Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Microarrays and Deep Sequencing in Clinical Microbiology Moving from PCR testing to more sophisticated nucleic acid testing opens up great opportunities as well as considerable challenges Charles Chiu and Steve Miller ucleic acid detection technology is revolutionizing microbiological diagnostic testing. With the exception of infectious prion proteins, all pathogenic microbes contain DNA and / or RNA and are thus targets for nucleic acid-based testing. Some clinical microbiology laboratories are largely shifting to nucleic acid testing, which is particularly useful for detecting slow-growing or unculturable organisms such as viruses, mycobacteria, and fastidious bacterial pathogens. However, significant challenges remain before this technology can be applied universally in clinical laboratories. The core of nucleic acid tests in clinical N Summary • High-density microarrays such as the Virochip pan-viral microarray and deep sequencing can test for thousands of potential pathogens simultaneously. • Identifying a novel or unusual infectious agent from clinical samples does not prove that it causes a disease, and further investigation is required to show that it is a pathogen • Interpreting data, setting quality control standards, and meeting regulatory requirements are among the challenges facing those who are adapting these technologies for diagnostic use. • These technologies are enabling microbiologists to conduct broad-based surveillance for novel infectious agents that elude conventional testing and in the near future are likely to radically transform the way clinical microbiology laboratories approach infectious disease diagnosis. microbiology laboratories is the polymerase chain reaction (PCR). Although PCR can rapidly provide clinically useful information regarding the presence of pathogens, its incredible specificity hinders its utility for broadspectrum detection. The adage, “you only find what you’re looking for” applies to PCR as well as most nucleic acid-based tests. Moreover, detecting an infectious agent yields only half the picture. Another part of the clinically useful information is based on susceptibility profiles. Genotyping through nucleic acid testing reliably predicts single-gene instances of drug resistance in both viruses and bacteria. More complicated resistance mechanisms involving multiple genes and interacting mutations require more comprehensive testing methods than PCR. Another reason why nucleic-acid detection methods other than PCR are needed is that clinical syndromes such as pneumonia, gastroenteritis, sepsis, and eucephalitis are not specific for any one pathogen or category of pathogens. The “one test-one pathogen” paradigm can prove costly, inefficient, and time-consuming. In addition, diagnostic microbiology tests routinely fail to detect novel or unusual pathogens. Thus, in the clinical microbiology laboratory, we do not identify the etiology for about 25% of cases of acute respiratory illness, 50% of diarrheal diseases, and more than 70% of encephalitis, despite extensive conventional testing. The emergence of novel pathogens further confounds diagnostic efforts in clinical microbiology. During the past century, Charles Chiu is an Assistant Professor in the Departments of Laboratory Medicine and Medicine (Infectious Diseases), University of California, San Francisco (UCSF) School of Medicine. Steve Miller is an Assistant Professor in the Department of Laboratory Medicine, University of California, San Francisco (UCSF) School of Medicine. Volume 6, Number 1, 2011 / Microbe Y 13 FIGURE 1 A B C D tect pathogens on a broad-spectrum basis, relying in part on high-density microarrays and deep sequencing. Rooted in metagenomics, the analysis of complex mixtures of nucleic acids, these technologies are being used for comprehensive surveillance and to identify novel pathogens. Although these approaches provide great opportunities, they also come with considerable challenges to be met amidst an increasingly complex and stringent regulatory landscape. Microarray Technology in Clinical Diagnostics DNA microarrays use probes to detect sequences that are complementary to those probes, which are immobilized on solid surfaces or attached to small beads and can interrogate millions of E sequences in a single assay. In microbiNP P/V M E HN L ology, such