INTRODUCTION
The prevalence of tick-borne illnesses and the pathogenic agents, which cause them are of growing concern. In the USA, the Centers for Disease Control estimate over 300 000 cases of Lyme disease per year, making it the most widespread arthropod-borne disease in the nation (CDC, 2015). Improved detection and diagnostic approaches for tick-borne diseases are sought for patient, public health and environmental surveillance. Droplet Digital™ PCR (ddPCR) (Bio-Rad Laboratories, Inc., Hercules, CA) is a promising technology for this task because it allows for sensitive, specific detection of single template molecules as well as precise quantification of target DNA.
Borrelia burgdorferi and Borrelia miyamotoi are infectious spirochetes (Tilly et al. Reference Tilly, Rosa and Stewart2008; Dietrich et al. Reference Dietrich, Schmidgen, Maggi, Richter, Matuschka, Vonthein, Breitschwerdt and Kempf2010; Krause et al. Reference Krause, Narasimhan, Wormser, Rollend, Fikrig, Lepore, Barbour and Fish2013; Vayssier-Taussat et al. Reference Vayssier-Taussat, Moutailler, Michelet, Devillers, Bonnet, Cheval, Hébert and Eloit2013) transmitted in the United States primarily by the black-legged deer tick Ixodes scapularis (Tilly et al. Reference Tilly, Rosa and Stewart2008; Vayssier-Taussat et al. Reference Vayssier-Taussat, Moutailler, Michelet, Devillers, Bonnet, Cheval, Hébert and Eloit2013). Borrelia burgdorferi is the causative agent of Lyme disease, whereas B. miyamotoi causes a form of relapsing fever (CDC, 2013, 2014a ). In recent years, several cases of B. miyamotoi infection have been found in areas where Lyme disease is endemic in the USA and other parts of the world (CDC, 2013, 2014a ; Gugliotta et al. Reference Gugliotta, Goethert, Berardi and Telford2013; Krause et al. Reference Krause, Narasimhan, Wormser, Rollend, Fikrig, Lepore, Barbour and Fish2013; Padgett et al. Reference Padgett, Bonilla, Kjemtrup, Vilcins, Yoshimizu, Hui, Sola, Quintana and Kramer2014).
The molecular diagnostic method currently employed for the detection of pathogens in ticks is endpoint PCR (Ullmann et al. Reference Ullmann, Gabitzsch, Schulze, Zeidner and Piesman2005; CDC, 2013, 2014b ; Padgett et al. Reference Padgett, Bonilla, Kjemtrup, Vilcins, Yoshimizu, Hui, Sola, Quintana and Kramer2014) or quantitative PCR (qPCR) (Hindson et al. Reference Hindson, Ness, Masquelier, Belgrader, Heredia, Makarewicz, Bright, Lucero, Hiddessen, Legler, Kitano, Hodel, Petersen, Wyatt, Steenblock, Shah, Bousse, Troup, Mellen, Wittmann, Erndt, Cauley, Koehler, So, Dube, Rose, Montesclaros, Wang, Stumbo and Hodges2011; Bio-Rad Laboratories, Reference Bio-Rad Laboratories2014). In one iteration, real-time detection of PCR products is achieved in each qPCR reaction through the attachment of a fluorescent reporter molecule to specific TaqMan® (Life Technologies, Carlsbad, CA) probes which, upon amplification of target DNA products, becomes cleaved and fluoresces. This fluorescence can be measured and a DNA quantity interpolated from a constructed standard curve. In essence, an increased amount of fluorescence corresponds to an increased amount of target DNA product (Bio-Rad Laboratories, Reference Bio-Rad Laboratories2014). Extrapolating backwards can yield an estimate of original DNA concentration.
