Download Microarrays and Deep Sequencing in Clinical Microbiology

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