Development and evaluation of an off-the-slide genotyping technique for identifying Giardia cysts and Cryptosporidium oocysts directly from US EPA Method 1623 slides

Authors


Correspondence

Eric N. Villegas, National Exposure Research Laboratory, US Environmental Protection Agency, Mailstop: 587, 26 W. Martin Luther King Drive Cincinnati, OH 45268, USA. E-mail: villegas.eric@epa.gov

Abstract

Aims

This study developed and systematically evaluated performance and limit of detection of an off-the-slide genotyping procedure for both Cryptosporidium oocysts and Giardia cysts.

Methods and Results

Slide standards containing flow-sorted (oo)cysts were used to evaluate the off-the-slide genotyping procedure by microscopy and PCR. Results show approximately 20% of cysts and oocysts are lost during staining. Although transfer efficiency from the slide to the PCR tube could not be determined by microscopy, it was observed that the transfer process aided in the physical lysis of the (oo)cysts likely releasing DNA. PCR detection rates for a single event on a slide were 44% for Giardia and 27% for Cryptosporidium, and a minimum of five cysts and 20 oocysts are required to achieve a 90% PCR detection rate. A Poisson distribution analysis estimated the relative PCR target densities and limits of detection, it showed that 18 Cryptosporidium and five Giardia replicates are required for a 95% probability of detecting a single (oo)cyst on a slide.

Conclusions

This study successfully developed and evaluated recovery rates and limits of detection of an off-the-slide genotyping procedure for both Cryptosporidium and Giardia (oo)cysts from the same slide.

Significance and Impact of the Study

This off-the-slide genotyping technique is a simple and low cost tool that expands the applications of US EPA Method 1623 results by identifying the genotypes and assemblages of the enumerated Cryptosporidium and Giardia. This additional information will be useful for microbial risk assessment models and watershed management decisions.

Introduction

Waterborne outbreaks of cryptosporidiosis and giardiasis have occurred and continue to occur worldwide; therefore, exposure to these parasites can pose a significant health risk (Craun et al. 2010; Brunkard et al. 2011; Hlavsa et al. 2011). Currently, there are 26 Cryptosporidium species and more than 50 genotypes described in the literature (Morgan-Ryan et al. 2002; Fayer and Santin 2009; Plutzer and Karanis 2009; Fayer 2010; Fayer et al. 2010; Robinson et al. 2010; Smith and Nichols 2010; Elwin et al. 2012b; Ren et al. 2012). The two primary agents of human cryptosporidiosis are C. hominis and C. parvum, although others including: C. meleagridis, C. ubiquitum, C. cuniculus, C. canis, C. felis, C. suis, C. andersoni-like and the skunk, horse, monkey and pig II genotypes have been reported in humans (Leoni et al. 2006; Xiao and Fayer 2008; Chalmers et al. 2011; Elwin et al. 2012a). For giardiasis, there are six described Giardia species, but only G. duodenalis (syn. G. lamblia or G. intestinalis) is known to cause disease in humans. Giardia duodenalis is further subdivided into genetic assemblages (A-H) of which only A and B infect humans (Xiao and Fayer 2008; Feng and Xiao 2011). The Long Term 2 Enhanced Surface Water Treatment Rule (LT2) was promulgated to reduce the risk of waterborne microbial contamination of drinking-water systems with Cryptosporidium, Giardia, enteric viruses and other emerging pathogens. The LT2 requires drinking-water utilities to monitor their source water by US EPA Method 1622 or 1623. Based upon these monitoring results, utilities deemed at higher risk for oocyst contamination were required to enhance their water treatment to achieve appropriate removal of Cryptosporidium oocysts (US EPA 2006).

US EPA Method 1623 was developed to detect Giardia cysts and Cryptosporidium oocysts from water. The method involves concentrating and isolating parasites from a sample by filtration, elution, centrifugation and selective isolation by immunomagnetic separation. The isolated parasites are dried onto a well slide, stained with genus-specific immunofluorescent antibodies and 4′, 6 diamidino-2-phenylindole (DAPI). After staining, the parasites on the slide are microscopically identified and counted. Although many Giardia and Cryptosporidium species share respective sizes, morphologies and antibody staining properties, they cannot always be identified by internal structures, and therefore, all cysts and oocysts detected by Method 1623 must be considered potentially pathogenic (Fall et al. 2003; US EPA 2005; Feng and Xiao 2011). Although Method 1623 results are used in the United States to determine whether additional water treatment, such as enhanced disinfection, is required (US EPA 2006), additional applications such as improved and more accurate quantitative risk assessment models and watershed management decisions are limited because the parasites detected are identified only to the genus level.

