A tongue-in-cheek view of philosophy sometimes seen on T-shirts presents a pithy summary of the essentialist and existential positions, as well as the unofficial but more popular Cool School: “To be is to do.”—Socrates; “To do is to be.”—Jean Paul Sartre; and “Do be do be do.”—Frank Sinatra.
It is fitting that the discipline of microbiology, having originated in France, has historically taken an existential or phenomenological approach to identifying microorganisms, what we call phenotyping. Identification has depended on the growth and appearance of colonies on agar, the shape of organisms under a microscope, their uptake of chemical stains, and biochemical reactions with sugars and other metabolites.
In the past few decades, however, with the development of molecular tools, it has become possible to analyze the essential nature of microorganisms, their DNA, and molecular methods have entered clinical microbiology laboratories. First was PCR, but more recently sequencing instruments have become available for diagnostic work.
Cathy A. Petti, MD, laboratory director at the Palo Alto Medical Foundation, says: “Sequencing is truly advancing the field of microbiology [Petti CA. Clin Infec Dis. 2007;44:1108–1114]. From my perspective, DNA target sequencing has revolutionized our ability to understand the breadth of microorganisms responsible for human disease.” Dr. Petti acknowledges that biochemical tests and semiautomated instruments are efficient for characterizing standard isolates. Only now, however, has it become possible to identify microorganisms accurately and find new and emerging pathogens, she says. “It is a misnomer for people to publish some of these microorganisms as ‘novel’ species, assuming they’ve never before caused human disease,” she says. “We just didn’t have tools 20 to 25 years ago to discriminate routinely Weissella sp from Lactobacillus sp or viridans group streptococci, for instance. DNA target sequencing has enabled us better to identify microorganisms and study virulence factors in these microorganisms in a particular clinical syndrome.”
Even more important, Dr. Petti adds, sequencing has called into question the basic identity of even well-studied microorganisms. “Depending on the target sequence you use, it is difficult to know what is E. coli and what is Shigella sp,” she says. “If you use the 16S ribosomal RNA [rRNA] gene, they are identical. Take it a step further—full genome sequencing has revealed that some strains of E. coli share only 40 percent of their set of proteins.”
Michel Drancourt, MD, PhD, cites the main practical advantage of sequencing: It is definitive. “When you have the sequence of an organism, there is nothing more that you can get,” says Dr. Drancourt, professor of clinical microbiology at Marseille Medical School and the University of the Mediterranean.
A second major advantage is that sequences can be stored in large databases. “There are now several huge databases that are easily accessible on the Internet,” Dr. Drancourt says. “Anyone in the world can compare a sequence they get in their laboratory to a database. That cannot be done with other methods.” In the United States the best known and most important sequence database is GenBank.
Augmenting the value of sequencing for bacterial identification has been the proliferation of bacterial genomes sequenced. Dr. Drancourt estimated that by mid-2007, 479 genomes from 352 species had been sequenced, covering all notable human pathogens (Fournier PE, et al. Lancet Infect Dis. 2007;7:711–723). This number has continued to grow.
At the Centers for Disease Control and Prevention, Hans Peter Hinrikson, PhD, FAMH, is working to build a comprehensive reference database of DNA sequences and other diagnostic information for microorganisms. He points out that the quality of the sequencing approach to identification depends on the quality of the reference database used. “Free databases like GenBank, while containing sequences from a wide range of microorganisms, are not all curated,” says Dr. Hinrikson, chief of the Special Bacteriology Reference Laboratory in the Coordinating Center for Infectious Diseases. Commercial databases are curated but tailored for only the most common pathogens. “What we are trying to do,” Dr. Hinrikson says, “is to assemble a high-quality reference database, MicrobeNet, that contains sequences and other diagnostic information for all clinically relevant microorganisms as well as their phylogenetic relatives, so that if we see anything for the first time we can identify it.”
In the diagnostic microbiology laboratory, sequencing is first being applied to mycobacteria. “We can identify virtually all mycobacteria by sequencing,” says Nancy L. Wengenack, PhD, D(ABMM), consultant in the Division of Clinical Microbiology at Mayo Clinic and associate professor of laboratory medicine and pathology at Mayo College of Medicine. Sequencing has a real advantage over nonmolecular methods because many mycobacteria grow very slowly. Using typical biochemical methods, it could take several weeks to two months for definitive identification. “The beauty of sequencing,” Dr. Wengenack says, “is that once an isolate grows in culture, we can accomplish identification in 24 hours. This short turnaround time with sequencing is a tremendous benefit to patients.”
