The groEL sequences of S. parasanguinis strains and closely related Streptococcus species (Glazunova et al., 2009) were downloaded from the Chaperonin Sequence Database in March 2017¹ and subjected to a multiple alignment with ClustalX version 2.1 provided by the European Bioinformatics Institute². The region of 552 bp of the complete groEL gene named universal target (UT) sequences was used to manually identify the discriminative nucleotides. Three primers, two forward primers Spa146f and Spa93f, and one reverse primer Spa525r were designed, and they composed two primer pairs, Spa146f-Spa525r and Spa93f-Spa525r (Table 1).

The genomes of 30 S. parasanguinis strains deposited in the NCBI genome collection database³ were downloaded. The sequence of each designed primer was blast against the 30 S. parasanguinis genomes using NCBI Basic local alignment search tool⁴. Each primer was also blast against all the groEL gene sequences of the Chaperonin Sequence Database¹ online with the “Primer Blast” program, and then against all sequences in NCBI nucleotide collection (nr/nt) database with the blastn tool⁴.

Eight hundred and sixty six groEL universal target (UT) sequences of 64 Streptococcus spp., which included 8 sequences of 7 S. parasanguinis strains, were downloaded from the Chaperonin Sequence Database¹ as templates. The sequences of these templates and each of the two primer pairs, Spa146f-Spa525r and Spa93f-Spa525r, were introduced into SPCR (Cao et al., 2005). The product amplification coefficient, increase of which corresponds to the enhancement of the annealing temperature in experimental PCR, was set to 0.80 (the recommended value according to the SPCR developer) and 0.90, respectively, for individual primer pair, and the SPCR algorithm output the template sequences that can be amplified by the tested primer pair under each coefficient.

Strains S. parasanguinis F278 and S. salivariusF286 were previously isolated from human breast milk in our laboratory, and the accession numbers of the 16S rRNA gene sequence of the two strains were KY038191 and KY038192, respectively (Wang et al., 2017). S. sanguinis ATCC 10556, S. mutans UA159 and S. gordonii ATCC 10558 were obtained from the American Type Culture Collection (ATCC). All strains were cultured in liquid M17 medium (Hopebiol, Qingdao, China) in an anaerobic chamber (DG500, Don Whitley Scientific, United Kingdom) at 37°C for 10 h.

The feces of 22 Chinese children, which contained varied abundance ofS. parasanguinis according to our previous metagenomic sequencing (Zhang et al., 2015) were used in the present study. These 3 to 6-year-old children were diagnosed with Prader–Willi syndrome or simple obesity, and were recruited previously to study the role of gut microbiota in genetic and simple obesity (Zhang et al., 2015). The cohort study was performed under the approval of the Ethics Committee of the School of Life Sciences and Biotechnology, Shanghai Jiao Tong University (No. 2012-016). Written informed consents were obtained from the guardians of the children.

The saliva DNA of 28 newly diagnosed periodontitis patients and 26 periodontally healthy volunteers, which were stored at −80°C after extraction (Chen et al., 2015), were used in the present study. The periodontitis patients aging 29–67 years and the periodontally healthy volunteers aging 21–55 years were all Chinese Han people, and recruited previously to compare the oral microbiota composition by doing illumina sequencing of 16S rRNA gene V3-V4 region (Chen et al., 2015). This study was approved by the Ethics Committee of Shanghai Ninth People’s Hospital affiliated to Shanghai Jiao Tong University, School of Medicine, China (Document No. 201262). Written informed consent was obtained from all the participants.

Genomic DNA from the bacterial cultures was extracted as described previously (Wang et al., 2017). DNA extraction from fecal samples was conducted as previously described (Godon et al., 1997). Both exaction processes included chemical lysis and bead beating to break the bacterial cells. The integrity of the DNA was assessed by using 0.8% agarose gel electrophoresis gels stained with ethidium bromide, and the concentration was quantified with PicoGreen fluorescent dye (Thermo Fisher Scientific, Sunnyvale, CA, United States) by using SpectraMax M5 microplate reader (Molecular Devices, San Francisco, CA, United States).