Though qPCR has proven to be a relatively successful diagnostic tool, the methods can be time-consuming, costly and imprecise (Hindson et al. Reference Hindson, Ness, Masquelier, Belgrader, Heredia, Makarewicz, Bright, Lucero, Hiddessen, Legler, Kitano, Hodel, Petersen, Wyatt, Steenblock, Shah, Bousse, Troup, Mellen, Wittmann, Erndt, Cauley, Koehler, So, Dube, Rose, Montesclaros, Wang, Stumbo and Hodges2011; Staff, Reference Staff2012). Another drawback to using qPCR methods in diagnostics is that the accuracy of this technique is limited by the fact that amplification efficiencies and cycle threshold (C T) values generated can vary greatly between runs (Hindson et al. Reference Hindson, Ness, Masquelier, Belgrader, Heredia, Makarewicz, Bright, Lucero, Hiddessen, Legler, Kitano, Hodel, Petersen, Wyatt, Steenblock, Shah, Bousse, Troup, Mellen, Wittmann, Erndt, Cauley, Koehler, So, Dube, Rose, Montesclaros, Wang, Stumbo and Hodges2011). Unlike the relative measurements of DNA concentration obtained through qPCR methodologies, droplet digital PCR (ddPCR) enables the absolute quantification of target DNA molecules (Hindson et al. Reference Hindson, Ness, Masquelier, Belgrader, Heredia, Makarewicz, Bright, Lucero, Hiddessen, Legler, Kitano, Hodel, Petersen, Wyatt, Steenblock, Shah, Bousse, Troup, Mellen, Wittmann, Erndt, Cauley, Koehler, So, Dube, Rose, Montesclaros, Wang, Stumbo and Hodges2011; Bio-Rad Laboratories, Reference Bio-Rad Laboratories2014; Sze et al. Reference Sze, Abbasi, Hogg and Sin2014). This capability is a defining characteristic of ddPCR and is potentially advantageous to a wide variety of applications including molecular diagnostic testing (Hindson et al. Reference Hindson, Ness, Masquelier, Belgrader, Heredia, Makarewicz, Bright, Lucero, Hiddessen, Legler, Kitano, Hodel, Petersen, Wyatt, Steenblock, Shah, Bousse, Troup, Mellen, Wittmann, Erndt, Cauley, Koehler, So, Dube, Rose, Montesclaros, Wang, Stumbo and Hodges2011; Staff, Reference Staff2012; Bio-Rad Laboratories, Reference Bio-Rad Laboratories2014).
The ddPCR technique is based upon the partitioning of a single DNA sample into an emulsion containing tens of thousands of smaller individual reaction liposomes, known as ‘droplets’ (Bio-Rad Laboratories, Reference Bio-Rad Laboratories2014; Sze et al. Reference Sze, Abbasi, Hogg and Sin2014). Dilution of the sample into an appropriate range results in the majority of droplets containing zero or one target molecule. PCR is subsequently performed to amplify the target DNA template within each droplet. The droplets are then counted as either positive or negative based on the presence or absence of amplicons as determined by target-dependent fluorescence signals in individual droplets (Sze et al. Reference Sze, Abbasi, Hogg and Sin2014). The ‘digital’ aspect of the system refers to the simple readout of droplet partitions as a binary code of either ‘positive’ or ‘negative’. Since the presence of a target molecule in a given droplet is a random event, the associated data can be fit to a Poisson distribution, which allows estimation of the number of positive droplets containing more than one target molecule (Bio-Rad Laboratories, Reference Bio-Rad Laboratories2014; Sze et al. Reference Sze, Abbasi, Hogg and Sin2014). This allows for direct calculation of DNA copy number in a given sample without the obligation of a standard curve, which is an improvement upon qPCR requirements in terms of time, money and accuracy (Hindson et al. Reference Hindson, Ness, Masquelier, Belgrader, Heredia, Makarewicz, Bright, Lucero, Hiddessen, Legler, Kitano, Hodel, Petersen, Wyatt, Steenblock, Shah, Bousse, Troup, Mellen, Wittmann, Erndt, Cauley, Koehler, So, Dube, Rose, Montesclaros, Wang, Stumbo and Hodges2011; Staff, Reference Staff2012; Bio-Rad Laboratories, Reference Bio-Rad Laboratories2014; Sze et al. Reference Sze, Abbasi, Hogg and Sin2014).
