DTIC ADA585214: A Host Transcriptional Signature for Presymptomatic Detection of Infection in Humans Exposed to Influenza H1N1 or H3N2

0.68 MB·English

by  Defense Technical Information Center

A Host Transcriptional Signature for Presymptomatic
Detection of Infection in Humans Exposed to Influenza
H1N1 or H3N2
ChristopherW.Woods1,2,3*.,MicahT.McClain2,3.,MinhuaChen4,AimeeK.Zaas1,2,BradlyP.Nicholson3,
JayVarkey2,TimothyVeldman1,StephenF.Kingsmore5,YongshengHuang6,RobertLambkin-Williams7,
Anthony G. Gilbert7, Alfred O. HeroIII6, Elizabeth Ramsburg8, Seth Glickman9, Joseph E. Lucas1,
Lawrence Carin4, Geoffrey S. Ginsburg1*
1InstituteforGenomeSciencesandPolicy,DukeUniversityMedicalCenter,Durham,NorthCarolina,UnitedStatesofAmerica,2DivisionofInfectiousDiseases,Duke
UniversityMedicalCenter,Durham,NorthCarolina, UnitedStatesofAmerica,3DurhamVeteran’sAffairsMedicalCenter,Durham,NorthCarolina,UnitedStatesof
America,4DepartmentofElectricaland ComputerEngineering,DukeUniversity, Durham,NorthCarolina, United StatesofAmerica,5NationalCenterforGenome
Resources,SantaFe,NewMexico,UnitedStatesofAmerica,6CenterforComputationalBiologyandBioinformatics,UniversityofMichigan,AnnArobor,Michigan,United
StatesofAmerica,7RetroscreenVirology,London,UnitedKingdom,8DukeHumanVaccineInstitute,DukeUniversityMedicalCenter,Durham,NorthCarolina,United
StatesofAmerica,9DepartmentofEmergencyMedicine,UniversityofNorthCarolina-Chapel-Hill,ChapelHill,NorthCarolina,UnitedStatesofAmerica
Abstract
There is great potential for host-based gene expression analysis to impact the early diagnosis of infectious diseases. In
particular,theinfluenzapandemicof2009highlightedthechallengesandlimitationsoftraditionalpathogen-basedtesting
forsuspectedupperrespiratoryviralinfection.WeinoculatedhumanvolunteerswitheitherinfluenzaA(A/Brisbane/59/2007
(H1N1) or A/Wisconsin/67/2005 (H3N2)), and assayed the peripheral blood transcriptome every 8 hours for 7 days. Of 41
inoculatedvolunteers,18(44%)developedsymptomaticinfection.Usingunbiasedsparselatentfactorregressionanalysis,
wegeneratedagenesignature(orfactor)forsymptomaticinfluenzacapableofdetecting94%ofinfectedcases.Thisgene
signatureisdetectableasearlyas29hourspost-exposureandachievesmaximalaccuracyonaverage43hours(p=0.003,
H1N1)and38hours(p-value=0.005,H3N2)beforepeakclinicalsymptoms.Inordertotesttherelevanceofthesefindingsin
naturallyacquireddisease,acompositeinfluenzaAsignaturebuiltfromthesechallengestudieswasappliedtoEmergency
Department patients where it discriminates between swine-origin influenza A/H1N1 (2009) infected and non-infected
individualswith92%accuracy.ThehostgenomicresponsetoInfluenzainfectionisrobustandmayprovidethemeansfor
detection before typical clinicalsymptoms are apparent.
Citation:WoodsCW,McClainMT,ChenM,ZaasAK,NicholsonBP,etal.(2013)AHostTranscriptionalSignatureforPresymptomaticDetectionofInfectionin
HumansExposedtoInfluenzaH1N1orH3N2.PLoSONE8(1):e52198.doi:10.1371/journal.pone.0052198
Editor:HermanTse,TheUniversityofHongKong,HongKong
ReceivedJune25,2012;AcceptedNovember15,2012;PublishedJanuary9,2013
Copyright: (cid:2) 2013 Woods etal. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricteduse,distribution,andreproductioninanymedium,providedtheoriginalauthorandsourcearecredited.
Funding: Funding for this study was providedby the US Defense Advanced Research Projects Agency (DARPA)through contract N66001-07-C-2024 (P.I.,
Ginsburg).Thefundershadnoroleinstudydesign,datacollectionandanalysis,decisiontopublish,orpreparationofthemanuscript.
CompetingInterests:TheviralsignaturedevelopedfromthisworkisindeedthesubjectofapendingpatentheldbyDr’sGinsburg,Woods,Zaas,Lucas,Hero
andCarin:PCTApplicationNo.PCT/US2010/036257,‘‘METHODSOFIDENTIFYINGINFECTIOUSDISEASEANDASSAYSFORIDENTIFYINGINFECTIOUSDISEASE’’(File
date 5/2010). However, this does not conflict with PLOS One policies and the research data will be made publically available according to the journal’’s
policy.Dr.GilbertandDr.Lambkin-WilliamsareemployedbyRetroscreenVirology,asmentionedintheAuthors’’affiliationsportionofthemanuscript.This
companywasinvolvedinoversightoftheconductionoftheviralchallengestudiesutilizedinthecurrentmanuscript.Thisdoesnotinanywayaltertheauthors’
adherencetoallofthePLOSONEpoliciesonsharingdataandmaterials.
*E-mail:[email protected](CW);[email protected](GG)
.Theseauthorscontributedequallytothiswork.
Introduction expression commonacross symptomaticindividuals fromall viral
challenges [4]. Furthermore, an analysis of publically available
Infectious disease diagnostics traditionally rely heavily on
peripheral blood-based gene expression data indicated that the
pathogen detection [1,2,3]. However, the development of repro-
respiratory viral signature could distinguish patients with symp-
duciblemeansforextractingRNAfromwholeblood,coupledwith
tomaticviralinfectionsfromthosewithbacterialinfectionsaswell
advanced statistical methodsforanalysisof complexdatasets,has
as fromhealthycontrols [4,5].
created the possibility of classifying infections based on host gene
TheemergenceofpandemicH1N1influenzain2009highlights
expressionprofiling.Werecentlydevelopedarobustwholeblood
the need for new approaches to diagnosis of respiratory tract
mRNA expression classifier for human respiratory viral infection
infections.Adiagnostictestthatcouldidentifypatientsbeforethe
atthetimeofmaximalsymptomsusingdatafromthreehumanviral
onset of symptoms (but after exposure) who will later become ill
challenge cohorts (rhinovirus, respiratory syncytial virus, and
would be an indispensible tool for guiding individual treatment
H3N2influenzaA)[4].Sparselatentfactoranalysisofperipheral
decisions when antiviral supplies may be limited. Furthermore,
blood mRNA expression data revealed a pattern of gene
PLOSONE | www.plosone.org 1 January2013 | Volume 8 | Issue 1 | e52198
Report Documentation Page Form Approved
OMB No. 0704-0188
Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and
maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information,
including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington
VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if it

does not display a currently valid OMB control number.