To better characterize Cryptosporidium oocysts from Method 1623 slides, several studies have described or evaluated off-the-slide genotyping methods from clinical or environmental samples (Amar et al. 2001, 2002; Ruecker et al. 2005, 2007, 2011; Nichols et al. 2006, 2010; Di Giovanni et al. 2010). They all use a similar framework which involves removing the coverslip, adding a lysis or transfer buffer to the sample well, scraping the well mechanically and transferring the scraped sample to a tube. The transferred oocysts in the tube are lysed, amplified by single or nested PCR and finally genotyped by a molecular method such as RFLP, melt curve or sequence analysis. To date, one laboratory has described off-the-slide genotyping for both Cryptosporidium and Giardia from a single sample slide (Almeida et al. 2010a,b). Other identification approaches, such as FISH, have not as yet been adapted for Method 1623 slides (Dorsch and Veal 2001; Lemos et al. 2005).

Based upon our experience, most environmental samples analysed for Cryptosporidium and Giardia using Method 1623 typically have between zero and three (oo)cysts per slide. Thus, it is critical that any slide genotyping protocol effectively transfers all cysts and/or oocysts for molecular analysis. In a study determining the genotypes of microscopically examined water sample slides, Nichols et al. successfully genotyped slides with one oocyst 44·5% of the time and did not achieve greater than 90% genotyping success until at least eight oocysts were present on the slide (Nichols et al. 2010). Ruecker et al. also report similar results from water sample slides (Ruecker et al. 2011). These were environmental samples, and therefore, the oocyst species and quality likely varied both within and between slides. Similarly, Reucker et al. reported a 60% detection rate from slides with a single flow-cytometry-sorted oocyst, and a greater than 90% detection rate with three oocysts for both C. andersoni and C. parvum (Ruecker et al. 2011). Di Giovanni et al. evaluated C. muris and C. parvum recoveries from slides with a single flow-sorted oocyst and reported recoveries of 83% and approximately 50% in a single and multilaboratory study, respectively (Di Giovanni et al. 2010)(George Di Giovanni, personal communication). Two studies report a sensitivity of 50 or 100 Giardia cysts using an off-the-slide approach; however, lower cyst numbers were not studied (Almeida et al. 2010a,b).

Although several studies describe the sensitivity of off-the-slide genotyping for Cryptosporidium with low oocyst concentrations, similar data have not been published for either Giardia cysts or for both parasites from the same slide. In addition, the critical steps at which parasites are lost during the process have not been thoroughly evaluated. This study systematically evaluated the sensitivity of off-the-slide genotyping for both Giardia cysts and Cryptosporidium oocysts from the same slide at low concentrations using flow-cytometry-enumerated cysts and oocysts. Recoveries and losses at each critical steps of the off-the-slide method were evaluated including: (i) after staining, (ii) during the removal of the coverslip, (iii) after removal of the coverslip and mounting media, and (iv) on the well slide after scraping. Off-the-slide detection rates for 1, 2, 3, and 5 cysts and the same number of oocysts per slide were also determined by endpoint PCR. Results from this study address the overall efficiency and sensitivity of an off-the-slide approach for genotyping both Giardia and Cryptosporidium obtained from a single slide. It also provides some insights on the reliability of genotyping results, and how to better interpret and address issues like false-negative genotyping results from slides with very low concentrations of cysts or oocysts.

Materials and methods

Parasites

Cryptosporidium parvum (Iowa strain) oocysts were propagated in immunosuppressed mice and purified by sucrose and caesium chloride flotations (Sigma, St. Louis, MO, USA) (Ware and Villegas 2010). Oocysts were used within 3 months of collection. Giardia duodenalis (H3 strain; Assemblage B) cysts were propagated in Mongolian gerbils and purified by sucrose flotation followed by Percoll (GE Healthcare Life Sciences, Piscataway, NJ, USA) sedimentation (Belosevic et al. 1983; Sauch 1984). Cysts were used within 14 days of collection. All animal studies were approved and overseen by the Cincinnati US EPA Institutional Animal Care and Use Committee. Some cysts and oocysts were heat inactivated for 10 min at 80°C (Ware et al. 2003).

Flow cytometry

Sorts were performed by either a FACS Vantage SE or a FACS Aria II (BD Bioscience, San Jose, CA, USA) using Isoton II or 1 × phosphate-buffered saline (PBS) as sheath fluid, respectively. The parasites were gated by forward-side scatter profiles (FSC/SSC). Parasites were sorted onto well slides with 50-μl reagent water. The slides were either dried at room temperature, on a 35°C slide warmer, or overnight at 4°C, as described in Method 1623 without methanol fixation (US EPA 2005). The total number of parasites sorted ranged from 1 to 200. For the blinded slides, an equal number of cysts and oocysts were sorted onto randomly numbered prelabelled slides. All sorts were verified by sorting 100 parasites into a tube containing 100-μl 0·01% Tween-20 which was then filtered through a 0·8-μm polycarbonate filter (GE Healthcare Life Sciences, USA), stained and counted using fluorescence microscopy as previously described (Ware and Schaefer 2005). The sort event was used only if the mean of the triplicate control was between 93 and 101 (oo)cysts with a relative standard deviation ≤5%.