Droplet partitioning during ddPCR reduces bias from PCR amplification efficiency and inhibitors, which reduces error rates and enables accurate quantification of DNA template (Bio-Rad Laboratories, Reference Bio-Rad Laboratories2014). Droplet partitioning also decreases effects of competitive amplification, allowing the sensitivity of template detection to increase by an order of magnitude when compared with qPCR under certain conditions (Hindson et al. Reference Hindson, Ness, Masquelier, Belgrader, Heredia, Makarewicz, Bright, Lucero, Hiddessen, Legler, Kitano, Hodel, Petersen, Wyatt, Steenblock, Shah, Bousse, Troup, Mellen, Wittmann, Erndt, Cauley, Koehler, So, Dube, Rose, Montesclaros, Wang, Stumbo and Hodges2011). Although the quality of the results yielded by ddPCR are distinct from those obtained through qPCR, the technology uses assay chemistries essentially identical to those widely used for qPCR applications (i.e. TaqMan), facilitating simple adaptation of existing qPCR protocols to ddPCR.
While ddPCR detection of certain pathogens may have notable advantages over existing technologies, its application in the detection of Borrelia spp. has not yet been validated in a controlled laboratory setting. In this study, we adapted previously published TaqMan-based qPCR assays for the detection of B. burgdorferi and B. miyamotoi to the ddPCR platform, in order to test the efficacy of the system (Ullmann et al. Reference Ullmann, Gabitzsch, Schulze, Zeidner and Piesman2005). We first determined the accuracy and sensitivity of each assay using purified bacterial genomic DNA (gDNA). We then investigated the potential for PCR inhibition from tick DNA using total I. scapularis DNA extracts that had been previously tested and found to be negative for Borrelia spp. infection. These whole-tick extracts were then spiked with known quantities of appropriate Borrelia gDNA and tested. Finally, we applied the technique to other ticks collected from the same heavily B. burgdorferi-infected population from Massachusetts and estimated the number of B. burgdorferi spirochetes contained within infected nymphs and adult females.
MATERIALS AND METHODS
Sample collection
Borrelia burgdorferi B31 (35210™) gDNA was obtained from the American Type Culture Collection (Manassas, VA). Borrelia miyamotoi strain US178 (Rhode Island) DNA was kindly provided by the Center for Disease Control and Prevention (CDC).
Ixodes scapularis ticks were collected in late May and early June of 2014 near North Truro, MA by standard flagging techniques. Live samples were transported to UNTHSC and frozen at −80 °C prior to extraction. Details on life stage and presence of B. burgdorferi DNA are shown in Table 1.
Table 1. Life stages and B. burgdorferi infection status of ticks used in this study
DNA extraction
DNA from ticks (n = 10) was extracted using previously described methods (Williamson et al. Reference Williamson, Billingsley, Teltow, Seals, Turnbough and Atkinson2010; Mitchell et al. Reference Mitchell, Williamson, Billingsley, Seals, Ferguson and Allen2016). All extractions were conducted in a pre-PCR clean room to limit risk of contamination. Briefly, ticks were laterally bisected with a sterile razor and pulverized by bead beating for 10 min, followed by digestion with proteinase K for 1 h. at 60 °C. Extractions were carried out using the E.Z.N.A.® Mollusc DNA Kit (Omega Bio-tek, Norcross, GA) per the manufacturer's instructions with an elution volume of 140 µL.
Conventional PCR and preliminary screening
Tick extracts were tested for amplifiable DNA using primers targeting the tick mitochondrial 16S rRNA gene as previously described (Mitchell et al. Reference Mitchell, Williamson, Billingsley, Seals, Ferguson and Allen2016). The presence of Borrelia was detected using a nested PCR assay, which targeted a portion of the flaB gene with Borrelia genus-specific primers. Reactions were performed in duplicate with positive (plasmid DNA) and no template controls (NTC) as previously described (Williamson et al. Reference Williamson, Billingsley, Teltow, Seals, Turnbough and Atkinson2010; Mitchell et al. Reference Mitchell, Williamson, Billingsley, Seals, Ferguson and Allen2016). Of the ten tick samples initially screened, two (of four total) negative and six positive ticks were utilized for later ddPCR experiments. All I. scapularis extracts were quantified prior to ddPCR analysis using the Qubit® dsDNA HS Assay Kit (ThermoFisher Scientific, Waltham, MA) to determine total concentration of double-stranded DNA present in each sample.
