Swimmer Risk of Gastrointestinal Illness from Exposure to Tropical

ARTICLE pubs.acs.org/est

Swimmer Risk of Gastrointestinal Illness from Exposure to Tropical Coastal Waters Impacted by Terrestrial Dry-Weather Runoff Emily J. Viau, Debbie Lee, and Alexandria B. Boehm* Environmental and Water Studies, Department of Civil & Environmental Engineering, Stanford University, Stanford, California 94305, United States

bS Supporting Information ABSTRACT: This study used molecular methods to measure concentrations of four enteric viruses (adenovirus, enterovirus, norovirus GI, and norovirus GII) and fecal source tracking markers (human, ruminant, and pig Bacteroidales) in land-based runoff from 22 tropical streams on O’ahu, Hawai’i. Each stream was sampled twice in the morning and afternoon during dry weather. Viruses and human Bacteroidales were widespread in the streams. Watershed septic tank densities were positively associated with higher occurrence of human Bacteroidales and norovirus. There were no associations between occurrence of viruses and fecal indicator concentrations. Virus concentrations and previously reported culturable Salmonella and Campylobacter were used as inputs to a quantitative microbial risk assessment (QMRA) model to estimate the risk of acquiring gastrointestinal (GI) illness from swimming in tropical marine waters adjacent to discharging streams. Monte Carlo methods were used to incorporate uncertainties in the dilution of stream discharge with seawater, swimmer ingestion volumes, pathogen concentrations, and dose
response parameters into the model. Median GI illness risk to swimmers from exposure to coastal waters adjacent to the 22 streams ranged from 0 to 21/1000. GI illness risks from viral exposures were generally orders of magnitude greater than bacterial exposures. Swimming adjacent to streams positive for norovirus or adenovirus resulted in the highest risks. The median risk adjacent to each stream was positively, significantly correlated to the concentration of Clostridium perfringens in the stream. Although a number of important assumptions were made to complete the QMRA, results suggest land-based runoff in the tropics as a potential source of GI illness risk, with pathogens coming from both human and nonhuman nonpoint sources including septic tanks.

’ INTRODUCTION In much of the world, surface waters are regulated for concentrations of fecal indicator bacteria (FIB) to control the spread of recreational waterborne illness. Standards adopted by nations are guided by epidemiology studies that document a correlative, but not causative, relationship between gastrointestinal (GI) disease and FIB such as Escherichia coli or enterococci.1 Most of these studies have been carried out in temperate climates in waters impacted by point sources of wastewater. This is despite the fact that point sources of wastewater are well-regulated in developed countries leaving nonpoint sources, such as storm runoff from urban and agricultural land, as the main contributors of pathogens to surface waters. Additionally, in the US, tropical beaches receive more visitors than all temperate beaches combined.2 Because tropical soils harbor high concentrations of FIB,3 researchers and managers have called into question the applicability of standards developed in temperate regions to the tropics.4 Efforts are needed to more fully understand the health risks associated with exposure to terrestrial runoff in the tropics and their relation to concentrations of fecal indicator organisms. r 2011 American Chemical Society

This research need is highlighted in an expert report commissioned by the U.S. Environmental Protection Agency (EPA) to inform the development of new recreational water quality criteria.5 Epidemiology studies are the gold standard for understanding health risks associated with environmental exposures to pathogens.4 Unfortunately, these studies are expensive; the cohort design favored in the US may require enrollment of over 10 000 individuals.6 A disadvantage of epidemiology studies is that they must use statistical constructs to relate illness outcomes to environmental exposures; thus, a mechanistic understanding of exposure routes and associated risks is difficult to obtain. Quantitative microbial risk assessment (QMRA) is gaining popularity as a method for estimating health risks from environmental exposures to pathogens.7 QMRA models are mechanistic in that they define exposures quantitatively. A limitation to their Received: March 24, 2011 Accepted: July 22, 2011 Revised: June 21, 2011 Published: July 22, 2011 7158

dx.doi.org/10.1021/es200984b | Environ. Sci. Technol. 2011, 45, 7158–7165

Environmental Science & Technology use is the lack of dose
response information for many pathogens as well as for susceptible populations. QMRA can be implemented in both static8 and dynamic9 modes, which ignore or consider, respectively, secondary transmission and immunity.10 In the field of recreational waterborne illness, QMRA has been used to explore the potential benefits of treatment technologies for surface water discharges11 and the differing health risks associated with exposure to pathogens from various animal feces.8,12 The present study uses static QMRA modeling to estimate the risk of acquiring GI illness from swimming in tropical marine waters adjacent to stream discharges. Using molecular methods, we measured concentrations of four enteric viruses in 22 streams on O’ahu, Hawai’i (n = 4 samples per stream): enterovirus, adenovirus, norovirus genogroup I (GI), and norovirus genogroup II (GII). We used the viral concentrations and previously reported data on culturable Salmonella and Campylobacter in streams13 as inputs to QMRA models. We tested the hypothesis that risk is correlated to fecal indicator concentrations, including enterococci, Escherichia coli, Clostridium perfringens, and F+ coliphage, which were previously reported,13 and new measurements of the ruminant-, pig-, and human-specific fecal Bacteroidales bacteria. We also explore associations between viruses, indicators, and septic tank density in stream watersheds.

