Selective BET bromodomain inhibition as an antifungal therapeutic strategy

15 Pages·2017·3.57 MB·English

by  Flore Mietton

ARTICLE
Received4Oct2016 |Accepted 31 Mar2017 |Published 18May 2017 DOI:10.1038/ncomms15482 OPEN
Selective BET bromodomain inhibition as an
antifungal therapeutic strategy
Flore Mietton1,2,*, Elena Ferri3,*, Morgane Champleboux1,*, Ninon Zala1, Danie`le Maubon4,5, Yingsheng Zhou3,
Mike Harbut6, Didier Spittler2, Ce´cile Garnaud4,5, Marie Courc¸on1, Murielle Chauvel7, Christophe d’Enfert7,
Boris A. Kashemirov3, Mitchell Hull6, Muriel Cornet4,5, Charles E. McKenna3, Je´roˆme Govin1 & Carlo Petosa2
Invasive fungal infections cause significant morbidity and mortality among immunocompro-
mised individuals, posing an urgent need for newantifungal therapeutic strategies. Here we
investigate a chromatin-interacting module, the bromodomain (BD) from the BET family of
proteins,asapotentialantifungaltargetinCandidaalbicans,amajorhumanfungalpathogen.
WeshowthattheBETproteinBdf1isessentialinC.albicansandthatmutationsinactivatingits
two BDs result in a loss of viability in vitro and decreased virulence in mice. We report
small-moleculecompoundsthatinhibitC.albicansBdf1withhighselectivityoverhumanBDs.
CrystalstructuresoftheBdf1BDsrevealbindingmodesfortheseinhibitorsthataresterically
incompatible with the human BET-binding pockets. Furthermore, we report a dibenzothia-
zepinone compound that phenocopies the effects of a Bdf1 BD-inactivating mutation on
C. albicans viability. These findings establish BET inhibition as a promising antifungal
therapeutic strategy and identify Bdf1 as an antifungal drug target that can be selectively
inhibited without antagonizing human BETfunction.
1InstitutdeBiosciencesetBiotechnologiesdeGrenoble,LaboratoireBiologiea`GrandeE´chelle,Universite´deGrenobleAlpes,CEA,Inserm,38000Grenoble,
France.2InstitutdeBiologieStructurale(IBS),Universite´deGrenobleAlpes,CEA,CNRS,38044Grenoble,France.3DepartmentofChemistry,Danaand
DavidDornsifeCollegeofLetters,ArtsandSciences,UniversityofSouthernCalifornia,UniversityParkCampus,LosAngeles,California90089,USA.
4LaboratoiredeParasitologie-Mycologie,InstitutdeBiologieetPathologie,CHUGrenobleAlpes,38043Grenoble,France.5LaboratoireTIMC-IMAG-
TheREx,UMR5525CNRS,Universite´GrenobleAlpes,38058Grenoble,France.6CaliforniaInstituteforBiomedicalResearch,11119NTorreyPinesRd,
LaJolla,California92037,USA.7Unite´BiologieetPathoge´nicite´Fongiques,InstitutPasteur,INRA,75015Paris,France.*Theseauthorscontributedequallyto
thiswork.CorrespondenceandrequestsformaterialsshouldbeaddressedtoC.E.M.(email:[email protected])ortoJ.G.(email:[email protected])
ortoC.P.(email:[email protected]).
NATURECOMMUNICATIONS|8:15482|DOI:10.1038/ncomms15482|www.nature.com/naturecommunications 1
ARTICLE
NATURECOMMUNICATIONS|DOI:10.1038/ncomms15482
I
nvasive fungal infections are a major global health concern, inFig.1a.TheBDsfromC.albicansBdf1(CaBdf1)share31–46%
with B2 million cases and 4800,000 deaths estimated sequence identity with human BET BDs and 58–66% identity
annually worldwide1. Candida species such as C. albicans with those from ScBdf1 and hence are likely to bind multi-
and C. glabrata are among the most significant human fungal acetylatedH3andH4tails(Fig.1b).Toverifythishypothesis,we
pathogens, with invasive candidiasis yielding 30–40% purifiedrecombinantCaBdf1BD1andBD2andassessedbinding
mortality2,3. An increase in drug-resistant fungal strains and to a microarray of human histone tail peptides bearing diverse
the limited repertoire of available drugs has led to an urgent post-translational modifications. While H2A and H2B tail
need for novel therapeutic agents1,4–6. Promising results have sequences differ considerably between human and C. albicans,
emerged from the study of chromatin-interacting proteins as thoseofH3andH4arenearlyperfectlyconserved,justifyinguse
antifungal targets, including histone acetyltransferases and of the array (Fig. 1c). As expected, CaBdf1 BDs showed weak
deacetylases7,8. Histone deacetylase inhibitors have weak binding to mono-acetylated peptides and stronger binding to
