Relationship and Variation of qPCR and Culturable Enterococci

Lake Michigan Ecological Research Station, Great Lakes Science Center, .... LA; Greenbriar State Park, MD; Fall Lake, Shagawa Lake, Trout Lake, Pike L...
0 downloads 0 Views 451KB Size
Download PDF
Environ. Sci. Technol. 2010, 44, 5049–5054

Relationship and Variation of qPCR and Culturable Enterococci Estimates in Ambient Surface Waters Are Predictable R I C H A R D L . W H I T M A N , * ,† Z H O N G F U G E , † MEREDITH B. NEVERS,† ALEXANDRIA B. BOEHM,‡ EUNICE C. CHERN,§ RICHARD A. HAUGLAND,§ ASHLEY M. LUKASIK,† MARIROSA MOLINA,| KASIA PRZYBYLA-KELLY,† DAWN A. SHIVELY,† EMILY M. WHITE,| RICHARD G. ZEPP,| AND MURULEEDHARA N. BYAPPANAHALLI† Lake Michigan Ecological Research Station, Great Lakes Science Center, United States Geological Survey, 1100 North Mineral Springs Road, Porter, Indiana 46304, Department of Civil and Environmental Engineering, Stanford University, Stanford, California 94305, Microbiological and Chemical Exposure Assessment Research Division, National Exposure Research Laboratory, U.S. Environmental Protection Agency, 26 West Martin Luther King Drive, Cincinnati, Ohio 45268, and Ecosystems Research Division, National Exposure Research Laboratory, U.S. Environmental Protection Agency, 960 College Station Road, Athens, Georgia 30605

Received September 23, 2009. Revised manuscript received May 5, 2010. Accepted May 13, 2010.

The quantitative polymerase chain reaction (qPCR) method provides rapid estimates of fecal indicator bacteria densities that have been indicated to be useful in the assessment of water quality. Primarily because this method provides faster results than standard culture-based methods, the U.S. Environmental Protection Agency is currently considering its use as a basis for revised ambient water quality criteria. In anticipation of this possibility, we sought to examine the relationship between qPCRbased and culture-based estimates of enterococci in surface waters. Using data from several research groups, we compared enterococci estimates by the two methods in water samples collected from 37 sites across the United States. A consistent linear pattern in the relationship between cell equivalents (CCE), based on the qPCR method, and colony-forming units (CFU), based on the traditional culturable method, was significant (P < 0.05) at most sites. A linearly decreasing variance of CCE with increasing CFU levels was significant (P < 0.05) or evident for all sites. Both marine and freshwater sites under continuous influence of point-source contamination tended to reveal a relatively constant proportion of CCE to CFU. The consistency in the mean and variance patterns of CCE versus * Corresponding author e-mail: [email protected]; phone: (219) 926-8336 (Ext. 424); fax: (219) 929-5792. † United States Geological Survey. ‡ Stanford University. § Microbiological and Chemical Exposure Assessment Research Division, U.S. EPA. | Ecosystems Research Division, U.S. EPA. 10.1021/es9028974

 2010 American Chemical Society

Published on Web 06/08/2010

CFU indicates that the relationship of results based on these two methods is more predictable at high CFU levels (e.g., log10CFU > 2.0/100 mL) while uncertainty increases at lower CFU values. It was further noted that the relative error in replicated qPCR estimates was generally higher than that in replicated culture counts even at relatively high target levels, suggesting a greater need for replicated analyses in the qPCR method to reduce relative error. Further studies evaluating the relationship between culture and qPCR should take into account analytical uncertainty as well as potential differences in results of these methods that may arise from sample variability, different sources of pollution, and environmental factors.

