|Home | About | Journals | Submit | Contact Us | Français|
The release of untreated sewage introduces non-indigenous microbial populations of uncertain composition into surface waters. We used massively parallel 454 sequencing of hypervariable regions in rRNA genes to profile microbial communities from eight untreated sewage influent samples of two wastewater treatment plants (WWTP) in metropolitan Milwaukee. The sewage profiles included a discernable human fecal signature made up of several taxonomic groups including multiple Bifidobacteriaceae, Coriobacteriaceae, Bacteroidaceae, Lachnospiraceae, and Ruminococcaceae genera. The fecal signature made up a small fraction of the taxa present in sewage but the relative abundance of these sequence tags mirrored the population structures of human fecal samples. These genera were much more prevalent in the sewage influent than standard indicators species. High-abundance sequences from taxonomic groups within the Beta- and Gammaproteobacteria dominated the sewage samples but occurred at very low levels in fecal and surface water samples, suggesting that these organisms proliferate within the sewer system. Samples from Jones Island (JI – servicing residential plus a combined sewer system) and South Shore (SS – servicing a residential area) WWTPs had very consistent community profiles, with greater similarity between WWTPs on a given collection day than the same plant collected on different days. Rainfall increased influent flows at SS and JI WWTPs, and this corresponded to greater diversity in the community at both plants. Overall, the sewer system appears to be a defined environment with both infiltration of rainwater and stormwater inputs modulating community composition. Microbial sewage communities represent a combination of inputs from human fecal microbes and enrichment of specific microbes from the environment to form a unique population structure.
Sewage overflows release human pathogens into surface waters, creating a serious public health risk and environmental concern. Older cities in the US, particularly in the northeast and Great Lakes region, have combined sewer systems that collect both wastewater and stormwater. These systems can exceed capacity following heavy rainfall events and subsequently overflow to surface waters. The USEPA estimates that communities with combined sewers annually release more than 3 trillion liters of sewage into the nation’s waterways (USEPA, 2004). Human exposure to sewage contaminated water can occur through drinking water or recreational activities. Surveillance for fecal pollution relies upon the detection of indicator organisms such as Clostridium perfringens, enterococci, fecal coliforms, or Escherichia coli (USEPA, 2000; Fujioka, 2002; NRC, 2004). Selective culturing methods have made these assessments simple, relatively fast (1–2 days) and inexpensive (Fujioka, 2002). Epidemiological studies have shown a correlation between gastrointestinal illness and elevated fecal indicator organism levels, but sometimes the relationship between indicators and disease is poor or nonexistent (Dufour, 1984; Haile et al., 1999; Wade et al., 2006; Colford et al., 2007). Fecal indicator bacteria can enter surface waters from agricultural runoff, urban stormwater, wildlife, and other non-human sources (Reeves et al., 2004; Shanks et al., 2006; Surbeck et al., 2006; Vereen et al., 2007; Lee et al., 2008). Discerning the source of pollution is important for evaluating health risk and formulating mitigation strategies, and is an ongoing challenge facing water resource managers and public health officials (Field and Samadpour, 2007; Stewart et al., 2008).
Alternative host-specific indicators might have the capacity to differentiate sewage from other sources of fecal pollution and provide a more accurate assessment of human health risk due to pathogens (Dick et al., 2005; Savichtcheva and Okabe, 2006; Guzman et al., 2007). Molecular-based methods have identified new microorganisms that can serve as host-specific indicators of fecal pollution including certain members of the order Bacteroidales, Methanobrevibacter smithii, and more recently Faecalibacterium sp. (Bernhard and Field, 2000a; Carson et al., 2005; Dick et al., 2005; Ufnar et al., 2006; Lamendella et al., 2007; Lee et al., 2008; Layton et al., 2009; Zheng et al., 2009). Yet, these taxa represent only a small fraction of the diversity in human fecal microbial communities or in other host sources of fecal contamination. Published reports describing complex sewage communities have focused on microbial population structures of sludge and pilot scale bioreactors (Chouari et al., 2005; Franklin and Mills, 2006) without performing detailed surveys of sewage influent entering treatment plants. Anticipated differences between raw sewage and activated sludge (Liu et al., 2007) will likely reflect shifts in microbial community composition from original influent through aeration and anaerobic digestion processes. Detailed inventories of microbial populations in raw sewage have the potential to improve water management activities through the identification of novel indicator taxa for human fecal contamination.