The two designed primer pairs, Spa146f-Spa525r and Spa93f-Spa525r, were respectively used to do experimental PCR under gradient annealing temperatures using the genomic DNA of S. parasanguinis F278, S. salivarius F286, S. sanguinis ATCC 10556, S. mutans UA159 and S. gordonii ATCC 10558, respectively, as template. The PCR program included an initial denaturing step of 5 min at 94°C, 30 cycles of 95°C for 30 s, a certain annealing temperature between 58 and 63°C for 20 s, and 72°C for 45 s, and a final extension at 72°C for 7 min. Each 25 μl PCR mixture contained 1 × PCR buffer, 2mM MgCl₂, 0.2 mM of each dNTP, 0.25 μM of each primer, 0.2 U Taq polymerase (TaKaRa, Dalian, China), and 20 ng of template DNA. Amplifications were performed with the ABI PCR thermal cycler (Applied Biosystems, United States). PCR products were assessed by electrophoresis on 1.5% (wt/vol) agarose gel.

PCR was performed with primer pair Spa146f-Spa525r and Spa93f-Spa525r, respectively, using the saliva DNA of each of six human volunteers as template. The 25 μl PCR mixture contained 1 × PCR buffer, 2 mM MgCl₂, 0.2 mM of each dNTP, 0.25 μM of each primer, 0.2 U Taq polymerase (TaKaRa, Dalian, China), and 2 μl saliva DNA. The PCR program was as follows: 94°C for 5 min; 30 cycles of 95°C for 30 s, 61°C and 62°C for Spa146f-Spa525r and Spa93f-Spa525r, respectively, for 20 s, and 72°C for 45 s; and finally 72°C for 7 min. The size and specificity of PCR products were checked by electrophoresis on 1.5% (wt/vol) agarose gel.

The products of individual PCR reactions were excised from the 1.5% agarose gel and purified using the Gel Extraction Kit 200 (Omega, United States) as recommended by the manufacturer. Purified amplicons were ligated into the pGEM-T easy vector (Promega, Madison, WI, United States), and then transformed into competent E. coli DH5α cells. From each library, 15 recombinant clones were randomly selected and sequenced (Life Technologies, Shanghai, China).

The cloned sequences were blast against the nr database of GenBank using the basic local alignment search tool (BLAST)⁵ to determine the closest relative bacteria species. The cloned sequences and the reference groEL gene of Streptococcus spp. were aligned, and the reference gene sequences were trimmed to the same length of clone sequences. Neighbor-joining phylogenetic trees containing the cloned sequences and the trimmed groEL gene of Streptococcus spp. were constructed with the Molecular Evolutionary Genetics Analysis package 7 (MEGA 7) using the Maximum Composite Likelihood method. The phylogenetic robustness was assessed by bootstrap analysis with 1000 replicates.

Quantitative real-time PCR (qPCR) was performed with primer pair Spa93f-Spa525r and Spa146f-Spa525r, respectively, for each human fecal or saliva sample. The PCR was done in 96-well optical plates on LightCycler^® 96 Real-Time PCR System sequence detector (Roche, United States). The 20-μl reaction mixture contained 1 × FastStart SYBR green I PCR Mix (iQ^TM SYBR^® Green, Bio-Rad, United States), 0.5 μM of each primer, and 2 μl fecal/saliva DNA template. The PCR program was as follows: 94°C for 5 min; 40 cycles of 95°C for 30 s, 61°C and 62°C for Spa146f-Spa525r and Spa93f-Spa525r, respectively, for 20 s, 72°C for 45 s, and melting temp 82°C for 5 s for fluorescence detection. To confirm the specificity of the PCR reaction, melting curve analysis was performed after amplification by increasing the temperature at a rate of 0.2°C per second from 65°C to 97°C with continuous fluorescence measurement. PCR reactions were performed in triplicate. Two recombinant plasmids, SH3-7-146f and SH3-9-93f, which were picked from the clone libraries of the S. parasanguinis groEL gene of human saliva, were used to construct the standard curve for the qPCR with primer pair Spa146f-Spa525r and Spa93f-Spa525r, respectively. The S. parasanguinis groEL gene copy number was quantified using standard curves constructed from known concentrations of the plasmid DNA ranging from 5 × 10¹ to 5 × 10⁸ copies/μl. The abundance of S. parasanguinis was expressed as copies/ng DNA.

