Highly accurate measurement of the relative abundance of oral pathogenic bacteria using colony-forming unit-based qPCR

Abstract

Purpose

Quantitative polymerase chain reaction (qPCR) has recently been employed to measure the number of bacterial cells by quantifying their DNA fragments. However, this method can yield inaccurate bacterial cell counts because the number of DNA fragments varies among different bacterial species. To resolve this issue, we developed a novel optimized qPCR method to quantify bacterial colony-forming units (CFUs), thereby ensuring a highly accurate count of bacterial cells.

Methods

To establish a new qPCR method for quantifying 6 oral bacteria namely, Porphyromonas gingivalis, Treponema denticola, Tannerella forsythia, Prevotella intermedia, Fusobacterium nucleatum, and Streptococcus mutans, the most appropriate primer-probe sets were selected based on sensitivity and specificity. To optimize the qPCR for predicting bacterial CFUs, standard curves were produced by plotting bacterial CFU against Ct values. To validate the accuracy of the predicted CFU values, a spiking study was conducted to calculate the recovery rates of the predicted CFUs to the true CFUs. To evaluate the reliability of the predicted CFU values, the consistency between the optimized qPCR method and shotgun metagenome sequencing (SMS) was assessed by comparing the relative abundance of the bacterial composition.

Results

For each bacterium, the selected primer-probe set amplified serial-diluted standard templates indicative of bacterial CFUs. The resultant Ct values and the corresponding bacterial CFU values were used to construct a standard curve, the linearity of which was determined by a coefficient of determination (r²) >0.99. The accuracy of the predicted CFU values was validated by recovery rates ranging from 95.1% to 106.8%. The reliability of the predicted CFUs was reflected by the consistency between the optimized qPCR and SMS, as demonstrated by a Spearman rank correlation coefficient (ρ) value of 1 for all 6 bacteria.

Conclusions

The CFU-based qPCR quantification method provides highly accurate and reliable quantitation of oral pathogenic bacteria.

Graphical Abstract

INTRODUCTION

The oral cavity is the main habitat of a diverse array of microbes, including beneficial and pathogenic bacteria. A balanced interaction between oral microbes and the host is crucial for maintaining oral health. For example, excessive growth of red complex (Porphyromonas gingivalis, Treponema denticola, Tannerella forsythia) or orange complex (Prevotella intermedia and Fusobacterium nucleatum) has been believed to trigger the initiation of periodontitis [1,2,3,4], a chronic inflammatory disease that has affected up to 18.3% of the global population over the past decade [5,6]. In addition, excessive growth of Streptococcus mutans due to oral microbial dysbiosis leads to the development of dental caries [7,8]. Notably, several oral pathogens, including P. gingivalis and F. nucleatum, have been reported to enter the bloodstream through the gingival sulcus and circulate throughout the body, potentially causing systemic inflammation and chronic diseases [9].

The symptoms of periodontitis are typically assessed through radiographic evaluation and physical measurements of periodontal indices, including pocket probing depth (PD), clinical attachment level, bleeding on probing, and gingival index. However, these radiographic and clinical evaluations are invasive, time-consuming, and labor-intensive. As an alternative, periodontal risk can be assessed by quantifying oral microbes, since microbial biofilm-induced inflammation is an important epidemiological definition of periodontitis [10,11,12].

Several companies have already commercialized the quantitative polymerase chain reaction (qPCR) method for measuring oral bacteria due to its cost-effectiveness and efficiency. However, these methods have significant limitations. First, the standard curves are generated using a standard template that contains only a fragment of bacterial DNA, such as the 16S rRNA gene. This approach can lead to biased quantification of bacterial cells due to the multiple copies of the 16S rRNA gene present in bacterial genomes [11,13,14,15]. Second, the cut-off values for bacterial composition that predict the onset of oral diseases are based on absolute abundances [11,13,14,16,17,18,19,20]. This can cause inconsistencies between studies, as results for the absolute abundance are often affected by the methods used to collect samples, unlike relative abundance [21].

Here, we aimed to develop an optimized qPCR method for the accurate and reliable quantification of 6 representative oral pathogenic bacteria: P. gingivalis, T. denticola, T. forsythia, P. intermedia, F. nucleatum, and S. mutans.

