Emerging Technologies for the Clinical Microbiology Laboratory

Nucleic acid amplification using thermostable polymerase (PCR) was initially reported in 1988, and this method remains largely unchanged as it forms the backbone of molecular diagnostics in clinical microbiology laboratories today (6). Properties such as high sensitivity and specificity, an extremely low limit of detection (1 to 10 copies of the target), and rapid results have led to proposed changes in the definition of the “gold standard” method for detection and identification of microorganisms in clinical specimens, especially for those that are difficult to culture, including fastidious bacterial or viral pathogens (7 – 10). While the basic principle of nucleic acid amplification tests (NAATs) has not changed, technologies surrounding this core, including amplification strategy, amplicon detection, multiplexing of reactions, and automation of the entire process into sample-to-result platforms, have provided a large menu of options for the molecular microbiology laboratory to choose from. One such modification is the departure from PCR-based amplification to transcription-mediated amplification (TMA) of a nucleic acid target. This method differs from PCR in that the target sequence is typically an RNA molecule (mRNA or rRNA), which may be present at a high copy number in the cell. Reverse transcriptase and engineered oligonucleotide primers are used to simultaneously generate a cDNA template and incorporate a promoter sequence recognized by a highly efficient, phage-encoded RNA polymerase enzyme. This enzyme enables isothermal synthesis of 100 to 1,000 copies of each starting template cDNA, which are in turn used as the template for subsequent rounds of amplification (11) (Fig. 1). The multicopy nature of the RNA target and ability to amplify beyond a log-linear rate without the need for thermocycling theoretically increase the speed and sensitivity of TMA compared to that of standard PCR. To date, the most widely used molecular assays target a single or few analytes, employing one or few oligonucleotide primer sets (11 – 13). Using target amplification coupled with fluorescence probe-based detection, these tests provide a mechanism for rapid and sensitive diagnostic tests.

The majority of molecular tests in use today are qualitative tests. Qualitative tests are best suited for the detection of microorganisms in specimens whose presence, at any level, is associated with a disease state. This includes microorganism that are not regarded as normal flora, as well as any organism isolated from a sterile site. A prime illustration is the use of NAATs for the detection of microorganisms associated with sexually transmitted illnesses, including Neisseria gonorrhoeae, Chlamydia trachomatis, Trichomonas vaginalis, and Mycoplasma genitalium. Culture of these organisms is either impractical or unreliable due to loss of viability during transport, which further decreases the sensitivity of culture methods. NAATs have demonstrated increased sensitivity compared to that of culture methods and dramatically reduced turnaround time for detection of these pathogens (12, 14 – 16). This enables more rapid, accurate identification of the pathogen(s) responsible for nonspecific symptoms of urethritis or pelvic inflammatory disease and also may aid in limiting the spread of these organisms by identifying asymptomatic carriers. Additionally, the increased sensitivity of NAATs can enable the analysis of specimens obtained by less invasive techniques or of patient-collected specimens, including urine and self-collected vaginal swabs, without affecting the accuracy of the test (12, 13, 17 – 19). The ability to use these types of specimens can contribute to higher participation in routine screening exams (12, 13, 17 – 19). Other pathogens commonly identified using qualitative NAATs include respiratory viruses, herpesviruses, Clostridium difficile, Staphylococcus aureus, methicillin-resistant S. aureus (MRSA), Streptococcus pyogenes, Streptococcus agalactiae, Bordetella pertussis, and bacterial pathogens associated with atypical pneumonia. Another use of qualitative tests is to obtain a rapid result for preoperative screening or for infection control purposes. A recent randomized trial compared targeted screening and decolonization of intensive care unit (ICU) patients to a universal decolonization program to reduce the rate of MRSA infection in hospital ICUs (20). While universal decolonization of all patients was associated with the lowest hazard ratio for infection (0.62), targeted screening and decolonization also demonstrated a reduced hazard ratio (0.75). Although screening and targeted decolonization of patients may not be as effective as universal decolonization, studies have demonstrated that sensitive detection of MRSA can significantly reduce the rate of postsurgical infection by identifying those patients who will benefit from preoperative prophylaxis and decolonization (21, 22). As a result, reduced rates of postsurgical infection resulting from molecular screening methods have been shown to reduce the cost of health care to both the hospital and third-party payer (23). Likewise, rapid and accurate detection of MRSA, vancomycin-resistant Enterococcus (VRE), or carbapenem-resistant Enterobacteriaceae (CRE) may aid infection control efforts by identifying those patients requiring contact isolation.

A potential drawback to the use of NAATS is the interpretation of positive results from asymptomatic patients or those who have received appropriate therapy following an initially positive result. While other technologies, including direct microscopy and antigen-based tests, are not immune from this shortcoming, the exquisite sensitivity of the PCR and TMA-based methods used for qualitative NAATs makes these methods most susceptible to potential overreporting of positive results. For these assays, any amount of nucleic acid detected in a specimen is reported as positive, regardless of whether it represents an infectious process due to a live organism, low-level or asymptomatic colonization, or free nucleic acid in the absence of a viable organism. This concern has been highlighted recently by several publications focused on selection of the most appropriate test or algorithm for diagnosis of C. difficile infection (24, 25). Because high rates of asymptomatic carriage of C. difficile are reported among elderly residents in long-term care facilities, it has been proposed that positive NAATs be followed by a direct test for the presence of C. difficile toxins (tcdA and tcdB) to differentiate between carriage and infection (26, 27). Supporting this notion, detection of toxin from patients following a positive NAAT has been correlated with worse prognosis than for patients with a positive NAAT alone (28). Furthermore, NAATs for C. difficile were positive up to 4 weeks following appropriate antibiotic treatment and resolution of symptoms in >50% of patients tested (29). These concerns were addressed directly in a study which demonstrated significantly reduced specificity of molecular modalities when patient symptoms were included as criteria for “gold standard” positive results (30). An excellent review of the diagnostic assays available and difficulties in interpretation of results pertaining to C. difficile has been published (31). Similarly, “pseudo-outbreaks” of Bordetella pertussis have been reported due to the use of NAATs that target the multicopy IS481 chromosomal element (32, 33). In both pseudo-outbreaks, NAAT-positive results could not be confirmed when using an alternative NAAT targeting a single copy genetic target, and 92% to 100% of patients did not meet clinical criteria for pertussis, or were seronegative for antipertussis toxin IgG (32). The effect of these “false-positive” results was the unnecessary prescription of antimicrobial therapy for the patient as well as close contacts and temporary isolation of patients, which constitute a needless financial and social burden on those affected (32). It has been established that NAATs that target the IS481 gene in B. pertussis are capable of detecting <1 organism per sample and that detection at PCR cycle thresholds of >35 has <50% correlation with clinical pertussis disease (33, 34). Therefore, it may be useful to incorporate clinical symptoms and results of other testing modalities when defining a positive cycle threshold for molecular tests (34, 35).

