Single-Cell Genomics and the Oral Microbiome

Abstract

The human oral cavity is one of the first environments where microbes have been discovered and studied since the dawn of microbiology. Nevertheless, approximately 200 types of bacteria from the oral microbiota have remained uncultured in the laboratory. Some are associated with a healthy oral microbial community, while others are linked to oral diseases, from dental caries to gum disease. Single-cell genomics has enabled inferences on the physiology, virulence, and evolution of such uncultured microorganisms and has further enabled isolation and cultivation of several novel oral bacteria, including the discovery of novel interspecies interactions. This review summarizes some of the more recent advances in this field, which is rapidly moving toward physiologic characterization of single cells and ultimately cultivation of the yet uncultured. A combination of traditional microbiological approaches with genomic-based physiologic predictions and isolation strategies may lead to the oral microbiome being the first complex microbial community to have all its members cultivable in the laboratory. Studying the biology of the individual microbes when in association with other members of the community, in controlled laboratory conditions and in vivo, should lead to a better understanding of oral dysbiosis and its prevention and reversion.

The Oral Microbiome

The human oral cavity harbors a highly diverse and unique microbial community. Individual oral microorganisms engage in direct interactions with other microbial species and with the human host in niche-specific consortia (Diaz and Valm 2019; Mark Welch et al. 2019). When the fine balance or homeostasis that maintains a healthy microbial composition is perturbed, microbial “dysbiosis” can lead to various oral diseases, such as caries and gum disease (gingivitis and periodontitis). It is well established that oral diseases are linked to gradual changes in the proportion of species in the microbiome that harbor a greater pathogenic potential (Diaz et al. 2016; Tanner et al. 2016; Lamont et al. 2018). Various intermicrobial and host-microbe interactions that lead to oral dysbiosis have been uncovered, but many remain unknown (Lamont et al. 2018; Sultan et al. 2018).

The exact number of microbial species that colonize the human oral cavity and its connected aerodigestive niches is still unknown but likely numbers in the thousands (Escapa et al. 2018). The Human Oral Microbiome Database (HOMD) was established in 2010 to further catalogue this microbial diversity (Dewhirst et al. 2010). The HOMD provides a curated, comprehensive platform to house genomic and phylogenetic information pertaining to the human oral microbiome. The recently expanded HOMD, eHOMD (Escapa et al. 2018), contains genetic information of approximately 771 microbial species and strains: 687 in the oral cavity and 84 connected to the aerodigestive microbiota outside the mouth. Of these, >70% of the microorganisms have been cultured in the laboratory, while a quarter remain uncultured. The diversity of characterized isolates and the uncultured phylotypes, based on small subunit rRNA genes, belong to 10 recognized bacterial phyla and several candidate groups, including TM7 (Saccharibacteria), GN02 (Gracilibacteria), and SR1 (Absconditabacteria), recently grouped into phylum Patescibacteria (Dewhirst et al. 2010; Parks et al. 2018; Fig. 1). The archaeal component of the oral microbiota is much less diverse with only 1 isolated representative, the methanogen Methanobrevibacter oralis (Euryarchaeota), although recent sequence-based studies suggest that other uncultured lineages are present as well (Wade 2013; Dame-Teixeira et al. 2020). The human oral mycobiome, which encompasses nearly 100 genera of fungi (Ghannoum et al. 2010), has been less studied, except for Candida species, which are well documented to play important roles in oral diseases, directly and through interactions with bacteria (Baker et al. 2017; Chevalier et al. 2018; Bertolini and Dongari-Bagtzoglou 2019).

Figure 1.

Phylogenetic tree (FastML) of human oral bacteria based on the SSU rRNA gene (HOMD V14.51 sequences). Phylum-level classification is indicated on the outer ring by individual color blocks with some of the major genera identified. Red dots indicate currently uncultured organisms, and blue squares indicate organisms for which single-cell genomes have been sequenced. The tree intends to provide a general view on the cultivation status at high taxonomic level. For the names of all oral microorganisms, and their status, see http://homd.org/.

