Part 2 examines reverse transcriptase-polymerase chain reaction, pulse field gel electrophoresis, and F-specific coliphage typing.
To read the first part of this series please click here.

By Mary Catherine Hager

Water-quality supervisors and stormwater managers face a daunting challenge in determining and eliminating sources of fecal bacterial contamination, especially when the contamination occurs in coastal waters. Charles McGee, microbiology laboratory supervisor for California’s Orange County Sanitation District, expresses the overwhelming issues stormwater managers confront in resolving bacterial contamination in coastal areas: "In our marine environment we have significant contributions to our beaches from urban runoff, cats, dogs, birds, humans … all kinds of E. coli and enterococci. Unless you understand the relative contributions of those sources, it’s hard to know how many samples to collect. Bacterial source detection is practically impossible in terms of costs and the work involved."

Various bacterial source tracking (BST) methods are currently in use, and more are emerging. Project managers must make informed assessments of their needs and decide which methods provide the best fit for their situations. BST methods vary widely, from what questions they answer to how much they cost and how long they take to perform. Part 1 of this article featured descriptions of three BST methods: ribotyping, toxin biomarkers, and antibiotic resistance analysis. Here we profile three more BST approaches: reverse transcriptase-polymerase chain reaction, pulse field gel electrophoresis, and F-specific coliphage typing.

Reverse Transcriptase-Polymerase Chain Reaction

Reverse transcriptase-polymerase chain reaction (RT-PCR) is a molecular source tracking technique that can be used to detect the RNA of any organism whose genome has been sequenced. Rachel Noble, Ph.D., a senior scientist with the Southern California Water Research Project, has applied the method extensively to enteroviruses in coastal waters. The method employs primers that are complementary to conservative RNA sequences found in the viruses, those that are shared within an entire viral family. The "reverse transcriptase" step of the process basically transcribes the detected RNA back into DNA that can then be amplified by PCR. This amplification, which was described in more detail in Part 1, essentially allows the investigator to obtain large numbers of copies of small DNA sequences. The scientific community first learned of PCR about 15 years ago, but RT-PCR has been applied to the study of environmental samples more recently: "Within the past 10 years," says Noble, "and there are still a lot of bugs being worked out."

Noble detects enteroviruses as indicators or tracers of human fecal contamination, the known source of these particular viruses. She and other researchers do not know if the enteroviruses found in coastal water bodies are infective but suspect they are, or have been, because RNA degrades rapidly in seawater. Unlike bacteria, these viruses do not reproduce once outside of the human body.

RT-PCR is not location specific–it can be used anywhere–and it does not depend on an established database, an advantage over some other BST methods. RT-PCR confirms the presence of the targeted organism, such as the enteroviruses Noble detects in coastal waters. Another advantage of RT-PCR is the volume representation afforded by the sampling design. Unlike the BST methods that culture bacterial colonies from small water samples (generally 50-100 ml), the RT-PCR method starts with large water samples of about 20 liters. Noble explains that the sample is then concentrated down to a very small volume: "We filter the water to remove anything large that will interfere with PCR. We concentrate everything in that remaining volume of water down to a small volume, typically less than about 5 milliliters, sometimes even 1 milliliter. We have about a 20,000-fold concentration." Although this extreme concentration keeps sample volume representation high, it has its drawbacks: "Concentrating storm-drain water or seawater to this volume, it’s a mess. It’s very difficult to work with some of these concentrates. They are gelatinous." Following concentration, RNA is extracted from a very small portion of the sample, meaning that many tests can be run from a single sample.

