Environ. Sci. Technol. 2002, 36, 2754-2759
PCR Detection of Specific Pathogens in Water: A Risk-Based Analysis F R A N K J . L O G E , * ,† DONALD E. THOMPSON,‡ AND DOUGLAS R. CALL§ Department of Civil and Environmental Engineering, Washington State University, P.O. Box 642910, Pullman, Washington 99164-2910, Department of Civil and Environmental Engineering, 1 Shields Avenue, University of California at Davis, Davis, California 95616, and Department of Veterinary Microbiology and Pathology, Washington State University, Pullman, Washington 99164-7040
The relative concentration of pathogens in water samples collected from storm drains and adjacent surfaces was evaluated using established PCR-based protocols. Out of the 58 samples collected from 21 different storm drains, 22% were PCR positive for Escherichia coli ETEC, Salmonella, or adenovirus. The risk of swimming related illnesses associated with detection of E. coli ETEC and Salmonella ranged from 0.39 to 30:100 000 and 0.3-25:1000, respectively. The detection limit corresponding to a negative-PCR result was evaluated in reference to water quality standards developed using a risk-based approach that integrates human dose-response data with acceptable levels of risk promulgated by the U.S. EPA for recreational contact. The percent of samples with an acceptable detection limit ranged from 0% for Giardia lamblia and Shigella to 100% for E. coli ETEC. The principal factor influencing the detection limit of G. lamblia and Shigella was sample volume. The principal factor influencing the detection limit of the remaining bacteria and protozoa, including E. coli O157:H7, Salmonella, and Cryptosporidium parvum, was the presence of inhibitory compounds in the purified nucleic acid extracts. Both recovery and inhibition adversely impacted the detection limit of viruses. Ambient water quality standards based on the occurrence of specific pathogens enumerated with PCR-based assays could serve as a method of evaluating the biological quality of water but only after significant improvements in filtration and purification protocols. The riskbased methodology developed in this study can be used to evaluate future improvements in filtration and purification protocols.
Introduction There has been increasing public concern with the quality of recreational waters within the United States (U.S.), stimulated to a large extent by findings from increased * Corresponding author phone: (509)335-3227; fax: (509)335-7632; e-mail:
[email protected]. † Department of Civil and Environmental Engineering, Washington State University. ‡ University of California at Davis. § Department of Veterinary Microbiology and Pathology, Washington State University. 2754
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 12, 2002
monitoring efforts at state and local levels. Monitoring efforts have resulted in a significant number of beach closures within the past several years. For example, there were 6160 days of beach closures and advisories in the U.S. in 1999 alone, a 50% increase over 1997 (1). The beach closures or advisories were based on concentrations of indicator organisms that exceeded state ambient water quality criteria. Recent studies of the correlation between indicator organisms and nonhuman fecal pollution raise questions about using existing standards to assess biological quality and associated health risks of water (2). Alternatively, advances in molecular techniques have stimulated interest in directly monitoring for specific pathogens in nonpoint source runoff and recreational waters (3, 4). The underlying objective of most molecular-based studies has been to use polymerase chain reaction (PCR) to evaluate the presence of selected pathogens, commonly in a ( reporting scheme. These studies make no explicit statement about detection limits, sources of variation, or health risks associated with positive or negative results. Interpreting the result from a PCR assay is more complicated than simple presence or absence of a pathogen. A positive result certainly provides an indication that some microbial contamination is present in the sample, although we cannot assess health risks without additional information. More importantly, a negative result is very difficult to interpret without knowing the precise detection limit for the assay. Detection limit is complicated by a series of factors including the volume of water that is filtered, the efficiency of target recovery, and the presence of inhibitory compounds in the PCR reaction. Unfortunately, all three of these factors will vary between each sample, and if the analysis fails to consider these factors, the results will be very difficult to assess in the context of risk to human health. Some interpretative aspects of PCR assays could be improved by using quantitative methods such as MPN-PCR (5) or Taqman assays (6). Nevertheless, interpreting results from quantitative methods still requires an explicit assessment of the effects of sample volume and template recovery. Furthermore, all PCR assays will be affected to some extent by inhibitory compounds, which include substances that squelch fluorescent reporter systems (6). Ultimately, these variables must be coupled with a risk-based analysis before we can ascertain the biological significance of PCR positive and negative results. The goal of this study was to develop a framework for interpreting PCR results in the context of risk to human health. The specific objectives were to (i) develop water quality standards for selected pathogens using a risk-based approach that integrates human dose-response data with acceptable risks of illness promulgated by the U.S. EPA for recreational contact; (ii) develop a systematic methodology for quantifying detection limits of PCR-based assays; and (iii) test for the presence of specific pathogens in nonpoint source runoff and assess results in the context of risk to human health.