Slide staining and enumeration

All studies used Dynal Spot-on (Invitrogen, Carlsbad, CA, USA) or Superstick slides (Waterborne, Inc., New Orleans, LA, USA). Slide staining and mounting were performed using Aqua-Glo (Waterborne, Inc.) as described in Method 1623, except the wash steps after antibody and DAPI staining were omitted. The stained slides were mounted with laboratory-made 2% DABCO mounting media described in Method 1623 (US EPA 2005). The slides were sealed with nail polish and analysed by microscopy using the criteria described in Method 1623 (US EPA 2005). Solutions containing formalin were not used in this study.

Microscopy

After identification and enumeration, the slide was placed on a clean laboratory tissue, the nail polish was removed using an applicator stick without solvent, and the coverslip was removed using a sterilized razor blade by gently prying it up from a corner. The freed coverslip was immediately counted by microscopy after it was inverted onto a glass slide with 2-μl mounting media. Likewise, the well slide was counted after removing the mounting media by inverting the slide onto a laboratory tissue and then immediately returning it to the upright position. Thirteen additional slides were used to evaluate removing the mounting media with a 50-μl 1 × PBS wash (Nichols et al. 2006; Di Giovanni et al. 2010).

The effects of scraping on cyst and oocyst quality and number were evaluated with three slides sets in triplicate with 200 cysts and 200 oocysts. The slides were allowed to air dry and 50 μl of water was added to each well. One set was scraped with one passage by a bacterial loop, and a second set was scraped by one passage by a foam cell (3 × 3× 2 mm, Wrightway Sports, Sandy, UT, USA) (Nichols et al. 2006; Di Giovanni et al. 2010). All were allowed to air dry again, stained, mounted and counted by microscopy as previously described. The cysts and oocysts were counted and classified based on morphology and appearance using differential interference contrast microscopy (DIC) and were considered intact if there was a complete oo(cyst) wall with either visible or amorphous internal structures.

Twelve slides were used to microscopically determine the overall transfer efficiency of the loop scraping method. In this study, 200 oocysts were sorted onto slides and scraped by loop as described by Nicols et al. except that sample was transferred through a 0·8-μm polycarbonate filter (Nichols et al. 2006). The filter was stained, mounted and counted. The cleared wells and coverslips were also counted (Ware and Schaefer 2005).

Polymerase chain reactions

For these studies, a modified procedure described by Di Giovanni et al. was used which scraped the slides with a closed cell foam instead of a loop (Di Giovanni et al. 2010). A 1·5-ml screw cap centrifuge sample tube was prepared by adding 30-μl of a 1:1 Chelex-100 per water slurry (v/v), pH 7·0 (Bio-Rad, Hercules, CA, USA). The coverslip was removed, placed face up on the tissue, carefully rinsed with 5-μl sterile 0·01% Tween-20, and the rinsate was transferred to the sample tube. Twenty μl of 0·01% Tween-20 was added to the well and scraped with four up and down followed by two left and right passages across the well surface. The 0·01% Tween-20 on the slide was transferred to the sample tube. Another 20 μl of 0·01% Tween-20 was then added to the well, and it was scraped with two additional passages across the well surface. The foam and the final 0·01% Tween-20 solution on the slide were transferred to the sample tube. The collected sample was then lysed by eight freeze thaw cycles (liquid nitrogen to 95°C). The entire sample was transferred using a large bore pipette tip to a Spin-X filter tube (Corning, Lowell, MA, USA), and the sample tube was rinsed with 5-μl molecular-grade water. The Spin-X sample tube was centrifuged for 1 min at 21 000 g, and then the insert containing the Chelex-100 and the sponge was discarded. The final volume in the sample tube was approximately 60 μl. Samples were either immediately analysed by PCR or stored at −20°C.