Noble has applied RT-PCR to samples taken from Newport Bay in southern California, working with that region’s Regional Water Quality Control Board (RWQCB). Linda Candelaria, Ph.D., director of the Newport Bay bacterial study for the RWQCB, describes how the RT-PCR method was selected: "We are working on the implementation of the fecal coliform total maximum daily load [TMDL] for Newport Bay. In essence, we are trying to define and control the sources of bacterial and viral pollution. That TMDL calls for the analysis of fecal and total coliform and enterococci. These are traditional indicators of pathogens in the system. In addition, we wanted to determine, by ribotyping or viral testing, if human waste was present." Candelaria explains that viral testing, unlike ribotyping, is a direct measure of pathogens in the system, and RT-PCR is one method of determining the presence of viruses. "Currently RT-PCR gives you a positive or a negative test result for virus particles; however, many virologists believe if you’re finding viral fragments, then live viruses are also present. That’s why sampling for enteroviruses is probably a better indicator of the presence of viruses than using traditional bacterial indicators for viruses." Stakeholders got together to discuss sampling in Newport Bay and, says Candelaria, "it was a group decision to go with viral sampling rather than trying to identify [specific bacterial sources] of E. coli," which is the goal of ribotyping.

Sampling took place in summer 2000, at nine sites selected from more than 30 stations monitored weekly by the Orange County Health Care Agency (OCHCA). At that time, OCHCA monitored E. coli, total coliform, and enterococci and has since added fecal coliform to its monitoring program. Candelaria believes OCHCA will sample Newport Bay again in summer 2001 because dry-season sampling is the target of the current study.

Noble’s RT-PCR work indicated possible enteroviruses at one site, but the presence of enteroviruses demonstrated no apparent relationship to the presence of fecal bacteria, possibly because the fecal bacteria could be coming from local bird populations. Candelaria points out that, because of limited funds, the study was designed to show trends and hot spots, not to give statistically significant results, with respect to the relationship between traditional indicators and viral testing. "We were looking for trends to identify the highest-priority areas. We wanted to screen areas that had the most bacterial hits, based on OCHCA’s weekly monitoring, to determine if bacterial exceedances corresponded to viral hits." For now, the study’s results remain unconfirmed. "This is a sensitive area, and it’s important ecological work," says Candelaria.

One of the potential problems with RT-PCR is inhibition of the PCR process caused by certain components of samples, which makes it difficult to compare samples. Many positive and negative controls are required to determine when and which samples are being inhibited. "For example," explains Noble, "if I run a PCR and get a negative result, we spike that with a known amount of enterovirus in a side-by-side test and make sure that comes out positive so we’re not getting a false negative."

Other limitations of RT-PCR are its expense, as the necessary laboratory equipment and chemicals are costly, and the technical expertise it requires. Noble points out that many BST methods are expensive, however, particularly if budgets allow the desired replication of sampling from a given site. With PCR’s expense and technical difficulty comes rapid sample analysis–sample results can be obtained in less than 24 hours and sometimes in less than 10 hours.

RT-PCR is reproducible and accurate and offers high specificity, as it is able to determine presence or absence of a targeted organism. Once a laboratory step known as hybridization confirms a positive result, researchers can be virtually certain the targeted organism existed in the sample. Sample analysis can be performed on a specific type of virus or organism, or a group of organisms, such as enteroviruses. Ways of improving the reliability of using RT-PCR for analysis of environmental samples include increasing the number of samples taken, replicating samples taken from a single location and at a single time, and reducing the likelihood of false negative results by "cleaning" samples to remove substances that inhibit the PCR reactions.

Like most source tracking techniques, RT-PCR does not allow for quantitative assessment of how many different species contribute to fecal contamination at a sampling site, or at what concentrations. Noble can run RT-PCR in a most probable number fashion, which produces estimates with very wide confidence intervals and is considered semiquantitative.

Candelaria finds that all source tracking methods have their strengths and shortcomings, but "state-of-the-art techniques, and the accuracy of those techniques, are being refined every day." For now, she feels that all methods are "in the research mode," with no one technique standing out as the ultimate method. Methods for detecting viruses are much more expensive per sample than bacterial detection methods, and Noble cautions that the two types of approaches should not be compared directly because "they tell you different things." Noble and her lab co-workers have made great strides in reducing the cost of viral testing by optimizing the method and the time required to perform it.

Noble advises managers faced with choosing and comparing BST methods to "evaluate what you have as a budget and very well define the question that you are interested in answering." Many managers share Candelaria’s goals for Newport Bay: "We work with both dischargers and environmental groups to get meaningful solutions; we want to come to a solution as a community. Those are the best long-term solutions."

