Factors affecting the relationship between quantitative polymerase chain reaction (qPCR) and culture-based enumeration of Enterococcus in environmental waters




To determine the extent to which discrepancies between qPCR and culture-based results in beach water quality monitoring can be attributed to: (i) within-method variability, (ii) between-method difference within each method class (qPCR or culture) and (iii) between-class difference.

Methods and Results

We analysed 306 samples using two culture-based (EPA1600 and Enterolert) and two qPCR (Taqman and Scorpion) methods, each in duplicate. Both qPCR methods correlated with EPA1600, but regression analyses indicated approximately 0·8 log10 unit overestimation by qPCR compared to culture methods. Differences between methods within a class were less than half of this and were minimal for between-replicate within a method. Using the 104 Enterococcus per 100 ml management decision threshold, Taqman qPCR indicated the same decisions as EPA1600 for 87% of the samples, but indicated beach posting for unhealthful water when EPA1600 did not for 12% of the samples. After accounting for within-method and within-class variability, 8% of the samples exhibited true between-class discrepancy where both qPCR methods indicated beach posting while both culture methods did not.


Measurement target difference (DNA vs growth) accounted for the majority of the qPCR-vs-culture discrepancy, but its influence on monitoring application is outweighed by frequent incorrect posting with culture methods due to incubation time delay.

Significance and Impact of the Study

This is the first study to quantify the frequency with which culture-vs-qPCR discrepancies can be attributed to target difference - vs - method variability.


Beach water quality monitoring is presently conducted using culture-based methods that have achieved widespread usage because they are cost effective, easily implemented, and correlate well with health risk (Wade et al. 2003). However, sample processing for culture-based methods requires 18–96 h, which means that health warnings are issued days after the samples are originally collected, inconsistent with beach contamination that is often of short duration (Leecaster and Weisberg 2001). EPA has recently allowed (U.S. EPA 2012) quantitative polymerase chain reaction (qPCR) as a new measurement method that reduces processing time to as little as 2 h by directly measuring genetic material and eliminating the time-consuming incubation step (Noble and Weisberg 2005). qPCR-based measurements have been found to relate well to health risk (Wade et al. 2006, 2008, 2010; Colford et al. 2012). Initial beach applications of qPCR have demonstrated success in technology transfer to monitoring agencies and in implementing health protection decisions the same day samples were collected (Griffith and Weisberg 2011; Ferretti et al. 2013).

Several studies have found that qPCR and the culture-based methods they are intended to replace correlate well, although qPCR often produces higher values (Haugland et al. 2005; He and Jiang 2005; Morrison et al. 2008; Lavender and Kinzelman 2009; Abdelzaher et al. 2010; Whitman et al. 2010). This has been largely attributed to differences in target, as qPCR measures an intracellular genetic target rather than the cells' ability to grow on selective media (Noble et al. 2010). However, there are many other factors that can lead to such differences, including method variability and calibration biases that affect converting gene copy numbers to cell equivalents (Cao et al. 2013; Shanks et al. 2012). Most of the comparison studies have been conducted processing a single replicate from a single culture-based and a single qPCR method, making it difficult to differentiate potential contributing sources to the discrepancy between qPCR and culture results.

Here, we extend those previous method comparison studies by simultaneously processing two qPCR and two culture-based methods, each in duplicate, for a series of marine beach samples. This design allows us to examine results discrepancy (and/or agreement) between two classes (qPCR and culture-based) of Enterococcus enumeration methods, between two methods within each class and between two replicates within each method. Such multiple levels of discrepancies can then be contrasted to reveal contributions from several potential causes of qPCR vs culture results discrepancy: (i) within-method variability, (ii) difference in enumeration approaches within each class (culture or qPCR) of methods and (iii) between-class target difference (i.e. growth vs genetic endpoints for culture and qPCR, respectively).

Materials and methods

Sample and laboratory analysis

Nine Orange County, CA, sites were sampled daily from 6 July 2010 to 31 August 2010. Eight of the sites were open coast beaches, while the ninth was a high-salinity embayment site located in Newport Bay (Site BNB24E, Fig. 1). Samples were collected between 7 and 9 am at c. 0·5 m depth on incoming water in acid-washed 1·5 l polypropylene bottles. Water was then filtered through polycarbonate 47-mm diameter, 0·4-μm pore-size filters (100 ml per filter) for qPCR analysis. Water from the same sample was also filtered through 47-mm mixed cellulose HA filters (10 and 1 ml volumes filtered) or used directly for culture-based analysis. The Orange County Sanitation District, the South Orange County Wastewater Authority and the Orange County Health Department were responsible for collecting samples from three sites each and performing the culture methods. The Southern California Coastal Water Research Project Authority performed the qPCR methods.