Culture-independent, molecular surveys of PCR amplicons for ribosomal RNA (rRNA) genes have documented considerable microbial diversity in the human gut (Ley et al., 2006a). Some of these microbes might serve as alternative indicator organisms. The most comprehensive studies of nearly full-length rRNA genes from human intestinal microbial flora report 13,000–18,000 sequences (Eckburg et al., 2005; Ley et al., 2006b), with one study noting great differences between microbial communities of mucosal surfaces and feces. Several smaller surveys of gut flora describe correlations with disease states (Hopkins et al., 2001; Zoetendal et al., 2002; Bibiloni et al., 2006; Eckburg and Relman, 2007; Frank et al., 2007), the establishment of gut flora (Favier et al., 2003; Magne et al., 2006), and effects of antibiotics (Shoemaker et al., 2001). Person to person and temporal variability appear to be great (Eckburg et al., 2005; Turnbaugh et al., 2009). When these communities leave the gut, the members will vary in their ability to survive and persist because of differences in cellular physiology (e.g. ability to adapt to low nutrients, temperature, low iron, etc.) and physical cell structures (e.g. spore formers vs. gram negative organisms). The selective persistence of specific subpopulations from fecal-associated microbial communities may exert a significant influence on sewage microbial population structures including shifts in composition of high and low abundance taxa.
To our knowledge, there are no reports of large molecular surveys for microbial communities from untreated human sewage. Conventional DNA sequencing of a few thousand 16S rRNA gene PCR amplicons in an environmental DNA sample will generally fail to detect sequences from most of the low-abundance organisms in a sewage microbial community, some of which correspond to human fecal populations. In contrast, massively-parallel pyrosequencing (MPSS) technology (Margulies et al., 2005) can generate hundreds of thousands of short sequences (MPSS tags) from hypervariable regions in rRNA genes (Sogin et al., 2006; Huber et al., 2007) which increases the molecular sampling effort required to detect low abundance taxa. These tag sequences serve as proxies for individual operational taxonomic units (OTUs). The frequency of different rRNA gene tags provides a first-order description of the relative abundance (i.e. evenness) of phylotypes in a population. The variable nature of the MPSS tags and paucity of positions do not allow direct inference of phylogenetic frameworks. However, matches of tag sequences to a comprehensive reference database of hypervariable regions from full-length sequences of known phylotypes provide information about taxonomic identity and novel microbial diversity (i.e. richness) (Huse et al., 2008). The technology has successfully documented shifts in human and mouse fecal associated microbial population structures in response to antibiotic treatment (Dethlefsen et al., 2008; Antonopoulos et al., 2009) and differences in microbial population structures between lean and obese twins and their mothers (Turnbaugh et al., 2009). In this study we examined the microbial community at high resolution for complex sewage communities. Our study site represents a typical urban area. Two WWTPs service the Metropolitan Milwaukee area. One WWTP includes inputs from combined sewers, which collects both sanitary sewage and stormwater. The second WWTP primarily receives inputs from separated sewers that contain only sanitary sewage; stormwater discharges directly to rivers (untreated) in a separate system. Our primary objective is to gain insights about the composition of low-abundance and dominant populations in microbial communities released into the environment as a result of sewage overflows and to identify a suite of new alternative fecal indicators.
We obtained WWTP influent samples of the two major facilities servicing Metropolitan Milwaukee, WI on multiple dates in 2005 and 2007 (Table 1). JI WWTP, located in downtown Milwaukee, receives sewage from the northern sector of the metropolitan area. The service area includes approximately 50 km2 of combined sewers that consist of both sanitary sewage and stormwater from Milwaukee’s densest urban area. SS WWTP services primarily residential areas of the metropolitan area. The surface water sample was taken from the junction of three major rivers leading to Lake Michigan (e.g. the estuary). These waters receive sewage overflows when they occur, but sample collection took place during base-flow conditions when minimal fecal pollution occurs.
Sequencing of WWTP influent samples yielded between 17,338 and 34,080 V6 region sequence tags for each sample (Table 1) with a total of 215,090 sequences (17,310 unique tags) recovered from all eight samples. The analysis of surface waters provided 38,368 sequences (4,603 unique tags). We compared these datasets to a total of 1,229,403 sequences (38,264 unique tags) collected from human fecal samples isolated from five individuals before, during and after treatment with antibiotics as well as samples from 11 families of lean and obese twins and their mothers (Dethlefsen et al., 2008; Turnbaugh et al., 2009). We used the GAST algorithm to assign taxonomy to individual tags (Huse et al., 2008). The human and surface water samples exhibited very different taxonomic compositions (Fig. 1). Proteobacteria, specifically those that normally occur in the environment, dominated the sewage samples (59% of the total number of tags). Gammaproteobacteria accounted for 38% of the tags while 15% and 4% resolved to Beta- and Epsilonproteobacteria. Members of Actinobacteria, Bacteroidetes, and Firmicutes made up 37.5% of the sewage dataset, collectively (Fig.1), but accounted for 97% of the tags in the human fecal samples.