Genospecies-specific PCR primers and probes
Widely accepted qPCR primer design guidelines also apply to the design of ddPCR primers and probes (Bio-Rad Laboratories, Reference Bio-Rad Laboratories2014); therefore, Borrelia species-specific primer and probe sets used during this research were selected from a study by Ullmann et al. (Reference Ullmann, Gabitzsch, Schulze, Zeidner and Piesman2005). These included oligonucleotide primers and probe sequences specific for the B. burgdorferi ospA and the B. miyamotoi glpQ genes. Primer sequences specific for the B. burgdorferi ospA gene were MOspA-F (5′-GYAAAGTAAAATTAACART) and MOspA-R (5′-TGTTTTRCCATCTTCTTT), which yield a 74-bp amplicon. The TaqMan probe was synthesized as MBurg-P and labeled with a 5′ 6-FAM™ (blue) dye and a 3′ MGB/nonfluorescent quencher (MGBNFQ) (5′−6-FAM-GACGATCTAGGTCAAACC-MGBNFQ). Primer sequences for the hard tick relapsing fever group Borrelia (i.e. B. miyamotoi) glpQ gene were MglpQ-F (5′-GATAATATTCCTGTTATAATGC) and MglpQ-R (5′-CACTGAGATTTAGTGATTTAAGTTC), which yield a 100-bp amplicon. The TaqMan probe for relapsing fever group Borrelia was MglpQ-P (5′-VIC-CCCAGAAATTGACAACCAC-MGBNFQ) and labelled with a 5′ VIC® (green) dye. Sequences were evaluated for specificity, length of amplicon, secondary structuring from internal primer binding, G–C content and the melting temperature of primers and probes using the Oligonucleotide Properties Calculator (http://basic.northwestern.edu/biotools/OligoCalc.html) (Kibbe, Reference Kibbe2007).
Droplet digital PCR
ddPCR for both ospA and glpQ assay formats was performed with a PCR reaction volume of 20 µL using the ddPCR™ Supermix for Probes (no dUTP) master mix (Bio-Rad, Hercules, CA). Each reaction included 10 µL of ddPCR Probe Supermix, forward and reverse primers at 900 nm each, probes at 250 nm and template DNA. The PCR reaction mixture was loaded into an eight-well DG8™ Cartridge (Bio-Rad) and droplets were formed with the Bio-Rad QX200™ Droplet Generator, following the manufacturer's instructions. During emulsion, the QX200 droplet generator partitions the samples into 20 000 nanolitre-sized droplets. The droplets were then transferred to a 96-well plate and sealed with a Bio-Rad PX1™ PCR Plate Sealer, as recommended by the manufacturer.
Optimal ddPCR annealing temperatures for the ospA and glpQ assays were determined by incorporating a temperature gradient from 46 to 60 °C into the annealing–extension step of the thermal cycling conditions.
Borrelia burgdorferi ospA assays were amplified using the following cycling conditions: an initial denaturation step at 95 °C for 10 min, followed by 50 cycles consisting of denaturation at 94 °C for 30 s and an annealing–extension step at 49 °C for 1 min, followed by a final extension step at 98 °C for 10 min and a 4 °C indefinite hold. The overall ramp rate was set at 2 °C s−1.
For B. miyamotoi detection, the glpQ assay reactions were similarly amplified with the following cycling conditions: an initial denaturation step at 95 °C for 10 min, followed by 50 cycles of denaturation at 94 °C for 30 s and an annealing–extension step at 52 °C for 1 min, followed by a final extension step at 98 °C for 10 min and a 4 °C indefinite hold. Overall ramp rate was set at 2 °C s−1. After cycling, droplets were immediately analysed or stored at 4 °C until analysis on the QX200™ Droplet Reader.