Experimental Section Selection of Pathogens for Analysis in Nonpoint Source Runoff. The following pathogens were selected for analysis because of their frequent association with waterborne-disease from drinking and recreational waters (1986-1998; CDC): E. coli ETEC, E. coli O157:H7, Shigella spp., Salmonella spp., enterovirus, rotavirus, hepatitis A, adenovirus, C. parvum, and G. lamblia. In addition, poliovirus was selected for analysis to subdivide the enterovirus group. 10.1021/es015777m CCC: $22.00
2002 American Chemical Society Published on Web 05/16/2002
Development of Recreational Water Quality Standards for Selected Pathogens. Human dose-response data was available for a subset of selected pathogens, including E. coli ETEC (7), Shigella (7), Salmonella (7), G. lamblia (8), C. parvum (9), and rotavirus (10). The dose-response model developed for Shigella was applied to E. coli O157:H7 (7). The risk of infection or illness was functionally characterized with either a beta-Poisson or an exponential model (see ref 7 as a general reference to microbial risk assessment)
(
Pbp(d) ) 1 - 1 +
d 1/R (2 - 1) N50
Pe(d) ) 1 - e-rd
)
-R
(1) (2)
where Pbp(d) and Pe(d) are the probability of either infection or illness based on a beta-Poisson or exponential model, respectively, for a given exposure, d, of a particular pathogen. N50, R, and r are model coefficients. The response endpoint for E. coli ETEC and Shigella was illness. For the remaining organisms, including Salmonella (11), G. lamblia (8), C. parvum (12), and rotavirus (13), the values of Pbp and Pe, based on an endpoint of infection, were multiplied by the appropriate morbidity ratio to obtain the corresponding probability of illness. The morbidity ratio was observed to be independent of dose for G. lamblia, C. parvum, and rotavirus (8, 12, 13). Recreational water quality standards (defined as an acceptable concentration of a particular pathogen) were developed using the above dose-response models with probabilities of illness equal to 8:1000 and 19: 1000 for fresh and marine waters (14), respectively, and the following assumptions (7): (i) 50 mL of water are consumed during each hour of swimming activity, (ii) 2.6 h/swim, and (iii) seven swims per year. The basis (e.g., annual or daily) for the acceptable probabilities of illness promulgated by the U.S. EPA (14) are not clearly delineated. These values were originally derived from indicator organisms that were enumerated from a limited number of epidemiological studies. Consequently, the EPA probabilities most likely represented daily risks of illness. For the present analysis, however, we chose to extrapolate these values to an annual basis (seven swims per year) and thereby produced a more conservative water quality standard. Sample Collection and Preservation. Water samples were collected from storm drains and adjacent surfaces at 21 locations throughout southern California. A brief description of seven selected locations is provided in Table S-1 (Supporting Information). Runoff was generated from the paved and unpaved surfaces by washing a 9 m2 area of the specified surface with 50 L of locally supplied tap water. All storm drains were sampled during dry weather flows. At each location water samples were collected in three 1-L Nalgene plastic containers for analyses of hardness, suspended solids, total and fecal coliforms, and enterococci as per Standard Methods, 20th edition (15), methods 2340C, 2540D, 9221B, 9221E, and 9230B, respectively. In addition, water samples, ranging in volume from 20 to 50 L, were collected in sterile 20-L Nalgene plastic containers for analysis of selected pathogens. These samples were concentrated onsite by pumping the water (stainless steel progressive cavity pump, Ryan-Hercoproducts Corp, Sacramento, CA) through a 1.0 µm pore size, 293 mm FALP filter (Millipore, Bedford, MA). Filters were prewetted with methanol and mounted on a sterilized stainless steel filter holder (Millipore). Filtrate was collected in sterile 20-L Nalgene plastic containers. The pH of the filtrate was adjusted to 5 with 1 N HCl, and the liquid was pumped (flexible impeller pump, Ryan-Hercoproducts Corp, Sacramento, CA) through a 1-MDS cartridge filter (CUNO, Inc., Meridian, CT). The 1-L water samples
and filters were shipped on ice to the University of California at Davis for analyses. Filters were eluted and concentrated within 24 h of arrival, and the resulting solutions were stored at -20 °C. The 1-L water samples were processed immediately for the specified water quality characteristics. Elution of FALP and 1-MDS Filters. The 293 mm FALP filters were cut in half and laid flat in a sterile plastic tray containing 100 mL of 1.5% (w/v) beef extract in 1X PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 2 mM KH2PO4, pH 7.4). The filter was eluted by scrubbing the surface with a sterile nylon bristle brush for 10 min. Extracts were collected in 50 mL centrifuge tubes; the brush was rinsed with elution buffer, and the liquid was pooled with the extracts. Tubes were centrifuged at 3000×g for 12 min. After removing the supernatant, the pellet was weighed and distributed to 2 mL screw-capped microcentrifuge tubes. The maximum mass allowed per tube was 300 mg. The 1-MDS cartridge filters were eluted and organically flocculated as per (16). The resulting pellet was dissolved in 8 mL of buffer (0.15 M Na2HP04‚7H2O, pH 7.5) and concentrated using a Biomax100K microconcentrator (Millipore) to a final volume of approximately 400 µL. Nucleic Acid Extraction and Purification. Bacterial and protozoan nucleic acids were extracted from the FALP pellet using mechanical glass bead lysis (17, 18) with a Mini-Bead Beater 8 (BioSpec Products, Bartlesville, OK) operated at 2510 rpm for 2 min. The DNA pellets resulting from the bead lysis procedure were redissolved in 100 µL of sterile doubledistilled water, and the solution was sequentially passed through PVPP (Sigma Chemical Co., St. Louis, MO) and Sephadex G200 (Amersham Pharmacia Biotech AB, Uppsala, Sweden) spin columns to remove substances that inhibit PCR (19). Viral nucleic acid was extracted using a QIAmp Viral RNA Kit (Qiagen Inc., Valencia, CA). The resulting nucleic acid extracts were passed through Sephadex G200 spin columns to remove substances that inhibit PCR. Nucleic Acid Amplification and Detection. The molecular techniques used to identify selected bacteria, viruses, and protozoa are summarized in Tables S-2 and S-3 (Supporting Information). Primers were synthesized by Life Technologies (Gibco BRL, Grand Island, NY) and stored in TE buffer at -20 °C. All PCR reactions were performed using a GeneAmp PCR System 9700 thermocycler (PE Biosystems, Foster City, CA). For each PCR assay, positive and negative controls were included. The PCR products were analyzed on a miniaturized microcapillary electrophoresis chip (Agilent Technologies Inc., Germany) using a 2100 Bioanalyzer chip reader (Agilent Technologies Inc., Germany). Positive-PCR results were confirmed with either multiple primer sets (see Table S-2 in reference to E. coli ETEC, adenovirus, or C. parvum) and the Bioanalyzer as the endpoint detector or Southern blots. Southern blots were performed by transferring the PCR product separated on a 1.5% agarose gel to a positively charged nylon membrane (Roche Molecular Biochemicals, Germany) using a Model 785 Vacuum Blotter (Bio-Rad, Hercules, CA) as per