For each sample, two separate PCRs were performed: a duplex Cryptosporidium PCR and a Giardia PCR. All PCRs were performed in duplicate using the ABI 9700 (Applied BioSystems, Carlsbad, CA, USA) and used the same cycle profile (95°C for 10 min; 55 cycles of 95°C for 30 s, 60°C for 1 min and 72°C for 30 s; 72°C for 10 min). The duplex Cryptosporidium PCR used 20 μl of sample template per reaction and amplified a 346-bp segment of the heat-shock protein 70 (hsp70, primers CPHSPT2F and CPHSPT2R (Di Giovanni and LeChevallier 2005)) and a 435-bp segment of the small subunit rRNA (SSU, primers CPB-DIAGF and CPB-DIAGR (Johnson et al. 1995)) targets (Di Giovanni et al. 2010). Each PCR consisted of 20 μl of template DNA into a total reaction volume of 50 μl with: 3 mmol l−1 MgCl2, 200 μmol l−1 dNTP mix, 2·5 U of AmpliTaq Gold polymerase (Applied Biosystems), 200 nmol l−1 of both forward and reverse primers and 0·75 μg μl−1 nonacetylated bovine serum albumin (Sigma). The Giardia PCR assay that amplified a 300-bp SSU rRNA target used the AL4303 and RH4 primers (Hopkins et al. 1997; Sulaiman et al. 2003). Each PCR consisted of 7 μl of template DNA into a total reaction volume of 50 μl with: 2 mmol l−1 MgCl2, 200 μmol l−1 dNTP mix, 2·5 U of AmpliTaq Gold polymerase (Applied Biosystems), 200 nmol l−1 of forward and reverse primers and 5% DMSO (v/v) (Sigma). All PCR products were analysed by 2% ethidium bromide e-gels (Invitrogen). The gel results were used for scoring PCR results. A PCR was considered positive if either replicate had an appropriate band size for the respective PCR. Positive PCR controls were slides containing 100 or 20 cysts and 100 or 20 oocysts. Likewise, negative controls were no template controls and blank slides processed through the extraction protocol.

Data and statistical analyses

Data and statistical analyses were performed using Excel 2007 (Microsoft, Redmond, WA, USA). Student's t-test analysis was used to determine statistical significance among the different conditions tested. A P-value of less than 0·05 was considered significant. The 95% rate of PCR positives was estimated by four parameter logistic regressions using SigmaPlot (version 11.0, Systat Software, Inc. San Jose, CA, USA). PCR target density was estimated with presence/absence data for 1, 2 and 3 (oo)cysts via the Poisson distribution function eqn 1.

display math(1)

where x denotes the number of events in a fixed interval, and λ denotes the known average rate of occurrence. Confidence limits were estimated by MPN software implemented in Excel (Jarvis et al. 2010). The geometric mean density (G) per (oo)cyst was calculated by eqn 2.

display math(2)

where N denotes the total number of flow-cytometry experiments, mi denotes the number of (oo)cysts on flow cytometry prepared slides of the ith experiment and δi denotes the PCR target density/lysate of the ith experiment. The Poisson distribution was used to estimate the PCR target density (number of amplicons), which expresses the probability of the occurrence of a given number of events (x) in a fixed interval if these events occur with a known average rate (λ). Single-dilution density estimates can be determined with the Poisson distribution by eqn 3.

display math(3)

where n and x denote the total number of replicates and the number of PCR positives, respectively, and n –x is the number of negatives (Shapiro 1999). For example, if the probability, P (x = 0), that is, zero sequences of the target are present is 0·05, then 0·05 = e−λ, which gives ln (0·05) = −λ; therefore, λ = 3. There are an average of 3 sequences (because λ is the mean of a Poisson distribution) of the target nucleic acid in a PCR aliquot. The PCR target density (δ) per slide was calculated by eqn 4.

display math(4)

where n and x denote the total number of replicates and the number of PCR positives, respectively, nx is the number of negatives, ν is the volume of lysate used for PCR and V is the total volume of the lysate. The number of replicates required to have a least one positive was calculated by eqn 5.

display math(5)

where nmin is the minimum number of replicates, pmin is the probability of a least one positive, ps is the probability of success/trial and ceiling (x) is the nearest integer greater than or equal to x (Ellison et al. 2006).

Results

Cryptosporidium and Giardia losses during the staining procedure

To evaluate the losses that may occur during staining of Cryptosporidium oocysts and Giardia cysts, (oo)cysts were flow-sorted directly onto microscope slides and not methanol fixed. When 200 Cryptosporidium oocysts were sorted onto slides and stained, the numbers of oocysts retained on the slide after staining were significantly less than the initial number sorted (8–16% average losses, < 0·05) (Table 1). Similar average staining losses were also observed with 200 sorted Giardia cysts (12–16%, < 0·05). To determine if these staining losses occurred with numbers of (oo)cysts closer to levels found in the environment, 20 (oo)cysts were sorted onto slides. Results revealed similar average losses of 15–24% for Cryptosporidium and 21–24% for Giardia. Results were not significantly different in respect to species or number of (oo)cysts sorted (P > 0·05).

Table 1. Microscopic counts from slides with flow-sorted parasites following staining and after coverslip and mounting media removal
ParasiteSlide1 n Sort totalNo. on stained well2 (avg ± CI)aNo. on well after coverslip removal (avg ± CI)aNo. on coverslip (avg ± CI)aAverage stain loss (% ± CI)aAverage on well after coverslip removal (% ± CI)aCalculated % remaining in mounting media (avg)
  1. 1D, Dynal Spot-on; W, Waterborne, Inc. Superstick.

  2. 2Column significantly different than previous (P < 0·05).

  3. a

    CI, 95% confidence interval.

