Genetic and Genomic Architecture of the Evolution of Resistance to Antifungal Drug Combinations

22 Pages·2013·3.22 MB·English

Genetic and Genomic Architecture of the Evolution of
Resistance to Antifungal Drug Combinations
Jessica A. Hill1, Ron Ammar1,2, Dax Torti3, Corey Nislow1,2, Leah E. Cowen1*
1DepartmentofMolecularGenetics,UniversityofToronto,Toronto,Ontario,Canada,2DonnellyCentreforCellularandBiomolecularResearch,UniversityofToronto,
Toronto,Ontario,Canada,3DonnellySequencingCentre,UniversityofToronto,Toronto,Ontario,Canada
Abstract
Theevolutionofdrugresistanceinfungalpathogenscompromisestheefficacyofthelimitednumberofantifungaldrugs.
Drugcombinationshaveemergedasapowerfulstrategytoenhanceantifungalefficacyandabrogatedrugresistance,but
theimpactontheevolutionofdrugresistanceremainslargelyunexplored.TargetingthemolecularchaperoneHsp90orits
downstreameffector,theproteinphosphatasecalcineurin,abrogatesresistancetothemostwidelydeployedantifungals,
the azoles, which inhibit ergosterol biosynthesis. Here, we evolved experimental populations of the model yeast
SaccharomycescerevisiaeandtheleadinghumanfungalpathogenCandidaalbicanswithazoleandaninhibitorofHsp90,
geldanamycin, or calcineurin, FK506. To recapitulate a clinical context where Hsp90 or calcineurin inhibitors could be
utilizedincombinationwith azolestorenderresistantpathogensresponsiveto treatment,theevolutionexperiment was
initiatedwithstrainsthatareresistanttoazolesinamannerthatdependsonHsp90andcalcineurin.Ofthe290lineages
initiated, most went extinct, yet 14 evolved resistance to the drug combination. Drug target mutations that conferred
resistance to geldanamycin or FK506 were identified and validated in five evolved lineages. Whole-genome sequencing
identified mutations in a gene encoding a transcriptional activator of drug efflux pumps, PDR1, and a gene encoding a
transcriptional repressor of ergosterol biosynthesis genes, MOT3, that transformed azole resistance of two lineages from
dependentoncalcineurintoindependentofthisregulator.Resistancealsoarosebymutationthattruncatedthecatalytic
subunitofcalcineurin,andbymutationinLCB1,encodingasphingolipidbiosyntheticenzyme.Genomeanalysisrevealed
extensiveaneuploidyinfouroftheC.albicanslineages.Thus,weidentifymoleculardeterminantsofthetransitionofazole
resistancefromcalcineurindependencetoindependenceandestablishmultiplemechanismsbywhichresistancetodrug
combinations evolves,providing a foundation forpredictingand preventing the evolutionofdrug resistance.
Citation: Hill JA, Ammar R, Torti D, Nislow C, Cowen LE (2013) Genetic and Genomic Architecture of the Evolution of Resistance to Antifungal Drug
Combinations.PLoSGenet9(4):e1003390.doi:10.1371/journal.pgen.1003390
Editor:JustinC.Fay,WashingtonUniversitySchoolofMedicine,UnitedStatesofAmerica
ReceivedDecember12,2012;AcceptedJanuary30,2013;PublishedApril4,2013
Copyright:(cid:2)2013Hilletal.Thisisanopen-accessarticledistributedunderthetermsoftheCreativeCommonsAttributionLicense,whichpermitsunrestricted
use,distribution,andreproductioninanymedium,providedtheoriginalauthorandsourcearecredited.
Funding:JAHwassupportedbyaNaturalSciencesandEngineeringResearchCouncilofCanada(NSERC)PGSDoctoralAward,anOntarioGraduateScholarship,
andaUniversityofTorontoOpenFellowship;RAbyanOntarioGraduateScholarship;DTbygrantsfromtheCanadaFoundationforInnovationtoBrenda
AndrewsandJackGreenblatt,whichestablishedtheDonnellySequencingCentre;CNbyCanadianCancerSocietygrant#20380;andLECbyaCareerAwardin
theBiomedicalSciencesfromtheBurroughsWellcomeFund,byaCanadaResearchChairinMicrobialGenomicsandInfectiousDisease,byaMinistryofResearch
andInnovationEarlyResearcherAward,andbyNSERCDiscoveryGrant355965-2009.Thefundershadnoroleinstudydesign,datacollectionandanalysis,
decisiontopublish,orpreparationofthemanuscript.
CompetingInterests:Theauthorshavedeclaredthatnocompetinginterestsexist.
