Massive dispersal of Coxiella burnetii among cattle across the United States

Q-fever is an underreported disease caused by the bacterium Coxiella burnetii, which is highly infectious and has the ability to disperse great distances. It is a completely clonal pathogen with low genetic diversity and requires whole-genome analysis to identify discriminating features among closely related isolates. C. burnetii, and in particular one genotype (ST20), is commonly found in cow’s milk across the entire dairy industry of the USA. This single genotype dominance is suggestive of host-specific adaptation, rapid dispersal and persistence within cattle. We used a comparative genomic approach to identify SNPs for high-resolution and high-throughput genotyping assays to better describe the dispersal of ST20 across the USA. We genotyped 507 ST20 cow milk samples and discovered three subgenotypes, all of which were present across the entire country and over the complete time period studied. Only one of these sub-genotypes was observed in a single dairy herd. The temporal and geographic distribution of these sub-genotypes is consistent with a model of large-scale, rapid, frequent and continuous dissemination on a continental scale. The distribution of subgenotypes is not consistent with wind-based dispersal alone, and it is likely that animal husbandry and transportation practices, including pooling of milk from multiple herds, have also shaped the patterns. On the scale of an entire country, there appear to be few barriers to rapid, frequent and large-scale dissemination of the ST20 subgenotypes.


Summary
The purpose of this document is to report on the validation and testing of two TaqMan assays, Cox1336818 and Cox1910969, designed for subgenotyping of Coxiella burnetii ST20 genotypes. These assays (Table 1) provide additional resolution among samples of C. burnetii that are sequence type 20 (ST20) [1][2][3]. Both assays are located on the branch separating the genomes ESFL1 and CMCA1c1 from all other ST20s as well as all other STs of Coxiella ( Figure 1). We tested the following parameters: precision, limits of quantitation, linearity, selectivity, limits of detection, and robustness. In addition, we compared the ΔCt values between the assays presented here, which target a single copy per genome, to the ΔCt values from an IS1111 assay [4], which targets multiple copies per genome and is used for detecting C. burnetii. The goal being to determine at what ΔCt values for IS1111 we see negative amplification with our single-target assays.

Figure 1: Phylogenetic tree indicating MST genotypes drawn according to Hornstra et al. [2] and rooted according to Pearson et al. [3] (boxed). The blue arrow indicates the location of ST20 in relation to all other sequence types of Coxiella.
Enlarged phylogenetic tree for ST20 genomes only is shown. Location of the assays used in this study are indicated; subgenotypes described in the manuscript are indicated by colors on branches. The ancestral to derived SNP change for both assays is C/T.

PCR Conditions
Unless otherwise noted in a specific validation test, the following conditions were used for the Cox1910969 assay: 1 μL of DNA was added to form a total reaction volume of 10 μL that contained 5 µL of 2x TaqMan® Universal PCR Master Mix (Life Technologies, CA, USA; p/n 4304437), 0.20 µL of each 20 µM primer (Table 1), 0.10 µL of the 20 µM FAM MGB-NFQ probe (Table 1), 0.15 µL of the 20 µM VIC MGB-NFQ probe (Table 1), and 3.35 µL of sterile, molecular grade water. For the Cox1336818 assay, reaction conditions were identical to those above but with the following exceptions: 0.06 µL of each 20 µM MGB-NFQ probe (Table 1) and 3.48 µL of sterile, molecular grade water were added. Unless otherwise specified in this document, thermal cycling conditions for both assays were as follows: 50°C for 2 min., 95°C for 10 min., followed by 45 cycles of 95°C for 15 sec., and 60°C for 1 min. PCR was performed on an Applied Biosystems 7900HT Fast real-time PCR system with SDS v2.4 software. All data were analyzed using SDS v2.4 software with a manual threshold of 0.1 and an automatic baseline.