MATERIALS AND METHODS

Bacterial cultures

P. gingivalis (American Type Culture Collection [ATCC] 33277), T. denticola (ATCC 35405), T. forsythia (ATCC 43037), P. intermedia (ATCC 25611), and F. nucleatum (ATCC 25586) were purchased from ATCC and anaerobically cultured on blood medium consisting of tryptic soy (30 g/L), yeast extract (5 g/L), sheep blood (50 mL/L), hemin (2.5 mg/L), cysteine HCl (125 μg/L), and vitamin K1 (0.5 μL/L). S. mutans (ATCC 25175) was also purchased from ATCC, and aerobically cultured as described by Park et al. [8].

qPCR

Genomic DNA (gDNA) was extracted using Bacteria Genomic DNA Isolation Kit (LaboPass, COSMO Genetech, Seoul, Korea) according to the given instructions. The sequences of primers and probes tested for each of 6 bacteria were oligomerized by Macrogen Corp. (Seoul, Korea) (Table 1). To evaluate the sensitivity of 3 sets of primers and probes for each bacterium, the limit of detection (LoD) was determined as the lowest amount of gDNA that can be detected by qPCR. The qPCR reactions were performed using a QuantStudio 3 (Thermo Fisher Scientific, Waltham, MA, USA) with the following conditions: Step 1, 95°C for 2 minutes (1 cycle); Step 2, 95°C for 1 minutes, 60°C for 30 seconds, and 72°C for 45 seconds (40 cycles); Step 3, 95°C for 5 minutes and 4°C until completion (1 cycle). The primers and probe used to detect total bacteria targeted conserved sequences on the 16S rRNA gene. The sequences for the forward primer, reverse primer, and probe were CCAGCAGCCGCGGTAATACG, CCGTCAATTCMTTTRAGTTT, and FAM-TACCAGGGTATCTAATCC, respectively.

Table 1. Primers and TaqMan probes tested to establish the new quantitative polymerase chain reaction system.

Bacterial strain	Set	Forward primer (5′-3′)	Reverse primer (5′-3′)	Probe (5′-3′)	Target gene	Amplicon size (bp)	Reference
P. gingivalis (ATCC 33277)	1a)	CTGCGTATCCGACATATC	GGTACTGGTTCACTATCG	AGACATCCTGTGTGAATTGGCG	23S	134	[18]
	2	TGCAACTTGCCTTACAGAGGG	ACTCGTATCGCCCGTTATTC	AGCTGTAAGATAGGCATGCGTCCCATTAGCTA	16S	344	[15]
	3	GCGCTCAACGTTCAGCC	CACGAATTCCGCCTGC	CACTGAACTCAAGCCCGGCAGTTTCAA	16S	68	[16]
T. denticola (ATCC 35405)	1	TGGTGAGTAACGCGTGGGTGACCT	TTCACCCTCTCAGGCCGGA	CCTGAAGATGGGGATAGCTAGTAGA	16S	204	[11]
	2a)	CCGAATGTGCTCATTTACATAAAGGT	GATACCCATCGTTGCCTTGGT	ATGGGCCCGCGTCCCATTAGC	16S	122	[19]
	3	TGAATACGTTCCTGGGCCTT	ACGGCTACCTTGTTACGACT	ACCGCCCGTCACACCATCCG	16S	143	This study
T. forsythia (ATCC 43037)	1	GGGTGAGTAACGCGTATGTAACCT	CCCATCCGCAACCAATAAA	CCCGCAACAGAGGGATAACCCGG	16S	127	[22]
	2a)	ATCCTGGCTCAGGATGAACG	TACGCATGCCCATCCGCAA	ATGTAACCTGCCCGCAACAGAGGGATAAC	16S	225	[17]
	3	GGTGTAGCGGTGAAATGCAT	TCCTGTTTGATACCCACGCT	CAGAACTCCGATTGCGAAGGC	16S	106	This study
P. intermedia (ATCC 25611)	1	CCACATATGGCATCTGACGTG	CACGCTACTTGGCTGGTTCA	ACCAAAGATTCATCGGTGGAGGATGGG	16S	232	[20]
	2a)	AGTCGAGGGGAAACGGCAT	CTTTGGTGGTCCACGTCAGATGC	AGACGGCCTAATACCCGATGTTGT	16S	160	[14]
	3	ACGGCCCTATGGGTTGTAAA	TAACAGACCGCCTACACTCC	ATTCCGTGCCAGCAGCCGCG	16S	177	This study
F. nucleatum (ATCC 25586)	1	GGATTTATTGGGCGTAAAGC	GGCATTCCTACAAATATCTACGAA	CTCTACACTTGTAGTTCCG	16S	162	[22]
	2	GGCTTCCCCATCGGCATTCC	AATGCAGGGCTCAACTCTGT	AGTTCCGCTTACCTCTCCAGTAC	16S	123	[18]
	3a)	ACCTTACCAGCGTTTGACAT	TGTAGCCCAGCGTATAAGGG	CCTAAAGACAGGTGGTGCATGGCT	16S	253	This study
S. mutans (ATCC 25175)	1	ACAGCTCAGAGATGCTATTCTTAA	ATTATTGAAGTGACGCCATACAC	AATGACGGTCGCCGTTATGAAAATGG	gtfB	124	[23]
	2a)	CAGCGATAAGACAGCCTATGCTAA	TGGTTTACCCGTTTGACTGGTT	CCGATTACCGTCTTTTGAACCGCACA	gtfD	76	[24]
	3	ATTCGAAGCAACGCGAAGAA	GCGGGACTTAACCCAACATC	ACCATGCACCACCTGTCTCCGA	16S	142	This study

