The basic goals of a clinical microbiology laboratory are to establish the presence of a pathogen in a clinical sample, to identify the pathogen, and, when possible, to provide other information that can help guide clinical management and even prognosis, such as antibiotic susceptibility profiles or the presence of virulence factors. To date, clinical microbiology laboratories have largely approached these goals phenotypically by growth-based assays and biochemical testing. Bacteria, for instance, are algorithmically grouped into species by their characteristic microscopic appearance, nutrient requirements for growth, and ability to catalyze certain reactions. Antibiotic susceptibility is determined in most cases by assessing growth in the presence of antibiotic.
With the sequencing revolution paving the way to easy access of complete pathogen genomes (Fig. 146-1), we are now able to more systematically clarify the genetic basis of these observable phenotypes. Compared with traditional growth-based methods for bacterial diagnostics that dominate the clinical microbiology laboratory, nucleic acid–based diagnostics promise improved speed, sensitivity, specificity, and breadth of information. Bridging clinical and research laboratories, adaptations of genomic technologies have begun to deliver on this promise.
HISTORICAL LIMITATIONS AND PROGRESS THROUGH GENETIC APPROACHES
The molecular diagnostics revolution in the clinical microbiology laboratory is well under way, borne of necessity in the effort to identify microbes that are refractory to traditional culture methods. Historically, diagnosis of many so-called unculturable pathogens has relied largely on serology and antigen detection. However, these methods provide only limited clinical information because of their suboptimal sensitivity and specificity as well as the long delays that diminish their utility for real-time patient management. Newer tests to detect pathogens based on nucleic acid content have already offered improvements in the select cases to which they have been applied thus far.
Unlike direct pathogen detection, serologic diagnosis—measurement of the host’s response to pathogen exposure—can typically be made only in retrospect, requiring both acute- and convalescent-phase sera. For chronic infections, distinguishing active from latent infection or identifying repeat exposure by serology alone can be difficult or impossible, depending on the syndrome. In addition, the sensitivity of serologic diagnosis varies with the organism and the patient’s immune status. For instance, tuberculosis is notoriously difficult to identify by serologic methods; tuberculin skin testing using purified protein derivative (PPD) is especially insensitive in active disease and may be cross-reactive with vaccines or other mycobacteria. Even the newer interferon γ release assays (IGRAs), which measure cytokine release from T lymphocytes in response to Mycobacterium tuberculosis–specific antigens in vitro, have limited sensitivity in immunodeficient hosts. Neither PPD testing nor IGRAs can distinguish latent from active infection. Serologic Lyme disease diagnostics suffer similar limitations: in patients from endemic regions, the presence of IgG antibodies to Borrelia burgdorferi may reflect prior exposure rather than active disease, while IgM antibodies are imperfectly sensitive and specific (50% and 80%, respectively, in early disease). The complex nature of these tests, particularly in view of the nonspecific symptoms that may accompany Lyme disease, has had substantial implications on public perceptions of Lyme disease and antibiotic misuse in endemic areas. Similarly, syphilis, a chronic infection caused by Treponema pallidum, is notoriously difficult to stage by serology alone, requiring the use of multiple different nontreponemal (e.g., rapid protein reagin) and treponemal (e.g., fluorescent treponemal antibody) tests in conjunction with clinical suspicion. Complementing serology, antigen detection can improve sensitivity and specificity in select cases but has been validated only for a limited set of infections. Typically, structural elements of pathogens are detected, including components of viral envelopes (e.g., hepatitis B surface antigen, HIV p24 antigen), cell surface markers in certain bacteria (e.g., Streptococcus pneumoniae, Legionella pneumophila serotype 1) or fungi (e.g., Cryptococcus, Histoplasma), and less specific fungal cell-wall components such as galactomannan and β-glucan (e.g., Aspergillus and other dimorphic fungi).