The sewage and human fecal samples contained many different Firmicute populations. Nearly 54% of the Firmicute tags in sewage samples mapped to Clostridia and 43% to Bacilli. In contrast, Clostridia tags comprised 98% of the Firmicutes in human fecal samples. The Bacteroidia accounted for the majority of the Bacteroidetes in sewage (54%), but we also detected Flavobacteria, Sphingobacteria, and several tags that did not resolve beyond the class level. Bacteroidia completely dominated the human fecal samples and accounted for 97% of the Bacteroidetes). The two dominant Actinobacteria families in the human fecal communities (Bifidobacteriaceae and Coriobacteriaceae) represented only 1.2% of the sewage tags.
The surface waters exhibited a taxonomic profile that differed from both human fecal and sewage samples. Betaproteobacteria (51%) and Alphaproteobacteria (19%) dominated the surface water microbial community. Actinobacteria of non-human origins and Bacteroidetes, mostly Sphingobacteria, accounted for 5.4% and 7.3% of the tags, respectively. Fewer than 0.5% of the surface water tags corresponded to Firmicutes.
To estimate the diversity of the surface water, human fecal, and a pooled dataset of all sewage influent samples, we identified OTUs corresponding to 3 and 6 percent sequence divergence. The combined sewage and human fecal datasets were randomly subsampled to the size of the smallest dataset (i.e. surface water) to generate subsets of equivalent size (~38,000 tags). After using MUSCLE (Edgar, 2004) to align tags from these subsets, DOTUR (Schloss and Handelsman, 2005) identified clusters using the furthest neighbor algorithm that allowed 3% and 6% sequence variation and calculated a variety of diversity indices. We identified a total of 3045, 1950, and 2833 OTUs (3% DOTUR clusters) in sewage, human fecal, and surface water, respectively. Sewage had the highest richness, followed closely by surface water at both the 3% and 6% sequence variation criteria (Fig. 2). Human fecal samples displayed considerably lower richness, which is consistent with the assessments of overall taxonomic composition; sewage contains taxa common to humans in addition to many taxa of environmental origin. The ACE value at the 3% criteria for sewage was 5014 and 4946 for surface waters compared with 3081 for human fecal samples. Chao values were similar, with sewage, human fecal, and surface water estimated as 4911, 2908, and 4891, respectively.
The Venn diagram reports the number of shared OTUs among the different sample types (Fig. 3). The majority of OTUs (84%) represented community members that were specific to only one of the sewage, human fecal, or freshwater microbial communities. As expected, the surface water and human fecal communities had few OTUs in common (i.e. 27 shared OTUs). In contrast, the overlap for human fecal and sewage influent included 611 shared OTUs while sewage and surface water had 479 OTUs in common. The number of tags observed within the shared OTUs was disproportionate for the different microbial communities. For example, 89% of the human fecal tags occurred within the 611 OTUs shared with sewage, whereas only 31% of the sewage tags were within these shared OTUs (Table 2). Microbes of environmental origin account for the balance of the sewage tags. The sewage and human shared OTUs may include candidates for new indicator organisms.
We also examined species richness endemic to individual sewage samples. ACE richness estimates for individual samples ranged from 1878 to 3709 (Table 1), which is considerably lower than estimates for the composite dataset of 5014. Table 2 summarizes the OTUs shared among all samples, and among a subset of samples. The majority of the tag dataset (71%) occurred within the OTUs common to all eight samples. The sewage OTUs that overlapped with the human dataset generally appeared in seven or eight of the sewage samples. OTUs that occurred in one sample at a very low abundance accounted for a large proportion of the richness (e.g., >50%), but a very minor portion of the tags.
Many of the tag sequences from sewage, fecal, and surface water samples can provide characteristic profiles describing the relative abundance for each distinct microbial population. By using the human fecal tag profile as a reference, the distribution of individual tags coupled to their taxonomic assignments can serve as a signature for sewage-derived fecal pollution. Comparisons of taxa assigned by GAST show there were 20 dominant human taxa that accounted for 94% of the human tags shared with sewage (Table 3), with other lower-abundance taxa accounting for the remaining overlap. The 20 most dominant taxa in sewage include tag sequences for the human-associated taxa Bacteroides, Lachnospiraceae, and Faecalibacterium. Bifidobacterium was also discernable in sewage profiles (Table 4). However, a greater proportion of the sewage tags belonged to taxonomic groups within Gammaproteobacteria (i.e. Acinetobacter, Aeromonas, and Pseudomonas) in addition to Arcobacter (Epsilonproteobacteria), and Trichococcus (Bacilli). Tag sequences for these taxonomic groups occurred at very low frequencies in surface waters.