ATCC: American Type Culture Collection.

^a)Selected primer-probe sets satisfying both sensitivity and specificity.

Calculation of colony-forming units (CFUs) for bacterial liquid cultures

For each of the 6 bacteria, liquid cultures were either diluted or concentrated until their absorbance at 600 nm reached a value of 1. These cultures were then diluted 10-fold 6 times. Subsequently, 100 μL from each of the first to sixth dilutions was cultured on an agar medium. The number of CFUs was counted on plates with dilutions that yielded between 30 and 200 CFUs. For each bacterium, the number of CFUs in 100 μL of liquid culture with an optical density at 600 nm (OD₆₀₀) of 1 was calculated by multiplying the count (x) by 10 raised to the power of the dilution factor (y).

Construction of a standard template and a standard curve

Each 100 μL of 6 liquid cultures with an OD₆₀₀=1 was combined to create a 600 μL mixture. Total bacterial gDNA was extracted from these mixtures as described by Park et al. [25] and weighted. Quantities of the gDNA (variable a) were equated to bacterial CFUs (variable b) as follows: Using P. gingivalis as an example, b = a × (P. gingivalis CFU in Total Mixtures)/(Total gDNA Amount in Total Mixture). A standard curve was constructed using a scatter plot where the x-axis represented bacterial CFUs and the y-axis represented the corresponding Ct values. The linearity of the standard curve was assessed by calculating the coefficient of determination (r²) using GraphPad Prism v5 (GraphPad Software Inc., San Diego, CA, USA).

Spiking study

The gDNA extracted from an anonymously collected mouthwash sample was spiked with 4 different quantities of the standard template (0, 25, 100, and 250 pg, respectively). The qPCR yielded the predicted CFU values corresponding to the spiked standard templates. Recovery rates were calculated by dividing the predicted CFU values by the true CFU values. We evaluated whether the recovery rates fell within the acceptable range of 85%–110%, as specified by the Association of Official Analytical Chemists (AOAC) International validation guidelines [26].

Shotgun metagenome sequencing (SMS)