In both examples, positive results may have been due to persistence of nucleic acid or nonviable organism in the specimen. This reinforces the point that molecular assays should be interpreted in the context of clinical presentation and should not be used as a test of cure.

To maximize the benefits of molecular testing, developers of diagnostics have begun to focus on technologies that employ both simplified technology and simplified specimen preparation in an attempt to bring molecular assays closer to the patient. These technologies have the potential to further reduce TAT, which may positively impact patient care and reduce the overall cost of health care. Isothermal amplification methods, including loop-mediated isothermal amplification (LAMP) and helicase-dependent amplification (HDA), effectively remove the need for expensive thermocyclers and technical optimization of cycling conditions. These methods can be coupled to alternative detection technologies (i.e., fluorescent probe-independent detection methods) that eliminate the need sophisticated optics. This further reduces the cost of instrumentation and enables these tests to be used outside today's “molecular laboratory” and closer to the point of care (POC).

LAMP utilizes 4 primers and 6 recognition (annealing) sites per target to create high levels of amplicon in <60 min. An “inner” set of primers initiates target amplification, while a second, “outer” set of primers initiates a round of replication that displaces the initial product, thus regenerating a single-strand template without the need for heat denaturation (36) (Fig. 2). The use of 6 primers and 4 recognition sites provides specificity higher than that of standard PCRs that utilize only 2 primers. The increased specificity eliminates the need for expensive fluorescence-labeled probes and accompanying optics and allows detection of amplified product based on by-products of DNA replication (37). Pyrophosphate ion, generated by target amplification, can be precipitated by the addition of magnesium ion to the reaction mixture. This enables visual inspection of the reaction tube for turbidity as an indication of a positive result. An increase in the turbidity of the reaction mixture can also be measured in real time using comparatively simple optics to permit the use of LAMP in quantitative assays (38). There are a number of FDA-cleared and laboratory-developed tests that utilize the LAMP technology. FDA-cleared tests utilizing LAMP include those for C. difficile, group A and B Streptococcus, Mycoplasma pneumoniae, and B. pertussis (39, 40). Clinical evaluations of the C. difficile and group A Streptococcus tests have demonstrated sensitivity and specificity similar to those of traditional real-time PCR, though a slight decrease in sensitivity for C. difficile has been noted (39 – 44). Laboratory-developed and commercially available research-use-only (RUO) tests using LAMP have targeted diverse groups of microorganisms, including Plasmodium spp., Giardia lamblia, Leishmania, Mycobacterium spp., and hepatitis viruses (45 – 50). Specifically, LAMP-based testing for Plasmodium spp. and Plasmodium falciparum demonstrated >97% sensitivity and >98% specificity compared to nested PCR in patients with parasitemia of ≥1 parasite/μl and was significantly more sensitive than standard microscopy (49, 50). The use of heat-treated whole blood rather than extracted nucleic acid, a simple heat block or water bath to maintain 60 to 65°C for isothermal target amplification, and visual determination of a positive result based on turbidity give LAMP an advantage over traditional PCR methods in resource-limited regions of the world, including many countries where malaria is endemic (51, 52). Further, the use of a pocket warmer (exothermic chemical reaction pouch) to drive LAMP maintained 90.5% sensitivity for detection of Mycobacterium ulcerans compared the same test run using a powered heat block (48). A major limitation of LAMP is the inability to multiplex. This is due to the nonspecific and indirect turbidity-based detection of the amplicon. Still, the noted advantages of inexpensive reagents, simple instrumentation, and “moderate complexity” designation make LAMP technology an emerging player in the field of molecular diagnostics.

Helicase-dependent amplification (HDA) is another isothermal amplification technology that could be adapted to point-of-care testing. This technology utilizes UvrD (DNA helicase) and MutL enzymes isolated from E. coli and single-strand binding proteins to create and maintain a single-stranded template for primer annealing and subsequent rounds of amplification (53) (Fig. 3). An initial heat-based denaturation is required for optimal efficiency; however, reliance on a single reaction temperature without initial denaturation maintains 40% to 60% efficiency and is adequate to generate sufficient amplicon for endpoint detection assays (53). Like LAMP, the isothermal amplification can be carried out using simple instrumentation in the absence of electricity (54). HDA has been applied to identification of C. difficile, Plasmodium spp., and S. aureus (55, 56). An advantage of HDA is that detection of target can be achieved by incorporation of fluorescein or digoxigenin into the amplicon, followed by capture and visualization of the amplicon as a colored line on an enzyme immunoassay (EIA) lateral-flow strip (56 – 58). This maintains the ability to utilize these assays without sophisticated instrumentation but also allows the detection of multiple targets in a single reaction. A test developed to detect and differentiate herpes simplex virus 1 (HSV-1) and HSV-2 using this approach has demonstrated 100% sensitivity compared to viral culture, with a limit of detection as low as 5.5 copies per reaction (59). Further, this test could be performed on oral and genital cutaneous or mucocutaneous sources without the need for nucleic acid extraction and could be completed within 75 min.

Qualitative NAATs vary widely in the level of automation, ranging from largely manual (offline nucleic acid extraction, manual preparation of master mix, and addition of template) to fully automated “sample-to-result” platforms (Table 1). Full automation is typically focused on high-volume or screening tests such as those used for N. gonorrhoeae and C. trachomatis, C. difficile, MRSA, VRE, and HSV. These highly automated sample-to-result platforms decrease technologist hands-on time and may provide more consistency by reducing the risk of cross-contamination of specimens, pipetting error, or other preanalytic errors attributable to human labor. Despite these obvious advantages, there are still impediments to maximizing the value of molecular testing when using these systems. Until recently, the majority of molecular tests have been considered “high complexity” and as such have been confined to molecular laboratories staffed with skilled technologists. This requires that specimens be transported to the laboratory for analysis. For inpatients, the delay resulting from transport of a specimen may not be significant; however, for outpatient clinics, the time between collection of the specimen and receipt by the laboratory may be several hours. This delay due to transport abrogates one of the key advantages of molecular tests, namely, rapid TAT. Additionally, some molecular assays are best suited for batch testing due to multistep processing or efficiency factors related to batching of specimens on automated platforms. Finally, the large capital expenditure for high-capacity fully automated instruments must be considered.