There were a total of 57,020 unique V6 tags in the combined datasets from sewage, human and surface waters. We examined the taxonomic profiles for the 4000 most abundant V6 tags (e.g. at the sequence level) in all samples combined. Fig. 4 shows tag profiles for human taxa that are prominent in sewage. Each panel represents 401 distinct tag sequences initially sorted by taxonomic assignment and then by decreasing tag frequency. The examination of the 4000 most abundant tag sequences captured the similarities and differences in the tag profiles for Bacteroidales (Fig. 4a) and Firmicutes (Fig. 4b and c). For the entire dataset of 57,020 unique tags, 3172 resolved to Bacteroides with the majority of these representing low abundance taxa in sewage, humans or surface water. Eighty-eight unique Bacteroides sequences accounted for nearly 85% of the total Bacteroides tags (shown in Fig. 4a). The remaining 15% of the Bacteroides tags distributed evenly at low levels across ~3100 Bacteroides tag sequences (data not shown). Lachnospiraceae and Ruminococcaceae, two other families of interest, had similar numbers of unique tags and distributions.
There were similarities in the relative abundance of tag sequences for specific populations within Bacteroidales and the Firmicutes in the human fecal and all of sewage samples. Many of these tag sequences were absent from surface water and therefore may represent good candidates for alternative indicator organisms. The eight WWTP samples had very similar abundance profiles of human-associated taxa (e.g. Lachnospiraceae in Fig. 4c) as well as taxa of environmental origin (e.g. Flavobacteriaceae in Fig. 4a). Human fecal and sewage microbial communities shared many high-abundance Bacteroides tags (Fig. 4a), including some that resolved to the species level, e.g. B. fragilis, B. massiliensis, and B. ovatus. We also observed a number of low-abundance, shared sequences from fecal and sewage samples but many did not resolve beyond the order level. Within the family Porphyromonadaceae, the surface waters contained some low-abundance tags that mapped to Paludibacter, while a single high-abundance Paludibacter tag occurred only in sewage (Fig. 4a taxa designation 215). Not all Bacteroidales tags in sewage appeared in human fecal samples. For example, several high-abundance Parabacteroides tags (Fig. 4a taxa designations 241–247) appeared in sewage but only one also occurred in high abundance in humans (Fig. 4a taxa designation 243), and other Parabacteroides tags were absent or in low levels. Similarly, sewage had multiple tags that mapped to Prevotella (Prevotellaceae) whereas humans contained a single example. The presence of tags from several closely related Bacteroidales taxa in sewage and their absence in humans suggest sewage contains non-human fecal sources or there may be geographic specificity of human derived Bacteroidales within these taxa.
The Firmicutes provided the richest profile of tags that were common to humans and sewage (Fig. 4b and c). Within Clostridiales, few tags belonged to Clostridiaceae. Lachnospiraceae tags were the most common; the majority of these resolved only to the family level. The genera Dorea, Lachnospira, and Roseburia all contained high abundance tags common to human and sewage. There were a few examples of low abundance tags (e.g. <200 in the normalized dataset) within Dorea that were absent in sewage. Tags within the family Ruminococcaceae followed a similar pattern; many tags resolved to the family level. Faecalibacterium, Ruminococcus, and Subdoligranulum tags were predominate features in humans and proportionally reflected in sewage.
In examining the V6 tags recovered from eight WWTP samples, we rarely encountered the standard fecal indicator organisms. Overall, only two tags out of the entire dataset of WWTP samples (215,090 tags) resolved to E. coli. Only three tags mapped to Enterococcus faecalis (n=1) or E. faecium (n=2) and no tags matched C. perfringens. A greater number of V6 tags matched at least one reference sequence from an indicator organism, but they resolved to higher taxonomic levels (see Methods on annotation criteria). Even when considering tags that resolved to more general taxonomic designations, standard indicators rarely occurred in the WWTP samples.