’ MATERIALS AND METHODS Sample Collection. Water samples were collected from 22 O’ahu coastal streams during two dry weather sampling campaigns in December (14
18 December 2009) and March (28 March to 3 April 2010) as described previously.13 The streams discharge into coastal waters adjacent to popular beaches (Table 1 of Viau et al.,13 Figure 1) used for swimming and other water activities, like surfing. Briefly, 20 L water samples were aseptically collected twice (before sunrise and after high noon) upstream of the stream’s ocean outfall in triple rinsed, 10% hydrochloric acid-washed plastic containers (DNA/RNA free). Samples were placed on ice and processed within 6 h of collection. Two 3 L aliquots of water were membrane filtered to collect (1) viruses for downstream RNA/DNA extractions and (2) bacteria for downstream DNA extractions. Water was filtered through 0.45 μm-pore size nitrocellulose filters (HA type filters, Millipore, Billerica, MA). In most cases three filters were needed to achieve filtration of 3 L (1 L per filter), but more turbid samples used up to 12 filters. For virus filtrations, molecular grade MgCl2 (JT Baker, Hot Springs, AR) was added to the water prior to filtration to a final concentration of 50 mM to improve viral recovery from stream samples (which had a range of salinities).14 Directly after filtration, filters were stabilized by adding 500 μL of RNAlater (Ambion, Austin, TX) to cover each filter, allowing it to sit for 5 min, and vacuum aspirating.15 The filter was rolled into a 5 mL DNA/RNA-free tube and immediately frozen at
80 °C until analysis (3
6 month storage time before extraction). DNA/RNA-free field laboratory procedures were maintained throughout sample processing as described in the Supporting Information (SI). Negative filtration blanks were generated daily for molecular bacterial and viral analyses by filtering rinsewater (MgCl2 was added to viral blanks). Molecular Analyses. Nucleic acid extractions, quantitative PCR (qPCR) and reverse-transcriptase qPCR (RT-qPCR) analyses were performed in accordance with the minimum

ARTICLE

information for publication of quantitative real-time PCR experiments (MIQE) guidelines.16 Nucleic Acid Extractions. A simultaneous viral RNA/DNA extraction method was applied to viral filters using the MoBio PowerWater RNA kit (MoBio Laboratories, Carlsbad, CA) according to the manufacturer’s instructions with some modifications (see SI for details). RNA/DNA extracts were analyzed for viruses immediately or stored in small volumes at
80 °C. Extraction blanks were run daily. To estimate viral nucleic acid extraction efficiencies, a subset of stream virus filters were spiked with 107 plaque forming units (PFU) of ssRNA bacteriophage MS2 (ATCC 15597-B1) prior to RNA/DNA extraction. A positive recovery control was included as just the MS2 spike alone with no filter. MS2 recoveries were determined with the MS2 RT-qPCR assay in Table 1. Percent recoveries were calculated by comparing stream MS2 recoveries to the ‘no filter’ spiked recovery control. Bacterial DNA was extracted from filters with the MoBio PowerWater DNA kit according to the manufacturer’s instructions with slight modifications (see SI for details). An extraction blank was performed daily. Bacterial DNA extraction efficiencies were determined by spiking 107 cells of Enterococcus faecium (ATCC 19434) pure culture into the 5 mL tubes with filters for a subset of stream samples. The spiked stream filters and a recovery control (with just cells) were extracted using the DNA extraction protocol. Extraction efficiencies were quantified using the Enterococcus genus qPCR assay in Table 1. Average DNA extraction efficiencies were calculated by dividing the amount of E. faecium genomes recovered from the stream matrix by the amount extracted from the recovery control. Nucleic acid concentrations and purities were determined using a spectrophotometer (see SI). PCR Assays. Each stream sample was tested for human enterovirus, norovirus GI, and norovirus GII using previously published RT-qPCR primers and hydrolysis probes (Table 1). Samples were also tested for human adenovirus and human-specific Bacteroidales (humbac) using published hydrolysis qPCR assays (Table 1), and the presence of ruminant and pig-specific Bacteroidales using conventional PCR (CF193 and PF163, respectively, Table 1). Hydrolysis probes were 50 labeled with a 60 FAM reporter dye and 30 labeled with a Black Hole Quencher-1 (BHQ-1) (Biosearch Technologies, Novato, CA). For viral analyses, replicates were performed for each stream sample until at least 100 mL equivalent stream volume were analyzed (n = 3
9 replicates). For humbac, triplicate reactions were run. One reaction was performed per sample for the pig and ruminant-specific Bacteriodales conventional PCR assays (equivalent to ∼20 mL of streamwater). One reaction was run for each relevant lab and extraction blank in all assays tested. A no template control and between one and three different concentrations of positive calibration control were included in each plate run. Details of molecular assays can be found in the SI. Generation of Standards and Calibration Curves. Standards included in vitro transcribed RNA products, recombinant plasmid DNA, or genomic DNA as specified in Table 1. Details on standard generation can be found in the SI. A “pooled” calibration curve of target RNA or DNA was used for absolute quantification with each quantitative assay17 (see SI). Calibration curves were used to relate a quantification cycle (Cq) to a concentration (genes copies/reaction) if the Cq was within the range of the lowest and highest standards 7159

dx.doi.org/10.1021/es200984b |Environ. Sci. Technol. 2011, 45, 7158–7165

7160

Mixed bases in degenerate primers and probe are as follows: Y, C or T; R, A or G; W, A or T; N, any. F denotes sense primer, R denotes antisense primer, and P is hydrolysis probe sequence. Concentrations were determined with the following formula: CT = slope  log(conc.) + y intercept. qPCR were determined by PCR efficiency = 10
1/slope
1. c Standard curve calculations account for six 23S gene copies in each E. faecium cell.