Some of the bacterial genera in the oral community include species with well-established pathogenic potential (e.g., Porphyromonas, Treponema, Tannerella, Fusobacterium, Aggregatibacter, Streptococcus, Lactobacillus) that have been studied extensively, resulting in robust experimental models of gum disease and caries (Graves et al. 2012; Hajishengallis 2014; Aberg et al. 2015; Diaz et al. 2016; Tanner et al. 2016; Lamont et al. 2018). It has been widely accepted that tooth caries and gum disease are complex oral diseases with a polymicrobial etiology that results from disturbances (dysbiosis) of the healthy community structure and subsequently its interaction with the human immune system (Darveau 2010; Diaz et al. 2016; Tanner et al. 2016; Lamont et al. 2018; Fine et al. 2019). Consequently, an increasing number of species, including uncultured lineages, have been associated with the disruption of microbial-host homeostasis and disease progression (Griffen et al. 2011; Abusleme et al. 2013; Wade 2013; Diaz et al. 2016; Tanner et al. 2016). Oral dysbiosis can be linked to numerous factors, including nutrition, hygiene, and genetic predisposition, and can be instigated by systemic diseases that trigger an inflammatory response (Graves et al. 2019). Other studies have indicated that some oral microbes may be linked to cardiovascular disease (Beck and Offenbacher 2005), diabetes (Genco et al. 2005), cancer (Koliarakis et al. 2019), and other diseases. Therefore, it is crucial to understand not only the overall oral microbial diversity and its dynamics in health and disease but also the biology of its individual components and how key species integrate in the community to affect other microbes and the human host. As a significant number of oral microorganisms currently remain uncultured, sequencing-based approaches at the single-cell and population levels can provide insights into their physiologic potentials and provide clues for their isolation in laboratory culture. This review summarizes recent advances in the study of uncultured bacteria and archaea at the single-cell level with an emphasis on the oral microbiota.

Cultivation-Independent Microbiology: SAGs and MAGs

Amplicon sequencing of the 16S rRNA gene enables a rapid way to determine the composition of oral microbiota to identify variations in 1) community structure among individuals, 2) oral niches, and 3) health and disease. However, it provides limited physiologic information of the organisms, which is especially true for uncultured organisms. Two main approaches have been developed to provide access to the microbial genomic information of these organisms to investigate their physiologic potential in the absence of cultivation. This can be achieved, first, at the single-cell level following amplification and sequencing of genomic content from an individual cell (single-amplified genomes [SAGs]) and, second, through shotgun metagenomic sequencing of the entire community genomic material, followed by assembly and binning of individual genomes (metagenome-assembled genomes [MAGs]). Both approaches have been reviewed and compared extensively (Lasken and McLean 2014; Woyke et al. 2017; Alneberg et al. 2018), each providing a different window into the physiologic potential of individual organisms and their communities. However, each has its limitations. In single-cell genomics, the genomic content of an individual cell that has been isolated by flow cytometry or microfluidic separation is enzymatically amplified and sequenced (Fig. 2). The resulting SAGs capture the sequence of these unique microbial cells and, if applied to multiple cells, can provide important insights into the genomic variability and population structure of individual species (McLean et al. 2013; Kashtan et al. 2014). If specific cells are isolated on the basis of phylogenetic markers (e.g., rRNA probes and fluorescence in situ hybridization [FISH]; Campbell, Campbell, et al. 2013) or specific physiologic characteristics (Doud and Woyke 2017), SAGs can provide a more focused investigation of target organisms regardless of their abundance within the target communities. However, SAGs have various degrees of completeness and can contain chimeric, artifactual genomic rearrangements generated during the enzymatic amplification step. MAGs, however, can achieve higher genomic coverage with fewer risks of artifacts, although their assembly is dependent on sequencing depth and community complexity, which could miss low-abundance taxa. Because they represent a consensus of sequences originating from many nonclonal cells, MAGs represent pangenomes of a species or strain, depending on the degree of genetic variation in that organism’s population. As such, each methodology has advantages and disadvantages that should be considered when experiments are designed specific to the questions at hand.