Karen M. Frank, MD, PhD, director of the clinical microbiology and immunology laboratories and assistant professor in the Department of Pathology at the University of Chicago Medical Center, uses sequencing to identify acid-fast bacilli and is validating it for other microorganisms that are hard to identify by standard methods. Examples include nonfermenters and anaerobes. “Validation for these organisms is a bit trickier than it was for mycobacteria,” Dr. Frank says. She now sends out hard-to-identify organisms for sequencing. By sequencing these organisms in her laboratory as well, and then comparing her results with those from the reference laboratory, she will obtain validation data.
Jill E. Clarridge, III, PhD, D(ABMM), has been sequencing selected clinical specimens according to defined criteria for more than 15 years. “We found that if the decision to perform 16S rRNA gene sequencing of an isolate is based on sound criteria, it is a necessary, sufficient, and cost-effective method for determining the identity of certain clinically significant organisms,” says Dr. Clarridge, chief of the microbiology, serology, and molecular microbiology laboratory service at the Seattle VA and professor of laboratory medicine at the University of Washington.
Molecular methods can also be helpful with fungi. “We can identify a large number of fungi easily and rapidly by conventional methods, such as microscopy,” Dr. Wengenack says. Candida albicans and C. glabrata are good examples. However, for fungi that grow slowly or don’t sporulate well, sequencing can be very helpful. Shorter turnaround time is a major advantage. “Some esoteric Candida species and some molds take a long time to identify,” Dr. Wengenack says. “Dermatophytes are one area where we, as a reference laboratory, use sequencing fairly heavily.”
Gary W. Procop, MD, MS, chair of clinical pathology and director of molecular microbiology, mycology, and parasitology at the Cleveland Clinic, has worked extensively with molecular methods for fungi. “Mycology is one field where molecular methods will likely prove superior to traditional techniques,” Dr. Procop says. One application is rapidly characterizing yeast in blood cultures. “Yeasts are an important cause of fungemia in the hospital,” Dr. Procop says. “They have a high mortality rate. Identification is important because it directs antifungal therapy.” He hopes eventually to be able to identify yeast directly in clinical specimens.
One molecular method—protein nucleic acid FISH, or PNA FISH—has been particularly helpful with fungi, says Michael Pfaller, MD, professor emeritus of pathology at the University of Iowa College of Medicine. Dr. Pfaller calls PNA FISH “probably the most significant advance in rapid identification of fungi since the introduction of DNA probes for endemic fungi.” PNA FISH probes are more cost-effective than traditional methods for differentiating Candida species.
However, Dr. Pfaller cautions against expecting too much from molecular methods. “Although we would like to be able to make direct diagnosis and even identification from clinical specimens, that is not going to happen anytime soon,” he says. (See “First, tackle the basics,” above.)
Expatiating on the basic knowledge gained from sequencing, Dr. Petti says, “What molecular tools have done is to further our understanding of the rapid exchange of genetic material between species that makes them so complicated and that can affect antimicrobial resistance.” For instance, with conventional tests, extended-spectrum beta-lactamases, or ESBLs, appear limited to a few organisms. “However,” Dr. Petti says, “now that we can determine genetically whether a resistance factor or gene is present, we sometimes find ESBLs in organisms in which we would not have looked for them.”
Another example of genetic exchange is the demonstration of convergence of Campylobacter coli and C. jejuni (Sheppard SK, et al. Science. 2008;320:237–239). “For decades we distinguished those two species,” Dr. Petti says. Now there appears to be devolution of the species, arguing that perhaps they should be one species. “We never would have had this understanding without molecular tools,” Dr. Petti says.
Basic research like this uses whole-genome sequencing. In clinical diagnostic work, however, cost and the need for high throughput dictate use of only one segment of a DNA target. For bacteria, the consensus sequence is usually the first 500 base pairs of the 16S rRNA gene; for fungi, it is the internal transcribed spacer 1 and 2 regions (ITS1 and ITS2). “Those two targets have brought us a greater awareness of the heterogeneity within species, as well as a greater awareness of similarities between genera,” Dr. Petti says. “When we brought on sequencing at ARUP Laboratories, the technologists said, ‘If we use the semiautomated instrument we always get an answer. With DNA target sequencing we often cannot distinguish among the genera Citrobacter, Pantoea, and Enterobacter.’” Dr. Petti’s interpretation is that getting an answer doesn’t always mean it is truth: “What DNA target sequencing demonstrates is that there is a lot more phylogenetic complexity among bacteria than we thought.”