arrays are used in several ways, including to speciate bacteria and fungi, to investigate microbial diversity in complex environmental samples, to 0 2 4 6 8 15 kb analyze transcripts, and to identify Virochip identification of human parainfluenza virus 4 (HPIV-4) infection in a critically ill pathogens. patient. (A) Chest radiograph showing bilateral infiltrates. (B) Computed tomography scan Various microarrays are being used revealing a “tree-in-bud” appearance suggestive of bronchiolitis. (C) Open lung biopsy for clinical microbiology diagnostics, showing an organizing bronchiolitis but no direct evidence of virus infection. (D) Positive indirect immunofluorescence assay for HPIV-4 from patient’s serum at day 24. The and the panels of infectious agents that cytoplasm of infected cells stains green against a background of uninfected cells in red. they can detect are continually expand(E) The locations of the six parainfluenza virus probes used to detect HPIV-4 on the ing. For example, Luminex markets a Virochip are mapped onto the !15 kB genome (arrows). Figure modified from Chiu, et al., (2006) Clin Infect Dis., with permission. panel of probes for detecting 12 viral agents of upper respiratory tract infection (the Luminex RVPTM Assay); this microbead-based panel is licensed by the Food more than 300 previously uncharacterized inand Drug Administration (FDA), while its larger fectious agents emerged in the human populapanel is already used in Europe. Meanwhile, tion. New pathogens such as the SARS coroAutogenomics has several research-use-only navirus and pandemic 2009 H1N1 influenza (RUO) probe panels for identifying or typing swiftly spread to susceptible populations respiratory viruses, human papillomavirus worldwide. Climate change is altering the geo(HPV), mycobacteria, and agents that cause graphic distribution of pathogens, while vaginitis. A number of companies, including greater numbers of immunocompromised Akonni, TessArae, and Veredus, produce RUO hosts are leading to infections by emerging assays for detection of influenza virus. Typiagents, including atypical mycobacteria and cally, these microarray-based technologies are JC / BK polyomaviruses. Meanwhile, crosslimited to approximately 100 probes, detecting species zoonotic transmissions of novel pathoonly the most common pathogens or strains but gens such as HIV are fueling devastating pannot meant for comprehensive analysis. demics. Our group at University of California, San In the face of those needs, microbiologists Francisco (UCSF) is developing microarrays for are adapting nucleic acid technologies to de- 14 Y Microbe / Volume 6, Number 1, 2011 dom amplification technique to enable unbiased detection. Initially, probes chosen for the ViroA Encephalomyocarditis chip were based on nucleic acid seVirus (EMCV) quences from the most conserved reHuman gions of viral families. Probes were later Cardioviruses added to cover other nodes in the viral taxonomy at the family, genus, and species levels. Moreover, the Virochip probes are updated regularly, with the v5.0 version from October 2009 containing about 36,000 probes. Its sensiTheiler’s Murine tivity is superior to direct fluorescent Encephalomyelitis antibody tests and comparable to PCR Virus (TMEV) for detecting respiratory genomes at 0.2 nucleotide substitutions levels as low as 100 genome copies based on testing for rhinoviruses. B Day 0 C We used the Virochip to diagnose the cause of a severe respiratory infection in 2.0 an immunocompetent, 28-year-old Convalescent woman, who was admitted to the hos1.5 pital in 2005 with a 10-day history of fever, cough, and night sweats. Her chest radiograph revealed nonspecific 1.0 bilateral infiltrates, and a CT scan had a Day 10 tree-in-bud appearance, suggesting bronchiolitis (Fig. 1A and B). After be0.5 ing treated with broad-spectrum antibiAcute otics, she developed acute respiratory 0 failure requiring mechanical ventilation. Conventional testing for over 30 infectious agents proved negative, and Serum dilution an open-lung biopsy was unrevealing (Fig. 1C). However, based on our ViroFollow-up investigation to link a novel cardiovirus with disease. (A) Phylogeny of the chip analysis (Fig. 1E), we determined human cardioviruses. The fully sequenced genomes of 6 human cardioviruses were aligned with the genomes of 4 animal cardioviruses using Geneious (Biomatters, Inc., that