Part I -Precision, Limits of Quantitation, and Linearity
The precision, limits of quantitation, and linearity of both Coxiella assays were measured using a single plate setup for each assay. For both assays, a panel consisting of a ten-fold serial dilution series of whole-genome amplified (WGA) product from two ancestral and two derived templates were used. Eleven dilution points were made with point one having the highest DNA concentration (i.e. WGA product that would yield a Ct (cycle threshold) of ~20) and point 11 the lowest DNA concentration (i.e. 1x10 10 dilution of point 1). Four assay replicates were used for each template (ancestral and derived) at every dilution point (points 1-11); 16 negative template controls were also used.

Precision
Precision, or agreement among multiple replicates of a sample, was measured for each dilution point by 1) examination of the C T standard deviation (SD) among the four replicates at each dilution point and 2) observation of the amount of positive amplifications per dilution point. Standard deviations ranged from 0-1.82 C T s for both assays. Amplification of all replicates only occurred out to the 5 th or 6 th dilution points, depending on the assay (Figure 2). There was no amplification in negative template controls. Cross hybridization between the ancestral and derived probes was observed for the Cox1336818 assay only, but the difference in C T between the allele-specific signal and the mismatch signal was always greater than 2.81 C T s and therefore large enough to provide discriminatory ability ( Figure 2). The assays were precise, therefore, out to the 5 th or 6 th dilutions when the original template had a starting C T of ~20. SNP genotyping calls for ancestral and derived WGA templates were 100% accurate at all dilution points which had amplification.

Limits of Quantitation
Limits of quantitation, or the lowest and highest concentrations of target DNA that can be measured with reasonable precision and accuracy, were measured by looking at the amount of correct SNP calls at each dilution point. As SNP calls were 100% accurate and there was no amplification of NTCs, the limits of quantitation were determined by the precision of each assay, which were accurate out to the 5 th or 6 th dilutions when the original template has a C T of ~20. In many cases, SNP calls were also accurately made at dilutions 7-9 (approximately C T s of 36), with only one or two replicates of four amplifying. The upper limit of quantitation was undetermined as we did not test samples more concentrated than dilution point 1.

Linearity
Linearity, or the ability to elicit results proportional to the concentration of target DNA in the sample, was measured by plotting the average C T across all amplifying replicates, per dilution point, per template, per assay and attributing r 2 values to the range of the linear dilutions ( Figures  3 and 4). The difference in the average C T values from one dilution point to the next ranged from 1.13 to 3.91 for both assays. The r 2 values for both assays where the linear range was determined was ≥0.9904. As we were able to determine a range where the change in C T values per ten-fold dilution increases by ~ 3.0 C T s and the r 2 values in this range are ≥0.9904, we can say that each assay has a linear range. For Cox1336818, the linear range of the assay falls between ~22-37 C T s (Fig. 3). For Cox1910969 the linear range of the assay falls between ~26-39 C T s (Fig. 4).

Part II -Selectivity
To test selectivity, or the ability to accurately genotype in the presence of mixed samples, one ancestral and one derived WGA template (point 1 from Part I above) were used. Mixtures of ancestral:derived template were combined at ratios of 10:90, 25:75, 50:50, 75:25, and 90:10 to test the selectivity of each assay. Un-mixed samples were used as positive template controls; molecular grade water was used as a negative template control. Four replicates were tested for each control or mixture for each assay. The ∆C T between the ancestral and derived probes was calculated for all replicates and the mixed and control samples were compared.
For both assays, the positive template controls (unmixed samples) did not show any cross hybridization of probes ( Figure 5, panels "100 % Ancestral" and "100% Derived"). The mixtures had ∆C T values ranging from 0.78 to 3.93 C T s. The 50:50 mixtures did not have an average ∆C T value of zero as would be expected if they were exactly 50:50 and/or the binding efficiencies of the probes were equal ( Figure 5 panels "50:50 ANC/DER"). Both assays appear to skew towards the ancestral SNP state (FAM-labeled probe for both assays) in mixtures as the 25:75 mixtures were called as ancestral despite the fact that there should have been much more derived template in the mixtures ( Figure 5). This should not affect the ability of the assays to accurately detect the presence of mixtures in samples when accurate ancestral and derived controls are used to provide a reference point when evaluating mixtures. It does, however, preclude the ability for these assays to be used to accurately estimate the ratio of mixtures in a sample.

