Gene expression profile of Scardovia spp. in the metatranscriptome of root caries

: A few investigations of caries biofilms have identified Scardovia spp.; however, little is known about its involvement in caries pathogenesis. The purpose of this study was to assess the gene expression profile of Scardovia spp. in root caries, and compare it with other microorganisms. Clinical samples from active root caries lesions were collected. Microbial mRNA was isolated and cDNA sequenced. The function and composition of the Scardovia were investigated using two methods: a) de novo assembly of the read data and mapping to contigs, and b) reads mapping to reference genomes. Pearson correlation was performed (p < 0.05). Proportion of Scardovia inopinata and Scardovia wiggsiae sequences ranged from 0-6% in the root caries metatranscriptome. There was a positive correlation between the transcriptome of Lactobacillus spp. and Scardovia spp. (r = 0.70; p = 0.03), as well as with other Bifidobacteriaceae (r = 0.91; p = 0.0006). Genes that code for fructose 6-phosphate phosphoketolase (the key enzyme for “Bifid shunt”), as well as ABC transporters and glycosyl-hydrolases were highly expressed. In conclusion, “Bifid shunt” and starch metabolism are involved in carbohydrate metabolism of S. inopinata and S. wiggsiae in root caries. There is a positive correlation between the metabolism abundance of Lactobacillus spp., Bifidobacteriaceae members, and Scardovia in root caries.


Introduction
Dental biofilms are microbial communities with a complex structure, in which each organism has specific requirements for growth and survival. Consequently, many species, metabolic pathways for utilisation of different nutrients and cell protection against stress factors are found in a biofilm. 1 This diversity of species and metabolism conserves the homeostasis within the biofilm, 2 which can be disrupted by an environmental change that leads to progression to caries. 3 Many species have been found in dental caries biofilms, 4 but studies still focus on well-established cariogenic microorganisms, Streptococcus mutans and Lactobacillus spp. S. mutans is well-documented as an important (but not necessary) organism for caries. Lactobacillus spp. are associated with root caries progression, but were absent or rarely observed in individuals without root caries, 5 in accordance with the concept that they are weakly adherent and depend on ideal niches with retention to be able to colonize. 6 Some studies have unveiled a massive prevalence of the Bifidobacteriaceae family in both coronal 7,8 and root caries. 9,10 Similarly to lactobacilli, the Bifidobacteriaceae family has no ability to form biofilms by itself, 11 requiring a retentive niche to colonize. Some species of this family have been consistently found in root caries lesions, and amongst them, Scardovia spp. has been identified as a member of caries lesion biofilms, especially in severe early childhood caries. 12,13,14 Scardovia spp. was one of the members of the oral microbiota with low abundance (< 0.01%) that significantly influenced the microbial community structure and could be associated with disruption of homeostasis. 15 Prevalence of this genus was associated with root caries activity. 9,10,16 It is still unknown whether Scardovia spp. metabolism may play a major role in dental caries biofilms. Their cariogenic potential was investigated in a few studies 11,17 and their acid production was shown to be higher when in association with S. mutans in vitro. 11 As heterofermentative organisms, Scardovia spp. can metabolize several carbohydrates with lactate and acetate as major end products of fermentation. 18 These products could help in the biofilm acidification leading to dysbiosis. The purpose of this study was to explore the gene expression profiles of Scardovia inopinata and Scardovia wiggsiae in root caries, and compare it with that of Streptococcus mutans, Lactobacillus spp., which are well-established cariogenic microorganisms, and other Bifidobacteraceae members.

Methodology Clinical collection and experimental approach
Volunteers to the study were patients who attended the dental clinics for any dental treatment in two centres: a) Faculty of Dentistry, Federal University of Rio Grande do Sul, Porto Alegre, Brazil; and b) Leeds School of Dentistry, University of Leeds, Leeds, United Kingdom. This study was approved by the ethics committee of the Federal University of Rio Grande do Sul (process no 427.168) and by the ethics committee of the NRES Committee Yorkshire & The Humber -Leeds West (protocol nº 2012002DD). Ethical approval was obtained for the samples collection and all volunteers signed a formal consent.
Details of the clinical collection and experimental approach are described elsewhere. 19 Briefly, clinical samples (biofilm + carious dentin) were collected from volunteers (n = 30) presenting an active cavitated root caries lesion. The diagnosis of root caries was done according to the activity criteria, using visual-tactile examination. 20 Excavated carious dentin containing soft and infected tissue was collected using a sterile spoon excavator (SSWhiteDuflex, Rio de Janeiro, Brazil). The mean age of the volunteers who donated root caries samples was 60.1 ± 11.6 years (range 40-90 years).
After collection, samples were immediately placed in a reagent for RNA stabilization (Qiagen Inc.. Manchester, UK). The total RNA was extracted using the UltraClean® Microbial RNA Isolation kit (Mo-bio, San Diego, USA) with on-column DNAse digestion (Qiagen Inc., Manchester, UK). Samples with a total RNA concentration < 30ng/RNA were pooled, leading to a final sample of seven pools and two unique samples, with good mRNA quantity and quality. The total RNA was depleted of ribosomal RNA using the Ribo-Zero™ Meta-Bacteria Kit (Epicentre, Illumina, San Diego, USA). Illumina®TruSeq™ library prep protocols (Illumina, San Diego, USA) were used for library preparation from the enriched mRNA. Sequencing was performed on the Illumina HiSeq2500 (Illumina Inc., San Diego, USA) sequencer to obtain 2 x 100bp sequence reads.