antifungal activity when used alone but synergize with antifungal peptides bearing two or more acetylation marks (Fig. 1d,e and
drugs such as azoles and echinocandins8,9. Deletion of either the Supplementary Data 1). Of the two BDs, BD1 exhibited more
histone deacetylase (HST3) or histone acetyltransferase (RTT109) promiscuous binding, recognizing 14 distinct H3 and H4
regulating histone H3 Lys56 acetylation (H3K56ac) in C. albicans acetylation patterns, versus only 7 for BD2. We replaced a con-
increases susceptibility to genotoxic and antifungal agents10. In a served tyrosine residue in the binding pocket of each BD by
studyoftheMediatorcomplexsubunitMed15,whichinteractsvia phenylalanine, a mutationknownto compromise ligand binding
its KIX domain with a transcription factor (Pdr1) implicated in in ScBdf1 (refs 22,23) (Supplementary Fig. 1a,b). These two ‘YF’
pleiotropic drug resistance in C. glabrata, drug-resistant strains mutations(Y248FandY425F)abolishedbindingtoallacetylated
were re-sensitized to antifungals by a small-molecule inhibitor peptides, confirming interaction specificity (Fig. 1d,e). For both
targetingtheKIXdomain–Pdr1interface11.Thesefindingspointto BD1 and BD2, the strongest binding was observed with an H4
animportantroleofchromatin-interactingproteinsinfungaldrug peptide tetra-acetylated on lysines 5, 8, 12 and 16 (hereafter
susceptibility. denoted H4ac4). A pull-down assay confirmed H4ac4 peptide
Here we investigated an epigenetic reader module, the recognition by both BDs, which was abolished by the YF
bromodomain (BD) from the bromo- and extra-terminal domain mutation (Fig. 1f). Interestingly, CaBdf1 BD1 and BD2 bound
(BET) family, as a potential antifungal target in C. albicans. BET tetra-acetylated H4 peptides with comparable strength (Fig. 1e),
proteins are chromatin-associated factors that regulate transcrip- in contrast with mammalian Brd2, Brd4 and Brdt, which bind
tion and chromatin remodelling12. Human BET family members tetra-acetylatedH4moretightlythroughBD1thanthroughBD2
are Brd2, Brd3, Brd4 and Brdt. BET proteins bind chromatin (refs 13,14), highlighting a certain redundancy in the ligand-
through their two BDs (BD1 and BD2), which specifically binding activity of the two CaBdf1 BDs.
recognize histones acetylated on lysine residues. Whereas
canonical BDs bind mono-acetylated histone peptides, BET BDs
possess a wider ligand-bindingpocketallowingthem torecognize Bdf1BDfunctionalityisessentialinC.albicans.Wenextasked
diacetylated peptides13,14. Small-molecule inhibitors, such as JQ1 whether Bdf1 BD function is important for the viability of
and IBET, which selectively target BET BDs have been used to C. albicans. Although a heterozygous BDF1 deletion mutant
validate BET inhibition as a therapeutic strategy against cancer, generated in strain SN152 (derived from SC5314) exhibited no
cardiovascular disease, inflammatory disorders and other medical significant phenotype, we were unable to obtain a homozygous
conditions,withseveralinhibitorscurrentlyinclinicaltrials12,15–21. bdf1D/D mutant, suggesting that BDF1 is essential. To confirm
ThefungalBETproteinBdf1hasbeencharacterizedasaglobal essentialityweplacedtheremainingalleleoftheBDF1geneinthe
transcriptional regulator in Saccharomyces cerevisiae, where it heterozygous strain under the control of a conditional promoter
regulates over 500 genes22. S. cerevisiae Bdf1 (ScBdf1) associates and evaluated survival under repressive conditions. We used
with acetylated histones H3 and H4 (refs 22–24) and with the either a methionine-repressible promoter or a newly engineered
general transcription factor TFIID25, and is a subunit of the tetracycline (Tet)-regulatable cassette compatible with animal
SWR1 chromatin remodelling complex24,26,27. ScBdf1 is also studies. Tet-dependent gene expression in C. albicans is usually
important for chromatin compaction during sporulation28 and achieved by integrating a chimeric transactivator protein and a
for the salt stress response29. In addition to BDF1, S. cerevisiae Tet-responsive promoter independently into the genome33,34.