Figure 2.

Single-cell genomics and cultivation workflow for targeted and untargeted approaches. Dotted lines indicate approaches incompatible with viable cell isolation/cultivation. FISH, fluorescence in situ hybridization.

Single-Cell Genomic Sequencing of the Oral Microbial “Dark Matter”

Several high-level taxonomic lineages from the human oral microbiota have resisted cultivation for decades and have been important initial targets for genomic characterization. These included candidate bacterial phyla (TM7, GN02, SR1, WPS-2) as well as oral Chloroflexi and Chlorobi, all present at low abundance and generally displaying reduced species diversity (Camanocha and Dewhirst 2014; Wade et al. 2016). These bacterial lineages are highly prevalent in open environments, ranging from hot spring mats and soils, anoxic sediments, and aquatic niches. The first genomic data for oral TM7 bacteria (more recently renamed Saccharibacteria) was obtained by sequencing amplified genomes from 3 cells isolated from oral subgingival plaque with a microfluidic chip device (Marcy, Ouverney, et al. 2007). Even though the original assemblies were contaminated with other bacterial sequences (Albertsen et al. 2013), their analyses indicated capacity for sugar and peptide utilization and provided insights into the membrane architecture in those organisms. Subsequent cultivation of the first TM7 bacterium (TM7x) revealed that these organisms depend on direct association with oral Actinobacteria and act as ectobionts/parasites (He et al. 2015). The single-cell TM7 genomes generated by Marcy et al. provided information for designing a targeted antibody-based approach to label, isolate, sequence and cultivate additional members of that oral lineage (Cross et al. 2019).

Single-cell genomic investigation of 1 of the 3 recognized phylotypes of human oral SR1 bacteria, isolated by flow sorting, led to the discovery of a novel type of genetic recoding where the opal stop codon (TGA) was reassigned to encode for glycine (Campbell, O’Donoghue, et al. 2013). SR1 bacteria have been recently assigned within the class Gracilibacteria, phylum Patescibacteria (Parks et al. 2018), and are widely distributed in anoxic environment. This is similar to GN02, a related lineage, that also uses TGA to glycine recoding (Rinke et al. 2013) and has small genomes, suggesting that these organisms have a strict dependency on other species. Another rare oral lineage that belongs to a nonphotosynthetic group of Chloroflexi (class Anarolineae) was first characterized by genomic sequencing of 2 cells isolated stochastically from oral subgingival samples in patients with periodontitis (Campbell et al. 2014). Subsequent culturing showed this organism to be dependent on yet-unidentified soluble factors produced by a Fusobacterium species (Vartoukian, Adamowska, et al. 2016). This Chloroflexi, Anaerolineae HOT-439, resembles its free-living relatives in that it relies on an anaerobic fermentative metabolism yet has acquired unique genomic adaptations in the human oral environment, including the acquisition of potential virulence factors that may be linked to its association with periodontitis (Abusleme et al. 2013; Szafranski et al. 2015).

Several members of the Bacteroidetes genus associated with health or disease were first characterized by single-cell sequencing. Tannerella sp. BU063 (HOT286), a relative of the periodontal pathogen Tannerella forsythia, is predominantly present in healthy oral communities and lacks some of the virulence genes present in disease-associated species (Beall et al. 2014). Further cultivation has revealed that this species also requires soluble factors produced by other oral bacteria (Vartoukian, Moazzez, et al. 2016). Tannerella sp. BU045 (HOT808), however, is a periodontitis-associated species for which a dozen single-cell genomes have been obtained; this breadth of information can help to better define health-disease markers within the Tannerella genus (Beall et al. 2018). In Porphyromonas gingivalis, analyses of single cells recovered from a hospital sink biofilm identified polymorphisms potentially linked to natural variations in virulence (McLean et al. 2013).

