 2020.6 (Fig. S1). The alpha diversity indices of Chao1, ACE, Shannon, Simpson and phylogenetic diversity (PD whole tree) were calculated using Mothur software (version v.1.30) with coverage over 99%. Beta diversity analysis was performed by non-metric multi-dimensional scaling (NMDS) analysis based on the weighted unifrac and unweighted unifrac algorithms. Linear discriminant analysis (LDA) effect size (LEfSe) was conducted to define the biomarkers of these groups. The threshold on the logarithmic LDA score for the distinguishing features was set to 4.0. Microbial functions were predicted using PICRUSt2 (v1.0.0) software following the online protocol and aligned to the Clusters of Orthologous Groups of proteins (COG) database. The profiles of Kyoto Encyclopedia of Genes and Genomes (KEGG) levels were generated based on the KEGG database. Analysis of Variance (ANOVA) is mainly used to test whether the mean of multiple samples is equal, so as to judge the significant difference of the average of two or more samples. Level of statistical significance (P value) was set as <0.05 for all comparisons. The feature sequences of Top 80 abundance ratio at OTU/ASV taxonomic level were selected, multiple sequence comparison was performed with QIIME command line and a taxa tree was constructed. Then the R package was used to combine the taxa tree and species taxonomic abundance data.

Species identification using full-length 16S rRNA gene amplicon sequence

A total of 1,344 single colonies from NEMM-agar plates and 1,312 single colonies from SHI-agar plates were picked up and inoculated into NEMM and SHI liquid media, cultured in aerobically or anaerobically, respectively. DNA from these clinical isolates was extracted using hexadecyl trimethyl ammonium bromide (CTAB) methods (Zheng, Gao & Deng, 2012). The full-length of 16S rRNA gene was amplified via PCR reaction. 27F (AGAGTTTGATCCTGGCTCAG) and 1492R (GGTTACCTTGTTACGACTT) served as forward and reverse primers, respectively. PCR reaction conditions were set as follows: predenatured at 94 °C for 5 min, 35 cycles, including denaturation at 94 °C for 30s, annealing at 54 °C for at 30s, elongation 72 °C for 90s, and extended at 72 °C for 10 min. The full-length of 16S rRNA gene amplicons were sequenced using the Sanger method. The sequences of 16S rRNA gene were blasted in either NCBI (https://blast.ncbi.nlm.nih.gov/Blast.cgi) or EzBioCloud (https://www.ezbiocloud.net/) to establish taxonomic relationships based on currently recognized prokaryotic reference species. Sequences with homology below 98.65% were defined as potential new strains (Kim et al., 2014).

Statistical analysis

The t-test was performed on the species abundance data between the groups using Metastats software. Principal component analysis was used to analyze the composition of the microbiome. BugBase was used to evaluate microbial phenotype followed by the Mann–Whitney U test. The species leading to the composition difference between the two groups of samples were determined by P value or Q value (correcting P value). The level of statistical significance (P value or Q value) was set as <0.05.

Comparison of bacterial community between S-ECC clinical dental plaque samples and their cultured bacteria

To investigate the difference in bacterial community of S-ECC among the original dental plaque, and cultured bacteria, we recruited nine S-ECC patients, half of the samples were used for culture and the other half for 16S rRNA gene high-throughput sequencing directly without culture. Alpha diversity indices of ACE, Chao1, Shannon and Simpson showed clinical dental plaque samples had greater alpha diversity than either all their cultured counterparts (Fig. 1A) or the cultured bacteria under normoxia and anaerobic conditions (Fig. 1B), although the Venn diagram showed that the OTU numbers of cultured bacteria at species level under anaerobic and normoxia conditions were higher than that in clinical samples (Fig. 2A). Core flora of clinical S-ECC samples had 14 common features, including two unclassified Streptococcus, one unclassified Granulicatella, two unclassified Veillonella, one unclassified Fusobacterium, one unclassified Aggregatibacter, one unclassified Kingella, one unclassified Lachnoanaerobaculum, one unclassified Lactobacillales, one unclassified Abiotrophia, one unclassified Rothia, one Rothia aeria, and one Capnocytophaga endodontalis (Fig. S2).

To visualize the distribution of samples in coordinate system, NMDS analysis showed that anaerobic cultured bacteria and normoxia cultured bacteria samples had similar culture features. However, both had a markedly different bacterial structure in comparison with clinical sample group (Fig. 2B).