Five mouthwash specimens were provided by the Biobank of Apple Dental Hospital, which is part of the Korea Biobank Network. These specimens were approved for use by the Korea National Institute for Bioethics Policy (approval number: P01-202111-31-002). Each specimen was treated with varying volumes of 6 bacterial liquid culture mixtures (0.0, 0.01, 0.1, 1, and 10 μL, respectively). gDNA was then extracted from each of the 5 samples and subjected to SMS (Theragen Health Corp., Seongnam, Korea). The quality and quantity of the extracted gDNA were evaluated using fluorometry (Qubit, Invitrogen, Waltham, MA, USA) and gel electrophoresis. Each sample, containing 100 ng of gDNA, was fragmented using acoustic shearing with a Q800R2 instrument (Qsonica Inc., Newtown, CT, USA). Following fragmentation, Illumina adapter sequences were annealed to the fragments, and PCR amplification was conducted. This was followed by a clean-up step using AMPureXP beads to remove impurities. The final library size was approximately 350 bp. Library quantification was carried out using the TapeStation 4200 instrument (Agilent Technologies, Santa Clara, CA, USA) and the KAPA Library Quantification Kit (KK4824, Kapa Biosystems, Wilmington, MA, USA). The prepared libraries were then pooled and loaded onto an Illumina flow cell for cluster generation. Sequencing was performed using 150 bp paired-end reads on an Illumina NovaSeq 6000 sequencer (Illumina, San Diego, CA, USA), following the manufacturer’s protocols. The raw sequencing data were processed as described by Lee et al. [27] with minor modifications. Briefly, adapter trimming was performed using Trimmomatic v0.38 [28] with the default options. Contaminated human nucleotide sequences were removed using Bowtie2 alignment against the GRCh38 reference genome [29]. Taxonomic profiles were obtained by using MetaPhlAn3 v3.1.0 against ChocoPhlAn databases [30].

Spearman rank correlation coefficient

In terms of the relative abundance of 6 bacteria, the relationship between qPCR and SMS was investigated by calculating Spearman rank correlation coefficient (ρ), a common method for examining the direction of the association between 2 variables [31]. The minimum sample size of 5 and the significance of correlations were calculated using GraphPad Prism v5.

Statistical analysis

All experiments were conducted in triplicate and subjected to statistical analysis to assess the significance of P values or 95% confidence intervals using GraphPad Prism v5.

RESULTS

Establishment of the qPCR method for measurements of 6 oral pathogenic microbes

We oligomerized 3 sets for each of 6 bacteria (P. gingivalis, T. denticola, T. forsythia, P. intermedia, F. nucleatum, and S. mutans) based on the literature and our original design (Table 1) and searched for primer-probe sets that exhibited excellent performance in both sensitivity and specificity. Each bacterium-extracted gDNA was serially diluted from 100 pg to 0.003 pg and subjected to qPCR using the primer-probe sets presented in Table 1, yielding scatter plots comprising log-scaled gDNA quantities on the x-axis and Ct values on the y-axis (Figure 1). The primer-probe set that detected the minimal LoD per bacterium was regarded as having the best sensitivity. To assess target-specificity, we tested each primer-probe set for potential cross reactivity with non-target bacterial gDNA (Table 2). Primer-probe sets that achieved detection of the minimal LoD and showed exclusive reactivity to their specific gDNA were marked with a footnote symbols (^a)) in Table 1. Next, to increase the efficiency of qPCR reactions, we consolidated 6 monoplex reactions into 2 multiplex reactions using FAM, JUN, and TAMRA dyes: P. gingivalis, F. nucleatum, and S. mutans in the first reaction; T. forsythia, T. denticola, and P. intermedia in the second reaction. We confirmed that there was no interference among the monoplex reactions during the multiplex reactions, as shown in Figure 2. The primer-probe set for detecting total bacteria remained in a monoplex reaction because its signal was interfered with by other primer-probe sets in a multiplexed format (data not shown).

Figure 1. Scatter plots showing the sensitivity of 3 primer-probe sets for (A) P. gingivalis, (B) T. denticola, (C) T. forsythia, (D) P. intermedia, (E) F. nucleatum, and (F) S. mutans. The log-scaled DNA quantities and the resultant Ct values are represented on the x- and y-axes, respectively. The coefficient of determination (r²) values were calculated using GraphPad Prism v5. The quantitative polymerase chain reaction reactions were repeated 3 times, and the error bars indicate 95% confidence intervals.

^a)For each bacterium, the primer-probe set satisfying both the minimal limit of detection and the highest r² value is indicated with black and marked with an footnote symbol (^a)).