The relative abundance of standard indicators was also determined using cultivation plus direct enumeration of total cell numbers. The total cell counts for sewage were approximately six orders of magnitude greater than culturable E. coli and Enterococcus levels, and nearly seven orders of magnitude greater than C. perfringens levels (Table 5). Overall, determinations made by culture combined with direct counts suggested that indicators were even more rare than tag sequencing would suggest. Several factors could contribute to this variance. E. coli and enterococci (E. faecalis and E. faecium) have seven and four copies of the 16S rRNA gene respectively, which is comparable to the 4.15 average for the 904 Bacterial genomes (Lee et al., 2009) and would not account for the higher estimation of relative abundance using the tag sequencing compared with the direct counts plus culture methods. Enumeration of indicator organisms by cultivation will not detect the presence of non-viable cells, but direct staining and tag sequencing will detect these cells, which most likely accounts for the estimates of relatively low abundance of indicator organisms using direct staining plus culture methods. In addition, total cell counts will include Archaea that molecular surveys of Bacteria may not detect, which would further lower the ratio of cultured indicator cells compared to direct counts. Overall, these results demonstrate that standard fecal indicators are very minor populations in sewage.
Analysis of eight WWTP samples over a two and a half year period revealed that the microbial community composition is very stable. The same human taxa dominated each of the eight samples (Fig. 3). The profiles of individual tag sequences for the eight samples were also very similar (Fig. 4). Comparisons of microbial profiles show greater similarity amongst samples collected at the two treatment plants on the same day than samples from the same WWTP collected on different days (Fig. 5). The relative proportion of human-associated taxa in individual samples ranged from 13% to 28% of the total tags. There was no significant difference between SS and JI WWTPs, despite the fact that SS receives primarily sanitary sewage and JI receives both sanitary sewage and stormwater (e.g. combined sewage). Further, rain events did not correlate with changes in the proportion of human-associated taxa at either WWTP. However, on one date (21 Aug 2007) with 3.38 cm of rainfall, influent flow increased at both WWTPs, and this corresponded to higher diversity in the JI and SS communities (Table 1). This suggests that rainfall modulates the community throughout the sewer system and may have an equal or even greater influence on composition than direct stormwater inputs into the combined sewer system.
Sewage overflows release a complex array of microorganisms into the environment, including pathogenic bacteria, viruses, and protozoa. The presence of fecal bacteria is a major cause of water quality impairment in the nation’s waterways and coastal regions (USEPA, 2006; Dorfman and Rosselot, 2008; Stewart et al., 2008; USEPA, 2009); however, the source of fecal pollution is unknown in the majority of cases. Fecal pollution may be derived from humans, agricultural or domestic animals, or wildlife. It is necessary to differentiate between human and other sources of fecal contamination for accurate assessments of risk to public health and prioritization of mitigation efforts.
Standard water quality measurements rely upon culture of bacteria and are not specific to a host source. These culturing methods rely upon biochemical reactions on media such as modified membrane-Thermotolerant Escherichia coli (modified m-TEC) agar (USEPA, 2002), membrane-Enterococcus Indoxyl-β-D-Glucoside (mEI) agar (USEPA, 2000), and membrane-Clostridium perfringens (mCP) agar (Emerson and Cabelli, 1982). Culture methods limit the range of organisms that can serve as indicators, and this study illustrates that fecal indicators represent minor populations in water quality assessments that have major regulatory and economic consequences. More recently, molecular techniques have expanded the range of organisms that can be used for fecal indicators (Field et al., 2003), however, these methods are based upon the presence or absence of a single target, and have been developed without a comprehensive view of the complex microbial community associated with different fecal pollution sources.
Monitoring for multiple organisms that are in higher abundance and specific to human fecal contamination would greatly enhance water quality assessment capabilities. MPSS allowed the sequencing of many thousands of PCR amplicons for homologous hypervariable rRNA coding regions from a wide range of taxa in sewage and surface water (~230,000 tag sequences), which were compared to human fecal samples (~1.2M tag sequences). We were able to describe radically different microbial communities in humans, sewage and surface water and identify specific taxa shared among sewage and humans. Within named taxa, we observed further resolution by cataloguing the collection of sequences that mapped to the same taxonomic levels. For example, sewage contained between 2000–3000 unique V6 sequences for the taxonomic groups Bacteroides, Lachnospiraceae, and Faecalibacterium (Table 3); however, fewer than twenty unique V6 sequences accounted for the majority of tags for each of these taxonomic groups. This illustrates the high resolution that can be achieved with MPSS. The occurrence of shared sequences, their taxonomic identity, and their relative abundance in sewage from a large metropolitan area and in human fecal samples provides a complex signature of human fecal pollution.
The complex signatures determined by MPSS contained few tags that corresponded to traditional indicator organisms. The reference database used for V6 tag annotation contained several distinct sequences each for E. coli, E. faecalis, E. faecium, and C. perfringens, as well as other sequences that do not differentiate between genera within a family. For example, one V6 tag sequence matched reference sequences for various genera of Enterobacteriaceae (primarily Klebsiella and Shigella) as well as of E. coli. Even if we included these less specific V6 sequences, indicator organisms were extremely rare in the sewage samples (<0.7%) of the total tags). This compares to cultivation experiments that detected ~ 2 indicator organisms/million total cells.