1. REPORT DATE 3. DATES COVERED
09 JAN 2013 2. REPORT TYPE 00-00-2013 to 00-00-2013
4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER
A Host Transcriptional Signature for Presymptomatic Detection of
5b. GRANT NUMBER
Infection in Humans Exposed to Influenza H1N1 or H3N2
5c. PROGRAM ELEMENT NUMBER
6. AUTHOR(S) 5d. PROJECT NUMBER
5e. TASK NUMBER
5f. WORK UNIT NUMBER
7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION
Duke University ,Department of Electrical and Computer REPORT NUMBER
Engineering,Durham,NC,27708
9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR’S ACRONYM(S)
11. SPONSOR/MONITOR’S REPORT
NUMBER(S)
12. DISTRIBUTION/AVAILABILITY STATEMENT
Approved for public release; distribution unlimited
13. SUPPLEMENTARY NOTES
14. ABSTRACT
There is great potential for host-based gene expression analysis to impact the early diagnosis of infectious
diseases. In particular, the influenza pandemic of 2009 highlighted the challenges and limitations of
traditional pathogen-based testing for suspected upper respiratory viral infection. We inoculated human
volunteers with either influenza A (A/Brisbane/59/2007 (H1N1) or A/Wisconsin/67/2005 (H3N2)), and
assayed the peripheral blood transcriptome every 8 hours for 7 days. Of 41 inoculated volunteers, 18 (44%)
developed symptomatic infection. Using unbiased sparse latent factor regression analysis we generated a
gene signature (or factor) for symptomatic influenza capable of detecting 94% of infected cases. This gene
signature is detectable as early as 29 hours post-exposure and achieves maximal accuracy on average 43
hours (p = 0.003 H1N1) and 38 hours (p-value = 0.005, H3N2) before peak clinical symptoms. In order to
test the relevance of these findings in naturally acquired disease, a composite influenza A signature built
from these challenge studies was applied to Emergency Department patients where it discriminates
between swine-origin influenza A/H1N1 (2009) infected and non-infected individuals with 92% accuracy.
The host genomic response to Influenza infection is robust and may provide the means for detection before
typical clinical symptoms are apparent.
15. SUBJECT TERMS
16. SECURITY CLASSIFICATION OF: 17. LIMITATION OF 18. NUMBER 19a. NAME OF
ABSTRACT OF PAGES RESPONSIBLE PERSON
a. REPORT b. ABSTRACT c. THIS PAGE Same as 9
unclassified unclassified unclassified Report (SAR)
Standard Form 298 (Rev. 8-98)
Prescribed by ANSI Std Z39-18
HostGenomicSignaturesDetectH1N1Infection
these early results may forecast epidemic/pandemic spread, (Fig.1)atanaverageof49.3hoursafterinoculation(range24–84
potentially mitigating pandemic intensity [6]. Although previous hours, median 48hrs). Subjects who became ill experienced
studies with dengue, melioidosis, tuberculosis, candidiasis, and maximalsymptomsonaverage90.6hoursafterinoculation(range
sepsis have focused on diagnosis in patients as they present with 60to108hours,median96hours).Forthesesubjectstheaverage
active disease [7,8,9,10,11,12], we utilized human influenza total5daysymptomscorewas21.1(range6–43)withanaverage
challenge cohorts with a defined inoculation event coupled with daily peak of 7.3(range 2–13).
dense serial sampling to explore the ability of modern genomic For both challenge studies, only those individuals achieving
and statistical techniques to accurately classify individuals with both clear clinical and virologic endpoints were analyzed as true
multiple subtypes of influenza infection as early as possible influenza ‘infection’ (see Methods, Table s3). In our challenge
following viral exposure. Through this method, we have demon- studies there were four major outcome groups despite historical
strated the potential for a robust host gene response signature in and immunologic screening and similar inoculations [13]. Most
pre-symptomatic human infection and suggest the utility of this individuals fall within our two analysis groups – those who are
approachfordetectingpandemicH1N1infectioninanacutecare symptomatic-infected or asymptomatic-uninfected. However, a
setting. few individuals demonstrate mixed phenotypes and are either
symptomatic-uninfected(symptomsbutnoviralsheddingdetected,
Results see Methods) or asymptomatic-infected individuals (never symp-
tomatic but clear viral shedding on multiple days (Table s3). We
Healthy Volunteers Demonstrate Variable Clinical
have focused this analysis on those subjects with the clear
Responses to Inoculation with Seasonal Influenza H1N1 phenotypes of ‘infected’ and ‘uninfected’ (see Methods for
and H3N2 phenotyping criteria). The development of biomarkers for
FortheH1N1challengeweinoculated24volunteersage20–35 asymptomatic-infected and symptomatic-uninfected and a under-
with influenza A (A/Brisbane/59/2007). Nine (38%) of the 24 standingtheirunderlyingbiologywouldbeinvaluable,andcould
inoculated subjects developed symptoms consistent with viral potentially inform our ability to forecast and track epidemics.
upper respiratory infection with confirmed shedding of challenge However,thenumbersofsuchindividualsfromthecurrentstudies
virus (Fig. 1). This infection rate is similar to previous viral are insufficient formeaningfulanalysis at thistime.
challenge studies [13], and occurs despite similar patient profiles, Influenza-induced host gene expression groups into
vaccination history, and baseline influenza hemagglutination and unbiased time-evolving factors Whole blood RNA was
neutralization titers (Sup Tables s1 and s2). Subjects exhibited isolated from each individual every 8 hours from inoculation
variability of time to initiation of symptoms as well as maximal through day 7 and assayed by Affymetrix U133a 2.0 human
severityofsymptomsachieved(Fig.s1),butsymptomonsetbegan microarrays.Co-expressedgenetranscriptfactorsweregenerated
an average of 61.3 hours after inoculation (range 24–108hrs, through sparse latent factor regression analysis to provide an
median 72hrs). Subjects who became ill experienced maximal unbiased (unlabeled) examination of gene expression [15]. This
symptomsonaverage102.7hoursafterinoculation(range60–120 methodology specifically selects gene ‘factors’, with each factor
hours, median 108hrs). For symptomatic subjects, the average effectively defining a specific, limitedsubset of genes that are up-
total5daysymptomscorewas19.7(range6–34)withanaverage or down-regulated in a given condition. Sparse latent factor
daily peak of 7.4(range 4–13,Table s3). regression analysis permits an unbiased selection of these co-
For the H3N2 challenge (A/Wisconsin/67/2005) reported regulated genes while simultaneously filtering the tremendous
previously [4,14], we inoculated 17 volunteers (mean age 27 number of genes tested into smaller, more manageable, biologi-
years; range 22–41 years). For the 9 (53%) symptomatic-infected callyconnectedsubsets(seeMethods).Baseduponthequantitative
subjects,symptomonsetbeganearlierthanintheH1N1challenge level of over- or under-expression of the individual genes in a
Figure 1. Clinical response to viral challenge. Average symptom scores over time of individuals with both clinical and microbiologically
confirmedinfection(symptomatic-infected)followingexperimentalviralinoculationwithH1N1(blue)andH3N2(red).