Cryptosporidium parvum D6200185 ± 3131 ± 252·8 ± 3·98 ± 271 ± 1428
W18200167 ± 9137 ± 138·2 ± 4·116 ± 582 ± 613
D332015·2 ± 1·412·1 ± 1·80·1 ± 0·224 ± 776 ± 823
W332017·0 ± 0·812·7 ± 1·30·8 ± 0·615 ± 475 ± 721
Giardia duodenalis D5200158 ± 1392 ± 2711·6 ± 10·716 ± 760 ± 1133
W5200176 ± 12131 ± 196·0 ± 7·612 ± 674 ± 723
D102013·7 ± 1·96·0 ± 1·35·1 ± 2·124 ± 763 ± 1116
W102015·8 ± 1·511·7 ± 2·70·4 ± 0·721 ± 774 ± 1623

Another potential factor that may contribute to the staining losses observed is the type of microscope slide used. Table 1 also compares staining losses, using a modified Aqua-glo staining procedure, for two commonly used slides in Method 1623, Waterborne, Inc. Superstick (W) and Dynal Spot-on (D). No significant differences in average losses after staining were observed between the two slides with 20 sorted Cryptosporidium (D 24 and W 15%) and 20 sorted Giardia (oo)cysts (D 24 and W 21%). The average Giardia losses with 200 sorted cysts were also similar for both slides. In contrast, when 200 Cryptosporidium oocysts were sorted, significantly more oocysts were lost on average using waterborne slides when compared to Dynal slides (W 16 and D 8%, = 0·002). Overall, the average staining losses for C. parvum oocysts and G. duodenalis cysts ranged between 8 and 24% and 12–24%, respectively. Similar staining losses were observed with heated (oo)cysts (data not shown).

Parasite retention following removal of the microscope coverslip and mounting media

Removal of the coverslip and mounting media could result in (oo)cyst losses, and the number retained on the slide was determined after removal of the coverslip and mounting media. Results indicated that there was not a significant difference between slides that were blotted or washed with 50 μl of PBS (data not shown) (Nichols et al. 2006; Di Giovanni et al. 2010). As shown in Table 1, on average 71–82% of the Cryptosporidium oocysts and 60–74% of the Giardia cysts were retained on the well after coverslip and mounting media removal, which was significantly less than the stained average (P < 0·05). The removed coverslips were also examined to determine if parasites were present. Results revealed that less than 7% of the initially stained cysts or oocysts were retained by the coverslip, except for one group of Giardia which retained 14% of the cysts (Table 1). Based upon the numbers initially detected on the slides, an average of 18 and 22% of the oocysts and cysts, respectively, were lost while removing the mounting media.

Evaluation of two scraping procedures used to remove Cryptosporidium and Giardia from slides

To evaluate the effects of scraping on cysts and oocysts from Method 1623 slides, the number and DIC appearance of parasites on unscraped control slides were compared to slides which had been scraped using two previously published procedures (bacterial loop and foam cell) (Nichols et al. 2006; Di Giovanni et al. 2010). In these studies, the slides were scraped by one passage, and the slides were then re-evaluated after staining without a transfer step. All slides initially had 200 cysts and 200 oocysts and were stained with fluorescent antibody after the second air-drying step. Control samples contained an average of 165 oocysts and 169 cysts remaining on the slide after staining, similar to results reported in Table 1 (Fig. 1). By contrast, the average number on scraped slides was significantly less than the control sample average (Giardia: Loop = 72, Foam = 118; Cryptosporidium: Loop = 114, Foam = 123, all P < 0·05) (Fig. 1).

Figure 1.

Microscopic analysis evaluating two scraping procedures on well slides containing 200 flow-sorted Cryptosporidium parvum oocysts (a) and Giardia duodenalis cysts (b) flow-sorted onto the slide and allowed to dry. Slides were unscraped (Control), scraped with a bacterial loop (Loop), or with a foam-cell scraper (Foam) and allowed to dry. Slides were then stained with fluorescent antibodies and examined by epifluorescence and differential interference contrast (DIC) to determine cyst and oocyst number and structural integrity. Results are the average of 4 independent experiments. Error bars indicate 95% confidence intervals. *indicates significance different than from control slides (< 0·05).

The proportion of intact C. parvum oocysts decreased from 88% on control slides to 50 or 46% after being scraped by a bacterial loop or a foam-cell scraper, respectively. Although the numbers of cysts and oocysts remaining on scraped slides were both significantly less than the control group, the proportion of intact Giardia cysts only decreased from 100 to 88% in loop scraped and 84% in foam-cell scraped samples. Microscopic inspection of the cyst and oocyst appearance found intact (oo)cyst walls and either detectable or amorphous internal structures on unscraped controls, and empty cysts and oocysts on a foam scraped slides (Fig. 2). Similar results were observed with bacterial loop-scraped slides (data not shown).

Figure 2.