Furthermore, phylogenetic targeting with FISH, followed by flow sorting and single-cell genome sequencing, led to the characterization of 2 previously uncharacterized oral Deltaproteobacteria genera that are enriched in periodontitis: Desulfovibrio and Desulfobulbus (Campbell, Campbell, et al. 2013). A single Desulfobulbus phylotype is present in the oral microbiota and, based on its inferred sulfate-reducing physiology, was enriched for and subsequently isolated in pure culture (Cross et al. 2018; Fig. 3). The cultivation of Desulfobulbus oralis represents one of the few cases where predicted physiologic characteristics based on single-cell genomic data has led to isolation of a specific microorganism in the laboratory. D. oralis was characterized as a novel periodontal pathobiont with virulence and proinflammatory properties acquired by lateral gene transfer from other oral bacteria and is dependent on nutrients produced by Fusobacterium for growth (Cross et al. 2018).

Figure 3.

Human oral cultured bacteria resulting from the use of single-cell genomic data. (A) A coculture of Fusobacterium nucleatum (DAPI stain, blue) and Desulfobulbus oralis (stained with a green fluorescent rRNA probe). (B) Scanning electron micrograph of a pure D. oralis culture. (C) A coculture of Actinomyces sp. and TM7 HOT952 stained with rRNA probes.

The genomic diversity of oral fungi has remained vastly unexplored, with the exception of Candida species. One study identified a dozen or so yet-uncultured fungal genera in oral samples from healthy individuals (Ghannoum et al. 2010). While the distribution of the different types of fungi across the human population and their correlation with heath and disease have not been studied extensively, molecular methods to identify oral fungi based on internal transcribed spacer sequences have been developed (Diaz et al. 2017). We have not found, however, any report of single-cell genomic sequencing of oral fungi. This may be due in part to difficulties in lysing fungal cells under the conditions typically used for bacteria. A recent study that generated improved protocols to enable genomic amplification and sequencing of nonhuman fungi may open the field of single-cell genomics for the human oral mycobiome (Ahrendt et al. 2018).

Experimental and Computational Advances in Microbial Single-Cell Sequencing

In recent years, microbial single-cell genomics has achieved a number of significant advances in throughput, in reduction of sequence amplification bias and artifacts, as well as in eliminating spurious data from genomic assemblies. As several recent reviews have covered this topic (Gawad et al. 2016; Woyke et al. 2017), we focus here primarily on the recent advancements.

Isolation of individual cells has traditionally been performed by high-speed flow cytometry sorting into 96- or 384-well plates or by lower-throughput micromanipulation and microfluidic separation into homemade chips. Flow-sorted cells are lysed and their genome amplified in multimicroliter-scale reaction volumes, which has enabled automation that has generally kept the cost high while presenting increased risks of contamination and amplification artifacts. Improvements were achieved by special cleaning protocols and the use of piezo liquid–dispensing equipment that reduced the reaction volumes to microliter levels (Rinke et al. 2014). Additionally, early use of microfluidic nanoliter reactors have continued to reduce amplification bias and contamination (Marcy, Ishoey, et al. 2007). More recently, several microfluidic approaches, such as combining handling of cells in closed systems, nanoliter droplets, and genomic barcoding, have enabled processing of many thousands of microbial cells (Gole et al. 2013; Xu et al. 2016; Lan et al. 2017).