doi:10.1371/journal.pone.0052198.g001
PLOSONE | www.plosone.org 2 January2013 | Volume 8 | Issue 1 | e52198
HostGenomicSignaturesDetectH1N1Infection
factor, a factor score is computed for a given factor in a given found that the signature is highly conserved across the two
sampleatagiventime.Ineachindividual,thefactorscoreforeach differentviruses.ForH1N1,thederivedfactorcorrectlyidentifies
group of co-expressed genes evolve as they progress through the our phenotypically confirmed individuals exposed to H1N1 as
various stages of disease (Fig. 2a, b). Furthermore, within each symptomatic-infected or asymptomatic-uninfected with only a
factor,theindividualgenesthemselvesexhibitvariableexpression single misclassification, whereas the H3N2 factor correctly
overtime(Fig.2c,d,Fig.s2),andthereforeeachgene’sindividual identifies100%ofindividualsexposedtoH3N2withaconfirmed
contribution to a single factor score continuously evolves, phenotype.
highlighting the complexity of the temporal dynamics of the host Theperformanceoftwoseparateclinicalchallengestudieswith
response to influenza challenge. The factor score provides a closelyrelatedvirusespermitsthevalidationoftheindependently
coherent representation of the aggregate of these co-expressed derived gene signatures by testing them on the subjects from the
genesatagiventime-pointallowingforamoremanageablemeans alternatestudy.WhenthefactorloadingsforH1N1areappliedto
of expressing biologically relevant genomicvariance overtime. thesubjectsfromtheH3N2study,theH1N1factor iscapable of
accurately discriminating between symptomatic-infected or
AWholeBloodRNA-basedGeneSignatureDifferentiates asymptomatic-uninfected H3N2 subjects 100% of the time (Fig.
Symptomatic Influenza A H1N1 or H3N2 Infection from s3). Similarly, when applied to the H1N1 data set, the H3N2
factorcorrectlyidentifies93%(14/15)ofthesubjectsintheH1N1
Asymptomatic Individuals
studyassymptomatic-infectedorasymptomatic-uninfected.Thus,
Similar to our previous work [4], in each challenge a single
each of the independently derived factors for H1N1 or H3N2
factor emerged as best able to discriminate symptomatic-infected
performs well when applied to a completely separate data set
subjects from asymptomatic-uninfected subjects at the time of
comprised of individualswith asimilar yet distinct pathogen.
maximal symptoms (Fig. 2). We derived this gene signature or
‘‘Influenza Factor’’ individually for both H1N1 and H3N2 and
Figure2.Geneexpressionsignaturesexpressedthroughfactorscores.Aninfluenzageneexpressionsignature,orfactor,evolvesovertime
insymptomaticindividuals(bluedots)anddistinguishesbetweensymptomaticandasymptomaticindividuals(reddots)forbothH1N1(A)andH3N2
(B)virusesatlatertimepoints.Heatmapsofthetop50genesinthediscriminativefactorforH1N1(c)andH3N2(d)astheydevelopovertimeare
shown.
doi:10.1371/journal.pone.0052198.g002
PLOSONE | www.plosone.org 3 January2013 | Volume 8 | Issue 1 | e52198
HostGenomicSignaturesDetectH1N1Infection
Figure3.Geneexpressionsignaturetrajectoryovertime.ThemagnitudeoftheInfluenzaFactorvariesfrominoculationthroughresolutionof
disease,forbothH1N1(A)andH3N2(B)patients.Theaveragefactorscoreateachtimepointforbothsymptomatic(blue)andasymptomatic(red)
individualsareshown.Theaveragetimeofsymptomonset(graydashedline)andmaximalsymptoms(blackdashedline)aredepicted.
doi:10.1371/journal.pone.0052198.g003
Discriminatory Factors for H1N1 and H3N2 are Similar Influenza Factor shares only 65–69% of its genes with factors
and Include Genes Involved in the Antiviral Response describinginfectionwiththeseotherrespiratoryviruses,suggesting
both common ‘viral URI’ pathways as well as some degree of
Thegenesignaturesderivedindependentlyforthetwodifferent
strainsofinfluenzaarehighlysimilar,sharing44outofthetop50 etiologic specificity. The majority of the top 50 predictive genes
genes (88%, Table s4). However, the importance of these few containedineachfactor areknowntocharacterize hostresponse
disparities isunclear, as the discordant genes are not sufficient to to viral infection, and include RSAD2, the OAS family, multiple
allow for differentiation between the two viruses in our analysis. interferonresponseelements,themyxovirus-resistancegeneMX1,
When compared to our previous work with HRV and RSV, the cytokineresponsepathwaysandothers[16,17,18].Many(butnot
PLOSONE | www.plosone.org 4 January2013 | Volume 8 | Issue 1 | e52198
HostGenomicSignaturesDetectH1N1Infection
all) of the components of these gene sets can be combined into hours (H3N2, p-value=0.005) and 60 hours (H1N1, p-val-
networks that putatively describe interactions between factor- ue=0.003)post-inoculation.
derived genes in canonical inflammatory and antiviral pathways WedevelopedReceiverOperatingCharacteristic(ROC)curves
(Fig. s4). Furthermore, the high degree of similarity and cross- ateachtimepointtovisualizetheabilityoftheInfluenzaFactorto
applicability of the two signatures permit the mathematical discriminate between symptomatic- infected and asymptomatic-
imputation of a combined ‘‘Influenza Factor’’ that retains the uninfectedsubjects(Figures7).ForH3N2infection,thefactorscan
discriminatory characteristics of the individual factors when distinguish between symptomatic and asymptomatic individuals
applied tobothcohorts (Fig.s5). with a sensitivity of 89% without false positives at 53 hours post-
exposure.By69hourspost-inoculationthesensitivityisincreased
TheInfluenzaFactorTracksCloselywithSymptomScores to100%.ForH1N1,thisoccursslightlylaterbutby60hourspost-
over Time and is Capable of Identifying Symptomatic- exposure the Influenza Factor demonstrates a sensitivity of 89%
infected Individuals Before the Time of Maximal Illness without false positives. These time points that the gene signature
first effectively discriminates symptomatic vs. asymptomatic
We next sought to define the clinical performance of the
subjectsusuallyprecedeorcoincidewiththetimeofaveragefirst
Influenza Factor over time. Just as symptom scores, time of peak
symptom onset (49 hrs for H3N2 and 61 hours for H1N1), and
symptoms, and symptom progression vary over time between
occurwell beforeclinically significantsymptoms(38hoursbefore
individuals (Fig. 1), the rise and fall of the gene expression based
maximal symptoms forH3N2and43hours forH1N1).
factorscorevariesaswell,andacommonfactortrajectorycanbe
mathematically imputed for all symptomatic subjects (Fig. 3a–b).