Data acquisition and analysis
The QX200 Droplet Reader analyses each droplet individually, in a single-file fashion, using a two-dye, two-channel detection system. The blue dye channel (channel 1) detected ospA PCR products, while the green dye channel (channel 2) detected glpQ amplicons. Droplets were classified as positive or negative according to a threshold manually set across all wells within a single run based upon results of the NTC sample. Positive droplets contain at least one copy of the target DNA molecule and display increased fluorescent amplitude compared with the negative controls.
The number of positive and negative droplets read in each channel is used by the QuantaSoft™ v.1.7.4.0917 (Bio-Rad) software to calculate the concentration of the target DNA sequences, along with their Poisson-based 95% confidence intervals (Hindson et al. Reference Hindson, Ness, Masquelier, Belgrader, Heredia, Makarewicz, Bright, Lucero, Hiddessen, Legler, Kitano, Hodel, Petersen, Wyatt, Steenblock, Shah, Bousse, Troup, Mellen, Wittmann, Erndt, Cauley, Koehler, So, Dube, Rose, Montesclaros, Wang, Stumbo and Hodges2011). The number of template copies per unit volume was estimated from the number of positive events detected by the droplet reader in the corresponding channel (channels 1 and 2 for FAM™ and VIC® dyes, respectively), and the number of total droplets by maximum-likelihood (Strain et al. Reference Strain, Lada, Luong, Rought, Gianella, Terry, Spina, Woelk and Richman2013). The distribution of templates among droplets was assumed to follow a Poisson distribution, and the number of positive droplets was assumed to follow a binomial distribution. 95% confidence intervals were estimated under the same assumptions. The droplet size was assumed to be 0·91 nL (Strain et al. Reference Strain, Lada, Luong, Rought, Gianella, Terry, Spina, Woelk and Richman2013). The concentration reported by QuantaSoft™ equals copies of template per μL of the final 1 × ddPCR reaction. This value was multiplied by the reaction volume to calculate the total number of template copies detected per sample.
Limit of detection and absolute quantification
A 1:100 dilution of each control Borrelia spp. gDNA sample was made and total double-stranded DNA (dsDNA) concentration quantified via Qubit® dsDNA HS Assay Kit (ThermoFisher Scientific). From this, a 7-sample standard dilution series was generated to test detection limits and quantification accuracy via ddPCR. Serial dilutions of control DNA were prepared separately for each Borrelia spp. in order to determine the sensitivity of the ddPCR instrument and protocol for each target species. Serial dilutions ranged from 10 to 100 000 copies per reaction of control B. burgdorferi DNA, and 6 to 165 000 copies per reaction of control B. miyamotoi DNA. Fresh dilutions were prepared daily for each experiment. PCRs were prepared according to the ddPCR conditions previously detailed with an added template volume of 1 µL diluted Borrelia control DNA. Expected genomic copies were calculated at each sample concentration according to the equation below. Expected copies were then compared with measured copies generated by the QuantaSoft™ (Bio-Rad) software in order to evaluate the ability of the ddPCR system to provide absolute quantification of known concentrations of Borrelia DNA. Expected copy calculations were based upon the following equation:
$$\eqalign{{\rm Genome \; copies} =\, & ({\rm ng \; template}) \times (6 \!\cdot\! 022 \times 10^{23} ) \cr& (\# {\rm bp} \times 650\,{\rm Da/bp}) \times (1 \times 10^9 )}$$
The linear Borrelia chromosomes were initially estimated to be 910 724 and 907 294 bp in length for B. burgdorferi and B. miyamotoi, respectively (Fraser et al. Reference Fraser, Casjens, Huang, Sutton, Clayton, Lathigra, White, Ketchum, Dodson, Hickey, Gwinn, Dougherty, Tomb, Fleischmann, Richardson, Peterson, Kerlavage, Quackenbush, Salzberg, Hanson, van Vugt, Palmer, Adams, Gocayne, Weidman, Utterback, Watthey, McDonald, Artiach and Bowman1997; Hue et al. Reference Hue, Ghalyanchi Langeroudi and Barbour2013), 650 Da is the average molecular weight for a DNA base pair (bp), and 109 is the unit conversion from nanograms to grams. Later, copy number calculations for glpQ were made using a revised B. miyamotoi genome of size of 1 715 670 bp (see the Discussion).