Figure 1.

Map of the nine sampling sites in Orange County, California, USA. Eight of the sites were open coast beaches, while the ninth (BNB24E) was a high-salinity embayment site.

Each sample was analysed for Enterococcus spp. in duplicate for each of two qPCR methods and each of two culture-based methods. The culture methods were membrane filtration following EPA Method 1600 (U.S. EPA 2002) and Enterolert (IDEXX Laboratories, Westbrook, ME, USA) following the manufacturer's instructions. EPA 1600 quantifies Enterococcus based on their ability to grow atop solid media (i.e. mEI agar) and produces a blue halo through metabolizing the chromogen indocyl-β-glucoside in the media. Enterolert uses Defined Substrate Technology® and quantifies Enterococcus based on their ability to grow in liquid media and metabolize 4-methylumbelliferone-β-glucoside to produce fluorescence. Studies have shown that both culture methods have false positives but to various degrees depending on the study (i.e. Enterolert and EPA1600 detect non-Enterococcus 4–26 and 11–26% of the time, respectively), and the two culture-based methods also appear to select for different Enterococcus species (Ferguson et al. 2013).

Both of the two qPCR assays are based on the assay outlined by Ludwig and Schleifer (2000), target a similar region of the multiple copy 23S rRNA gene and were designed to be specific to the Enterococcus genus. The first assay uses a forward and a reverse primer with a Taqman™ probe following EPA Method A (U.S. EPA 2010). The second uses a forward primer and a Scorpion® primer–probe complex following Noble et al. (2010). The Taqman assay primers are located at positions 818–837 and 888–909 on the 23S rRNA gene (Enterococcus faecalis, accession number NC019770) and yield a 92-bp amplicon. In silico amplification using the probe check program [no mismatch allowed, (Loy et al. 2008)] with the SILVA 23S rRNA database showed that, while the Taqman probe offers no specificity to Enterococcus, the forward primer mostly amplifies Enterococcus and Streptococcus and the reverse primer mostly amplifies Enterococcus. The Scorpion method is proprietary and similar analysis could not be carried out to confirm method specificity. However, epidemiology studies have demonstrated comparable performance between the Taqman and the Scorpion qPCR methods (Colford et al. 2012).

For each sample, DNA was extracted from the polycarbonate filter using an extraction buffer (i.e. AE buffer spiked with salmon testes DNA) per the standard bead beating crude lysate protocol as described in EPA Method A (U.S. EPA 2010). Both qPCR methods were run on Bio-Rad CFX96 (Bio-Rad, Hercules, CA, USA) in duplicate 25 μl reactions with 5 μl of the sample DNA from the same filter. The Taqman reactions contained 1× Taqman Universal PCR Master Mix (Applied Biosystems, Carlsbad, CA, USA), 0·2 mg ml−1 bovine serum albumin (Sigma, St Louis, MO, USA), 1 μmol l−1 of each primer and 0·08 μmol l−1 of the probe. The Scorpion reactions contained 1× OmniMix HS (Cepheid, Sunnyvale, CA, USA), 0·25 μmol l−1 of primer and 0·25 μmol l−1 of primer–probe complex. The thermal conditions were 2 min at 50°C, 10 min at 95°C, followed by 40 cycles of 15 s at 95°C and 60 s at 60°C for the Taqman method, and 2 min at 95°C followed by 40 cycles of 5 s at 95°C and 43 s at 62°C for the Scorpion method. As per the common practice in recreational water monitoring, both Enterococcus qPCR methods had companion qPCR assays (i.e. Sketa) to measure the salmon testes DNA (i.e. that spiked into the extraction buffer) in duplicate reactions as sample processing and inhibition control. While the Sketa controls are sensitive to DNA loss during sample processing, they may be less susceptible to inhibition than the Enterococcus qPCR assays (Cao et al. 2012). Detailed protocols and evaluation of the Sketa control assays are described elsewhere (Cao et al. 2012).