Pulse Field Gel Electrophoresis

Pulse field gel electrophoresis technique

Pulse field gel electrophoresis (PFGE), a technique used in the field of genetics, is a molecular BST method that provides DNA fingerprints of sources of fecal bacterial contamination in a water body. PFGE is similar to ribotyping (see Part 1 of this article), although ribotyping analyzes ribosomal RNA of E. coli strains, whereas PFGE works with the whole DNA genome of E. coli strains. Developed commercially by Bio-Rad Laboratories, the PFGE technique was pioneered as a BST technique around 1994 by George Simmons, Ph.D., of Virginia Polytechnic Institute and State University and Stephen Edberg, Ph.D., of the Yale University School of Medicine. As with ribotyping, PFGE uses restriction enzymes to cut E. coli DNA at specific locations. The resulting segments are then run through electrophoresis to generate banding patterns that can be compared against known patterns.

Modifications incorporated by the PFGE method set it apart from other approaches to electrophoresis. A specially designed gel setup, called the Genepath apparatus, sends electric current through a gel in different directions for several hours, which allows for superior band separation. Bacterial DNA analyzed through PFGE are embedded in agarose plugs. These plugs are placed in hollow combs of the electrophoresis gel, where they become part of the gel as the gel moves over the combs. Following electrophoresis, banding patterns become apparent after the gels are stained. Embedding the DNA in the agarose plugs essentially eliminates the potential for sample contamination, a common problem with molecular BST approaches.

PFGE is a database-dependent methodology, as researchers employing the technique seek to subtype isolates of water and bacterial sources by matching them with previously identified isolates stored in an established library. Because sources of bacterial contamination may vary from place to place, PFGE is location specific in terms of requiring isolate libraries to represent E. coli strains specific to each sampling region.

PFGE is a highly sensitive BST method, which enhances its effectiveness but can increase costs incurred through subtyping and identifying the many isolates that the method can detect. "PFGE is a much more sensitive method than ribotyping," says Mansour Samadpour, Ph.D., of the University of Washington, who actively employs ribotyping for source tracking. "With ribotyping, we might find 5,000 E. coli [strains]; because PFGE is more sensitive, it will find between 40,000 and 50,000 E. coli." Samadpour points out that funding limitations usually prevent the identification of so many E. coli strains, which in turn could somewhat reduce identification standards of PFGE. The method’s sensitivity also makes it prone to detect random mutations, which are not likely to be useful additions to an isolate database. Still, PFGE maintains an accuracy rate of 70% or better.

The Northern Virginia Regional Commission (NVRC) contracted with Simmons at Virginia Tech to explore sources of bacteria in the Four Mile Run watershed, a 20-mi.² urban stream system in the metropolitan Washington, DC, area. Don Waye, senior water resources planner at NVRC and coordinator of the Four Mile Run study, explains, "We went with PFGE because of Simmons’ success in applying it to Virginia’s Eastern Shore." He also notes the well-documented procedures and quality assurance/quality control trail for this BST technique.

Sample turnaround time for PFGE analysis remains relatively slow: up to a month or longer. As with other molecular BST methods, PFGE can be costly, and it requires personnel with specific laboratory training. Sample volume representation is relatively low, with small samples (100 ml) and only a few strains tested per sample. Waye points out a shortcoming of BST methods in general: "Another limitation with all these studies is that they are based on a limited number of grab samples, which means that you are collecting extreme minutia about a single point at a single instant in time. Thus sampling bias becomes extremely important to control against." In the Four Mile Run study, Waye explains, "Our main hedge against sampling bias was to collect samples along as many different nooks and crannies across our study watershed as possible. We also collected data across all four seasons and collected samples from both the water column and bottom sediments."