The Bacteroidales and Clostridiales groups exhibited high levels of diversity in both the sewage and human fecal communities. Previous reports have proposed Bacteroides as an alternative indicator (Kreader, 1995) and characterization of the broader group of Bacteroidales in multiple hosts has identified phylotypes that demonstrate host specificity for humans or other animals (Dick et al., 2005; Layton et al., 2006; Okabe et al., 2007; Lamendella et al., 2007). Bernhard and Field first described a human-specific Bacteroides marker (Bernhard and Field, 2000b), which has been used in numerous field studies to detect human sources of fecal pollution (Bernhard and Field, 2000b; Bower et al., 2005; Okabe et al., 2007; Santoro and Boehm, 2007). However, there is no clear host-specific structure across the Bacteroidales group as a whole; Bacteroidales from different host animals are found within the same clade (Dick et al., 2005; Lamendella et al., 2007; Jeter et al., 2009). These assessments of Bacteroidales populations have relied upon comparisons between a relatively small number (97–1200) of rRNA sequences (Dick et al., 2005; Lamendella et al., 2007; Jeter et al., 2009). Other studies have identified a Faecalibacterium phylotype using subtractive hybridization that appears to be specific to humans (Zheng et al., 2009). All of these studies have focused on identification of a marker that can be used to detect the presence of fecal pollution from a human or animal source.
Our approach is more comprehensive than identifying a single marker. We detected thousands of Bacteroidales; however, a relatively small subset occurred at a high frequency (e.g. 84% of the Bacteroides were represented in the 88 Bacteroides tags shown in Fig. 4a). We found relative abundance patterns of these tags in sewage mirrored that of human fecal communities. Taxonomic group profiles within different hosts may contain unique signatures, and these might serve as complex indicators of fecal pollution versus reliance on the presence or absence of a single phylotype. Evaluating host sources by detecting signatures based on multiple taxa (which may co-occur because of their presence in the host) and their relative abundance would not be influenced by transient colonization, or low levels of cross colonization to other hosts. Utilization of a single marker requires a high level of host specificity, and any cross over between hosts confounds assessments. A similar approach to discriminating fecal pollution from humans, pigs, and cattle has utilized qPCR to detect the relative abundance of dominant bacterial groups: Bacteroides/Prevotella, Clostridium coccoides, and Bifidobacterium (Furet et al., 2009), however the qPCR experimental design did not target specific populations within these groups. Employing community profiling by 454 sequencing enables high resolution within each of these taxa of interest (Fig. 4).
A large majority of the sewage tags were not shared with human fecal material and belonged to taxonomic groups that are generally not considered human comensual organisms. This finding is not surprising considering the nature of the sewer environment, which contains high nutrient levels in an aqueous state. Previous studies have shown that this environment is rich in biofilms (Chen et al., 2003; Leung et al., 2005; Ort and Gujer, 2008), which may act as a reservoir for resident organisms. Several taxa, including Acinetobacter, Aeromonas, Pseudomonas, and Trichococcus, were highly abundant in sewage, but not present in surface water, or only present at very low levels, suggesting these organisms may be specific to the sewer environment. Only one or two specific V6 sequences represented each of these high abundance taxa, which further suggests that these organisms comprise a relatively narrow population that could be propagating within the sewer system itself. The two WWTPs are relatively isolated from one another, so the observation of the same high abundance environmental organisms in JI and SS WWTP samples on the same day and over time would indicate that the sewer system behaves as a defined environment and the sewage community is subjected to a common ecological forcing throughout the system.
Alternative indicators may also include taxa that may be resident organisms growing in the sewer system, but not associated with human fecal material. Acinetobacter, Aeromonas, and Trichococcus might be useful adjuncts to the fecal specific organisms since they occur in very high abundance. Both Aeromonas and Arcobacter have been associated with sewage-contaminated water (Collado et al., 2008; Khan et al., 2009). Concurrent detection of human signature organisms with the sewage specific organisms would provide an additional level of confidence for assessing sources of fecal pollution. Further work is necessary to determine if these high abundance taxa are specific to the sewer system, or if they propagate in other environments, particularly those associated with fecal pollution such as contaminated soils, sediments, or beach sand. Some of these environments also appear to support the growth of indicator bacteria (Alm et al., 2003; Whitman and Nevers, 2003; Byappanahalli and Fujioka, 2004; Ishii et al., 2006).