Introduction Most ambient water quality criteria around the world are based on densities of indicator bacteria E. coli and enterococci that are estimated by standard culture-based methods. The standard methods require 24 h to complete because they rely on cultivation and enumeration of colony forming units. Indicator bacteria densities are variable over time scales that are shorter than 24 h owing, for example, to variations in mixing of ambient waters, bacterial photoinactivation, and unsteady sources (1-3). This is problematic because during the 24-h lag between water sample collection and the receipt of test results, water quality is likely to have changed. The 24-h analytical delay has been shown to give rise to misrepresentation of true water quality for recreational use resulting in inaccurate management decisions (4-6). Two approaches for addressing the time lag problem include development of (1) predictive models to nowcast water quality and (2) rapid detection technologies. With regard to the latter, quantitative polymerase chain reaction (qPCR) is the technology that has received the most attention from researchers and regulators. Technical details and procedures of the application of qPCR in estimating enterococci densities in beach water samples, as well as a comparison with a standard culture method, were presented in Haugland et al. (7). The U.S. Environmental Protection Agency (USEPA) further incorporated qPCR assays for enterococci into several epidemiological studies at beaches where the measurement has been well correlated to swimmer health effects (8). In the United States, epidemiological studies linking indicator densities to health effects represent the cornerstone of ambient water quality criteria (9), thus, the USEPA is considering the qPCR technique as one of its options in plans to promulgate new criteria by 2012. Although it is well-accepted that qPCR measurements can be obtained more quickly than culture-based measurements, overcoming the described time lag problem, and researchers have demonstrated the utility of the method in ambient waters (7, 8, 10), the comparability and possible relationships between qPCR and culture-based measures of enterococi have not been adequately described. Aside from a small number of studies (7, 11-13) where mean and variance of the two end points and their correlation were discussed, few researchers have statistically related these two end points to one another or explored what factors might be affecting the relationship. The statistical relationship between the two measures may be important; for example, if qPCR is used for real-time beach monitoring and culture methods are used for discharge permitting, 303-d listings, total maximum daily load calculations, and water quality standard violations. Additionally, linking the old and new measures may allow merging of historical with new water quality data sets. VOL. 44, NO. 13, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

5049

TABLE 1. Sampling Sites and Projects Included in the Present Study location

sampling dates

sample size

culturable enterococci mean/range (log10 CFU/100 mL)

water type

Indiana West Beach Salt Creek Chesterton wastewater effluent

5/31/2003 - 8/03/2003 3/19/2009 3/19/2009

320 84a 100a

1.03/0-3.57 1.38/0.13-2.83 1.91/0.70-3.06

freshwater freshwater (river) sewage

Ohio Huntington Beach

7/27/2003 - 9/14/2003

393

1.52/0-4.68

freshwater

California Avalon Beachb

8/19/2008 - 8/22/2008

72

1.81/0.30-2.93

marine water

various states additional 32 sitesc

2006

1d

1.63/0-3.99

freshwater

a

The sample size was generated by diluting the original water sample, replicating, and reanalyzing generated samples over time. b 72-h continuous hourly samples. See Boehm et al. (2) for details. c Salt River (two locations), AZ; Santa Ana River, Lake Elsinore, CA; Lake Sidney-Lanier - Burton Mill, Lake Sidney-Lanier - Buford Dam, GA; Manoa Stream, HI; Backbone State Park, Coralville Lake, IA; Lake Michigan - Dunes Creek (two locations), Lake Michigan - West Beach, Brookville Lake, IN; Pontchartrain - Bayou Lacombe, LA; Greenbriar State Park, MD; Fall Lake, Shagawa Lake, Trout Lake, Pike Lake, MN; Harry Wright Lake, Manahawkin Lake, Lake Carasaljo, Pine Lake, NJ; Chautauqua Lake, NY; Neuse River Flanners Beach, Neuse River - Storm Water, NC; Lake Waco, TX; Lake Washington - Thornton Creek, WA; Lake Winnebago, Mendota Lake, Lake Wingra, Lake Michigan - South Shore Beach, WI. d One pair of CCE and CFU estimates were obtained for each site.

In this paper, we compare estimates obtained by qPCR and a standard culture-based method for the densities of enterococci, a representative fecal indicator bacterial group routinely used for both fresh and marine water quality worldwide, in ambient waters from 36 different locations across the United States and from one sewage effluent. Rather than reassess the technical foundation of the methods, we are more concerned with the statistical behavior of the qPCRbased results compared to those by the culture-based method at a wide variety of geographical locations. Our results can supplement previous studies, such as Lavender and Kinzelman (12), with implications that may be relevant to beach monitoring and recreational water quality criteria development.

