We used 137 human stool specimens to evaluate the DNA extraction and PCR methods. The human stool specimens had been sent to the Centers for Disease Control and Prevention (CDC) for diagnostic confirmation or as part of outbreak investigations during 2004–2015 and were used in accordance with the CDC Human Subjects Research protocol entitled “Use of residual diagnostic specimens from humans for laboratory methods research.” The stool samples were unpreserved or had been collected in commonly used fixatives for parasitology, including Zn-PVA, EcoFix, or TotalFix.

Thirty-two of the 137 human stools were classified as positive for C. cayetanensis, based on positive UV fluorescence microscopy (n = 30) or DNA sequencing confirmation of PCR products obtained by the nested PCR assay (n = 2) (Berlin et al., 1998; Relman et al., 1996). One of the microscopy-positive specimen, selected because of a large volume (about 100 ml) and sufficiently high concentration of oocysts (an oocyst count of 1.3 × 10⁴ oocysts per ml, as determined using a hemocytometer), was used to assess the analytical sensitivities (detection limits) of the four PCR assays: four aliquots of this stool were serially diluted in parasite-free stool to obtain four separate dilution series down to 1.3 oocysts per ml. The definition of the detection limit was the lowest number of oocysts that an assay detected in all four aliquots.

For specificity analysis, we used 105 of the 137 human stool specimens, 15 stool samples from non-human primates, 3 stool samples from rats, and 2 DNA samples. The human stool specimens were positive by microscopy for either Entamoeba histolytica/dispar (n = 24 specimens); Enterocytozoon/Encephalitozoon spp. (n = 13); Giardia duodenalis (n = 6); Cryptosporidium spp. (n = 3); Blastocystis hominis, Dientamoeba fragilis, or Iodamoeba butschlii (each n = 2); or for Balantidium coli, Chilomastix mesnili, Entamoeba coli, Entamoeba hartmani, hookworm, or Trichomonas hominis (each n = 1); or were negative by microscopy for parasites and microsporidia (n = 47). The non-human primate samples, positive for simian Cyclospora spp. as described elsewhere (Eberhard et al., 1999; Eberhard et al., 2001), included eight Cyclospora papionis-positive samples from Papio anubis in Ethiopia, two C. papionis-positive samples from P. anubis in Kenya, two Cyclospora colobi-positive samples from Colobus angolensis in Kenya, and three Cyclospora cercopitheci-positive samples from Cercopithecus aethiops in Kenya. The rat stools (collected during an environmental investigation unrelated to this study) were positive by microscopy for Eimeria spp. (n = 2 samples) and Hymenolepis nana (n = 1). DNA samples extracted from Eimeria tenella and Eimeria acervulina were also included.

Two methods for extraction of total genomic DNA from stool were compared using fifty human stool samples (25 C. cayetanensis-positive and 25 specificity controls). Method 1 used a modification of the FastDNA^® method (da Silva et al., 1999). Aliquots of ~0.3–0.5 ml of each stool were subjected to bead beating in a FastPrep-24 cell disruptor instrument (MP Biomedicals, Santa Ana, CA, USA). An internal DNA quality control plasmid (Duffy et al., 2013), here called pIC, was added immediately after the bead-beating step to allow for detection of DNA extraction failure (see below). Potential inhibitors remaining after the FastDNA extraction were removed by further purification with the QIAquick PCR purification kit, according to the manufacturer’s instructions (QIAGEN Inc., Valencia, CA, USA).

Method 2 was based on the commercially available UNEX (Universal Nucleic Acid Extraction) buffer. An aliquot of ~0.5 ml of stool was added to a matrix E bead-beating tube (MP Biomedicals), along with 60 µl of proteinase K (QIAGEN) and 600 µl of UNEX buffer (Phthisis Diagnostics, Charlottesville, VA, USA). The tube was incubated for 15 min at 56°C for protein digestion, followed by disruption in the FastPrep-24 cell disruptor instrument at a speed of 6.0 m/s for 1 min. One pg of the internal control plasmid pIC was added, and the tube was then centrifuged at maximum speed (>13,000 × g) for 1 min to pellet the debris. The supernatant was passed through a DNA-binding column (DNeasy mini spin column from QIAGEN). After two wash cycles with ethanol-containing wash buffers, the DNA was eluted from the column in 80 µl of AE buffer (QIAGEN). To purify the eluate further, it was passed through a Zymo-Spin IV-HRC column (Zymo Research Corp., Irvine, CA, USA).