TheInfluenzaFactorAccuratelyIdentifiesPandemic2009
The trajectory of the Influenza Factor for symptomatic, infected
individuals first begins to diverge from asymptomatic, uninfected H1N1 Infections in a Clinical Cohort
individuals at 35–40% of the elapsed time between inoculation In order to assess the validity of the experimentally derived
andthetimeofmaximalsymptomsforeachindividual(38hours Influenza Factor to perform in a free-living (non-experimental)
post-inoculation for H1N1 and 29 hours for H3N2, Fig. 3a–b). settingweusedacohortofindividualsenrolledduringthe2009–
Even in this controlled challenge study among young, healthy 10Influenzaseason.Atthattime,weidentified36individualswho
individuals,wefindvariabilityinthistemporalrelationship,similar presentedtotheDukeUniversityHospitalemergencydepartment
to the individual variability seen with symptom scores. In most with symptomatic H1N1 infection (confirmed by RT-PCR), and
symptomaticindividuals,therise,peak,andfallofthefactorscore 45healthycontrols.PeripheralbloodRNAsampleswereobtained
trajectorytendstomimicincharacterbutprecedethechangesin fromthesymptomaticindividualsatthetimeofpresentationwith
theclinicalscore(Fig.s6).Evenwiththisvariabilityandrelatively symptomaticrespiratoryviralinfection.TheInfluenzaFactorwas
limited sample size (9 symptomatic-infected individuals in each applied to the microarray data derived from the blood RNA
study), the symptomatic-infected factor trajectory diverges by 53 samples and correctly identifies 92% (33/36) of the subjects as
N
Figure4.ValidationoftheInfluenzaFactorinareal-worldcohortofindividualspresentingwithconfirmedswine-origin2009A/
H1N1 infection. The Influenza Factor scores distinguish individuals with RT-PCR proven H1N1 infection ( ) from healthy individuals (#) as
demonstratedbothbyfactorscoreandbyROCcurveforhealthyvs.H1N1(insert,AUC0.98).
doi:10.1371/journal.pone.0052198.g004
PLOSONE | www.plosone.org 5 January2013 | Volume 8 | Issue 1 | e52198
HostGenomicSignaturesDetectH1N1Infection
infectedwithNovelH1N1,andcorrectlyidentified93%(42/45)of symptoms and disease transmission [24,25,26]. Furthermore, we
the healthy controls (Fig. 4). Overall, the Influenza Factor show that the overall trajectory of the Influenza Factor tracks
performed with an accuracy of 92.3% in the setting of a real- closelywithsymptomscoresovertime,butalsothattheobserved
world, independent cohortwithpandemic 2009H1N1infection. genomicresponsetendstosignificantlyprecedechangesinclinical
scoresinsymptomaticindividuals.Noneofouraffectedindividuals
Discussion developed severe infection, but the characteristics of the timing
anddevelopmentofthesesignaturessuggestthat,similartorecent
We performed two independent human viral challenge studies work with Dengue infections [9], genomic signatures may
(using influenza H1N1 and H3N2) to define the host-based potentially prove invaluable for predicting clinical outcomes.
peripheral blood gene expression patterns characteristic of the However,furtherlongitudinalstudieswithpatientswhoeventually
responsetoinfluenzainfection.Theresultsprovideclearevidence exhibit more severe disease will be required to fully assess this
that a biologically relevant peripheral blood gene expression potential.
signature can distinguish influenza infection with a remarkable The nature of the individual components of the genomic
degreeofaccuracyacrossthetwostrains.Wehavealsodefinedthe response to influenza infection and the biological pathways they
performance of the blood gene expression signature over time represent lend plausibility to this discovery. In particular,
throughout the complete course of human influenza infection. interferon stimulated pathways such as those including RSAD2,
Furthermore, despite arising from a controlled experimental IRF7,MX1, OAS3,MDA-5, RIG-Iand othersare incorporated
challenge setting, we demonstrate that an influenza signature is andthoughttodrivebothinnateand,toalesserdegree,adaptive
able to accurately identify individuals presenting with naturally- immuneresponsestoviralinfection[18,27,28,29].Manyofthese
occurring, RT-PCR confirmed H1N1 infection during the 2009 pathwaysareconsistentwiththoseidentifiedinacutelyillpediatric
pandemic. influenzasubjects[30]andrecentstudiesofthegenomicresponse
Definingtheetiologyofclinicalsyndromesinwhichinfectionis following vaccination with live, attenuated influenza vaccine
suspected remains challenging. Currently available influenza reported a profile of ‘immune activation’ which shares a number
diagnostic tests exhibit highly variable sensitivity, ranging from of genes with the Influenza Factor described here [31,32].
53 to 100% in various studies [19,20]. Importantly, even those Interestingly, a few genes which consistently feature prominently
withpowerfultestcharacteristicssuchasRT-PCRaredependent in the Influenza Factor are not clearly tied to inflammatory or
upon sampling technique and inclusion of virus-specific compo- immunologic pathways, and their significance remains unclear.