Materials and Methods Water samples were collected from West Beach, Indiana; Huntington Beach, Ohio; Avalon Beach, California; Salt Creek, Indiana; and a wastewater final effluent in Chesterton, Indiana. A mixed data set containing observations from 32 other sites in 15 states of the United States was also examined. Detailed characteristics of the sampling projects are given in Table 1. Huntington Beach and West Beach. The sites, sampling, and analysis procedures have been detailed in Haugland et al. (7). Briefly, water samples were collected on Saturdays, Sundays, and holidays from May 31 through August 3, 2003 at West Beach and from July 27 through September 14, 2003 at Huntington Beach. At both beaches, samples were collected at waist-level (1 m deep) and at shin-level (0.3 m deep) along multiple transects at 8 a.m., 11 a.m., and 3 p.m. each day. In the context of comparing the culturable and qPCR methods, all samples are treated as independent ones, while in Haugland et al. (7) the geometric mean of the results per visit was reported for their particular purpose of characterizing beach water quality and health risks. Samples with no detectable culturable or target DNA densities, which were assigned artificial values for calculating geometric means and the associated standard deviations (7), are excluded in the present work. Viable enterococci were enumerated by U.S.EPA Method 1600 on mEI agar plates (14); qPCR analyses were performed using fluorogenic 5′ nuclease (TaqMan) system. Sample processing, standard curve generation with different DNA calibrators, and sensitivity of qPCR assays have also been discussed in detail (7). 5050

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 13, 2010

Salt Creek (Surface Water) and Chesterton Wastewater Effluent. To assess the relationship of the two analytical methods over a wide range of indicator bacteria densities, we filtered different volumes of samples from two mesocosm experiments over time for analysis by both qPCR and culture as detailed below. Approximately 6 L of water were collected from Salt Creek (surface water) (41°28.137′ N, 87°4.779′ W) and Chesterton wastewater effluent (41°37.153′ N, 87°3.830′ W) on March 19, 2009. The Salt Creek sampling site was downstream of the outfall of the treatment plant. The Chesterton Wastewater Treatment Plant uses an anaerobic digestion-activated sludge treatment followed by UV disinfection. Each of these original water samples was homogenized in the laboratory for use in the mesocosms. Samples from each mesocosm were filtered for qPCR and culture analyses as described in the respective methods (7, 14) at four time points over a 3-week period, referred to as week 0, the sample collection time; week 1, the end of the first week; through week 3, the end of the 3-week period. Due to the decrease in culturable counts over the course of the experiment, the number of different sample volumes from each mesocosm that were filtered for this method decreased in the later weeks of the study: three volumes in weeks 0 and 1 (4, 11, 33 mL for Salt Creek; 1, 4, 11 mL for the effluent), two volumes in week 2 (11, 33 mL for Salt Creek; 4, 11 mL for the effluent), and one volume in week 3 (100 mL for Salt Creek; 25 mL for the effluent). Three replicate samples of each of the indicated volumes were filtered for culture analysis. For qPCR analysis, five replicate samples of five volumes (1, 4, 11, 33, 100 mL) were produced for weeks 0, 1, and 2; and two volumes (33 and 100 mL) of Salt Creek water for week 3 were filtered. The culture counts for volumes that did not correspond to some of the volumes filtered for qPCR analysis were extrapolated from the mean count values and relative volumes of the samples that were filtered for culture analysis for comparison of results. Mesocosms were stored in the dark at 4 °C between weekly sampling. Samples were analyzed for culturable enterococci by the membrane filtration method 1600 (14). Enterococci qPCR assays were performed per Haugland et al. (7). Samples were filtered through 47-mm, 0.4-µm-pore-size polycarbonate filters (GE Osmonics Labstore, Minnetonka, MN); filters were transferred to 2-mL microcentrifuge tubes containing 0.3-g glass beads. Genomic DNA extractions were performed using 600 µL of AE Buffer (Qiagen, Valencia, CA), containing 0.2

