Development and evaluation of a multi-target droplet digital PCR assay for highly sensitive and specific detection of Yersinia pestis

Background Plague, caused by the bacterium Yersinia pestis, is a zoonotic disease that poses considerable threats to human health. Nucleic acid tests are crucial for plague surveillance and the rapid detection of Y. pestis. However, inhibitors in complex samples such as soil and animal tissues often hamper nucleic acid detection, leading to a reduced rate of identifying low concentrations of Y. pestis. To address this challenge, we developed a sensitive and specific droplet digital polymerase chain reaction (ddPCR) assay for detecting Y. pestis DNA from soil and animal tissue samples. Methods Three genes (ypo2088, caf1, and pla) from Y. pestis were used to develop a multi-target ddPCR assay. The limits of detection (LoD), reproducibility, and specificity were assessed for bacterial genomic DNA samples. The ability of the assay to detect low concentrations of Y. pestis DNA from simulated soil and mouse liver tissue samples was respectively evaluated and compared with that of quantitative real-time PCR (qPCR). Results The results showed that the ddPCR LoDs ranged from 6.2 to 15.4 copies/reaction for the target genes, with good reproducibility and high specificity for Y. pestis. By testing 130 soil and mouse liver tissue samples spiked with Y. pestis, the ddPCR assay exhibited a better sensitivity than that of the qPCR assay used in the study, with LoDs of 102 colony forming units (CFU)/100 mg soil and 103 CFU/20 mg liver. Moreover, the assay presented good quantitative linearity (R2 = 0.99) for Y. pestis at 103–106 CFU/sample for soil and liver samples. Conclusion The ddPCR assay presented good performance for detecting Y. pestis DNA from soil and mouse tissue samples, showing great potential for improving the detection rate of low concentrations of Y. pestis in plague surveillance and facilitating the early diagnosis of plague cases.


Introduction
Plague is a zoonotic disease spread by the bacterium Yersinia pestis through rodent fleas [1].Human infection with Y. pestis usually occurs via direct contact with infected animals or through flea bites [2].Three major plague pandemics in history have caused millions of human deaths [3].The 2017 plague epidemic in Madagascar resulted in more than 2,400 plague cases and hundreds of deaths, indicating that the plague remains a major public health concern [4][5][6].Early detection and surveillance of Y. pestis prevalence in animals and the environment in natural epidemic foci is essential for preventing and controlling plague outbreaks [7].
The polymerase chain reaction (PCR) is widely used for the rapid and sensitive detection of Y. pestis [4,8].Most methods select specific genes located on virulence plasmids of Y. pestis as detection targets, such as the caf1 gene on the pMT1 plasmid and the pla gene on the pPCP1 plasmid [9][10][11].However, this strategy carries a risk of false negatives because the corresponding plasmids may be lost from Y. pestis cells.Y. pestis strains lacking the pPCP1 or pMT1 plasmid have been isolated in nature [12,13].To address this issue, researchers have developed detection methods targeting chromosome-encoded genes specific to Y. pestis [14,15].In 2021, the World Health Organization (WHO) revised the international definition of plague cases and recommended a combination of at least two different gene targets (including pla, caf1, ypo2088, and ypo2486) to detect Y. pestis by PCR-based methods; two of these genes must be positive for the laboratory confirmation of Y. pestis [16].
In general, the most common sample types encountered in plague surveillance are animal organs or tissues and contaminated environmental samples.These samples usually contain many PCR inhibitors (such as humic acid, heparin, and hemoglobin), which may affect the detection performance of PCR [17].In our previous study [18], we established a recombinase polymerase amplification assay coupled with the clustered regularly interspaced short PLOS NEGLECTED TROPICAL DISEASES palindromic repeats technique (RPA-CRISPR) to detect Y. pestis and evaluated its sample tolerance to soil, mouse blood, and mouse lung tissue.The detection sensitivity of tissue samples (10 5 colony forming units (CFU)/20 mg mouse lung tissue) was significantly decreased compared with that of soil samples (10 3 CFU/100 mg soil) and blood samples (10 2 CFU/100 μL mouse blood).We hypothesized that this reduction in sensitivity may be due to the inhibitors (such as heparin) from tissue samples that remain in the DNA extract.
As a new generation of PCR technology, droplet digital PCR (ddPCR) possesses the advantages of absolute quantification, high sensitivity, and strong sample tolerance over traditional PCR methods [19,20].In the ddPCR reaction, the sample is partitioned into over ten thousand separate droplets, and the fluorescence signal of each droplet is collected after PCR amplification.These individual droplets mitigate the influence of inhibitors on PCR amplification, as the positive signal can be still retained even if moderate PCR inhibition is present in the droplet [21,22].Therefore, ddPCR is suitable for highly sensitive and specific Y. pestis detection in complex sample matrices.In a previous study [23], researchers developed a ddPCR detection method for Y. pestis with a high sensitivity of ~10 2 CFU/100 mg soil.However, this assay only targeted one gene on the Y. pestis chromosome, which did not meet the latest WHO recommended guidelines for the plague detection.
In this study, we established a multi-target ddPCR assay to detect Y. pestis DNA using three target genes (ypo2088, caf1, and pla).The reaction can be performed in a single tube and requires a digital PCR instrument with two-color fluorescent channels.The performance of the assay was comprehensively evaluated, including detection sensitivity, reproducibility, quantitative accuracy, and specificity.Moreover, the assay's ability to detect low concentrations of Y. pestis DNA from soil and mouse liver tissue samples was assessed and compared with that of quantitative real-time PCR (qPCR).