To display the composition of dominant bacteria, the stacked column plot showed the percentage of each high-abundance bacteria. In clinical samples, except Veillonella, Leptotrichia and Streptococcus, other bacteria occupied the most of bacterial community. No matter under anaerobic and normoxia conditions, the percentages of other bacteria, Veillonella and Leptotrichia were clearly decreased. Meanwhile, there were observed increases in Desemzia, Actietobacter, Lactobacillus, Limosilactobacillus, Ligilactobacillus, Vagococcus and Streptococcus (Fig. 3A). The results indicated that there were more bacterial categories in clinical samples than cultures by media. NEMM and SHI media only could cultivate part of clinical bacteria in S-ECC plaques. There is still a lot of room for improvement in the culture medium.

To detect the difference among the groups, ANOVA was used to display the relative difference of bacteria at the genus level. The results showed that both anaerobic and normoxia cultures were beneficial for the growth of Streptococcus, Bacillus (including other unclassified Bacillus) and Bifidobacterium. The relative higher abundance of most other bacteria, including Leptotrichia, Veillonella, Corynebacterium, Actinomyces, Fusobacterium, Neisseria, Campocytophaga, Prevotella, Selenomonas, Campylobacter and Aggregatibacter in clinical samples presented marked lower percentage in both anaerobic and normoxia cultures. It suggested that most of the above-mentioned bacteria categories could not be successfully enriched by NEMM and SHI media (Fig. 3B).

To assess the predicted function of the bacterial community in clinical samples and the cultured bacteria samples, microbial phenotype prediction was evaluated by BugBase followed by the Mann–Whitney U test. The results showed that there were decreases in Gram-positive bacteria, facultatively anaerobic bacteria and stress tolerance ability, but increases in Gram-negative bacteria, anaerobic respiration, biofilms forming and potential pathogenicity in clinical samples compared with cultured samples under aerobic and anaerobic conditions (Fig. 4). The results indicated that uncultured Gram-negative, anaerobic, potentially pathogenic and biofilm-forming bacteria may contribute to S-ECC development.

NEMM cultivates more abundant bacteria species compared to SHI medium

All cultured microbiota were examined in terms of both species proportion and abundance, with a specific focus on the enrichment of species. Of the organisms presented in the inoculum, all 32 phyla, 87 classes, 260 orders, 501 families, 1,107 genera and 1,502 bacterial species were captured. We identified 3,719, 5,605, 3,705, and 3,460 OTUs in the aNEMM, nNEMM, nSHI and aSHI groups, respectively. Among them, 608 OTUs were uniform, occupying 4.9% of all the OTUs detected, indicating that a steady facultative anaerobe microorganism composition was found in all four fluid media. The other OTUs were not shared in all groups. Among these variable microbiomes, 2,061 OTUs were unique to the aNEMM group, while 1,848 OTUs could be detected only in aSHI. Moreover, 4,010 OTUs could be found only in nNEMM groups, while 2,096 OTUs could be found only in nSHI groups.

To compare NEMM with SHI, which are suitable for cultivating specific bacterial types, we further analyzed 16S rRNA gene high-throughput sequencing data. Alpha diversity indices of ACE and Chao1 showed no significant difference between NEMM and SHI under the normoxia condition. However, under anaerobic culture conditions, NEMM had greater alpha diversity indices of ACE and Chao1 than SHI (Fig. 5A). NMDS analysis showed that regardless of whether anaerobic or normoxia, cultured bacteria samples had similar features (Fig. 5B). Venn diagram showed that the OTU numbers of NEMM cultured bacteria under anaerobic and normoxia conditions were higher than that in SHI cultured bacteria under the same conditions (Fig. 5C).

Differential predominant bacteria between NEMM and SHI were analyzed in the relative abundance of bacteria at the genus level by ANOVA. Of the 19 most predominant bacterial genera found in NEMM and SHI media, Ligilactobacillus, Limosilactobacillus, Lactobacillus, and Desemzia showed the highest abundance levels. Based on the relative abundance of cultured bacteria, NEMM was suitable for cultivating bacteria including Lactobacillus, Desemzia, Veillonella, Psychrobacterium and Leptotrichia, while SHI was good for culturing Ligilactobacillus, Limosilactobacillus, Bacillus, Neisseria, Leuconostoc, Cryptobacterium, Aggregatibacter and Kingella (Fig. 6).