b

a

Table 1. Quantitative PCR Toolbox for Oahu Stream Survey—Primers, Probes, “Pooled” Calibration Curve Parameters, and Controls

Environmental Science & Technology ARTICLE

dx.doi.org/10.1021/es200984b |Environ. Sci. Technol. 2011, 45, 7158–7165

Environmental Science & Technology

ARTICLE

(parameters provided in Table 1 and SI Figure S1). If the Cq was higher than that of the lowest standard, the sample was considered positive but below the limit of quantification (BLOQ) for that assay. It was subsequently assigned the lowest quantifiable value in a QPCR reaction, 1 gene copy, as a low-end estimate. For quantification of viruses, the number of gene copies detected in all replicates was summed and normalized by the equivalent volume of water tested to obtain a concentration in gene copies/100 mL. For humbac, gene copies in each replicate reaction were normalized to the equivalent volume of streamwater analyzed in each reaction, log-transformed, and averaged to obtain a concentration in gene copies/100 mL. If one or two of the three humbac replicates were negative it was assigned a target copy number of 0.5 gene copies prior to analysis. If all three replicates were negative, then the target was deemed “not detected”. Characterization of Inhibition. Spiking procedures were used to characterize inhibition in a subset of samples for each viral assay. One or more samples from 21 (adenovirus), 16 (enterovirus), 8 (norovirus G1), and 12 (norovirus GII) streams were spiked with 104 gene copies/reaction of positive control RNA or DNA for a given assay. Cq values were converted to concentrations and the percent recovery was calculated.

Ancillary Data. A previous study13 provided the following for

each stream sample: salinity, temperature, dissolved oxygen percent saturation, chlorophyll a (chla), turbidity, E. coli, enterococci, F+ coliphage, and Clostridium perfringens concentrations, as well as presence of Salmonella and Campylobacter in 1 L of water. The average density of septic tanks in each watershed of the 22 streams was determined using ArcGIS (ESRI, Redlands, CA), watershed boundary data from the Hawai‘i Statewide GIS program,18 and a georeferenced map of septic tank density.19 Quantitative Microbial Risk Assessment (QMRA). A static QMRA was used to estimate GI illness risk from swimming in marine waters adjacent to where streams discharge to the sea using Matlab version R2009b (The Mathworks, Natwick, MA); the influence of immunity and secondary transmission was not considered. Monte Carlo simulations (n = 10 000) were used to estimate the risk of GI illness from exposure to discharge from each stream while considering uncertainties associated with (1) dose
response parameters when possible, (2) probability of illness upon infection, (3) volume of water ingested, (4) dilution of streamwater in coastal marine waters, and (5) pathogen concentration (Tables 2, 3, and S1). It was assumed that enterovirus had infectivity of echovirus 12, adenovirus had infectivity of adenovirus 4,20 Salmonella had infectivity of nontyphi Salmonella, and Campylobacter had infectivity of C. jejuni
point estimates of dose
response parameters were used in the dose
response equations with exception of those for Salmonella (Table 2). The norovirus dose
response relation was assumed to apply to both GI and GII and to all swimmers even though it was developed for GI and Se+ individuals. It was assumed that norovirus was not aggregated in the environment because viral coagulation in natural waters has been shown to be negligible21 so the nonaggregated version of the dose
response relation was used. The range of probabilities of illness upon infection for each pathogen is also provided in Table 2; uniform distributions between minimum and maximum probabilities were assumed following the approach of others.10,22 Note that a range was not available for adenovirus, so that of enterovirus was adopted. Table 3. Variables Used As Inputs to QMRA Model

Figure 1. Occurrence of viruses in 22 O’ahu coastal streams. Adv, entero, noro I, and noro II are adenovirus, enterovirus, norovirus I, and norovirus II, respectively, and are represented by different quadrants of a circle (upper right corner). If a virus's quadrant is shaded, it indicates that the stream tested positive at least 1 out of 4 times for the virus.

variable

distribution used

V, volume ingested (mL)

ln-normal distribution with mean = 2.92 and stdev = 1.43 25,26

dilution (
)

log-uniform distribution with range between 1 and 0.01

Table 2. Dose-Response Relations for Enteric Pathogens Enumerated in This Study organism Salmonella

P

surrogate organism nontyphi Salmonella

a

Pill|inf (distribution)

Pill = 1
exp(
exp(
a+2.148  ln(μ)))

26

NA

a = 29
50 (uniform) C. jejuni

Pinf = 1
1
1F1(0.145,0.145 + 7.59,
μ) 8

0.1
0.6 (uniform) 22

enterovirus adenovirus

echovirus 12 adenovirus 4

Pinf = 1
exp(
μ/78.3) Pinf = 1
exp(
μ/2.397) 7

0.25
0.75 (uniform) 10 0.25
0.75 (uniform)

norovirus

norwalk virus, nonagg

Pinf = 1
1F1 (0.04, 0.04+ 0.055, -μ) 49,50

0.3
0.8 (uniform) 22

Campylobacter

7

μ is the dose, Pinf is probability of infection, Pill is probability of illness. For each organism, a surrogate organism is chosen for the dose-response. “Nonagg” is non-aggregated. 1F1 is the hypergeometric function. Pill|inf are represented by a range of parameters, as indicated, drawn from a uniform distribution. There is no Pill|inf for Salmonella because the dose response is for illness. All dose-response parameters are point estimates with the exception of a for Salmonella which is represented by a number drawn from a uniform distribution within the range indicated. a