Bioinformatics analysis
RNA sequencing data are available from the National Center for Biotechnology Information (NCBI) Sequence Read Archive, under the accession number SRS779973. Two different pipelines were used to carry out the bioinformatics analysis, as described below.
Pipeline a) Annotation of de novo assembled contigs Read data were obtained as FASTQ files and were quality trimmed using cutadapt (https:// github.com/marcelm/cutadapt). The taxonomic profiling was carried out by denovo assembly of the reads into contigs using the default parameters at CLC Genomics Workbench 7.5.1 software (CLC Bio, Qiagen, Manchester, UK). Species and gene annotation of the contigs was carried out using diamond (https://ab.inf.uni-tuebingen.de/software/ diamond/) with the NCBI nr protein database (downloaded in December 2017), with settings of 70% similarity to the reference database. Reads from each sample were then mapped against the contigs using the CLC Genomics Workbench 7.5.1 software and a read count table was obtained.
The number of reads assigned to Scardovia spp., Bifidobacteriaceae members, Lactobacillus spp. and S. mutans in the total metatranscriptome was assessed and plotted (RStudio for Mac, version 1.1.463, Plotly R package, ggplot2 R package). The proportion of species in the total metatranscriptome was calculated as follows: [sum of unique reads by species or genera / total number of mapped reads *100]. This was considered as the relative abundance of each species/genus in each sample. Pearson correlation was performed to evaluate the degree of correlation (and the direction of correlation -whether positive or negative) between the appraised organisms (RStudio, base R package; p < 0.05), after checking the normality of the data with the Shapiro-Wilk test.

Pipeline b) Read mapping to 162 oral bacterial genomes
Scardovia spp. gene expression was analysed by mapping reads to 162 oral bacterial reference genomes, including S. inopinata JCM 12537 and S. wiggsiae F0424. Quality-trimmed FASTQ files for each sample were imported into the CLC Genomics Workbench 7.5.1 software (CLC Bio, Qiagen, Manchester, UK). Genomes and their associated information were downloaded from the DNA Data Bank of Japan, NCBI, the Broad Institute and HOMD database, and were used as reference genomes for short read mapping data and a count table was obtained. Then, the putative presence of Scardovia spp. in each sample was estimated by the following method: [sum of reads/total number of genes ≥ 1]. Based on this calculation, one sample was excluded from this analysis due to the low number of reads mapped to Scardovia. The correlation between samples gene expression was plotted (RStudio for Mac, version 1.1.463, Plotly R package, ggplot2 R package, corrplot R package). Gene expression levels were normalized by dividing the number of reads mapping to a single gene from a bacterium by the total reads assigned to that corresponding bacterium and presented as percentages. In order to compare the gene expression of specific genes related to carbohydrate metabolism, differences in means of expression level for particular genes were tested (Wilcoxon Rank Sum Test, corresponding to the Mann-Whitney U test; RStudio, base R package; p < 0.05) after checking the distribution of the data (Shapiro-Wilk and Kolmogorov-Smirnov tests). The number of reads assigned to Scardovia spp., Bifodobacterium spp., Lactobacillus spp. and S. mutans in the total metatranscriptome of root caries was assessed and plotted (RStudio for Mac, version 1.1.463, Plotly R package).

