Relationships between Microbial Indicators and Pathogens in Recreational Water Settings

Fecal pollution of recreational waters can cause scenic blight and pose a threat to public health, resulting in beach advisories and closures. Fecal indicator bacteria (total and fecal coliforms, Escherichia coli, and enterococci), and alternative indicators of fecal pollution (Clostridium perfringens and bacteriophages) are routinely used in the assessment of sanitary quality of recreational waters. However, fecal indicator bacteria (FIB), and alternative indicators are found in the gastrointestinal tract of humans, and many other animals and therefore are considered general indicators of fecal pollution. As such, there is room for improvement in terms of their use for informing risk assessment and remediation strategies. Microbial source tracking (MST) genetic markers are closely associated with animal hosts and are used to identify fecal pollution sources. In this review, we examine 73 papers generated over 40 years that reported the relationship between at least one indicator and one pathogen group or species. Nearly half of the reports did not include statistical analysis, while the remainder were almost equally split between those that observed statistically significant relationships and those that did not. Statistical significance was reported less frequently in marine and brackish waters compared to freshwater, and the number of statistically significant relationships was considerably higher in freshwater (p < 0.0001). Overall, significant relationships were more commonly reported between FIB and pathogenic bacteria or protozoa, compared to pathogenic viruses (p: 0.0022–0.0005), and this was more pronounced in freshwater compared to marine. Statistically significant relationships were typically noted following wet weather events and at sites known to be impacted by recent fecal pollution. Among the studies that reported frequency of detection, FIB were detected most consistently, followed by alternative indicators. MST markers and the three pathogen groups were detected least frequently. This trend was mirrored by reported concentrations for each group of organisms (FIB > alternative indicators > MST markers > pathogens). Thus, while FIB, alternative indicators, and MST markers continue to be suitable indicators of fecal pollution, their relationship with waterborne pathogens, particularly viruses, is tenuous at best and influenced by many different factors such as frequency of detection, variable shedding rates, differential fate and transport characteristics, as well as a broad range of site-specific factors such as the potential for the presence of a complex mixture of multiple sources of fecal contamination and pathogens.


Introduction
Approximately 39% of the United States (US) population and more than 50% of the global population live near a coastal area [1,2]. Coastal tourism accounts for 85% of all tourism revenue Even though the concepts of FIB, alternative indicators, and MST markers were developed to indicate fecal contamination and its sources, the same paradigm is often employed to indicate pathogen presence under the assumption that indicators consistently covary with pathogen presence. The goals of this review are: (1) to examine reported relationships between various indicators and pathogen species to determine the feasibility of indicators as pathogen sentinels in recreational waters; and (2) to identify factors that affect this relationship (or lack thereof). In addition, we also queried epidemiological studies to determine which indicator(s) most commonly correlated with illness in recreational waters. Our search criteria mandated that each study measured at least one indicator (FIB or alternative) or MST marker along with at least one pathogen. We focused on studies conducted in waters intended for primary human contact (e.g., swimming, wading, diving, and surfing) such as beaches and swimming pools, but also included ambient waters used for secondary or non-contact (e.g., boating, fishing) human activities. Our methodology for collecting the manuscripts involved querying "PubMed" (www.ncbi.nlm.nih.gov/pubmed/) and "Google Scholar" (https://scholar.google.com/) databases for following keywords: "recreational water pathogens", "recreational water viral pathogens", "recreational water bacterial pathogens", "recreational water protozoan pathogens", "recreational water fungal pathogens", "swimming pools pathogens", "swimming pools viral pathogens", "swimming pools bacterial pathogens, "swimming pools protozoan pathogens" and "swimming pools fungal pathogens" regardless of the year published. For the purposes of this review, a relationship is identified as a significant correlation (e.g., Pearson Product Momentum Correlation, Wilcoxon signed-rank tests) and/or significant predictive relationship (e.g., binary and other logistic regression modelling). Assumptions made in our analyses included the following: (1) all measurement strategies yielded equivalent results (e.g., various culture-based, molecular and microscopy were equally sensitive); and (2) data were not affected by characteristics of the water samples (e.g., we assumed that the water chemistry did not influence performance of the methods). In total, we collected 73 papers spanning over four decades of research from 25 countries: Argentina, Australia, Bolivia, Brazil, Canada, China, Cyprus, Democratic Republic of Congo, France, Greece, Germany, Hungary, Iceland, Italy, Japan, Luxembourg, Netherlands, New Zealand, Poland, Portugal, South Africa, Taiwan, United Kingdom, US, and Venezuela. The majority of studies were conducted in freshwaters (lakes, rivers and streams), followed by marine/brackish waters and swimming pools ( Figure 1). Since some studies were conducted in both fresh and marine/brackish waters, they were included in each water type in Figure 1. This resulted in a total of 126 observations (i.e., report on a relationship between indicator and pathogen) since some studies were conducted in both, marine and freshwater, and/or measured more than one type of indicator or pathogen. The majority of observations (n = 52) did not report any relationship between indicator(s) and pathogen(s), while those that did, were split into relationships that were statistically significant (n = 30) and those that were not (n = 44). Statistically significant relationships (or the lack thereof) and rationale for the observed trends are further examined in the following sections. the feasibility of indicators as pathogen sentinels in recreational waters; and (2) to identify factors that affect this relationship (or lack thereof). In addition, we also queried epidemiological studies to determine which indicator(s) most commonly correlated with illness in recreational waters. Our search criteria mandated that each study measured at least one indicator (FIB or alternative) or MST marker along with at least one pathogen. We focused on studies conducted in waters intended for primary human contact (e.g., swimming, wading, diving, and surfing) such as beaches and swimming pools, but also included ambient waters used for secondary or non-contact (e.g., boating, fishing) human activities. Our methodology for collecting the manuscripts involved querying "PubMed" (www.ncbi.nlm.nih.gov/pubmed/) and "Google Scholar" (https://scholar.google.com/) databases for following keywords: "recreational water pathogens", "recreational water viral pathogens", "recreational water bacterial pathogens", "recreational water protozoan pathogens", "recreational water fungal pathogens", "swimming pools pathogens", "swimming pools viral pathogens", "swimming pools bacterial pathogens, "swimming pools protozoan pathogens" and "swimming pools fungal pathogens" regardless of the year published. For the purposes of this review, a relationship is identified as a significant correlation (e.g., Pearson Product Momentum Correlation, Wilcoxon signed-rank tests) and/or significant predictive relationship (e.g., binary and other logistic regression modelling). Assumptions made in our analyses included the following: (1) all measurement strategies yielded equivalent results (e.g., various culture-based, molecular and microscopy were equally sensitive); and (2) data were not affected by characteristics of the water samples (e.g., we assumed that the water chemistry did not influence performance of the methods). In total, we collected 73 papers spanning over four decades of research from 25 countries: Argentina, Australia, Bolivia, Brazil, Canada, China, Cyprus, Democratic Republic of Congo, France, Greece, Germany, Hungary, Iceland, Italy, Japan, Luxembourg, Netherlands, New Zealand, Poland, Portugal, South Africa, Taiwan, United Kingdom, US, and Venezuela. The majority of studies were conducted in freshwaters (lakes, rivers and streams), followed by marine/brackish waters and swimming pools ( Figure 1). Since some studies were conducted in both fresh and marine/brackish waters, they were included in each water type in Figure 1. This resulted in a total of 126 observations (i.e., report on a relationship between indicator and pathogen) since some studies were conducted in both, marine and freshwater, and/or measured more than one type of indicator or pathogen. The majority of observations (n = 52) did not report any relationship between indicator(s) and pathogen(s), while those that did, were split into relationships that were statistically significant (n = 30) and those that were not (n = 44). Statistically significant relationships (or the lack thereof) and rationale for the observed trends are further examined in the following sections.