she was infected with the human NZ). (B) Cytopathic effect from a human cardiovirus is observed at day 10 (bottom panel) parainfluenza 4 virus (HPIV-4). This after infection of monkey kidney cells at day 0 (upper panel). (C) A cardiovirus-specific finding was unexpected because ELISA (enzyme linked immunosorbent assay) demonstrates an acute seroconversion to cardiovirus infection in a child with diarrhea and vomiting. Figure modified from Chiu, et HPIV-4 is primarily associated with al. (2010) J Virol., with permission. mild upper respiratory infections in children and adults. The diagnosis was confirmed by direct RT-PCR amplificatwo main purposes, conducting broad-spectrum tion and sequencing, as well as serological analsurveillance and discovering viral pathogens. ysis (Fig. 1D). The Virochip, which we use, and GreeneChip, a similar platform designed by Ian Lipkin and his Clinical Applications of Pathogen colleagues at Columbia University, New York, Discovery N.Y., are microarrays containing thousands of probes to target all viral families known to infect The Virochip is actively being used to identify humans. The probes on these arrays contain other viral agents associated with disease. In elongated, 70-mer oligonucleotides to increase 2006, for example, we reported its use to identheir sensitivity for diverse viral strains. Both tify the XMRV retrovirus in prostate cancer platforms employ a primer-independent ranpatients. Other research groups have subse1: 20 1: 40 1: 80 1: 16 0 1: 32 0 1: 64 1: 0 12 8 1: 0 25 6 1: 0 51 20 Absorbance (450 nm) FIGURE 2 16 Y Microbe / Volume 6, Number 1, 2011 quently detected this virus in associaFIGURE 3 tion with chronic fatigue syndrome. The Virochip also helped to identify the A H3N2 Virochip H1N1 (human) H1N1 (swine) probes Virochip probes Virochip probes SARS coronavirus, a new clade of rhinoviruses, a divergent human metapseasonal influenza H3N2 neumovirus infection in a patient with nasal swab samples critical respiratory illness, avian bornaseasonal influenza H1N1 virus (the viral agent of an HIV-like nasal swab samples illness in parrots), and a novel human cardiovirus. The clinical utility of this approach comes from detecting novel or previpandemic 2009 influenza H1N1 nasal swab samples ously uncharacterized agents that truly cause disease. For example, our unanticipated finding of severe HPIV-4 infection in a healthy young adult expanded the documented clinical spectrum of this pathogen, supporting B specific testing for HPIV-4 in individu0 2000 4000 6000 8000 10000 12000 als with severe respiratory illness of unPB2 PB1 PA HA NP NA M NS known cause. Subsequent research confirmed that HPIV-4 may indeed be an All 17 pandemic underappreciated cause of respiratory 2009 influenza H1N1 infections in nursing home, hospital, nasal swab samples and daycare settings. (21 de novo However, the discovery of a novel assembled contigs, virus or bacterium may not necessarily 89.5% coverage) be clinically relevant. Merely detecting a virus in clinical samples does not Investigation of patients infected with pandemic 2009 H1N1 influenza by microarrays and deep sequencing. (A) Heat map (cluster analysis) showing increased intensity of Virochip prove whether it is part of normal flora, probes derived from swine, not human, influenza, in patients infected with 2009 a bystander, or a true pathogen. In cases influenza H1N1. Note that the Virochip was designed prior to the emergence of the where the pathogen, clinical syndrome, pandemic. (B) De novo assembly of !90% of the genome of 2009 influenza H1N1 from 1.6 million deep sequencing reads in the absence of a reference, including the nearly and epidemiology findings are linked, a full-length sequences of the PB1, HA, NP, and NA genes. Figure modified from Chiu, et strong case can be made for causality. al, (2010) PloS ONE, with permission. For example, we recently used the Virochip to detect a novel human cardiovirus in children with acute respiratory Validation, Quality Control, and and diarrheal illness (Fig. 2A). Subsequently, we Regulatory Issues cultured the cardiovirus in human cell lines (Fig. It is not feasible to verify acceptable test re2B) and detected an acute seroconversion event sults for hundreds to thousands of individual in a child with diarrhea and vomiting in a housetargets in broad-spectrum microarrays. Inhold infected with