µg/mL salmon testes DNA (Sigma-Aldrich, St. Louis, MO). Enterococci were quantified using an adaptation of the comparative cycle threshold method developed previously by Applied Biosystems (15). Avalon Beach. Avalon Beach is a marine site located on Santa Catalina Island in the Southern California Bight. Water quality at this location appears to be strongly influenced by leaking sewage lines located within 10 miles of the land-sea interface (2). Water samples were collected at ankle depth hourly from 0400 h August 19 to 0300 h August 22, 2008 10 meters north of the Pleasure Pier (33°20.9′ N, 118°19.5′ W) and were stored in the dark in a cooler. Samples were processed within 6 h of collection. Enterococci were analyzed using membrane filtration with mEI media (U.S.EPA method 1600) (14). Total enterococci were quantified in 100 mL of sample water using the Taqman qPCR assays targeting 23S genes (2, 7). Results were corrected for inhibition and extraction efficiency as described by Yamahara et al. (16). Whole genomic DNA from E. faecium (ATCC 19434) was used to generate standard curves for qPCR assays. More details of the sampling project and methods can be found elsewhere (2, 16). 32 Additional Freshwater Sites. Water samples were collected by collaborators from the sampling locations listed in Table 1. Two replicate subsamples from each location were filtered on site for analysis by U.S.EPA method 1600 within 6 h of collection and an additional subsample was shipped on ice to the U.S.EPA laboratory for next morning delivery. Shipped samples were filtered and processed for the qPCR analysis in duplicate according to Haugland et al. (7). Replicated qPCR analyses of each sample were performed with either TaqMan probes or Scorpion probes (DxS Limited, Manchester, UK) and using three different instrument and reagent systems as described in the Supporting Information. Other qPCR analytical details, including the determination of amplification efficiencies for each assay calibration standards and the cycle threshold (CT) can be found elsewhere (7, 17). No significant differences were observed between overall enterococci measurements by the different probe and instrument systems. Therefore results from all systems were treated as replicate analyses (N ) 8). The geometric mean of the eight replicates for each location was reported as the qPCR estimate. Individual analyses showing no detectable enterococcus DNA were assigned a log10 value of 0. Data Processing. Culturable and qPCR results are reported as colony-forming units (CFU) and calibrator cell equivalents (CCE), respectively, per 100 mL of water. At all sites except Salt Creek and the Chesterton wastewater effluent (for selected cases as described previously), the measurement of CFU densities was not replicated so no estimate of within sample variance was available. Throughout this study, we mostly focused on the within sample variance in the qPCR results. These samples are from diverse locations that are possibly affected by different environmental conditions. The only assessment of the measurement errors was attempted in the analyses of the mesocosm samples from Salt Creek and the Chesterton wastewater effluent, where replicate samples were collected for both methods as described above. Statistical analyses were performed using SPSS 12.0 and SAS 9.0. CCE and CFU values were log10-transformed to meet the assumption of normality although still simply referred to as CCE and CFU, respectively, in the following sections. Unless otherwise stated, statistical significance was set at R ) 0.05.

Results Mean and Variance Patterns between CCE and CFU. Comparisons of CCE and their corresponding CFU numbers are given in Figure 1 for data collected from Huntington Beach and West Beach. The mean trend between the logarithmic CCE and CFU counts is evidently linear for