the manufacturer guidelines. Hybridization and colorimetric detection of bound probe (Table S-2) was performed using DIG Nucleic Acid Detection Kit (Roche Molecular Biochemicals, Germany). Baseline Sensitivity of PCR Assays. Nucleic acids were prepared from stock cultures of bacteria, virus, and protozoa for use as positive controls in PCR assays (Table S-2). Bacterial nucleic acid was prepared from a 200 µL aliquot culture (grown 24 h in nutrient broth, 37 °C) using a series of four rapid freeze-thaw cycles. Stock cultures of protozoa were supplemented with Amphotericin B to retard fungal growth and stored at 4 °C. Protozoa nucleic acid was extracted using the above freeze-thaw procedure. Stock cultures of the viruses were divided into small aliquots and stored at -20 °C. Viral nucleic acid was obtained using the QIAmp Viral VOL. 36, NO. 12, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2755
TABLE 1. Potential Range in Concentrations and the Corresponding Risks of Illness Associated with Samples that Produced a Positive Result in the PCR-Based Assaysa E. coli ETEC sample
concn (cfu/L)b
risk of illnessc,d
1 2 3 4 5 6 7 8 9 10 11 12 13
287.5-2875 338-3880 42.5-425 333-3330
2.6-26 3.1-31 0.39-3.9 3-30
Salmonella
adenovirus
concn (cfu/L)b
risk of illnessc,e
43.8-438 87.5-875 87.5-875 14.9-149 87.5-875 87.5-875 117-1170
0.9-9 1.8-18 1.8-18 0.3-3 1.8-18 1.8-18 2.5-25
concn (TCID50/L)b
risk of illness
5.19-51.9 51.9-519
NAf NAf
a Acceptable levels of risk promulgated by the U.S. EPA are 19:1000 and 8:1000 (1900:100 000 and 800:100 000) for marine and freshwaters, respectively. In this study, the acceptable levels of risk are assumed to be on an annual basis. b The range reflects the terminal two dilutions producing a positive and negative result in the 10-fold serial dilution sequence. c Annual risks were estimated as Pa ) 1 - (1-P)N, where Pa is the annual risk of illness, P is is the probability of illness based on a single exposure (e.g, either Pbp or Pe in eqs 1 and 2, respectively, adjusted with a morbidity ratio where appropriate), and N is the number of exposures per year (assumed to be 7 in this study). The parameter P (e.g., Pbp or Pe) was calculated based on the range of concentrations specified in the preceding column. See Table 2 for a summary of model parameters and morbidity ratios. d Per 100 000 individuals. e Per 1000 individuals. f Dose-response data is not available for transmission via a fecal-oral route.
RNA Kit (Qiagen) according to manufacturer’s directions. All nucleic acids were stored at -20 °C. A baseline detection limit for the PCR assay was established in RNase-free reagent-grade water for each pathogen. Serial dilutions were made with the nucleic acid stocks. The concentrations of organisms from original stock cultures were coupled with the volume used in the PCR reaction tube to estimate the minimum number of organisms necessary to produce a detectable band on the Bioanalyzer. The values obtained with the above procedure are accurate within an order of magnitude and represent a conservative estimate of the minimum detection limit. Inhibition of PCR. The relative impact of inhibitory substances on PCR was evaluated in the water samples collected from each location. An aliquot of the nucleic acid extract obtained from either the FALP filter or the 1-MDS filter was spiked with ca. 40 cells of Staphylococcus aureus prior to purification on the PVPP/Sephadex G200 spin column. The PCR detection limit was then determined using 10-fold serial dilutions of the purified samples. Inhibition of RT-PCR was evaluated using the GeneAmp RNA PCR Control Kit (Perkin-Elmer). An aliquot of each nucleic acid extract (obtained from the 1-MDS filter) was