cardiovirus (Fig. 2C), thus stead, quality control relies on validating each linking cardiovirus infection with diarrheal step of the assay while verifying only the most symptoms. common infectious agents. Thus, it becomes Even if a novel agent causes a disease, develnecessary to confirm unusual results with oping and validating diagnostic tests for that some other method, such as pathogen-specific agent might not be justified if the disease is PCR, serology, or sequencing, while the clinibenign or if diagnosis has no impact on clinical cal significance of atypical or novel pathogens management. Nevertheless, as in the 2009 has to be carefully considered with clinicians H1N1 influenza pandemic, identifying novel ordering the tests and caring for particular pathogens can rapidly spur efforts to develop patients. tests to monitor outbreaks as well as effective Clinical laboratories validate tests before drugs and vaccines to treat the disease. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 Volume 6, Number 1, 2011 / Microbe Y 17 approved test was available to detect this specific strain of the influenza virus. Then diagnostic laboratories and Experimental observations test manufacturers played “catch-up” as the FDA began approving tests for the virus on an emergency use authorization (EUA) basis. Soon after the 2009 H1N1 outbreak began, the Virochip system identified the Environmental or RNA / Microarray Hybridization clinical sample DNA pattern H1N1 virus as a novel strain that was most similar to a swine influenza strain (Fig. 3A). While “homebrew” assays Virus1 such as the Virochip may be useful for Virus2 detecting emerging pathogens, their disVirus3 Probe tribution is very limited because the test is selection Virusk not FDA-approved. One approach to exi pediting FDA review might be to validate Ranked viral Pattern to profile identities and comparisons targeted microarrays containing specific probability categories of pathogens such as biothreat estimates agents or retroviruses or tailored to specific disease states, such as those causin1 grespiratory illnesses, diarrhea, sepsis, or encephalitis. We are actively exploring 2 ATTGCGTTAT this strategy forgaining FDA approval for ATTACGACAT such microarray assays. k An additional complication arises Environment GenBank Alignment to Theoretical microarray probes energy profiles from dependence on algorithms to interpret results from microarrays and Predicted observations deep sequencing analyses. In 2006, FDA officials issued a draft guidance, The E-Predict algorithm. Nucleic acid from an environmental or clinical sample is labeled and hybridized to the Virochp microarray. The resulting microarray hybridization pattern is “In Vitro Diagnostic Multivariate Incompared with a set of theoretical microarray profiles computed for every viral species dex Assays (IVD-MIA),” which recomof interest. Energy profiles with statistically significant comparison scores suggest the mends formal reviews for any diagnospresence of the corresponding virus in the sample. Figure modified from Urisman, et al. (2005) Genome Biology, with permission. tic device that combines multiple variables to yield a single “index” or “score” that “cannot be independently they are used for guiding patient care. Diagderived or verified by the end user.” In short, nostic tests typically adhere to FDA-approved this guidance urges that algorithms be validated, protocols, use standardized reagents, and are a process that would significantly increase the approved based on their ability to diagnose time and resources needed to validate such sysdisease. Clinical laboratories that are certified tems and could further delay their commercial according to the Clinical Laboratory Improvedevelopment as well as increase their cost to ment Amendments (CLIA) are permitted to users. perform in-lab validated or “homebrew” tests. Because such tests are not FDA-apDeep Sequencing Technology proved, CLIA-certified labs are required to show clinical utility of those tests and to stanDeep sequencing involves highly parallel DNA dardize their performance characteristics. sequence analysis, yielding thousands to milTypically, the FDA requires patient outcome lions of sequence reads per run. Typically, each data for each pathogen included in an infecDNA sequence contains 36 –500 base pairs, tious disease test panel. The weakness of this