Relationships between Indicators and Pathogens in Recreational Waters
We queried studies conducted in marine, brackish, freshwater, and swimming pool waters meeting our search criteria for FIB, alternative indicators, MST markers, and pathogen data. FIB levels were typically reported as colony forming units (CFU), or most probable number (MPN), depending whether studies measured concentrations using membrane filtration on selective-differential media or defined substrate technology (e.g., Enterolert and Colilert), respectively (Tables 1 and 2). However, a few studies quantified general FIB using qPCR and expressed concentrations usually as gene copies per unit volume (Tables 1 and 2). Alternative indicators of fecal pollution such as C. perfringens and different bacteriophages [34] were also of interest for inclusion to assess their reliability for estimating pathogen presence compared to general FIB. C. perfringens was measured using membrane filtration on selective-differential media with concentrations expressed as CFU per unit volume, while bacteriophage concentrations were usually measured via single or double agar layer (SAL, DAL) techniques, with data expressed as plaque forming units (PFU) per unit volume (Table 3).
We also gathered data for various bacterial, viral, and protozoan pathogens. For bacterial pathogens, we collected data on 10 genera (Vibrio, Salmonella, Shigella, Mycobacteria, Pseudomonas, Escherichia, Aeromonas, Campylobacter, Legionella, and Listeria). Measurement strategies ranged from culture-based (data reported as CFU, MPN or presence/absence) to end-point PCR (presence/absence) and qPCR (gene copies) (Tables 1-4). For viral pathogens, we collected data on eight different species including enteroviruses, adenoviruses, noroviruses, hepatitis A and E, astroviruses, rotaviruses, reoviruses, and sapoviruses. Viral data were expressed as MPN (infectious viruses, ICC-[RT]-PCR), presence/absence (PCR) or gene copies (qPCR) (Tables 1-4). The most frequently measured protozoan pathogens, Cryptosporidium spp. and Giardia spp. (oo)cysts, were usually enumerated using immunomagnetic separation, followed by staining, although in some instances, qPCR was also performed (Tables 1-4). Enterocytozoon bieneusi was measured using similar detection methods to that of Cryptosporidium and Giardia (oo)cysts (Table 2), while two pathogenic amoeba species (Acanthamoeba spp. and Naegleria fowleri) were reported as presence/absence (i.e., PCR) (Table 1). Lastly, Candida spp. were enumerated using membrane filtration on selective-differential media and reported as CFU (Table 2). Sections below summarize our results regarding relationships that FIB, alternative indicators and MST markers have with pathogens and waterborne illness occurrence in freshwater, marine/brackish waters and swimming pools.
The frequency of significant relationships of FIB with bacterial or protozoan pathogens was similar; however, significant relationships with viral pathogens were less frequent (Fisher's exact test, p: 0.0005-0.0022). While lack of relationship between FIB and pathogenic viruses is not surprising given the enormous differences between the two groups in terms of persistence in the environment and low levels of viral pathogens typically found in ambient waters, correlations between FIB and protozoan pathogens are more difficult to understand. Interestingly, when Cryptosporidium and Giardia spp. were detected, they were generally present in higher concentrations compared to viral pathogens. Significant relationships were reported more commonly for rivers and streams compared to lakes [58][59][60]62], but this trend was not statistically significant (Fisher's exact test, two-tailed p = 0.085). Correlations appeared to be influenced by weather conditions, as most occurred during wet seasons and/or following rainfall events [53,61,63,67,75]. Not surprisingly, correlations were also more likely following sewage spills and/or wastewater discharges [54,56,57,61] and in waters impacted by agricultural operations [57,61], likely due to elevated FIB concentrations, greater likelihood of pathogen presence and the potential for location to be dominated by single fecal source. Significant correlation commonly observed in waters classified as "poor" and "sufficient" but also seen in waters classified as "good" or "excellent" Higher geometric mean of FIB in Salmonella spp. positive samples than in Salmonella spp. negative samples. Even though significant correlation was reported, Salmonella was also detected in water samples with "good" and "excellent" water quality.
[59] [47] Total coliforms e L. pneumophila d Puzih River and hot spring recreational areas, Taiwan Total coliforms and L. pneumophila significantly correlated.
L. pneumophila detected in > 90% of samples. L. pneumophila and total coliforms also correlated with turbidity.
The fraction of samples that contained an indicator when pathogen was detected was highest for the protozoan parasites.
Relationships dependent on season and site.