possesses a second BET gene, BDF2, which is partly functionally Hereweconstructedacassetteallowingintegrationofallrequired
redundant with BDF1 (refs 24,25,30,31). Disruption of BDF1 components in a single step. The cassette contains the
causesseveremorphologicalandgrowthdefects,whiledeletionof transactivator (TetR-VP16), a selective marker (ARG4) and
both BDF1 and BDF2 is lethal22,23. Point mutations that abolish seven tandem Tet-operator elements, which we inserted
ligand binding by ScBdf1 BD1 and BD2 cause growth and upstream of the BDF1 open reading frame (ORF) to generate
sporulationdefects24andaffectthemajorityoftranscriptsaltered strain Dbdf1/pTetO-BDF1 (Fig. 2a). Immunoblotting with a
by disruption of the entire gene22. Many pathogenic fungal polyclonal antibody developed in this study to allow specific
species (including C. albicans, C. glabrata, Aspergillus fumigatus, CaBdf1detection(Supplementary Fig. 1c,d)confirmedthatBdf1
Cryptococcus neoformans and Pneumocystis jirovecii) lack BDF2, was expressed from the pTetO promoter in the absence of
suggesting that inhibition of the sole BET family protein Bdf1 doxycycline (Dox), albeit at a weaker level than from the
might significantly reduce viability and virulence32. endogenousBDF1promoter,andwaseffectivelyrepressedinthe
HerewedemonstratethatBdf1BDfunctionalityisessentialin presenceofDox(Fig.2b).Strikingly,thegrowthofstrainDbdf1/
C.albicansandidentifysmall-moleculeinhibitorsthattargetBdf1 pTetO-BDF1 mirrored these expression levels: compared to wild
BDs without inhibiting human BET proteins, establishing Bdf1 type (WT), growth was reduced in the absence of Dox and
inhibition as a potential antifungal therapeutic strategy. abrogatedinitspresence(Fig.2c).Thephenotypewasrescuedby
re-introducingafunctionalcopyofBDF1(strainBDF1-R/pTetO-
BDF1), confirming that BDF1 is essential in C. albicans.
Results ToverifytheimportanceofBDfunctionforfungalgrowth,we
C. albicans Bdf1 BDs bind multi-acetylated histone tails. generatedstrainsinwhichoneorbothBdf1BDswereinactivated
AphylogenetictreeofhumanandfungalBETproteinsisshown by domaindeletion orbytheYFpointmutationwhiletheother
2 NATURECOMMUNICATIONS|8:15482|DOI:10.1038/ncomms15482|www.nature.com/naturecommunications
ARTICLE
NATURECOMMUNICATIONS|DOI:10.1038/ncomms15482
a b
% identity within BD1
Human BET paralogs human S.cer C.alb C.gla
Bromodomains Extra-terminal domain Brd2 Brd3 Brd4 Brdt Bdf1 Bdf1 Bdf1
BBrdr4d3 11 22 726 1,362 within BD2 Human BBBBrrrrdddd234t 887-212- 887-711- 787-914- 777-443- 33335253 33331121 33335341
Brd2 1 2 801 entity SC..caelbr BBddff11 3494 3482 4404 4416 5-8- 6-6- 8641
Brdt 1 2 947 % id C.gla Bdf1 40 38 37 38 71 57 --
c
Fungal Bdf1 orthologs 686 10 20 30
S. cerevisiae 1 2
H2A
C. glabrata
732
C. albicans 1 2 H2B
A. fumigatus Y248 Y425
C. gattii H3
C. neoformans
P. jirovecii H4
e
d
BD1 BD2
CaBdf1-BD1 CaBdf1-BD2 Acetylation Spot wt YF wt YF 4 9 1418 27 36
1..4..7..10..13..16..19..22.. 1. .4. .7. .10. .13. .16. .19. .22. .
H3ac0 None A1 1 19
A- 4 A9
-
D 9 A16
-- H3ac1 14 A19
G- 18 J11 7 26
wt - 27 K3 16 35
J
- 36 L11 26 45
-
M 4/9 F3
-- H3ac2 9/14 D4
P 14/18 J14
581216
A
- H4ac0 None L12 1 19
-
D 5 L16
-- 8 L17 % intensity
YF G- H4ac1 12 L18 45–100
J- 16 L19 15–45
-- 5/8 M3 5–15
M- H4ac2 8/12 M4 2–5
- 8/16 M5 0–2
P 12/16 M6
5/8/12 M11
Control H4 H4ac3 8/12/16 M12
Peptides: H3 – + ac0 ac4 H4ac4 5/8/12/16 M17
H4 wt
BD1 f
H2A Mut YF Input Bound
H2B wt CaBdf1 Brd4 CaBdf1 Brd4
BD2 wt YF wt wt YF wt
Controls Mut YF
BD1
BD2
H4ac0 – – – + – + – + –
H4ac4 – – – – + – + – +
1 2 3 4 5 6 7 8 9
Figure1|CaBdf1BDsbindmulti-acetylatedhistonetails.(a)PhylogenyanddomainorganizationoffungalBdf1andhumanBETproteins.Theneighbour-
joiningtreebasedonanalignmentofBETproteinsequenceswasgeneratedusingprogramsClustalXandNJPlot.(b)Sequenceidentityamonghumanand
fungalBETBDs.(c)Human(Hs)andC.albicans(Ca)histonetailsequences.(d)BindingofWTandmutant(YF)CaBdf1BDstoahistonepeptide
microarray.Thearraycomprisedcontrolpeptides(greyoutline)orN-terminalpeptidesfromhistonesH3,H4,H2AandH2B(blue,cyan,darkorangeand
lightorangeoutlines,respectively).Signalsforbackgroundandpositive-bindingcontrols(boxedinlightanddarkgreen)orforH4ac0andH4ac4peptides
(boxedinmagentaandpurple)areshownmagnifiedbelowthearray.MicroarraydataaresummarizedinSupplementaryData1.(e)Bindingintensitiesfor
H3andH4peptides.Peptidesareshownasblue(H3)andcyan(H4)rectangles,withacetylationmarksindicatedbyblackboxes.(f)Pull-downassay.