The ability of the PFGE method to accurately represent which E. coli strains are in a given watershed depends on the number of strains identified relative to the total number of E. coli in the sample and the number of samples taken. These factors also affect the ability of this method to be considered quantifiable. "Simmons’ application of PFGE is based on statistical probabilities," says Waye. Comparing the Four Mile Run results to a ribotyping analysis of a similar creek nearby, Waye reports, "We were pleased to see that our study found many of the same sources in roughly the same proportions."

Many scientists, such as Noble, consider PFGE results to be reproducible. The method can provide high specificity, allowing the identification of many species that contribute fecal bacteria to water bodies. In the Four Mile Run watershed study, PFGE matched, at 80% similarity, 350 E. coli isolates with particular animal species in the isolate database. Preliminary results of the study indicate that 85% of isolates originated with nonhuman sources, including many wildlife species typically associated with urban streams. Waterfowl contributed more than a third of the matched bacteria, dogs accounted for 11% of the matched strains, and raccoons 15%. Deer and rats were also identified as animals contributing to bacterial contamination of the watershed. Simmons and Waye believe that this BST profile has potential implications for many other metropolitan areas.

F-Specific Coliphage Typing

Bacteriophages are viruses that infect bacteria, and coliphages are bacteriophages that are specific pathogens of E. coli. Because coliphages almost always come from fecal material, their presence in water bodies can indicate bacterial contamination. Mark Sobsey, a microbiologist at the University of North Carolina, tracks bacterial contamination sources through typing certain coliphage groups that contain RNA, those known as the "RNA male-specific" or "RNA F+" group of coliphages. "We have found the typing of those [coliphages] to often be useful in identifying pollution sources in a general way," notes Sobsey, "and in particular to distinguish between human and nonhuman fecal contamination."

E. coli cells can be male or female, and male-specific coliphages infect only male E. coli cells through bacterial appendages called "pili," which are unique to male cells. Male E. coli cells transfer genetic information to female cells through these pili. An "F+" designation–"F" signifying "fertility"–represents the genetic information responsible for producing the pili on male E. coli. There are four groups of F+ coliphages: Group 2 is the human-specific coliphage group in North America; Group 4 is the animal-specific coliphage group, virtually never found in human wastes; Group 3 contains human-specific coliphages common in other parts of the world; and Group 1 consists of coliphages found in both humans and animals.

Sobsey explains that the preferred method for typing coliphage groups found in water is genotyping, using a nucleic acid probe technique–gene probing–common to many labs. This approach puts coliphage typing in the "molecular BST technique" category, but until recently, the common approach to distinguishing coliphage groups was serotyping, a biochemical method. With serotyping, Sobsey describes, "We made antisera against the four groups of phages, then you could use the antisera to prevent infection of host cells by the corresponding serum. If you had different phages, then you could show that the antisera against Group 1 prevented the growth of Group 1 phages, and so on." Japan developed the first coliphage serotyping technique in the 1980s. In the early 1990s, Sobsey’s lab and a lab in the Netherlands became involved in coliphage typing, each working independently to develop a genotyping approach. The two labs eventually combined efforts, publishing a joint paper on the genotyping approach in 1995.

Coliphages collected in water samples grow in a Petri dish on a "lawn" of E. coli. Coliphages can be spotted in circular areas called "plaques," where phages have lysed bacterial cells. The number of plaques represents the number of coliphages contained in the volume of water put into the Petri dish. The method is fairly sensitive, giving a clear number of coliphages affecting E. coli. Some of the phages are removed from the plaques and placed in bacterial suspension, then incubated again in Petri dishes on lawns of E. coli, where they form lysis zones, discrete areas of lysed cells that contain many coliphages. From each individual plaque and single drop of the suspension, lab workers can create multiple lysis zones. Sobsey explains how the genetic probe occurs with the F+ coliphages: "With areas of lysis on the lawn of host bacteria, you can lay a piece of filter paper over the plate and the phages will stick to the filter paper. You can then, by chemical methods plus heat, release the nucleic acids from the bacteriophages and have the nucleic acid (RNA) adhere to the filter very strongly. It’s done in such a way that the RNA is actually chemically bonded to the filter paper. The filter paper is placed into a solution that has a DNA probe that is complementary to the RNA, or to a length of RNA, in the phage. If the DNA finds this complementary RNA site in the lysis zone, it will bind by the usual DNA binding mechanism. The probe sticks to the corresponding viral RNA." In the lab, technicians prepare the DNA with some kind of visual tag that allows it to be easily detected. There are four different probes, one for each of the coliphage groups, although it is possible to combine probes of different types.