FIB and Pathogens in Marine and Brackish Water
Significant relationships with FIB were most commonly reported for Salmonella spp., followed by adenoviruses, and Campylobacter spp., Vibrio spp., S. aureus, and protozoan pathogens ( Table 2). The most significant relationships with pathogens were reported for enterococci (n = 11), followed by E. coli (n = 4), and fecal coliforms (n = 2). No statistically significant relationships were reported for total coliforms. Significant relationships were reported between enterococci and adenoviruses (n = 8), Salmonella spp. (n = 4), Cryptosporidium/Giardia (oo)cysts (n = 4), Campylobacter spp. (n = 3), and Candida spp., Vibrio spp., S. aureus, noroviruses, and E. bieneusi (one observation each) (Table 2). E. coli formed significant relationships with Salmonella spp. and Vibrio spp. (n = 1 each) and adenoviruses (n = 2). Statistically significant relationships with fecal coliforms were reported only for adenoviruses (28.6%, n = 7). The methodology employed did not appear to influence the outcome; significant relationships were not more likely when both indicator and pathogen were measured by a similar technique ( Table 2).
As expected, FIB had more significant relationships with bacterial pathogens compared to viral pathogens (Fisher's exact, p = 0.0069), but there was no significant difference in other comparisons (i.e., FIB relationships with bacterial compared to protozoan pathogens, or FIB relationships with protozoan compared to viral pathogens). Of note, FIB most likely to correlate with pathogens were enterococci, which supports its recommended use to monitor marine recreational water quality. No clear trend for different marine water types (e.g., brackish waters and coastal beaches) was observed with respect to statistically significant indicator/pathogen relationships [59,96,[98][99][100][101], suggesting that hydrological factors play less of a role compared to freshwaters. Similar to freshwater, the common trend among the studies reporting significant relationships was that they were conducted in waters impacted by fecal contamination [96,98,99], and when bather numbers were high [97], conditions likely to result in elevated FIB and pathogen levels.  Levels of FIB correlated with presence of Salmonella spp., especially in waters deemed "poor" or "sufficient" compared to "excellent".

Alternative Indicators and Pathogens in Marine, Brackish and Freshwater
Ten studies conducted in freshwater and fourteen studies conducted in brackish and marine waters measured at least one alternative indicator and one pathogen. In freshwater, four studies measured C. perfringens, five studies measured bacteriophage, and one study measured both (Table 3). In brackish/marine waters, the majority (n = 12) of studies measured coliphage (somatic, F-specific), followed by C. perfringens (n = 7), and phages infecting Bacteroides thetaiotaomicron (n = 1) ( Table 3). Similar to FIB, more statistically significant relationships were reported in freshwater compared to brackish/marine waters (Fisher's exact test, p = 0.0057). Please see Table 3 ("relationship" and "comments" columns) for summary of relationships and other comments regarding studies that found no significant relationship or those that did not report it.
In marine and brackish waters, approximately, half of studies (n = 6) did not report any statistical analysis [76,77,81,82,87,89], while three reported statistically significant relationships [79,90,100], and five reported non-significant relationships [64,88,91,92,94]. Two studies found significant relationships between F-specific coliphage and pathogens; one reported it with methicillin resistant S. aureus (MRSA), and S. aureus at a marine beach affected by fecal-impacted freshwater intrusion [79], while a second reported it for adenoviruses in water impacted by urban run-off [90] (Table 3). No studies noted a significant relationship between somatic coliphage and pathogens (Table 3). Only a single study conducted in Hawaii [100], a state that recommends using C. perfringens for monitoring ambient waters [104], found a relationship between this indicator, and two pathogens (Campylobacter spp., and V. parahaemolyticus) ( Table 3). The methodology employed did not appear to influence the outcome in marine or freshwaters; significant relationships were not more likely when both indicator and pathogen were measured by a similar technique (Table 3). While there were insufficient data to perform statistical analyses regarding relationship of alternative indicators and different pathogen groups, F-specific coliphage tended to perform better compared to somatic coliphage and C. perfringens. Higher statistical correlation observed between enteric viruses than between enteric viruses and coliphages. Noroviruses most abundant, followed by rotaviruses. Wet weather concentration of coliphage and viruses higher than dry weather concentration, but difference is not statistically significant. [