The remaining 87 human stool specimens and the 18 animal stool samples were extracted with the UNEX-based method only, since that method performed better than the Fast DNA method for extraction of C. cayetanensis DNA from stool (see results).

The nested PCR assay contained 0.3 µM of each primer (Relman et al., 1996) and the AmpliTaq Gold PCR Master Mix (Life Technologies, Grand Island, NY, USA). The total volume was 50 µl, with 5 µl of extracted DNA added to the first reaction and 3 µl of the undiluted product from the first step added to the nested reaction. The cycling parameters were 95°C for 5 min to activate the polymerase, followed by 30 cycles of 95°C for 15 s, 57°C for 15 s, and 72°C for 1 min for the first step; and 35 cycles of 95°C for 15 s, 65°C for 15 s, and 72°C for 1 min for the nested step. The PCR products were visualized on 1.5% agarose gels stained with ethidium bromide.

We compared four real-time PCR assays: two TaqMan assays we refer to as the Verweij (Verweij et al., 2003) and the Varma (Varma et al., 2003) assays, which target different parts of the 18S rRNA gene; a SYBR Green assay (Lalonde & Gajadhar, 2011) that targets the 18S rRNA gene; and an EvaGreen assay (Marangi et al., 2015) that targets the internal transcribed region 2 (ITS2). The TaqMan real-time PCR assays were performed in a Mx3000P™ thermocycler, whereas the DNA binding dye assays were performed in an AriaMx thermocycler to facilitate high-resolution melt (HRM) analysis. A 2-µl aliquot of template DNA was added to each reaction.

We used previously described reaction conditions for the EvaGreen assay (Marangi et al., 2015). For the SYBR Green assay, we only used the Cyclospora-specific primers (CycloF and CycloR) (Lalonde & Gajadhar, 2011). We performed the Varma assay with the conditions described in the original publication (Varma et al., 2003) and with various modifications in attempts to improve amplification performance. The results presented here were obtained using QuantiTect Probe PCR Master Mix (QIAGEN) instead of AmpliTaq Master Mix with BSA. We improved the specificity of the Verweij assay by making two modifications: 1) increasing the annealing/extension temperature from 60°C to 67°C, and 2) correcting the reverse primer sequence to 5’-AAT GCC ACG GTA GGC CAA TA-3’ (the reverse primer sequence in the original publication by Verweij et al. was missing a base, in comparison with the target gene). We also replaced Platinum QPCR Supermix (Life Technologies) for the reagents from Eurogentec and decreased the primer and probe concentrations to 0.5 µM and 0.1 µM, respectively.

The pIC is a recombinant plasmid with the cDNA of an Arabidopsis thaliana gene (GenBank accession number NM_114612.3) inserted into vector pZErO-2. The pIC plasmid was linearized by PstI digestion and diluted to 0.1 pg/µl before adding it to stool during DNA extraction. DNA samples that contained the pIC were analyzed in a real-time PCR assay containing 0.1 µM each of the primers IAC Fw and IAC Rv (Duffy et al., 2013) and 0.05 µM of the TaqMan probe ICP2 (5’-HEX-CCACTGCTAAAGGTAGCCCACGTC-BHQ1-3’), using standard TaqMan cycling structure.

The proportions of samples that were correctly identified were tabulated and confidence intervals were calculated using the Clopper-Pearson method (Clopper & Pearson, 1934). We compared the performances of the FastDNA and UNEX-based extraction methods by calculating the mean Ct value for each method. We assumed that the lower the mean Ct value the better the performance of the DNA extraction method. To evaluate operator-associated variability, two laboratorians processed the same specimens independently. In the statistical analyses, the negative samples, which, by definition, were not assigned Ct values, were considered as right-censored outcomes, in an adaptation of a previously described model for left-censored data (Jin et al., 2011). Thus, a model was fit to the data in SAS software version 9.3 (SAS Institute, Inc., Cary, NC, USA) to estimate the mean difference in the Ct values between the two methods, for each laboratorian; an interaction term between the method and laboratorian was included in the model, and a 5% level of significance was used.