7161

dx.doi.org/10.1021/es200984b |Environ. Sci. Technol. 2011, 45, 7158–7165

Environmental Science & Technology To estimate dose, we assumed that swimmer exposure occurred in the coastal ocean adjacent to the stream discharge in a region where streamwater was diluted up to 1:100 by surrounding pathogen-free coastal waters. A log-uniform distribution of dilutions was assumed between 0.01 and 1 (Table 3). The spatial scale over which such dilution will actually be observed is a complex function of stream momentum and buoyancy, as well as tides and waves.23 As a bench mark, a study in Huntington Beach, California indicated that the length of shoreline over which microbial pollutants decayed to 1/e of their initial concentrations from a momentumless point source was on the order of 1000 m.24 The distribution of water volume ingested (Table 3) was based on a study of pool swimmers which determined that the volume of water ingested (in mL) during a 45 min swimming event by a group of adults and nonadults was best represented by a loge-normal distribution with mean of 2.92 and standard deviation of 1.43.25,26 Virus concentrations in the ingested water were the concentrations measured herein (SI Table S1); concentrations of norovirus GI and GII were added to obtain total concentration of norovirus. In streams where detected, concentrations of Campylobacter and Salmonella were assumed to be 1 MPN/L based on results of Viau et al.13 This concentration represents the lowest detectable concentration given the enumeration methods used by the authors. To estimate risks from exposure to each pathogen in each stream, the pathogen concentration (SI Table S1) was randomly drawn from the 4 concentrations measured in the 4 replicate stream samples. The mathematical expressions for the probability of illness of each pathogen i (Pill_i), and the cumulative risk (Pill) due to all pathogens can be found in the SI. The Monte Carlo simulations provided 10 000 estimates of Pill_i and Pill for each stream (n = 22); these are represented in box plots, cumulative distribution functions, and by medians and percentiles. A sensitivity analysis following the methods outlined in Julian et al.27 was conducted to determine which parameters the model is most sensitive to (see SI). Statistical Methods. PASWStatistics version 18.0.2 (IBM Corporation, Somers, NY) was used for all statistical analyses. The lowest detected concentration of humbac (1.2 gene copies/ 100 mL) was substituted for nondetects in stream samples for statistical analysis. Microbial concentrations, watershed septic tank density, chla, and turbidity were log10-transformed prior to statistical analysis to achieve normality. Generalized estimating equations (GEEs) with a binary logistic linking function were used to examine associations between the presence of viruses and indicators as well as ancillary data. GEEs were used to account for repeated measures at each stream.28 The coefficients β and their p values are reported for statistically significant associations. Correlations and Kruskal
Wallis one way analysis of variance were used to investigate relations between predicted risks and microbial concentrations, as the risks were not normally distributed. Results were deemed statistically significant if p e 0.05; however some results for p e 0.1 are also presented.

’ RESULTS Molecular Assay Efficiency, Inhibition, And Controls. Extraction efficiency for viruses (as measured with MS2 for six samples) average(standard deviation 7(10%. For bacterial extractions (n = 8 samples), E. faecium extraction efficiency was 17(6%. We acknowledge that extraction efficiencies for

ARTICLE

Bacteroidales may differ from Enterococcus, particularly given their different cell membrane structures. Additionally, efficiencies for MS2 may differ from the human viruses tested herein. Assay inhibition was characterized for a given RT-qPCR/qPCR assay with a subset of O’ahu stream samples (one or more samples from between 8 and 21 streams, depending on assay). Inhibition was overall low (between 2 and 95% of spiked controls were recovered, see SI for more details). Although we included a number of streams in the inhibition study, we did not include them all, so it is possible that inhibition could be higher in an untested stream. No effort was made to adjust quantitative measures for extraction efficiency or inhibition, thus the measurements reported may be viewed as conservative. For each environmental sample tested, an associated negative control was carried through the (1) field lab, (2) the nucleic acids extraction procedure, and the (3) qPCR/RT-qPCR/PCR plate (no template control). In all cases, negative controls were negative. Virus Occurrence and Concentrations. Figure 1 shows the occurrence of viruses within streams; SI Table S1 shows concentrations. Enterovirus was detected in the fewest number of samples; it was present in 5 of 88 (6%) samples and ranged in concentration from 0.8 to 4.2 gene copies/100 mL. Adenovirus was detected in 13 of 88 (15%) samples and ranged in concentration from 0.4 to 4.8 gene copies/100 mL. Norovirus GI was most commonly detected. It was found in 19 of 88 (22%) samples and ranged in concentration from 1.2 to 1441 gene copies/100 mL. Norovirus GII was detected in 11 of 88 (12.5%) samples and ranged in concentration from 0.9 to 62.4 gene copies/100 mL. In only one stream, Moanalua, were all four viruses detected. Five streams had three of the four viruses detected (Kalihi, Makua, Nanakuli, Kahana, and Kiikii). Six streams had two virus types detected, six streams had only one virus type detected, and the remaining four streams had no virus detection. Bacteroidales Occurrence and Concentrations. Humanspecific Bacteroidales (humbac) were detected in 62 of the 88 samples and at least once in each of the 22 streams. Concentrations ranged from 1 to 11 000 gene copies/100 mL (SI Table S1). The log-mean concentration across all samples (with the lowest detected concentration substituted for the 26 nondetects) was 1.0 log gene copies/100 mL. Ruminant-specific Bacteroidales (CF192) were not detected in any samples (0/88). Pig-specific Bacteroidales (PF163) were detected in 8 of 88 samples (10%); they were detected in one of four samples from Waimea and Anahulu streams, and in three of four samples collected from Kahana and Punalu’u streams (SI Table S1). Co-Variation of Virus and Indicator Organisms. We used generalized estimated equations (GEE) to examine associations between virus occurrence and humbac, PF163, as well as previously reported concentrations of enterococci, E. coli, Clostridium perfringens, and F+ coliphage.13 E. coli was negatively associated with adenovirus (β =
0.73, p = 0.02). Adenovirus and norovirus GI were less frequently detected when humbac was present (adenovirus: β =
1.6, p = 0.02; norovirus GI: β =
1.3, p = 0.02). There were no other significant associations. Previous work reported the presence of culturable Salmonella and Campylobacter in the same samples as those analyzed here for virus and Bacteroidales indicators; we compared the presence of these pathogens to newly reported Bacteroidales indicators (humbac and PF163). Humbac occurrence was positively 7162