Differential interference contrast (DIC) photomicrograph of Cryptosporidium parvum oocysts and Giardia duodenalis cysts on control and scraped slides. C. parvum and G. duodenalis organisms were flow-sorted onto the slide, allowed to dry and 50 μl of water was added to the well. One slide was scraped with a foam-cell and both were allowed to air dry again. Slides were then examined by DIC to determine (oo)cyst integrity. Panel (a) illustrates Giardia cysts (arrow) and Cryptosporidium oocysts (arrowhead) from control slide samples showing intact or amorphous internal structures for both organisms. Panel (b) shows Giardia cysts (arrow) and Cryptosporidium oocysts (arrowhead) slides that have been scraped with a foam cell. Empty (oo)cyst walls and unidentified debris likely from destroyed cysts and oocysts were evident throughout the slide. Scale bar = 5 μm.

Detection rates of the off-the-slide PCR-based typing using low amounts of flow-sorted Cryptosporidium oocysts and Giardia cysts

For the slides seeded with low numbers of Giardia cysts, the PCR detection rates were 44, 77, 86 and 95% for 1, 2, 3 and 5 cysts, respectively (Fig. 3). Although the duplex C. parvum PCR detects both hsp70 and SSU rRNA genes, the percentage of slides that were hsp70 PCR positive was far less than the percentage of SSU rRNA PCR positive (22 and 41%, respectively). The results achieved with the hsp70 amplification were inconsistent (data not shown). Therefore, the Cryptosporidium positive samples were scored using only the presence of SSU rRNA PCR amplicon. The Cryptosporidium results were lower than the Giardia results with PCR detection rates of 27, 48, 57 and 81% for 1, 2, 3 and 5 oocysts, respectively. Results also revealed that positive PCR detection of greater 95% was achieved when seed levels were above five Giardia cysts or 20 Cryptosporidium oocysts. Overall, these results confirm other studies reporting PCR detection rates from a single Cryptosporidium oocyst on a slide (Di Giovanni et al. 2010; Nichols et al. 2010; Ruecker et al. 2011); George Di Giovanni, personal communication).

Figure 3.

Off-the-slide PCR detection rates by parasite number from flow-cytometry prepared slides. Specific amounts of Cryptosporidium oocysts (a) and Giardia cysts (b) ranging from 1 to 100 organisms were flow-sorted onto the slide, stained, scraped and analysed by PCR. Results are the per cent of slides PCR positive for the SSU rRNA of Cryptosporidium and Giardia. The number in the bar graph denotes the number of slide replicates for each sample.

The PCR results presented in Fig. 3 were further analysed using the Poisson distribution to estimate the SSU rRNA PCR target densities (Table 2). The Cryptosporidium PCR target densities formed a positive, linear relationship (slope = 0·637, R2 = 0·950) with the quantity of flow-sorted oocysts on the microscope slides. A positive, linear relationship was also observed for Giardia (slope = 4·827, R2 = 0·979). A comparison of the ratios of the regression slopes and the geometric means suggested that there were 7·57 and 8·8-fold, respectively, more detectable SSU rRNA per Giardia cyst than for a Cryptosporidium oocyst. A comparison of the 95% confidence SSU rRNA PCR detection rates of Giardia (Fig. 4A, 9·2 cysts) and Cryptosporidium (Fig. 4B, 53·8 oocysts) echoed these findings. Combined, these comparisons suggested that PCR detection of Giardia was several times more sensitive than Cryptosporidium.

Table 2. PCR target density for Cryptosporidium and Giardia SSU rRNA genes
Parasite(Oo)cyst (No.)Density/slidea(δ)Lower CL-95%Upper CL-95%Density per (Oo)cystGeometric meana
  1. a

    See 'Methods and Results' for calculation.

C. parvum
SSU rRNA10·520·330·810·520·60
21·411·051·920·71
31·791·352·370·60
G. duodenalis
SSU rRNA15·294·036·945·295·30
211·339·4314·575·66
314·9412·0018·864·98
Figure 4.

Logistic regression analysis of off-the-slide Giardia and Cryptosporidium PCR target densities. The dotted lines show predicted number of (oo)cysts that gave a 95% PCR detection rate. (a) Giardia cysts (R2 = 0·973), (b) Cparvum oocysts (R2 = 0·966).

To better estimate the limits of detection for this off-the-slide genotyping assay, eqn (5), was used to calculate the minimum number of PCR trials (nmin) required for a single positive PCR (Ellison et al. 2006). Assuming a 95% probability of detection (i.e. pmin = 0·95), the nmin for 1, 2 and 3 (oo)cysts were 18, 7 and 6 for Cryptosporidium, and 5, 3 and 2 for Giardia.