spiked with control RNA, purified on a Sephadex G200 spin column, and amplified using the appropriate RT-PCR protocol. For each assay the concentration of cells in the spiked sample (estimated to within an order of magnitude) was compared to the baseline sensitivity of the PCR assay to assess the relative impact of inhibitory substances on the minimum detection limit. Baseline Recovery from FALP and 1-MDS Filters. A baseline recovery was established using 10 L of dechlorinated tap water spiked with each of the appropriate stock cultures. The initial titer of each organism was estimated using an end-point serial dilution sequence with the appropriate PCRbased assay. The liquid was then sequentially filtered through the FALP and 1-MDS filters. Filters were eluted, and concentrations were estimated using the procedure outlined above. Percent recovery of each organism was established by comparing estimated concentrations before and after filtration. Recovery from FALP Filters Processed at Each Sample Location. Recovery of bacteria from the FALP filters was evaluated by comparing the concentration of fecal coliforms in the original sample with filter eluate. Original samples 2756
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 12, 2002
were processed using the multiple tube fermentation test (as per ref 15, method 9221E). Filter eluate was processed using an MPN-PCR assay (20). Recovery from 1-MDS Filters Processed at Each Sample Location. At each sample location, bovine enterovirus was spiked into the filtrate of the FALP filter. The 1-MDS filter was then processed as outlined above. Recovery of bovine enterovirus in each sample was established by comparing the concentration estimated from the filter eluate with the initial titer. Recovery of all the viruses were then corrected from the initial baseline values established in dechlorinated tap water to reflect the percent difference in recovery of the bovine enterovirus at each sample location.
Results and Discussion Health Risks Associated with Detectable Concentrations of Pathogens. Out of the 58 samples collected from 21 different storm drains, 22% (13 samples) produced a positive result for E. coli ETEC, Salmonella, or adenovirus. As with other studies that used a ( reporting scheme (e.g., ref 4), these results would indicate the presence of fecal pollution and the potential for health risks associated with recreational contact. We extended this finding, however, by comparing the range of possible pathogen concentrations from PCR positive samples with acceptable levels of risk promulgated by the U.S. EPA (Table 1). The risks of illness associated with the detectable concentrations of E. coli ETEC and Salmonella potentially exceeded the recommended regulatory values for marine (19:1000) and freshwaters (8:1000) in 9% and 55% of samples, respectively. The differential risks of illness associated with marine and freshwaters highlight potential difficulties when interpreting water quality data. Assuming the recommended regulatory values are appropriate, the detectable concentration of pathogens arguably poses a minimal risk to public health in reference to marine waters but a substantial risk in reference to freshwater. The discrepancy highlights the need to revisit the recommended risks of illness when the occurrence of specific pathogens is used to assess biological quality of water. Nevertheless, the risk-based methodology outlined above provides a framework for interpreting positive results of PCR-based assays that are used to assess the biological quality of recreational waters. Further refinement in quantitative PCR techniques will reduce the potential range of concentrations reported in this study.