The coliphage typing approach is most useful when a question arises of whether fecal contamination in a water body originates from human or animal sources, but it is not capable of distinguishing particular animal sources. "Coliphage typing is a rapid, less expensive way to make the initial cut, is it human or animal?" says Sobsey. "If it’s animal, we can’t tell if it’s cattle, pig, sheep, et cetera. More sophisticated analysis, more advanced nucleic acid genetic analysis, would have to be done to get more definitive information." He recommends the higher specificity of such methods as RT-PCR, PFGE, ribotyping, antibiotic resistance analysis, and toxin biomarkers, all of which are described within the two parts of this article.

Coliphage genotyping is very simple and inexpensive to perform. The gene probes–short lengths of synthetic DNA–are widely available. The method can be performed fairly rapidly, producing results in about two days. Coliphage typing is not database dependent or location specific, only "waste-source specific," to human or animal contamination.

Sobsey describes investigations in which coliphage typing provided straightforward answers. Joan Rose, Ph.D., of the University of South Florida determined through coliphage analysis that instances of fecal contamination in the Florida Keys came from faulty septic tanks along the shore. Once coliphage typing determined the presence of human fecal bacteria, the septic systems were pinpointed through dye studies, where dye flushed down toilets went straight to the contaminated water. Sobsey’s lab also performed similar studies on freshwater lakes with contaminated beaches.

Sobsey also used coliphage typing to determine the source of fecal bacteria in drinking-water reservoirs for New York City as part of an investigation conducted for the New York City Department of Environmental Protection. Evidence at the reservoirs suggested that E. coli likely came from waterfowl, particularly Canada geese, but regulatory agencies needed to confirm that the contamination wasn’t human, resulting from sewage release. Coliphage typing of samples taken from different parts of the reservoir at different times indicated both that coliphages appeared primarily when the geese were present and that they were not from human sources. Coliphages found in goose excrement matched those collected in the reservoir. Reservoir managers implemented programs to actively discourage the birds’ use of the reservoir, and that controlled the E. coli problem. Contamination was reduced, points out Sobsey, through a "low-cost, effective solution accepted by regulatory agencies."

In an unpublished comparison of various BST methods, Betty Olson, Ph.D., of the University of California, Irvine names substantial sample volumes as another plus of the coliphage typing approach. As with the toxin biomarker method, sample volumes of water of 1-10 liters processed in coliphage analysis "give a better statistical representation of the types of contamination present in the water." Olson cites some problems with crossover between animal and human coliphage groups that could affect resolution between species. She shares Sobsey’s own assessment that, for now, the method is best limited to identifying sewage pollution rather than specific sources.

Olson also notes that the potential for the method to be quantifiable depends on how well the phage population that is randomly picked for analysis represents the coliphages in the total sample and in the environment sampled. Sobsey believes, "More work needs to be done with our system to try and come up with better performance criteria, for being confident in getting enough samples to distinguish human versus animal contamination." He does not restrict his concern to coliphage analysis, however: "Neither we nor anyone else has come up with specific quantitative criteria on that basis. That’s probably something that should be done for all these [BST] systems at this point. Right now a lot of judgment gets exercised in interpreting the results."

Conclusions

Although no one BST method offers a perfect fit for all situations, many are well suited to address particular issues associated with bacterial contamination of coastal waters. The "toolbox" approach advocated by Charles Hagedorn of Virginia Tech and described in Part 1–combining various methods to capitalize on the best features of each–might lead to stronger, more reliable BST approaches. The field of bacterial source tracking continues to evolve rapidly, and researchers see promising developments emerging. 

Mary Catherine Hager is a biologist, writer, and editor in Lafayette, LA.

To read the first part of this series please click here.