Given the impracticality of culture and the lack of sensitivity or sufficient clinical information afforded by serologic and antigenic methods, the push toward nucleic acid–based diagnostics originated in pursuit of viruses and fastidious bacteria, becoming part of the standard of care for select organisms in U.S. hospitals. Such tests, including polymerase chain reaction (PCR) and other nucleic acid amplification tests (NAATs), are now widely used for many viral infections, both chronic (e.g., HIV infection) and acute (e.g., influenza). This technique provides essential information about both the initial diagnosis and the response to therapy and in some cases genotypically predicts drug resistance. Indeed, progression from antigen detection to PCR transformed our understanding of the natural course of HIV infection, with profound implications for treatment (Fig. 146-2). In the early years of the AIDS pandemic, p24 antigenemia was detected in acute HIV infection but then disappeared for years before emerging again with progression to AIDS (Fig. 146-2B). Without a marker demonstrating viremia, the role of treatment during HIV infection prior to the development of clinical AIDS was uncertain, and monitoring treatment efficacy was challenging. With the emergence of PCR as a progressively more sensitive test (now able to detect as few as 20 copies of virus per milliliter of blood), viremia was recognized as a near-universal feature of HIV infection. This recognition has been transformative in guiding the initiation of therapy as well as adjustments in therapy and, together with the development of less toxic therapies, has helped to shape guidelines that now favor earlier introduction of antiretroviral therapy for HIV infection.
A. Timeline of select milestones in HIV management. Genomic advances are shown in bold type. The approvals and recommendations indicated apply to the United States. ARV, antiretroviral; AZT, zidovudine; NRTI, nucleoside reverse transcriptase (RT) inhibitor; NNRTI, non-nucleoside RT inhibitor; PI, protease inhibitor. B. Viral dynamics in the natural history of HIV infection. Three diagnostic markers are shown: HIV antibody (Ab), p24 antigen (p24), and viral load (VL). Dashed gray line represents limit of detection. (Adapted from data in HH Fiebig et al: Dynamics of HIV viremia and antibody seroconversion in plasma donors: Implications for diagnosis and staging of primary HIV infection. AIDS 17:1871, 2003.)
As they are for viruses, nucleic acid–based tests have become the diagnostic tests of choice for fastidious bacteria, including the common sexually transmitted intracellular bacterial pathogens Neisseria gonorrhoeae and Chlamydia trachomatis as well as the tick-borne Ehrlichia chaffeensis and Anaplasma phagocytophilum. More recently, nucleic acid amplification–based detection has offered improved sensitivity for diagnosis of the important nosocomial pathogen Clostridium difficile; NAATs can provide clinically relevant information on the presence of cytotoxins A and B as well as molecular markers of hypervirulence such as those characterizing the recently recognized North American pulsotype 1 (NAP1), which is found more frequently in cases of severe illness. The importance of genomics in selecting loci for diagnostic assays and in monitoring test sensitivity was recently highlighted by the emergence in Sweden of a new variant of C. trachomatis containing a deletion that includes the gene targeted by a set of commercial NAATs. By evading detection through this deletion (which would have prompted the initiation of treatment), this strain came to be highly prevalent in some areas of Sweden. While nucleic acid–based tests remain the diagnostic approach of choice for fastidious bacteria, this example serves as a reminder of the need for careful development and ongoing monitoring of molecular diagnostics.
In contrast, for typical bacterial pathogens for which culture methods are well established, growth-based assays followed by biochemical tests still dominate in the clinical laboratory. Informed by decades of clinical microbiology, these tests have served clinicians well, yet the limitations of growth-based tests—in particular, the delays associated with waiting for growth—have left open opportunities for improvements. Molecular diagnostics, greatly informed by the vast quantity of microbial genome sequences generated in recent years, offers a way forward. First, sequencing studies may identify key genes (or noncoding nucleic acids) that can be developed into targets for clinical assays using PCR or hybridization platforms. Second, sequencing itself may eventually become inexpensive and rapid enough to be performed routinely on clinical specimens, with consequent unbiased detection of pathogens.
In order to adapt nucleic acid detection to diagnostic tests and thus to identify pathogens on a wide scale, sequences must be identified that are conserved enough within a species to identify the diversity of strains that may be encountered in various clinical settings, yet divergent enough to distinguish one species from another. Until recently, this problem has been solved for bacteria by targeting the element of a bacterial genome that is most highly conserved within a species: the 16S ribosomal RNA (rRNA) subunit. At present, 16S PCR amplification from tissue specimens can be performed by specialty laboratories, though its sensitivity and clinical utility to date have remained somewhat limited because, for instance, of inhibitory molecules often found in clinical tissue samples that prevent reliable, sensitive PCR amplification. As such barriers are reduced through technological advances and as the causes of culture-negative infection are clarified (perhaps in part through sequencing efforts), these tests may become both more accessible and more helpful.