The trends toward consolidation/centralization of laboratories and bundled care reimbursement structures favors highly automated systems with large-throughput batch processing of specimens to achieve a low cost per test (60). Systems like the m2000 (Abbott), and Cobas AmpliPrep (Roche) feature a two-step system whereby automated nucleic acid extraction is followed by automatic addition of all reagents required for an RT-PCR on one instrument. These instruments can process up to 96 specimens per run; however, prepared specimens must be physically moved to a thermocycler within 30 to 150 min to complete analysis of the specimen. The need for human intervention and a narrow window for transfer of specimens to a thermocycler limit the walkaway capability and present difficulty for laboratories not well staffed on all shifts. Other batch-type platforms such as the BD Max and BD Viper (BD), Tigris (Hologic Gen-Probe), and Cobas AmpliPrep/Cobas TaqMan system (Roche) are true sample-to-result platforms. Most of these platforms are classified as high-complexity molecular assays; however, the BD Max offers FDA-cleared moderate-complexity in vitro diagnostic (IVD) tests as well. These systems incorporate thermocyclers capable of RT-PCR and result reporting along with sample preparation. In addition to simplifying workflow, sample-to-result instruments may also reduce contamination or labeling errors by reducing the number of times that specimens are manipulated by technologists. A major disadvantage of batch platforms is the delay in availability of results compared to on-demand NAATs. In the case of outpatient surgeries, some institutions maintain presurgical clinics scheduled 1 to 2 weeks prior to the scheduled surgery, while in other institutions more than 80% of patients may be admitted on the day of surgery (21). In these cases, a point-of-care or on-demand test may be a better solution to benefit the patient rather than batched molecular assays. For example, real-time on-demand screening for colonization with MRSA or VRE could potentially alter presurgery prophylaxis or infection control measures (20 – 22, 61).

The advantages of point-of-care (POC) testing have reviewed by Robinson et al. and include a reduction in repeat and unnecessary test orders, a reduced length of stay, and shorter times to appropriate therapy; however, the authors acknowledge the lack of published studies objectively examining quantifiable outcomes related to the use of POC testing (60). There are several on-demand sample-to-result molecular testing platforms, including the GeneXpert (Cepheid, Sunnyvale, CA), Verigene (Nanosphere, Northbrook, IL), Portrait (Great Basin, Salt Lake City, UT), and FilmArray (BioFire, Salt Lake City, UT) (Table 1). Currently, these platforms are best suited to the laboratory; however, movement toward use as point-of-care (POC) tests is being pursued through miniaturization, automation, and simplification of the testing process. Other platforms, including Illumigene (Meridian Bioscience, Cincinnati, OH), and AmpliVue (Quidel Molecular, San Diego, CA) lack automation but have been simplified to potentially enable use as POC or “near-POC” diagnostics. Fully automated on-demand or single-test formats are often significantly more expensive on a per-test basis than batched testing formats; however, the rapid results provided by these systems often enable patient management decisions that can reduce the total cost of care. Many of the studies that demonstrate this principal utilize on-demand molecular tests for the identification of S. aureus and MRSA in positive blood culture broths. One such study compared cohorts of patients with Gram-positive bloodstream infection (BSIs) pre- and postimplementation of an on-demand molecular test that identified S. aureus and differentiated S. aureus from MRSA. The authors reported a 1.6-day reduction in time to optimal antibiotic therapy and a 6.2-day reduction in hospital stay for the cohort of patients tested using the NAAT (62). This also translated to >US$21,000 reduction in the total cost of care for these patients. In contrast, a similar study conducted using a lower-cost-per-test batch-format NAAT to identify S. aureus and MRSA in positive blood culture broths failed to demonstrate such savings (63). Importantly, failure to actively report laboratory values also decreased the benefits of NAAT results in the latter study. Another area of great interest is the use of rapid and accurate molecular assays for the identification of Mycobacterium tuberculosis in patient specimens (64). The recent availability of an FDA-cleared NAAT (Xpert TB/RIF; Cepheid) for the identification of M. tuberculosis, including strains resistant to rifampin, has prompted studies assessing the cost-effectiveness of such a test. The cost of sputum smear as a primary diagnostic test for patients suspected of having active tuberculosis is <7% the cost of Xpert TB/RIF; however, overall savings of US$2,278 per admission could be realized when considering the reduced occupation of isolation rooms for patients with a negative result (65). Other studies have reported up to a 94% decrease in unnecessary antituberculosis treatment and an average 1.5-month reduction in unnecessary therapy as well as a reduction in time in isolation for patients who were smear positive but culture negative for M. tuberculosis when an NAAT was used (66, 67). Importantly, these data were based on a studies conducted in high-prevalence populations (6% to 37% positive for M. tuberculosis). For hospitals and laboratories serving low-prevalence populations, implementation of a more costly molecular test for all smear-positive specimens may increase the overall cost of care for these patients. In these cases, communication between the laboratory and clinician to establish the patient history and risk of M. tuberculosis may be beneficial to reduce unnecessary cost of a molecular test.

In all cases, to reap the greatest benefit from these technologies, the assays must be able to be conducted and results reported in a true “real-time” 24-h-per-day, 7-days-per-week fashion, or the benefit of rapid TAT to patient care will be lost.

The introduction of platforms equipped with optics capable of excitation and detection of multiple fluorophores in a closed system in real time (real-time PCR [RT-PCR]) made multiplex pathogen detection a simple and viable option for molecular diagnostics in routine clinical laboratories. Laboratory-developed tests (LDTs) have taken advantage of the different probes and platforms in the design of multiplex tests for the detection of a variety of analytes. Only recently have larger multiplex panels begun to be available as FDA-cleared tests for use in clinical diagnostics. These PCR-probe-based tests are typically capable of low-density multiplexing of 4 to 6 unique targets. This limitation is imposed by the number of optical channels and ability to differentiate between fluorescent dyes with similar emission wavelengths. The optics on early platforms, including SmartCycler II (Cepheid, Sunnyvale, CA) and first-generation BD Max (BD, Sparks, MD) were limited to a maximum of 4 channels. Newer platforms, including the GeneXpert (Cepheid), LightCycler 2.0 (Roche, Indianapolis, IN), second-generation BD Max (BD), and ABI 7500 Fast Dx and ABI QuantStudio (ABI, Foster City, CA) are capable of detection in up to 6 different channels. Compared to more recently developed multiplexing technologies, including solid and liquid microarray (discussed below) methods, the ability to multiplex 4 or 6 targets can be a limitation. This is especially true for specimen types in which there are numerous, diverse microorganisms capable of causing similar symptoms or syndromes such as upper respiratory illness, gastroenteritis, or bacterial and fungal sepsis. Despite the limitations in the number of targets that can be detected simultaneously, numerous FDA-cleared tests using these platforms have been favorably evaluated and are applicable in the clinical laboratory.

The SmartCycler II and LightCycler 2.0 are open platforms for RT-PCR. Both require preextraction of nucleic acids to obtain template and manual pipetting of each PCR component or master mix into individual RT-PCR tubes. Multiplex assays using analyte-specific reagents (ASRs) for influenza viruses A and B, respiratory syncytial viruses (RSV) A and B, and HSV-1 and -2 have demonstrated high sensitivities compared to other rapid tests, and results are available days earlier than with viral culture methods (10, 69 – 72). A recently developed and FDA-cleared test for the detection of bacterial causes of enteritis demonstrated 100% sensitivity and >99% specificity for 5 targets (Salmonella spp., Shigella spp., Campylobacter coli/jejuni, stx ₁, and stx ₂) compared to culture and an alternative molecular assay (73). A drawback to this test is the need for offline nucleic acid extraction and the necessity to set up parallel reactions for each specimen to accommodate all 5 assay targets due to limitations of the SmartCycler II optics.