FIGURE 1. Logarithmic CCE versus CFU at (a) Huntington Beach and (b) West Beach. Solid line: linear mean trend between CCE and CFU; dotted line: 95% confidence interval of CCE; dashed line: reference for CCE ) CFU. Huntington Beach with a moderate R2 (coefficient of determination) of 0.47. The variance of CCE relative to the mean trend is estimated in each of a series of CFU bins that correspond to various CFU ranges, such as from 0 to 0.5 and from 0.5 to 1.0, and is hypothesized to have a linear relation with the central CFU value of the bin (bins were combined when there would be otherwise too few data points in each). For Huntington Beach, the linear decrease of the variance of CCE with increasing CFU was found significant by ANOVA (F ) 11.4, P ) 0.01), with a coefficient of determination (denoted as r2 hereinafter to avoid confusion with R2) of 0.62. It should be noted that a high r2 does not exclude the case of constant variance of CCE with CFU. The ANOVA is used to test whether the variance of CCE is significantly related to CFU counts. The 95% confidence interval is obtained at approximately 1.96 times the local standard error, assuming normality. Note that the uncertainty in the CCE value is generally large. For example, the logarithmic CCE range given by the 95% confidence interval for Huntington Beach at CFU ) 2 is approximately from 1.4 to 3.4, corresponding to an original CCE count from 25 to 2512 CCE/100 mL. The data range for CCE is wider for lower CFU values. For West Beach, the strength of the relationship between CCE and CFU counts is poor (R2 ) 0.12). The variance of CCE does not decrease significantly with increasing CFU (F ) 0.61, P ) 0.47). The weaker linearity in both mean and variance trends might have resulted from the much fewer high CFU occurrences compared to Huntington Beach. The noise could have been from either the CCE or CFU measurements. However, it still is evident that a larger mean and a smaller variance of CCE are consistent with increasing CFU counts at West Beach. Controlled Experiment. CCE (five replicates for each filtered volume) versus CFU (only the mean of the triplicates or the extrapolated means are taken) counts for the water samples from Salt Creek and the Chesterton wastewater effluent are shown in Figure 2. The collective CCE data for different dilutions and weeks at both sites form consistent patterns with one another and with the two beaches shown in Figure 1 with mean and variance of CCE estimates being linearly and inversely related to those of CFU counts. The mean trend of CCE with CFU yielded R2 ) 0.83 for Salt Creek and R2 ) 0.59 for the effluent. The variance of CCE relative to CFU was significantly linear according to ANOVA (r2 ) 0.80, F ) 15.3, P ) 0.02 for Salt Creek and r2 ) 0.71, F ) 9.2, P ) 0.03 for the effluent). VOL. 44, NO. 13, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

5051

FIGURE 2. Logarithmic CCE versus CFU at (a) Salt Creek, (b) Chesterton wastewater effluent, and (c) Avalon Beach. Solid line: linear mean trend between CCE and CFU; dotted line: 95% confidence interval of CCE; dashed line: reference for CCE ) CFU. In (a) and (b): circle: week 0; square: week 1; triangle: week 2; *: week 3. Consistency of Patterns. Figures 1 and 2 have shown that the mean and variance trends of CCE with increasing CFU are similar regardless of the site. The two sets of 95% confidence intervals established in Figure 1 are also shown in Figure 3. Despite the variety in the geographical location, all but three (or two) data points from the 32 sites fall within the 95% confidence interval for Huntington Beach (or West Beach). This is consistent with the hypothesis of a similar pattern between CCE and CFU among all of these sites. The data pattern at Avalon Beach is shown in Figure 2c. The mean trend of CCE versus CFU levels at Avalon Beach is also best described by a linear model with an R2 of 0.55. The variance distribution with CFU is also linear (r2 ) 0.64). This 3-day, hourly time series data set does not encompass as wide a range of CFU as, e.g., those at Huntington Beach or West Beach. This rendered the effect of the CFU value on the CCE variance insignificant (F ) 3.55, P ) 0.20). The decreasing trend of CCE variance with increasing CFU, however, is still evident when CFU is larger than 2 (log10CFU/ 100 mL). There is high similarity between the mean trend of CCE relative to the CCE ) CFU line at Avalon Beach to those at Salt Creek and Chesterton wastewater effluent. The mean trend of all these cases does not cross the CCE ) CFU line within their respective data ranges as it does for Huntington Beach and West Beach. This characteristic is likely due to the fact that Avalon Beach is strongly affected by sewage input (2), which provides a relatively persistent source of both viable and nonviable enterococci cells as at Salt Creek (surface water) and Chesterton sewage effluent. Therefore, the mean and variance patterns between CCE and CFU at the marine beach are consistent with those at freshwater sites. Measurement Error: Controlled Experiment. It has been shown at various sites that the variance of CCE increases evidently or significantly when CFU values become small, which could signify undesired characteristics such as the loss of accuracy, precision, or sensitivity of either of the methods. To help compare the measurement errors of the qPCR and culturable methods for water samples over a wide range of enterococci levels, we extracted DNA from five replicate aliquots for CCE and plated triplicate aliquots for culturable enterococci for each serially diluted water sample from Salt Creek and from the Chesterton wastewater effluent over three consecutive weeks. While the sample numbers are not large enough to infer any statistical significance, the 5052

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 13, 2010

FIGURE 3. Logarithmic CCE versus CFU at 32 additional sites in 15 states of the U.S. compared to 95% confidence intervals for Huntington Beach and West Beach (Figure 1).