*E-mail:[email protected]
Introduction combating infections [5]. Even with current treatment options,
mortalityratesduetoinvasivefungalinfectionsoftenexceed50%,
The evolution of drug resistance is a ubiquitous phenomenon and fungal pathogens kill as many people as tuberculosis or
thathasaprofoundimpactonhumanhealth.Withthewidespread malaria [6,7]. Thus, there is a pressing need to develop new
deployment of antimicrobial agents in both clinical and environ- strategies to enhance the efficacy of antifungal drugs and to
mental settings, the rate at which resistance evolves in pathogen minimize theemergenceof drug resistance.
populations far outpaces the rate at which new drugs are A powerful strategy to extend the life of current antimicrobial
developed[1,2].Drugresistancethreatenstheutilityofthelimited agentsisdrugcombinationtherapy[8].Combinationtherapyhas
arsenalofantimicrobialagents.Theeconomiccostsarestaggering thepotentialtominimizetheevolutionofdrugresistancebymore
and exceed $33 billion in the United States alone to cover effectively eradicating pathogen populations and by requiring
treatment of drug-resistant infections in patients, eradication of multiplemutationstoconferdrugresistance[9].Greatsuccesshas
resistantpathogensinagriculture,andcroplossestoresistantpests beenachievedwithcombinationtherapyinthetreatmentofHIV
[3].Theevolutionofresistancetoantifungaldrugsisofparticular [10–12], and it is currently the recommended strategy for
concerngiventheincreasingincidenceoflife-threateninginvasive treatment of tuberculosis and malaria [13,14]. Combination
fungalinfections,andthelimitednumberofantifungaldrugswith therapies have been less well explored in the clinic for fungal
distinct targets [4]. Unlike for antibacterials, fungal-specific drug pathogens. However, targeting cellular regulators of fungal stress
targets are limited, in part due to the close evolutionary responses has emerged as a promising strategy to enhance the
relationships of these eukaryotic pathogens with their human efficacyofantifungaldrugsandtoabrogatedrugresistance[5,15].
hosts,renderingmosttreatmentstoxictothehostorineffectivein Two key cellular regulators that are critical for orchestrating
PLOSGenetics | www.plosgenetics.org 1 April2013 | Volume 9 | Issue 4 | e1003390
EvolutionofResistancetoDrugCombinations
stress response regulators enhances azole sensitivity of diverse
Author Summary
clinical isolates, and compromises azole resistance of isolates that
Fungal infections are a leading cause of mortality evolved resistance in a human host [16,18,23,31]. Inhibition of
worldwide and are difficult to treat due to the limited Hsp90 or calcineurin with molecules that are well tolerated in
number of antifungal drugs, whose effectiveness is humans can impair the evolution of azole resistance [16,20],
compromised by the emergence of drug resistance. A though the potential for evolution of resistance to the drug
powerful strategy to combat drug resistance is combina- combinations remainsunknown.
tiontherapy.InhibitingthemolecularchaperoneHsp90or
Azole resistance mechanisms have been studied most exten-
its downstream effector calcineurin cripples fungal stress
sivelyintheopportunisticfungalpathogenCandidaalbicansandthe
responsesandabrogatesdrugresistance.Hereweprovide
modelyeastSaccharomycescerevisiae.C.albicansistheleadingcauseof
thefirstanalysisofthegeneticandgenomicchangesthat
deathduetofungalinfection[32],andthefourthleadingcauseof
underpin the evolution of resistance to antifungal drug
hospital-acquiredinfectiousdisease[7,32].Itisanaturalmember
combinations in the leading human fungal pathogen,
of themucosal microbiotaof healthy humans, butcan cause life-
Candida albicans, and model yeast, Saccharomyces cerevi-
threatening illness in immunocompromised individuals, such as
siae.Weevolvedexperimentalpopulationswithcombina-
tions of inhibitors of Hsp90 or calcineurin and the most transplantrecipientsandthoseinfectedwithHIV[7,33,34].Drug
widely used antifungal in the clinic, the azoles, which resistancecanreadilyevolveinC.albicansinthelaboratoryandthe
inhibit ergosterol biosynthesis. We harnessed whole- clinic,andmolecularstudieshaverevealedadiversityofresistance
genome sequencing to identify diverse resistance muta- mechanisms [35]. Molecular studies with C. albicans are hindered
tionsamongthe14lineagesthatevolvedresistancetothe by its obligate diploid state, lack of meiotic cycle, unusual codon
drug combination. These included mutations in genes usage, and inability to maintain plasmids [36], thus complemen-
encoding the drug targets, a transcriptional regulator of taryexperimentsareoftenperformedwithitsgeneticallytractable
multidrug transporters, a transcriptional repressor of relative, S. cerevisiae, with which it often shares drug resistance
ergosterol biosynthesis enzymes, and a regulator of phenotypesandunderlyingmolecularmechanisms[37].Forboth
sphingolipid biosynthesis. We also identified extensive species,inhibitionofHsp90orcalcineurinreducesazoleresistance
aneuploidies in several C. albicans lineages. Our study acquired by diverse mutations [16,18,22,38]. With short genera-
reveals multiple mechanisms by which resistance to drug
tion times and relatively small genomes, these organisms provide
combination can evolve, suggesting new strategies to
tractableandcomplementarysystemstoexplorethedynamicsand
combat drug resistance.
mechanisms underpinning the evolution of resistance to drug
combinations.