MST Markers and Pathogens in Marine, Brackish and Freshwater
The number of studies that measured MST marker(s) along with at least one pathogen is considerably smaller (n = 19; eight in freshwater, and 11 in brackish/marine/waters) compared to studies measuring FIB or alternative indicators ( Table 4). The majority of MST measurements reported were for human-associated marker(s) (76.1%), followed by general MST markers (7.0%), cattle and dog associated MST markers (5.6%), and seagull and swine-associated MST markers (2.8%) ( Table 4). Most frequently measured pathogens were viruses (adenovirus, enterovirus, noroviruses, hepatitis, and infectious enteric viruses) and bacteria (E. coli, Campylobacter spp., Salmonella spp., V. vulnificus, S. aureus, and Legionella spp.) with 22 measurements each, while Cryptosporidium and Giardia (oo)cysts were reported less frequently (n = 6) ( Table 4). Irrespective of the water type, nine of these studies did not report statistical analyses for relationships between MST marker(s) and pathogens [51,76,77,79,80,83,86,88,105], and another seven reported non-significant relationship [48,56,57,70,72,98,99]. The remaining two studies reported statistically significant relationship [93,106] (Table 4). Please see Table 4 ("relationship" and "comments" columns) for summary of relationships and other comments regarding studies that found no significant relationship or those that did not report it.
Significant relationships between pathogens and human-associated MST markers were reported for HF183 and adenoviruses, at a marine beach impacted by non-point source(s) [93], and between HF183/HF134 and Campylobacter spp. in freshwater affected by livestock operations [106]. In the same freshwater study, cattle-associated MST markers (CF128, CF193) correlated with E. coli O157:H7, and Salmonella spp., while a general Bacteroidales MST marker (Bac32F) correlated with all three pathogens [106]. The methodology employed did not appear to influence the outcome; in other words, significant relationships were not more likely when both indicator and pathogen were measured by a similar technique (Table 4). There were insufficient data regarding relationship of MST markers and different pathogen groups (e.g., bacterial, viral and protozoan) to perform statistical analyses. While it may seem counter-intuitive that MST markers (particularly human-associated subset), were not generally correlated with pathogens, it is important to note that sensitivity and specificity of MST markers varies greatly [14]. Furthermore, many pathogens reported in these studies are zoonotic, making this relationship even more tenuous.

Various Indicators and Pathogens in Swimming Pools
Our search of literature for paired measurements of indicator(s) and pathogen(s) recorded for swimming pools yielded considerably fewer studies (n = 3), compared to ambient waters. None of the studies reported statistical analyses on relationships between indicators and pathogens. Two studies, both conducted in Italy, were performed on pools that were in compliance with microbiological requirements for E. coli, enterococci, P. aeruginosa, and S. aureus [107,108]. However, one study detected infectious Simkania negevensis, a bacterium related to Chlamydia, in nearly 43% of samples, while the second one measured Papillomaviruses in 64% of samples [108]. Interestingly, HPyVs were co-detected with Papillomaviruses in all the samples [108]. L. pneumophila and enteric viruses (adenovirus, norovirus, and enteroviruses) [108] were not detected. Examination of wading pools in Finland during a gastroenteritis outbreak detected Norovirus GII and astrovirus in~83% and~33% of samples, respectively [109]. E. coli was absent from samples collected~2 weeks before the outbreak, but high concentrations (370-24,000 CFU/100 mL) were measured in two samples taken during the outbreak [109].

Relationship of Indicators with Illness
To identify associations between the presence of general FIB, alternative indicators or MST markers with that of waterborne illness occurrence, various epidemiologic studies were collected from existing literature dating back to the early 1990s. For inclusion, it was required that the study measured at least one FIB, alternative indicator or MST marker (culture or molecular) in combination with an epidemiological survey of resulting illness from the recreational water exposure. In total, 17 studies [76,79,86,[110][111][112][113][114][115][116][117][118][119][120][121][122][123][124] met these criteria and were included in analyses. One study each was conducted in Europe and Africa, and fifteen studies were conducted in the US. Since some of these studies were conducted in more than one water type, this resulted in the inclusion of 20 freshwater sites and 29 brackish/marine sites. Thirteen different microbiological assays were reported including those targeting: enterococci, fecal and total coliforms, E. coli, somatic and F+ coliphage, as well as various general and human-associated MST markers (Figure 2). In addition to gastrointestinal illnesses characterized by symptoms of diarrhea, vomiting, and stomach cramps, other waterborne illnesses included skin, ear and sinus infection [76,79,86,[110][111][112][113][115][116][117][118][119][120][121][122]124]. For epidemiological studies, assays targeting enterococci were the most commonly recorded, with 25 instances of measurements of either culture based or molecular enterococci targets, followed by human-associated MST markers, F+ coliphage, fecal coliforms, general MST markers, total coliforms, culturable E. coli, somatic coliphage and finally E. coli qPCR signal (Figure 2).

