Escherichia coli cultures were maintained at 37°C in Luria Bertani medium (LB; Difco, Sparks, MD) with aeration or grown aerobically on LB agar plates. All Veillonella cultures and agar plates were grown at 37°C in anaerobic conditions (85% nitrogen, 10% carbon dioxide, and 5% hydrogen) using either Todd Hewitt medium (Difco, Sparks, MD) supplemented with 0.3% wt/vol yeast extract (Fisher, Fair Lawn, NJ) and 0.6% sodium lactate (Spectrum Chemical, Gardena, CA) (together referred to as THL medium) or a novel medium referred to as SK medium. SK medium composition is as follows: 10 g L⁻¹ tryptone (Fisher, Fair Lawn, NJ), 10 g l⁻¹ yeast extract (Fisher, Fair Lawn, NJ), 0.4 g l⁻¹ disodium phosphate (Fisher, Fair Lawn, NJ), 2 g l⁻¹ sodium chloride (Fisher, Fair Lawn, NJ), and 10 ml l⁻¹ 60% (wt/vol) sodium lactate (Spectrum Chemical, Gardena, CA). For antibiotic selection, cultures and/or agar plates were supplemented with the following antibiotics: 150 μg ml⁻¹ ampicillin (E. coli), 9 μg ml⁻¹ tetracycline (Veillonella), and 350 μg ml⁻¹ spectinomycin (Veillonella).

Prior to the study, the clinical sample collection and strain isolation protocol were reviewed by the Oregon Health and Science University Institutional Review Board (IRB) and deemed not human subjects research. All clinical specimens were collected by clinicians in the OHSU Pediatric Dental clinic using material generated during routine treatment procedures. Specimens were removed of all Protected Health Information (PHI) identifiers as part of the collection protocol. Clinical specimens were derived from a pediatric cohort and consisted of 67 unique buccal molar plaque samples. Initially, Veillonella clinical isolates were obtained using a previously published protocol (Liu et al., 2012). Briefly, plaque samples obtained from the clinic were vortexed to disperse the cells and then 100 μl was plated onto TH plates containing a confluent lawn of S. mutans strain UA159 previously grown for 24 h. Plates were incubated anaerobically at 37°C for 48 h. before inspecting the plates for the presence of colonies forming on top of the S. mutans lawn. Cells from candidate colonies were observed by microscopy for the expected cellular morphology. Promising clones were restreaked and then further tested by 16S rRNA sequencing and, in some cases, rpoB PCR amplification. In an effort to improve the efficiency of Veillonella isolation, a new enrichment strategy was developed and implemented. For this approach, each unique clinical plaque sample was used to inoculate two separate 1 ml cultures of SK medium +15.6 μg ml⁻¹ erythromycin as well as SK medium +0.5 μg ml⁻¹ chloramphenicol. Antibiotic enrichment cultures were incubated until reaching stationary phase (typically 24 h) and then quadrant streaked onto SK agar plates. Cells from isolated colonies were observed by microscopy for the expected cellular morphology. Candidate clones were restreaked and then further tested by 16S rRNA sequencing and, in some cases, rpoB PCR amplification.

All primers used in the study are listed in Table 1. Pure culture isolates of candidate Veillonella species were tested by amplifying the 16S rRNA gene using PCR and the primers 16S Uni F/R. The resulting PCR products were purified using the DNA Clean and Concentrator kit (Zymo, Irvine, CA) and then sequenced using the 16S Uni F/R primers. Species assignments were made by performing a BLASTN search of 16S sequence data in both the Human Oral Microbiome Database (www.homd.org) and the Green Genes database (greengenes.lbl.gov). In some cases, it was necessary to further verify species assignments using PCR amplification of the Veillonella rpoB gene (Igarashi et al., 2009). 5 unique rpoB forward primers (DENF, PARF, ROGF, ATYF, and DISF) were combined in a single PCR reaction with the Veillonella rpoB universal reverse primer VR. Based upon amplicon sizes, it was possible to determine the species of the tested isolates. All identified Veillonella clinical isolates are listed in Table 2.