Table 2. Specificity testing for the 3 primer-probe sets listed in Table 1 .

gDNA	Primer-probe set
	Pg			Td			Tf			Pi			Fn			Sm
	1	2	3	1	2	3	1	2	3	1	2	3	1	2	3	1	2	3
Pg	+	+	+			+
Td				+	+	+
Tf			+				+	+	+			+
Pi			+			+			+	+	+	+
Fn			+						+				+		+
Sm						+										+	+

The occurrence of gDNA polymerization by the respective primer-probe set was indicated by the symbol “+.”

gDNA: genomic DNA, Pg: P. gingivalis; Td: T. denticola; Tf: T. forsythia; Pi: P. intermedia; Fn: F. nucleatum; Sm: S. mutans.

Figure 2. Bar graphs comparing monoplex and multiplex qPCR reactions; one group of the multiplex consisted of (A) Pg, (B) Fn, and (C) Sm and the other group consisted of (D) Tf, (E) Td, and (F) Pi. For each bacterium, 2 quantities (100 pg and 0.2 pg) of gDNA were subjected to monoplex and multiplex qPCR, respectively, repeated 3 times. The differences between monoplex and multiplex were not significant according to GraphPad Prism v5.

qPCR: quantitative polymerase chain reaction; gDNA: genomic DNA; Pg: P. gingivalis; Td: T. denticola; Tf: T. forsythia; Pi: P. intermedia; Fn: F. nucleatum; Sm: S. mutans.

Construction of a standard template

To optimize qPCR for quantifying bacterial CFUs rather than bacterial DNA copy numbers, it was necessary to construct a standard template that reflects bacterial CFUs. Each bacterium was cultured in liquid, adjusted to an OD₆₀₀ of 1.0, serially diluted tenfold, and 100 µL of each dilution was cultured on agar anaerobically to yield 30 to 200 CFUs. The number of CFUs in 100 µL of liquid culture at an OD₆₀₀ of 1.0 was calculated by multiplying by the dilution factors, as shown in Table 3. The standard template was created by extracting gDNA from 600 µL of liquid culture mixtures, each containing 100 µL (OD₆₀₀=1) from 6 different bacteria. This process yielded 420 ng of total bacterial gDNA, representative of the CFU of each of the 6 bacteria, as indicated in Table 3.

Table 3. CFU determination of bacterial liquid cultures and DNA quantity of the standard template.

Variables	P. gingivalis	T. denticola	T. forsythia	P. intermedia	F. nucleatum	S. mutans
Dilution factor from (100 μL, OD600=1.0)	106	106	106	105	106	106
CFUs (100 μL, diluted)	99	80	41	290	59	220
CFUs (100 μL, OD600=1.0)	9.9×107	8.0×107	4.1×107	2.9×107	5.9×107	2.2×108
CFUs (600 μL, OD600=1.0)	5.28×108
DNA quantity of the standard template	420 ng

CFU: colony-forming unit, OD₆₀₀: optical density at 600 nm.

Construction of standard curves

The quantities of the standard template, serially diluted, corresponded to each bacterial CFU as presented in Table 4, following the equation described in the Materials and Methods section. For each bacterium, a standard curve was generated by plotting the log-scaled CFU values against the corresponding Ct values (Figure 3). For all standard curves, the coefficient of determination (r²) values exceeded 0.99, demonstrating a strong linear relationship between the CFU values and their corresponding Ct values.

Table 4. Serial dilutions of the standard template and the corresponding number of bacterial CFUs.

Standard template (pg)	No. of CFUs
	P. gingivalis	T. denticola	T. forsythia	P. intermedia	F. nucleatum	S. mutans	Total bacteria
500	117,857	95,238	48,810	34,524	70,238	261,905	628,571
250	58,929	47,619	24,405	17,262	35,119	130,952	314,286
100	23,571	19,048	9,762	6,905	14,048	52,381	125,714
25	5,893	4,762	2,440	1,726	3,512	13,095	31,429
1.562	368	298	152	108	219	818	1,964
0.195	46	37	19	13	27	102	245
0.097	23	18	9	7	14	51	122

CFU: colony-forming unit.

Figure 3. Standard curves of (A) P. gingivalis, (B) T. denticola, (C) T. forsythia, (D) P. intermedia, (E) F. nucleatum, (F) S. mutans, and (G) total bacteria. The coefficient of determination (r²) values and standard equations were calculated using GraphPad Prism v5. The quantitative polymerase chain reaction reactions were repeated 3 times, and the error bars indicate 95% confidence intervals.

CFU: colony-forming unit.