ImmobilizedH4ac0andH4ac4peptideswereincubatedwithGST-taggedCaBdf1BDsorwiththecorrespondingYFmutants.Afterwashing,boundproteins
wereelutedandvisualizedbywesternblottingwithananti-GSTantibody.ThefullblotsareshowninSupplementaryFig.13a.
NATURECOMMUNICATIONS|8:15482|DOI:10.1038/ncomms15482|www.nature.com/naturecommunications 3
ARTICLE
NATURECOMMUNICATIONS|DOI:10.1038/ncomms15482
a b
– Dox ‡ = pTetO-BDF1
ON
WT bdf1Δ / ‡ BDF1-R / ‡
ARG4 TetR - VP16 TetO x7 BDF1 Dox: – + – + – +
Single step insertion Bdf1
+ Dox( ) H4
OFF
ARG4 TetR - VP16 TetO x7 BDF1
c e
– Dox + Dox
pTetO ON pTetO OFF
BD1BD2
WT – Dox (pTetO ON)
+ Dox (pTetO OFF)
bdf1Δ / ‡
OD at 600 nm
BDF1-R / ‡ 0 0.5 1
d WT
BDF1-R / ‡
bdf1Δ / ‡ BDF1-R / ‡
bdf1-bd1Δ / ‡ bdf1Δ / ‡
bdf1-bd1Δ / ‡ *** ***
bdf1-bd2Δ / ‡ bdf1-bd2Δ / ‡
bdf1-bd1Δ-bd2Δ / ‡ bdf1-bd1Δ-bd2Δ / ‡ ***
bdf1-Y248F / ‡
BDF1-R / ‡
bdf1Δ / ‡ bdf1-Y425F / ‡
bdf1-Y248F-Y425F / ‡
bdf1-Y248F / ‡
‡ = pTetO-BDF1
bdf1-Y425F / ‡
bdf1-Y248F-Y425F / ‡
‡=pTetO-BDF1 Y248F Y425F
f g
– Dox (pTetO ON) + Dox (pTetO OFF)
+ Dox (pTetO OFF) ***
***
* **
106 106
–1g) –1g)
U ** U
CF 104 CF 104
d ( d (
a a
o o
gal l 102 gal l 102
n n
u u
F F
1 1
Dox: – + – + – + WT bdf1Δ BDF1 bdf1 bdf1
WT bdf1Δ / ‡ BDF1-R / ‡ / ‡ -R / ‡ -bd1Δ -Y248F
-bd2Δ -Y425F
‡ = pTetO-BDF1
/ ‡ / ‡
‡ = pTetO-BDF1
Figure2|Bdf1BDsareessentialforC.albicansviabilityandvirulence.(a)Tet-OFFconstructusedinthisstudy.DoxinhibitsthebindingofTetR-VP16to
theTetO,preventingBDF1transcription.(b)Bdf1proteinexpressionindifferentstrains.ThefullblotsareshowninSupplementaryFig.13b.(c,d)Colony
formationassaysshowingtheeffectof(c)Bdf1repressionand(d)Bdf1BDinactivationongrowth.Thisexperimentwasrepeatedthreetimeswithsimilar
results.(e)Growthassaysinliquidmedia.Anequalfungalloadwasseededforeachstrainandgrowthmonitoredbyopticaldensityat600nm.Meanand
s.d.valuesareshownfromthreeindependentexperiments.***Pr0.01.Pvaluesweredeterminedinatwo-sidedWelcht-test.(f)Kidneyfungalloadof
miceinjectedwithstrainsshownin(c)measured7dayspostinfection,showingthatBdf1isrequiredforvirulence.Datashownaremeanands.d.values
(n¼6).*Pr0.07;**Pr0.05.Pvaluesweredeterminedusingatwo-sidedWilcoxonranksumtestwithcontinuitycorrection.(g)Kidneyfungalloadof
miceinfectedwithstrainsshownin(d),showingthelossofvirulenceonBDinactivation.Datashownaremeanands.d.values(n¼6).***Pr0.01;Pvalues
weredeterminedusingatwo-sidedWilcoxonranksumtestwithcontinuitycorrection.