Moving to culture-independent methods for identification implies becoming more independent of culture in looking for resistance as well. “With nucleic acid sequencing we will be able to detect genetic regions for resistance and virulence directly,” Dr. Petti says. “In the future we won’t care so much what the name of an organism is. What we really want to know is whether it carries a shiga-like toxin, for instance. That’s a property that causes disease.” She cites genetic analysis of the pneumococcus as “a great example” of this principle. Streptococcus pneumoniae carries the gene for pneumolysin, thought to be an important virulence factor. However, sequencing has revealed that some S. mitis isolates also carry the pneumolysin gene and can cause severe respiratory symptoms and invasive disease (Whatmore AN, et al. Infect Immun. 2000;68:1374–1382; Neeleman C, et al. J Clin Microbiol. 2004;42;4355–4357; Ip M, et al. Antimicrob Agents Chemother. 2007;51:2690–2700). “If an isolate is identified as S. mitis, the clinician won’t think it’s a serious infection,” Dr. Petti says. “But if you report ‘pneumolysin gene present,’ the physician will be more alert.” In the future, Dr. Petti predicts, laboratory reports will highlight not the name of an organism but which virulence factors it carries.
Dr. Drancourt has been using DNA sequencing for microbial identification since the mid-1990s, giving his laboratory at Marseille Medical School and the University of the Mediterranean one of the longest experiences with this technology in the world. He finds it useful for fastidious organisms that are hard to isolate from specimens, such as some mycobacteria and intracellular bacteria including Rickettsia and Coxiella. A second use is for bacteria not in their usual tissue or specimen, such as pneumococci from joints. “In these situations it is worth confirming accurate identification using sequencing,” he says. Sequencing is also helpful with an organism that is “new,” in Dr. Petti’s sense. “We have described 60 to 70 organisms that had not been previously isolated from humans,” Dr. Drancourt says. In his area of expertise, mycobacteria, he has identified 10 new species in five years, many in M. avium-intracellulare complex, or MAC.
Since some bacterial species harbor almost the same gene for 16S rRNA, Dr. Drancourt is now developing secondary gene targets. One candidate is the gene for the beta subunit of RNA polymerase (rpoB) (Adekambi T, et al. Trends Microbiol. 2009;17:37–45). Initially Dr. Drancourt and his colleagues intended rpoB to be used when two species have the same 16S rRNA sequence. Now their data suggest that it could be the most useful gene for identifying mycobacterium, as well as E. coli and other organisms.
Dr. Drancourt and his colleagues are also exploring multispacer sequence typing, or MST (Djelouadji Z, et al. PLoS ONE. 2008;3:e2433). Spacers, which are noncoding regions, are generally located between genes and come in groups of three to 10. Because they are not under as intense selection pressure as coding regions, the sequences of spacers may be valuable for genotyping. “Spacers have larger freedom to become variable,” Dr. Drancourt explains. His initial work with Yersinia pestis verified that MST is useful. They have since applied it to Rickettsia and mycobacteria.
Sequencing instruments are commercially available, making it feasible to do this test in a clinical laboratory. Reimbursement can be an issue, particularly in the United States. Putting the question of money aside, Dr. Drancourt says, “From a technical viewpoint it is not a difficult method to put into a laboratory right now. Several companies sell sequencing packages,” which include a sequencer, reagents, software, and a database.
For the past two years, Dr. Frank has been using gene sequencing to identify acid-fast bacilli in clinical specimens. Sputum samples are inoculated into liquid broth in an automated culture tube system. “When the system flags a positive culture, traditionally we would subculture onto plate and slant media and set up biochemical assays,” Dr. Frank says. “It can take literally weeks to get identification this way.” To accelerate identification, they had been using a combination of traditional biochemistry and Gen-Probe assay, but Gen-Probe identifies only four mycobacterial species—gordonae, tuberculosis, avium-intracellulare, and kansasii. “We were spending lots of tech time and re-agents getting other IDs,” Dr. Frank says. “That was the impetus to bring in sequencing, which is faster, identifies a wider range of species, and in the end is cheaper.”
Next she plans to introduce sequencing for a wider range of organismssuch as grampositive rods, nonfermenters, and anaerobes. “However,” Dr. Frank says, “some species are not easily distinguished by 16S rRNA alone,” so she will select a second gene when indicated. More than one gene is necessary even for further speciation of mycobacteria, but this is not a problem because Dr. Frank’s laboratory reports out complexes, such as “Mycobacterium avium/intracellulare complex.” “That’s all that’s necessary for clinical purposes,” she says. Other potential target genes are in databases such as RIDOM (Ribosomal Differentiation of Medical Microorganisms; www.ridom-rdna.de/).