There is a high potential to be able to track signature taxa from sewage in surface water during contaminant events. The surface water we analyzed was from the junction the three main rivers (freshwater estuary) that lead to Lake Michigan, which serves as the receiving waters for sewage overflows in the Milwaukee area. The freshwater estuary system is highly diverse and contains microorganisms originating from the rivers, as well as Lake Michigan (Mueller-Spitz et al., 2009). We found very little overlap between sewage and the estuary surface water community profiles. Taxa found in humans were not detected. The taxa that appeared to be derived from the sewage environment were nearly absent in surface water. Sewage and surface water both contained Betaproteobacteria, primarily members of Comamonadaceae, but the genera composition was completely different between the two sample types: sewage contained Acidovorax, Simplicispira, and Polaromonas while surface water contained Curvibacter, Malikia, and Variovorax.
The JI WWTP services both separated and combined sewer systems, therefore, sewage influent contains human fecal microbial populations from households and non-human fecal pollution from stormwater runoff during rain events. The SS WWTP services primarily separated sewer systems and its influent contains fecal populations mainly from human sources. We hypothesized there would be significant differences between the JI and SS microbial communities and that the variance would increase during rain events when non-human fecal material might enter the JI system. Surprisingly, WWTP samples were most similar on the same collection date, and rainfall did not affect the similarity. This is surprising given that very large amounts of stormwater entered the combined sewer system on 21 Aug 2007; stormwater from 50 km2 of an urbanized area is conveyed to JI through this system. However, SS WWTP also was found to have high flows into the plant. Significant infiltration of rainwater through leaking or cracked sewer pipes occurs in the Milwaukee system as well as some direct inflow through openings in sewer manhole covers. The high diversity in both JI and SS samples suggests that additional community members enter the system during this time, such as soil bacteria from direct runoff or through infiltration. Further sampling would be necessary to assess if there are seasonal patterns in sewage communities, and discern specific taxa that correlate with sewer system configuration or environmental parameters. Such information will be important to establish which taxa provide a consistent signature that can be used to track sewage contaminated water.
The 454 sequencing platform allows us to collect sequence information for thousands of microbial organisms in sewage, including human commensals and closely related variants. This approach would completely redefine how we identify new indicators and track pollution. Instead of looking for a single indicator organism, suites of alternative indicators, and their relative abundance patterns could be determined in sewage and subsequently tracked in contaminated water. Using DNA-based detection methods, it is possible to test for a broader range of organisms, including fecal anaerobes, and to target multiple phylotypes within these groups that may be associated with a specific source of fecal pollution. These multi-taxa signatures could also be used to differentiate between human and non-human fecal sources. We identified tags shared between human feces and sewage, which are potentially new alternative indicators. New data generated as part of the human microbiome project, along with comparisons of microbial communities from other animals (Ley et al., 2008) will increase and refine the database of human fecal pollution signatures. This approach provides new insight into the composition and complexity of sewage-derived microorganisms.
We obtained WWTP influent samples of the two major facilities servicing Metropolitan Milwaukee, WI on multiple dates in 2005 and 2007 (Table 1). The entire service area is 1065 km2. JI WWTP is located in downtown Milwaukee and receives sewage from 122 km of metropolitan interceptor sewer (MIS), the northern sector of the metropolitan area. A portion of the city serviced by JI WWTP (50 km2) is comprised of combined sewers that consist of both sanitary sewage and stormwater from Milwaukee’s most densely urban area. SS WWTP services primarily residential areas of the metropolitan area and receives sewage from 201 km of MIS. An additional 143 km of MIS can be diverted to either SS or JI. Samples consisted of 1 L of 24-hour flow-weighted samples collected from 6 am on the preceding day until 6 am on the stated collection day. Flow into the WWTP was averaged between measurements taken at the 6 am time points each day. Sewage samples were filtered through a 47 mm, 0.45 µm nitrocellulose membrane (Millipore, Billerica, MA) at 100 ml or until the filter clogged. Surface water was collected at the Milwaukee estuary on 19 June 2007 using a 5L grab sampler. Three samples were pooled prior to filtering 200 ml as described above. All samples were stored at −80°C until processing for DNA extraction. DNA extractions were carried out using the Fast DNA spin kit for Soil (MP Biomedicals, Solon, OH).