Molecular tests have also been developed for detection of bacterial and fungal pathogens associated with bloodstream infection (BSI) (Table 2). Initial tests were developed using the SmartCycler II or LightCycler 2.0 for low-density multiplexing (63, 74, 75). The SeptiFast assay (Roche) is unique among these tests in that it is intended for use with whole blood specimens prior to broth culture enrichment. This assay has not received FDA clearance for use in the United States. Although run on the LightCycler 2.0, the use of 3 parallel real-time PCRs with different primer/probe combinations and postamplification melt curve analysis expanded the number of bacterial and fungal targets that could be detected using SeptiFast to 20 (75). Importantly, the low number of organisms per ml in direct whole-blood specimens limited the sensitivity to 42% to 79% compared to culture (75 – 77). The specificity of this test was reported as 95.0 to 97.1% in patients without clinical signs of sepsis but was 74% in symptomatic patients (76). A positive SeptiFast result was confirmed by culture in only 67% of specimens; however, the detected organism was recovered from other clinically relevant samples in approximately half of the discordant cases (75). Together this suggests that a NAAT may be more sensitive than culture in patients with clinical symptoms of sepsis; however, additional studies are needed to correlate positive NAAT results with clinical outcomes. Because of the difficulties in molecular analysis of whole blood, more recent molecular tests have focused on analysis of positive blood cultures. The StaphSR test is performed on positive blood cultures containing Gram-positive cocci. This test is designed to detect and differentiate methicillin-susceptible and -resistant strains of S. aureus (63, 74). Initial studies reported sensitivity and specificity for identification of S. aureus of 96.7% to 99.4% and sensitivity for MRSA of 100% (63, 78); however, subsequent studies report sensitivities as low as 50% depending on the type of SCCmec cassette present in circulating strains (79, 80) (Table 2). Another drawback of this assay is the requirement for offline extraction and manual setup of individual RT-PCRs, which lends to batching of specimens. In the case of positive blood cultures, batching of specimens contributes to delays in reporting of results, which can abrogate the benefit that rapid molecular diagnostics can have for patient care (63). Finally, while S. aureus is of major concern in BSIs, it comprises only about 20% of positive cultures (81). The limited number of fluorophores that can be differentiated in a single reaction using standard RT-PCR platforms prevents the inclusion of additional targets required to make this type of test applicable to the majority of positive blood cultures. For laboratories with larger specimen volumes or limited staffing, the offline processing and manual setup of reactions can complicate assay setup and strain resources and may also be a potential source for cross-contamination of specimens.

Miniaturization of singleplex reactions can overcome some of the limitations to traditional PCR-probe-based multiplexing. Conducting singleplex real-time PCR in multiple individual wells enables simultaneous amplification and detection of different targets, but all within a single test device. This can be accomplished using a thermocycler capable of real-time quantitative PCR such as the ABI 7500 FastDx or ABI QuantStudio, which can accommodate 96- or 384-well microplates and can interrogate each well separately. Importantly, these platforms are not sample-to-result platforms, and this approach still requires extraction and manual setup of multiple real-time PCR wells per specimen. In contrast, the FilmArray system (BioFire, Salt Lake City, UT) is a sample-to-result multiplex PCR system contained within a single test pouch. In addition to simplifying workflow, this methodology also enables the assay to be classified as a moderate-complexity IVD test. The clinical specimen is diluted and added directly to a sample port. The specimen then passes through multiple chambers containing reagents for lysis and extraction of nucleic acids from the specimen. Once extracted, the nucleic acids undergo a nested PCR in which the first reaction utilizes degenerate primers to broadly amplify target sequences. Products from the first PCR are then diluted and inoculated into 102 microwells, each of which contains reagents for singleplex amplification and detection of a specific target sequence (82). Each well can be individually interrogated for fluorescence, allowing the use of a single fluorophore for detection of amplicon. Tests using this approach are available or under development for the detection of respiratory viruses, bacteria, and fungi in positive blood cultures and bacterial, viral, and protozoan pathogens in stool (82 – 85).

Many studies have evaluated the FilmArray respiratory panel (RP), and the performance in these studies has been reviewed by Babady (86). In general, evaluation of the FilmArray respiratory assay in adult and pediatric populations has demonstrated 80% to 100% agreement with alternative molecular tests, with notable deficiencies in detection of specific adenoviruses (83, 84, 87, 88). This deficiency has been addressed in a more recent version of the assay (version 1.7), which has demonstrated an increase in sensitivity from 43% to 66% to 88% to 91% for detection of 39 clinically relevant adenovirus serotypes (89). Compared to other molecular tests for respiratory viruses, the FilmArray had the highest cost per test, but this was countered by the full sample-to-result capability, highest number of targets detected (n = 20), and fastest total time to result (1 h) (83). In addition to the relatively high per-test cost, a second potential drawback to the use of the FilmArray as a mainstream method for analysis of respiratory specimens is the limited throughput. Each FilmArray is capable of analyzing only a single specimen per run. This can be a significant bottleneck for larger laboratories, which may receive hundreds of respiratory specimens per day in peak respiratory illness season. Therefore, use of the FilmArray with its broadly inclusive panel may be best suited to critically ill or immunocompromised patients rather than for routine testing of all community patients suffering from respiratory symptoms during “influenza season.”

Initial clinical evaluations of the FilmArray BCID blood culture assay demonstrated overall sensitivity of 91% to 99%, including 98.5%, 96.7%, and 100% for 11 Gram-negative, 8 Gram-positive, and 2 yeast targets, respectively, with specificity of 97% to 100% for each of the individual targets on the panel (85, 90) (Table 2). A potential weakness of the assay is the inclusion of a single “Enterococcus spp.” target which is unable to differentiate between E. faecalis and E. faecium. This distinction can be helpful when considering antimicrobial therapy because of differences in susceptibility patterns between the two species. Specifically, resistance to ampicillin and vancomycin are rare in E. faecalis, 1.3%, and 0.5%, respectively, while 82.4% and 9.6% of E. faecium isolates are resistant to ampicillin and vancomycin, respectively (91) A second potential shortcoming is the failure to reliably detect all components present in polymicrobial cultures. Overall, the FilmArray BCID detected all microorganisms present in just 71% of polymicrobial cultures. While many of these were organisms not present on the BCID panel, E. faecalis was missed in two polymicrobial cultures, while E. coli and a viridans group Streptococcus spp. were missed in two other polymicrobial cultures (85). Finally, while the assay includes a total of 24 genus or species targets commonly associated with bloodstream infection, up to 8% of blood cultures contain organisms not present on the BCID panel (85, 90). Therefore, a primary Gram stain of all positive blood culture broths as well as routine culture of broths which are both positive and negative by BCID is prudent before finalizing results.