4 NATURECOMMUNICATIONS|8:15482|DOI:10.1038/ncomms15482|www.nature.com/naturecommunications
ARTICLE
NATURECOMMUNICATIONS|DOI:10.1038/ncomms15482
WTalleleisexpressedfromtheDox-repressiblepromoter.Strains BETiresistancebyCaBdf1isduetosmallerpocketresidues.To
in which both BDs were inactivated grew as poorly as the identify features differentiating the ligand-binding pockets of
conditional deletion mutant, whereas strains in which only BD1 humanandfungalBETBDs,wedeterminedcrystalstructuresfor
or BD2 was inactivated displayed milder growth defects, with CaBdf1BD1andBD2(SupplementaryTable2).Thesestructures
BD2 inactivation yielding the more pronounced defect (Fig. 2d). exhibit the canonical BD fold (comprising helices Z,A,B and C,
Additionalassaysevaluatingstressresistanceorcellwallintegrity with the ZA and BC loops defining the ligand-binding pocket;
did not reveal any significant phenotype. Growth rates in liquid Fig. 4a) and closely resemble those of their human BET coun-
media recapitulated the phenotypes observed in the colony terparts (mean pairwise root-mean-square deviations (RMSDs)
formation assay (Fig. 2e). Analogousresults were obtainedwhen r1.5Å; Supplementary Table 3). The ligand-binding loops are
BDF1expressionwasrepressedusingthemethionine-regulatable structurally well conserved, including nearly all water molecules
promoter (Supplementary Fig. 2). Thus, C. albicans viability implicatedinligandrecognition(SupplementaryFig.3),although
invitrorequiresthepresenceofatleastonefunctionalBDwithin theoverallbindingsurfacesarelessnegativelychargedinCaBdf1
Bdf1. (Fig. 4b). Superimposing the CaBdf1 BD structures onto that of
JQ1-boundBrd4BD1revealsnomajorstericclashesbetweenJQ1
and the fungal BDs. However, whereas JQ1 fits snugly into the
Bdf1 BDs are required for virulence in a mouse model. Using
Brd4-binding pocket, the fit with the CaBdf1-binding pockets is
our Tet-OFF system for Dox-repressible Bdf1 expression, we
poor.InBrd4thefusedringsystemofJQ1issandwichedbetween
verified the role of Bdf1 in a murine model of invasive candi-
the ‘WPF’ shelf (Trp81, Pro82 and Phe83) and Leu residues 92
diasis.Initially,micewereinjectedwitheitherWTC.albicans,the
and 94, with the thiophene ring positioned within the ‘ZA
conditional BDF1 deletion strain (bdf1D/pTetO-BDF1) or the
channel’ defined by Trp81 and Leu92 (ref. 41). Replacement of
rescue strain (BDF1-R/pTetO-BDF1). Mice injected with the WT
Trp81 by a Val or Phe residue in CaBdf1 BD1 and BD2,
or rescue strains exhibited a high fungal load
respectively, widens the ZA channel, resulting in a poorer fit
(4110,000CFUg(cid:2)1)inthekidney7dayspostinfection(Fig.2f).
(black diamonds in Fig. 4c). Replacement of Brd4 residue Leu94
The fungal load observed with strain bdf1D/pTetO-BDF1 was
reducedby30-fold(B3,500CFUg(cid:2)1)intheabsenceofDoxand by a Val or Ile residue further contributes to the poor fit. In
addition, the p-chlorophenyl group of JQ1 occupies a groove in
was abolished in its presence, mirroring the growth phenotypes
Brd4, which is absent (BD1) or shallower (BD2) in CaBdf1
observed in vitro (Fig. 2c). Subsequently, we evaluated the viru-
(asterisks in Fig. 4c). A similar lack of complementarity between
lence of strains in which both Bdf1 BDs were deleted or inacti-
theligandandbindingpocketexplainstheinsensitivityofCaBdf1
vatedbytheYFpointmutations(Fig.2g).Thesestrainsexhibited
markedly reduced fungal loads (by B8,800- and 1,600-fold, BDs towards PFI-1, IBET-151 and bromosporine.