Sewage samples were examined for culturable E. coli, enterococci, and C. perfringens levels to compare with total cell counts. Sewage influent samples were diluted to a range of 10−3 to 10−5 and 1 ml was filtered through a 0.45 µm pore size 47 mm nitrocellulose filter. Filters were placed on modified m-TEC (Difco, Sparkes, MD) for E. coli or mEI (Difco, Sparkes, MD) agar according to the US Environmental Protection Agency (USEPA) methods for E. coli and enteroocci enumeration (USEPA, 2000). Levels of C. perfringens were enumerated using membrane filtration and mCP media was prepared according to previously published methods (Armon and Payment, 1988). After anaerobic incubation, mCP plates were exposed to vapors of ammonium hydroxide for 20−30 sec and the straw colored colonies that turned pink to red were counted as C. perfringens. The direct counts were preformed on sewage following the acridine orange method presented by Jjemba et al. (2006). Both 10−3 to 10−4 dilutions were made from sewage influent. Twenty fields per slide were counted.
We used a Roche Genome Sequencer FLX System (GS-FLX) to characterize populations of PCR amplicons from hypervariable regions in rRNA genes. A cocktail of five fused primers at the 5’ end of the V6 region (E. coli positions 967–985) and four primers at the 3’ end (E. coli positions 1046-1028) captured the full diversity of rRNA sequences represented in molecular databases (Sogin et al., 2006; Huber et al., 2007). The 5’ fused primer includes a 5 nt “key” inserted between the Life Sciences primer A and the 967F rRNA primer. The 5 nt key is permuted for each sample and permits the bioinformatics binning of tags according to sample source. We prepared amplicon libraries from at least three independent PCR cocktails for each sample to minimize the impact of potential early-round PCR errors. After passage through a Qiagen MinElute PCR purification kit (Qiagen, Valencia, CA) and monitoring on an Agilent Bioanalyzer 2100, the Life Science A primer sequence on each fragment anneals to complementary oligonucleotides that are tethered onto micron-scale beads. The annealing conditions favor one fragment per bead. The beads were emulsified in a PCR mixture-in-oil and PCR amplification occurs in micro-reactors generating ~ten million copies of a unique DNA template. After breaking the emulsion, the DNA strands are denatured and beads carrying single-stranded DNA clones are deposited into wells on one of the two regions of the PicoTiterPlate™ and sequenced on the GS-FLX, providing approximately 400,000 reads per run.
We trimmed the resulting sequences of their 5 nt keys and primers then filtered out as low-quality all reads without an exact match to the forward primer, without a recognizable reverse primer, with fewer than 50 bases, or with any ambiguous base calls (Ns) (Huse et al., 2007). Reads were binned to samples using the 5 nt keys. We used the rRNA indexing algorithm Global Assignment of Sequence Taxonomy (GAST) (Huse et al., 2008) for comparing tag sequences to known rRNA genes that have already been placed in a phylogenetic framework of more than 500,000 nearly full length reference SSU rRNA genes (RefSSU) based on the SILVA database (Pruesse et al., 2007). GAST compares each tag against a database of V6 sequences excised from the RefSSU, selects the nearest reference V6 match or matches, and assigns taxonomy to the tag using a consensus of the taxonomy assigned to all RefSSU sequences containing that V6 region. GAST is more computationally intensive than relying upon BLAST score ratios (Rasko et al., 2005) but results in a better match in 5–10% of comparisons (Huse et al., 2008).
To create OTUs across the three environments (sewage, human, surface water), we randomly selected 38,368 sequences from each (the number of sequences in the smallest set, surface water). These sequences were combined, duplicates removed, and aligned with MUSCLE (with the –diags and –maxiters 2 options) (Edgar, 2004) and all pairwise sequence distances were calculated with quickdist (Sogin et al., 2006). We used DOTUR (Schloss and Handelsman, 2005) to create furthest neighbor clusters at 3%, 6%, and 10% pairwise distances, all duplicate sequences were reassigned to their corresponding clusters. DOTUR output also calculates richness estimates (ACE and Chao) and rarefaction. For richness estimates of individual datasets, we likewise subsampled to the smallest set (17300 from JI 20 April 2005), created a multiple sequence alignment with MUSCLE on the unique sequences, clustered with DOTUR, reassigned all duplicate sequences to the clusters and used DOTUR to estimate the richness.
The OTU overlap among samples was plotted using The Venn Diagram Plotter provided by Pacific Northwest National Laboratory, Department of Energy, which is made available OMICS.PNL.GOV
We calculated the Bray-Curtis distances between WWTP samples based on the GAST taxonomic assignments made to each sample using the VAMPS website (http://vamps.mbl.edu). A similarity matrix of these distances was used to construct a UPGMA tree using the R package (Paradis et al., 2004). Differences in human taxa abundance in WWTP samples were evaluated using a student’s t-test with a significance level of p ≤ 0.05.
Sequences were deposited in GenBank under the study accession ID SRP000905.
This work was funded by NIH grant 1 R21 AI076970-01A1, and by NSF/OCE grant 0430724.