Several approaches have been explored to expand the number of targets detectable in a single multiplex nucleic acid test. Collectively, these are referred to as microarrays. Microarrays can be broadly broken into two classes: solid arrays, which rely on spatial detection of targets arranged on a solid surface, and liquid arrays, which utilize target-specific capture probes conjugated to microspheres which can be detected using flow cytometry. For a thorough review of microarray technologies, the reader is referred to the article by Miller and Tang (92). Microarrays are attractive in diagnostics because they can reduce the cost per target tested and allow simultaneous testing for multiple pathogens associated with similar symptoms.

Traditional microarrays are composed of synthetic oligonucleotides or peptides (capture probes) immobilized on a solid substrate such as a glass slide or nitrocellulose membrane. The number of unique capture probes on a single array can range from 100 on low-density printed arrays to >1 million on in situ-synthesized high-density arrays. The probes on high-density arrays are typically shorter (20 to 25 nucleotides [nt]) and are designed to have target redundancy to increase the specificity of target detection (92). Because of the large number of probes, these arrays are most commonly used for whole-genome expression profiling or for other genome-wide comparisons such as mutations or deletions. Low-density arrays consist of longer probes, typically 50 to 800 nucleotides in length, which may be chemically synthesized or created as amplicons by PCR. The use of PCR amplicons and liquid spotting of probes makes this type of array comparatively inexpensive to manufacture. The relatively long length of amplicon probes increases target sensitivity because several polymorphisms can be tolerated during hybridization steps; however, this can also results in decreased specificity for the target (92, 93). Therefore, each probe is typically spotted in replicate on a single array to increase test specificity (92). Each synthesized oligonucleotide or amplicon probe corresponds to a single gene and is spotted or printed to the array solid surface. Inexpensive manufacturing and high sensitivity make low-density printed arrays a reasonable choice for diagnostic tests designed for use in clinical microbiology laboratories.

The commercially available and FDA-cleared Verigene system (Nanosphere, Northbrook, IL) (Fig. 4) has offers microarray-based tests for identification of respiratory viruses (RV+), C. difficile (CDF), blood cultures containing Gram-positive bacteria (BC-GP) or Gram-negative bacteria (BC-GN), and identification of genetic variants, including Factor V Leiden and CYP450 2C19 *2 and *3, which impact patients with coagulation disorders (94 – 102).

The RV+ test simultaneously tests specimens submitted in viral transport medium for influenza viruses A and B, including subtypes H1, H3, and 2009 H1N1, and RSV A/B. Clinical evaluations have reported sensitivities of 96.6% to 100% for influenza virus A, 96.8% to 100% for influenza virus B, and 89.8% to 91.7% for RSV, with specificities of >96.5% for all targets (100, 103, 104), which in one comparative study was superior to results for a traditional RT-PCR test (103).

A clinical evaluation of the CDF assay demonstrated 98.7% sensitivity and 87.5% specificity for detection of toxigenic C. difficile based on the presence of tcdA and/or tcdB, the primary toxin-encoding genes present in toxigenic strains of C. difficile (94). In addition, the CDF assay contains capture probes for detection of the Δ117 deletion in tcdC, which encodes the repressor of tcdA and -B expression, and genes encoding binary toxin (cdtA and cdtB) (94). Strains with the Δ117 deletion produce up to 23-fold more toxin than wild-type strains (105). Additionally, the Δ117 deletion and the presence of binary toxin are characteristic of strains of the O27/NAP1 ribotype, which has been associated with more severe disease (105, 106).

Direct detection of microorganisms from positive blood cultures is an area of great interest because of the potential benefits of rapid identification to patient care, antimicrobial stewardship, and health care cost. Clinical performance of the BC-GP assay has been evaluated in several studies, including large multicenter efforts encompassing all commercially available blood culture systems. These studies have reported sensitivities of >96% for most of the 12 identification targets on the panel; however, lower sensitivities of 67.0% to 94.8% were reported for E. faecalis and E. faecium, and false-positive Streptococcus pneumoniae results were reported for isolates of Streptococcus mitis/Streptococcus oralis (95 – 97, 107 – 109). Importantly, the sensitivity of BC-GP for detection of mecA in methicillin-resistant S. aureus (MRSA) in these studies was >99%. Additionally, the performance does not appear to be affected by the type of blood culture broth used (i.e., aerobic versus anaerobic, pediatric versus adult, containing charcoal or resin, etc.). Literature regarding the performance of the more recently FDA-cleared BC-GN test is currently limited to studies including a small number of prospectively collected clinical specimens or studies using primarily simulated specimens (101, 102). In a multicenter evaluation of 104 clinical specimens, the overall sensitivity of BC-GN was 91%, with sensitivities of 67% to 100% for the 9 Gram-negative genus or species targets (101). In a larger study of 397 blood cultures (75% simulated specimens), sensitivity was >98% for 7 of the 8 targets present on the FDA-cleared panel. The single target demonstrating poor performance was Klebsiella pneumoniae, which was reported to be 86.1% sensitive (102). Interestingly, all of the specimens with false-negative results were phenotypically identified as Klebsiella pneumoniae; however, 16S rRNA gene sequence analysis identified these isolates as K. variicola. A distinguishing characteristic of the BC-GN compared to the FilmArray BCID is the inclusion of the resistance markers bla _CTX-M, bla _IMP, bla _VIM, and bla _OXA in addition to bla _KPC and bla _NDM, which are present in both assays. The sensitivity of BC-GN for these 6 genetic markers of antibiotic resistance is reported to be 100% compared to sequence analysis of the strains (102). Importantly, additional studies to demonstrate phenotypic correlation with detection of these markers are needed.

A potential strength of solid-array technology is the ability to correctly identify multiple targets in the same specimen; however, in studies involving the BC-GP and BC-GN, all targets in a polymicrobial culture were correctly identified in only 60.0% to 81.3% of specimens (96, 97, 101, 102). This limitation is similar to that observed with the FilmArray BCID (discussed above). Additionally, unlike the FilmArray, the Verigene blood culture assays are restricted to Gram-positive or Gram-negative targets. Selection of the correct test depends on accurate reading of the primary Gram stain. These limitations again underscore the importance of primary Gram staining as well as routine culture of all positive blood culture broths prior to finalizing the culture.

Liquid-array technology, typified by the xTAG assays (Luminex, Toronto, Canada), involves an initial multiplexed PCR step, followed by target-specific primer extension that incorporates a unique nucleic acid “tag” and biotin label into each target amplicon. Tagged amplicons are then incubated with microbeads of various fluorescent potential, each type coated with a unique antitag sequence. Amplified target sequences with incorporated tags complementary to those on a specific bead will hybridize. Finally, a streptavidin-fluorophore conjugate is added and hybridizes to biotin-labeled amplicons immobilized on the beads. Detection of a target is accomplished using two lasers that interrogate each bead for (i) the presence of a captured amplicon as indicated by streptavidin-fluorophore and ii) the identity of amplicon as indicated by fluorescence of the bead specific for each antitag (Fig. 5) (110). The xTAG test for agents of gastroenteritis (xTAG GPP) includes targets for 15 bacterial, viral, and protozoan pathogens associated with gastroenteritis. Few clinical evaluations of the assay have been published, but initial reports demonstrate sensitivity and specificity ranging from 82 to 100% depending on the comparator used as gold standard (111, 112). A larger number of studies have evaluated the xTAG assay for respiratory pathogens (xTAG RVP), which detects 12 to 19 viruses (FDA-cleared versus CE-Mark targets) associated with respiratory illness. These studies have found 92 to 100% agreement of xTAG RVP with other molecular platforms and sensitivities of 91 to 100% with specificities of >99% for individual targets on the panel (87, 113, 114).