TABLE 2. Detection Limits of PCR Assays Corresponding to Acceptable Health Risks Associated with Recreational Waters
organism
dose/response parameters (ref)a
E. coli O157:H7 E. coli ETEC Shigella Salmonella G. lamblia C. parvum Rotavirus
R ) 0.21; N50 ) 1120 (7) R ) 0.18; N50 ) 8.6 × 107 (7) R ) 0.21; N50 ) 1120 (7) R ) 0.31; N50 ) 2.36 × 104 (7) r ) 0.01982 (8) r ) 0.00419 (9) R ) 0.265; N50 ) 5.6 (10)
detection limit resulting in an acceptable risk of illnessc fresh marine
morbidity ratio (ref)b
1.6 8.8 × 104 1.9 380 0.89 6.1 0.015
0.22 (11) 0.5 (8) 0.39 (12) 0.57 (13)
4.4 2.1 × 105 4.3 900 2.2 13 0.034
percent of samples with an acceptable PCR detection limitd fresh marine 5-17 100 0 69-87 0 5-17 22-71
10-71 100 0 81-100 0 10-71 22-73
a R and N are parameters in the Beta-Poisson dose-response model; r is a parameter in the exponential dose-response model. b The dose50 response parameters for E. coli O157:H7, E. coli ETEC, and Shigella are based on illness as an endpoint; infection is used as the endpoint for the remaining organisms. c The U.S. EPA promulgates an acceptable risk of illness corresponding to 8:1000 and 19:1000 for primary recreational contact in fresh and marine waters, respectively (14). The acceptable risks of illness, assumed to be on an annual basis, were used in the following equation (see footnote (c) in Table 1 for definitions) to estimate the probability of illness based on a single exposure event (P): Pa ) 1 - (1-P)N. The parameter P was used in the appropriate dose-response model (e.g., eq 1 or 2) to estimate an acceptable dose (d). The dose was then adjusted to reflect a concentration using the following assumptions (as per 7): (i) 50 mL of water consumed during each hour of swimming activity and (ii) 2.6 h/swim. d Range reflects the terminal two dilutions producing a positive and negative result in the 10-fold serial dilution sequence.
Evaluation of Detection Limits Relative to Risk-Based Water Quality Standards. The concentration of pathogens was below the detection limit of the PCR-based assays in the majority (78%) of samples. The health risks associated with a negative result were evaluated in reference to an acceptable detection limit defined herein as the maximum concentration of a pathogen that results in an acceptable risk of illness (19:1000 and 8:1000 for marine and freshwaters, respectively). The percentage of PCR negative results having sufficient detection limits ranged from 0% for G. lamblia and Shigella to 100% for E. coli ETEC (Table 2). Although many of the samples resulted in an acceptable detection limit on an intermittent basis, the wide range in detection limits associated with a negative-PCR result precluded an inclusive statement about health risks. The above analysis provides a means to evaluate health risks associated with a negative-PCR result on a case-bycase basis. Nevertheless, the potential range of detection limits observed in this study indicates that existing PCRbased assays are not suitable for routinely monitoring microbial quality of water. Sensitivity can be affected by several factors including sample volume, preparation, and the PCR assay itself. Factors Influencing Detection Limits. Variables controlling overall detection limit included volume of sample processed through the filters (Vf), recovery of organisms from FALP and 1-MDS filters (Re), inhibition of PCR assays (I), and fraction of concentrated sample analyzed with PCR (%Vp). The relative influence of each factor on the overall detection limit can be expressed as
detection limit ) (I)(S) RI S I S ) ‚ (3) ) ‚ (Vf)(%Vp)(Re) (Re)(Vf) %Vp Vf %Vp where S is the baseline sensitivity of the PCR assay (see Table S-4 for values obtained in this study). Both S and Vp (the volume of concentrated sample used in the PCR reaction vial) were constants for a particular pathogen. Inhibition (I) was sample specific and represents the dilution necessary to produce a positive-PCR result. (I) was expressed as the inverse of the dilution factor and ranged from 1 to 1000 in this study. Recovery (R) represents the fraction of organisms in the filter eluate relative to the concentration in original unfiltered sample. The fraction of bacteria and protozoa recovered from the FALP filter ranged from 0.1 to 0.8, and the fraction of viruses recovered from the 1-MDS filter ranged from 0.001 to 0.56. The volume of water processed through the FALP and 1-MDS filters (Vf) ranged from 20 to 50 L.
TABLE 3. Threshold Values of Sample Volume (Vf) and RI Necessary To Achieve an Acceptable Detection Limit in Fresh and Marine Waters minimum volumea,b (L)
maximum RIa,c
organism
fresh
marine
fresh
marine
E. coli O157:H7 E. coli ETEC Shigella Salmonella G. lamblia C. parvum Rotavirus
28.7 5.23 × 10-3 484 4.24 × 10-1 129 18.9 3.93
10.5 2.19 × 10-3 214 1.79 × 10-1 52.3 8.85 1.72
1000