Dr. Frank’s hospital uses an ABI capillary electrophoresis sequencer and primers that the molecular diagnostics laboratory director made from published sequences and validated in their laboratories. “It was not too big of a switch” to go from using the sequencer for oncology applications, with which they were already familiar, to microbiology. “The biggest issue was interpreting databases and establishing criteria for identification,” she says. “There were no guidelines when we started, so I reviewed the literature and talked to colleagues who were already sequencing.” A group that included Drs. Petti, Clarridge, Procop, and others worked under the auspices of the Clinical and Laboratory Standards Institute and wrote a guideline on Interpretative Criteria for Identification of Bacteria and Fungi by DNA Target Sequencing, which will facilitate this process for future adopters.
In a small study of about 100 samples, Dr. Clarridge asked whether she could establish cost-effective criteria to select isolates for identification by sequencing. “We wanted to see if we could determine from growth and maybe initial results that it would be impossible for us to identify an isolate by phenotypic tests,” she says. Criteria for sequencing included discordant results on phenotypic tests or failure of the semiautomated instrument, among others. Sequencing had a one-day turnaround time.
“Even with gram-negative rods, which are generally considered easier to identify, we selected organisms whose phenotypes would have yielded a different incorrect identification more than 80 percent of the time.” With selected gram-positive rods and streptococci, Dr. Clarridge and Steven Mahlen, PhD, a postdoctoral fellow at her laboratory, showed even more necessity for sequencing for a correct identification. “We were right to sequence a high percentage of the time,” Dr. Clarridge says. Phenotypic methods would either have failed or been wrong in more than half of the cases selected for sequencing.
“Right now sequencing is too expensive for most clinical microbiology laboratories,” Dr. Clarridge says. Low volume is the major issue. “We need sequence identification for only about five to 15 isolates per week,” she says. In her pilot study the cost of sequencing was only about $16 per isolate because it was done in a shared core facility.
Dr. Hinrikson is working to establish a high-quality sequence database as part of his laboratory’s basic mission at the CDC, which he says is “to identify what is not in the book.”
“An isolate may resemble something known, but never has been identified in humans before. That is the discovery aspect. Also we assist whoever has identification problems, not necessarily because of a new pathogen, but maybe a slightly different strain or one that has lost a defining biochemical feature.” In these cases a molecular approach can be helpful.
Dr. Hinrikson has assembled a “top quality” reference database containing some of his laboratory’s rare or unusual pathogens and information from the work of others. In addition to 16S rRNA and rpoB genes, it has a limited amount of information on other diagnostic targets such as the gyrase B gene, a topoisomerase gene, which codes for an enzyme subunit involved in DNA replication. Quinolone antibiotics are directed against DNA gyrase; its mutations can be used to predict quinolone resistance.
An ongoing commitment is needed to maintain any sequence database. “Databases have been published and then taken down,” Dr. Hinrikson says. “All the work done with reference to those databases is now obsolete. There was a gyrase database that I used to use in Switzerland. The professor retired and it doesn’t exist anymore.” A European database on 16S rRNA genes was once at the forefront, but it is now “frozen,” Dr. Hinrikson says, perhaps because the funds needed to update it are not available.
In the field of clinical mycology, Dr. Procop believes that molecular methods such as sequencing are likely to prove superior in part because traditional methods take days to produce results. “Molecular methods, some commercially available, can identify fungi within hours,” he says. One nonsequencing molecular method, PNA FISH, can distinguish C. albicans from C. glabrata directly on specimens from blood culture bottles; another kit adds susceptibility to fluconazole (Shepard JR, et al. J Clin Microbiol. 2008;46:50–55). Dr. Procop calls the newer kit the “traffic light” assay—C. tropicalis is yellow, C. albicans and C. parapsilosis are green, and C. krusei and C. glabrata are red, meaning “stop” using fluconazole. “For C. krusei and C. glabrata, you go to a drug from the echinocandin class, such as caspofungin,” Dr. Procop says.
“This is almost fungal pharmacogenomics,” he points out. “Providing genomic information on the fungus tells you what treatment to use.” New antifungal drugs have been in the pipeline for many years. “Now,” Dr. Procop says, “we have several, but each has limitations.” For instance, C. parapsilosis has elevated MICs to echinocandins, and voriconazole doesn’t cover zygomycetes. “So identification has become more important than ever in helping to direct antifungal therapy,” he says.
More rapid and informative methods for identifying fungi have taken on greater importance as the possibility of fungal infections in clinical samples has become greater. “A suspicion of fungi is pretty much moving to encompass the majority of patients in hospitals today,” Dr. Pfaller says, “because we have so much greater acuity of illness these days. We are discharging a lot of people from hospitals that we would have kept in before, and folks who would previously have been in ICUs are now more often in general wards.” ICUs and organ transplant units are still “hot spots” for fungi, he notes, but there are other places in hospitals where patients can be at high risk for fungal infections.