dx.doi.org/10.1021/es200984b |Environ. Sci. Technol. 2011, 45, 7158–7165

Environmental Science & Technology

ARTICLE

to mean Clostridium perfringens concentration in the stream (Spearman’s F = 0.47, p = 0.03). A sensitivity analysis (described in detail in the SI and in SI Table S4) indicated the model outputs of cumulative illness risk are most sensitive to stream dilution, volume of water ingested, and pathogen concentrations. The model was insensitive to dose
response parameters.

’ DISCUSSION Human Viruses and a Human-Specific Bacteroidales Fecal Marker Were Widespread in O’ahu Streams. Sources of

Figure 2. Risk of GI illness from exposure to discharge from each stream. Boxes show the interquartile range (25th to 75th percentile), the line within the box shows the median. SI Table S3 provides the numerical values of medians and percentiles. SI Figure S2 shows results for each pathogen individually.

associated with Campylobacter (β = 1.4, p = 0.02). There were no other significant associations. Relationship between Virus Presence and Ancillary Data. Using GEE analyses, we found no significant variations in salinity, temperature, dissolved oxygen, chla, or turbidity when viruses were present or absent with one exception; adenovirus was positively associated with chla (β = 1.0, p = 0.03). Watershed septic tank density (SI Table S2) was positively associated with norovirus GII presence (β = 1.4, p = 0.011), and negatively associated with the presence of enterovirus (β =
1.1, p = 0.001). The presence of humbac was positively associated with septic-tank density at the 10% confidence level (β = 0.59, p = 0.06). Estimation of GI Illness Risk. Across all streams, cumulative GI illness risk estimated from the Monte Carlo simulations ranged from 0 to 0.8 with a median of 10
5. Risks due to exposures to adenovirus (range across all streams of 0
0.7), norovirus (range of 0
0.5), and echovirus (range of 0
0.2) were higher than the risks from Campylobacter exposures (range of 0
0.01) and much higher than risks from Salmonella exposures (range of 0
10
13) (SI Figure S2). Figure 2 and SI Table S3 show the cumulative risks of illness from exposure to discharge from each of the streams. The median cumulative risk at each stream ranged from 0 to 0.02. One of the streams had a median risk higher than 0.019, which is the U.S. EPA tolerable level of enteric illness risk from exposure to marine waters.4 The median risk at each stream was significantly, positively correlated

viruses and humbac could include open defecation by homeless or others, leaking sewage infrastructure, and/or septic tanks. Although quantitative information on homeless populations within each watershed is not available, field technicians observed large homeless encampments near some streams, specifically Moanalua, Kalihi, Nanakuli, and Kiikii, all of which had three or four virus types detected. O’ahu watersheds have septic tanks as well as separated storm and sewage systems. An association between the presence of septic tank density and both norovirus GII and humbac is consistent with septic tanks being a source of these targets. Interestingly, there was an inverse association between both adenovirus and norovirus GI and humbac. Differential persistence and transport of these targets could explain their lack of positive correlation in field samples downstream from their human sources. Differential persistence may also explain the negative association between E. coli and adenovirus. Humbac has been detected in nonhuman sources;29 this may also explain lack of association with viruses. Lack of positive correlations between viruses and bacterial indicators has been reported previously.30 An Association Between Campylobacter and HumanSpecific Fecal Bacteroidales Suggests Humans Contribute this Enteric Pathogen to Streams. Salmonella did not correlate to the presence of humbac or PF163, but was weakly associated with septic tank density (β = 0.62, p = 0.083) suggesting septage could be its source. It is important to note that Campylobacter and Salmonella have a number of animal reservoirs12 including pigs which are common in these watersheds. Median Gastrointestinal (GI) Illness Risk from Swimming in Discharge from O’ahu Streams was 0.01%. The median risk was highest at 2% adjacent to Makua stream. A comparison of individual pathogen risks to cumulative risks suggests that viruses are likely the main etiologies of GI. Median risks were higher for exposures to stream waters that contained adenovirus (Kruskal
Wallis, p < 0.05) or either norovirus (Kruskal
Wallis, p < 0.1). This finding supports previous reports that norovirus is a main driver of recreational waterborne illness.8 There was a significant positive correlation between GI illness risk and C. perfringens; no other microbial indicators correlated to risk. Thus monitoring waters for C. perfringens, which is routinely done by the Department of Health on O’ahu, appears to be a valid method for assessing health risk in coastal waters impacted by stream discharge. There Are Several Important Limitations of This Study. We included 5 enteric pathogens in our estimates of GI illness risk, but there are many more. In particular, we did not include measurements of rotavirus, Cryptosporidium, Giardia, or E. coli O157:H7 which have been used as reference pathogens in other QMRA studies of recreational waterborne illness.12 7163