DNA extraction
Y. pestis strain 201, harboring four plasmids (pPCP1, pCD1, pMT1, and pCRY), was used in the study.This strain belongs to the biovar Microtus and is highly virulent in mice but avirulent in humans [24,25].The bacteria were inoculated in 5 mL of Luria-Bertani (LB) medium at 26˚C for about 10 hours, with shaking at 200 r/min.500 μL Y. pestis culture was centrifuged at 8,000 × g for 5 min to collect the bacterial pellet.DNA was then extracted using the QIAamp DNA Mini kit (51304) (Qiagen, Germany) according to the manufacturer's instructions.Finally, DNA was eluted with 100 μL double distilled water (ddH 2 O).The DNA concentration was measured using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, USA).

Primers and probes
The primers and TaqMan probes for ypo2088 were obtained according to a previous report [26].The primers and TaqMan probes for the caf1 and pla genes were designed using the Primer Express software version 3.0 (Thermo Fisher Scientific).The sequences are shown in Table 1.Sangon Biotech (Shanghai, China) synthesized all primers and probes.
Next, 20 μL ddPCR reaction mixture and 70 μL Droplet generation Oil For Probe (Bio-Rad) were used to generate droplets with the QX200 ddPCR system (Bio-Rad).The droplets were amplified using a T100 Thermal Cycler (Bio-Rad) under the following conditions: initial denaturation at 95˚C for 10 min, 40 cycles of denaturation (94˚C, 0.5 min), and annealing/ extension (60˚C, 1 min), enzyme deactivation (98˚C, 10 min), and incubation at 12˚C for 30 min.After amplification, the products were detected and analyzed with a droplet reader in the QX200 ddPCR system.

Real-time PCR assay
For single-target qPCR, the reaction mixture (20 μL) contained 10 μL qPCR Probe SuperMix (Takara, Japan), 400 nM forward primer, 400 nM reverse primer, 200 nM probe, 2 μL DNA template, and ddH 2 O.The primer and probe sequences were the same as those listed in Table 1.
Real-time PCR was performed on a CFX96 Real-Time System (Bio-Rad) under the following conditions: 95˚C for 5 min, followed by 40 cycles of 95˚C for 10 s and 60˚C for 60 s.The cycle threshold (Ct) was determined using the single threshold method provided in the Bio-Rad CFX Maestro software (Bio-Rad).

Evaluation of the ddPCR assay using diluted DNA samples
The DNA templates were diluted to concentrations ranging from 1 ng/μL to 1 fg/μL and were detected using the ddPCR assay as described above.According to the Clinical and Laboratory Standards Institute (CLSI) EP17-A2 guideline, the limit of blank (LoB) for each target gene was calculated by testing 35 blank samples (ddH 2 O) (details in S1 Table ).A ddPCR readout above the LoB value is considered positive.The lowest DNA concentration that could be consistently detected (n = 3) was the limit of detection (LoD).The coefficient of variation (CV) of the repeated tests was calculated to evaluate reproducibility.The quantification curve was plotted using the logarithm of the genomic equivalent (GE) number of Y. pestis cells per reaction [Log 10 (GE /reaction)] as the x-axis and the logarithm of the ddPCR readouts [Log 10 (copy number/reaction)] as the y-axis.The coefficient of determination (R 2 ) was calculated using Origin software (OriginLab).