Bugbase was used to predict the function of bacterial community in NEMM and SHI under anaerobic and normoxia culture conditions. The results showed that the microbiota developed in NEMM were predicted stronger anaerobic ability, whether under anaerobic conditions or normoxic conditions. Under normoxic culture conditions, there were no significant differences in Gram-positive bacteria among the four groups (aNEMM, aSHI, nNEMM and nSHI). However, under anaerobic conditions, NEMM supported the growth of more Gram-negative bacteria and less Gram-positive bacteria. Dental plaque samples cultured in NEMM were more likely to be stress tolerant and less likely to be potential pathogens and form biofilms. Moreover, it had weaker facultatively anaerobic ability when under anaerobic conditions (Fig. 7).

To assess the effects of the two media on the predicted gene categories (COGs), we compared the predicted COGs between aNEMM and aSHI, nNEMM and nSHI (Fig. 8A). NEMM showed more ability in carbohydrate transport and metabolism than SHI medium under both anaerobic conditions and normoxia conditions. Additionally, those in NEMM displayed an increased relative abundance of protein functions linked to defense metabolism, inorganic ion transport and metabolism under anaerobic conditions. Furthermore, pathways of membrane transport and cellular community-prokaryotes were associated with aNEMM group (Fig. 8B).

NEMM and SHI media were suitable for culturing specific bacteria respectively

The predominant bacteria were largely consistent among the four groups (aNEMM, aSHI, nNEMM, nSHI), but different relative abundances could be observed (Fig. S3). More genera, species and uncultured bacteria were captured in aNEMM groups at day 1 (Table S2) under anaerobical condition compared with those in aSHI Day1 group. For normoxia conditions, there was a significant difference of genera and uncultured bacteria between NEMM and SHI media (Table 1). Table S3 showed that genera, species and uncultured bacteria detected in nNEMM were higher compared to nSHI at each time point.

Principal component analysis of the development of the oral microbiome shows distinct shifts in the composition of the microbiome. We could see the changes in bacterial community structure over time in Figs. 9A and 9B. The proportions of Streptococcus were reduced along with time in the four groups. In this study, the NEMM as a growth and colonization substrate resulted in a dynamic microbial succession, characterized by an initial proportional reduction in Veillonella, corresponding to an increase of Lactobacillus and unclassified_Bacilli proportions under anaerobic conditions. Bifidobacterium reached the highest percentage on the 15th day and then decreased in aNEMM group. In the aSHI group, the proportions of Lactobacillus and unclassified_Bacilli increased gradually before the 15th day. While Ligilactobacillus increased rapidly from day one to day three in both aSHI and nSHI groups. In nNEMM group, Ligilactobacillus, unclassified_Bacilli, Lactobacillus, Lacticaseibacillus and Lactiplantibacillus showed the time-dependent accumulation, while Acinetobacter exhibited the opposite tendency.

The taxa that most likely explains the differences between NEMM and SHI media from different time points are defined by LEfSe. Figures 10A and 10B showed cladograms representing the potential biomarkers of different groups. At the genus level, Veillonella was significantly enrichment in the aNEMM cultured Day1 group (aNEMM-Day1), while Streptococcus, Gardiobacterium and Granulicatella exhibited relatively higher abundance in the aSHI-Day1 group. Bifidobacterium, Akkermansia and Desemzia were significantly more abundant in the aNEMM-Day15, aNEMM-Day21 and aNEMM-Day28 groups, respectively. While Ligilactobacillus, unclassified_Bacilli and Lacticaseibacillus were most abundant in aSHI-Day6, aSHI-Day21 and aSHI-Day28 groups, respectively. For samples grown aerobically, significant bacterial differences were also detected, with Veillonella, Desemzia and Lacticaseibacillus were remarkably enriched in the nNEMM-Day1, nNEMM-Day9 and nNEMM-Day28 groups at the genus level, while Streptococcus exhibited higher abundance in the nSHI-Day1 group (Figs. 11A and 11B, LDA > 4, P < 0.05).

To study the difference in microbial community abundance between two groups of samples, Metastats software was used to conduct t-test on genus richness data between groups (Fig. 12). At the genus level, Actinobacillus, Pseudoxanthomonas were significantly enriched in aSHI medium while Bilophila was significantly enriched in aNEMM medium (Fig. 12A). Unclassified Clade III, NS3a marine group and unclassified Saprospiraceae were more abundant in nNEMM suspension (Fig. 12B). The same tendency can also be seen in Fig. S4. From the taxa tree, aNEMM was suitable for culturing Anaerolinea, Aerococcus, Lactobacillus and Negativicutes. Planococcus, Desemzia and Psychrobacter were enriched in NEMM no matter under normoxia or anaerobic condition. Geobacillus, Lactococcus, Clostridiaceae, Hyphomicrobium, Pseudolabrys, Paracoccus, Rhodobacter, Rickettsiaceae, Sphingobium, Legionella, Moraxella, Luteimonas and Lysobacter were suitable for growth in nNEMM. Bacillus, Pediococcus, Brevundimonas, Caulobacter, Phenylobacterium, Sphingomonas and Pseudoxanthomonas were suitable for growing in aSHI. Blastomonas were enriched in nSHI. Abiotrophia grew well in SHI both under normoxia and anaerobic conditions (Fig. S4).