To determine the quantitative limit of the qPCR with the primer pairs, three fecal samples, which was collected from three obese children respectively and contained no S. parasanguinis according to metagenomic sequencing (Zhang et al., 2015), were used in the spiking experiments. Aliquots of 200 mg feces were spiked with 10-fold serial dilutions of the culture of S. parasanguinis F278 strain ranging from 2 to 9 log₁₀ cells. The concentrations of S. parasanguinis in the spiked samples were numerated by plate counting on M17 agar medium in triplicate. DNA was extracted from spiked feces as described above.

Data statistics was done with GraphPad Prism version 6.0. Shapiro-Wilk test was used to check the normal distribution of the data. Mann–Whitney U test was used to compare the abundances of saliva S. parasanguinis between periodontitis patients and orally healthy people. The Pearson’s correlation test was used to examine the correlation of S. parasanguinis abundances in fecal or saliva quantified with different techniques.

The partial groEL gene sequences of S. parasanguinis cloned from human saliva were deposited in GenBank under accession numbers MK608386 – MK608559, MK616660 – MK616661 and MK637615 – MK637617.

Primer Spa146f, Spa93f, and Spa525r showed 100% similarity with the groEL gene of all 7 S. parasanguinis strains deposited in Chaperonin Sequence Database⁶ (Table 2 and Supplementary Tables S1–S3). Besides, we downloaded the genome sequences of 30 S. parasanguinis strains from NCBI genome database⁷, and compared the primer sequences to the groEL gene sequence of each of the 30 S. parasanguinis strains. Spa146f 100% matched with 28 of 30 S. parasanguinis strains, and showed one base mismatch with 2 strains in the middle position closer to 5′ terminal where the mismatches had no significant effect on PCR amplification (Kwok et al., 1990). Spa93f fully matched with 18 of 30 S. parasanguinis strains, and had one base mismatch with three strains in the middle position or 5′ end, and had one base mismatch with nine strains at 3′ terminal. Spa525r 100% matched with 27 S. parasanguinis strains, and had one base mismatch with three strains in the middle position close to 5′ terminal (Supplementary Tables S1–S3).

Meanwhile, the sequences of the three primers were compared to the groEL gene of Streptococcus species other than S. parasanguinis (Table 2). The two forward primers, Spa146f and Spa93f, showed multiple mismatches with the sequences of non-S. parasanguinis strains within the last five bases at the 3′ terminal of the primer where the mismatches significantly prevent the amplification of non-specific sequences (Kwok et al., 1990). The reverse primer, Spa525r, 100% matched with the groEL gene of two (but not all) S. oralis strains and two S. alactolyticus strains, but had multiple mismatches with other non-S. parasanguinis strains.

The three primers were respectively blast against the groEL gene sequences of the Chaperonin Sequence Database⁶ using the Primer Blast tool of the database, and this tool calculated the score value by taking into account of identical bases and gaps of two aligned sequences to reflects the identity of the primer sequence with reference sequences (Altschul et al., 1990). The higher the score is, the more identical the primer is with the reference sequence. The two forward primers, Spa146f and Spa93f, had highest score (46 and 40, respectively) with only S. parasanguinis, and showed significantly lower scores (no higher than 32 and 34, respectively) with non-S. parasanguinis bacteria. The reverse primer, Spa525r, had the highest score 46 with S. parasanguinis strains and two S. alactolyticus strains, and lower scores (no higher than 44) with other bacterial strains.

Each of the three primers was then subjected to online BLAST analysis against the DNA sequences of the NCBI nucleotide collection (nr/nt) database with the NCBI blastn tool, and the alignment score values were also given to evaluate the identity of the primer sequence with the reference sequences. In accordance with the results of Primer Blast in Chaperonin Sequence Database, Spa146f and Spa93f had highest score (46.1 and 40.1, respectively) with only S. parasanguinis but significantly lower scores (no higher than 36.2 for both primers) with non-S. parasanguinis bacteria; and Spa525r had the highest score 46.1 with S. parasanguinis strains, two S. alactolyticus strains, and two S. oralis strains, and showed much lower scores (no higher than 38.2) with other non-S. parasanguinis strains.