High-risk patients in ICUs and transplant units will have multiple types of samples. “We get blood, normally sterile body fluids such as CSF and tissue biopsies,” Dr. Pfaller says. To a lesser extent they receive respiratory specimens such as invasive bronchoscopies for culture if the clinician suspects fungi. “It is becoming pretty clear,” he says, “that we really have to push to identify the organisms that are infecting these patients, and not just say ‘fungus.’” Fungi growing in culture have classically been thought of as contaminants. “We can now clearly show that they are responsible for fatal infections,” Dr. Pfaller says. “It doesn’t help a clinician who is taking care of a critically ill patient who may have been on an antifungal drug like fluconazole to say ‘Candida, not-albicans.’ He needs to know more than that because we have a variety of ways now of treating these infections. The laboratory is abdicating its role as helper of clinicians if we don’t make an effort to identify these organisms.”
The PNA FISH test is one molecular tool for achieving this objective. “It has been shown to be very accurate and relatively easy to use,” Dr. Pfaller says. “From the laboratory’s standpoint it costs a lot more than standard tests. But you have to look at it from the institutional perspective, which is driven largely by the high cost of echinocandins that are used very prominently as first-line therapy.” If the laboratory can make a call in a couple of hours or within the same day and say whether an isolate is albicans or glabrata, he says, it will direct therapeutic management of that patient and allow the clinician to make an informed decision to either keep the patient on an echinocandin, in the case of glabrata, or switch to fluconazole, which is now much less expensive, if it is albicans. “That is an important advance,” says Dr. Pfaller, who has been lobbying to have PNA FISH approved for his laboratory. “We are getting reasonably close,” he says.
In Dr. Wengenack’s laboratory, DNA sequencing is now the gold standard for mycobacterial identification in conjunction with morphology. “For fungi we use it selectively as an adjunct,” she says. “Fungi may be easy and quick to identify with traditional methods. If an isolate turns out to be more difficult, we may turn to sequencing.” They use a commercial sequencing kit from a major vendor. “We have also developed our own library of supplementary sequences,” Dr. Wengenack says, “which improves specificity.” For mycobacteria they sequence a 500-base pair portion of the 16S rRNA gene. Fungal sequencing involves the D1/D2 region of the large ribosomal subunit; the ITS region can also be used.
A number of major manufacturers provide sequencing platforms, Dr. Wengenack says. Some are based on traditional Sanger dideoxy sequencing, which she calls “tried-and-true.” In addition, several platforms are coming out using pyrosequencing, or sequencing by synthesis. “Pyrosequencing gives shorter reads, about 100 base pairs compared to 400 to 800 base pairs by the Sanger method,” Dr. Wengenack says. “But pyrosequencing is also a bit faster and cheaper.”
For mycobacteria, Dr. Wengenack rates the specificity of sequencing “excellent.” Sequence libraries for mycobacteria are good and, using one or two DNA targets, sequencing can generally differentiate to species level. “Sensitivity is not an issue, since we are working with organisms grown in culture,” she says. For fungi, libraries can be a problem. “Lab-developed libraries are a bit less robust for fungi,” Dr. Wengenack says. “We know what groups of organisms we have good libraries for.” Dermatophytes fall in this category. “In our lab, the average time for dermatophyte identification is nine days faster using sequencing,” Dr. Wengenack says.
She notes that sequencing is still a fairly complex method and requires upfront equipment cost. “So it is probably limited to larger molecular laboratories and reference laboratories,” she says. “It is not realistic for all hospital laboratories to aspire to do microorganism identification by sequencing, mostly because of cost and complexity.” In addition, smaller laboratories may not have adequate volume of mycobacteria and fungi to warrant this type of testing. “For those labs, it may make more sense to send out for sequencing,” she says. Currently, her laboratory performs about 1,000 sequences per month, with that volume split fairly evenly among fungi, mycobacteria, and bacteria (including anaerobes).
In the future, Dr. Procop hopes to see the advent of a nonsequencing molecular methodpanfungal PCR—for detecting fungi. “The real question,” he says, “is which patients with low WBC counts will get an invasive, potentially fatal fungal infection. Right now there is no way to tell until disease is pretty far along.” Dr. Procop proposes that “PCR with panfungal primers may be used to monitor respiratory secretions or blood in these patients for fungal loads similar to how we monitor patients with viral loads.” When a patient turns positive, post-amplification analysis can be used to identify the fungus and thereby direct therapy.
William Check is a medical writer in Wilmette, Ill.