V. parvula strains SKV17B and SKV38 were streaked onto both SK and THL plates and incubated anaerobically at 37°C for 48 h. Cells were scraped from each of the plates and resuspended in 1 ml of their respective liquid media (SK or THL) and then vortexed to disperse. The cell densities of the cultures were subsequently adjusted to OD₆₀₀ 0.4. Next, 10 μl of the cell suspensions was spotted onto the respective agar plates (SK or THL) and then immediately followed by adding an additional 800 ng of the E. coli-Veillonella shuttle plasmid pBSJL1 (dissolved in 3 μl TE buffer) (Liu et al., 2012) to the spots. Plates were allowed to dry in the anaerobic chamber before incubating anaerobically at 37°C for 48 h. Afterward, cells growing in the 10 μl spots were scraped from the plates and resuspended in 250 μl of their respective liquid media (SK or THL) and then vortexed to disperse. 10 μl of each cell suspension was spotted onto SK plates supplemented with tetracycline and incubated anaerobically at 37°C for an additional 48 h. Candidate pBSJL1 transformants of SKV17B and SKV38 were inoculated into SK medium supplemented with tetracycline and then tested for the presence of pBSJL1 using PCR and the primers pBS F/R.

Plasmid extraction from V. parvula was performed using the typical alkaline lysis solutions I, II, and III (Green and Sambrook, 2016) with minor modifications as described below. 12 ml cultures of SK medium + tetracycline were inoculated with pBSJL1 transformants of V. parvula strains SKV17B and SKV38 and grown to stationary phase. Separate 12 ml non-selective SK medium cultures were also inoculated with wild-type plasmid-free SKV17B and SKV38 to serve as negative controls. The cultures were each divided into 3 × 4 ml aliquots and pelleted by centrifugation. Cells were resuspended in 250 μl of solution I containing 2.5 mg lysozyme and incubated in a 37°C water bath for 1.5 h. with shaking every 30 min. Next, 250 μl of solution II was added to lyse the cells followed by an additional 350 μl of solution III to neutralize the lysate. Cellular debris was pelleted by centrifugation at 16,000 × g for 8.5 min. and the cleared supernatants were collected. Supernatants were treated with phenol/chloroform solution, precipitated with 2-propanol, and resuspended in 70 μl of TE buffer. Purified pBSJL1 was used to retransform E. coli along with the negative control samples prepared from wild-type plasmid-free SKV17B and SKV38.

All primers used for mutation construct assembly are listed in Table 1. All primers and mutagenesis constructs were designed using Serial Cloner software (https://serial-cloner.en.softonic.com). Allelic exchange mutation constructs were assembled using overlap extension PCR ligation (Xie et al., 2013) with Phusion DNA polymerase (New England Biolabs, New Ipswich, MA) and strain SKV38 chromosomal DNA and pBSJL1 as PCR templates. The recA construct was assembled by first amplifying the upstream and downstream homologous fragments from strain SKV38 using the primers recA up F/R and recA dn F/R. 0.5 μl of both resulting PCR amplicons was mixed with a PCR amplicon of the tetM tetracycline resistance cassette amplified using pBSJL1 as a template and the primers TetM F/R. Regions of complementarity between the recA homologous fragments and the tetM amplicon facilitated their subsequent PCR ligation using the primers recA up F/recA dn R. A similar strategy was used to assemble constructs creating both large and small deletions within the competence gene locus. For the large deletion, about 5.7 kb of the 8 kb locus was deleted. Primers used to amplify the upstream and downstream homologous fragments for this construct were gspE up F3/TMP up R3 and 4715 dn F/4725 dn R, respectively. For the small deletion, 1 kb was deleted from the middle of the locus. The upstream and downstream homologous fragments were amplified with the primers gspF dn F2/pulO dn R2 and pulO dn F2/964 dn R, respectively. 0.5 μl of the resulting PCR amplicons was mixed with a PCR amplicon of the tetM tetracycline resistance cassette amplified with the primers TetM F/R. Regions of complementarity between the competence locus homologous fragments and the tetM amplicon facilitated their subsequent PCR ligation using the primers gspE up F3/4725 dn R for the large deletion construct and gspF dn F2/964 dn R for the small deletion construct. The resulting PCR products were transformed into wild-type V. parvula strain SKV38 and selected on agar plates supplemented with tetracycline. Additional competence locus mutation constructs creating large deletions were assembled using template gDNA extracted from a confirmed competence gene locus large deletion mutant and amplifying with the primers gspE up F4/4725 dn R2 as well as gspE up F3/4725 dn R. These PCR products were subsequently transformed into wild-type V. parvula strain SKV38 and selected on agar plates supplemented with tetracycline. Additional competence locus mutation constructs creating small deletions were assembled using template gDNA extracted from a confirmed competence gene locus small deletion mutant and amplifying with the primers gspF dn F3/pulO dn R3 as well as gspE up F3/4725 dn R. These PCR products were subsequently transformed into wild-type V. parvula strain SKV38 and selected on agar plates supplemented with tetracycline.