cellular responses to drug-induced stress are Hsp90 and calci- Here, we provide the first analysis of the genetic and
neurin. The molecular chaperone Hsp90 regulates the stability genomic architecture of the evolution of resistance to drug
andfunctionofdiverseclientproteins[16,17],andcontrolsstress combinationsinfungi.Torecapitulateaclinicalcontextwhere
responses required for drug resistance by stabilizing the protein Hsp90 or calcineurin inhibitors could be used in combination
phosphatase calcineurin [16,18–21]. Compromise of Hsp90 or with azoles to render azole-resistant fungal pathogens respon-
calcineurin function transforms antifungals from fungistatic to sive to treatment, we initiated an evolution experiment with
fungicidalandenhancestheefficacyofantifungalsinmammalian strainsthatareresistanttoazolesinamannerthatdependson
models of systemic and biofilm fungal infections [15,22–24], Hsp90 and calcineurin. We evolved populations of S. cerevisiae
suggestingthatcombinationtherapywithazolesandinhibitorsof and C. albicans that were resistant to azoles due to loss of
Hsp90orcalcineurinmayprovideapowerfulstrategytotreatlife- function of Erg3 with a combination of an azole and an
threatening fungal infections. inhibitor of Hsp90, geldanamycin, or calcineurin, FK506, to
Targeting fungal stress response regulators holds particular identify the mechanisms by which resistance evolves to the
therapeuticpromiseforenhancingtheefficacyoftheazoles,which drug combinations. Of 290 lineages initiated, most went
aretheclassofantifungaldrugthathasbeenusedmostwidelyin extinct, yet 14 evolved resistance. We identified mechanisms
the clinic for decades. Azoles block the production of ergosterol, of resistance in the evolved lineages using a hypothesis-driven
the major sterol of fungal cell membranes, by inhibition of approach based on cross-resistance profiling and a comple-
lanosterol demethylase, Erg11, resulting in a depletion of mentary unbiased approach using whole genome sequencing.
ergosterol and the accumulation of the toxic sterol intermediate, Resistance mutations in the drug target of FK506 or
14-a-methyl-3,6-diol, produced by Erg3 [25]. The azoles are geldanamycin were identified and validated in five lineages.
generallyfungistatic,causinginhibitionofgrowthratherthancell Non-synonymous substitutions conferring resistance were
death, and thus impose strong selection for resistance on the identified in a transcriptional activator of drug efflux pumps,
surviving fungal population [26]; as a consequence, resistance is Pdr1, and in a regulator of sphingolipid biosynthesis, Lcb1.
frequently encountered in the clinic [27]. Azole resistance Resistancealsoarosebyprematurestopcodonsinthecatalytic
mechanismsfallintotwobroadclasses:thosethatblocktheeffect subunit of calcineurin and in a repressor of ergosterol
of the drug on the fungal cell and those that allow the cell to biosynthesis genes, Mot3. Several of the mutations conferred
toleratethedrugbyminimizingitstoxicity[5].Theformerclassof resistance to geldanamycin or FK506, while other mutations
resistancemechanismsincludesupregulationofdrugeffluxpumps transformedazoleresistancefromdependentoncalcineurinto
[28], or mutation of the azole target that prevents azole binding independent of thisstress responseregulator. Genomeanalysis
[29].Thelatterclassincludesloss-of-functionmutationsinERG3, also identified extensive aneuploidy in four of the C. albicans
which encodes a D-5,6-desaturase in the ergosterol biosynthesis lineages. Thus, we illuminate the molecular basis for the
pathway;Erg3loss-of-functionblockstheaccumulationofatoxic transition of azole resistance from calcineurin dependence to
sterol intermediate, conferring azole resistance that is contingent independence, and establish numerous mechanisms by which
on cellular stress responses [16,30]. Azole resistance acquired by resistance to drug combinations can evolve, providing a
lossoffunctionofErg3orbymanyothermutationsisexquisitely foundationforpredictingandpreventingtheevolutionofdrug
dependent on Hsp90 and calcineurin [16]; inhibition of these resistance.
PLOSGenetics | www.plosgenetics.org 2 April2013 | Volume 9 | Issue 4 | e1003390
EvolutionofResistancetoDrugCombinations
Results calcineurin inhibitor). Cross-resistance profiles can be used to
predictcandidateresistancemechanismsbasedonanunderstand-
Experimental evolution of C. albicans and S. cerevisiae ing of how these inhibitors bind to and inhibit their targets
yields resistance to the combination of an azole and an (Figure 3).
inhibitor of Hsp90 or calcineurin Lineages evolved with azole and FK506 were assayed for
Inhibition of Hsp90 or calcineurin has emerged as promising resistancetoazoleandgeldanamycin(ana¨ıveinhibitor)aswellas
strategy to enhance the efficacy of azoles against resistant fungal to azole and cyclosporin A (a structurally dissimilar calcineurin
pathogens, motivating our study to monitor the evolution of inhibitor) [39,40] (Figure 3A). FK506 inhibits calcineurin by
resistancetothedrugcombinationsinazole-resistantpopulations. formingacomplexwiththeimmunophilinFpr1,anditisthedrug-
To do so, we used an experimental evolution approach starting immunophilincomplexthatbindstoandinhibitscalcineurin[41].