PCR Inhibition by background host DNA
To evaluate potential inhibitory effects on the performance of ddPCR detection and absolute quantification, a series of diluted concentrations of control Borrelia DNA was spiked into pre-determined quantities of total DNA extracted from I. scapularis ticks that had previously tested negative for Borrelia infection. PCR assays included 10 µL of ddPCR Probe Supermix, forward and reverse primers at 900 nm each, probes at 250 nm and 4·9 µL of uninfected I. scapularis total DNA spiked with 1 µL of various concentrations of control Borrelia gDNA. The concentrations of control gDNA were identical to the template serial dilutions tested previously. Once ddPCR was completed, results of assays with and without the presence of tick DNA extract were evaluated for concordance.
Validation and estimation of B. burgdorferi carriage levels in ticks
Six I. scapularis ticks (three adults, three nymphs) previously determined to be positive for B. burgdorferi were selected for ddPCR testing in order to validate the ospA assay. Validation of developed ddPCR protocols was carried out only for the ospA assay because B. miyamotoi was not detected among available tick samples. Each tick sample was tested using the ddPCR ospA assay as previously described with 5·9 µL of B. burgdorferi-infected tick DNA extract added as template at various concentrations. Results were analysed with QuantaSoft™ (Bio-Rad) software. The B. burgdorferi carriage level was estimated by calculating the number of spirochetes detected per infected tick. This value was found by multiplying the elution volume of tick sample extract (140 µL) by the number of template copies detected per μL template (a product of input template amount and generated copy number), assuming 100% extraction efficiency. All experiments were conducted in replicates of three unless otherwise noted.
RESULTS
Optimization of annealing/extension temperatures in ddPCR
The resulting droplet digital data from the investigation of optimal annealing temperature for the B. burgdorferi-targeting assay, ospA, are shown in Fig. 1A. No amplification of bacterial DNA was detected with the ospA assay above an annealing temperature of 54·8 °C. The optimal ddPCR result had a clustering of positive droplets at a fluorescent amplitude above 6000 rfu (relative fluorescent units) with minimal ‘rain’-down of positive droplets from this positive-clustering corresponding to a calculated annealing/extension temperature of 48·8 °C. Subsequent experiments were therefore carried out with an annealing/extension temperature of 49 °C.
Fig. 1. Optimal annealing temperatures for Borrelia DNA assays. (A) ddPCR results of annealing temperature gradient for amplification of B. burgdorferi target ospA and (B) B. miyamotoi target glpQ. Event number (X-axis) reflects cumulative number of droplets counted per experiment. Yellow vertical lines demarcate samples from individual wells. The calculated annealing temperature tested per sample is indicated for each ddPCR well in °C. Amplitude on Y-axis refers to relative fluorescence of individual droplets (blue and green, positive; black, negative).
The ddPCR data from the B. miyamotoi-targeting glpQ assay are shown in Fig. 1B. Positive amplification droplets were generated at each annealing temperature investigated. However, the sample which generated the optimal ddPCR data (with a positive-droplet line around 7000 rfu and little droplet rain) had a calculated annealing temperature of 51·8 °C. Therefore, 52 °C was selected as the optimal annealing temperature for the glpQ ddPCR assay.
Estimation of detection limits
Target DNA could be detected for both B. burgdorferi and B. miyamotoi sample sets over the entire dilution series spanning five orders of magnitude (Fig. 2). Copy number detection for B. burgdorferi ospA at the lowest dilution of 10 fg of gDNA was 10·6 copies µL−1, which corresponds well to the expected copy number of 10·17 copies µL−1. Copy number detection for B. miyamotoi glpQ at 10 fg gDNA was lower than expected, giving 5·4 copies µL−1 compared with a calculated copy number of 10·21 copies µL−1(Table 2).
Fig. 2. Detection limit estimation by ddPCR for (A) B. burgdorferi ospA and (B) B. miyamotoi glpQ assays. Pink horizontal bar indicates manually assigned cutoff for positive droplets based on no template control (NTC). Template concentration is listed above each well.
