Comparison of two novel one-tube nested real-time qPCR assays to detect human-infecting Cyclospora spp.

ABSTRACT Human-infecting Cyclospora spp. currently include three coccidian parasites that cause the gastrointestinal disease cyclosporiasis in humans. They are often spread through contaminated produce, including leafy greens and berries. The increased availability of sensitive molecular tests for the diagnosis of cyclosporiasis is an important advancement, allowing public health agencies to better understand the scope and source of cyclosporiasis outbreaks. To improve the diagnosis of infected patients, rapidly detect outbreaks, and keep the food supply safe, it is important to continue to develop sensitive, reliable, and inexpensive tests to detect human-infecting Cyclospora spp. In this report, we describe the development and evaluation of two novel one-tube nested qPCR assays for the detection of human-infecting Cyclospora spp. in clinical stool samples, one targeting cytb and the other targeting coxI. Of these, the assay targeting the cytb mitochondrial locus possessed strong performance characteristics compared to a routinely used 18S assay, including a markedly improved (approximately 10-fold lower) relative detection limit of 0.613 oocysts per gram of feces. This is compared to coxI that has a relative detection limit equal to that of the 18S assay. Given the strong performance characteristics of the cytb assay, we propose that it may be useful to diagnostic laboratories wishing to screen clinical fecal specimens suspected of containing human-infecting Cyclospora spp. IMPORTANCE Human-infecting Cyclospora spp. cause gastrointestinal distress among healthy individuals contributing to morbidity and putting stress on the economics of countries and companies in the form of produce recalls. Accessible and easy-to-use diagnostic tools available to a wide variety of laboratories would aid in the early detection of possible outbreaks of cyclosporiasis. This, in turn, will assist in the timely traceback investigation to the suspected source of an outbreak by informing the smallest possible recall and protecting consumers from contaminated produce. This manuscript describes two novel detection methods with improved performance for the causative agents of cyclosporiasis when compared to the currently used 18S assay.

Laboratory-confirmed cases of cyclosporiasis are thought to represent only a fraction of the actual number of U.S. infections given that some persons do not seek medical care for GI illness (4).In addition, patients may not be correctly diagnosed with cyclosporiasis because the symptoms are non-specific and standard parasitology screening tests such as ova and parasite microscopy may fail to detect human-infecting Cyclospora spp.(5).A correct diagnosis is important not only to facilitate appropriate treatment of patients but also to better understand the scope and source of cyclosporiasis outbreaks.
Diagnostic methods for the detection of Cyclospora in human fecal specimens include morphological and PCR-based methods.Morphologic detection of human-infecting Cyclospora spp.oocysts using ultraviolet fluorescence microscopy or a modified acid-fast staining technique is low cost but requires a highly trained and experienced technician for accurate results (5,6).In the years after the BioFire FilmArray GI Panel (BioFire assay) became available for clinical use, the number of laboratory-confirmed cyclosporiasis cases has increased, owing at least in part to the use of this new diagnostic test; this phenomenon was especially pronounced during the 2018 cyclosporiasis peak period where the case count approximately doubled compared to the previous year (7,8).However, other sensitive tests are needed to fill the gap when the BioFire assay is either unavailable or too expensive to apply to a situation.
For detection of Cyclospora spp.many U.S. public health laboratories currently utilize a TaqMan real-time PCR originally published in 2003 (9).This assay has a traditional real-time PCR design with one primer pair and a TaqMan probe, targeting the 18S rRNA gene.The rRNA genes are common targets for diagnostic tests due to their high copy number and their tendency to be highly conserved within species; however, other markers that have not been explored may yield a tool with improved performance.Mitochondrial genes may represent useful targets for the design of novel assays due to similarly high copy numbers compared to rRNA genes.We selected the mitochondrial targets cytb and coxI for this study, as they appeared to be both sensitive (i.e., present in many copies in each oocyst) and specific (i.e., sufficiently diverse from other intesti nal pathogens) based on in silico comparisons to homologous sequences from related protozoa.Additionally, to increase sensitivity, we utilize a single-tube nested real-time PCR fomat (10).The performance of our novel single-tube nested qPCR assays targeting the mitochondrial genes cytb and coxI was evaluated against that of the widely used 18S TaqMan assay.