with C. albicans and S. cerevisiae strains that harbour erg3 loss-of- The structurally unrelated calcineurin inhibitor cyclosporin A
functionmutationsordeletions,renderingthemresistanttoazoles binds to a distinct immunophilin, Cpr1, to form a complex that
inamannerthatdependsonthestressresponseregulatorsHsp90 binds to calcineurin and inhibits its function [40]. Geldanamycin
andcalcineurin[5].Propagationofthesestrainsinthepresenceof inhibitsHsp90bybindingdirectlytotheunconventionalBergerat
azole and the Hsp90 inhibitor geldanamycin or azole and the nucleotide-binding pocket of Hsp90 [42,43]. The level of
calcineurininhibitorFK506atconcentrationsthatexertselection resistance to these specific drug combinations suggests several
pressure for resistance to the drug combination could lead to the candidate mechanisms of resistance (Figure 3B). For example,
evolution of resistance to geldanamycin or FK506, or the resistancetothecombinationofazoleandFK506butnottoazole
evolution of an azole resistance mechanism that is independent andotherinhibitorstestedsuggestsanFK506-specificmechanism
of Hsp90 or calcineurin among extant lineages (Figure 1A). of resistance such as mutation of FPR1. If resistance was also
Lineages were propagated by serial transfer for between 33 and observed to the combination of azole and cyclosporin A, this
100 generations until robust growth in the presence of the drug would suggest that calcineurin has been altered in a way that
combination was observed in extant lineages (Figure 1B). The prevents the binding of both immunophilin-drug complexes, or
effective population size per lineage was ,4.66106, given that thatacalcineurin-independentmechanismofazoleresistancehas
cultures reachedsaturation (,107cells/ml) between transfers. Of evolved. If resistance was also observed to the combination of an
the 290 lineages initiated, the majority went extinct. Fourteen azole and the na¨ıve inhibitor geldanamycin, this would suggest
lineages evolved resistance to the combination of azole and thatresistanceemergedbyamechanismthatisindependentofthe
inhibitor of Hsp90 or calcineurin (Figure 1C); seven of these stress response regulators Hsp90 and calcineurin; candidate
lineages are C. albicans and seven are S. cerevisiae (Table 1). Six C. mechanisms include those that block the effect of the azoles on
albicans lineages evolved resistance to azole and FK506 (Ca-F theirtarget,suchasup-regulationofthedrugeffluxpumpPdr5in
lineages), and only one evolved resistance to azole and S. cerevisiae [28], or alteration of the azole target Erg11 that
geldanamycin (Ca-G lineage). Four S. cerevisiae lineages evolved prevents azolebinding [29].
resistance to azole and geldanamycin (Sc-G lineages) and three Lineagesevolvedwithazoleandgeldanamycinwereassayedfor
evolved resistancetoazoleand FK506(Sc-F lineages). resistance to azole and radicicol, a structurally unrelated Hsp90
Resistance levels to the drug combinations of all fourteen inhibitor. Like geldanamycin, radicicol binds to the unusual
evolved lineages were evaluated by performing minimum inhib- nucleotide-binding pocket of Hsp90, inhibiting its chaperone
itory concentration(MIC) assaysin thepresenceoftheinhibitors function [42] (Figure 3C). These lineages were also assayed for
with which they were evolved, azole and FK506 (Figure 2A and cross-resistance to azole and FK506, a na¨ıve inhibitor to these
2B)orazoleandgeldanamycin(Figure2C–2E).Becausetheazole strains.Resistancetoazoleandgeldanamycinalonesuggeststhata
resistance phenotypes of the starting strains were abrogated by mutation in HSP90 occurred that prevents the binding of
geldanamycin or FK506, resistance of the evolved lineages was geldanamycin (Figure 3D). This cross-resistance profile is also
monitored with a fixed concentration of azole and a gradient of consistentwithaspecificincreaseingeldanamycinmetabolismor
concentrations of geldanamycin or FK506. Resistance was efflux.Cross-resistancetoazoleandFK506suggeststhatanazole
monitored for a population of cells from each archived lineage, resistance mechanism evolved that is independent of the stress
and for four clones isolated from the evolved population. In all response regulatorsHsp90 andcalcineurin.
cases, the clones reflected the resistant phenotype of the Variationinthepatternsofcross-resistancetothedistinctdrug
population(datanot shown),suggestive ofstrongselectivesweeps combinations was observed among the evolved strains (Figure 4,
as mutations were rapidly fixed in the population. For each Figure 5, Figure 6, Figure 7), implicating a multitude of distinct
population, a clone was archived and further analyses were resistance mechanisms. Even within a cross-resistance category
performed on that strain. The lineages evolved distinct levels of variation was observed in the level of resistance to the drug
resistancetothedrugcombinations(Figure2),indicatingthatthey combinations between strains, indicating that different mutations
acquired different mutations conferringresistance. wereresponsibleforresistance.Thisisconsistentwiththevariation
inlevelsofresistancewiththenativedrugcombinationwithwhich
Cross-resistance assays as a strategy to predict distinct thepopulation wasoriginally evolved (Figure2).