Discussion

Combining US EPA Method 1623 with the off-the-slide genotyping assay provides the number of cysts and oocysts in a sample which are then further identified to species and/or genotype. This study systematically evaluates the slide manipulation steps associated with both Method 1623 and off-the-slide transfer for Giardia cysts and Cryptosporidium oocysts using microscopy and molecular assays. Parasite retention on well slides is critical because there are multiple stain and wash steps in Method 1623 and in off-the-slide genotyping procedures previously described (Ruecker et al. 2005; US EPA 2005; Nichols et al. 2006; Almeida et al. 2010a,b; Di Giovanni et al. 2010). The initial microscopic results in this study demonstrated that approximately 20% the parasites can be lost during staining with unfixed slides (Table 1). Similar losses during staining were observed with methanol-fixed slides (data not shown). The losses in this study, however, were minimized because the washes between the stain steps described in Method 1623 were eliminated (US EPA 2005). We also observed that during microscopic examination of Method 1623 slides, some cysts and oocysts tend to drift in the well, especially when using oil immersion. These results suggested that some of the performance variability and recovery issues observed in both US EPA Method 1623 and off-the-slide genotyping procedures could be due to cyst and oocyst losses during the application and removal of liquids to and from well slides; and thus when possible, wash steps should be eliminated. For the same reason, the DABCO mounting media should also be transferred along with the organisms for genotyping analysis.

Evaluation of off-the-slide transfer of whole oocysts revealed that the transfer efficiency was difficult to quantify by microscopy as over half of the oocysts were not observed in the transferred sample (data not shown). Further microscopy revealed that many cysts and oocysts dried onto a slide were destroyed or lost their structural integrity after being scraped by a bacterial loop or foam cell (Figs 1 and 2). This marked reduction in the total numbers of intact (oo)cysts retained on the slides suggested that the scraping procedure aided in the physical disruption of the organisms for more efficient nucleic acid extraction procedures necessary for improving PCR analysis and detection limits. Indeed, previous studies reported that efficient amplification from low concentrations of intact oocysts required multiple freeze thaw cycles and glass-bead beating to release DNA, but the current off-the-slide procedures described in our study do not use this approach because the drying and scraping steps effectively disrupted the cyst and oocyst walls releasing their DNA contents (Lindergard et al. 2003; Hill et al. 2007).

Some off-the-slide genotyping methods use only the Cryptosporidium SSU rRNA gene to characterize the oocysts (Ruecker et al. 2005; Almeida et al. 2010a,b). While using a single locus to genotype may be appropriate for many samples, some, especially those with multiple G. duodenalis isolates, may require the results from multiple loci to obtain more precise identification (Caccio et al. 2008; Bonhomme et al. 2011). Multiplex PCRs for two or more loci are difficult to optimize and require further optimization for multiple PCR platforms. Multiple nested PCRs would also allow for multilocus genotyping but require rigorous quality control.

Although this study used only the SSU r RNA Giardia loci, the hsp 70 and SSU rRNA Cryptosporidium genes were performed using a duplex PCR(Di Giovanni et al. 2010). Our duplex PCR results showed that hsp70 gene was less frequently detected than the SSU rRNA gene despite C. parvum containing both genes. This could in part be due to differences in gene copy numbers between the two loci, the hsp 70 is a single copy gene while the SSU rRNA gene has at least five copies in the Cryptosporidium genome (Khramtsov et al. 1995; Le Blancq et al. 1997). In addition, this duplex reaction was optimized for the Rotor-Gene 6000 (Qiagen) which was not used in this study. Di Giovanni et al. have reported variable results with different PCR platforms (Di Giovanni et al. 2010). Our results demonstrated that there were off-the-slide genotyping false-negative results for slides with 1–10 oocysts. The proportion of genotyping negative results from slides with <5 oocysts in this study were similar to other studies, without using the entire sample or nested PCR (Ruecker et al. 2005; Nichols et al. 2006; Di Giovanni et al. 2010). More importantly, our off-the-slide genotyping procedure also genotyped Giardia cysts present on the same slide.