Table 2. Limits of detection for ospA and glpQ ddPCR assays. In both cases the entire dynamic range of approximately 5 orders of magnitude down to 10 fg could be detected. For glpQ, an alternative calculation was performed based on increased estimated genome size (see text).
Estimation of B. burgdorferi carriage levels in I. scapularis nymphs and female adults
Ixodes scapularis nymphs (n = 3) and female adults (n = 3) previously determined as positive for B. burgdorferi by nested PCR and sequencing of the flaB gene fragments were analysed for the B. burgdorferi ospA gene copy numbers per sample, assuming a theoretical 100% extraction efficiency of target DNA. In all cases, ticks identified as positive by nested PCR were also found to be positive by the ddPCR assay. Gene copy number in nymphs varied with an average of 2292 and range from 231 to 5268 (Table 3). Adult female I. scapularis ticks were found to average 51 179 with a range of 5647–115 797 copies of ospA per tick.
Table 3. Estimation of Borrelia burgdorferi genome copies per infected nymph or adult female Ixodes scapularis tick
DISCUSSION
Optimal annealing temperatures determined for ddPCR were considerably lower than that employed in the multiplex qPCR by Ullmann et al. (55 °C). While ddPCR is theoretically more immune to suboptimal amplification conditions, we did detect increased ‘rain’ (i.e. droplets of intermediate intensity between the average of positive and negative droplets) in samples outside the optimal annealing temperature determined here. The presence of Borrelia-negative tick gDNA did not appear to interfere with detection limits or copy number calculations.
Calculation of copy numbers using dilutions of B. burgdorferi gDNA for ospA assays corresponded well with the theoretical copy number predicted, particularly at the lowest dilution (10·6 detected vs 10·17 expected copies µL−1, Table 2); however, differences in detected vs theoretical copy number did increase with higher target concentration. The maximum difference was seen at the highest concentration (95·6 pg µL−1 gDNA, 97·3 million detected copies µL−1 vs 122 million theoretical expected copies per μL). Nevertheless, the performance of the assay was strongly linear over five orders of magnitude of gene copy abundance (R 2 = 0·9996) (Fig. 3A).
Fig. 3. Relationship of expected theoretical copy numbers to actual detected copy numbers for (A) ospA and (B) glpQ using B. burgdorferi (Bb) and B. miyamotoi (Bm) control DNA, respectively. (C) Corrected copy number calculation based on estimated B. miyamotoi whole genome size of 1 715 670 bp. See the text for details.
Copy number calculations made using dilutions of B. miyamotoi DNA for glpQ assays were not as concordant with the estimated theoretical copy numbers, despite strong linearity of the assay over the dynamic range tested (R 2 = 0·9999) (Table 2, Fig. 3B). Results indicated a pattern of consistent underestimation of the target numbers by approximately 50% across the entire range of concentrations. The consistency of the copy number inaccuracy suggested a potential systematic error in our process with respect to the B. miyamotoi glpQ assay that was not present for B. burgdorferi ospA. Upon further investigation it was revealed that the B. miyamotoi gDNA had been purified using a commercial genomic prep kit that employed a silica spin column (A. Replogle, personal communication 2015). This is in contrast to the density ultracentrifugation purification methods utilized for B. burgdorferi DNA acquired from ATCC (ATCC, personal communication 2015). The spin column chemistry would be expected to co-purify gDNA as well as any large plasmids. Reports in the literature on genomes of relapsing fever Borrelia frequently note the presence of linear megaplasmids of 160 kb or larger, in addition to numerous other plasmids (Miller et al. Reference Miller, Porcella, Raffel, Schwan and Barbour2013). In order to adjust the estimated genome size for B. miyamotoi to include possible megaplasmids, we back-calculated the total genome size (i.e. chromosome plus purified plasmids) from this DNA preparation using the average detected number of gene copies over the range of dilutions and estimate the total to be 1 708 315 bp. Using this value, we then re-calculated the expected and theoretical copy numbers from the glpQ assays (Table 2, Fig. 3C). The re-calculated detected copy numbers were similar to the revised theoretical expected copy numbers and consistent with the performance seen in the B. burgdorferi ospA assay. However, it should be noted that the predicted effective genome size of 1·7 Mb based on ddPCR glpQ abundance as estimated here is considerably larger than that extrapolated from Miller et al. based on a linear chromosome size of ~950 kb and inclusion of large megaplasmids common in relapsing fever group Borrelia of ~160–170 kb. One possible reason that could give rise to an artificially inflated genome size prediction would be contamination by foreign DNA, which in turn could arise from components in the medium or by contamination of the culture by other microbes, or both. To address the former, we performed PCR using primers targeting the mammalian mitochondrial cytochrome B gene (Kocher et al. Reference Kocher, Thomas, Meyer, Edwards, Paabo, Villablanca and Wilson1989). No amplicon was detected following gel electrophoresis (data not shown). Next we performed PCR of the bacterial 16S ribosomal RNA gene using universal primers 515F and 806R (Caporaso et al. Reference Caporaso, Lauber, Walters, Berg-Lyons, Lozupone and Turnbaugh2011), and Sanger sequenced the resulting amplicons. While a mixture would be expected to yield multiple N's upon sequencing, the resultant sequence was clean, consistent with a pure culture (data not shown). Further investigation will be required to address this inconsistency.