with some technologies producing paired-end system was exposed during the 2009 H1N1 reads that contain information from both ends influenza pandemic. At the outset, no FDAof a DNA molecule. The instruments now typiFIGURE 4 18 Y Microbe / Volume 6, Number 1, 2011 cally used for identifying and analyzing pathogens include the Roche 454 pyrosequencing system and the Illumina Genetic Analyzer. While single runs can cost several thousand dollars, “barcode” adapters reduce costs by enabling simultaneous analyses of multiple samples. Costs are expected to continue to fall as these technologies continue to evolve. While deep sequencing is not currently being used for diagnostic tests, it has been used to identify several novel viruses, including the hemorrhagic fever Lujo virus from South Africa; the Dandenong virus, an arenavirus associated with fatal diseases in transplant recipients; the Merkel cell polyomavirus, associated with a rare skin cancer; and several viruses associated with gastroenteritis such as cosavirus and klassevirus / salivirus. Deep sequencing was also applied to examine minority quasispecies of HIV-1 to detect rare resistance determinants and to test vaccines for low-level contaminants such as the porcine circoviruses that were found in the RotaTeq vaccine. Researchers are also using deep sequencing as a metagenomics approach to discern organism profiles in clinical samples. For example, the Human Microbiome Project relies on deep sequencing to follow changes in human flora associated with diseases affecting the mouth, skin, vagina, gut, and respiratory tract. We recently applied deep sequencing technology to 17 respiratory samples collected from individuals infected with the 2009 H1N1 influenza virus early during the pandemic. Our analysis confirmed that the virus was the sole infectious agent common to all 17 cases. We also tested whether we could use a deep sequencing approach to identify novel viruses by attempting to de novo assemble the influenza genome after removing all influenza sequences from the reference database. Indeed, we were able to assemble the nearly complete sequence of all eight segments of the influenza genome without relying on a priori knowledge of the influenza (Orthomyoviridae) family (Fig. 3B). Thus, a diagnostic strategy of rapid Virochip-based testing followed by deep sequencing could prove to be a useful public health response to infectious disease outbreaks in the future. Challenges Interpreting Bioinformatics Data Correctly interpreting the clinical significance of microarray and sequencing data remains a formidable challenge. For example, where samples are polymicrobial or agents are present at low levels, supportive clinical and epidemiological information might be required to draw clinically meaningful conclusions. Another challenge is to determine which algorithms and what analytic methods will yield the most clinically meaningful results. When using the Virochip, we typically perform three different sets of analyses. Rank analysis by raw intensity data or Z-scores highlights the probes with high signal intensity, thus pinpointing the virus that is present. Hierarchical cluster analysis reveals a visual pattern of probe signals corresponding to a known virus or viral strain (Fig. 3A). E-Predict and related algorithms compare the microarray hybridization pattern from a clinical sample to profiles from viruses in the GenBank sequence database (Fig. 4). Because deep sequencing provides no visual check of the raw data, we depend on sequence alignments to identify species. Some instruments yield longer DNA reads, which are useful for identifying novel and/or divergent pathogens and de novo genome assembly, while others have better depth of coverage (more sequences), which increases sensitivity as well as accuracy. Clinical samples can be difficult to analyze. For instance, we sometimes need to treat respiratory samples from influenza patients with DNase to remove host background DNA. Moreover, having two closely related species in the same sample can cause assembly errors, leading to false chimeric assemblies. Thus, deep sequencing analysis requires careful consideration of the sample type, processing method, and background flora. SUGGESTED READING Briese, T., J. T. Paweska, L. K. McMullan, S. K. Hutchison, C. Street, G. Palacios, M. L. Khristova, J. Weyer, R. Swanepoel, M. Egholm, S. T. Nichol, and W. I. Lipkin. 