The same natural transformation protocol described for pBSJL1 transformation was used to transform the allelic exchange mutagenesis constructs created by overlap extension PCR ligation. Transformation reactions both contained an equivalent molar ratio of a saturating amount of transforming DNA. For the recA construct 75 ng was transformed (7.5 μg ml⁻¹ final concentration), while 145 ng of the competence locus construct was transformed (14.5 μg ml⁻¹ final concentration). Transformants were selected on SK agar plates supplemented with tetracycline, confirmed via PCR, and finally tested for their natural competence abilities using the previously described natural transformation protocol. Wild-type and mutant strains were transformed with isogenic genomic DNA derived from a spontaneous spectinomycin resistant isolate of SKV38. The spectinomycin resistant mutant was isolated by plating 500 μl of a stationary phase culture of SKV38 onto SK agar plates supplemented with spectinomycin. 500 ng of spectinomycin resistant mutant genomic DNA (50 μg ml⁻¹) was added to each transformation reaction and transformants were selected on SK agar plates supplemented with 350 μg ml⁻¹ spectinomycin.

The same previously described natural transformation protocol was used to transform wild-type SKV38 with either 300 ng pBSJL1, 145 ng of competence locus deletion PCR product, or 500 ng genomic DNA. To assess transformation efficiency, cells were plated on both non-selective and selective SK agar plates. Transformation efficiency is defined as the ratio of transformants to total CFU. Data are expressed as the average and standard deviations of at least three independent experiments performed in triplicate.

The same natural transformation protocol described for pBSJL1 transformation was used to transform a panel of wild-type V. parvula strains with 500 ng genomic DNA derived from a tetracycline resistant competence locus deletion mutant of strain SKV38. In some cases, it was necessary to transform gDNA derived from a spontaneous spectinomycin resistant mutant of strain SKV38 due to the inherent tetracycline resistance of the parent strain. In such cases, 500 ng of gDNA conferring spectinomycin resistance was transformed and the transformants were selected on agar plates supplemented with spectinomycin. Data are presented as the number of tetracycline resistant transformants generated from 10 μl of transformation reaction cell suspension. Data are expressed as the average and standard deviations of three independent experiments performed in triplicate.

Recently, several mutations have been introduced into V. atypica strain OK5 using electrotransformation of the suicide vector pBST (Zhou et al., 2015a,b, 2016). These studies successfully demonstrated the potential genetic tractability of veillonellae. While this approach has proven to be a reliable mechanism to generate targeted mutations in strain OK5, the entire procedure can be laborious, and thus far, only single crossover insertion duplication mutations have been achieved despite previous attempts to create double crossover allelic exchange mutations (Zhou et al., 2015a). One of the principal benefits of allelic exchange mutagenesis is that the constructs can be easily assembled using solely PCR, circumventing the cloning requirement for suicide vector construction. This greatly increases the speed at which constructs can be assembled and avoids issues related to construct toxicity in E. coli. Since it was already established that allelic exchange mutagenesis is ineffective via electrotransformation of OK5, we were curious whether veillonellae might be naturally transformable, and thus, likely amenable to allelic exchange mutagenesis. To test this, we first isolated a collection of veillonellae from clinical plaque samples in a pediatric cohort from the OHSU pediatric dental clinic. In total, 67 unique buccal molar plaque specimens were collected from separate patients, resulting in a total of 43 unique Veillonella isolates. Initially, we employed a previously published Veillonella isolation technique in which strains are isolated on confluent lawns of S. mutans (Liu et al., 2012). While this approach did yield isolates, our success rate was lower than desired. Therefore, we developed a new simple and efficient protocol to improve the reliability of Veillonella isolation from plaque samples. As described in Materials and Methods, clinical specimens are first directly enriched for veillonellae by incubating in duplicate cultures of SK medium + chloramphenicol and SK medium + erythromycin. We found this single enrichment step to substantially increase the proportion of Veillonella in one or both cultures. Consequently, we were able to simply streak the enrichment cultures on non-selective plates and identify candidate veillonellae from the resulting colonies. Of the 43 unique isolates, we obtained multiple strains of the three most common Veillonella species found in dental plaque (V. atypica, dispar, and parvula) as well as a lone isolate of the less common species V. denticariosi (Table 2). Species assignments were based upon 16S rRNA sequence data. Since it was difficult to definitively distinguish between V. dispar and V. parvula using 16S rRNA sequences, each of these isolates was further confirmed by amplification of the rpoB gene (Igarashi et al., 2009). Next, each of the strains was tested for natural competence by transforming the E. coli-Veillonella shuttle vector pBSJL1 using an agar plate transformation approach with SK and THL media. Surprisingly, two strains of V. parvula (SKV17B and SKV38) yielded candidate transformants using SK medium, whereas no candidates were detected from any of the strains using THL medium (Figure 1A). Three candidates from the SKV17B and SKV38 transformations were further tested for the presence of pBSJL1 using plasmid-specific PCR primers and all appeared to contain the vector (Figure 1B). As a final confirmation, plasmid DNA was extracted from these transformants and then retransformed into E. coli, resulting in the expected acquisition of antibiotic resistance (Figure 1C).