Determination of bacterial load was made for B. burgdorferi in infected I. scapularis ticks. The numbers revealed infection levels spanning an approximately 20-fold range of 231–5268 gene copies per nymph, and 5647–115 797 gene copies per adult female. This indicates a wide variance in infection levels in ticks. A range of spirochete levels in nymphs and adults has been previously reported based upon microscopic methods, as well as temporal changes in spirochete number following blood feeding (Piesman et al. Reference Piesman, Oliver and Sinsky1990). However, more conclusive estimates will require further analysis of other variables including evaluation of DNA extraction efficiencies, as well as tick host physiology and the role of time since feeding or infection. Still, these data should serve as a starting point and conservative lower estimate of carriage levels of B. burgdorferi in these ticks.
Concluding remarks
The ddPCR assays tested here resulted in data comparable to that reported using qPCR in terms of sensitivity, dynamic range and detections limits, and confirm the utility of ddPCR assays for the detection of Borrelia spp. in I. scapularis. One advantage of ddPCR over qPCR may include reduced need for technical replication and standard curves. The ddPCR technique also provides absolute quantification allowing for more accurate estimation of bacterial loads in infected samples. The discrepancy found between predicted and calculated genome copy numbers with B. miyamotoi glpQ highlights the importance of properly defining ‘genome size’ in relation to total DNA, as opposed to simply using published chromosome size data. It should be noted that while the B. miyamotoi sample was tested for the possibility of contamination, the tests conducted were not exhaustive and foreign DNA could still have led to the overestimation of the total genome size of this organism, and this value should be viewed with caution. Nevertheless, performance of the ddPCR assay led to linear output over a range of 10–100 000 copies per sample with an R 2 of 0·9999.
The ddPCR technique is still relatively new. Although not specifically tested, one limitation is the reduced ability to perform multiplex assays using a two-channel system such as the one tested here. Presumably future iterations of the platform may have additional capacity for multiplexing with other dyes, provided the dye chemistries can be made compatible with the droplet oil. Another complication with ddPCR is the inherent fragility of the droplets themselves. Careless handling can lead to rupture of the droplets, and full automation of droplet formation may prove more difficult than procedures designed for traditional PCR or qPCR. Under the current format, the process is also limited to the preparation of eight samples at a time. While the impact of this constraint is somewhat lessened by the reduced need for standard curves and large numbers of technical replicates, it necessarily limits the platform to a low-to-medium range sample number throughput. Nevertheless, the accuracy and precision generated by the technique suggests that it will find useful niches in molecular diagnostics.
ACKNOWLEDGEMENTS
We would like to thank Dr Martin Schriefer and Adam Replogle of the CDC Division of Vector-Borne Disease for kindly providing B. miyamotoi DNA, as well as Diamond Rogers for her help in sample preparation.
FINANCIAL SUPPORT
This work was supported in part by NIH NHLBI SMART grant 2R25HL007786-23 (for D. R.) and the State of Texas.