dx.doi.org/10.1021/es200984b |Environ. Sci. Technol. 2011, 45, 7158–7165

Environmental Science & Technology The dose
response relations used herein for echovirus and adenovirus were developing using cultured virus counts while the norovirus dose
response curve was developed using RT-qPCR enumerated viruses. Using RT-qPCR-based measures of enterovirus and adenovirus as inputs to the dose
response curves may yield either over- or underestimates of risk because RT-qPCR may over- or underestimate the number of infectious viruses.31 Two previous studies of marine bathing waters showed higher detection frequency of enterovirus by infectivity assays compared to molecular assays.32,33 On the other hand, adenoviruses were detected more frequently by qPCR than infectivity assays in urban rivers.34 Well-established cell lines exist for adenovirus and enterovirus. However, there is no robust cell line available for cultivation of norovirus,35 so detection with molecular methods is presently the only option. At the time of this study there were no established dose
response curves for norovirus GII, so norovirus risks may represent and under- or overestimation depending on the infectivity of this group. Bacterial concentrations used herein are likely underestimates. Viau et al.13 reported presence of Salmonella and Campylobacter (including C. jejuni) in one liter stream samples, so we conservatively assumed that their concentration was 1 MPN/L. Thus, the true risk attributable to these pathogens may be higher than those presented here. We assumed that when stream discharge enters the coastal ocean, pathogens are diluted and no die-off occurs. The pathogens considered here may undergo inactivation while undergoing dilution. Future work with sophisticated process-based models to predict stream plume and microorganism fate would better inform the risk assessment. Indeed, the sensitivity analysis indicated that dilution of streamwater has a great effect on model outputs. In order to apply static QMRA, we made simplifying assumptions about exposure and used published dose
response curves. There is a large degree of uncertainty in these parameters and relationships that were not considered herein (e.g., the effect of immunity, the assumptions that adenovirus have the same infectivity as adenovirus 420, a respiratory virus, and that norovirus GII has the same infectivity of norovirus GI). Although our risk estimates are for gastrointestinal illness, the stream waters contain pathogens that cause other illnesses, such as Staphylococcus aureus.13 Furthermore, other nonpoint sources of pathogens may also contribute to ocean swimmer risks (e.g., groundwater discharge, beach sands), so the reported stream risks may only characterize a fraction of cumulative swimmer risks. Additionally, sampling was conducted during dry weather, so risks could be different and potentially higher during and after storm events. Regardless, we have provided estimates of gastrointestinal illness risk from exposure to tropical marine waters polluted with pathogens from terrestrial runoff and given the information presently available, these are the best estimates possible.

’ ASSOCIATED CONTENT

bS

Supporting Information. Analytical details of methods and results, Tables S1
S3, and Figures S1
S2. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Phone: (650)724-9128; fax: (650)723-7058; e-mail: aboehm@ stanford.edu.

ARTICLE

’ ACKNOWLEDGMENT Alissa Rogers, Jeff Soller, Lauren Sassoubre, Mike Cooke, Ari Mesa, Mark Mench, Patie Boehm, Blythe Layton, Kevan Yamahara, Shimi Rii, Hsin-I Tong, Simon Wong, Siobhan Burns, and Yuanan Lu are acknowledged for their assistance. Two anonymous reviewers provided comments that improved the manuscript. This work was funded by NSF Oceans and Human Health: Pacific Research Center for Marine Biomedicine Grant No. OCE/0910491. ’ REFERENCES (1) Boehm, A. B.; Soller, J. A., Risks associated with recreational waters: pathogens and fecal indicators. In Encyclopedia of Sustainability Science and Technology; Meyers, R. A., Ed.; Springer: 2011, doi: 10.1007/ 978-1-14419-0851-3. (2) Leeworthy, V. R.; Wiley, P. C., Marine recreation participation and use, National Survey on Recreation and the Environment. In National Ocean Service; S. P. O., Ed.; National Oceanic and Atmospheric Administration: Silver Spring, MD, 2001. (3) Hardina, C. M.; Fujioka, R. S. Soil: the environmental source of Escherichia coli and enterococci in Hawai‘i’s streams. Environ. Toxicol. Water Qual. 1991, 6, 185–195. (4) 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, 9–20. (5) Report of the Experts Scientific Workshop on Critical Research Needs for the Development of New or Revised Recreational Water Quality Criteria, 823-R-07-006; United States Environmental Protection Agency (USEPA): Washington, DC, 2007. (6) Colford, J. M., Jr.; Wade, T. J.; Schiff, K. C.; Wright, C. C.; Griffith, J. F.; Sandhu, S. K.; Burns, S.; Hayes, J.; Sobsey, M.; Lovelace, G.; Weisberg, S. Water quality indicators and the risk of illness at non-point source beaches in Mission Bay, California. Epidemiology 2007, 18 (1), 27–35. (7) Haas, C.; Rose, J.; Gerba, C. P. Quantitative Microbial Risk Assessment; John Wiley and Sons: New York, 1995. (8) Schoen, M. E.; Ashbolt, N. J. Assessing pathogen risk to swimmers at non-sewage impacted recreational beaches. Environ. Sci. Technol. 2010, 44, 2286–2291. (9) Eisenberg, J. N. S.; Soller, J. A.; Scott, J.; Eisenberg, D. M.; Colford, J. M., Jr. A dynamic model to assess microbial health risks associated with beneficial uses of biosolids. Risk Anal. 2004, 24 (1), 221–236. (10) Soller, J. A.; Eisenberg, J. N. S. An evaluation of parsimony for microbial risk assessment models. Environmetrics 2008, 19 (1), 61–78. (11) Soller, J. A.; Olivieri, A. W.; Crook, J.; Cooper, R. C.; Tchobanoglous, G.; Parkin, R. T.; Spear, R. C.; Eisenberg, J. N. S. Risk-based approach to evaluate the public health benefit of additional wastewater treatment. Environ. Sci. Technol. 2003, 37 (9), 1882–1891. (12) Soller, J. A.; Schoen, M. E.; Bartrand, T.; Ravenscroft, J. E.; Ashbolt, N. J. Estimated human health risks from exposure to recreational waters impacted by human and non-human sources of faecal contamination. Water Res. 2010, 44 (16), 4674–4691. (13) Viau, E. J.; Goodwin, K. D.; Yamahara, K. M.; Layton, B. A.; Sassoubre, L. M.; Burns, S.; Tong, H.-I.; Wong, S.; Boehm, A. B. Human bacterial pathogens and fecal indicators in tropical streams discharging to Hawaiian coastal waters. Water Res. 2011, 45 (11), 3279–3290. (14) Victoria, M.; Guimaraes, F.; Fumian, T.; Ferreira, F.; Vierira, C.; Leite, J. P.; Miagostovich, M. Evaluation of an adsorption-elution method for detection of astrovirus and norovirus in environmental waters. J. Virol. Methods 2009, 156 (1
2), 73–79. (15) Keating, D. T. K.; Malizia, A. P.; Sadlier, D.; Hurson, C.; Wood, A. E.; McCarthy, J.; Nolke, L.; Egan, J. J.; Doran, P. P. Lung tissue storage: optimizing conditions for future use in molecular research. Exper. Lung Res. 2008, 34 (8), 455–466. (16) Bustin, S. A.; Benes, V.; Garson, J. A.; Hellemans, J.; Huggett, J.; Kubista, M.; Mueller, R.; Nolan, T.; Pfaffl, M. W.; Shipley, G. L.; 7164