Two major improvements have been recently achieved for the single-cell genomic amplification step. First, the inclusion of a DNA primase from Thermus thermophilus (PrimPol) has eliminated the requirement for exogenous primers to be added to the MDA reaction (multiple displacement amplification). PrimPol generates oligonucleotides de novo with the target genomic DNA, which are then used by the phi29 polymerase for long extension and strand displacement reactions (Picher et al. 2016). The enzyme system, commercialized as TruePrime (Expedeon Inc), was reported to significantly reduce amplification bias and provide more uniform amplification across the genome (Picher et al. 2016). Second, a mutant thermostable phi29 polymerase has been engineered, allowing MDA reaction to proceed at elevated temperature (45 °C vs. the standard 30 °C; Stepanauskas et al. 2017). This has resulted in more efficient amplification, especially for genomes with high GC content, which has been difficult to recover with traditional MDA. The novel polymerase is being commercialized as EquiPhi29 DNA polymerase by ThermoFisher Scientific and requires the addition of exonuclease-resistant oligonucleotide primers, as with the protocol for native phi29 polymerase. A combination of primer-independent and elevated-temperature MDA has not yet been reported but may combine the advantages of the 2 independent advances.

The presence of sequences originating from microbes other than that of the main amplified genome has always plagued microbial single-cell genomics. Interpreted as resulting from 1) contamination, 2) artifactual introduction by any of the multiple experimental steps and reagents, and 3) the presence of free DNA in the biological sample or multiple cells/cell fragments isolated as a single particle. New bioinformatic methods have been developed to identify such “contaminants.” The presence of >1 organism in “single cell” genomic data may also indicate real interspecies interaction (Podar et al. 2013; Jarett et al. 2018). Single-cell sequence data are fragmentary, with uneven genomic coverage and co-occurring genomes (contamination or not), which introduces nucleotide composition differences (characteristics shared with metagenomic data). Assembly has most often been performed with dedicated software such as SPAdes (Bankevich et al. 2012) and IDBA-UD (Peng et al. 2012). To identify compositional heterogeneity in the assembled contigs, sequences are binned on the basis of tetranucleotide frequencies, followed by phylogenetic assignment and analysis of single-copy genes, similar to approaches used in metagenomic sequence analyses (Dick et al. 2009; Parks et al. 2015; Sieber et al. 2018). A set of standards has been proposed for the quality and reported information for single-cell genomic as well as for metagenomic data deposited in public sequence repositories (Bowers et al. 2017).

From Single-Cell Genomes to Phenomes and Cultures

Although untargeted single-cell isolation and genomic sequencing are powerful in uncovering the diversity of the microbial world and enabling inferences about microbial function, it falls short of revealing the metabolic state and physiologic capabilities of those cells. Moreover, because the cells isolated for genome sequencing are destroyed in the process, downstream biological studies of those lineages are unfeasible. Over the last several years, increased efforts have been directed to identify or label microbial cells from environmental samples based on specific enzymatic activities or functional markers to garner a more robust view of their physiologic potential. Conceivably, if labeling cells for a specific functional trait is nondestructive, it could also be combined with cultivation and open the door to direct biological characterization. As a recent review (Doud and Woyke 2017) summarized many of the concepts, results, and directions for function based single-cell genomics, we focus here on 3 novel approaches that are generally applicable to any microbial community, including oral microbiota, and are poised to significantly affect the field.