With the wealth of sequencing data now available, other regions beyond 16S rRNA can be targeted for bacterial species identification. These other genomic loci can provide additional information about a clinical isolate that is relevant to patient management. For instance, detection of the presence—or potentially even the expression—of toxin genes such as those for C. difficile toxins A and B or Shiga toxin may provide clinicians with additional information that will help distinguish commensals or colonizing bacteria from pathogens and thus aid in prognostication as well as diagnosis.
While amplification tests such as PCR exemplify one approach to nucleic acid detection, other approaches exist, including detection by hybridization. Although not currently used in the clinical realm, techniques for detection and identification of pathogens by hybridization to microarrays are being developed for other purposes. Of note, these different detection techniques require different degrees of conservation. Highly sensitive amplification methods require a high degree of sequence identity between PCR primer pairs and their short, specific target sequences; even a single base-pair mismatch (particularly near the 3′ end of the primer) may interfere with detection. In contrast, hybridization-based tests are more tolerant of mismatch and thus can be used to detect important regions that may be less precisely conserved within a species, thus potentially allowing detection of clinical isolates from a given species with greater diversity between isolates. Such assays take advantage of the predictable binding interactions of nucleic acids. The applicability of hybridization-based methods toward either DNA or RNA opens up the possibility of expression profiling, which can uncover phenotypic information from nucleic acid content.
Both PCR and hybridization methods target specific, known organisms. At the other extreme, as sequencing costs and turnaround times decrease, direct metagenomic sequencing from patient samples is becoming increasingly feasible. This shotgun sequencing approach is unbiased—i.e., is able to detect any microbial sequence, however divergent or unexpected. This new approach brings its own set of challenges, however, including the need to recognize pathogenic sequences against a background of expected host and commensal sequences and to distinguish true pathogens from either colonizers or laboratory contaminants. In a powerful example of this new frontier of sequencing-based clinical diagnosis, investigators diagnosed neuroleptospirosis in a child with an unexplained encephalitis syndrome by finding sequences corresponding to the Leptospira genus in cerebrospinal fluid from the patient. Rapid (<48-h) sequencing and analysis informed the patient’s care in real time, leading to life-saving targeted antibiotic therapy for an unexpected diagnosis that was impossible to make through standard laboratory testing. The diagnosis was retrospectively confirmed through both convalescent serologies and PCR using primers designed on the basis of sequencing data.
In addition to clinical diagnostic applications, novel genomic technologies, including whole-genome sequencing, are being applied to clinical research specimens with a goal of identifying new pathogens in a variety of circumstances. The tremendous sensitivity and unbiased nature of sequencing is also ideal in searching clinical samples for unknown or unsuspected pathogens.
Causal inference in infectious diseases has progressed since the time of Koch, whose historical postulates provided a rigorous framework for attributing a disease to a microorganism. According to an updated version of Koch’s postulates, an organism, whether it can be cultured or not, should induce disease upon introduction into a healthy host if it is to be implicated as a causative pathogen. Current sequencing technologies are ideal for advancing this modern version of Koch’s postulates because they can identify candidate causal pathogens with unprecedented sensitivity and in an unbiased way, unencumbered by limitations such as culturability. Yet, as direct sequencing on primary patient samples greatly expands our ability to recognize associations between microbes and disease states, critical thinking and experimentation will continue to be vital to establishing causality.
Virus discovery in particular has been greatly facilitated by new nucleic acid technology. These frontiers were first notably explored with high-density microarrays containing spatially arrayed sequences from a phylogenetically diverse collection of viruses. Although biased toward those with homology to known viruses, novel viruses in clinical samples were successfully identified on the basis of their ability to hybridize to these prespecified sequences. This methodology famously contributed to identification of the coronavirus causing severe acute respiratory syndrome (SARS). Once discovered, this SARS coronavirus was rapidly sequenced: the full genome was assembled in April 2003, less than 6 months after recognition of the first case. This accomplishment illustrated the advancing power and speed of new diagnostic technologies.