mechanisms of resistance
Mutations in HSP90 confer resistance to azole and
To gain insight into mechanisms of resistance to the drug
combinations, we assessed cross-resistance profiles. Cross-resis- geldanamycin in two S. cerevisiae lineages and one C.
tance assays were performed in the presence of a fixed albicans lineage
concentration of an azole and a gradient of concentrations of TwoS.cerevisiaelineagesevolvedwithazoleandgeldanamycin,
thestructurallydissimilarcounterparttotheHsp90orcalcineurin Sc-G-12 and Sc-G-14, displayed different levels of resistance to
inhibitorwithwhichthepopulationwasevolved(nativeinhibitor), azoleandgeldanamycinrelativetotheancestralstrain(Figure2C
as well as with an azole and an inhibitor of the other stress and 2D). Both Sc-G-12 and Sc-G-14 showed increased cross-
responseregulatornottargetedintheevolutionexperiment(na¨ıve resistance to azole and radicicol, although Sc-G-14 was able to
inhibitor;i.e.Hsp90inhibitorifthepopulationwasevolvedwitha grow with higher concentrations of azole and geldanamycin as
PLOSGenetics | www.plosgenetics.org 3 April2013 | Volume 9 | Issue 4 | e1003390
EvolutionofResistancetoDrugCombinations
Figure1.Designandoutcomeoftheexperimentalevolutionofresistancetodrugcombinations.A)Experimentalpopulationswere
initiatedwithS.cerevisiaeandC.albicansstrainsresistanttoazolesduetoerg3lossoffunction.ThisresistancemechanismiscontingentonHsp90
andcalcineurin,suchthatinhibitionofeitherofthesecellularstressresponseregulatorsresultsincelldeath(t ).Propagationofthesestrainsinthe
0
presenceofazoleandtheHsp90inhibitorgeldanamycinorazoleandthecalcineurininhibitorFK506atconcentrationsthatexertselectionpressure
PLOSGenetics | www.plosgenetics.org 4 April2013 | Volume 9 | Issue 4 | e1003390
EvolutionofResistancetoDrugCombinations
forresistancetothedrugcombinationresultsintheevolutionofresistancetogeldanamycinorFK506(t )ortheevolutionofanazoleresistance
1a
mechanism that is independent of Hsp90 or calcineurin (t ) among extant lineages. B) Single colony founders were used to initiate evolution
1b
experimentsin24-or96-wellplatescontainingcontrolandtreatmentwells.Controlsconsistedof:nodrug,azolealone,geldanamycinalone,or
FK506 alone, where drug concentrations were not inhibitory. Treatment wells consisted of combinations of azole and geldanamycin or FK506,
selectedbasedondoseresponsematrices(seeFigureS2).C)ExperimentalevolutionofresistancetoazoleandgeldanamycinorazoleandFK506
yielded14resistantlineagesoutof290initiated.Ca=Candidaalbicans;Sc=Saccharomycescerevisiae.
doi:10.1371/journal.pgen.1003390.g001
well as azole and radicicol than Sc-G-12 (Figure 4A). Neither non-functional [45]. Surprisingly, deletion of HSC82 in the
lineage showed any cross-resistance to azole and FK506. This parental strain confers a slight increase in resistance toazole and
suggestsdistinctmutationsinHSP90mightconferresistancetothe geldanamycin that phenocopies the resistance of Sc-G-12
drug combinations in these lineages. In S. cerevisiae, Hsp90 is (Figure 4C), suggesting that HSC82K385* is indeed non-functional
encodedbytwogenes,HSC82,whichisexpressedatconstitutively and confers resistance to the combination of azole and
high levels, and HSP82, which is induced by high temperature geldanamycin inSc-G-12.
[44]. Sequencing of HSC82 and HSP82 in Sc-G-14 identified a The C. albicans lineage Ca-G-10 exhibited increased resistance
non-synonymous point mutation that maps to the N-terminal to azole and geldanamycin with no cross-resistance to azole and
domain of HSC82, T1350A. This leads to the amino acid FK506orazoleandradicicol(Figure4D).Itwascross-resistantto
substitution I117N, a residue located in the groove lining the azole and 17-AAG (Figure S1), a derivative of geldanamycin,
nucleotide-binding pocket of Hsp90 to which geldanamycin and suggesting a mode of resistance specific to ansamycin benzoqui-
radicicolbind.Thisresidueishighlyconservedandthoughttobe noneHsp90inhibitors.Sequencingidentifiedaheterozygous,non-
90–100%buried[43].TheimpactofHSC82I117Nonresistanceto synonymousmutationinHSP90,G271T.Thismutationcausesa
azole and geldanamycin was confirmed by performing an allele D91Y amino acid substitution at a residue in the Hsp90
swap,whereHSC82wasdeletedfromtheancestralstrainandthe nucleotide-binding pocket that is thought to be 60–90% buried.
evolved allele was introduced on a plasmid, and reciprocally, This residue is conserved in human Hsp90 although not in S.