The SPCR algorithm predicts in silico whether individual pairs of PCR primers produce amplicons with template DNA sequences at certain product amplification coefficient, increase of which corresponds to the increase of annealing temperature of actual PCR (Cao et al., 2005). The 866 groEL gene sequences of 64 Streptococcus species, including the 8 sequences of 7 S. parasanguinis strains, were downloaded from the Chaperonin Sequence Database, and input as templates into SPCR algorithm. According to the prediction of SPCR, Spa146f-Spa525r amplified only S. parasanguinis sequences at the product amplification coefficient 0.8, while Spa93f-Spa525r did so only when the product amplification coefficient was increased to 0.9 (Supplementary Table S4 and Supplementary Figure S1). This suggests that the both primer pairs are capable to specifically amplify the groEL gene of S. parasanguinis under PCR conditions stringent enough, and that compared to Spa93f-Spa525r, Spa146f-Spa525r requires less stringent condition to achieve specific amplification.

Actual PCR was performed using each of the two primer pairs and the genomic DNA of strains belonging to S. parasanguinis, and S. salivarius, S. sanguinis, S. mutans, and S. gordonii that are common human commensal streptococci. Spa146f-Spa525r yielded amplicons of the expected size only for S. parasanguinis at annealing temperature 58 – 63°C, whereas Spa93f-Spa525r did so at annealing temperatures no less than 62°C (Supplementary Figures S2, S3). Spa93f-Spa525r produced an amplicon with the genomic DNA of S. salivarius as templates at annealing temperature 59 – 62°C, but it was about 100 bp smaller in size than S. parasanguinis amplicon (Supplementary Figure S3). The lower annealing temperature of Spa146f-Spa525r compared to Spa93f-Spa525r for specific amplification of S. parasanguinis under the conventional PCR condition is in accordance with the SPCR prediction. However, Spa93f-Spa525r did not generate any amplicon with the S. salivarius genomic DNA as templates in qPCR assays in which the annealing temperature was 62°C (Supplementary Figure S4).

In order to further evaluate the specificity of the two primer pairs, we constructed clone libraries with PCR products amplified from the saliva DNA with primer pair Spa146f-Spa525r and Spa93f-Spa525r, respectively, for each of 6 people (Supplementary Figure S5). Saliva samples were used because they were reported to harbor diverse and abundant Streptococcus spp. (Sakamoto et al., 2000; Belstrom et al., 2017).

Fifteen clones were randomly selected from each library and sequenced. The Blastn analysis against the DNA sequences of the nucleotide collection (nr/nt) database of NCBI showed that the nearest neighbor bacteria of all the sequenced clones obtained with Spa146f-Spa525r and Spa93f-Spa525r were S. parasanguinis, with the similarity being 95–100%.

Phylogenetic trees were constructed with the cloned sequences obtained with primer pair Spa146f-Spa525r and Spa93f-Spa525r, respectively (Figures 1, 2). Regardless of the primer pair, all cloned sequences clustered with known S. parasanguinis strains, and separated from other Streptococcus spp. In both trees, the S. parasanguinis sequences, including cloned sequences and reference sequences of known S. parasanguinis isolates, formed different lineages, and the cloned sequences from the same human subjects distributed in at least two different lineages (Figures 1, 2, and Supplementary Figure S6).

For each of the six people, the sequences cloned with both primer pairs were aligned, and the Spa93f-Spa525r sequences were trimmed so that they are of the same length with Spa146f-Spa525r sequences, and then the processed sequences of both primer pairs were used to build the phylogenetic tree (Figure 3). For four people, the sequences cloned with the two primer pairs distributed among one another in the tree (Figures 3A–D). For one person (SP15), 10 of 15 Spa146f-Spa525r sequences and all Spa93f-Spa525r sequences distributed among one another, and five Spa146f-Spa525r sequences located in clusters distinct from others (Figure 3E). For only one person (SH10), the Spa146f-Spa525r sequences and Spa93f-Spa525r sequences formed separate lineages (Figure 3F). This indicates that in a few but not all humans, primer pair Spa146f-Spa525r and Spa93f-Spa525r may preferentially detect different S. parasanguinis strains.