FIGURE 4. Relative errors in the measurement of CCE and CFU for selected filtration volumes in three consecutive weeks at (a) Salt Creek and at (b) Chesterton wastewater effluent. relative error in the measurement for each filtered volume, defined as the standard deviation divided by the mean of five replicates (for CCE) or triplicate (for CFU), is considerably higher for the qPCR method except when the mean log10 CFU level drops to approximately 1 at the Chesterton wastewater effluent or below 1.5 at Salt Creek (Figure 4). In much of the data range (e.g., log10 CFU from 1.5 to 2.4 at Salt Creek), therefore, the qPCR method yielded a higher relative error or larger statistical uncertainty. The uncertainty due to the qPCR approach was large even at high mean CFU levels. The relative error in the CCE measurement at Salt Creek, for example, was 5-10 times higher than that in the CFU measurement when the mean CFU was above 1.8.

Discussion Characterization of capabilities and limitations of testing procedures is an important component of defining the application and use of a proposed analytical method. A thorough understanding of a new analytical method should include its technical properties (biases, upper/lower detection limits, and calibration curves) and comparisons to results based on standard or widely accepted methods (18). QPCR detects fundamentally different targets (DNA sequences) than culture methods and the CCE reporting unit should not be construed as necessarily providing accurate estimates of cell densities in different environmental samples (19). In addition,

the qPCR method may detect DNA from substantially different cell populations (both live and dead as well as a potentially wider spectrum of different species) than culture methods. Therefore, while it is both of interest and appropriate to compare the results of the qPCR and culture methods to determine the patterns and strengths of the relationships between their results in different waters and conditions, which is the focus of the present study, the differences between these two methods may create challenges in demonstrating equivalent empirical results on a consistent basis. In this paper we studied the variation of the qPCR enterococci signal in different water samples as a function of the densities of culturable enterococci. In addition to a positive trend between mean CCE and CFU estimates, we noted that there was less variability in the qPCR estimates among samples with higher CFU densities (typically log10CFU above 2.0/100 mL). The decrease in variability in the relationship between results of the two methods with increasing CFU values was further found to be linear for all cases but West Beach. These relationships are based on statistical analyses of data sets that cover a variety of water types, ambient conditions, and geographical locations in the United States. This consistent pattern between CCE and CFU suggests an overall analytical commonality, independent of the source waters tested, that is related to the densities of culturable enterococci that are present. It is important to note, however, that the consistency of the patterns shown here may result from smoothing of individual environmental, sampling, and analytical variation. At Huntington Beach and West Beach (Figure 1), the data points associated with the largest CFU values appear to converge onto the reference line for CCE ) CFU, although the fitted linear trend of the mean CCE deviates from this asymptote due to extrapolation at this end of data. For the Salt Creek and Chesterton wastewater effluent data sets, the fitted mean CCE line is approximately parallel to the CCE ) CFU line, suggesting a nearly constant proportion of CCE to CFU. This relatively constant relationship, which is also strongly suggested at Avalon Beach, might be intimately related to the influence of nearby point sources and hence site-specific because this characteristic is not appreciable at Huntington Beach or West Beach which may be more influenced by changing amounts of effluent. Controlled mesocosm experiments over time showed a more stable pattern of CCE variance with respect to CFU densities for sewage water (Chesterton wastewater effluent) than was observed for ambient water (Salt Creek), which was particularly evident from week 2 to week 3. This difference may be primarily associated with differential effects of the two water types and environmental conditions on the persistence of viable cells. The overall similarity of the CCE and CFU relationship patterns seen in these experiments with those of the other data sets suggest that the relationship is reasonably predictable over time but becomes more uncertain as CFU values decrease to a very low level (e.g., < 1 log10CFU/100 mL). The similarity of the data patterns, in both freshwater and at the marine beach, suggests a mathematical expression for CCE: CCE ) R + βCFU + ε(CFU)