HSC82wasdeletedfromtheevolvedstrainandtheancestralallele cerevisiae, where the native amino acid is glutamic acid [43]. To
was introduced. Expression of the HSC82I117N allele in the assesstheimpactofHSP90D91Yonresistancetothecombinationof
ancestral strainconferreda levelofresistancetothecombination azole and geldanamycin we performed an allele swap, replacing
of azole and geldanamycin equivalent to the evolved Sc-G-14 one allele of HSP90 in the ancestral strain with the HSP90D91Y
lineage(Figure4B).Reciprocally,expressionofonlytheancestral allelefromtheevolvedstrain,andreplacingtheHSP90D91Yallele
HSC82alleleintheevolvedstrainabrogatedresistancetothedrug in the evolved strain with the ancestral HSP90 allele. Replacing
combination (Figure 4B). This confirms that HSC82I117N confers HSP90D91YinCa-G-10withtheancestralHSP90alleleabrogated
resistancetothecombinationofazoleandgeldanamycinintheSc- resistance in two independent transformants (Figure 4E). Recip-
G-14 lineage, perhaps by blocking geldanamycin-mediated rocally, replacing a native allele of HSP90 in the ancestral strain
inhibition ofHsp90 function. with the HSP90D91Y allele conferred resistance that phenocopied
SequencingofHSC82andHSP82inSc-G-12identifieda4 bp thatofCa-G-10.ThisindicatesthatHSP90D91Yconfersresistance
insertion in HSC82 that results in a frameshift mutation and a toazoleandgeldanamycinandisresponsibleforresistanceofCa-
premature stop codon in the middle of the coding sequence G-10. Thus,distinctmutations inHsp90canblocktheimpactof
(HSC82K385*). This mutation is expected to render HSC82K385* geldanamycin on azole resistance in both C. albicans and S.
Table1. Evolutionexperiment treatmentsand conditions.
Fluconazole(FL)or
miconazole(M) FK506orgeldanamycin Numberof
Strain Drugcombination concentration (GdA)concentration Numberoftransfers wellsinplate
name Ancestor evolvedin evolvedin(mg/ml) evolvedin(mM) (generations) evolvedin
Sc-F-1 ScLC7 FLandFK506 32 2.5 13(,86) 24
Sc-F-2 ScLC7 FLandFK506 64 0.03 13(,86) 24
Sc-F-3 ScLC7 MandFK506 75 0.06 6(,40) 24
Ca-F-4 CaLC660 FLandFK506 256 20 9(,60) 96
Ca-F-5 CaLC660 FLandFK506 256 20 9(,60) 96
Ca-F-6 CaLC660 FLandFK506 256 1.2 13(,86) 24
Ca-F-7 CaLC660 FLandFK506 4 2 5(,33) 24
Ca-F-8 CaLC660 FLandFK506 4 2 5(,33) 24
Ca-F-9 CaLC660 MandFK506 64 1.2 5(,33) 24
Ca-F-10 CaLC660 FLandGdA 0.1875 0.16 13(,86) 24
Sc-G-11 ScLC10 FLandGdA 256 0.6 13(,86) 24
Sc-G-12 ScLC7 FLandGdA 256 0.6 13(,86) 24
Sc-G-13 ScLC10 FLandGdA 16 2.5 5(,33) 24
Sc-G-14 ScLC10 FLandGdA 16 2.5 5(,33) 24
doi:10.1371/journal.pgen.1003390.t001
PLOSGenetics | www.plosgenetics.org 5 April2013 | Volume 9 | Issue 4 | e1003390
EvolutionofResistancetoDrugCombinations
Figure2.Thepopulationsevolveddistinctresistanceprofiles.LevelsofresistancetoazoleandFK506(A,B)orazoleandgeldanamycin(C–E)
ofevolvedstrainsofS.cerevisiae(A,C,D)andC.albicans(B,E),relativetotheirancestors.Resistancewasmeasuredwithaconstantconcentrationof
azole and a gradient of geldanamycin or FK506 in YPD at 30uC for 2 days (B) or 3 days (A, C–E). Optical densities were averaged for duplicate
measurementsandnormalizedrelativetodrug-freecontrols(seecolourbar).GdA=geldanamycinandFL=fluconazole.
doi:10.1371/journal.pgen.1003390.g002
PLOSGenetics | www.plosgenetics.org 6 April2013 | Volume 9 | Issue 4 | e1003390
EvolutionofResistancetoDrugCombinations
Figure3.Cross-resistanceprofilesprovide a strategyto predictresistance mechanisms. (A)Strainsevolvedin azoleandFK506 were
testedforcross-resistancetoazoleandthecalcineurininhibitorcyclosporinAaswellasazoleandtheHsp90inhibitorgeldanamycin.(B)Candidate
resistance mechanisms based on specific cross-resistance profiles of strains evolved with azole and FK506. (C) Strains evolved in azole and
geldanamycinweretestedforcross-resistancetoazoleandtheHsp90inhibitorradicicolaswellasazoleandthecalcineurininhibitorFK506.(D)
Candidateresistancemechanismsbasedonspecificcross-resistanceprofilesofstrainsevolvedwithazoleandgeldanamycin.GdA=geldanamycin;
RAD=radicicol;andCsA=cyclosporinA.