Detection of Y. pestis DNA from pure bacterial cultures
Y. pestis cells were washed three times and resuspended in sterile normal saline (NS) solution (0.9% NaCl).The bacterial number was determined with the plate counting method.Serial dilutions of Y. pestis in NS solutions were then prepared from ~10 CFU/mL to 10 8 CFU/mL.For each concentration, 0.5 mL bacterial suspension was subjected to DNA extraction using the QIAamp DNA Mini kit (51304) (Qiagen, Germany) as described above.The extracted DNA was eluted with 50 μL ddH 2 O. Finally, 2 μL DNA was used as the template and detected with the ddPCR and qPCR methods, respectively.The NS solution was used as the negative control (NC).Each test was repeated in triplicate.The LoB was obtained by testing the NS solutions without spiking Y. pestis.

Establishment of the multi-target ddPCR assay for Y. pestis detection
We established three single-target ddPCR reactions for the ypo2088, caf1, and pla genes of Y. pestis.The fluorescence amplitude of positive droplets gradually increased with increasing probe concentration from 100 to 1500 nM (Fig 1).Among the three targets, the amplification results of the pla gene presented the "rain" phenomenon (a subset of the droplets exhibits a faint fluorescence intensity, making the delineation between positive and negative droplets not clear) [28,29], although we performed multiple optimizations (such as adjusting the annealing temperature, the number of amplification cycles, and the temperature ramp rate in the cycling).To avoid possible interference, the labeling fluorescence of the pla probe was designed to be different from that of the ypo2088 and caf1 probes when establishing the multiplex ddPCR assay.To detect ypo2088 and caf1 simultaneously in one channel, we adopted the "amplitude-based multiplex strategy", where the final probe concentrations of ypo2088 and caf1 were 500 nM and 250 nM, respectively.Under these conditions, the difference in fluorescence amplitudes between ypo2088 and caf1 is significant enough to distinguish the two targets.Finally, the multi-target ddPCR assay for Y. pestis detection was optimized as follows: the pla probe was labeled with FAM at a final concentration of 250 nM; the caf1 and ypo2088 probes were both labeled with HEX at a final concentration of 250 nM and 500 nM, respectively.After optimization, the multi-target ddPCR assay could simultaneously detect three gene targets of Y. pestis in a single tube with a common two-color ddPCR instrument.

Sensitivity, linearity, and reproducibility of the ddPCR assay
By testing 35 blank samples (ddH 2 O), the LoBs for ypo2088, caf1, and pla were determined to be 11.4,6.4, and 2.2 copies/reaction, respectively (S1 Table ).Then, serially diluted Y. pestis genomic DNA solutions were tested in triplicate to assess detection sensitivity (Fig 2A and  2B).The results shows that the LoDs for ypo2088 and caf1 were 15.4 and 8.9 copies/reaction, respectively; while the LoD for pla was the lowest, reaching 6.2 copies/reaction (Fig 2C).The ddPCR assay also shows a good linearity (R 2 = 0.99) for quantifying these three gene targets, with dynamic ranges spanning four orders of magnitude (Fig 2D).In the dynamic range, all intra-assay CV values were within 15% (S1 Data), demonstrating good test reproducibility.The x-axis unit in Fig 2D, genomic equivalent (GE) number of Y. pestis cells per reaction, were estimated by the amount of genomic DNA template (pg) and the genome mass of Y. pestis (213 GEs/pg) [10].The copy numbers of ypo2088 and caf1 determined by ddPCR were more consistent with the estimated GE number, whereas the copy number of pla was approximately 10-fold higher than the GE number.This result is reasonable because the pPCP1 plasmid that harbors the pla gene is multi-copy in Y. pestis [30].