Results
Pipeline a) Composition of the metabolically active microbiota and comparison with established cariogenic species and other Bifidobacteriaceae members Figures 1 and 2 shows descriptive analyses of the number of reads in each sample in the genomes of S. wiggsiae and S. inopinata (reads of both species were summed and presented as Scardovia spp.), as well as species/genera conventionally recognized as cariogenic pathogens. The analyses showed a prevalence of reads assigned to S. inopinata and S. wiggisae ranging from 0 to 6% in all metatrascriptomes of root caries. Two samples presented no gene expression of S. inopinata at all and only 0.04-0.07% of S. wiggsiae. What stands out from these charts is the very high but variable gene expression associated with Lactobacillus spp. This group represented 0. reuteri; L. rhamnosus; L. salivarius; amongst others. Other members of the Bifidobacteriaceae family gene expression in root caries ranged from 0.4% to 8.4% of the total metatranscriptome. This group included reads of Bifidobacterium sp., B. dentium, B. brevis, B. longum, B. termophilum and Parascardovia denticolens. S. mutans was very active in some samples (19.32% of all metatranscriptome in RC_7, for example). The most striking result to emerge from the data is that samples with S. inopinata gene expression representing > 5% of the total microbial metatranscriptome tended to present lower expression of S. mutans, but higher expression of Lactobacillus sp. genes (RC_A, RC_B, RC_H).
The correlation between the proportion of Scardovia spp. and other cariogenic species in the total metatranscriptome was tested using Pearson's correlation after checking data normality. A consistent positive correlation between Scardovia spp. and Lactobacillus spp. was observed (r = 0.70; p = 0.03), as well as a positive correlation with other Bifidobacteriaceae members (r = 0.91; p = 0.0006). In contrast, there was a non-significant negative weak correlation of Scardovia spp. and S. mutans (r = -0.34; p = 0.35) (Figure 3).

Pipeline b) Gene expression analysis
Considering in this analysis the cut-off of the putative presence of Scardovia spp. assigned to each sample was the number of reads/number of genes ≥ 1, only one library had no expression of S. inopinata or S. wiggsiae (RC_E; a pool of samples of two lesions from two patients, originated from center 2). In this sample, high gene expression of Lactobacillus spp. and S. mutans was observed. One sample showed no expression of S. wiggsiae (RC_8; non-pooled sample, from a single patient, originated from centre 1); this was the sample with higher predominance of         Gene expression analysis suggests the utilization of the "Bifid shunt" (oxidative pentose-phosphate (OPP) pathway) by S. inopinata and S. wiggsiae in root caries. There were differences in the ratios of expression of glucose-6-phosphate 1-dehydrogenase (median = 0.06%, 25th-75 th = 0.0%-0.1%), and phosphoketolase (fructose 6-phosphate; median = 0.83%; 25th-75 th = 0.2%-1.3%) (Wilcoxon test, p = 0.000). Looking at the pentose phosphate pathway enzymes ( Figure 5), it appears that phosphoketolase (EC 4.1.2.2) was the most expressed enzyme in both, S. inopitata and S. wiggsiae.
In the top 20 most expressed genes of S. inopinata, were observed the ones that code peptide ABC transporter, ABC transporter ATP-binding protein and sugar ABC transporter permease, while for S. wiggsiae ABC transporter ATP-binding protein (the highest median of reads in the genome), multidrug ABC transporter ATP-binding protein, ABC transporter, ABC transporter substrate-binding protein, sugar ABC transporter permease, and peptide ABC transporter ATP-binding protein featured in the list of most highly expressed genes. A gene that codes a collagen adhesion protein was also among the most expressed genes in both S. inopinata and S. wiggsiae.