qPCR calibrators were prepared using Ent. faecalis (ATCC #29212) cells. Cells were cultured overnight at 37°C in 1% brain heart infusion broth in phosphate-buffered saline (PBS) following standard protocol (U.S. EPA 2010). Cells were enumerated spectrophotometrically after 18 h, and cell suspensions were diluted with PBS. Calibration standards were prepared by filtering 100 000 cells onto 47-mm diameter, 0·4-μm pore-size polycarbonate filters. Filters were stored at −80°C until sample DNA isolation for qPCR analyses. A four-point duplicate standard curve (concentration range equivalent to 105–102 cells per filter) was included in each qPCR run using the calibrator and three serial 10-fold dilutions.

Results for both qPCR methods were calculated using the dCt quantification model as described previously (Noble et al. 2010; Griffith and Weisberg 2011; Cao et al. 2013) and expressed in cell equivalents (CE). In the dCt quantification model, the salmon control assays were only used for flagging samples that failed the sample processing and inhibition control criteria, but not for adjusting the Enterococcus quantification results. Samples where the control assay in either of the duplicate reactions yielded Ct values >1·7 cycles higher than that from the calibrator were rerun after 1 : 5 dilution with DNA free water (Griffith and Weisberg 2011; Cao et al. 2012). Among the 306 samples processed, 67 samples failed this 1·7 Ct quality control criteria even after dilution and were excluded from data analysis. While it is possible samples that failed the qPCR quality control criteria may exhibit different culture-vs-qPCR relationship, future studies employing qPCR methods more robust against inhibition (such as the TaqEnviron method in Cao et al. 2012) are needed for further investigation. As data from the majority (80%) of the samples were retained in this study, data analyses were expected to reveal the primary trends in the culture-vs-qPCR comparison.

As the Taqman qPCR method development was based on ddCt quantification model, in which Enterococcus results are adjusted based on performance of the salmon DNA control assay (Haugland et al. 2005), the Taqman qPCR results were also calculated using the ddCt model. As described elsewhere (Haugland et al. 2005; U.S. EPA 2010), the ddCt model uses a more relaxed quality control criteria based on the salmon DNA control assay. If the salmon assay Ct shift between the sample and the calibrator was outside the ±3·0 cycles, then the sample was excluded. Among the 306 sample processed, 16 samples failed this 3 Ct quality control criteria and were excluded from data analysis.

Data analysis

Enterococcus results were compared between two classes (culture and qPCR) of enumeration methods to assess whether and why qPCR yielded higher Enterococcus concentrations than culture methods. The between-class comparison was placed into context of within-class comparison (EPA1600 vs IDEXX within the class of culture-based method, and Taqman vs Scorpion within the class of qPCR method) and within-method (i.e. between duplicates within each of the four methods) to assess contribution of within-class and within-method variability to the observed discrepancy between qPCR vs culture results.

The comparisons were conducted either continuously, via regression analysis, or categorically, via contingency tables. Simple linear regression was performed to assess repeatability between duplicates within each method: a unit slope and zero intercept would indicate little variability within a method and little contribution of within-method variability to culture-vs-qPCR discrepancy. Similarly, regression was performed between methods within each class (culture or qPCR) and between either Taqman or Scorpion qPCR and EPA 1600: a unit slope and zero intercept would indicate strong agreement. ancova was used to compare slopes and intercepts among three regression lines: Enterolert, Taqman and Scorpion vs EPA 1600: different slopes would indicate different proportional biases against EPA 1600, while different intercepts would indicate different constant biases (i.e. over-or underestimation by a constant amount) against EPA 1600. Enterococcus concentrations were log10-transformed and all values <33 ENT per 100 ml, which is the approximate qPCR detection limit (Haugland et al. 2005), were excluded from the regression analyses.