Specificity of the ddPCR assay
Genomic DNA (0.1 ng/μL) samples, from seven other bacterial species within the Yersinia genus and four other zoonotic pathogens, were tested using the developed ddPCR assay for Y. pestis detection.According to the results (Table 2), no positive droplets were observed for ypo2088, caf1, and pla genes in these eleven bacteria, Additionally, Y. pestis strains lacking the pMT1 or pPCP1 plasmid could still be detected and identified as Y. pestis by the ddPCR assay.However, for strains that were missing both pMT1 and pPCP1 plasmids, the assay failed as expected because only the chromosomal gene ypo2088 could be detected.For this assay, samples were considered positive for Y. pestis when at least two of the three target genes of Y. pestis were detected.

Detection of Y. pestis DNA from pure bacterial cultures by ddPCR and qPCR
Genomic DNA was extracted from serial dilutions (10 1 -10 8 CFU/sample) of pure Y. pestis cultures and detected with ddPCR.As shown in Fig 3A -3C, the detection sensitivity of Y. pestis was 10 2 CFU/sample (1 mL NS buffer) when targeting ypo2088 or caf1, whereas it was 10 1 CFU/sample when targeting the pla gene.Consequently, the lowest bacterial concentration that could be reliably detected by ddPCR was set at 10 2 CFU/sample, with at least two gene targets testing positive.The same batch of DNA samples was also subjected to qPCR analysis, which was established for the three genes using the same primer and probe sequences as the ddPCR assay.The detection sensitivity by qPCR was set at 10 3 CFU/sample, considering that the lowest detectable bacterial concentration when targeting ypo2088, caf1, and pla was 10 3 , 10 3 , and 10 CFU/sample, respectively (Fig 3D -3F).This result demonstrates a 10-fold

Detection of Y. pestis DNA from simulated environmental and biological samples by ddPCR and qPCR
We further evaluated the performance of the ddPCR assay for detecting Y. pestis DNA from simulated soil and mouse liver tissue samples, which were artificially contaminated with a serial numbers of Y. pestis cells.DNA was extracted and analyzed according to the method described above.The lowest amount of Y. pestis that could be detected in soil sample (100 mg) was 10 3 , 10 2 , and 10 CFU/sample when targeting ypo2088, caf1, and pla, respectively (Fig 4A -4C).For liver tissue sample (20 mg), the lowest amount detected was 10 3 , 10 3 , and 10 CFU/ sample, respectively.Consequently, the ddPCR detection sensitivity of Y. pestis in soil and liver tissue samples was set at 10 2 and 10 3 CFU/sample, respectively.In comparison, the single-target qPCR assay exhibited a lower sensitivity (10 4 CFU/sample) for both soil and liver tissue samples (S1 Fig) .Therefore, the ddPCR assay presented a 10-100 times improvement in sensitivity compared to the qPCR assays employed in the study.Given its high sensitivity, the actual positive rate of ddPCR (71.43%, 50/70) was higher than that of qPCR (57.14%, 40/70) when analyzing the combined 130 samples of Y. pestis-spiked soil and liver tissue samples (S2 and S3 Tables).Moreover, the ddPCR sensitivity for soil samples remained consistent with that of pure Y. pestis cultures (10 2 CFU/sample), demonstrating a good tolerance to soil.Although the sensitivity for liver tissue sample was reduced by 10 times, it was still better than that of qPCR.Furthermore, the quantitative accuracy of ddPCR was not significantly affected by the soil and