With the advent of next-generation sequencing, unbiased pathogen discovery is now possible through a process known as metagenomic assembly (Fig. 146-3). Sequences of random nucleotide fragments can be generated from clinical specimens with no a priori knowledge of pathogen identity through a process called shotgun sequencing. This collection of sequences can then be computationally aligned to host (i.e., human) sequences, with aligned sequences removed and remaining sequences compared with other known genomes to detect the presence of known microorganisms. Sequence fragments that remain unaligned suggest the presence of an additional organism that cannot be matched to a known, characterized genome; these reads can be assembled into contiguous nucleic acid stretches that can be compared to known sequences to construct the genome of a potentially novel organism. Assembled genomes (or parts of genomes) can then be compared to known genomes to infer the phylogeny of new organisms and identify related classes or traits. Thus, not only can this process identify unanticipated pathogens; it can even identify undiscovered organisms. Some early applications of sequencing on clinical samples have centered around the discovery of novel viruses, including such emerging pathogens as West Nile virus, SARS coronavirus, and the Middle East respiratory syndrome coronavirus (MERS-CoV) that has caused severe respiratory illnesses in healthy adults, as well as viral causes of myriad other conditions, from tropical hemorrhagic fevers to diarrhea in newborns.
Workflow of metagenomic assembly for pathogen discovery. DNA is isolated from a specimen of interest (e.g., tissue, body fluid) containing a mixture of host DNA and nucleic acids from coexisting microbes, either commensal or pathogenic. All DNA (and RNA if a reverse transcription step is added) is then sequenced, yielding a mixture of DNA sequence fragments (“reads”) from organisms present. These reads are then aligned to existing reference genomes for the host or any known microbes, leaving reads that do not align (“map”) to any known sequence. These unmapped reads are then computationally assembled de novo into the largest contiguous stretches of DNA possible (“contigs”), representing fragments of previously unsequenced genomes. These genome fragments (contigs) are then mapped onto a phylogenetic tree based on their sequence. Some may represent known but as-yet-unsequenced organisms, while others will represent novel species. (Figure prepared with valuable input from Dr. Ami S. Bhatt, personal communication.)
More recently, metagenomic assembly has been successfully extended to bacterial pathogen discovery. Investigators identified a new bacterial species associated with “cord colitis”—a rare antibiotic-responsive, culture-negative colitis in recipients of umbilical cord-blood stem cells—by sequencing colon biopsy samples from affected patients and matched controls. A single dominant species emerged from metagenomic assembly in samples from patients that was absent from control samples. The presence of this species was confirmed by PCR and fluorescence in situ hybridization on primary tissue samples. On the basis of its similarity to other known species, the organism was named Bradyrhizobium enterica, a novel species from a genus that has proved difficult to culture and thus would have been hard to identify by other means. Correlation versus causation remains an open question; therefore, further efforts will be required to make such links.
As metagenomic sequencing and assembly techniques become more robust, this technology holds great promise for identifying microorganisms that are associated with clinical conditions of unknown etiology. Conventional methods already have unexpectedly linked numerous conditions with specific agents of infection—e.g., cervical and oropharyngeal cancers with human papillomavirus, Kaposi’s sarcoma with human herpesvirus 8, and certain lymphomas with Epstein-Barr virus. Sequencing techniques offer unprecedented sensitivity and specificity for identifying foreign nucleic acid sequences that may suggest other conditions—from malignancies to inflammatory conditions to unexplained fevers or other clinical syndromes—associated with organisms from viruses to bacteria to parasites. As sequencing-based discovery expands, microbes may be found to be associated with conditions not classically thought of as infectious. Studies of bowel flora in laboratory animals and even humans are already beginning to suggest correlations between microbe composition and various aspects of metabolic and cardiovascular health. Improved methods for pathogen detection will continue to uncover unexpected correlations between microbes and disease states, but the mere presence of a microbe does not establish causality. Fortunately, once the relatively laborious and computationally intensive metagenomic sequencing and assembly efforts have identified a pathogen, further detection can easily be undertaken with targeted methods such as PCR or hybridization, which are much more straightforward and scalable. This capacity should facilitate the additional careful investigation that will be required to progress beyond correlation and to draw causal inference.
At present, antibiotic resistance in bacteria and fungi is determined by isolating a single colony from a cultured clinical specimen and testing its growth in the presence of a drug. The requirement for multiple growth steps in these conventional assays has several consequences. First, only culturable pathogens can be readily processed. Second, this process requires considerable infrastructure to support the sterile environment required for culture-based testing of diverse organisms. Finally, and perhaps most significantly, even the fastest-growing organisms require 1–2 days of processing for identification and 2–3 days for determination of susceptibilities. Slower-growing organisms take even longer: for instance, weeks must pass before drug-resistant M. tuberculosis can be identified by growth phenotype. Given the clinical imperative in serious illness to begin effective therapy early, this inherent delay in susceptibility determination has obvious implications for empirical antibiotic use: broad-spectrum antibiotics often must be chosen up front in situations where it is later shown that preferred narrower-spectrum drugs would have been effective or even that no antibiotics were appropriate (i.e., in viral infections). With this strategy, the empirical choice can be incorrect, often with devastating consequences. Real-time identification of the infecting organism and information on its susceptibility profile would guide initial therapy and support judicious antibiotic use, ideally improving patient outcomes while aiding in the ever-escalating struggle with antibiotic resistance by reserving the use of broad-spectrum agents for cases in which they are truly needed.