dx.doi.org/10.1021/es200984b |Environ. Sci. Technol. 2011, 45, 7158–7165

Environmental Science & Technology Vandesompele, J.; Wittwer, C. T. The MIQE Guidelines: Minimum Information for publication of Quantitive real-time PCR Experiments. Clin. Chem. 2009, 55 (4), 611–622. (17) Sivaganesan, M.; Haugland, R. A.; Chern, E. C.; Shanks, O. C. Improved strategies and optimization of calibration models for real-time PCR absolute quantification. Water Res. 2010, 44 (16), 4726–4735. (18) State of Hawaii Hawaii Statewide GIS Program. http://hawaii. gov/dbet/gis/ (accessed July 28, 2011). (19) Whittier, R. B.; El-Kadi, A. I. Human and Environmental Risk Ranking of Onsite Sewage Disposal Systems; School of Ocean and Earth Science Technology, University of Honolulu: Honolulu, HI, 2009. (20) Crabtree, K. D.; Gerba, C. P.; Rose, J. B.; Haas, C. N. Waterborne adenovirus: a risk assessment. Water Sci. Technol. 1997, 35 (11
12), 1–6. (21) Grant, S. B. Virus coagulation in aqueous environments. Environ. Sci. Technol. 1994, 28 (5), 928–933. (22) USEPA. Quantitative microbial risk assessment to estimate illness in freshwater impacted by agricultural animal sources of fecal contamination; U.S. Environmental Protection Agency: Washington, DC, 2010. (23) Nash, J. D.; Jirka, G. H. Buoyant surface discharges into unsteady ambient flows. Dyn. Atmos. Oceans 1996, 24 (1
4), 74–84. (24) Boehm, A. B. Model of microbial transport and inactivation in the surf zone and application to field measurements of total coliform in Northern Orange County, California. Environ. Sci. Technol. 2003, 37 (24), 5511–5517. (25) Dufour, A. P.; Evans, O.; Behymer, T. D.; Cantu, R. Water ingestion during swimming activities in a pool: A pilot study. J. Water Health 2006, 4 (4), 425–430. (26) Soller, J. A.; Seto, E.; Olivieri, A. W. Microbial Risk Assessment Interface Tool: User Documentation; Water Environment Research Foundation, Alexandria, VA, 2008. (27) Julian, T. R.; Canales, R. A.; Leckie, J. O.; Boehm, A. B. A model of exposure to rotavirus from non-dietary ingestion iterated by simulated intermittent contacts. Risk Anal. 2009, 25 (5), 617–632. (28) Hardin, J. W.; Hilbe, J. M., Generalized Estimating Equations; Chapman and Hall/CRC: New York, 2003. (29) Wang, D.; Silkie, S. S.; Nelson, K. L.; Wuertz, S. Estimating true human and animal host source contribution in quantitative microbial source tracking using the Monte Carlo method. Water Res. 2010, 44 (16), 4760–4775. (30) Noble, R. T.; Fuhrman, J. D. Enteroviruses detected by reverse transcriptase polymerase chain reaction from the coastal waters of Santa Monica Bay, California: low correlation to bacterial indicator levels. Hydrobiologia 2001, 460 (1
3), 175–183. (31) Rodriguez, R. A.; Pepper, I. L.; Gerba, C. P. Application of PCRbased methods to assess the infectivity of enteric viruses in environmental samples. Appl. Environ. Microbiol. 2009, 75 (2), 297–307. (32) Moce-Llivina, L.; Lucena, F.; Jofre, J. Enteroviruses and bacteriophages in bathing waters. Appl. Environ. Microbiol. 2005, 71 (11), 6838–6844. (33) Reynolds, K. A.; Roll, K.; Fujioka, R. S.; Gerba, C. P.; Pepper, I. L. Incidence of enteroviruses in Mamala Bay, Hawaii using cell culture and direct polymerase chain reaction methodologies. Can. J. Microbiol. 1998, 44 (6), 598–604. (34) Choi, S.; Jiang, S. C. Real-time PCR quantification of human adenoviruses in urban rivers indicates genome prevalence but low infectivity. Appl. Environ. Microbiol. 2005, 71 (11), 7426–7433. (35) Duizer, E.; Schwab, K. J.; Neill, F. H.; Atmar, R. L.; Koopmans, M. P. G.; Estes, M. K. Laboratory efforts to cultivate noroviruses. J. Gen. Virol. 2004, 85 (Pt 1), 79–87. (36) Jothikumar, N.; Cromeans, T. L.; Hill, V. R.; Lu, X.; Sobsey, M. D.; Erdman, D. D. Quantitative real-time PCR assays for detection of human adenoviruses and identification of serotypes 40 and 41. Appl. Environ. Microbiol. 2005, 71 (6), 3131–3136. (37) Allard, A.; Albinsson, B.; Wadell, G. Rapid typing of human adenoviruses by a general PCR combined with restriction endonuclease analysis. J. Clin. Microbiol. 2001, 39 (2), 498–505. (38) Gregory, J. B.; Litaker, R. W.; Noble, R. T. Rapid one-step quantitative reverse transcriptase PCR Assay with competitive internal