The diagnostic performances of the assays were evaluated as described elsewhere (Griner et al., 1981). In brief, sensitivity was calculated as the probability that the result was positive when C. cayetanensis-positive specimens were tested, and specificity as the probability that the result was negative when C. cayetanensis-negative specimens/samples were tested.

To compare the FastDNA and UNEX-based methods for DNA extraction from stool, two different laboratorians processed 25 C. cayetanensis-positive and 25 specificity-control specimens. Both laboratorians performed both methods in parallel (i.e., for a total of 200 DNA extractions) without knowing which specimens were positive or negative. The results are summarized in Table 1. All of the results obtained with the UNEX-based method agreed with the microscopy results (i.e., no false-positive or false-negative results). In contrast, three C. cayetanensis-positive samples tested negative for C. cayetanensis after DNA extraction with the FastDNA method, even though the pIC was amplified from all three samples. The statistical modeling supported the conclusion that UNEX was better than FastDNA at recovering Cyclospora DNA from stool: the estimated mean difference in the Ct values between the two methods was significantly different from zero for both laboratorians, indicating that the UNEX-based method had a lower estimated Ct mean (Table 2). Furthermore, the FastDNA method had more manual pipetting steps and was therefore more labor-intensive than the UNEX-based method.

We compared the performances of the four real-time PCR assays with that of the conventional nested PCR assay using 32 Cyclospora-positive human stool specimens and 125 specificity controls (Table 3). The main difference among the assays was in their analytical sensitivities (i.e., detection limits). The EvaGreen assay was the most sensitive, with a detection limit of two oocysts per DNA extraction (i.e., seven oocysts per ml of stool). The Verweij and SYBR Green assays had the same detection limit as the conventional nested assay (i.e., 15 oocysts per DNA extraction, or 50 per ml), whereas the detection limit for the Varma assay was 200 oocysts per sample (~700 oocysts/ml). The Varma assay also failed to detect C. cayetanensis DNA in three of the Cyclospora-positive human stools, such that its diagnostic sensitivity was 93%.

The specificities were 100% for both the conventional nested PCR and the Varma assays, as neither assay amplified any of the specificity controls. The specificity of the Verweij assay was dependent on the annealing temperature; the assay amplified DNA from several of the simian Cyclospora samples when the previously published annealing temperature of 60°C was used (data not shown) but was specific for C. cayetanensis when the annealing temperature was increased to 67°C.

The PCR primers used in the SYBR Green assay were originally designed to detect and distinguish multiple coccidian parasites using HRM analysis. In this study, these primers amplified C. cayetanensis, simian Cyclospora spp., and Eimeria spp. DNA. Fig. 1 depicts the results of the HRM analyses for representative study samples. Eimeria spp. displayed a distinct melt curve profile with a main peak at ~86°C. Cyclospora spp. displayed a melt curve pattern with three peaks. All of the C. cayetanensis specimens had one melt peak in the range of 79.2–79.8°C. Whereas most (14 of 16) of the simian samples had a peak in the range of 80.2–80.8°C, the melt peaks were 79.4°C for one C. papionis sample and 79.6°C for one C. cercopitheci sample. The other two peaks in the three-peak pattern (at ~85.5°C and 88.5°C) overlapped between C. cayetanensis and the simian Cyclospora samples. Therefore, the SYBR Green assay could not reliably distinguish C. cayetanensis from the simian Cyclospora spp., resulting in an analytical specificity of 85%.

The EvaGreen assay was designed to specifically amplify C. cayetanensis DNA. All of the C. cayetanensis specimens in this study displayed a single melt peak at 85.0–85.5°C. Unfortunately, two of the human specificity controls (one stool positive for microsporidia and one parasite/microsporidia-free stool) also produced PCR products with similar melt curves as those of the C. cayetanensis-positive specimens (Fig. 2). Thus, the EvaGreen assay had a 98% diagnostic specificity in this study.