doi:10.1371/journal.pgen.1003390.g003
cerevisiae, providing a mechanism for resistance to this drug the native FPR1 allele had been deleted, and reciprocally, the
combination. ancestral FPR1 allele was expressed in Sc-F-3 in which the
FPR1V108Fallelehadbeendeleted.ExpressionoftheFPR1V108F
MutationsinFPR1conferresistancetoazoleandFK506in allele in the ancestral strain conferred resistance to azole and
two S. cerevisiae lineages FK506, while replacing FPR1V108F with the ancestral allele in
Sc-F-2 and Sc-F-3 were evolved with azole and FK506, and Sc-F-3 abrogated resistance (Figure 5C). This mutation likely
demonstrate no cross-resistance to azole and cyclosporin A or reducesbutdoesnotcompletelyblockbindingofFK506toFpr1
azoleandgeldanamycin(Figure5Aand5B),suggestingamutation given that complete deletion of FPR1 confers an even greater
in FPR1 may confer resistance to azole and FK506 in these level of resistance to azole and FK506. Consistent with this
lineages.Sequencingidentifiedanon-synonymousmutationinSc- mutation conferring resistance to FK506 rather than altering
F-3 FPR1, G322T. This mutation leads to a V108F amino acid the dependence of the azole resistance phenotype on calci-
substitution that was responsible for the azole and FK506 neurin, deletion of the regulatory subunit of calcineurin
resistance, as determined by an allele swap where the FPR1V108F required for its activation, CNB1, abrogated resistance of Sc-F-
allelewasexpressedfromaplasmidintheancestralstraininwhich 3 (Figure 5E).
PLOSGenetics | www.plosgenetics.org 7 April2013 | Volume 9 | Issue 4 | e1003390
EvolutionofResistancetoDrugCombinations
PLOSGenetics | www.plosgenetics.org 8 April2013 | Volume 9 | Issue 4 | e1003390
EvolutionofResistancetoDrugCombinations
Figure4.MutationsinHSP90conferresistancetoazoleandgeldanamycinintwoS.cerevisiaelineagesandinoneC.albicanslineage.
(A)Sc-G-12(rightpanel)andSc-G-14(leftpanel)arebothresistanttoazoleandgeldanamycinandslightlycross-resistanttoazoleandradicicol,
relativetotheirparentalstrains(above).(B)ResistancetoazoleandgeldanamycininSc-G-14isattributabletoHSC82I117N.Replacingtheancestral
allelewiththeHSC82I117NalleleexpressedonaplasmidincreasesresistanceoftheancestralstraintothelevelobservedinSc-G-14,whilereplacing
theHSC82I117NallelewiththeancestralalleleonaplasmidabrogatesresistanceofSc-G-14.(C)DeletionofHSC82inSc-G-12oritsparentalstrain
phenocopiesresistanceofSc-G-12,suggestingthatHSC82K385*confersresistancebylossoffunctionofHSC82.(D)Ca-G-10hasincreasedresistanceto
azoleandgeldanamycinbutnocross-resistancetoazoleandFK506orazoleandradicicol.(E)ResistancetoazoleandgeldanamycininCa-G-10is
attributable to HSP90D91Y. Replacing the native HSP90 allele in parental strain with HSP90D91Y phenocopied resistance of Ca-G-10. Conversely,
resistanceofCa-G-10wasabrogatedwhenHSP90D91YwasreplacedwiththeancestralHSP90allele.Resistanceassayswereperformedandanalyzedis
inFigure2,afterincubationat30uCfor2days(D)or3days(A–C,E).AssayswereperformedinYPD(A,C–E)orSDwithaminoacidsupplements(B).
GdA=geldanamycin;RAD=radicicol;andFL=fluconazole.
doi:10.1371/journal.pgen.1003390.g004
Sequencing FPR1 in Sc-F-2 revealed a tandem duplication of stop codon near the middle of the coding sequence, MOT3G265*,
nineaminoacidsthatmapstothemiddleofthecodingsequence, suggesting that this might be a loss-of-function allele. Deletion of
FPRdupG53-D61. Expressing FPRdupG53-D61 in the background of the MOT3 in the background of the ancestral strain or in Sc-F-1
ancestralstrainconferredincreasedresistancetoazoleandFK506 phenocopied the level of resistance of Sc-F-1, which is consistent
(Figure5D).ThatresistancewasnotasstrongasinSc-F-2islikely with MOT3G265* being a loss-of-function allele that confers
duetothedifferenceinexpressionlevelsofthenativegeneandthe resistance inSc-F-1(Figure 6C).