While the regression analysis examined the general trend and qPCR-vs-culture discrepancy across the range of observed Entercoccus concentrations, beach monitoring applications typically involve comparing Enterococcus results to numeric thresholds for making beach posting decisions. How results differ at selected numeric cut points (i.e. management decision thresholds) influence utility of qPCR for beach monitoring applications. Although recreational water quality criteria differ slightly around the globe, the numeric thresholds are generally of similar magnitude. For example, the European Union uses 100 CFU Enterococci per 100 ml as the criteria for excellent bathing water quality for coastal waters (European Parliament 2006). Contingency tables were therefore generated to assess agreement between Taqman qPCR and EPA Method 1600 relative to the 104 CFU per 100 ml threshold used for beach management decisions in California. This was initially carried out comparing a single replicate of each method (as commonly carried out in routine monitoring), and repeated using the average of the two replicates for each method, to assess how much of the observed discrepancy was due to within-method variability. The additional methods (i.e. Enterolert, Scorpion qPCR) were then incorporated into a categorical analysis to assess the percentage of samples for which there were within-class differences compared to those for which both methods within a class indicated the same beach management decision and both methods of the other class indicated the opposite management decision. This allowed assessment of what percentage of the observed discrepancy was due to within-class-between-method variability, as opposed to true discrepancy between the two classes (culture vs qPCR) of enumeration methods.


Regression analyses

qPCR methods significantly correlated with EPA 1600 over the range of concentrations detected (max: 3·58 log10 Enterococcus), but generally overestimated Enterococcus compared to EPA 1600. The overestimation appeared to be more consistent with a raised ‘baseline’ (i.e. intercept >0) instead of a proportional overestimation (i.e. slope >1) (Table 1). Both qPCR methods had unit (or near unit) slopes in regression with EPA 1600, but intercepts that were significantly larger than zero and indicated 0·3–1·0 log10 unit overestimation compared to EPA 1600. Despite the significant correlation, a relatively high level of data spreading (as indicated by moderate R2 values) around the regression line was observed.

Table 1. Results from regression analyses between different Enterococcus enumeration methods. All slopes and intercepts (95% confidence intervals in parenthesis) are significantly different from zero (P < 0·05)
x y n SlopeIntercept R 2
  1. x And y indicate independent and dependent variables, respectively, in regression analyses. n and R2 are sample size and coefficient of determination, respectively, for the regression analysis.

  2. a

    The slope was significantly different from 1 (P < 0·05).

EPA 1600Taqman420·76 (0·51, 1)0·81 (0·3, 1·31)0·48
Scorpion380·77 (0·55, 0·99)a0·75 (0·3, 1·2)0·57
EPA 1600Enterolert380·87 (0·7, 1·04)0·36 (0·01, 0·72)0·75
TaqmanScorpion1080·8 (0·69, 0·90)a0·36 (0·14, 0·58)0·69

There was generally good agreement between the two methods within each class (culture or qPCR). There was little apparent systematic over- or underestimation of Enterolert compared to EPA 1600, or of Scorpion compared to Taqman, as indicated by unit (or near unit) slopes and near zero intercepts (Table 1). There was also less discrepancy within a class (i.e. between culture methods or between qPCR methods), compared to between classes (i.e. between qPCR and culture methods). The within-class-between-method regression R2 values were much higher than those for between-class (Taqman or Scorpion vs EPA 1600). Nevertheless, slopes among the regressions (Enterolert, Taqman, or Scorpion vs EPA 1600) were not significantly different (ancova, P = 0·70), while intercepts were significantly different (ancova, P = 0·009), confirming overestimation by qPCR compared to culture-based methods was likely due to a raised ‘baseline’ (i.e. a constant number of perhaps stressed or dead Enterococcus cells detectable by qPCR but not by culture methods in the ambient samples) instead of proportional overestimation.

Replicate results within each method showed good agreement. There was no systematic overestimation of one replicate compared to the other, as indicated by unit (or near unit) slopes and zero (or near zero) intercepts (Table 2). Data values also more closely followed the regression lines (as indicated by relatively high R2 values and low random measurement errors), compared to a higher level of spreading in the between-class and between-method regressions.

Table 2. Results from regression analysis between duplicates within each method. All slopes and intercepts (95% confidence intervals in parenthesis) are significantly different than zero (P < 0·05) except the intercepts highlighted in grey
Method n SlopeIntercept R 2
  1. a

    The slope was significantly different from 1 (P < 0·05).

  2. n And R2 are sample size and coefficient of determination, respectively, for the regression analysis.