Validation of the optimized qPCR method for accuracy

To validate the accuracy of the newly optimized qPCR method, we conducted a spiking study according to International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use guideline Q2(R2) [32]. This involved spiking the standard template into a test template extracted from an anonymously collected mouthwash sample. Specifically, the test template was equally divided into 4 microtubes, each spiked with different quantities of the standard template (0, 25, 100, and 250 pg, respectively), and then subjected to the optimized qPCR. As indicated in Table 5, the ratios of predicted bacterial CFUs to true bacterial CFUs for the spiked standard templates ranged from 88.6% to 105.2%, which fell within the acceptable range of 85%–110% as stipulated by the international AOAC guidelines for validation [26].

Table 5. Recovery rates of the predicted CFU value to the true CFU value.

Microbe	Spiked (pg)	True CFU value	Predicted CFU value			Recovery rate
			1st	2nd	3rd	Mean (%)	RSD (%)
Pg	25	5,893	5,092	5,966	5,286	92.4	8.4
	100	23,571	23,474	25,134	24,647	103.6	3.5
	250	58,929	55,200	57,950	55,259	95.3	2.8
Td	25	4,762	4,742	4,485	4,305	94.7	4.9
	100	19,048	19,767	17,652	18,511	97.9	5.7
	250	47,619	45,707	48,984	47,286	99.4	3.5
Tf	25	2,440	2,340	2,597	2,762	105.2	8.3
	100	9,762	10,520	10,237	9,391	102.9	5.8
	250	24,405	25,397	23,669	23,833	99.6	3.9
Pi	25	1,726	1,682	1,791	1,843	102.7	4.6
	100	6,905	7,338	6,988	6,621	101.1	5.1
	250	17,262	17,024	17,721	16,223	98.4	4.4
Fn	25	3,512	3,140	3,282	3,471	93.9	5.0
	100	14,048	14,115	13,427	14,809	100.5	4.9
	250	35,119	35,761	32,645	36,083	99.2	5.5
Sm	25	13,095	11,925	11,696	11,199	88.6	3.2
	100	52,381	48,554	51,704	53,722	98.0	5.1
	250	130,952	122,353	131,832	137,087	99.6	5.7

The recovery rates were calculated 3 times independently. RSD was calculated by the formula as follows: Standard Deviation/Mean×100.

CFU: colony-forming unit, RSD: relative standard deviation, Pg: P. gingivalis; Td: T. denticola; Tf: T. forsythia; Pi: P. intermedia; Fn: F. nucleatum; Sm: S. mutans.

Evaluation of the reliability of the optimized qPCR method

The reliability of the predicted bacterial CFU values was investigated by comparing the optimized qPCR method with SMS, which is considered the most accurate method for measuring the relative abundance of bacterial composition among metagenome methods that utilize next-generation sequencing technologies [27]. A 10 mL mouthwash sample donated by an adult volunteer was divided into five 2 mL samples, each supplemented with mixtures of 6 bacterial liquid cultures in a dose-dependent manner, as illustrated in Figure 4. One hundred nanograms of gDNA extracted from each sample were analyzed using both the optimized qPCR method and SMS. The predicted CFU of each bacterium obtained using the optimized qPCR method was divided by that of the total bacteria, yielding the percentage of relative abundance as shown on Table 6. SMS provided the percentage of relative abundance by dividing each bacterial read by the total bacterial reads (Table 6). The consistency between the optimized qPCR method and SMS for 6 bacteria was evaluated using Spearman rank correlation coefficient (ρ) values for each of the 6 bacteria. These values were all 1; this significant result indicated that the bacterial CFU values predicted by the optimized qPCR method were reliable (Figure 5).

Figure 4. Schematic diagram of the process preparing MW samples for quantitative polymerase chain reaction and shotgun metagenome sequencing. In total, 10 mL of MW was aliquoted into 5 samples of 2 mL of MW. Thereafter, each MW sample was added with mixtures of 6 bacterial cultures in a dose dependent manner (0, 0.01, 0.1, 1, 10 μL, respectively).

MW: mouthwash, gDNA: genomic DNA.

Table 6. The relative abundance values (%) of the colony-forming unit-based optimized qPCR method and SMS.