Molecular diagnostics and sequencing offer a way to accelerate detection of a pathogen’s antibiotic susceptibility profile. If a genotype that confers resistance can be identified, this genotype can be targeted for molecular detection. In infectious disease, this approach has most convincingly come to fruition for HIV (Fig. 146-2A). (In a conceptually parallel application of genomic analysis, molecular detection of certain resistance determinants in cancers is beginning to inform chemotherapeutic selection.) Extensive sequencing of HIV strains and correlations drawn between viral genotypes and phenotypic resistance have delineated the majority of mutations in key HIV genes, such as reverse transcriptase, protease, and integrase, that confer resistance to the antiretroviral agents that target these proteins. For instance, the single-amino-acid substitution K103N in the HIV reverse transcriptase gene predicts resistance to the first-line nonnucleoside reverse transcriptase inhibitor efavirenz, and its detection thus informs a clinician to choose a different agent. The effects of these common mutations on HIV susceptibility to various drugs—as well as on viral fitness—are curated in publically available databases. Thus, genotypes are now routinely used to predict drug resistance in HIV, as phenotypic resistance assays are far more cumbersome than targeted sequencing. Indeed, current recommendations in the United States are to sequence virus from a patient’s blood before initiating antiretroviral therapy, which is then tailored to the predicted resistance phenotype. As new targeted therapies are introduced, this targeted sequencing–based approach to drug resistance will likely prove important in other viral infections (e.g., hepatitis C).
For several reasons, the challenge of predicting antibiotic susceptibility from genotype has not yet been met in bacteria to the same degree as in HIV. In general, bacteria have evolved diverse resistance mechanisms to most antibiotics; thus, the task cannot be reduced to probing for a single genetic lesion, target, or mechanism. For instance, at least five distinct modes of resistance to fluoroquinolones are known: reduced import, increased efflux, mutated target sites, drug modification, and shielding of the target sites by expression of another protein. Further, we lack a comprehensive compendium of genetic elements conferring resistance, and new mechanisms and genes emerge regularly in the face of antibiotic deployment. As bacteria have far more complex genomes than viruses, with thousands of genes on their chromosomes and the capacity for acquiring many more through horizontal gene transfer of plasmids and mobile genetic elements within and even between species, the task of not only defining all current but also predicting all future mechanisms at a genetic level is daunting or perhaps impossible.
Despite these challenges, in a few select cases where the genotypic basis for resistance has been well defined, genotypic assays for antibiotic resistance are already being introduced into clinical practice. One important example is the detection of methicillin-resistant Staphylococcus aureus (MRSA). S. aureus is one of the most common and serious bacterial pathogens of humans, particularly in health care settings. Resistance to methicillin, the most effective antistaphylococcal antibiotic, has become very common even in community-acquired strains. The alternative to methicillin, vancomycin, is effective against MRSA but measurably inferior to methicillin against methicillin-susceptible S. aureus (MSSA). Analysis of clinical MRSA isolates has demonstrated that the molecular basis for resistance to methicillin in essentially all cases stems from the expression of an alternative penicillin-binding protein (PBP2A) encoded by the gene mecA, which is found within a transferable genetic element called mec. This mobile cassette has spread rapidly through the S. aureus population via horizontal gene transfer and selection from widespread antibiotic use. Because resistance is essentially always due to the presence of the mec cassette, MRSA is amenable to molecular detection. In recent years, a PCR test for the presence of the mec cassette, which saves hours to days compared with standard culture-based methods, has been approved by the U.S. Food and Drug Administration.