EPA 1600330·75 (0·66, 0·85)a0·47 (0·27, 0·67)0·89
Enterolert340·94 (0·79, 1·09)0·11 (−0·23, 0·45)0·83
Taqman930·86 (0·77, 0·95)a0·29 (0·1, 0·48)0·8
Scorpion850·95 (0·88, 1·03)0·09 (−0·07, 0·25)0·88

Categorical analyses

Taqman qPCR and EPA 1600 results agreed with respect to California's beach management decision threshold for 86·6% of the 239 samples that passed the 1·7 Ct salmon control assay QC criteria (Table 3a). qPCR indicated a beach posting for unhealthful water should occur when EPA 1600 did not for 12·1% of the samples. In contrast, for only 1·3% samples, the culture method indicated a posting when qPCR did not. As the salmon control assay may be less susceptible to inhibition than the Enterococcus qPCR assay (Cao et al. 2012), it is possible that these few samples (Table S1) with qPCR results lower than that from EPA 1600 may have experienced inhibition but passed the QC criteria.

Table 3. Agreement relative to California's beach warning decision threshold (i.e. 104 Enterococcus per 100 ml) between EPA 1600 and Taqman qPCR when comparing a single replicate of each method (a) and when comparing the average of the two replicates for each method (b) (n = 239, i.e. number of samples included in the analysis)
 EPA1600 < 104EPA1600 > =104
  1. Percentages of disagreement are in bold.

(a) Single replicate comparison
qPCR < 10482·0% 1·3%
qPCR > = 104 12·1% 4·6%
(b) Comparison of averaged results
qPCR < 10482·8% 0·8%
qPCR > = 104 11·7% 4·6%

Consistent with the trend observed in the regression analysis, only a small portion of the culture-vs-qPCR discrepancy at the 104 cut-off point could be attributed to within-method variability (i.e. measurement error for either Taqman or EPA 1600). Discrepancy in beach management decision between the two replicates was the least (<1%) for EPA 1600 and approximately 3, 5, and 6% for Enterolert, Taqman and Scorpion qPCR methods, respectively (Table 4). When the average of the two replicates was used to compare Taqman qPCR and EPA 1600, only a slight increase in overall agreement (from 86·6 to 87·4%) and a slight decrease in the percentage of postings by qPCR but not by EPA 1600 (from 12·1 to 11·7%; Table 3) was observed.

Table 4. Agreement relative to California's beach warning decision threshold (i.e. 104 Enterococcus per 100 ml) between duplicate measurements within each method. n denotes the number of samples included in each analysis
 Rep 2 < 104Rep 2 > =104
  1. Percentages of disagreement are in bold.

EPA 1600 (n = 239)
Rep1 <10493·7% 0·4%
Rep1 > = 104 0·4% 5·4%
Enterolert (n = 239)
Rep1 < 10490·8% 2·1%
Rep1 > = 104 0·8% 6·3%
Taqman (n = 239)
Rep1 < 10480·8% 3·3%
Rep1 > = 104 2·1% 13·8%
Scorpion (n = 234)
Rep1 < 10482·9% 3·4%
Rep1 > = 104 2·6% 11·1%

A larger portion of the culture-vs-qPCR discrepancy could be attributed to differences between methods within a class (Table 5). Most of the within-class differences were associated with the two qPCR methods, which yielded different management decisions for 5·1% of the samples (i.e. row sum of the ‘qPCR mixed’ in Table 5). The two culture methods yielded different management decisions for only 1·7% of the samples (i.e. row sum of the ‘culture mixed’ in Table 5). However, even after accounting for within-class variability, there were still 7·7% samples for which both qPCR methods indicated a warning should be posted while both culture methods indicated it should not. In contrast, for only 0·9% samples, both culture methods indicated a warning should be posted when both qPCR methods did not.

Table 5. Agreement within and across method classes relative to California's beach warning decision threshold (n = 234, i.e. number of samples included in the analysis)
 EPA 1600 < 104 & Enterolert < 104Culture mixedaEPA 1600 > =104 & Enterolert > = 104
  1. Percentages of disagreement between classes are in bold.

  2. a

    Culture mixed: The measurement by one culture method is <104 Enterococcus per 100 ml, while measurement by the other culture method is > = 104; qPCR mixed: The measurement by one qPCR method is <104, while measurement by the other qPCR method is > = 104.