2009. Genetic detection and characterization of Lujo virus, a new hemorrhagic fever-associated arenavirus from southern Africa. PLoS Pathog 5:e1000455. Chiu, C. Y., A. L. Greninger, E. C. Chen, T. D. Haggerty, J. Parsonnet, E. Delwart, J. L. Derisi, and D. Ganem. 2010. Volume 6, Number 1, 2011 / Microbe Y 19 Cultivation and serological characterization of a human Theiler’s-like cardiovirus associated with diarrheal disease. J. Virol. 84:4407– 4414. Chiu, C. Y., A. L. Greninger, K. Kanada, T. Kwok, K. F. Fischer, C. Runckel, J. K. Louie, C. A. Glaser, S. Yagi, D. P. Schnurr, T. D. Haggerty, J. Parsonnet, D. Ganem, and J. L. DeRisi. 2008. Identification of cardioviruses related to Theiler’s murine encephalomyelitis virus in human infections. Proc. Natl. Acad. Sci. USA 105:14124 –14129. Chiu, C. Y., S. Rouskin, A. Koshy, A. Urisman, K. Fischer, S. Yagi, D. Schnurr, P. B. Eckburg, L. S. Tompkins, B. G. Blackburn, J. D. Merker, B. K. Patterson, D. Ganem, and J. L. DeRisi. 2006. Microarray detection of human parainfluenzavirus 4 infection associated with respiratory failure in an immunocompetent adult. Clin. Infect. Dis. 43:e71–76. Greninger, A. L., E. C. Chen, T. Sitter, A. Scheinerman, N. Roubinian, G. Yu, E. Kim, D. R. Pillai, C. Guyard, T. Mazzulli, P. Isa, Arias.F., J. Hackett, G. Schochetman, S. Miller, P. Tang, and C. Y. Chiu. 2010. A metagenomic analysis of pandemic influenza A (2009 H1N1) infection in patients from North America. PLoS ONE, in press. Ji, H., N. Masse, S. Tyler, B. Liang, Y. Li, H. Merks, M. Graham, P. Sandstrom, and J. Brooks. 2010. HIV drug resistance surveillance using pooled pyrosequencing. PLoS ONE 5:e9263. Tang, P., and C. Chiu. 2010. Metagenomics for the discovery of novel human viruses. Future Microbiol. 5:177–189. Urisman, A., K. F. Fischer, C. Y. Chiu, A. L. Kistler, S. Beck, D. Wang, and J. L. DeRisi. 2005. E-Predict: a computational strategy for species identification based on observed DNA microarray hybridization patterns. Genome Biol. 6:R78. Victoria, J. G., A. Kapoor, L. Li, O. Blinkova, B. Slikas, C. Wang, A. Naeem, S. Zaidi, and E. Delwart. 2009. Metagenomic analyses of viruses in stool samples from children with acute flaccid paralysis. J. Virol. 83:4642– 4651. Wang, D., L. Coscoy, M. Zylberberg, P. C. Avila, H. A. Boushey, D. Ganem, and J. L. DeRisi. 2002. Microarray-based detection and genotyping of viral pathogens. Proc. Natl. Acad. Sci. USA 99:15687–15692. 2011 Career Development Grants for Postdoctoral Women …to enhance the careers of postdoctoral women of outstanding scientific accomplishment and potential for significant research in microbiology… Any woman scientist in the United States holding a doctoral degree and performing postdoctoral work in the areas of microbiology represented by ASM Scientific Divisions may apply. Currently, three grants of $1200 are given annually. Nomination requirements: x A Candidate Statement & CV x Nominating and Seconding Letters of Support x Candidates & Nominators must be ASM members x Nominations must be postmarked by February 1st Email nominations to [email protected] or mail to: ASM Membership Board Selection Committee Career Development Grants for Postdoctoral Women American Society for Microbiology 1752 N Street, NW, Washington, DC 20036-2804 Additional eligibility requirements exist. Information can be found on the ASM Website at http://www.asm.org/index.php?option=com_content&view= article&id=37857&Itemid=199 20 Y Microbe / Volume 6, Number 1, 2011 ASM ASM INTERNATIONAL AFFAIRS GRANT PROGRAMS VISITING RESOURCE PERSON Planning an international trip within the next 6 months? UNESCO and ASM offer you the chance to extend your stay and share your knowledge with scientists from around the world! If you will be traveling to a developing nation within the next 6 months, the UNESCO/ASM International Visiting Resource Person (VRP) Program can offer you the opportunity to spend an extra day with colleagues at a university or research institute in a major city. UNESCO will provide funds to cover the cost of the extended stay, making this program a cost-effective way to share your knowledge with others while enhancing your experience of the country. Application Form and Procedure: To learn about the application procedures and to download the application form please visit: http://www.asm.org/International/vrp Deadline: Rolling