More generally, recognition of the four BETi compounds by
respectively), consistent with the in vitro growth rates associated
human BET BDs involves 16 residues, including the eight
withthesemutations(Fig.2d,eandSupplementaryFig.2).Hence,
signatureresiduespreviouslyusedtoclassifyhumanBD-binding
Bdf1 BD functionality appears critical for the virulence of
sites42(Fig.5).FiveoftheseresiduesinBD1andfourinBD2are
C. albicans in vivo.
invariant across human BET proteins but diverge in CaBdf1
(magentaarrowsin Fig.5),including theWPF-shelf Trpresidue
CaBdf1 BDs are resistant to human BET inhibitors. For Bdf1 and the Leu residue opposite this shelf, at signature positions
inhibition to be a feasible antifungal therapeutic strategy, small- 1and3,respectively.Remarkably,nearlyalloftheseresidueshave
moleculeinhibitorsneedtodiscriminatefungalfromhumanBET shorter side chains in CaBdf1 than in the human BET proteins,
BDs.Toprobethesimilaritybetweentheligand-bindingpockets implying that selective CaBdf1 inhibition could conceivably be
of C. albicans and human BET BDs, we asked whether human achievedviasmallmoleculesthatclashstericallywiththehuman,
BET inhibitors (BETi) could inhibit CaBdf1 BDs. We used a butnotfungal,sidechains.Interestingly,otherpathogenicfungal
homogeneous time-resolved fluorescence (HTRF) assay to eval- Bdf1 sequences also diverge from human BET proteins at these
uate the ability of four chemically diverse BETi compounds residue positions, suggesting that selective chemical inhibition
(JQ1,PFI-1,IBET-151andbromosporine)toinhibitBDbinding might also be feasible for these targets.
to an H4ac4 peptide (Fig. 3a). The four compounds inhibited
human Brd4 BD1 and BD2 with median inhibitory concentra-
tions (IC values) in the low nM range (3–95nM), consistent CaBdf1BDscanbeselectivelytargetedbysmallmolecules.We
50
with previous reports16,35,36. In contrast, the CaBdf1 BDs were next set out to identify small-molecule inhibitors that would
relativelyinsensitivetotheseinhibitors,withIC valuesbetween targetC.albicans Bdf1BDswithoutinhibitinghumanBETBDs.
50
0.3 and 410mM, or approximately a two- to threefold log- We used our HTRF inhibition assay to screen a library of
reductioninsensitivity(Fig.3b).Theseresultswereconfirmedfor B80,000 chemically diverse compounds, resulting in several
JQ1 by isothermal titration calorimetry (ITC): whereas JQ1 hundred hits for each BD (Fig. 6a). Dose–response curves mea-
boundBrd4 BD1tightly(K of 62nM), nobindingwasdetected suredforCaBdf1andBrd4BDsidentified125and44compounds
d
for CaBdf1 BD1 or BD2 (Fig. 3c and Supplementary Table 1). selective for the fungal BD1 and BD2, respectively. In particular,
None of the BETi compounds significantly inhibited growth of several compounds possessing a dibenzothiazepinone scaffold
C.albicansinvitroat10mMconcentration(Fig.3d).Thisfinding exhibited low-micromolar IC values towards CaBdf1 BD1 and
50
isconsistentwiththepoorIC valuesobservedforJQ1andPFI- no or only weak inhibition of Brd4 BD1 at the highest
50
1 towards CaBdf1 BDs. It also indicates that IBET-151 and concentration(20mM)tested.Onesuchinhibitorwascompound
bromosporine, which display (sub)micromolar IC values, have 1(Fig.6b).ITCmeasurementsshowedthat1bindsCaBdf1BD1
50
poor cellular potency, possibly due to mechanisms known to with a K of 5mM, consistent with the IC value (4.3mM)
d 50
reducedrugpotencyinC.albicans,includingcellwallandplasma determined by HTRF, whereas no or only modest binding and
membrane permeability barriers37,38 and the activity of efflux inhibitionweredetectedforBrd4BD1andCaBdf1BD2(Fig.6c,d
pumpsleadingtorapiddrugextrusion39,40.Insummary,CaBdf1 andSupplementaryTable1).Furthermore,aBROMOscanscreen
and human BET BDs have distinct BETi-binding activity, fromDiscoverXrevealedthatcompound1showednosignificant
supporting the feasibility of targeting CaBdf1 BDs with highly activity against 32 representative human (BET and non-BET)
selective small-molecule inhibitors. BDs (Supplementary Fig. 4); indeed, an ITC experiment
NATURECOMMUNICATIONS|8:15482|DOI:10.1038/ncomms15482|www.nature.com/naturecommunications 5
ARTICLE
NATURECOMMUNICATIONS|DOI:10.1038/ncomms15482
a c
– BETi High FRET JQ1 +
0.2
Acceptor CaBdf1-BD1 ( )
Donor 0
BD Anti-GST –1s) 0.2 CaBdf1-BD2 ( )
al 0
c 0.4
μ hBRD4-BD1 ( )
SA GST P (
Δ 0.2
H4ac4
0
0 25 50 75 100
Time (min)
+ BETi Low FRET Acceptor J) 0
μn ( –2
o
Donor Bd ecti –4
BETi GST at of inj ––68 NKd = = 612.0 ±0 1±6 0 n.0M3
e
SA H –10
0 1 2
Molar ratio (JQ1 / protein)
H4ac4
b
O O NH N N
S N NN N O O N
N N
H
N N N
S N N O O
Cl O O O O O NH O NH O NH S
JQ1 PFI-1 IBET-151 Bromosporine
al 100
n
g
si
RF 50
T
H
%
0
0.1 1 10 100 103 104 0.1 1 10 100 103 104 0.1 1 10 100 103 104 0.1 1 10 100 103 104
Inhibitor (nM) Inhibitor (nM) Inhibitor (nM) Inhibitor (nM)
CaBFD1 BD1 >10,000 >10,000 303 ± 120 >10,000
IC
valu5e0s " BD2 ~10,000 ~10,000 1470 ± 90 313 ± 2
(nM) hBRD4 BD1 18.7 ± 11.2 14.7 ± 3.3 2.9 ± 0.1 6.0 ± 0.7
" BD2 64.9 ± 8.1 94.8 ± 2.8 24.1 ± 1.1 16.5 ± 1.5
d
1.0
DMSO
JQ1
m
00 n 0.5 iBET-151
6 Bromosporine
D
O PFI-1
0
WT BDF1-R / ‡ bdf1-bd1Δ /‡ bdf1-bd2Δ /‡
‡ = pTetO-BDF1
Figure3|CaBdf1BDsareresistanttoBETi.(a)HTRFassay.Abiotinylatedtetra-acetylatedhistoneH4peptideisboundtostreptavidinbeadscoupledto
thedonorfluorophore.AGST-taggedBDisboundbyananti-GSTantibodycoupledtotheacceptorfluorophor.PeptidebindingbytheBDresultsinFRET.