Sample selection
The BioFire assay was utilized as a gold standard for the selection of true positive samples for assessing the sensitivity of our novel assays against the 18S assay (11).We selected 101 human stool specimens submitted to CDC for Cyclospora spp.genotyping in 2019 (4) and that had tested positive for human-infecting Cyclospora spp.using the BioFire assay (for details, see Table S1).All BioFire-positive specimens received for genotyping were examined with the 18S assay, and the Cq values were used as a proxy for parasite load.From these, specimens were selected to represent various genetic variants (as indicated by the genotyping results) as well as strong, medium, and weak parasite loads (including specimens that tested negative in the 18S assay despite being positive in the BioFire assay).The patients providing these samples are assumed to be symptomatic, as they sought medical care.

Primer design
We designed two candidate assays for evaluation: one targeting the coxI gene (coxI assay) and the other targeting the cytb gene (cytb assay).For each target, we designed six primers using 32 mitochondrial genomes available for human-infect ing Cyclospora spp.from the NCBI database (GenBank [GB] accession numbers: NC_038230, MG831586-MG831588, MN260345-MN260366, MN316534, MN316535, KP796149, KP658101, KP231180, and JROU01000145).To align the genomic scaffolds and identify the binding sites for the forward and reverse primers of the inner amplicon, we used the Geneious Prime v2020.0.5 (Auckland, New Zealand) software alignment tool with the default parameters.We used the Geneious Primer design tool to select the probe and outer primers with the following specifications; the priming sites had to flank the inner amplicon, and the outer primers had to have a theoretical annealing tempera ture at least 5°C higher than that of the inner primers.Primers were manually curated to reduce primer dimer and other unwanted secondary structures and to maintain proper annealing temperature differences between the inner primers, the probe, and the outer primers.In silico analysis of the primers and probes was done by utilizing the BLAST function of the Geneious Prime software, to assess specificity.Primers and probe sequences are listed in Table 1.

Extraction of DNA
DNA was extracted from stool samples using the UNEX method described by Qvarn strom et al. (11), following one to three washes with phosphate-buffered saline (PBS) at 2,500 × g to remove preservatives.

Assay preparation
All three assays (18S, coxI, and cytb) were evaluated for specificity (n = 108), sensitivity (n = 101), relative analytic detection limit, repeatability, and amplification efficiency using the equation (E = 10 (-1/slope) −1), where the slope of the standard curve is used, and by calculation of R 2 values.The specificity and sensitivity panels were run once on each assay, and the serial dilutions were run in duplicate.Both newly designed assays utilized a 20-µL reaction mixture consisting of Invitrogen Platinum Quantitative PCR SuperMix-UDG with ROX (Life Technologies Corporation, Rockville, MD), 200 nM of each primer (four per assay), 100 nM of the TaqMan probe, and 5 µL of the DNA template.The 18S assay used the previously described reaction mixture and cycling conditions published in Qvarnstrom et al. (11), with the only modification being the addition of 5 µL of the DNA template as opposed to 2.5 µL.The cycling conditions for the assays targeting the coxI gene (coxI assay) and the cytb gene (cytb assay) are listed in Table 2.All three real-time PCR assays were run on an Agilent Technologies AriaMX with software version v1.71 (Santa Clara, CA).

Standard curve
We created a standard curve to approximate the analytic sensitivity of each assay.To do this, we used 1:100 dilution of a fecal sample with a known quantity of human-infecting Cyclospora spp.oocysts in saline solution.We manually counted the number of oocysts present in a 10µl aliquot of the fecal solution under ultraviolet microscopy (Olympus BX51, Olympus Life Sciences) in triplicate, and the average was taken and used to calculate the number of oocysts in the undiluted sample.Following DNA extraction of the undiluted fecal sample, the DNA was serially diluted to a 1:1 × 10 6 dilution by factors of 10.Using the number of oocysts in the undiluted sample, we approximated the number in each dilution.