Discussion
In this study, we established a multi-target ddPCR assay for the simultaneous detection of three specific genes of Y. pestis, including ypo2088 (on the chromosome), caf1 (on the pMT1 plasmid), and pla (on the pPCP1 plasmid).For this assay, at least two positive gene detections are required to call a positive detection of Y. pestis, in order to enhance the assay specificity and avoid false negatives caused by plasmid deficiency.The performance of the assay was comprehensively evaluated; it exhibited low LoDs ranging from 6.2 to 15.4 copies/reaction for the targeted genes and displayed high specificity in identifying Y. pestis.Furthermore, its capacity to detect low concentrations of Y. pestis (10 2 -10 3 CFU/sample) DNA from soil and mouse liver tissue samples was also demonstrated, revealing a 10-100 times advantage in sensitivity over the qPCR method (10 4 CFU/sample) employed in the study.
For detecting Y. pestis, pla is the most sensitive gene marker among the three gene targets because the pPCP1 plasmid that encodes pla is multi-copy in Y. pestis [30].However, the detection of Y. pestis using only pla gene as the target is not reliable because this gene is not specific enough for Y. pestis [4,31].The presence of pla sequences has been found in other bacterial strains such as Amphibacillus jilinensis Y1 (100% similarity with the pla gene sequences from Y. pestis CO92, 939/939 identities), Citrobacter koseri (98.7% similarity, 927/939 identities), and Escherichia coli strain FHI29 (98.5% similarity, 925/939 identities).Consequently, the results for Y. pestis detection must be reported with caution when only pla gene is positive in samples.Considering that the pla gene encodes an important virulence factor (plasminogen activator) for Y. pestis, we still included this gene target in our ddPCR assay; at the same time, other two specific gene targets (ypo2088 and caf1) were included to ensure high specificity to Y. pestis.
Compared with earlier reported ddPCR and qPCR assays for Y. pestis detection, our assay provides the advantage of both comprehensive gene targets and high detection sensitivity (Table 3).The qPCR-based FilmArray Biothreat Panel test and GeneXpert tularensis-anthracis-pestis (TAP) assay [9,10] facilitate the sample-to-answer detection of Y. pestis; however, their target genes were only on the plasmids (pMT1 or/and pPCP1).The TaqMan Array Card (TAC) assay [11] developed for Y. pestis selects three distinctive gene targets and its sensitivity could reach 10 3 CFU/sample for mouse blood samples.Specifically, the above three qPCRbased assays for Y. pestis all pertain to blood samples and do not involve sensitivity evaluations for soil or rodent organ samples.Qu et al. [26] established two single-target qPCR assays (targeting the 3a sequence on the chromosome and the caf1 gene, respectively) and demonstrated a sensitivity of 10 4 CFU/200 mg soil for the detection of Y. pestis in soil samples.In contrast to this report, our ddPCR assay demonstrates an approximately 100-fold improvement in sensitivity, allowing for the detection of Y. pestis at a low concentration of 10 2 CFU/100 mg soil.
In this study, the performance of the ddPCR assay was evaluated using Y. pestis-spiked environmental soil and mouse liver samples, with pure culture samples as comparisons.The detection sensitivity was not significantly interfered with by the soil matrix (10 2 CFU/sample) but was 10-fold interfered with by the liver tissue matrix (10 3 CFU/sample).This result is similar to our previous study [18], which showed that lung tissue samples had a greater influence on PCR sensitivity than soil samples.In fact, the detection sensitivity of 10 2 CFU/sample is equivalent to approximately 4 CFU/reaction, as 2 μL of the 50 μL extracted DNA was used as the template.For such a small amount of DNA, a slight impact of residual inhibitors in the reaction is considered acceptable.Despite the interference, our method still displayed higher sensitivity than qPCR when detecting Y. pestis in these complex samples.
There are several limitations to our method.First, the detection primers and probes of the pla gene used in this study need further improvement in specificity [32]; other genes (such as pst) on the pPCP1 plasmid are an alternative choice to increase the specificity [33].Second, Y. pestis strains containing neither the pMT1 nor the pPCP1 plasmid may exist in nature.For these strains, our approach would not work because only one chromosomal target gene can be PLOS NEGLECTED TROPICAL DISEASES detected.Therefore, at least two different specific genes on the chromosome should be included in future ddPCR assays, such as ypo2486 [15] and ypo0392 [34].Moreover, while ddPCR presents higher sensitivity, it involves a more time-consuming process compared to qPCR approach, due to the inclusion of additional steps such as droplet formation and individual droplet signal analysis.Lastly, we estimated the direct cost of each DNA detection (for the amplification step only) in our laboratory: about 3 US dollars per qPCR test (overall cost for three gene targets) and 5 US dollars per ddPCR test.Therefore, reducing ddPCR assay costs and improving its availability in resource-limited environments are still warranted.
In conclusion, we developed a highly sensitive and specific ddPCR assay to detect Y. pestis DNA in this study.This approach can be applied to detect low concentrations of Y. pestis DNA from soil and animal tissue samples, which is of great potential for improving the detection rate of Y. pestis in plague surveillance and facilitating the early diagnosis of plague cases.