Multiplexed molecular panels containing up to 20 targets or more can simplify ordering for the physician and simplify workflow in the laboratory by consolidating what were previously individual tests into a single “complex panel” for patients with respiratory illness, gastroenteritis, or positive blood cultures. An obvious benefit of these large multiplex molecular tests is the ability to detect numerous pathogens in a specimen without having to rely on different methodologies, including culture, molecular, EIA, or direct staining procedures as appropriate for the various pathogens that may be present in a single specimen. Especially in the case of fully automated platforms, this can ease the burden on the laboratory and reduce the dependence on experienced technologists for such tasks as identification of protozoan pathogens in a trichrome stain. Large multiplex panels also simplify test ordering for physicians, who may miss a diagnosis because of failure to order the correct test. For example, the diversity of targets on the Luminex GPP test enabled detection of a pathogen that would have been missed in up to 65% of specimens because the appropriate routine test to detect these pathogens was not ordered (111). An additional potential benefit is the ability to detect multiple pathogens simultaneously. Up to 10% of stool specimens may be positive for multiple targets which can be an indication of coinfection; however, these results must be interpreted with caution, since the presence of nucleic acid does not always correlate with clinical illness (111). Asymptomatic carriage of C. difficile, which can be as high as 15 to 20%, asymptomatic shedding of adenoviruses, or residual nucleic acid in the absence of viable organisms following treatment are potential sources of false-positive results (115 – 117). Other considerations include the pretest probability for a given pathogen and the cost per test, which is often higher for densely multiplexed and fully automated tests. During peak respiratory illness season, use of a batched molecular test for influenza viruses A and B for the majority of clinic patients may be more economical than a large on-demand multiplexed panel.

Identification of the organism present in positive blood culture broths using multiplexed molecular assays has been the focus of several recent publications because of the potential to dramatically impact patient care and reduce the total cost of care for patients suffering from bloodstream infections. Studies using NAATs or fluorescent in situ hybridization (FISH) for rapid identification of 2 to 4 targets in a positive blood culture broth have demonstrated significant reductions in the time on suboptimal antimicrobial therapy, length of hospital or ICU stay, and overall cost of care for patients infected with S. aureus, MRSA, Enterococcus, or Candida species (62, 118 – 120). While these outcomes are impressive, each test is limited to a relatively small number of microorganism targets, making each applicable to only 5% to 50% of cultures having a Gram stain consistent with specific test targets (81). Larger multiplex panels containing 12 or more targets are more broadly applicable across all blood cultures. For example, the Verigene BC-GP test (12 Gram-positive identification targets) effectively identified the bacterium present in 92.5% of cultures containing Gram-positive organisms, and the FilmArray BCID test (8 Gram-positive, 11 Gram-negative, and 5 Candida targets) accommodated >90% of the microorganisms present in all positive blood cultures (85, 97).

The level of automation of multiplexed tests and the level of complexity (high complexity versus moderate complexity) are features that broadly divide these tests and must be considered when choosing the most appropriate test for a laboratory. Tests that require “offline” extraction of nucleic acids and manual pipetting to set up the PCR(s) are designated high-complexity tests and as a result may not be suitable for most laboratories with restricted staffing or expertise. Alternatively, sample-to-result platforms typically gain approval as moderate-complexity tests which can be adopted by laboratories which lack staff with appropriate training/certification or which are not designated “high-complexity” laboratories by CLIA. This may be an important factor for many laboratories when selecting a molecular platform that best suits their needs. Another factor that needs to be considered is the per-test cost. As discussed above, the cost per test may be reduced for batch-type platforms compared to sample-to-results tests; however, the turnaround time for reporting of results will suffer. Evaluations of total time to result, throughput, and cost of the xTAG, FilmArray, and Verigene have reported a total turnaround time of 7 to 8 h for xTAG, with up to 21 samples reported in this run time (87). The extended TAT for xTAG is a result of the requirement for offline extraction and manual setup, which require 3 to 5 steps and over 1 h of hands-on time. In contrast, the FilmArray and Verigene are true sample-to-result platforms that provide a reportable result in 1 to 2 h, with <5 min of hands-on time. As discussed above, a rapid TAT is essential to maximizing the clinical benefits for identification of microorganisms in blood culture but may not be as critical for other specimens types, such as stool specimens, received from outpatient clinics. The major drawback to both sample-to-result platforms is limited throughput, which is only 4 to 8 tests per 8-h shift (87). Therefore, the benefit of a rapid, on-demand result must be weighed against specimen throughput capabilities. These decisions may be affected by season (e.g., respiratory illness season) or by patient population (inpatient versus outpatient), so it is important for the clinical laboratory to fully assess the needs of its specific institution or clinics when deciding on a platform for multiplex testing of clinical specimens.

Originally described in 1976, DNA sequencing has undergone significant modifications, culminating in massive parallel or “next-generation” whole-genome sequencing (WGS) methods which are gaining favor for use in clinical microbiology laboratories. Allan Maxam and Walter Gilbert first achieved reliable DNA sequencing using a method that involved radioactive labeling of the DNA and chemical treatment to break DNA into small fragments. Fragments from each of four parallel reaction mixtures (one for each nucleotide) were electrophoresed side by side, with visualization using X-ray film autoradiography. Banding patterns in each lane corresponded to radiolabeled DNA fragments containing one of four radiolabeled nucleotides, from which the sequence could be inferred. Maxam-Gilbert sequencing quickly fell out of favor due to its technical complexity and use of hazardous chemicals, which complicated scale-up and prevented its use in standard molecular biology kits. Chain termination sequencing, or Sanger sequencing, improved upon the Maxam-Gilbert method by using dideoxynucleoside triphosphates (ddNTPs) as DNA elongation terminators. Sequencing reaction mixtures were again divided into four parallel vessels, each containing one of four ddNTPs (ddATP, ddGTP, ddCTP, or ddTTP) along with an excess of standard dNTPs. The resulting reaction mixtures contained DNA fragments of different lengths representing each size fragment produced by termination following inclusion of the given ddNTP. Fragments from the four reaction mixtures could then be separated by gel or capillary electrophoresis with a resolution of one nucleotide. The ability to radioactively or fluorescently label each ddNTP enabled detection of fragments and reading of sequence data by automated sequencing instruments. These advances in automation and analysis made nucleic acid sequencing a realistic option for diagnostics; however, several technical challenges still prevented the widespread use of Sanger sequencing in the clinical laboratory. Poor quality in the first 15 to 40 bases of the sequence and deteriorating quality of sequencing data after 700 to 900 bases limit its applicability to relatively short DNA fragments. Additionally, Sanger sequencing reactions are limited to sequence analysis of a single amplicon per reaction. This prevents the analysis of complex specimens such as sputum or abscess which contain multiple organisms (149).