Repeatability
To determine the repeatability of the cytb and coxI assays, we ran six positive samples and six negative samples on four separate runs over at least 7 days, at least 2 days apart, by two different operators using the reaction conditions described above.The positive samples were selected such that there were two weak, two medium strength, and two strong positives, based on Ct values from the 18S assay.The previously published repeatability for the 18S assay was used for comparison (11).

Accuracy
To assess the accuracy of the coxI and cytb assays, we tested 101 known positive samples and 108 known negative samples with each assay.The positive samples had previously been shown to contain human-infecting Cyclospora spp.using the BioFire assay.We

Data analysis
To assess differences in performance between the assays, we used chi-squared analyses and t-tests in R 4.0.3(12) with all pairwise comparisons: coxI vs. 18S, cytb vs. 18S, and cytb vs. coxI.We used the chi-squared test (H 0 : µ A = µ B ; H A : µ A ≠ µ B ; α = 0.05; β = 0.80) to determine whether the number of samples that each assay detected as human-infecting Cyclospora spp.positive was significantly different between the methods for each of the specificity and sensitivity evaluations and applied the Yates correction.We used a one-sided t-test (H 0 : µ A = µ B ; H A : µ A > µ B ; α = 0.05; β = 0.80) to assess whether the average Ct values obtained while assessing the sensitivity of the cytb and coxI assays were significantly different from the values obtained using the 18S assay.Specimens with no Ct value were excluded when calculating averages.A Bonferroni correction was used to correct the resulting P-values from the t-tests for multiple comparisons, using the equation "α new = α original / n. "

Assay design and limits of detection
This study used a one-tube nested real-time PCR assay design as a mechanism to increase assay sensitivity.The addition of the outer primer pair to facilitate the nested amplification resulted in a positive human-infecting Cyclospora spp.detection for 10 and 12 additional specimens, compared to using the inner primers and the TaqMan probe alone (the non-nested version) for the cytb and coxI assays, respectively.Furthermore, the nested version of the cytb assay was able to detect one replicate of the most diluted DNA sample in the standard curve (the 1:1 × 10 6 ), while the non-nested version failed to detect human-infecting Cyclospora spp. at this concentration.

Sensitivity
A panel of 101 specimens that tested positive by the BioFire assay for human-infecting Cyclospora spp.were used to analyze the sensitivity of the novel assays.In general, the cytb assay was the most sensitive among the three evaluated assays (Table 3)(Fig.1).We did not find a significant difference in the number of samples that were positive for human-infecting Cyclospora spp.when comparing the coxI assay to the 18S assay [X 2 = (1, N = 101) = 1.73,P = 0.19].However, we did find a significant difference in the number of samples testing positive for human-infecting Cyclospora spp.by the cytb assay versus the 18S assay [X 2 = (1, N = 101) =10.26,P = 0.001].
Finally, an approximate analytic sensitivity (detection limit) was ascertained by determining the lowest concentration of serial dilutions testing positive for humaninfecting Cyclospora spp.The detection limit was defined as the lowest level at which the assays detected a single replicate out of two of the same serial dilution on separate runs.The three separate counts of the original sample used for the serial dilutions yielded an average of 613,000 oocysts present per gram of stool.The cytb assay was able to detect human-infecting Cyclospora spp. in the 1:1 × 10 6 dilution in one out of two replicates and, therefore, detected an equivalent of 0.613 oocysts per gram.The coxI and 18S assays both detected human-infecting Cyclospora spp. at the 1:1 × 10 5 dilution, both yielding an approximate analytic sensitivity of 6.13 oocysts per gram.