Given the success with pBSJL1 transformation, we were next curious to test whether the same procedure could be used for targeted mutagenesis of V. parvula via allelic exchange mutagenesis with PCR products. We were particularly interested to mutagenize genes potentially responsible for exogenous DNA uptake. In most naturally competent bacteria, the DNA uptake machinery is encoded by loci containing genes homologous to those of T2SS and genes involved in type IV pilin biogenesis (Averhoff and Friedrich, 2003; Chen and Dubnau, 2003, 2004). A candidate locus was identified in the sequence data for the V. parvula genome reference strain DSM2008 (VPAR_RS05490—5435) (Figure 2A). The 12 genes in the locus appeared to be organized into one or two polycistronic operons. The first two genes in the locus (VPAR_RS05490 and 5485) encode putative traffic NTPases resembling those of T2SS with the second gene product exhibiting homology to PilT proteins. Most bona fide T2SS do not encode a second traffic NTPase, whereas this is a common feature of type IV pilin biogenesis loci in both Gram positive and Gram negative bacteria (Chen and Dubnau, 2004; Giltner et al., 2012). In addition, a variety of the downstream open reading frames (ORFs) between VPAR_RS05465—5435 have features that are highly reminiscent of type IV prepilins (Figure 2A). The ORFs in this locus were further analyzed using the PilFind webserver (http://signalfind.org/pilfind.html) (Imam et al., 2011) and a canonical type IV prepilin leader sequence was detected in VPAR_RS05465. This ORF is also located directly adjacent to the putative prepilin peptidase (VPAR_RS05470) likely responsible for cleaving such leader peptides (Giltner et al., 2012) (Figure 2A). These features were all suggestive of a type IV pilin biogenesis locus, many of which serve as competence loci (Chen and Dubnau, 2004; Giltner et al., 2012), and thus it was targeted for mutagenesis. As a control, we also targeted recA for mutagenesis, since this mutant was expected to lose its natural transformability due to severe impairments in homologous recombination. Allelic exchange mutagenesis constructs were assembled via PCR and transformed into SKV38 using both SK and THL media. Once again, SK medium yielded substantially higher numbers of transformants compared to THL (Figure 2B), further confirming our previous observations. Antibiotic resistant transformants were subsequently tested via PCR and confirmed to exhibit the expected mutant genotypes for both the recA and competence gene locus deletions (Figure 2C). Next, we assayed the natural competence abilities of the two allelic exchange mutant strains. Since only one antibiotic resistance cassette is available for veillonellae, we transformed genomic DNA (gDNA) derived from a spontaneous spectinomycin resistant mutant of SVK38. These transformants could then be selected on plates supplemented with spectinomycin. To our surprise, the wild-type control yielded a large number of transformants far beyond what was anticipated based upon the previous transformations with either pBSJL1 or PCR products (Figure 2D). However, this was not the case for both the recA and competence gene locus deletion strains, as no transformants were detected from either (Figure 2D). Thus, both deletions resulted in a minimum of 3-log reductions in transformability.