Various Indicators and Pathogens in Swimming Pools
Our search of literature for paired measurements of indicator(s) and pathogen(s) recorded for swimming pools yielded considerably fewer studies (n = 3), compared to ambient waters. None of the studies reported statistical analyses on relationships between indicators and pathogens. Two studies, both conducted in Italy, were performed on pools that were in compliance with microbiological requirements for E. coli, enterococci, P. aeruginosa, and S. aureus [107,108]. However, one study detected infectious Simkania negevensis, a bacterium related to Chlamydia, in nearly 43% of samples, while the second one measured Papillomaviruses in 64% of samples [108]. Interestingly, HPyVs were co-detected with Papillomaviruses in all the samples [108]. L. pneumophila and enteric viruses (adenovirus, norovirus, and enteroviruses) [108] were not detected. Examination of wading pools in Finland during a gastroenteritis outbreak detected Norovirus GII and astrovirus in ~83% and ~33% of samples, respectively [109]. E. coli was absent from samples collected ~2 weeks before the outbreak, but high concentrations (370-24,000 CFU/100 mL) were measured in two samples taken during the outbreak [109].

Relationship of Indicators with Illness
To identify associations between the presence of general FIB, alternative indicators or MST markers with that of waterborne illness occurrence, various epidemiologic studies were collected from existing literature dating back to the early 1990s. For inclusion, it was required that the study measured at least one FIB, alternative indicator or MST marker (culture or molecular) in combination with an epidemiological survey of resulting illness from the recreational water exposure. In total, 17 studies [76,79,86,[110][111][112][113][114][115][116][117][118][119][120][121][122][123][124] met these criteria and were included in analyses. One study each was conducted in Europe and Africa, and fifteen studies were conducted in the US. Since some of these studies were conducted in more than one water type, this resulted in the inclusion of 20 freshwater sites and 29 brackish/marine sites. Thirteen different microbiological assays were reported including those targeting: enterococci, fecal and total coliforms, E. coli, somatic and F+ coliphage, as well as various general and human-associated MST markers (Figure 2). In addition to gastrointestinal illnesses characterized by symptoms of diarrhea, vomiting, and stomach cramps, other waterborne illnesses included skin, ear and sinus infection [76,79,86,[110][111][112][113][115][116][117][118][119][120][121][122]124]. For epidemiological studies, assays targeting enterococci were the most commonly recorded, with 25 instances of measurements of either culture based or molecular enterococci targets, followed by humanassociated MST markers, F+ coliphage, fecal coliforms, general MST markers, total coliforms, culturable E. coli, somatic coliphage and finally E. coli qPCR signal (Figure 2).  Correlations between observed illness in these studies were most common with enterococci (10 studies out of 17) [79,86,110,111,113,114,116,117,120,121], followed by F+ coliphage (5 studies) [79,113,118,119,123] (Figure 2), suggesting that these two indicators may be better predictors of waterborne illness occurrence. Fecal coliforms, human-associated MST markers (Bsteri, BuniF2, and HF134), general MST marker (GenBac3), culturable E. coli, total coliforms, and somatic coliphage were correlated with illness less frequently (Figure 2). Twenty-seven indicator measurements across all studies were correlated with human illness, and 93% of these studies were conducted in waters with known point or non-point source contamination, contaminated surface/ground water flow or following wet weather events. Only six studies [79,86,[117][118][119]124], all of which found relationship between indicator and illness, measured pathogens, in addition to recording illness information, and indicator organism concentrations. Only one of the six studies found a relationship between pathogens and illness or indicator concentrations. This is not surprising since, in these studies, pathogens were detected infrequently and at low concentrations. This illustrates the potential challenges of detecting relationships between indicators and pathogens in the field even when health relationships were observed with fecal indictors.

Factors that Influence Indicator and Pathogen Relationships
Most recreational waters at any given time are impacted by many different sources of fecal contamination (e.g., treated and untreated wastewater, agricultural operations, stormwater, and domestic and wildlife animals) and these influences can change depending on many different factors including precipitation, tidal flow and wind direction. In addition, each fecal source has its own set of indicators and the potential for different types of pathogens. Therefore, the more fecal sources a recreational water is impacted by, the more challenging it will be to show correlations between indicators and pathogens. Preceding sections described our findings regarding relationships between indicators and pathogens in recreational waters, as well as relationships between indicators and illness. Overall, FIB were better predictors of bacterial and protozoan pathogen presence (compared to viral), relationships were more probable under scenarios where both indicator and pathogen were likely to be present at higher concentrations, and enterococci and F-specific coliphage tended to be better predictors of waterborne illness occurrence compared to other indicators. The following sections examine various factors that are likely to influence the observed trends.