Next-generation sequencing (NGS) refers to a high-throughput sequencing method that parallelizes the sequencing process, producing thousands or millions of sequences at once. Intentionally broad, next-generation sequencing encompasses several different sequencing technologies that have been adapted to high-throughput, low-cost sequencing. A thorough review and comparison of these methods has been published elsewhere (150), but we summarize the key differences and applications of the major NGS approaches (Table 3).

Pyrosequencing, licensed by 454 Life Sciences and later purchased by Roche, was the first next-generation sequencing method commercially marketed. Pyrosequencing employs a “sequence-by-synthesis” approach, meaning that it generates sequence data during DNA synthesis rather than analyzing nucleic acid amplicons postsynthesis as is the case with Sanger sequencing (151, 152) (Fig. 7). Amplified or chromosomal target nucleic acid is fragmented, and synthetic nucleic acid adaptors are enzymatically ligated to each end of the product. One adaptor serves as an adaptor for hybridization of the nucleic acid product to a microbead, and the other serves as a sequencing primer. Following a PCR to amplify the target sequence, microbeads coated with amplicon are segregated into microwells. Each well contains all the reagents required for sequencing, including DNA polymerase, luciferase, ATP sulfurylase, and apyrase. Each of the four dNTPs is individually added and washed away from the wells in repeating cycles. When a complementary dNTP is added, it is incorporated by DNA polymerase, with the concomitant release of pyrophosphate as a by-product of DNA synthesis. ATP sulfurylase converts the released pyrophosphate to ATP, which is used to drive luciferase activity, resulting in the production of light (153). Sequence data are generated by monitoring the microwell reactions for a pulse of light following addition of each dNTP. Since each microwell contains a single microbead harboring a unique region of chromosomal DNA, parallel sequencing of hundreds of regions of the chromosome achieves high sequence coverage in a single run. Additionally, because sequencing reactions are carried out in picoliter-volume reaction wells, this technology is capable of sequencing 400 to 600 megabases of DNA per 10-h run at a price per base up to 100-fold lower than that for Sanger sequencing (154) (Table 3). Pyrosequencing was initially capable of generating accurate reads of approximately 100 bases, with the limiting factor related to decreasing efficiency of apyrase in degrading unincorporated nucleotides in each successive cycle (155). Replacement of apyrase with thorough washing to remove unused nucleotides can extend the effective read length to approximately 400 bases. This is still a relatively short read in comparison to that with the Sanger method, but it is significantly longer than those of other NGS methods (155). An extended read length can be advantageous when attempting rapid whole-genome sequencing (WGS), especially when coupled with the speed of pyrosequencing technology and sophisticated software capable of assembling short individual reads into a confluent genome sequence. The overall accuracy of the sequence data generated is 99.51% to 99.96% (156, 157). A potential drawback to pyrosequencing is the inability to generate reliable sequences of homopolymers of >4 bases in length (156). In a study assessing the accuracy of sequences generated by pyrosequencing, 39% of errors were attributable to homopolymer sequences (156).

Semiconductor sequencing, typified by the Ion Torrent system (ABI), is a similar “sequence-by synthesis” technology. Parallel sequencing reactions are carried out in 1.2 million microwells on the surface of a low-cost semiconductor chip (158). Each picoliter well contains template and DNA polymerase, to which each of the four nucleosides is added in sequential order, however; Ion Torrent sequencing differs from pyrosequencing in that it uses production of hydrogen as the sole marker for determining the sequence (Fig. 7) (158). Release of hydrogen ions following incorporation of a complementary nucleotide is detected by a miniaturized ion sensor integrated into each reaction well. This technology is capable of generating up to 25 Mb of sequence data in a single run with a 2-h run time (158). Independence from the use of multiple enzymes, sensitive optics, or modified nucleotides dramatically reduces the cost of reagents and equipment compared to those with Sanger or other NGS methods. The reported cost of an Ion Torrent instrument is approximately US$50,000, excluding sample preparation equipment and a server for data analysis (159). The reported accuracy of semiconductor sequencing systems, including Ion Torrent, ranges from 98.4% to 98.9% (158, 160) (Table 3). The major limitations of this system are that it has difficulty in enumerating long repeats (homopolymers of >6 nt in length) and has a read length of 50 to 100 nt, which is relatively a short compared to that of Sanger sequencing or pyrosequencing (158).

Applications of pyrosequencing and semiconductor sequencing include whole-genome sequencing (WGS), amplicon sequencing, transcriptome sequencing, and metagenomics. The strength of pyrosequencing for WGS was demonstrated by Margulies et al., who sequenced the entire genome of M. genitalium (580,096 bp) with >99.9% accuracy and 96% genome coverage in a single run (157). More impressively, pyrosequencing was utilized to sequence the entire 6-gigabase human genome with 7.4× coverage in just 2 months (154). Similarly, the whole genomes of Escherichia coli and Vibrio fischeri were sequenced with 96.8 to 99.9% coverage with 98.9% accuracy in a single run using Ion Torrent (158). While the sequencing and assembly of an entire genome in days to months are remarkable, the most immediate use of NGS in clinical microbiology is likely amplicon sequencing. Amplicon sequencing is targeted to full sequencing of one or more genetic loci concurrently. This method is valuable when identification of multiple mutations or SNPs in a genetic locus is required to predict oncogenic potential or antimicrobial resistance. In addition to detection of multiple SNPs in a single locus, parallel sequencing offers the ability to generate sequences for multiple loci simultaneously. Next-generation sequencing is among the molecular technologies that can be applied to the identification of mycobacteria, including the prediction of resistance to antituberculosis therapies (64). Determination of resistance to first-line antituberculosis drugs (rifampin [RIF], isoniazid [INH], pyrazinamide [PZA], and ethambutol [EMB]) requires the analysis of several SNPs contained on 5 different genes (161). SNPs associated with resistance to rifampin are relatively conserved, with 3 mutations accounting for up to 75% of resistance (161). In this instance, routine probe-based amplification tests can be up to 98% sensitive (162). However, SNPs resulting in resistance to other first line antituberculosis drugs are considerably less conserved, rendering detection by a limited number of probes impractical. Pyrosequencing has been exploited for the simultaneous detection of resistance mutations in multiple genes to rapidly identify multidrug-resistant (MDR) strains of M. tuberculosis (163, 164). Resistance to rifampin, isoniazid, and fluoroquinolones was determined using 4 sequencing primers to identify multiple point mutations in rpoB, katG, and gyrA, with sensitivities of 96.7%, 63.8%, and 70%, respectively. The specificity of the pyrosequencing reaction was reported to be 97.3% to 100% (163). Variable sensitivity for predicting susceptibility to the 3 drugs reflects the lack of knowledge regarding the mutations and mechanisms which contribute to a resistant phenotype. This limitation is inherent to all molecular testing strategies and will be overcome only through continued research to characterize mutations conferring resistance and development of more complete reference libraries for sequence comparison.