nents leading to reduced effectiveness with emerging viral strains Previouslypublishedworkwithbacterialrespiratoryinfectionshas
[21]. In addition to being less susceptible to sampling error, yielded quite different genomic results [4,5] suggesting that some
genomic signatures are not viral antigen or nucleic acid- aspectsofthehostresponsearespecificatleastformajorclassesof
dependent,andunlikelytobeasstrain-specificaspathogen-based pathogens (e.g., viral vs. bacterial). The genomic pathways
platforms.Therefore,inadditiontohighsensitivityinthecohorts identified suggest we are largely measuring indicators of the
studied [92%(95%CI79–99%for2009H1N1)],influenzagene development/amplification of the immune response to the virus
signatureshavetheaddedpotentialofbeingabletoidentify,inthe similar to previous work [13,16,32], and that these indicators
acute phase of illness, likely cases of infection with emerging parallel (and usually precede) clinical symptom development in
influenzastrainsforwhichaspecificdiagnosticplatformhasyetto time. The immunologic pathways observed in these studies that
be developed and distributed. The nature of challenge studies areknowntobecommonlyactivatedearlyonattheprimarysite
limits our ability to make direct comparisons to other infected ofinfection(i.e.,respiratoryepithelium)[29,33],exhibitrelatively
states – however, our previous work has demonstrated that delayedappearanceintheperiphery.Thisdelayseemslogical,as
genomic signatures similarly derived from viral challenges are earlyinnateresponsesatthesiteofinfectionwouldbeexpectedto
capable of distinguishing upper respiratory viral infection from have an initially minor impact on global peripheral gene
pneumonia due to Streptococcus pneumoniae [4]. These findings are expression.Atveryearlytime-points(,53hrsfollowingexposure)
promising but additional testing of these signatures in other insufficient numbers of peripheral cells are undergoing the
models, including acute human cases of bacterial infection, will conserved stimulation required to produce a significant change
need tobeperformed tobetterdelineate theirspecificity. in global gene expression, at least as detected by microarray
The unique design and frequent sampling involved in two analysis.Thisraisesthepossibilitythatmoresensitivemethodsof
experimentalchallengestudieshasalsogivenusthesingularability detecting genomic changes, such as individual cell-type sampling
toexaminethedynamicsoftemporaldevelopmentofthegenomic orRT-PCRofselectgenes,willprovetobeevenmorepreciseat
responses following exposure to infectious virus. We have shown early time points in the evolution of viral infection. Additional
that when viewed through the lens of the genomic response, it is work will be essential (and is underway) to further define the
possible to correctly distinguish individuals as infected or natureandbiologicalimplicationsofthesedata,aswellastowork
uninfectedwithinfluenzawellbeforetheyhaveclinicallyrelevant towardsdevelopmentofamorepracticalmeansofassayingthese
symptomsorwouldbeillenoughtopresentforclinicalevaluation. changes in the clinical setting, such as RT-PCR of select ‘core’
The potential power of this approach is manifested by full genes fromsignatures like theonedescribed herein.
discriminativeabilityofthegenomicsignatureasearlyas53hours Clearly,greatcaremustbetakenwhenanalyzingandapplying
post-viral exposure, at a time when the average clinical score of hostgenomicdatafromhumanchallengestudieswherethemeans
symptomaticindividualsisonly2.4.Symptomsofthisnatureand oftransmissionofthevirusisexperimentallydesignedratherthan
severity are clinically vague and would be typical of very mild ‘natural’, and the degree of illness which follows is not always
allergies [22] or even symptoms due to sequelae of chronic typical of the severity seen in naturally acquired infection in
smoking [23]. Therefore, genomic analyses demonstrate the subjectswhopresentforclinicalcare,eventhoughitdoestendto
potentialtoidentifyviralinfectioneitherbeforesymptomsemerge mimictheoverallcharacterofnaturalclinicaldisease[13].Hosts
or among what otherwise are common, nonspecific upper in these studies are universally young, healthy individuals at
respiratory symptoms, when early intervention with antiviral minimalriskfordevelopingseverecomplications,whichmaylimit
medications could have profound impact on both individual thebroadapplicabilityofsuchfindings,althoughthisissomewhat
PLOSONE | www.plosone.org 6 January2013 | Volume 8 | Issue 1 | e52198
HostGenomicSignaturesDetectH1N1Infection
mitigatedbythestrongperformanceofthegenesignaturesdespite Clinical Case Definitions
significant clinical variability in infected subjects. It is also Symptoms were recorded twice daily using a modified
important to note that while this type of factor analysis allows standardized symptom score [35]. The modified Jackson Score
for description of conserved biological pathways indicative of requires subjects torank symptoms of upper respiratory infection
influenza infection, a given factor only represents a limited (stuffynose,scratchythroat,headache,cough,etc)onascaleof0–
interrelated subset of all genes that are globally up- or down- 3 of‘‘nosymptoms’’,‘‘just noticeable’’, ‘‘bothersome butcan still
regulated in response to a given condition, and thus does not do activities’’ and ‘‘bothersome and cannot do daily activities’’.
describe theentirety ofthegenomicresponse. For all cohorts, modified Jackson scores were tabulated to
Despitetheselimitations,wehaveforthefirsttimedefinedthe determine if subjects became symptomatic from the respiratory
temporal dynamics of a genomic signature driving the host viral challenge. Symptom onset was defined as the first of 2
response to influenza infection in humans. These molecular and contiguousdayswithscoreof2ormore.AmodifiedJacksonscore
statistical techniques combined with the ability to longitudinally of$6overaconsecutivefivedayperiodwastheprimaryindicator
study exposed human hosts have given us the opportunity to ofsymptomaticviralinfection[36]andsubjectswiththisscoreand
examine periods of human disease which have previously been a positive qualitative viral culture or quantitative RT-PCR for at
largely unexplored. Moreover, despite being developed in an least 2 consecutive days (beginning 24 hours after inoculation)
experimental challenge model, this host genomic signature were denoted as ’’symptomatic infection’’ and included in the
performs at a high level of accuracy in the setting of naturally signature performance analyses. [35,36,37]. Subjects were classi-
acquiredpandemic2009H1N1infection.Thisworkdemonstrates fiedas‘‘asymptomatic,notinfected’’ifthesymptomscorewasless
that analyses of the temporal development of gene expression than6overthefivedaysofobservationandviralsheddingwasnot
signatures shows promise both for creating diagnostics for early documented after the first 24 hours subsequent to inoculation as
detection,aswellasprovidinginsightintothebiologyofthehost above.Standardizedsymptomscoresweretabulatedattheendof
response toinfluenza andother pathogens. each study to determine attack rate and time of maximal
symptoms. Some subjects in each study (2 H3N2 and 8 H1N1
Materials and Methods subjects)demonstratedanoverallpicturethatfellinbetweenthese
twocategories.Theseindividualswereeither‘asymptomatic viral
Institutional Review Board Approvals shedders’ (2 H3N2 and 5 H1N1) or ‘symptomatic non-viral
The Influenza challenge protocols were approved by the East shedders’(0H3N2and3H1N1).Oneadditionalindividualinthe
London and City Research Ethics. Committee 1 (London, UK), H1N1 study was excluded due to additional infection acquired
an independent institutional review board (WIRB: Western. duringthestudy.Giventheheterogeneityoftheiroverall‘infected’
Institutional Review Board; Olympia, WA), the IRB of Duke statustheseindividualswerenotincludedinperformanceanalyses.