plasmidborneallele,whichisdrivenbytheGPD1promoter.Itis S. cerevisiae lineage Sc-G-13 was evolved with azole and
unlikelythatthereareothermutationsaffectingresistanceinSc-F- geldanamycinanddemonstratesonlyasmallincreaseinresistance
2 given that the resistance phenotypes of the ancestral strain and to this combination, with no cross-resistance to either azole and
Sc-F-2 with the plasmid borne FPRdupG53-D61 allele as the sole FK506orazoleandradicicol(Figure6D).Thisresistanceprofileis
sourceofFpr1wereidentical.Furtherconfirmingtheimportance consistent with a mutation in HSC82 or HSP82 that partially
of FPRdupG53-D61 for resistance to azole and FK506, replacing reduces binding of geldanamycin, however, no mutations were
FPRdupG53-D61 of Sc-F-2 with the ancestral FPR1 abrogated identifieduponsequencingHSC82andHSP82.Genomesequenc-
resistance (Figure 5D). As with the FPR1 mutation identified in ingofSc-G-13identifiedfivenon-synonymousmutations, aswell
Sc-F-3, the FPRdupG53-D61 mutation in Sc-F-2 likely reduces but as 130 that were synonymous or in non-coding regions (Table 2
does not block binding of FK506 to Fpr1 as deletion of FPR1 and Table S3); the best candidate for a mutation affecting
confers an even greater level of resistance to azole and FK506 resistance was a C2593G mutation in PDR1, which encodes a
(Figure 5D). Aswith Sc-F-3,deletion oftheregulatory subunit of transcription factor that regulates the expression of numerous
calcineurinrequiredforitsactivation,CNB1,abrogatedresistance multidrugtransporterssuchasPDR5.Gain-of-functionmutations
of Sc-F-2, consistent with this duplication in FPR1 conferring inPDR1areawell-establishedmechanismofazoleresistancethat
resistance to FK506 rather than altering the dependence of the is independent of Hsp90 and calcineurin [16,30,48]. The mild
azole resistancephenotype on calcineurin (Figure5F). resistance phenotype of Sc-G-13 suggested that the PDR1P865R
alleleinSc-G-13confersonlyaslightincreaseindrugeffluxpump
Whole-genome sequencing identifies candidate expression.Cross-resistancetoazoleandFK506wasnotobserved,
resistance mutations in additional S. cerevisiae evolved likely because FK506 inhibits Pdr5-mediated efflux [49]. To
evaluate the importance of the PDR1P865R allele in resistance to
lineages
azoleandgeldanamycinwedeletedPDR1fromtheancestralstrain
Hypothesis driven approaches did not uncover any candidate
and the evolved Sc-G-13 lineage and introduced the ancestral
resistance mutations for several evolved lineages. We therefore
PDR1alleleonaplasmiddrivenbytheGPD1promoter.Replacing
turned to whole genome sequencing to provide an unbiased
the PDR1P865R allele of Sc-G-13 with the ancestral PDR1 allele
approach to identify mutations that accompany the evolution of
reduced resistance of Sc-G-13 (Figure 6E). Resistance remained
resistance to the drug combinations on a genomic scale. For
slightly increased relative to the ancestral strain, likely due to
example,S.cerevisiaeSc-F-1wasevolvedwithazoleandFK506and
higherexpressionofPDR1fromtheGPD1promoterrelativetothe
demonstrated robust resistance to the combination of azole and
native promoter; consistent with this possibility, simply replacing
FK506 as well as azole and cyclosporin A (Figure 6A). This
the ancestral PDR1 allele in the ancestor with the same allele on
resistance profile suggested a possible mechanism of resistance
the plasmid conferred a small increase in resistance (Figure 6E).
involving alteration of calcineurin that prevents the binding of
Sincetherewasnodifferenceinresistancephenotypebetweenthe
bothprotein-drugimmunophilincomplexes,ortheemergenceofa
ancestralandevolvedstrainswhentheplasmidprovidedtheonly
calcineurin-independent azole resistance mechanism. Calcineurin
allele of PDR1, there are likely no other mutations conferring
is encoded by the redundant catalytic subunits CNA1 and CNA2
resistance inSc-G-13.
and the regulatory subunit CNB1 in S. cerevisiae [39,46].
Sequencing of CNA1, CNA2 and CNB1 did not reveal any
Whole-genome sequencing identifies extensive
mutations. Intriguingly, abrogating calcineurin function by dele-
aneuploidy in four of the C. albicans evolved lineages
tionofCNB1didnotreduceresistancetoazoleandFK506inSc-
F-1,indicatingacalcineurin-independentmechanismofresistance and additional candidate resistance mutations in two of
had evolved (Figure 6B). Whole genome sequencing at high the C. albicans lineages
coverage (Table S1) identified two non-synonymous mutations For the six C. albicans lineages evolved with fluconazole and
(Table2),aswellas58mutationsthatweresynonymousorinnon- FK506 (Ca-F-4, Ca-F-5, Ca-F-6, Ca-F-7, Ca-F-8, and Ca-F-9),
coding regions(Table S2andTable S3); thebest candidate fora candidateresistancemutationswerenotidentifiedbyhypotheses-
mutation for affecting resistance was a mutation in MOT3, a based cross-resistance profiles. These lineages shared the same
transcriptionalrepressorofergosterolbiosynthesisgenes[47].The cross-resistance profile of resistance to high concentrations of
non-synonymous substitution in MOT3 resulted in a premature FK506andincreasedresistancetocyclosporinAinthepresenceof
PLOSGenetics | www.plosgenetics.org 9 April2013 | Volume 9 | Issue 4 | e1003390
EvolutionofResistancetoDrugCombinations
PLOSGenetics | www.plosgenetics.org 10 April2013 | Volume 9 | Issue 4 | e1003390