Given the apparent disparities in the rates of transformation between PCR products and gDNA, we were curious to directly compare transformation efficiencies between the two sources of DNA. Typically, if the source of DNA strongly impacts its transformability, it is because the organism can distinguish foreign DNA through its methylation patterns (Kobayashi, 2001) and/or through the presence of DNA uptake sequences (DUS) (Smith et al., 1999; Chen and Dubnau, 2003, 2004; Mell and Redfield, 2014) Thus, we extracted gDNA from a confirmed competence gene locus deletion mutant and compared its transformation efficiency with that of a PCR product harboring the same mutation. As shown in Figure 3A, gDNA transformation was extremely efficient with transformants comprising >0.1% of the total population. However, we observed a ~2-log lower transformation efficiency with a PCR product, despite the fact that the copy number of the PCR product was higher than the gDNA. This confirmed our previous observations, as both the pBSJL1 and PCR product transformations were far less efficient than gDNA transformation (Figures 1A, 2B,D). Since both the PCR product and gDNA targeted the same region of the chromosome and had identical sequences, it seemed unlikely that the disparate transformation efficiencies could be attributed to a DUS. For example, in Haemophilus influenza, Neisseria meningitidis, and Neisseria gonorrhoeae, DUS occur at an average frequency of 1/kb and even a single DUS is sufficient to facilitate uptake of both long and short DNA fragments (Mell and Redfield, 2014). If V. parvula exhibited a similar bias for DUS throughout its genome, we would expect our mutagenesis construct to contain at least one DUS, since it contained nearly 2 kb of sequence identical to the chromosome. To exclude the possibility that construct design was the source of the lower transformation efficiencies of PCR products, we created a series of additional mutations within the competence gene locus using PCR products harboring either small (~0.9 kb) or large (>2.5 kb) homologous fragments and creating either small (~0.4 kb) or large (~5.7 kb) deletions within the locus. Interestingly, in all cases, the PCR products yielded significantly fewer transformants than their corresponding gDNAs with most constructs yielding similar 10-fold lower transformation efficiencies relative to gDNA, indicating that construct design had little impact on the results (Figure 3A). The only exception was the first construct, which contained both small homologous fragments and created a large chromosomal deletion. It yielded an additional 10-fold reduction in transformation efficiency. Thus, construct design could be a partial contributor to the exceptionally poor transformability of this specific construct. We also repeated this experiment using our recA deletion construct and observed a similar preference for gDNA, which indicated that this phenomenon is unlikely to be locus-specific (data not shown). Since construct design did not appear to be the primary source of the V. parvula preference for genomic DNA, we were next interested to test the role of DNA methylation. If methylation were indeed a factor, we would predict significantly higher rates of transformation from plasmids purified from V. parvula vs. E. coli, due to differences in host methylation patterns. However, there appeared to be no obvious difference in the transformation efficiencies using plasmids purified from either E. coli or V. parvula. We did note substantially lower overall transformation efficiencies for pBSJL1 transformations compared to PCR products and gDNA (Figure 3B). Thus, for unknown reasons, V. parvula natural competence exhibits a clear hierarchy of preference for the source/type of DNA with plasmids being the least efficiently transformed and genomic DNA being the most efficient.

Given the relatively low efficiency of plasmid transformations in V. parvula, we decided to revisit our original natural competence screen of V. parvula isolates, as those transformations had utilized pBSJL1. Interestingly, when retesting these isolates using gDNA extracted from an SKV38 mutant, we were able to reliably transform more than half of our V. parvula isolates (Table 3). This suggested that natural competence is actually quite common among V. parvula strains, even when using low-passage clinical isolates. Consequently, we were curious whether similar competence gene loci might be found among other veillonellae. Indeed, this did prove to be the case, as the genes required for DNA uptake in V. parvula are widely conserved within the Veillonellaceae family. All of the putative competence loci found among these organisms exhibit synteny and encode two copies of traffic NTPases with the second copy bearing homology to pilT genes (Figure 4), which is highly suggestive of a similar role in DNA uptake.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.