Taqman < 104 & Scorpion < 10480·8%0·4% 0·9%
qPCR mixeda4·3%0·4%0·4%
Taqman > = 104 & Scorpion > = 104 7·7% 0·9%4·3%

Similar trends were observed in the contingency table analysis when the ddCt quantification model was used for Taqman qPCR. For single replicate comparison, the overall management decision agreement of Taqman vs EPA 1600 was 72·8% and qPCR indicated a beach posting when EPA 1600 did not for 22·1% of the samples (Table 6a). Discrepancy in beach management decision between the two replicates of ddCt Taqman was relatively high (8·6%, compared to 5% for dCt Taqman). Nevertheless, when the average of the two replicates was used to compare ddCt Taqman qPCR and EPA 1600, a slight decrease in overall agreement (from 77·2 to 76·2%) and a slight increase in the percentage of posting by qPCR but not by EPA 1600 (from 22·1 to 23·1%; Table 6) was observed. However, accounting for the within-class method difference greatly reduced the percentage of posting by qPCR, but not culture methods to 13·7% (contingency table not shown).

Table 6. Agreement relative to California's beach warning decision threshold between EPA 1600 and Taqman qPCR using ddCt calculation model when comparing a single replicate of each method (a) and when comparing the average of the two replicates for each method (b) (n = 290, i.e. number of samples included in the analysis)
 EPA 1600 < 104EPA 1600 > = 104
  1. Percentages of disagreement are in bold.

(a) Single replicate comparison
qPCR < 10472·8% 0·7%
qPCR > = 104 22·1% 4·5%
(b) Comparison between averaged results
qPCR < 10472·1% 0·7%
qPCR > = 104 23·1% 4·1%


Our finding that qPCR overestimated Enterococcus relative to EPA Method 1600 is consistent with findings of previous comparative studies (Haugland et al. 2005; Byappanahalli et al. 2006; Lavender and Kinzelman 2009; Whitman et al. 2010). We focused on comparison of EPA 1600 with Taqman qPCR as the former is the predominant method presently in use and the latter is the method EPA is proposing for adoption as the molecular alternative. Nevertheless, our finding that qPCR generally yields higher concentration estimates compared to culture methods and would lead to a greater number of beach warnings than culture methods was unchanged regardless of which combination of methods between the two classes were compared or which quantification model was used for Taqman qPCR. However, we did observe a higher extent of overestimation by qPCR compared to EPA 1600 for the ddCt quantification model. This occurs because the ddCt model adjusts Enterococcus results by the salmon control assay results, which can produce up to an 8-fold increase in the qPCR results from that produced by the dCt model (Cao et al. 2013).

The most compelling explanation for the higher qPCR values is measurement target difference between qPCR and culture methods. Culture methods measure culturable cells that can achieve sufficient growth or metabolic activity on selected media (e.g. mEI or Enterolert media) within the specified incubation time (i.e. 18–24 h), while qPCR also measures stressed (e.g. viable but nonculturable) and dead-but-intact cells. This explanation was supported by both regression analyses and contingency table analyses, which demonstrated greater between-class difference than within-class difference, and even more so than within-method variability. Nevertheless, one challenge in reaching this conclusion is distinguishing measurement target difference from calibration error. A whole cell calibrator was used to relate gene copy number to cell equivalents and results would be systematically offset if the calibration was flawed, leading to a proportional over-or underestimation (Cao et al. 2013). However, this alternative explanation seems unlikely because the cell counts of the calibration material were verified using traditional methods, as well as by microscopy, and the unit slopes (Table 1) indicated absence of proportional overestimation by qPCR relative to culture methods. There is also the potential for the calibration material, which comes from one Enterococcus species, to contain different gene copies per cell than environmental cells that may represent a wide range of Enterococcus spp. However, this also seems an unlikely alternative explanation as the differences we observed between the two classes of methods (Table 1) were much greater than the potential gene copy differences among Enterococcus spp. For example, Ent. faecium contains six copies of the 23S rRNA gene per genome, while Ent. faecalis contains four copies per genome (Klappenbach et al. 2001), leading to only a 1·2-fold difference. Similarly, rapidly growing cells may have more than one genome per cell, which would change how gene copy numbers are related to cell equivalents. This scenario is also unlikely as the majority of calibrator and environmental cells would be at stationary phase with one genome per cell. This is because nutrient levels in the media used and in the natural environment are generally too low to sustain rapid exponential growth.