Species	Methods	Sample					Spearman ρ	P value
		A	B	C	D	E
P. gingivalis	qPCR	0.00451	0.00549	0.01376	0.12933	0.72873	1.0	P<0.05
	SMS	0.00040	0.00152	0.01230	0.15608	0.92915
T. denticola	qPCR	0.04433	0.05003	0.05227	0.13179	0.77624	1.0	P<0.05
	SMS	0.01229	0.01830	0.02158	0.12287	0.79179
T. forsythia	qPCR	0.06312	0.05687	0.06204	0.13493	0.40448	1.0	P<0.05
	SMS	0.03889	0.03137	0.03764	0.12068	0.44531
P. intermedia	qPCR	0.08089	0.08201	0.09074	0.11629	0.30744	1.0	P<0.05
	SMS	0.06751	0.06871	0.07202	0.12910	0.34329
F. nucleatum	qPCR	0.20182	0.19748	0.21998	0.22463	0.72818	1.0	P<0.05
	SMS	0.18843	0.15374	0.19252	0.23474	0.78497
S. mutans	qPCR	0.00600	0.01064	0.01998	0.14616	1.70013	1.0	P<0.05
	SMS	0.00066	0.00569	0.01819	0.17320	2.05461

Spearman rank correlation coefficient (ρ) and P values were calculated using GraphPad Prism v5.

qPCR: quantitative polymerase chain reaction, SMS: shotgun metagenome sequencing.

Figure 5. Spearman rank correlation coefficient (ρ) values between the CFU-based optimized qPCR and SMS were calculated for (A) P. gingivalis, (B) T. denticola, (C) T. forsythia, (D) P. intermedia, (E) F. nucleatum, and (F) S. mutans. The qPCR reactions were repeated 3 times, and the significance of the Spearman rank correlations was calculated with the significance threshold of P<0.05 using GraphPad Prism v5.

CFU: colony-forming unit; qPCR: quantitative polymerase chain reaction; SMS: shotgun metagenome sequencing.

To strengthen the findings regarding the significance of the optimized qPCR method, we also investigated a previously reported qPCR method that calculates bacterial DNA copy numbers [11,20,33] instead of bacterial CFUs, according to its own materials and methods using the same 5 mouthwash samples. The calculated relative abundance of the 6 bacteria was then compared to those obtained using SMS using Spearman rank correlation coefficients (Table 7, Figure 6). Four of the 6 bacteria namely, T. denticola, T. forsythia, P. intermedia, and F. nucleatum exhibited non-significant consistency between the bacterial DNA copy number-based qPCR and SMS, reflecting the significance of the currently developed optimized qPCR method that represents bacterial CFU values.

Table 7. The relative abundance (%) of the DNA copy number-based qPCR method and SMS.

Species	Methods	Sample					Spearman ρ	P value
		A	B	C	D	E
P. gingivalis	qPCR	0.10182	0.13758	0.32825	1.42812	8.04639	1.0	P<0.05
	SMS	0.00040	0.00152	0.01230	0.15608	0.92915
T. denticola	qPCR	0.24795	0.33667	0.31656	0.90106	11.96953	0.9	ns
	SMS	0.01229	0.01830	0.02158	0.12287	0.79179
T. forsythia	qPCR	0.07721	0.07584	0.10755	0.28144	1.20710	0.9	ns
	SMS	0.03889	0.03137	0.03764	0.12068	0.44531
P. intermedia	qPCR	0.26696	0.32914	0.30677	0.43373	1.36669	0.9	ns
	SMS	0.06751	0.06871	0.07202	0.12910	0.34329
F. nucleatum	qPCR	2.73155	3.76145	3.85632	3.88837	15.24100	0.9	ns
	SMS	0.18843	0.15374	0.19252	0.23474	0.78497
S. mutans	qPCR	0.01421	0.02497	0.06258	0.49047	6.68285	1.0	P<0.05
	SMS	0.00066	0.00569	0.01819	0.17320	2.05461

Spearman rank correlation coefficient (ρ) and P values were calculated by GraphPad Prism v5.

qPCR: quantitative polymerase chain reaction, SMS: shotgun metagenome sequencing, ns: not significant.