Detection Frequency and Concentrations of Indicators and Pathogens in Marine, Brackish and Freshwaters
The observed relationships between indicators and pathogens can be influenced by logistical factors that may confound determination of actual relationships, including study design and methodological limitations. Study design determines the frequency at which a target (FIB, alternative indicator, MST marker or pathogen) was measured (per study or cumulative multiple studies), while methodology employed influences likelihood of detection. We compiled studies that reported frequency of detection (or data that allowed calculation of frequency detection such as total number of samples and samples positive) for at least one FIB/alternative indicator/MST marker and at least one pathogen per sample(s), resulting in inclusion of 49 studies (Table 5). Microbial data collected were first grouped according to indicator (FIB, alternative or MST) or pathogen type (bacterial, viral or protozoan), and further organized according to the detection format employed (different types of culture-based or molecular). Table 5 describes the detection frequency (per study and per total cumulative samples) for microorganism targets (FIB, alternative indicators, MST, and pathogens) for both freshwater and brackish/marine waters. Table 5. Frequency (%) of detection of microorganisms over all eligible studies (those that included data on individual observations). Detection frequency is expressed per study and for cumulative samples across all studies. Studies with least one sample positive for the organism were scored positive in the "per study" column.  Each FIB was detected at least once in 100% of studies, which was true for most of the microbial targets. General FIB were also the most frequently detected on a per sample basis, as they were found in 94.2% of samples across 13,823 measurements in marine and freshwaters (Table 5). In freshwater, detection frequency of FIB per sample was 95% across 11,920 measurements and it was somewhat lower in brackish/marine waters (93% across 2203 measurements) ( Table 5). Detection frequency of alternative indicators per sample, irrespective of the water type, was considerably less than FIB, averaging 60.6% (across 2606 measurements); the difference between water types was also more pronounced than for FIB (freshwater detection frequency 83.7%/2366 measurements vs. marine 45.1%/240 samples) ( Table 5). While the 2705 samples analyzed for MST markers in marine and freshwaters were similar to alternative indicators, the overall detection frequency average (42.9%) was considerably lower ( Table 5). The frequency of detection and total number of samples collected in each water type (37.4%/1355 in freshwater vs. 47.3%/1350 in marine water) was similar ( Table 5).
Irrespective of the water type, bacterial pathogens were measured more often (9280 total samples), compared to viral (3462) and protozoan (3400) pathogens, although the frequency of detection across different pathogen groups was similar (33.8%, bacterial; 29.6%, viral; and 28.9%, protozoan (Table 5)). There also appeared to be no appreciable difference in detection frequency between the water types for any of the pathogen groups, although considerably more samples were collected in freshwater (Table 5). For bacterial pathogens, frequency of detection across 8936 total samples collected in freshwater was 35.2%, compared to 30.3% in 344 marine/brackish water samples (Table 5). Similarly, viral pathogens were detected in 33.1% freshwater samples (out of 2952) and 23.6% of 510 marine water samples (Table 5). Lastly, protozoan pathogens were detected in 34.6% of 3,134 freshwater samples and 24.4% of 266 marine water samples (Table 5).
A subset of studies examined (n = 33) reported concentration data in the body of the manuscript, tables or supplemental materials, allowing graphs to be created displaying average densities per organism and water type (Figures 3 and 4). Concentrations of indicators were on average 1-3 log 10 higher than pathogen concentrations for both water types. Both indicators and pathogens in marine waters were found at slightly lower levels (0.5-1 log 10 ) than those observed in freshwater (Figures 3  and 4 and Table 6). Within an indicator group, concentrations of FIB ranged from not detected (ND (observed only for enterococci)) to 5.39 log 10 per 100 mL (Table 6), and total coliform levels were the highest, followed by fecal coliforms, E. coli, and enterococci (Figures 3 and 4). Alternative indicator concentrations were lower than FIB (ranging from ND-3.29 log 10 per 100 mL (Table 6)), and C. perfringens levels were higher than somatic and F-specific coliphage (Figures 3 and 4). MST marker concentrations were reported less frequently and were more variable, ranging from ND-2.50 log 10 copies per 100 mL (Table 6 and Figures 3 and 4). Bacterial pathogen concentrations (range: ND-5.09 log 10 per 100 mL) were higher than viral (range: ND-1.58 log 10 per 100 mL) and protozoan pathogens (range: ND-1.93 log 10 per 100 mL), in both marine and freshwater ( Table 6, Figures 3 and 4).   As evidenced by the examples above, readily detected microorganisms are more likely to be measured and frequency of detection of a given microorganism is influenced by the concentration and distribution of the target in the sample types tested, as well as the limit of detection of the method   As evidenced by the examples above, readily detected microorganisms are more likely to be measured and frequency of detection of a given microorganism is influenced by the concentration and distribution of the target in the sample types tested, as well as the limit of detection of the method used. Culture methods such as membrane filtration can have a low limit of detection, e.g. 1 CFU/100  As evidenced by the examples above, readily detected microorganisms are more likely to be measured and frequency of detection of a given microorganism is influenced by the concentration and distribution of the target in the sample types tested, as well as the limit of detection of the method used. Culture methods such as membrane filtration can have a low limit of detection, e.g., 1 CFU/100 mL, and can reliably detect FIB in water samples with minimal contamination. Conversely, pathogens are generally present sporadically and in lower levels than fecal indicators. These types of targets require high-throughput filtration methods that can achieve large concentration factors, with the tradeoff that limits of detection are generally quite high. In addition, the volumes sampled, and the concentration strategy used can vary between studies and can affect the sensitivity of a given method. These logistical factors frequently result in unbalanced comparisons in which the indicator organism is frequently detected, but the pathogen is not. Therefore, the disconnect between indicators and pathogens may not be due to a true lack of relationship in many cases, but to methodology that is much more suited to detecting indicators than pathogens.