UniversityMedicalCenter.(Durham,NC),andtheSSC-SDIRB
(US Department of Defense; Washington, DC) and were Pandemic 2009 H1N1 Real-World Cohort
conducted in accordance with the Declaration of Helsinki. All
Subjects were recruited from the Duke University Medical
subjects enrolled in viral challenge studies provided written CenterEmergencyDepartment(DUMC-Level1TraumaCenter
informed consent per standard IRB protocol. Funding for this with annual census of 65,000). This study was approved by the
study was provided by the US Defense Advanced Research Institutional Review Board at each institution and written,
Projects Agency (DARPA) through contract N66001-07-C-2024 informed consent was obtained by all study participants or their
(P.I.,Ginsburg). legaldesignates.SubjectswerescreenedbetweenSeptember1and
December31,2009.Subjectswereconsideredfortheenrollmentif
Human Viral Challenges theyhadaknownorsuspectedinfluenzainfectiononthebasisof
IncollaborationwithRetroscreenVirology,Ltd(London,UK), clinicaldataatthetimeofscreeningandiftheyexhibitedtwoor
weintranasallyinoculated24healthyvolunteerswithinfluenzaA more signs of systemic inflammation (SIRS) within a 24-hour
H1N1 (A/Brisbane/59/2007). All volunteers provided informed period. Subjects were excluded if ,18 years old, if they had an
consentandunderwentextensivepre-enrollmenthealthscreening, imminently terminal co-morbid condition, if they had recently
includingbaselineantibodytiterstothespecificstrainsofinfluenza been treated with an antibiotic for a viral, bacterial, or fungal
utilized. After 24hrs in quarantine, we instilled one of four infection,oriftheywereparticipatinginanongoingclinicaltrial.
dilutions(1:10,1:100,1:1000,1:10000)of107TCID influenzaA Trainedstudycoordinatorsateachsitereviewedandabstracted
50
into bilateral nares of subjects (groups of 4–6 for each dilution) vitalsigns,microbiology,laboratory,andimagingresultsfromthe
using standard methods. [4] The virus was manufactured and initial ED encounter and at 24-hour intervals if patient was
processedundercurrentgoodmanufacturingpractices(cGMP)by admitted. Following hospital discharge, research personnel
BaxterBioScience,(Vienna,Austria).Atpre-determinedintervals abstracted the duration of hospitalization, length of ICU stay,
(q8hforthefirst5dfollowinginoculation),wecollectedbloodinto in-hospitalmortality,timingandappropriatenessofantimicrobial
RNAPAXGeneTMcollectiontubes(PreAnalytix;FranklinLakes, administration,andmicrobiologic-cultureresultsfromthemedical
NJ)accordingtomanufacturers’specifications.Weobtainednasal record.Inadditiontoresidualrespiratorysamplescollectedaspart
lavagesamplesfromeachsubjectdailyforqualitativeviralculture of routine care, an NP swab was collected from each enrolled
and and/or quantitative influenza RT-PCR to assess the success subject. Total nucleic acids were extracted from nasal swab or
and timing of infection [34]. Blood and nasal lavage collection washisolateswiththeEZ1BiorobotandtheEZ1VirusMiniKit
continuedthroughoutthedurationofthequarantine.Allsubjects v2.0(Qiagen).2009H1N1viruswasconfirmedin20uldetection
received oral oseltamivir (Roche Pharmaceuticals) 75mg by reactions, Qiagen One-Step RT-PCR (Qiagen) reagents on a
mouth twice daily as treatment or prophylaxis at day 6 following LightCycler v2.0 (Roche) using the settings and conditions
inoculation.Allsubjectswerenegativebyrapidantigendetection recommendedintheCDCRealtimeRTPCR(rRTPCR)Protocol
(BinaxNow Rapid Influenza Antigen; Inverness Medical Innova- for Detection and Characterization of Swine Influenza (version
tions, Inc) at time of discharge. Detailed methods of the H3N2 2009). The primers and probes were as described in the CDC
Challenge study havebeen reported previously[4,14]. protocol and obtained from Integrated DNA Technologies. We
PLOSONE | www.plosone.org 7 January2013 | Volume 8 | Issue 1 | e52198
HostGenomicSignaturesDetectH1N1Infection
included mRNA expression data obtained concurrently with the whether the sample will be symptomatic, but we underscore that
H1N1 cohortfrom45gender-matched, healthycontrols. this symptomatic/asymptomatic information is not employed in
themodel.
RNA Purification and Microarray Analysis
For each challenge, we collected peripheral blood at 24 hours Supporting Information
prior to inoculation with virus (baseline), immediately prior to
Figure S1 For the H1N1 Challenge Trial, individual
inoculation(pre-challenge)andatsetintervalsfollowingchallenge.
symptom scores of symptomatic infected patients from
RNA was extracted at Expression Analysis (Durham, NC) from
the time of inoculation (time 0) through the end of the
whole blood using the PAXgeneTM 96 Blood RNA Kit
study.
(PreAnalytiX, Valencia, CA) employing the manufacturer’s
(PDF)
recommended protocol. Complete methodology can be viewed
intheMethodsS1.Hybridizationandmicroarraydatacollection Figure S2 Variation over time of the expression of the
was also performed at Expression Analysis (Durham, NC) using top 30 individual genes which make up the Influenza
the GeneChipH Human Genome U133A 2.0 Array (Affymetrix, factor.
Santa Clara, CA). Microarray data used for this study will be (PDF)
deposited inGEO prior topublication.
Figure S3 Cross-validation of H1N1 (Top) and H3N2
(Bottom) derived factors.
Statistical Analyses
(PDF)
FollowingRMAnormalizationofrawprobedata,sparselatent
factor regression analysis was applied to each dataset Figure S4 Genes comprising the discriminative. Factor
[38,39,40,41]. This reduces the dimensionality of the complex for Influenza infection are involved in canonical antiviral
geneexpressionarraydatasetassumingthatmanyoftheprobesets pathways, suchas theSTAT-1 dependent portions of Interferon-
ontheexpressionarraychiparehighlyinterrelated(targetingthe responseanddsRNA-inducedinnatesignalingdepictedhere(top),
samegenesorgenesinthesamepathways).Dimensionreduction and the IRF-7 and RIG-I, MDA-5 dependent portions of
isperformedbyconstructingfactors(groupsofgeneswithrelated Interferon-responseandssRNA-inducedinnatesignaling(bottom,
expression values). These factors are used in a sparse linear www.genego.com). Pathways impacted by genes from the
regression framework to explain the variation seen in all of the discriminative Factorsare markedwith ared target symbol.
probe sets. By default, most of the coefficients in this linear (PDF)
regression are zero. Thus, a small number (e.g., 50) of factors Figure S5 Temporal development of the combined
explain variationseen inany singledataset. Influenza Factor applied to H1N1 (pp top) and H3N2
Factor loadings are defined as the coefficients of the factor (bottom) cohorts.
regression,and,toexplorethebiologicalrelevanceanyparticular (PDF)
factor,weexaminethegenesthatare"in"thatfactor–thegenes
FigureS6 InfluenzaFactorscorecomparedwithclinical
that show significantly non-zero factor loadings. ‘‘Factor scores"
symptomscoreovertimeforallindividualsinthestudy.