Additional molecular diagnostics are being implemented in the evaluation of bacterial antibiotic resistance. Vancomycin-resistant enterococci (VRE) harbor one of a limited number of van genes responsible for resistance to this important antibiotic by altering the mechanism for cell wall cross-linking that vancomycin inhibits. Detection of one of these genes by PCR indicates resistance. Identification of two carbapenemase-encoding plasmids—NDM-1 and KPC—responsible for a significant fraction of carbapenem resistance (though not for all such resistance) has led to the development of a multiplexed PCR assay to detect these important resistance elements. Because carbapenems are broad-spectrum antibiotics frequently reserved for multidrug-resistant bacteria (particularly enteric gram-negative bacilli) and are often used as antibiotics of last resort, the initial appearance of resistance and the subsequent increase in its prevalence have caused considerable concern. Therefore, even though other mechanisms for carbapenem resistance exist, a rapid PCR test for the two plasmids encoding these two carbapenemases has been developed to aid in both diagnostic and infection control efforts. Efforts are under way to extend this multiplexed PCR assay to other plasmid-borne carbapenemases and thus to make it more comprehensive.
The power and application of molecular genetic tests are not limited to high-income settings. With the increasing burden of drug-resistant tuberculosis in the developing world, a molecular diagnostic test has now been developed to detect rifampin-resistant tuberculosis. The genetic basis for rifampin resistance has been well defined by targeted sequencing: characteristic mutations in the molecular target of rifampin, RNA polymerase, account for the vast majority of rifampin-resistant strains of M. tuberculosis. A PCR assay that can detect both the M. tuberculosis organism and a rifampin-resistant allele of RNA polymerase from clinical samples has recently been approved. Since rifampin resistance frequently accompanies resistance to other antibiotics, this test can suggest the possible presence of multidrug-resistant M. tuberculosis within hours instead of weeks.
Despite differences in relative genome complexity, HIV genotypic screening for antiretroviral resistance offers one framework for broadening efforts at genotypic assays for nonviral antibiotic resistance. As whole-genome pathogen sequencing has become increasingly feasible and inexpensive (Fig. 146-1), significant efforts have been launched to sequence hundreds to thousands of antibiotic-sensitive and -resistant isolates of a given pathogen in order to more comprehensively define resistance-conferring genetic elements. In parallel with advancing sequencing technologies, progress in computational techniques, bioinformatics and statistics, and data storage as well as experimental confirmatory testing of hypotheses will be needed to move toward the ambitious goal of a comprehensive compendium of antibiotic resistance determinants. Open sharing and careful curation of new sequence information will be of paramount importance.
Yet no matter how thorough and carefully curated such a genotype-phenotype database is, history suggests that comprehensively cataloguing resistance in nonviral pathogens, with new mechanisms continuously emerging, will be challenging at best. Even identifying and itemizing current resistance mutations is a daunting prospect: nonviral genomes are much larger than viral ones, and their abundance and diversity are such that hundreds to thousands of genetic differences often exist between clinical isolates, of which perhaps only one may cause resistance. For example, increasing resistance to artemisinin, one of the most effective new agents for malaria, has prompted recent large-scale efforts to identify the basis for resistance. While such studies have identified promising leads, no clear mechanism has emerged; in fact, a single genetic lesion alone may not fully account for resistance. Especially with multiple possible resistance mechanisms for a given antibiotic as well as ongoing evolutionary pressure resulting in the development and acquisition of new modes of resistance, a genotypic approach to diagnosing antibiotic resistance is likely to be imperfect.
We have already observed the accumulation of new or unanticipated modes of resistance from ongoing evolutionary pressure caused by the widespread clinical use of antibiotics. Even with MRSA, perhaps the best-studied case of antibiotic resistance and a model of relative simplicity with a single known monogenic resistance determinant (mecA), a genotype-based approach to resistance detection proved flawed. One limitation was a recall of the initial commercial genotypic resistance assay that was deployed for the identification of MRSA. A clinical isolate of S. aureus emerged in Belgium that expressed a variant of the mec cassette not detected by the assay’s PCR primers. New primers were added to detect this new variant, and the assay was re-approved for use. More recently, an even more divergent but functionally analogous gene called mecC, which confers methicillin resistance but evades PCR detection by this assay, was found. This series of events exemplifies the need for ongoing monitoring of any genotypic resistance assay. A second limitation is that a contradiction can occur between genotypic and phenotypic evidence for resistance. Up to 5% of MSSA strains carry a copy of the mecA gene that is either nonfunctional or not expressed. Thus, the erroneous identification of these strains as MRSA by genotypic detection would lead to administration of the inferior antibiotic vancomycin rather than the preferred β-lactam therapy.