Analogous to the use of NGS methods to sequence multiple targets in a single organism is the utility of NGS to simultaneously sequence and identify multiple organisms in a single specimen. Many of these studies, known as metagenomics, have been conducted to characterize complex bacterial communities in environmental specimens. Clinically, NGS has been used characterize the microbial community present in the airways of patients with cystic fibrosis (CF) using sputum specimens (149, 165). An advantage of NGS is the detection of nonculturable or fastidious organisms that may be outcompeted and overlooked in routine CF cultures. In a cohort of 66 sputum specimens from CF patients, NGS identified 122 different microbial species, compared to only 18 identified by culture (149). In an analytic study, organisms representing as little as 0.25% of the total nucleic acid template in a specimen were reproducibly identified (149). This ability to better define the microbiological components of the CF lung could aid in a better understanding of the associated illness and inform therapeutic strategies. A potential drawback to this type of metagenomic study is the semiquantitative nature NGS, which prevents an accurate assessment of the proportion of each organism present at a single point or changes in the composition of microorganisms in serial specimens. Similarly, the presence of nucleic acid is not necessarily indicative of a viable organism and may represent residual nucleic acid from flora or exogenous sources entering the upper respiratory tract.

A final application of pyrosequencing and semiconductor sequencing is in epidemiological investigation of outbreaks. Most notably, Mellmann et al. used the Ion Torrent NGS to identify and characterize a novel strain of enterohemorrhagic E. coli (EHEC) responsible for a large outbreak in Germany in 2011 (166). Whole-genome sequencing of 4 isolates from geographically distinct cities along with relevant historical reference strains was conducted. Sequencing and analysis of the strains were completed in 2 to 3 days and enabled near-real-time phylogenetic linkage of these strains (166). Investigators were also able to propose a likely evolutionary pathway linking the outbreak strain to an earlier progenitor strain identified 20 years earlier. In a smaller study, investigators were able to examine 33 multidrug-resistant isolates of E. coli obtained from patients in a neonatal intensive care unit using Ion Torrent NGS (167). The authors reported a 5-day turnaround and a cost of US$300 per isolate for whole-genome sequencing. Sequencing resulted in 88% to 89% genome coverage, which was sufficient to link all strains phylogenetically and identify them as most closely related to multiresistant strains of the ST-131 multilocus sequence type (MLST). While approximately twice the cost of traditional strain typing using pulse-field gel electrophoresis (PFGE) or MLST, NGS provided additional useful information, including the specific identification of the bla _CTX-M-15 extended-spectrum beta-lactamase (ESBL) gene and the presence of other genes and point mutations associated with resistance to several classes of antimicrobials (167).

The term “ultradeep sequencing” (UDS) refers to amplicon sequencing designed to allow mutations to be detected at extremely low levels in a population. Initial PCR amplification of a genetic region of interest followed by segregation of each amplicon into a separate reaction well allows sequencing and identification of rare sequence variants. For example, ultradeep sequencing has been successfully used to detect HIV quasispecies and the emergence of resistant subpopulations. Analysis of blood samples from HIV-infected patients using pyrosequencing identified strains with mutations in the viral reverse transcriptase gene at levels of <0.1% of the total viral population (168, 169). Similarly, Ion Torrent sequencing was utilized to identify the emergence of mutations conferring resistance to nonnucleoside reverse transcriptase inhibitors (NNRTIs) and protease inhibitors (PIs) at a level of <1% of the total population through an average of 13,700× coverage of the gag-pol loci, though it was noted that coverage decreased significantly in a homopolymeric region containing five consecutive guanine residues (170). This is again in comparison to routine Sanger methods, which demonstrate a limit of detection of approximately 20 to 35% of the population (171). Accurate sequence data with error rates of <0.05% are easily achieved due to the high number of parallel reads, which provide highly redundant coverage of the target sequence (168, 169). Pretherapy resistance testing has been recommended to identify quasispecies with mutations known to confer resistance to antiretrovirals and is also recommended following a rise in HIV load attributed to therapy failure (172). Early detection of mutant alleles present at a low frequency is key in selection of antiretroviral therapy, since discontinuation of a specific antiviral can result in reversion of mutant populations to a susceptible, pretherapy genotype (168, 173). Future clinical applications of pyrosequencing include transcriptome sequencing, which aims to efficiently create RNA profiles and examine the effects of mRNA transcript expression. The majority of research using NGS for transcriptome analysis has involved the basic sciences; however, recent studies have utilized this method for comparison of mRNA expression in normal and malignant cell populations and for discovery of latent or cryptic viruses whose presence and expression may be associated with malignancies (174 – 176).

Other NGS platforms such as Illumina and SOLiD are capable of generating 1.5 to 4.0 Gb of data per single run at a cost of less than $0.10 per kilobase, which is significantly less expensive than Sanger or other NGS methods (177, 178). Illumina (Solexa) sequencing is based on reversible dye terminators. DNA molecules are first attached to primers on a slide and amplified so that local clonal colonies are formed. Four types of reversible terminator bases are added, and nonincorporated nucleotides are washed away. A camera takes images of the fluorescently labeled nucleotides, and then the dye along with the terminal 3′ blocker is chemically removed from the DNA, allowing the next cycle. In contrast, SOLiD (supported oligonucleotide ligation and detection) is a method of sequencing by ligation. A target-specific sequencing primer is used to initiate sequencing by the sequential addition of octamer probes, each containing 2 specific nucleotides at the 5′ terminus followed by 6 degenerate nucleotides. Each of the 16 possible combinations of two nucleotides is represented, and octamers are fluorescently labeled with one of 4 fluorophores. The 16 octamers are then grouped into 4 sets (each containing one each of the 4 fluorophores) and are added sequentially to the sequencing reaction mixture for 7 full cycles of the 4 groups. Fluorescence is measured after addition of each 4-member group of probes, and the 2-base sequence is determined by the fluorophore detected. Gaps in the sequence corresponding to the 6 degenerate nucleotides in each probe are filled in by repeating the reaction using additional sequencing primers, each offset by one nucleotide (n − 1, n − 2, etc.) from the initial primer (177). This results in short reads (26 nucleotides); however, the sequencing error rate is reduced to 0.001 because each nucleotide in the template is read twice (177). The disadvantage of this technology is turnaround time. The run time for a single sequencing reaction is 2.5 to 6 days, resulting in turnaround time for a full genome sequence of up to 2 weeks (178). Because of the large amount of sequence data generated per run, low cost per base sequenced, and extended TAT, these platforms are currently best suited to whole-genome sequencing projects rather than rapid identification of microorganisms or SNP polymorphisms in a clinical laboratory. Most recently, Illumina has begun offering full genome sequencing through it reference laboratory at a reported cost of $4,000.00 per genome.