are defined as the vector that best describes the co-expression of
(PDF)
the genes in a particular factor. Both factor loadings and factor
scores are fit to the data concurrently, and the full details of the Figure S7 Performance of the Influenza Factor. The
process can be found in the supplementary statistical analysis Influenza Factor develops accurate discriminative utility early in
section. While 50 factors were used for the results reported here, thecourseofinfluenzainfection,asillustratedbyROCcurvesfor
we also considered 20, 30 and 40, with minimal effect on the the Factor at each successive timepoint. Depicted are: H1N1-
significant factor loadings. Notably, the initial models built to derivedFactorappliedtoH1N1subjects(A),H3N2Factorapplied
determinefactorsthatdistinguishsymptomaticinfectedindividuals toH1N1subjects(B),H1N1FactorappliedtoH3N2subjects(C),
fromasymptomaticindividualswerederivedusinganunsupervised andtheH3N2 Factorapplied toH3N2subjects (D).
process(i.e.,themodelclassifiedsubjectsbasedongeneexpression (PDF)
patternalone,withoutaprioriknowledgeofinfectionstatus).
Table S1 Patient demographics and pre-challenge se-
Ourstatisticalmodelisunsupervised,andthusseekstodescribe
rology for HAI titers to challenge viruse (H1N1). Unique
the statistical properties of the expression data without using
ID’sin Blueindicate ‘symptomatic infected’ individuals.
labeleddata.Suchunsupervisedalgorithmsmayuncoverstatistical
(PDF)
characteristics that distinguish symptomatic and asymptomatic
subjects, but this relationship is inferred a posteriori. The Table S2 Patient demographics and pre-challenge se-
unsupervised models are not explicitly designed to perform rologyforHAItiterstochallengeviruse(H3N2).Unique
classification. The specific unsupervised model employed here ID’sinBlueindicate‘symptomaticinfected’individuals.
correspondstoBayesianfactoranalysis.Thismodelrepresentsthe (PDF)
gene-expression values of each sample in terms of a linear
TableS3 CompletesubjectlistforbothH1N1andH3N2
combination of factors. Within the model we impose that each
viral challenge trials, with total symptom scores and
factorissparse,meaningthatonlyarelativelysmallfractionofthe
clinical/virologic classifications.
genes have non-zero expression within the factor loading. This
(PDF)
sparseness seeks to map each factor to a biological pathway by
identifying genes which are co-expressed, and each pathway is Table S4 Comparison of the top 50 genes from the
assumedtoberepresentedintermsofasmallfractionofthetotal discriminative factors derived from H1N1 and H3N2
numberofgenes.Thenumberoffactorsappropriateforthedata challenge trials, ranked by order of individual contri-
isinferred,usingastatisticaltooltermedthebetaprocess[15].We butiontothestrengthoftheFactor(highestcontributors
have found that, for the virus data considered here, the factor atthe top).
score associated with one of these factors is a good marker as to (PDF)
PLOSONE | www.plosone.org 8 January2013 | Volume 8 | Issue 1 | e52198
HostGenomicSignaturesDetectH1N1Infection
MethodsS1 Additional material defining the statistical Author Contributions
models used are presented.
Conceived and designed the experiments: CWW GSG TV LC RL-W
(PDF)
AGG.Performedtheexperiments:CWWMTMBNJVRL-WAGGSG
ER.Analyzedthedata:CWWMTMMCAKZYHAOHJLLCGSG.
Contributed reagents/materials/analysis tools: SFK YH RL-W AGG
AOHERJLLC.Wrotethepaper:CWWMTMAKZGSG.
References
1. Chiarini A, Palmeri A, Amato T, Immordino R, Distefano S, et al. (2008) 22. Calpin C, Macarthur C, Stephens D, Feldman W, Parkin PC (1997)
DetectionofbacteriaandyeastspeciesbytheBACTEC9120automatedsystem Effectivenessofprophylacticinhaledsteroidsinchildhoodasthma:asystemic
withtheroutineuseofaerobic,anaerobic,andfungalmedia.JClinMicrobiol. reviewoftheliterature.JAllergyClinImmunol100:452–457.
2. LambertSB,WhileyDM,O’NeillNT,AndrewsEC,CanavanFM,etal.(2008) 23. WalkerFD,WhitePS(2000)Sinussymptomscores:whatistherangeinhealthy
Comparing nose-throat swabs and nasopharyngeal aspirates collected from individuals?ClinOtolaryngolAlliedSci25:482–484.
children with symptoms for respiratory virus identification using real-time 24. Siston AM, Rasmussen SA, Honein MA, Fry AM, Seib K, et al. (2010)
polymerasechainreaction.Pediatrics122:e615–620. Pandemic2009influenzaA(H1N1)virusillnessamongpregnantwomeninthe
3. RobinsonJL,LeeBE,KothapalliS,CraigWR,FoxJD(2008)Useofthroat UnitedStates.JAMA303:1517–1525.
swaborsalivaspecimensfordetectionofrespiratoryvirusesinchildren.Clin 25. Goldstein E, Cowling BJ, O’Hagan JJ, Danon L, Fang VJ, et al. (2010)
InfectDis46:e61–64. OseltamivirfortreatmentandpreventionofpandemicinfluenzaA/H1N1virus
4. ZaasAK,ChenM,VarkeyJ,VeldmanT,HeroAO3rd,etal.(2009)Gene infectioninhouseholds,Milwaukee,2009.BMCInfectDis10:211.
26. HaydenFG,TreanorJJ,FritzRS,LoboM,BettsRF,etal.(1999)Useofthe
expressionsignaturesdiagnoseinfluenzaandothersymptomaticrespiratoryviral
oral neuraminidase inhibitor oseltamivir in experimental human influenza:
infectionsinhumans.CellHostMicrobe6:207–217.
randomizedcontrolledtrialsforpreventionandtreatment.JAMA282:1240–
5. Ramilo O, Allman W, Chung W, Mejias A, Ardura M, et al. (2007) Gene
1246.
expression patterns in blood leukocytes discriminate patients with acute
27. Kochs G, Haller O (1999) Interferon-induced human MxA GTPase blocks
infections.Blood109:2066–2077.
nuclearimportofThogotovirusnucleocapsids.ProcNatlAcadSciUSA96:
6. Memoli MJ, Morens DM, Taubenberger JK (2008) Pandemic and seasonal
2082–2086.
influenza:therapeuticchallenges.DrugDiscovToday13:590–595.