Microbiota developed in NEMM and SHI media, under anaerobic and normoxia conditions exhibited distinct community structures. According to the relative abundance ratios of genera shown on taxa tree (Fig. S4), we constructed Table 2 for clearly showing the different microbial profiles cultured in NEMM medium and SHI medium. At the genus level, Anaerolinea, Planococcaceae, Acerococcus, Desemzia, Sphongopyxia, Psychrobacter and some not assigned species were significantly enriched in aNEMM. Planococcaceae, Geobacillus, Facklamia, Desemzia, Lactococcus, Blastomonas, Sphongopyxia, Legionella, Psychrobacter, Luteimonas, Lysobacter and Pseudoxanthomonas were more abundant in nNEMM conditions. The aSHI condition was suitable for Pseudograclibacillus, Sphingomonas and Pseudoxanthomonas, while Lactococcus, Blastomonas, Sphingomonas, Legionella, Luteimonas, Lysobacter and Pseudoxanthomonas were more abundant in nSHI condition.

NEMM helps to culture previously uncultivated oral bacteria

NEMM supported the growth of many oral species that have not been cultured from S-ECC clinical samples so far. All 128 previously uncultivated species were detected by 16S rRNA gene sequencing using SILVA as the reference database.

To study the difference in microbial community abundance between groups of samples, Metastats analysis was used to conduct t-test on species richness data between groups. Uncultured bacteria including methanogenic_archaeon, Truepera_sp., Acidobacterium_sp., Desulfovibrionaceae_bacterium and Nitrosomonadaceae_bacterium could only be detected in aNEMM group. Comparing relative abundance of these uncultured species between aNEMM and aSHI media, uncultured bacteria including forest_soil_bacterium, Clostridiales_bacterium, Desulfuromonadales_bacterium, Firmicutes_bacterium and Alphaproteobacteria_bacterium were observed higher in aNEMM group (Table S4). Discrepancies were also detected between the nNEMM and nSHI groups, with higher abundances of uncultured bacteria including Desulfuromonadales_bacterium, Bacteroidetes_bacterium, Clostridium_sp., Mollicutes_bacterium, Bacteroidales_bacterium, Holophaga_sp. and Acidobacteria_bacterium in the nNEMM group. Moreover, uncultured bacteria including methanogenic_archaeon, Sphingobacteriales_bacterium, Aminicenantes_bacterium, marine_bacterium, forest_soil_bacterium, Acidobacterium_sp., Xanthomonadaceae_bacterium, Sorangiineae_bacterium, marine_microorganism and candidate_division_SBR1093_bacterium were only observed in the nNEMM group (P < 0.05) (Table S5).

The serial diluted culturomics samples derived from above broth medium conditions were inoculated into the two media-agar plates and incubated under normoxia and anaerobic, respectively, for 1-5 days (Fig. S5). The bacterial diversity gradually decreased over time. With the extension of culture time, the difference of dominant bacteria in the two media gradually decreased, and from the morphological point of view, SHI and NEMM agar plates tended to cultivate similar bacterial diversity. After this process, the slow-growing colonies were subcultured and identified to isolate previously uncultivated phylotypes. Through culturomics, uncultivated bacteria belonging to the Prevotella, Actinomyces, Pseudomonas, Granulicatella, Lactobacillus, Campylobacter, Streptococcus and Bacillus genera were identified in the biofilm growth media. The oral taxa identified under the four conditions were shown in Fig. S6, displaying that both NEMM and SHI media could culture the uncultivated bacterium in aerobic and anaerobic conditions.

Also, it turned out that Achromobacter spanius, Lactobacillus delbrueckii subsp and Actinomyces naeslundii in aNEMM, Pseudomonas sp. and Veillonella parvula in aSHI, Brevibacillus agri in nNEMM, Leuconostoc lactis, Chryseobacterium indologenes, Cardiobacterium hominis and Kingella bonacorsii in nSHI may tend to be enriched in each growth condition.