ARTICLE

positive control for detection of enteroviruses in environmental samples. Appl. Environ. Microbiol. 2006, 72 (6), 3960–3967. (39) Walters, S. P.; Yamahara, K. M.; Boehm, A. B. Persistence of nucleic acid markers of health-relevant organisms in seawater microcosms; Implications for their use in assessing risk in recreational waters. Water Res. 2009, 43 (19), 4929–4939. (40) Svraka, S.; Duizer, E.; Vennema, H.; Bruin, E. d.; Veer, B. v. d.; Dorresteijn, B.; Koopmans, M. Etiological role of viruses in outbreaks of acute gastroenteritis in The Netherlands from 1994 through 2005. J. Clin. Microbiol. 2007, 45 (5), 1389–1394. (41) Kageyama, T.; Kojima, S.; Shinohara, M.; Uchida, K.; Fukushi, S.; Hoshino, F. B.; Takeda, N.; Katayama, K. Broadly reactive and highly sensitive assay for Norwalk-like viruses based on real-time quantitative reverse transcription-PCR. J. Clin. Microbiol. 2003, 41 (4), 1548–1557. (42) Loisy, F.; Atmar, R. L.; Guillon, P.; Le Cann, P.; Pommepuy, M.; Le Guyader, F. Real-time RT-PCR for norovirus screening in shellfish. J. Virol. Methods 2005, 123 (1), 1–7. (43) O’Connell, K. P.; Bucher, J. R.; Anderson, P. E.; Cao, C. J.; Khan, A. S.; Gostomski, M. V.; Valdes, J. J. Real-time fluorogenic reverse transcription-PCR assays for detection of bacteriophage MS2. Appl. Environ. Microbiol. 2006, 72 (1), 478–483. (44) 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. (45) Kildare, B. J.; Leutenegger, C. M.; McSwain, B. S.; Bambic, D. G.; Rajal, V. B.; Wuertz, S. 16S rRNA-based assays for quantitative detection of universal, human-, cow-, and dog-specific fecal Bacteroidales: A Bayesian approach. Water Res. 2007, 41 (16), 3701–3715. (46) Bernhard, A. E.; Field, K. G. Identification of nonpoint sources of fecal pollution in coastal waters by using host-specific 16S ribosomal DNA genetic markers from fecal anaerobes. Appl. Environ. Microbiol. 2000, 66 (4), 1587–1594. (47) Bernhard, A. E.; Field, K. G. A PCR assay to discriminate human and ruminant feces on the basis of host differences in BacteroidesPrevotella genes encoding 16S rRNA. Appl. Environ. Microbiol. 2000, 66 (10), 4571–4574. (48) Boehm, A. B.; Yamahara, K. M.; Walters, S. P.; Layton, B. A.; Keymer, D. P.; Strickfaden, R. M.; Knee, K. L.; Rosener, M. Dissolved inorganic nitrogen, soluble reactive phosphorous, and microbial pollutant loading from tropical rural watersheds in Hawai’i to the coastal ocean during non-storm conditions. Estuaries Coasts 2011, doi: 10.1007/s12237-010-9352-8. (49) Tuenis, P. F.; Moe, C. L.; Liu, P.; Miller, S. E.; Lindesmith, L.; Baric, R. S.; Pendu, J. L.; Calderon, R. L. Norwalk virus: How infectious is it? J. Med. Virol. 2008, 80 (8), 1468–1476. (50) Schoen, M. E. Personal Communication. 2011.

7165

dx.doi.org/10.1021/es200984b |Environ. Sci. Technol. 2011, 45, 7158–7165