28. OkumuraA,LuG,Pitha-RoweI,PithaPM(2006)Innateantiviralresponse
7. Zaas AK, Aziz H, Lucas J, Perfect JR, Ginsburg GS (2010) Blood gene
targetsHIV-1releasebytheinductionofubiquitin-likeproteinISG15.ProcNatl
expressionsignaturespredictinvasivecandidiasis.SciTranslMed2:21ra17. AcadSciUSA103:1440–1445.
8. Berry MP, Graham CM, McNab FW, Xu Z, Bloch SA, et al. (2010) An 29. Player MR, Torrence PF (1998) The 2–5A system: modulation of viral and
interferon-inducibleneutrophil-drivenbloodtranscriptionalsignatureinhuman cellularprocessesthroughaccelerationofRNAdegradation.PharmacolTher
tuberculosis.Nature466:973–977. 78:55–113.
9. NascimentoEJ,Braga-NetoU,Calzavara-SilvaCE,GomesAL,AbathFG,et 30. KawadaJ,KimuraH,KamachiY,NishikawaK,TaniguchiM,etal.(2006)
al.(2009)Geneexpressionprofilingduringearlyacutefebrilestageofdengue Analysisofgene-expressionprofilesbyoligonucleotidemicroarrayinchildren
infectioncanpredictthediseaseoutcome.PLoSOne4:e7892. withinfluenza.JGenVirol87:1677–1683.
10. TangBM,McLeanAS,DawesIW,HuangSJ,LinRC(2009)Gene-expression 31. ZhuW,HiggsBW,MorehouseC,StreicherK,AmbroseCS,etal.(2010)A
profilingof peripheral bloodmononuclear cells in sepsis.Crit Care Med 37: wholegenometranscriptionalanalysisoftheearlyimmuneresponseinducedby
882–888. liveattenuatedandinactivatedinfluenzavaccinesinyoungchildren.Vaccine28:
11. Ramilo O, Mejias A (2009) Shifting the paradigm: host gene signatures for 2865–2876.
diagnosisofinfectiousdiseases.CellHostMicrobe6:199–200. 32. NakayaHI,WrammertJ,LeeEK,RacioppiL,Marie-KunzeS,etal.(2011)
12. PanklaR,BuddhisaS,BerryM,BlankenshipDM,BancroftGJ,etal.(2009) Systemsbiologyofvaccinationforseasonalinfluenzainhumans.NatImmunol.
Genomic transcriptional profiling identifies a candidate blood biomarker 33. BaskinCR,Bielefeldt-OhmannH,TumpeyTM,SabourinPJ,LongJP,etal.
signatureforthediagnosisofsepticemicmelioidosis.GenomeBiol10:R127. (2009)Earlyandsustainedinnateimmuneresponsedefinespathologyanddeath
13. Carrat F, Vergu E, Ferguson NM, Lemaitre M, Cauchemez S, et al. (2008) innonhumanprimatesinfectedbyhighlypathogenicinfluenzavirus.ProcNatl
Timelinesofinfectionanddiseaseinhumaninfluenza:areviewofvolunteer AcadSciUSA106:3455–3460.
challengestudies.AmJEpidemiol167:775–785. 34. GharabaghiF,TellierR,CheungR,CollinsC,BroukhanskiG,etal.(2008)
14. WilkinsonTM,LiCK,ChuiCS,HuangAK,PerkinsM,etal.(2012)Preexisting Comparison of a commercial qualitative real-time RT-PCR kit with direct
influenza-specific CD4+ T cells correlate with disease protection against immunofluorescenceassay(DFA)andcellculturefordetectionofinfluenzaA
influenzachallengeinhumans.NatMed18:274–280. andBinchildren.JClinVirol42:190–193.
35. JacksonGG,DowlingHF,SpiesmanIG,BoandAV(1958)Transmissionofthe
15. ChenB,ChenM,PaisleyJ,ZaasA,WoodsC,etal.(2010)Bayesianinferenceof
commoncoldtovolunteersundercontrolledconditions.I.Thecommoncoldas
thenumberoffactorsingene-expressionanalysis:applicationtohumanvirus
aclinicalentity.AMAArchInternMed101:267–278.
challengestudies.BMCBioinformatics11:552.
36. TurnerRB(2001)Ineffectivenessofintranasalzincgluconateforpreventionof
16. JulkunenI,SarenevaT,PirhonenJ,RonniT,MelenK,etal.(2001)Molecular
experimentalrhinoviruscolds.ClinInfectDis33:1865–1870.
pathogenesis of influenza A virus infection and virus-induced regulation of
37. BarrettB,BrownR,VolandR,MaberryR,TurnerR(2006)Relationsamong
cytokinegeneexpression.CytokineGrowthFactorRev12:171–180.
questionnaireandlaboratorymeasuresofrhinovirusinfection.EurRespirJ28:
17. EhrhardtC,SeyerR,HrinciusER,EierhoffT,WolffT,etal.(2010)Interplay
358–363.
betweeninfluenzaAvirusandtheinnateimmunesignaling.MicrobesInfect12:
38. Carvalho C, Lucas J, Wang Q, Chang J, Nevins JR, et al. (2008) High
81–87.
dimensionalsparsefactormodelling:applicationsingeneexpressiongenomics.
18. Katze MG, Fornek JL, Palermo RE, Walters KA, Korth MJ (2008) Innate JournalofAmericanStatisticalAssociation.
immune modulation by RNA viruses: emerging insights from functional 39. LucasJE,CarvalhoCM,MerlD,West M(2008)In-Vitro toIn-Vivofactor
genomics.NatRevImmunol8:644–654. profiling in expression genomics. In: Dey D, Ghosh S, Mallick B, editors.
19. HurtAC,BaasC,DengYM,RobertsS,KelsoA,etal.(2009)Performanceof BayesianModelinginBioinformatics:Taylor-Francis.
influenza rapid point-of-care tests in the detection of swine lineage A(H1N1) 40. Lucas JE, Carvalho CM, Wang Q, Bild A, Nevins JR, et al. (2006) Sparse
influenzaviruses.InfluenzaOtherRespiViruses3:171–176. statisticalmodellingingeneexpressiongenomics.BayesianInferenceforGene
20. Faix DJ, Sherman SS, Waterman SH (2009) Rapid-test sensitivity for novel ExpressionandProteomics:CambridgeUniversityPress.155–176.
swine-origininfluenzaA(H1N1)virusinhumans.NEnglJMed361:728–729. 41. Wang Q, Carvalho CM, Lucas JE, West M (2007) BFRM: Bayesian factor
21. BoggildAK,McGeerAJ(2010)Laboratorydiagnosisof2009H1N1influenzaA regressionmodeling.BulletinoftheInternationalSocietyofBayesianAnalysis
virus.CritCareMed38:e38–42. 14:4–5.
PLOSONE | www.plosone.org 9 January2013 | Volume 8 | Issue 1 | e52198