Fig 1 .
Fig 1. Establishment of the multi-target ddPCR assay.(A-C) Amplification plots and (D) comparisons of the fluorescence amplitude of single-target ddPCR mixtures containing probes of different concentrations.(E) Two-dimensional fluorescence scatter plot (2D plot) of the ddPCR assay for the simultaneous detection of three different gene targets.The x-axis represents the fluorescence amplitude in the FAM channel, and the y-axis represents the fluorescence amplitude in the HEX channel.https://doi.org/10.1371/journal.pntd.0012167.g001

Fig 2 .
Fig 2. Performance of the multi-target ddPCR assay for detecting diluted genomic DNA samples of Y. pestis.(A, B) Amplification plots of the ypo2088, caf1, and pla genes by ddPCR.The grey dots in the graph refer to the droplets containing no target gene.The green dots represent the droplets containing either ypo2088 or caf1 or both genes, as detailed in the graph.The blue dots represent droplets containing the pla gene.(C) LoB and LoD values of the three gene targets.(D) Quantification linearity of the assay targeting different genes.The x-axis refers to Log 10 (estimated GE number/reaction), and the y-axis refers to Log 10 (copy number/reaction) by ddPCR.Each test was repeated in triplicate.https://doi.org/10.1371/journal.pntd.0012167.g002

Fig 3 .
Fig 3. Performance of the multi-target ddPCR and single-target qPCR assays in detecting genomic DNA samples of Y. pestis.(A-C) Serial dilutions of Y. pestis ranging from 10 2 to 10 6 CFU/mL were subjected to DNA extraction and ddPCR detection.The x-axis refers to the bacterial concentration [Log 10 (CFU/ sample]), and the y-axis refers to the ddPCR readout [Log 10 (copies/reaction)].(D-F) The same batch of DNA samples were also tested by single-target qPCR assays.Each test was repeated in triplicate.https://doi.org/10.1371/journal.pntd.0012167.g003

Fig 4 .
Fig 4. Performance of the ddPCR assay in detecting Y. pestis DNA from soil and mouse liver tissue samples.(A-C) The ddPCR readouts of the detection of ypo2088, caf1, and pla, respectively.(D-F) Linearity and dynamic ranges of the ddPCR assay for quantifying Y. pestis in soil and liver tissue samples with ypo2088, caf1, and pla as the target gene, respectively.The x-axis refers to the bacterial concentration [Log 10 (CFU/sample)], and the y-axis refers to the ddPCR readout [Log 10 (copies/reaction)].Each test was repeated five times.https://doi.org/10.1371/journal.pntd.0012167.g004

pestis DNA from simulated liver samples
Loam soil samples were collected from a flower bed in local outdoor environment and equally divided into 65 portions of 250 mg each.Then, 100 μL NS solution containing Y. pestis cells was added to 35 of the samples to prepare soil samples spiked with different concentrations of Y. pestis (10-10 6 CFU/sample, five replicates for each concentration).100μLsterile NS solution was added to another 30 samples as NCs.According to the manufacturer's instructions, DNA was extracted using the TIANamp Soil DNA Kit (TIANGEN Biotech, China).Finally, the DNA was eluted with 50 μL ddH 2 O, and 2 μL was used as the template for the ddPCR and qPCR assays, as mentioned above.The corresponding LoB was obtained by testing soil samples spiked without Y.pestis.Detection of Y.Blood & Tissue Kit (69504) (Qiagen, Germany) according to the manufacturer's instructions.Finally, the DNA was eluted with 50 μL ddH 2 O, and 2 μL was used as the template for the ddPCR and qPCR assays.The corresponding LoB was obtained by testing tissue samples spiked without Y.pestis.

Table 3 . Comparison of the ddPCR assay and other PCR-based assays for Y. pestis. PCR-based assays for Y. pestis Detection sensitivity (CFU/sample) Sample type (volume/weight) Ref. chromosome caf1 pla a FilmArray Biothreat Panel test
The detection targets include Y. pestis and 15 other biothreat pathogens.The detection targets include Y. pestis, F. tularensis, and B. anthracis.The detection targets include Y. pestis and four other biothreat pathogens.https://doi.org/10.1371/journal.pntd.0012167.t003 a b c