TheadditionofaBETireducesFRET.(b)HTRFassaysperformedonBDsfromCaBdf1andhumanBrd4inthepresenceoftheindicatedBETi.Inhibition
curvesareshownasclosed(BD1)andopen(BD2)circlesingreen(Brd4)andmagenta(Bdf1).IC valuesarelistedbeloweachgraph.Datarepresentthe
50
meanands.d.valuesfromthreeindependentexperiments.(c)RepresentativeITCexperimentsmeasuringthebindingofJQ1toCaBDF1BDs(magenta)and
tohumanBrd4BD1(green).ThevaluesindicatedforK andNrepresentthemeanands.d.fromthreeindependentexperiments.SeealsoSupplementary
d
Table1.(d)BETinhibitorsdonotaffectC.albicansgrowth,evenwhenBdf1BD1orBD2isdeleted.Inhibitorsweretestedat10mMconcentration.
ExperimentswereperformedinthepresenceofdoxycylinetorepressexpressionfromthepTetO-BDF1allele.Datarepresentthemeanands.d.valuesfrom
threeindependentexperiments.
6 NATURECOMMUNICATIONS|8:15482|DOI:10.1038/ncomms15482|www.nature.com/naturecommunications
ARTICLE
NATURECOMMUNICATIONS|DOI:10.1038/ncomms15482
a b
BD1 Brd4 BD2 Brd4 BD1 Brd4 BD2
CaBdf1
αB αB
αZ αZ αA
αA
αC αC
JQ1 JQ1
90° 90° CaBdf1 BD1 CaBdf1 BD2
αA αA
αZ αZ
αB αB
αC αC
c
Brd4 BD1 CaBdf1 BD1 CaBdf1 BD2
L94 V245 I422
L92
Y139
F290 F467
W81 V232 F409
D145 G296 D473
40° 40° 40°
L92
Y139 F290 F467
W81 V232 F409
D145 G296 D473
Figure4|StructuralbasisofBETiresistancebyCaBdf1BD1andBD2.(a)AlignmentofBD1(left)andBD2(right)structuresfromhumanBrd4
(green;PDBcodes3MXFand2OUO)andCaBdf1(magenta).Forclarity,the18and20N-terminalresiduesprecedinghelixaZareomittedfromtheBrd4
BD1andCaBdf1BD1structures,respectively.Brd4-boundJQ1isshownincyan.(b)PlotsofelectrostaticsurfacepotentialofBD1(left)andBD2(right)
structuresfromhumanBrd4andCaBdf1.Regionsofnegativeandpositivepotentialareshowninredandblue,respectively.JQ1boundtoBrd4isshownin
cyan.ThebindingsurfacesofthefungalBDsarelessnegativelychargedcomparedtohumanBETBDs,suggestingdifferencesinthebindingpartnersof
thesedomains.(c)Comparisonoftheligand-bindingpocketsofCaBdf1BD1andBD2withthatofBrd4BD1.JQ1wassuperimposedontheCaBdf1BDsviaa
structuralalignmentwithBrd4BD1.ThediamondandasteriskindicatetheZAchannelandsurfacegroovewhichinBrd4BD1interactwiththethienoand
p-chlorophenylgroupsofJQ1,respectively,andthecorrespondingpositionsintheCaBdf1BDs.Theviewintheupperpanelsisthatof(b)rotated50(cid:2)
counter-clockwise.