Figure 6. Spearman rank correlation coefficient (ρ) values between the DNA copy number-based qPCR and SMS were calculated for (A) P. gingivalis, (B) T. denticola, (C) T. forsythia, (D) P. intermedia, (E) F. nucleatum, and (F) S. mutans. The qPCR reactions were repeated 3 times, and the significance of the Spearman rank correlations was calculated with the significance threshold of P<0.05 using GraphPad Prism v5.

qPCR: quantitative polymerase chain reaction, SMS: shotgun metagenome sequencing; ns: non-significant.

DISCUSSION

Numerous recent studies have attempted to quantify oral pathogenic bacteria using qPCR methods [13,14,17,19,20,24,34], based on the hypothesis that qPCR might be an effective way to diagnose oral diseases such as periodontitis and dental caries. However, most studies employing qPCR have failed to demonstrate a clear relationship between microbial load and oral diseases [35,36]. These issues may be attributed to the lack of accuracy and reliability in qPCR measurements. In fact, variations in qPCR operational conditions, including primer-probe sequences and standard templates, have hindered the ability to conduct meta-analyses across studies. The present study aims to establish a gold standard for qPCR measurement of 6 oral pathogens: P. gingivalis, T. denticola, T. forsythia, P. intermedia, F. nucleatum, and S. mutans.

A study comparing the subgingival microbiome of 25 individuals with chronic periodontitis to 25 healthy individuals using 16S rRNA V1-V3 metagenome sequencing identified P. gingivalis, T. forsythia, T. denticola, and P. intermedia as ranking first, third, sixth, and tenth, respectively, among the bacteria most strongly associated with periodontitis [37]. Another study involving 29 chronic periodontitis patients and 29 healthy individuals, using 16S rRNA V1-V2 and V4 metagenome sequencing found that P. gingivalis, T. denticola, P. intermedia, and T. forsythia were ranked second, third, 10th, and 16th respectively, among the bacteria most strongly associated with periodontitis [38]. A further study comparing the subgingival microbiome of 24 peri-implantitis patients to 24 healthy individuals using SMS showed that P. gingivalis, T. forsythia, F. nucleatum, P. intermedia, and T. denticola were ranked first, third, fourth, and sixth, and seventh respectively, among the bacteria most strongly associated with peri-implantitis [39]. Although not yet published by our group, a recent study comparing the mouthwash microbiome of 70 periodontitis patients to 119 healthy individuals using 16S rRNA V3-V4 metagenome sequencing revealed that P. gingivalis, P. intermedia, T. denticola, and T. forsythia were ranked first, second, fourth, and 16th respectively, among the bacteria most strongly associated with periodontitis. Taken together, P. gingivalis, P. intermedia, T. denticola, T. forsythia, and F. nucleatum have been proven to be strongly associated with periodontitis and peri-implantitis, regardless of the methods used to collect samples and conduct metagenomic sequencing, reinforcing the relevance of targeting P. gingivalis, P. intermedia, T. denticola, T. forsythia, and F. nucleatum in the current study.

The purpose of the present study was to develop a new qPCR method for predicting bacterial quantities based on their CFU count, offering higher accuracy and reliability compared to another qPCR method that relies on bacterial DNA copy numbers. The primary achievement of this study was the construction of a standard template that reflects bacterial CFUs. This was facilitated by using an anaerobic culture system and an agar plate count method, both well-established in our laboratories. This approach addresses the issue found in other qPCR methods that quantify partial DNA, such as the 16S rRNA fragment, whose quantities do not accurately reflect bacterial cell numbers due to its redundancy. Furthermore, the new qPCR method, based on precise measurements of bacterial CFU, provided the relative abundance of 6 pathogenic bacteria, with results equivalent to SMS. This highlights the superiority of the optimized qPCR method over other methods that only yield absolute abundance. Two previous studies that investigated the correlation between the absolute abundance of P. intermedia and clinical PD values reported conflicting results; one study found a positive association [40], but another did not [34]. This suggests that relying solely on bacterial absolute abundance could lead to biased judgments in assessing the risk of oral diseases.

The newly optimized qPCR method described in this study can accurately and reliably predict bacterial CFU and relative abundance. The next step is to determine bacterial cut-off values using this optimized qPCR method to predict and diagnose oral diseases in comparison to healthy states.