(1)

where CCE is a random variable, a function of CFU which is here treated as a deterministic variable. ε is the stochastic part of CCE with a zero mean and a linearly decreasing variance (or standard deviation) with increasing CFU. R is always a non-negative number because CCE is no less than CFU in any water sample. The coefficient β is typically no larger than 1 so that it intersects (Figures 1, 2a, and 3) or is

nearly parallel to the CCE ) CFU line (Figure 2b) at high CFU levels. When β is close to 1, the mean trend of CCE, R + βCFU, does not intersect the CCE ) CFU line, and a large proportion of R reflects the persistent background level of CCE. This characteristic might be due to the influence of a relatively strong point source near the site (e.g., Salt Creek, Chesterton wastewater effluent, and Avalon Beach). Therefore, it should be noted that, although the pattern as described by eq 1 generally holds, the parameters in eq 1 and hence the particular form of the relationship between CCE and CFU can be influenced by environmental factors. For instance, different sources of fecal bacteria and proximity of these fecal sources to the sampling location will influence the fluidity of the particular form of the relationship. The direct comparison of measurement errors in replicated qPCR and culturable results for selected filtered volumes over three consecutive weeks (Figure 4) in this study indicated that the relative errors in the qPCR results were higher than the culturable results by a factor up to 10, except when the mean culturable enterococci densities were very low (e.g., log10CFU < 1/100 mL for Chesterton wastewater effluent or < 1.5/100 mL for Salt Creek). These results suggest that the uncertainty (or noise) in the culturable measurements in Figures 1 and 2 may be dominated by that in the qPCR measurements in the range of CFU higher than 1.2. This assumption partially supports the use of nonreplicated CFU numbers while the CCE estimate was examined as a random variable in Figures 1-3 and suggests a potentially greater need for averaging of replicated analyses in the qPCR method to reduce relative error. When the mean log10CFU was lower than 1.5, both methods yielded comparable relative errors, at a level from 0.05 to 0.15. Only in extreme cases where the mean log10CFU was lower than 0.9 (Figure 4a) did the error in the CFU measurement exceed that in CCE. It also should be noted, however, that this comparison was limited by the small sample sizes considered and hence simply suggestive. Further studies are needed to determine the relative contributions of analytical uncertainty, distributions of targeted cell populations, environmental conditions, and variation in the attributes of different fecal sources on the relationships between fecal indicator bacteria density estimates by qPCR and culture methods. Nevertheless, these results suggest that investigators can indeed develop empirical relationships between CCE and CFU, especially at higher densities of CFU, but care should be exercised when extrapolating CCE signals, particularly at low CFU densities.

Acknowledgments This article is Contribution 1591 of the USGS Great Lakes Science Center. We are grateful to Kevin Oshima and Larry Wymer of U.S.EPA for their comments that greatly improved our manuscript. This paper has been reviewed in accordance with the U.S. Environmental Protection Agency’s peer and administrative review policies and approved for publication. Mention of trade names or commercial products does not constitute an endorsement or recommendation for use by the U.S.EPA.

Supporting Information Available Details of the sampling and qPCR analytical methods for the 32 additional freshwater sites. This material is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited (1) Boehm, A. B.; Grant, S. B.; Kim, J. H.; Mowbray, S. L.; McGee, C. D.; Clark, C. D.; Foley, D. M.; Wellman, D. E. Decadal and shorter period variability of surf zone water quality at Huntington Beach, California. Environ. Sci. Technol. 2002, 36 (18), 3885– 3892. VOL. 44, NO. 13, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