Another critical component analysed in this study was the limits of detection by PCR for this off-the-slide procedure. This study employed PCR target density analysis to estimate the limits of detection of the SSU rRNA PCR assays used for Cryptosporidium oocysts and Giardia cysts. This analysis revealed a much higher PCR target density for Giardia than Cryptosporidium (geometric means 5·30 and 0·60, respectively) (Table 2, Fig. 4). To help explain why the Giardia off-the-slide PCR results were more successful than Cryptosporidium, we recalculated the number of SSU rRNA gene copies per Giardia cyst by using the copy number originally described by Boothroyd et al., and the12-Mb Giardia genome size described by Adam, et al. (Boothroyd et al. 1987; Adam 2000). This revised calculation showed that there are approximately 9 SSU rRNA gene copies per haploid genome, and given that mature Giardia cysts are 16N (Bernander et al. 2001), we estimated the SSU rRNA gene copies per cyst to be 146, which is greater than the SSU rRNA gene copies (20) for Cryptosporidium oocysts (Le Blancq et al. 1997). Thus, assuming similar recovery efficiencies, the PCR detection of Giardia cysts should be 7·3 times more sensitive than Cryptosporidium oocysts (i.e. 146/20). This ratio is remarkably similar to the ratio of the Giardia and Cryptosporidium PCR target density regression slopes (7·57) and the ratio of geometric means (8·8). It explains why the Giardia off-the-slide PCR results were more successful for 1, 2 and 3 cysts when compared to corresponding Cryptosporidium results, despite using only 7 μl of sample per Giardia PCR compared to 20 μl per Cryptosporidium PCR (Le Blancq et al. 1997). Our experimentally determined PCR target densities estimate that 5, 3 and 2 PCR replicates were required to obtain the 95% limit of detection for 1, 2 and 3 Giardia cysts, respectively, compared to 18, 7 and 6 PCR replicates required for 1, 2 and 3 Cryptosporidium oocysts, respectively. Using nested PCR could increase the amount of target DNA, but this can create the need for multiple PCRs and requires more rigorous quality control to minimize cross-contamination inherent of conducting nested PCRs. A promising solution is the whole genome amplification (WGA) procedure, which in one reaction, could provide sufficient DNA for archiving and multilocus genotyping of both Cryptosporidium and Giardia (Bouzid et al. 2010).

Off-the-slide genotyping is a relatively simple and low cost tool that provides utilities additional information obtained from Method 1623 slides. These include the ability to identify sources of contamination, improve microbial risk assessment models and improve watershed management practices associated with both Cryptosporidium and Giardia. Interpretation of Method 1623 results is complicated and profoundly affected by the sample matrix. Method 1623 uses genus-specific antibodies that have algal interferences and has known recovery issues (Rodgers et al. 1995; DiGiorgio et al. 2002; Francy et al. 2004). This study shows that Method 1623 staining and washing steps result in average significant losses of approximately 20% of the (oo)cysts.

Likewise, interpretation of Method 1623 off-the-slide genotyping results is complicated and also potentially influenced by the sample matrix. This study shows that DNA is not always recovered from slides even those with fresh parasites and no environmental inhibition. Naturally occurring compounds can inhibit PCR; however, Method 1623 slides have been processed by IMS to selectively isolate (oo)cysts from the debris. Environmental samples may contain more empty (oo)cysts likely devoid of DNA compared to this study, and Cryptosporidium genotyping has been positively correlated with the number of DAPI stained nuclei observed (Nichols et al. 2010). In addition, the reagents used in this study did not contain formalin because some formalin containing commercial staining reagents inhibit PCR (Di Giovanni et al. 2010).

Genotyping results of environmental samples demonstrate that some Method 1623 slides have multiple Cryptosporidium species/genotypes and/or Giardia Assemblages (Ruecker et al. 2005; Almeida et al. 2010a; Nichols et al. 2010). This study did not evaluate multiple strains of parasites; however, Ruecker et al. evaluated slides with mixtures of C. parvum and C. andersoni reported amplification bias with mixture ratios >1 : 10 resulting in losses of the less abundant species by nested PCR–RFLP (Ruecker et al. 2011). This bias can be alleviated by PCR amplification of limiting-dilution templates for ratios up to 1 : 20. Based upon these results, it is unlikely that there would be an amplification bias in samples with low (oo)cyst concentrations, such as most drinking-water samples. PCR biases may be a problem in wastewater or samples with very high concentrations. Similar mixture studies have not been performed with the PCRs used in this study (George Di Giovanni, personal communication). In samples using the internal (oo)cysts control ColorSeed™ (BTF Pty, North Ryde, NSW, Australia), off-the-slide genotyping may not be possible because of the potential to have a 1 : 100 amplification bias, which may not be alleviated by limiting dilutions.

Ideally, all results should be confirmed by sequencing to better identify the isolate(s) and to minimize genotyping false positives due to non-specific amplification. Our off-the-slide results show that only a proportion of slides containing 1, 2 and 3 (oo)cysts can be successfully amplified with Giardia being more sensitive than Cryptosporidium. PCR target density analysis and gene copy number affect the overall PCR sensitivity and explain the differences in results between the species. Incorporating PCR target density analysis into designing future PCR strategies would improve both Giardia and Cryptosporidium off-the-slides genotyping success rates.

Acknowledgements

We thank Drs. Sarah Staggs and Leah Villegas for critical reviews of the manuscript. The views expressed in this article are those of the authors and do not necessarily represent the views or policies of the U.S. Environmental Protection Agency. Mention of trade names, products or services does not convey, and should not be interpreted as conveying, official EPA approval, endorsement or recommendation. The authors have no conflicts of interest. MWW designed and conducted the experiments, analysed the results and wrote the manuscript; SPK analysed the results and wrote the manuscript; ENV designed the experiment, analysed the results and wrote the manuscript.

Ancillary