Cellular Labeling Based on Specific Physiologic Processes

Over a decade ago, Dieterich et al. (2006) discovered that mammalian cells can uptake amino acid surrogates containing azido groups (e.g., the methionine derivative L-azidohomoalanine) and incorporate them into newly synthesized proteins. These proteins could then be selectively and fluorescently labeled by a click chemistry reaction, a procedure referred to as bioorthogonal noncanonical amino acid tagging (BONCAT). Recently, Hatzenpichler et al. used BONCAT to identify translationally active cells within complex environmental samples, including an oral biofilm, and showed that these labeled cells can be coupled with FISH for taxonomic identification and further sorted for single-cell genomics with flow cytometry (Hatzenpichler et al. 2014; Hatzenpichler et al. 2016). The amino acid analogues do not appear to be toxic, and labeling can be performed under physiologic conditions. This provides immense opportunities to identify active populations of microorganisms in the environment, including those from complex environments such as the oral microbiome, that respond to perturbations and selectively isolate them for further characterization. Click chemistry has since been applied to metabolically label bacterial cell wall peptidoglycan and lipopolysaccharides, including those in active mammalian microbiomes (Hudak et al. 2017; Wang et al. 2017), enabling identification of actively dividing bacteria. As the range of clickable metabolic analogues increases, this approach may expand to target additional physiologic processes in microbial populations, and several companies have commercialized such substrates and labeling reagents (e.g., Jena Bioscience, Click Chemistry Tools).

Single-Cell Raman Spectroscopy

Single-cell Raman spectroscopy (SCRS) is a novel, noninvasive, and label-free technology that provides a phenotypic and intrinsic biochemical fingerprint of single cells. This facilitates differentiation of cell types, physiologic states, nutrient condition, and variable phenotypes of specific microorganisms following differential incorporation of stable isotopes. SCRS, in combination with cell-sorting techniques such as Raman-activated cell ejection and Raman-activated microfluidic sorting, makes it possible to isolate single cells based on their SCRS characteristics and amplify single-cell genomes from complex microbial communities (Li et al. 2014; Song et al. 2017; Jing et al. 2018). This method has been successfully employed to sort single cells from marine samples and to identify bacteria with novel functional genes, including a previously uncultured Cyanobacteria sp. (Song et al. 2017). More recently, Lee et al. (2019) combined microfluidics, optical tweezing, and Raman spectroscopy to develop an optofluidic platform to sort stable isotope–labeled bacterial cells involved in mucin degradation in mouse colon. Following sorting, the individual cells were suitable for downstream single-cell genomics, mini-metagenomics, and cultivation. Although this method has yet to be applied to oral microorganisms, it appears to be a promising approach in studying individual microorganisms as part of complex oral communities assembled in vitro or from oral samples.

Single-Cell Isolation and Cultivation through Targeted Reverse Genomics

The human oral microbiome is likely the first microbial community for which the goal of achieving cultivation of all its members is within reach. As many of its yet-uncultured microorganisms are linked to heath or disease status, laboratory isolates will be important to fully characterize their physiologic, microbial, and host interaction potential. Information obtained from culture-independent studies, including single-cell genomics and metagenomics, can be used to guide cultivation efforts, as shown for Desulfobulbus oralis (Cross et al. 2018). A recent study also demonstrated that genomic information may in fact be used to directly facilitate physical isolation of target organisms and lead to cultivation (Cross et al. 2019). The approach, referred to as “targeted reverse genomics” (Fig. 2), involves first identifying genes in the genome of the uncultured target organism that encode membrane proteins with extracellular domains, ideally having structurally and functionally characterized homologues in other bacteria. Polyclonal antibodies against fragments of those domains can then be raised with either peptides or proteins heterologously expressed on the basis of synthetic genes. Those antibodies can then be used to label oral samples containing the target organisms and single cells isolated by flow sorting. The cells can be either used to obtain additional genomic sequence data or inoculated in media for cultivation as the antibodies bind cells externally, rendering them viable. This approach led to the isolation of several novel human oral TM7 bacteria with their Actinobacteria hosts (Fig. 3) and has made progress toward the first reported cultivation of an oral SR1 bacterium (Cross et al. 2019). Targeted reverse genomic-based cultivation has the potential for a high impact in environmental microbiology, as many lineages that are known only by sequence may now be targeted for isolation and biological characterization.

Author Contributions

M. Balachandran, contributed to data analysis, drafted the manuscript; K.L. Cross, contributed to data analysis, critically revised the manuscript; M. Podar, contributed to conception and design, drafted the manuscript. All authors gave final approval and agree to be accountable for all aspects of the work.