Part III -Limits of Detection
The limits of detection, or the concentration at which amplification of one more replicates occurs, was measured for ancestral and derived probes in both Coxiella assays by testing six dilution points from Part I. The six dilution points were: two points that have successful amplification (all four replicates amplified in Part I) two points that have spotty amplification (less than four replicates amplified in Part I), and two points that failed to amplify (no amplification in Part I).
Twenty-four replicates were tested for each ancestral and derived template at each of the six points selected; Twenty-four negative template controls were also tested. The points from Part I chosen for each assay per derived and ancestral templates are givin in Table 2.
The limits of detection (LOD) were measured by counting the number of replicates out of 24 which amplified for each point chosen, per template, per assay ( Table 2). The LOD for both assays varied: for Cox1336818 the LOD occurred at point 7 for both probes with the ancestral probe amplifying 5/24 replicates and the derived probe amplifying 3/24 replicates (Table 2). For assay Cox1910969, the LOD varied by probe with the ancestral probe's LOD being point 6 (11/24 replicates) and the derived probes LOD = point 5 (3/24 replicates). No NTCs amplified for either assay.

Part IV -Robustness
The robustness of each assay, or the ability for the assay to amplify and genotype samples correctly at different annealing temperatures, was tested by taking three dilution points of an ancestral and derived WGA template determined from the results of Part I to be spotty (see also Part III) plus two, ten-fold dilution points after that and testing them with three different annealing temperatures. Four replicates were run for each dilution point along with eight NTCs, per assay. The annealing temperatures tested were ±5°C from 60°C and including 60°C. When tested with the standard temperature of 60°C, results were as expected (samples genotyped correctly) and little to no cross-hybridization was observed ( Figure 6). When the annealing temperature was decreased to 55°C, the specificity of the probes was reduced for both assays, thus increasing cross hybridization of the probes (Figure 6) making genotyping of samples more ambiguous. When tested with an annealing temperature of 65°C, the assays failed entirely across all samples; no amplification was observed (data not shown). In summary, these assays are not very robust to changes in annealing temperature and therefore PCR should always be performed with an annealing temperature of 60°C.

Part V -IS1111 Comparison
As an additional test of these assays, a comparison between the Coxiella-specific detection assay, IS1111 [4], and our genotyping assays was done in order to determine the C T value cut-off at which Coxiella burnetii DNA can be detected in a sample with IS1111 but no longer genotyped by our assays. For this comparison, we used all dilution points from Part I (described above) and plotted the average C T of all amplifying replicates (per template, per probe,) along with the average C T of the sample replicates when tested with IS1111. When running the IS1111 assay all SDS v2.4 settings were left as automatic and PCR was performed according to [4]. As shown in the plots below (Fig. 7), the last C T value where IS1111 amplifies and where genotyping results are obtained for each assay varies from a C T value of 29.96 to 35.65. This implies possible genotyping failures with our assay for samples that have an IS1111 C T value ~30 or greater and likely genotyping failures with a C T value above 35.65 ( fig. 7). We would expect that genotyping failures with these two assays will increase and/or occur at a lower C T value when the isolates being tested have a greater number of copies of IS1111 per genome. The tests below were performed using whole genome amplified product from strains CMSC1 and Q154 for the ancestral templates and ESFL1 and CMCA1c1 for the derived templates. We estimate that CMSC1, ESFL1, and CMCA1c1 have ~25 copies of IS1111, per genome and Q154 has ~46 copies, per genome (data not shown).
Figure 7: IS1111, Cox1336818, and Cox1910969 average C T comparison. Each plot depicts the average C T for all amplifying replicates across a ten-fold dilution series of either an ancestral or derived WGA template when tested with the IS1111 [4], Cox1336818, and Cox1910969 TaqMan assays. For all plots, the blue diamonds indicate the IS1111 average C T at each dilution point, the red squares indicate the average C T for the probe corresponding to the sample for Cox1336818, and the green triangles indicate the average C T for the probe corresponding to the sample for Cox1910969.