While overestimation relative to culture methods can be an impediment to adoption of qPCR, it is preferable to the alternative. Previous efforts indicate that when culture methods are used for posting beaches, c. 70% of the postings were incorrect due to the incubation time delay (Leecaster and Weisberg 2001). Health departments are more concerned with underestimation because it fails to protect swimmers (Griffith and Weisberg 2011), particularly because Enterococcus measurements are the screening tool for alerting them to a potential problem. In contrast, the increased number of beach postings (i.e. compared to culture methods) may be overprotective, but they usually trigger adaptive sampling that would allow the health departments to refine their warnings (Ferretti et al. 2013). It is also important to recognize that the term overestimation is used in a context relative to culture methods and should not be considered as absolute overestimation of total Enterococcus in the environmental waters. It is unclear whether qPCR is truly overprotective, as Wade et al. (Wade et al. 2006, 2008) found health risk relationships with qPCR to be equivalent to that of culture-based methods. Use of the term ‘false positive’ by qPCR compared to culture methods can be misleading and should be discouraged. In fact, studies have found that the survival of viruses is more similar to that of DNA than it is to culture-based measurements of Enterococcus, further supporting adoption of qPCR methods for microbiological monitoring (Walters et al. 2009).

It is also important to place any difference in outcomes from existing methods into context of other sources of error. This study was conducted as part of a technology transfer project (Griffith and Weisberg 2011) in which the collecting laboratories also processed the samples for the Scorpion qPCR assay and supplemental data checking procedures were employed to assess their rate of data recording errors. Four such recording errors were identified that would not have been discovered in typical monitoring applications. This is consistent with a previous evaluation that found data recording errors occurred for nearly 10% of samples in routine monitoring using culture methods, often by an order of magnitude, because of failure to account for sample dilution in the calculations (Griffith et al. 2006). This data recording error is comparable in frequency, and greater in magnitude, than the differences observed between culture and qPCR methods.

Additionally, implementing qPCR using the culture-based numeric threshold is a conservative approach used by early adopters (Griffith and Weisberg 2011; Ferretti et al. 2013). U.S. EPA has recommended different culture and qPCR management thresholds developed separately based on their corresponding indicator–illness relationships in epidemiology studies in the new Recreational Water Quality Criteria (U.S. EPA 2012). These illness-based thresholds, called beach action values (BAVs), account for the target difference between qPCR and culture methods and should remove discrepancy in management decision between culture and qPCR-based beach monitoring. Using these BAVs (70 colony-forming unit by EPA 1600 or 1000 calibrator cell equivalent by ddCt Taqman qPCR per 100 ml of water), we found that overall agreement on management decisions between culture and qPCR improved to over 92%, but qPCR results were much less likely to indicate a beach posting than EPA 1600 (Table 7). A recent study also reported less frequent beach posting based on qPCR BAV than based on culture BAV, and the authors attributed such discrepancy to the fact that the published BAVs were derived from epidemiology studies at point source impacted beaches and may not perform well at sites with diffusive sources of Enterococcus (Nevers et al. 2013). Site-specific conditions related to fate and transport of the source, prevalence of PCR-inhibitory compounds (Noble et al. 2010) and robustness of the chosen qPCR method against inhibition (Cao et al. 2012) can also affect the culture-vs-qPCR difference. USEPA recommends site-specific evaluation for determining qPCR-based management decision thresholds, which provides a mechanism for minimizing influence of culture-vs-qPCR target difference on monitoring applications.

Table 7. Agreement relative to EPA's new beach action values (BAVs) for EPA 1600 and ddCt Taqman qPCR (n = 290, i.e. number of samples included in the analysis). The BAVs are 70 colony-forming unit (CFU) or 1000 calibrator cell equivalent (CCE) per 100 ml of water
 EPA 1600 < 70 CFUEPA 1600 > = 70 CFU
  1. Percentages of disagreement are in bold.

qPCR < 1000cce91·0% 6·6%
qPCR > = 1000cce 0·7% 1·7%


The authors thank staff from SOCWA, OCSD and OCPHL who were instrumental in sample collection and processing of the traditional measures: Larry Honeybourne, Ann Harley, Keith Bacon, Laura Miller, Angie Richardson, Joe Guzman, Reza Mahallati, Charlie McGee, Mike von Winkleman, Art Diaz, Ben Ferraro, as well as support staff at SCCWRP who also helped with sample processing: Jen Topor, Elizabeth Scott, Nick Sadrpour and Johnny Griffith.

Conflict of interest

No conflict of interest declared.