Specificity
The specificity panel for this study consisted of 108 specimens containing a variety of different pathogens commonly found in human stool samples, as well as some closely related nonhuman parasites.Although in silico analysis of the oligonucleotides indicated no cross-reactivity, all three assays returned false-positive results on a few of the specificity controls (Table 4).
We obtained a positive result with the 18S assay from a sample containing Dipylidium and all eight of the tested Cystoisospora samples, resulting in an overall specificity of 91.7% (Table 4).The cytb assay obtained a positive result for three human fecal speci mens that were positive for Ascaris, while the coxI assay obtained a positive result for one of these Ascaris samples.However, most Ascaris-positive samples (total number of 14) tested negative in these two assays and there were no additional false positives among the human samples.The cytb assay did cross-react with six animal samples containing nonhuman-infecting primate Cyclospora spp.and Eimeria, while the coxI assay produced a positive detection for one animal sample containing Cyclospora cercopithecus (one of the nonhuman primate-infecting Cyclospora spp.).Therefore, the overall cytb assay specificity is 91.7%, while the coxI assay specificity is 98.1%.To determine the significance of these differences, an X 2 test was performed.We found that neither the cytb assay nor the coxI assay had a specificity that statistically differed from that of the 18S assay [cytb X 2 = (1, N = 108) = 0.06, P = 0.81, coxI X 2 = (1, N = 108) =3.45, P = 0.06].

Assay efficiency
When testing the standard curve samples, it was observed that the undiluted sample had a higher Ct value than that of the next more diluted sample in all three assays, indicating the presence of amplification inhibition.Thus, the undiluted sample was removed from the standard curve analysis.As seen in Fig. 2, the coxI assay's R 2 variable is 0.995, which is close to the desired value of 1.This assay did not exceed a Ct of 28 and was not able to detect the lowest dilution factor, as seen in Table 5.The cytb assay produced an R 2 of 0.98, well within the acceptable parameters (0.95-1).This assay was the only one to detect the most diluted sample in the serial dilution and did not have a Ct higher than 25 on the standard curve.The established 18S assay had an R 2 value of 0.963, was unable to detect the most dilute sample in the dilution series, and had Ct values near 35, as seen in Fig. 2.

Repeatability
The repeatability of both novel assays was shown to be 100%.This was done by two technicians, each running the repeatability panel described in the methods section twice each for a total of four runs, over at least 7 days, allowing at least 2 days (48 h) between each run.By comparison, the published 18S assay repeatability was reported to be 100% (11).