Ten-fold serial dilutions of pGEM-T Easy plasmids containing partial groEL gene sequence of S. parasanguinis were used to generate the standard curve for primer pair Spa146f-Spa525r and Spa93f-Spa525r. In the range of 1 × 10² to 1 × 10⁹ copies per PCR, the standard curves for both primer pairs were highly linear with the R² > 0.99. The PCR efficiency was 79–96% for the two primer pairs.

To determine the lowest groEL gene copy number that can be detected with the two primer pairs in qPCR assay, fecal samples of three persons, which contained no S. parasanguinis as shown by fecal metagenomic sequencing (Zhang et al., 2015), were spiked with 10-fold serial dilutions of S. parasanguinis cells. The DNA extracted from the spiked feces was used as templates in qPCR with each of the two primer pairs. For both primer pairs, the quantification limit in feces was 5–6 log₁₀ groEL copies/g feces depending on the PCR efficiency. When the PCR efficiency was higher than 90%, the limit was 5 log₁₀ groEL copies/g feces. When the PCR efficiency was between 80 and 90%, the limit became 6 log₁₀ groEL copies/g feces.

Twenty-two fecal samples that contained varied amounts of S. parasanguinis according to our previous metagenomic sequencing (Zhang et al., 2015) were selected, and S. parasanguinis in these samples were re-quantified as groEL gene copies/ng DNA with qPCR using primer pair Spa146f-Spa525r and Spa93f-Spa525r, respectively (Supplementary Table S5). The melting curves of fecal qPCR products of both primer pairs showed there was no non-specific amplicon and only amplicons corresponding to S. parasanguinis were quantified at the fluorescence reading temperature 82°C (Supplementary Figures S7A,B). Because Spa93f-Spa525r produced an amplicon with the genomic DNA of S. salivarius as templates under convention PCR condition (Supplementary Figure S3), the fecal qPCR products of Spa93f-Spa525r were also run on 1.5% agarose gel to check the specificity, and there was no non-specific amplicon (Supplementary Figure S7C).

As shown by Pearson’s correlation analyses, the fecal S. parasanguinis abundances measured with either primer pair were well correlated with those obtained with metagenomic sequencing (r = 0.99 and p < 0.0001 for both Spa146f-Spa525r and Spa93f-Spa525r, Figures 4A,C), and data obtained with the two primer pairs also showed strong and significant correlation (r = 0.99 and p < 0.0001, Figure 4E). After the removal of the sample that had extraordinarily higher level of S. parasanguinis than any other samples, the correlations were still strong and significant among the results of metagenomic sequencing, qPCR with Spa146f-Spa525r, and qPCR with Spa93f-Spa525r (r was between 0.77 and 0.83, and p < 0.0001, Figures 4B,D,F).

qPCR using Spa146f-Spa525r and Spa93f-Spa525r, respectively, was performed to quantify the abundances of saliva S. parasanguinis in 26 healthy people and 28 periodontitis patients (Supplementary Table S6). The melting curves of saliva qPCR products of both primer pairs showed there was no non-specific amplicon and only amplicons corresponding to S. parasanguinis were quantified at the fluorescence reading temperature 82°C (Supplementary Figures S8A,B). The 1.5% agarose gel of saliva qPCR products of Spa93f-Spa525r showed no non-specific amplicon was generated (Supplementary Figure S8C). No matter which primer pairs were used, the results showed that healthy people had significantly higher level (2.7 fold for Spa146f-Spa525r and 4.2 fold for Spa93f-Spa525r) of saliva S. parasanguinis than periodontitis patients (Figures 5A,B).

The saliva S. parasanguinis quantity obtained with the two primer pairs showed strong and significant correlation (r = 0.98, p < 0.0001, Figure 5C).