Discussion
This study set out to observe the gene expression profile of Scardovia spp. in natural samples of root caries, and compare it with that of other wellestablished cariogenic microorganisms. The most obvious finding to emerge from this study is a great relative prevalence of metabolism of S. inopinata and S. wiggisiae in root caries biofilms where Lactobacillus spp. were also highly relatively prevalent. The research has also shown that genes with functions in some important metabolic pathways were highly expressed in root caries, such as the "bifid shunt" and starch metabolism. These findings contribute to our understanding of Scardovia cariogenic potential Many species have been found in root caries lesions, but studies are still focusing on wellestablished cariogenic microorganisms, such as S. mutans and Lactobacillus spp. However, previous studies reported low or highly variable percentages of these predicted cariogenic species in root caries 16,21,22 . Although culture and DNA-based studies showed low abundance of Lactobacillus sp. and S. mutans, the RNA-seq showed great gene expression here. Such approaches have also failed to address Scardovia spp. prevalence. In this study, the proportion of each, S. inopinata and S. wiggisae, reached approximately 6% of the total metatranscriptome of root caries. Scardovia has been detected in caries lesions, 12,13,14 in association with root caries progression. 9,10,16 It was defined as one of the organisms that could be a 'keystone' for coronal caries. 15 It is interesting to note that only one sample had no expression at all of both species, and all other samples in the present study had activity of at least one species of Scardovia. In this sample, a great gene expression of Lactobacillus spp. and S. mutans was observed. We also sequenced biofilm samples from sound root surfaces; however, using the cut-off, only one out of 10 samples presented Scardovia (data not shown), suggesting that its prevalence is related to the root caries environment.
The correlation of the proportion of Scardovia within the metatranscriptome was evaluated and a significant positive correlation with Lactobacillus and Bifidobacteriaceae members was observed, while a non-significant negative correlation with S. mutans was found. Nyvad et al. 23  a single or few microorganisms may reflect a very high cariogenicity. This is in line with the results of our study, which showed a predominance of one or few species, and a pattern of predominance of Lactobacillus spp. with others, but S. mutans appeared to be active as the main cariogenic species in the community. Eriksson et al. 24 suggested that in subjects with caries experience, high levels of S. mutans were associated with the presence of a few saccharolytic species, including S. wiggsiae. Similarly, a culturebased study showed that the presence of both S. wiggsiae and S. mutans was associated with severe early childhood caries. 13 Another previous in vitro study showed that Scardovia acid production was higher when in association with S. mutans. 11 However, our results showed that clinically this association is not preferable from both species regarding their metabolism in root caries. Bifidobacteria possess a high number of enzymes involved in sugar metabolism. 25 Phosphoketolase ( Figure 5) is the key enzyme for the "bifid shunt" and both species of Scardovia possess it. This pathway allows the production of more energy from carbohydrates compared with that produced by the Embden-Meyerhof-Parnas pathway. 26 Phosphoketolase is also a taxonomic marker for the family of Bifidobacteriaceae. Our results showed that this enzyme was the highest expressed in the OPP pathway enzymes in both S. inopinata and S. wiggsiae ( Figure 5).
In general, high lactate production correlated with low amounts of acetate, ethanol and formate production and vice versa, 27 and it is caused by the rapid consumption of an energy source. 18,26 We observed high gene expression of lactate dehydrogenase (ldh) and pyruvate formate lyase (pfl) in Scardovia. Furthermore, the pfl was significantly higher expressed than acetate kinase (in both species) and acetaldehyde-CoA/ alcohol dehydrogenase (only at S. wiggsiae), which could result in a higher concentration of lactate than acetate and alcohol production by these species, and this may be related to root caries progress.
The expression of carbohydrate-modifying enzymes was also observed. As all samples presented an expression of at least 0.2% of the whole transcriptome of the species, it suggests that this is an essential gene to survive in vivo in root caries. Pullulan is a polysaccharide polymer consisting of maltotriose units, and it is produced from starch and, thus, Scardovia may be metabolizing starch that is cariogenic to root caries. 28 The literature has shown that the gene content of a bifidobacterial genome seems to reflect its adaptation, as indicated by the presence of genes that encode a variety of carbohydrate-modifying enzymes. As expected, many genes coding ABC transporters had high levels of expression in both species. ABC systems are responsible for the transport of nutrients such as mannose-containing oligosaccharides, while glucose is internalized using a glucose-specific PEP:PTS. 18,26 It could suggest that in root caries, Scardovia spp. use low glucose but high complexity carbohydrates because it expresses more ABC transporters than PEP:PTS systems. In contrast to S. mutans, which possess 14 PEP:PTS systems, only a minority of the sugars utilized by bifidobacteria are internalized via PEP:PTS. 18,26 Strengths of this study include the type of methodology used (RNA-seq). As the target molecule of this study was mRNA, the result represents the proportion of general metabolic activity. The metatranscriptome allows the investigation of a large number of genes; however, specific aspects of how some species collaborate to promote root caries progression, as showed here, would be unmanageable by showing the results of the metatranscriptome as a whole. This account must be approached with some caution because it was based on short read mapping and this could add some bias in analysing reads at the species-level. This led us to use two different analysis pipelines to try overcome this issue. The contigs with ~500bp were assembled and mapped to species.
A number of conclusions can be drawn from the present descriptive study. There is a positive correlation between the total metabolism of Lactobacillus spp., other Bifidobacteriaceae and Scardovia sp. in root caries, which could suggest a symbiotic relationship between these species. The Bifid shunt and starch metabolism are expressed in S. inopinata and S. wiggsiae in root caries. The understanding of Scardovia sp. as a protagonist in root caries might provide meaningful information for the development of future strategies of diagnosis and treatment.