DISCUSSION
The objective of this study was to develop an assay that provides more sensitive detection of human-infecting Cyclospora spp. in fecal specimens than that provided by the previously published real-time PCR assays.We designed two single-tube, nested real-time PCR assays targeting two mitochondrial loci (coxI and cytb), and these assays were evaluated for their performance in comparison to a published 18S TaqMan assay, using the BioFire assay as a gold standard.We selected the two mitochondrial targets on the basis that they exist at a relatively high copy number per cell (13) and are fairly diverse loci offering unique binding sites for specificity, while maintaining sufficient stability to ensure the detection of diverse populations of human-infecting Cyclospora spp.(6).Both designs resulted in assays with comparable (coxI assay) or improved (cytb assay) diagnostic performance characteristics when compared to the 18S assay.The  efficiency of both novel assays was higher than expected for a standard real-time PCR assay, which is a common observation in one-tube nested real-time PCR assays (14), and may be related to nonspecific binding in one or both of the primer pairs (15).This drawback is equalized by the nested format that confers increased specificity.The BioFire assay continues to be more diagnostically sensitive than either of the two novel assays described in this study.A recent study determined that the BioFire assay can detect as little as just one oocyst in the 200-µL stool used in the analysis (16).A direct comparison of the limit of detection between BioFire and the assays evaluated in this study is not possible because of the different methodologies used (the BioFire study used a cell sorter to count oocysts, and that technology was not available for this study).However, both novel assays detected Cyclospora in more of the samples included in the sensitivity panel than the 18S assay: the coxI assay detected 10 more (10%), and the cytb assay detected 22 more (22%).These assays also detected positive samples at an earlier Ct on average (µ 18S = 26.49± 3.69, µ coxI = 25.74 ± 6.71, µ cytb = 20.97 ± 4.9).The diagnostic sensitivity of the coxI and 18S assays did not significantly differ, but the cytb assay showed significantly improved diagnostic sensitivity relative to the 18S assay.Based on the assessment of the approximate limit of detection, the cytb assay had an approximately 10-fold lower limit of detection (LOD) relative to the 18S assay.The undiluted sample was not used in the serial dilution, since evidence of inhibition had been observed, as is common in nucleic acid extracted from fecal materials.A limitation of this study is that an internal control assay was not included, which could highlight instances of inhibition in a sample and clarify if the false negative samples for each assay were due to limit of detection differences or inhibition.
Our data suggest that the coxI assay cross-reacts with at least some nonhumaninfecting Cyclospora spp., although this has minimal human medical importance due to these organism's inability to cause human illness.Additionally, a fecal specimen containing Ascaris also returned a positive result with the coxI assay.Interestingly, this same specimen was also positive using the cytb assay, raising the question of whether this specimen could have contained a small amount of Cyclospora that may have been missed due to the comparatively high limit of detection of microscopy and the 18S assay.This could be resolved by running these samples on the BioFire assay in the future; however, the platform was not available to our group while data were being generated for this study and our study design relied on external partners performing the BioFire assay.Alternatively, this specimen may have contained something else that cross-reacted with both the coxI and cytb assays.Most Ascaris-positive samples remained negative in all Cyclospora-detecting assays in this study, so a true cross-reaction with Ascaris DNA seems unlikely.The cytb assay had a lower specificity than that of either of the other assays, reacting to two additional supposedly Cyclospora-free human specimens.In addition, this assay produced positive results for specimens containing Cyclospora and Eimeria species that are not known to cause human illness (17,18).The broader detection of coccidia shown here with the cytb and coxI assays may have important implications for other uses outside of the human clinical realm, such as in veterinary/wildlife or environmental studies.As such, positive results from samples that do not derive from humans in clinical settings should be analyzed with this in mind and utilize a secondary method as confirmation of correct detection.Interestingly, the 18S assay was the only evaluated method to cross-react with specimen containing Cystoisospora (n = 8), which is a human pathogen.
To conclude, we evaluated two novel single-tube, nested qPCR assays for detection of human-infecting Cyclospora spp. in clinical fecal specimens, targeting different loci (coxI and cytb) encoded in the mitochondrion of these parasites.Both novel assays performed better than a previously developed TaqMan assay targeting the 18S rRNA, with the cytb assay having significantly better sensitivity.We propose that these assays may be useful to other laboratories wishing to provide sensitive detection of human-infecting Cyclospora spp. in clinical fecal specimens that do not have the specialized and expensive equipment to run BioFire or that have do not need such highly multiplexed assays.

FIG 1
FIG 1 Specimens from our sensitivity panel obtaining a positive detection for each assay.This Venn diagram shows the overlapping positive detection of Cyclospora using three real-time PCR assays for 101 fecal specimens that previously tested positive for Cyclospora using the BioFire FilmArray assay.In this figure, the cytb assay is shown to have detected 81/101 specimens in total.Of these 81 specimens, 52 were detected by all three assays, and 59 were detected by the cytb and 18S assays.The cytb assay detected all the specimens that both other assays could detect.The 20 specimens undetected by the cytb assay were only detected by the gold standard BioFire assay.

FIG 2
FIG 2Standard curves for 185 (red), coxl (green), and cytb (blue) assays.Standard curves were plotted using duplicates of serially diluted oocysts ranging from 0.61 to 613,000 oocysts per gram of feces as shown in Table2.These curves were used to calculate R 2 and the efficiency (E) of the assay using the equation E = 10 (−1/slope) −1.The R 2 values calculated for each assay were 0.963, 0.995, and 0.98, respectively.The efficiency of each assay was 103.69, 109.16, and 116.37, respectively.

TABLE 1
Primer and probe information

TABLE 2
Cycling conditions for two novel assays the negative samples by microscopy.All negative samples contained other pathogens (see TableS2) or were classified as "no parasite found" (NPF). confirmed

TABLE 3
Results from testing true positive samples (samples that tested positive for human-infecting Cyclospora spp. in BioFire)

TABLE 4
Results from testing true negative samples (samples free of intestinal parasites or confirmed positive for other intestinal parasites besides humaninfecting Cyclospora spp.)

TABLE 5
Limit of detection of each assay in terms of oocysts detected per gram of feces