5053

(2) Boehm, A. B.; Yamahara, K. M.; Love, D. C.; Peterson, B. M.; McNeill, K.; Nelson, K. L. Covariation and photoinactivation of traditional, rapid, and novel indicator organisms and human viruses at a sewage-impacted marine beach. Environ. Sci. Technol. 2009, 43 (21), 8046–8052. (3) Whitman, R. L.; Nevers, M. B.; Korinek, G. C.; Byappanahalli, M. N. Solar and temporal effects on Escherichia coli concentration at a Great Lakes swimming beach. Appl. Environ. Microbiol. 2004, 70 (7), 4276–4285. (4) Frick, W. A.; Ge, Z.; Zepp, R. G. Nowcasting and forecasting concentrations of biological contaminants at beaches: A feasibility and case study. Environ. Sci. Technol. 2008, 42, 4218– 4824. (5) Hou, D.; Rabinovici, S. J. M.; Boehm, A. B. Enterococci predictions from partial least squares regression models in conjunction with a single-sample standard improve the efficacy of beach management advisories. Environ. Sci. Technol. 2006, 40 (6), 1737–1743. (6) Kim, J. H.; Grant, S. B. Public mis-notification of coastal water quality: A probabilistic evaluation of posting errors at Huntington Beach, California. Environ. Sci. Technol. 2004, 38 (9), 2497–2504. (7) Haugland, R. A.; Siefring, S. C.; Wymer, L. J.; Brenner, K. P.; Dufour, A. P. Comparison of Enterococcus measurements in freshwater at two recreational beaches by quantitative polymerase chain reaction and membrane filter culture analysis. Water Res. 2005, 39 (4), 559–568. (8) Wade, T. J.; Calderon, R. L.; Sams, E.; Beach, M.; Brenner, K. P.; Williams, A. H.; Dufour, A. P. Rapidly measured indicators of recreational water quality are predictive of swimming associated gastrointestinal illness. Environ. Health Perspect. 2006, 114 (1), 24–28. (9) Boehm, A. B.; Ashbolt, N. J.; Colford, J. M., Jr.; Dunbar, L. E.; Fleming, L. E.; Gold, M. A.; Hansel, J. A.; Hunter, P. R.; Ichida, A. M.; McGee, C. D.; Soller, J. A.; Weisberg, S. B. A sea change ahead for recreational water quality criteria. J. Water Health 2009, 7 (1), 9–20. (10) Wade, T. J.; Calderon, R. L.; Brenner, K. P.; Sams, E.; Beach, M.; Haugland, R.; Wymer, L.; Dufour, A. P. High sensitivity of children to swimming-associated gastrointestinal illness: Results using

5054

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 13, 2010

(11)

(12)

(13)

(14)

(15)

(16)

(17)

(18)

(19)

a rapid assay of recreational water quality. Epidemiology 2008, 19 (3), 375–383. He, J. W.; Jiang, S. Quantification of enterococci and human adenoviruses in environmental samples by real-time PCR. Appl. Environ. Microbiol. 2005, 71 (5), 2250–2255. Lavender, J. S.; Kinzelman, J. L. A cross comparison of QPCR to agar-based or defined substrate test methods for the determination of Escherichia coli and enterococci in municipal water quality monitoring programs. Water Res. 2009, 43 (19), 4967–4979. Morrison, C. R.; Bachoon, D. S.; Gates, K. W. Quantification of enterococci and bifidobacteria in Georgia estuaries using conventional and molecular methods. Water Res. 2008, 42 (14), 4001–4009. U.S. EPA Method 1600--Enterococci in water by membrane filtration using membrane-Enterococcus Indoxyl-B-D-Glucoside agar (mEI); EOA-821-R-02-022; U.S. Environmental Protection Agency: Washington, DC, 2002. Applied Biosystems. User Bulletin #2 ABI PRISU 7700 Sequence Detection System; Applied Biosystems Corporation: Foster City, CA, 1997. Yamahara, K. M.; Walters, S. P.; Boehm, A. B. Growth of enterococci in unaltered, unseeded beach sands subjected to tidal wetting. Appl. Environ. Microbiol. 2009, 75 (6), 15171524. Haugland, R. A.; Siefring, S.; Varma, M.; Atikovic, E.; Brenner, K. P.; Dufour, A. P. Evaluation of matrix effects from nationwide inland surface waters on quantitation polymerase chain reaction analysis for fecal indicator bacteria; National Beaches Conference, Niagara Falls, NY, 11-13 October, 2006. American Public Health Association. Standard Methods for the Examination of Water and Wastewater, 21st ed.; American Public Health Association: Washington, DC, 2005. Chern, E. C.; Brenner, K. P.; Wymer, L.; Haugland, R. A. Comparison of fecal indicator bacteria densities in marine recreational waters by QPCR. Water Qual. Expo. Health 2009, 1, 203–214.

ES9028974