These examples illustrate one of the prime challenges of moving beyond growth-based assays: genotype is merely a proxy for the resistance phenotype that directly informs patient care. One alternative approach currently under development attempts to circumvent the limitations of genotypic resistance testing by returning to a phenotypic approach, albeit one informed by genomic methods: transcriptional profiles serve as a rapid phenotypic signature for antibiotic response. Conceptually, since dying cells are transcriptionally distinct from cells fated to survive, susceptible bacteria enact different transcriptional profiles after antibiotic exposure that are different from the profiles of resistant strains, independent of the mechanism of resistance. These differences can be measured and, since transcription is one of the most rapid responses to cell stress (minutes to hours), can be used to determine whether cells are resistant or susceptible much more rapidly than is possible if one waits for growth in the presence of antibiotics (days). Like DNA, RNA can be readily detected through predictable rules governing base pairing via either amplification or hybridization-based methods. Changes in a carefully selected set of transcripts form an expression signature that can represent the total cellular response to antibiotic without requiring full characterization of the entire transcriptome. Preliminary proof-of-concept studies suggest that this approach may identify antibiotic susceptibility on the basis of transcriptional phenotype much more quickly than is possible with growth-based assays.
Because of its sensitivity in detecting even very rare nucleic acid fragments, sequencing is now permitting studies of unprecedented depth into complex populations of cells and tissues. The strength of this depth and sensitivity applies not only to the detection of rare, novel pathogens in a sea of host signal but also to the identification of heterogeneous pathogen subpopulations in a single host that may differ, for example, in drug resistance profiles or pathogenesis determinants. Future studies will be needed to elucidate the clinical significance of these variable subpopulations, even as deep sequencing is now providing unprecedented levels of detail about majority and minority members of this population.
While pathogen-based diagnostics continue to be the mainstay for diagnosing infection, serologic testing has long been the basis of a strategy to diagnose infection by measuring host responses. Here, too, the application of genomics is now being explored to improve upon this approach, given the previously described limitations of serologic testing. Rather than using antibody responses as a retrospective biomarker for infection, recent efforts have focused on transcriptomic analysis of the host response as a new direction with diagnostic implications for human disease. For instance, while pathogen-based diagnostic tests to distinguish active from latent tuberculosis infection have proved elusive, recent work shows that the transcriptional profile of circulating white blood cells exhibits a differential pattern of expression of nearly 400 transcripts that distinguish active from latent tuberculosis; this expression pattern is driven in part by changes in interferon-inducible genes in the myeloid lineage. In a validation cohort, this transcriptional signature was able to distinguish patients with active versus latent disease, to distinguish tuberculosis infection from other pulmonary inflammatory states or infections, and to track responses to treatment in as little as 2 weeks, with normalization of expression toward that of patients without active disease over 6 months of effective therapy. Such a test could play an important role not only in the management of patients but also as a marker of efficacy in clinical trials of new therapeutic agents. Similarly, other investigators have been trying to identify host transcriptional signatures in circulating blood cells that are distinct in influenza A infection from those in upper respiratory infections caused by certain other viruses or bacteria. These signatures also varied with phase of infection and showed promise in distinguishing exposed subjects who will become symptomatic from those who will not. These results suggest that profiling of host transcriptional dynamics could augment the information obtained from studies of pathogens, both enhancing diagnosis and monitoring the progression of illness and the response to therapy.
In this era of genome-wide association studies and attempts to move toward personalized medicine, genomic approaches are also being applied to the identification of host genetic loci and factors that contribute to infection susceptibility. Such loci will have undergone strong selection among populations in which the disease is endemic. By identifying the beneficial genetic alleles among individuals who survive in such settings, markers for susceptibility or resistance are being discovered; these markers can be translated into diagnostic tests to identify susceptible individuals in order to implement preventive or prophylactic interventions. Further, such studies may offer mechanistic insight into the pathogenesis of infection and inform new methods of therapeutic intervention. Such beneficial genetic associations were recognized long before the advent of genomics, as in the protective effects of the negative Duffy blood group or heterozygous hemoglobin abnormalities against Plasmodium infection. Genomic methods enable more systematic and widespread investigations of the host to identify not only people with altered susceptibility to numerous diseases (e.g., HIV infection, tuberculosis, and cholera) but also host factors that contribute to and thus might predict the severity of disease.