Microbial Levels in Fecal Material
The observed relationships between indicators and pathogens can also be affected by factors intrinsic to the organisms themselves, including levels in various hosts, as well as shedding frequencies and duration. FIB are commensal inhabitants of the GI tract of humans and other animals and as such are shed continually in feces. Levels of fecal coliforms, E. coli, and enterococci typically found in human feces range 10 5 -10 9 CFU per gram [13], while levels detected in untreated wastewater are somewhat lower (10 5 -10 8 CFU per 100 mL) [13,128]. The concentration of FIB in animal excreta is lower still, ranging from 10 4 to 10 7 CFU per gram, depending on the animal host [13]. Alternative indicators are also commensal organisms of the GI tract but are typically found in lower concentrations and are more influenced by diets and physiologies of the host [13,14,128]. For example, C. perfringens levels in animal and human feces range from undetectable to 10 8 CFU per gram [13], while coliphages were absent from some animal feces and primary wastewater effluents [128,129] and typically did not exceed 10 3 PFU/mL of untreated wastewater [128]. MST markers target different fecal microorganisms that are strongly associated with particular hosts [14] and the human-associated subset is reported to range from 10 3 to 10 10 gene copies per gram of feces or 100 mL of untreated wastewater, while animal-associated MST markers range from 10 4 -10 9 gene copies per gram of feces depending on the sensitivity of individual markers and geographic region [13].
Pathogens may cause symptomatic or asymptomatic infection of their human and animal hosts. Shedding rates can vary widely, although levels found in the wastewater are typically several orders of magnitude lower than any indicator species [128,[130][131][132][133] likely due to the sporadic nature of pathogen occurrence and detection compared to indicators. Additionally, only a small part of the population is infected with pathogens at any given time, resulting in considerable variation in the levels of pathogens, particularly when originating from relatively small populations. Differential shedding of pathogens from infected hosts is also contributing to the occurrence of pathogens in recreational waters. For example, shedding rates for human viral pathogens can be as high as 10 11 viral particles per gram of feces in the case of adenoviruses [134], while shedding rates of bacterial pathogens are typically lower [133], as cattle excreting >10 4 CFU per gram of feces E. coli O157:H7 are considered to be "super-shedders" [135]. Cryptosporidium and Giardia (oo)cyst shedding rates by the infected individuals can range from 10 6 to 10 11 per gram of feces [132] and are typically higher in animal hosts compared to human [131,136], although not all (oo)cysts excreted by animals are zoonotic [131,137].
Shedding duration of viruses can vary from weeks to months [133], and some viruses display distinct seasonal trends (e.g., infectious enteroviruses are more prevalent in wastewater in summer and early fall) [134]. Similar to pathogenic viruses, excretion of (oo)cysts is typically long term [132]. Shedding duration of bacterial pathogens is shorter with median values typically reported to bẽ 2 weeks, although in some instances it can last considerably longer [138,139]. Similar to viral and protozoan pathogens, shedding is affected by many different factors including diet and age of the host [140,141], temperature [140], as well as composition of gut microbiome [142]. Infectious dose of different pathogens is also variable and typically the lowest for viruses [134,143], medium range for protozoan pathogens and generally highest for bacterial pathogens [143], although E. coli O157:H7, with a low infectious dose, is an exception [144]. The infectious dose of viral, bacterial and protozoan pathogens is dependent on many factors, including individual strains and health status of the host [134,143].

Susceptibility to Environmental Stressors
While wastewater treatment processes generally result in some removal of indicators and pathogens [128,145], sanitary sewer and combined sewer overflows, along with other infrastructure failures can result in release of indicators and pathogens into ambient waters. In addition, different indicator and pathogen groups exhibit variable susceptibilities to disinfection strategies. Bacteria are generally susceptible to chlorination and UV treatment [146]. Protozoa and viruses are typically most susceptible to UV treatment [146,147], with the notable exception of adenoviruses [148]. Once indicators and pathogens are released into ambient waters, a new panoply of biotic and abiotic environmental factors affects fate and transport characteristics, including ambient sunlight, indigenous microbiota (i.e., predation and competition interactions), temperature, salinity, nutrient levels, location (water column vs. sediment), source of fecal pollution and resilience of individual organisms.
Ambient sunlight and associated UV radiation typically act to increase the decay rates, although the magnitude of this effect is influenced by the environmental conditions [149] and measurement strategies [150,151]. For example, viable cells and culturable/infectious organisms typically display the effects of UV damage more readily than their corresponding nucleic acids. Interactions with indigenous microbiota also increase decay rates, although this was predominantly shown for FIB, MST markers and some bacterial pathogens, (e.g., [152][153][154][155]) with inconclusive data for other organisms (e.g., various bacteriophages and C. parvum [156,157]). Influx of nutrients (in the form of organic carbon, nitrogen and phosphorus) can result in extended persistence [158][159][160], and potentially mitigate the effects of biotic interactions [161] but this assertion was not tested in detail for organisms other than culturable FIB. Temperature and location affect decay rates of most organisms tested (e.g., FIB, bacteriophage, viral pathogens, MST markers) almost unilaterally with greater persistence at lower temperatures [150,[162][163][164][165][166] and in the sediments and sands compared to the water column (recently reviewed in [167]).
Similar to the effect of ambient sunlight, salinity (and the associated ionic content of brackish and marine waters) affected the decay rates of culturable/infectious FIB, alternative indicators, MST markers and pathogens more so than their corresponding nucleic acids [168][169][170][171][172][173][174]. The effect of source of fecal pollution has been studied on FIB and MST markers, and indicators originating from ruminants are more persistent compared to those from other fecal sources (e.g., dog, seagull, and human) [169,[175][176][177][178][179], although different human sources (e.g., feces, septage, and sewage) elicit different decay rates [152]; analogous information for alternative indicators and pathogens is still missing. Finally, studies that compared decay of various indicators to pathogens directly under the same experimental conditions are rare and report conflicting results. For instance, in one study, E. coli O157:H7 persisted longer than FIB (e.g., E. coli and enterococci) in freshwater [180], but another group reported similar trend in the freshwater sediments but not the water column [181]. Another group reported no difference in decay between FIB, various MST markers and C. jejuni, S. enterica and adenovirus in freshwater [151]. Others reported considerably faster decay of C. jejuni (but not C. coli or Salmonella spp.) than FIB and MST markers, irrespective of the water type [171]. As exemplified above, variable responses of different indicator and pathogen groups to these stressors and the resulting differential decay rates further confound the indicator paradigm.

Conclusions
FIB and alternative indicator organisms (C. perfringens and coliphages) have been used for over a century, and continue to be used today as indicators of general fecal pollution in many applications, including the assessment of sanitary quality of recreational waters [8]. MST markers are used to identify source(s) of fecal pollution and are a more recent addition to the monitoring toolbox available to water quality managers and other practitioners in the field [14]. The goal of this review is to two-fold. Our primary objective was to examine reported relationships between various indicators and pathogens in recreational waters to determine the value of different indicators as surrogates for pathogen presence. Secondly, we aimed to more closely inspect different factors that may have an impact on this relationship.
The majority of the studies either did not report a relationship, or they reported a statistically non-significant relationship. Among the studies that observed statistically significant relationships, it was considerably more common in freshwater compared to marine waters. General FIB tended to form statistically significant relationships more commonly with bacterial and protozoan pathogens, (compared to viral pathogens) and this difference was statistically significant. Alternative indicators and MST markers correlated with pathogens less frequently, although it occurred more in freshwater than marine/brackish waters. Overall, statistically significant relationships were detected more frequently in waters known to be impacted by fecal pollution and following wet weather events, both scenarios under which indicators and pathogens are more likely to be co-detected.
Among factors influencing these relationships frequency of detection and variable concentrations of indicators and pathogens were identified as major contributing factor. Not surprisingly, general FIB were measured and detected more frequently than any other indicator or pathogen (generally in >90% of samples) and were also reported at higher concentrations, irrespective of the water type. Alternative indicators were also frequently detected in samples (>70%), while MST markers were measured and detected less frequently, and in lower concentrations than FIB or alternative markers (frequently in <10% of samples). Pathogen detection frequencies were similarly low. Low frequency of detection affects the ability to establish relationships between the frequently-detected and infrequently-detected analytes, as the dataset becomes left-censored (biased toward non-detects values). What looks like "absence" is frequently an artifact of comparing an analyte with high density (e.g., FIB) with one of low density (e.g., pathogen). Better concentration and recovery methods for the infrequently-detected analytes may provide a more realistic picture of the relationships among these various microorganisms in environmental waters. Finally, concentrations in feces and wastewater, shedding rates and patterns of various indicator and pathogen groups differ, as do their fate and transport characteristics in secondary habitats. Indicators are typically present in higher concentrations than any of the pathogen groups, and are also shed constantly, or more frequently, compared to pathogens. Upon entry into the secondary habitats, a host of biotic and abiotic factors differentially affects persistence of indicators and pathogens, further confounding the indicator-pathogen paradigm. Lastly, another important factor impacting the ability to establish relationships between indicators and pathogens is the realization that most locations are impacted by multiple sources of fecal contamination. Although it is difficult to measure the impact of multiple fecal inputs, tools such as sanitary surveys and GIS mapping have the ability to indicate potential point and non-point sources of fecal pollution and future MST studies should improve our understanding of the impacts of multiple fecal sources.
To further our understanding of indicator and pathogen relationships, future studies measuring these microorganisms in recreational waters should evaluate and report the existence (or lack thereof) of such relationships. Other considerations include careful selection of targeted pathogens and methodology used to quantify them. Furthermore, providing the data on a per sample basis (rather than descriptive statistics of a dataset) in at least supplementary materials, will enable metanalyses, which may yield a more robust estimate of a true state of indicator/pathogen paradigm. Lastly, while standardized and sensitive methods exist for FIB detection and enumeration in recreational waters, analogous procedures for alternative indicators, MST markers and pathogens are still missing. Standardization of detection and quantification methods suitable for each indicator/pathogen group can enable more accurate evaluation of any statistically significant relationships between these two groups.

Conflicts of Interest:
The authors declare no conflict of interest.

Disclaimer:
The United States Environmental Protection Agency through its Office of Research and Development funded and managed the research described here. It has been subjected to Agency's administrative review and approved for publication. The views expressed in this article are those of the author(s) and do not necessarily represent the views or policies of the U.S. Environmental Protection Agency. Mention of trade names or commercial products does not constitute endorsement or recommendation for use.