confirmedthatthetwohumanBDs(SMARCA2andSMARCA4) awayfromthepocket,adoptingdifferentorientationsinthefour
withthestrongestBROMOscanresponsedidnotdetectablybind complexes present in the crystal’s asymmetric unit
compound 1 (Supplementary Fig. 5a). Compound 1 also exhib- (Supplementary Fig. 7). Superimposing the ligand-bound struc-
ited low cytotoxicity towards HeLa and IMR90 (primary fibro- turewiththatofBrd4BD1resultsinastericoverlapbetweenthe
blast) cells as measured in an MTT colorimetric assay largerBrd4sidechainsatsignaturepositions1and3(Trp81and
(EC Z100mM; Supplementary Fig. 6). Leu94) and the oxadiazole and distal benzene groups of
50
To understand themolecular basis of selectivity we solved the compound 1, explaining why 1 fails to inhibit the human BD
crystalstructureofCaBdf1BD1boundto1.Compound1mimics (Fig. 6f,h). Steric hindrance from the corresponding residues in
the acetyllysine ligand by interacting with the conserved Tyr248 CaBdf1 BD2 (Phe409 and Ile422) also account for this domain’s
and Asn291 residues, forming direct and water-mediated insensitivity to compound 1.
hydrogen bonds via the thiazepinone carbonyl group (Fig. 6e–g). Similarly, we identified an imidazopyridine compound 2
The dibenzothiazepinone ring system engages hydrophobic (Fig.7a),whichinhibitsCaBdf1BD2withhighselectivity:HTRF
residues from the Val-Pro-Phe (VPF) shelf (Pro233), helix B and ITC assays yielded IC and K values of 1.1 and 2.1mM,
50 d
(Phe290) and the ZZ0 loop (Val238, Leu243 and Val245), while respectively, whereas human Brd4 BD2 and CaBdf1 BD1
the cyclopropyl group interacts with the VPF shelf (Pro233 and were only weakly inhibited (IC Z40mM) (Fig. 7b,c and
50
Phe234),helixB(Cys287)andthegatekeeper residueonhelixC Supplementary Table 1). Moreover, no significant inhibition of
(Ile297). In contrast, the oxadiazole and pyrazine groups point any human BDs was observed for compound 2 by BROMOscan
NATURECOMMUNICATIONS|8:15482|DOI:10.1038/ncomms15482|www.nature.com/naturecommunications 7
ARTICLE
NATURECOMMUNICATIONS|DOI:10.1038/ncomms15482
Signature res. 1 2 3 4 567 8 VdW contact
Bromosporine H-bond
PFI-1
HO-mediated
IBET-151 2
H-bond
JQ1, IBET-762
Divergent
αZ αZ’ 30 αB αC
BD1 res
Brd4 75 153
Human Brd3 51 129
BET Brd2 91 169
Brdt 44 124
C. albicans 226 304
S. cerevisiae 165 243
Fungal C. glabrata 152 230
Bdf1 A. fumigatus 296 374
H. capsulatum 294 371
C. neoformans 223 322
P. jirovecii 105 183
BD2
Brd4 365 446
Human Brd3 325 404
BET Brd2 361 442
Brdt 284 365
C. albicans 400 481
S. cerevisiae 329 410
Fungal C. glabrata 324 407
Bdf1 A. fumigatus 489 570
H. capsulatum 493 574
C. neoformans 414 495
P. jirovecii 271 347
Identical between BD1 & BD2:
Human BET
CaBdf1
Figure5|Structure-basedsequencealignmentofhumanandfungalBETproteinsshowingtheligand-bindingregion.Residueshighlightedingreyare
conservedinatleastthreehumanBETsequenceswithineachBD.TheconservedTyrandAsnresiduesinvolvedinwater-mediatedanddirecthydrogen
bondstotheligandareshownininversefont.HumanBETBDresiduesinteractingwithBETicompoundsthroughdirectorwater-mediatedhydrogenbonds
orthroughvanderWaalscontactsareindicatedbyasterisks,arrowheadsandgreencircles,respectively.Interactionsaretakenfromthecrystalstructures
ofBrd4BD1boundtoJQ1,IBET-762orPFI-1(PDBcodes3MXF,3P5Oand4E96),fromthatofBrd2BD1boundtoIBET-151(PDB4ALG),andfroma
structuralalignmentofBrd4BD1withhumanBRPF1BDboundtobromosporine(PDB4C7N).ThesignatureresiduesusedtoclassifyhumanBD-binding
sitesarenumbered1–8(ref.42).MagentaarrowsandboxesindicateBETi-contactingresidueswhichareconservedinhumanBETproteinsbutdifferin
CaBdf1andotherfungalsequences.BluedotsbelowthealignmentindicateresidueswhichareidenticalbetweenBD1andBD2acrosshumanBETproteins
orwithinCaBdf1.
profiling and by an ITC assay on the two most sensitive BDs shows that larger Brd4 side chains at four of these contact
identified by this screen (SMARCA2 and SMARCA4; positions, including signature positions 1,3 and 5 (Trp374,
Supplementary Figs 4 and 5b); while only mild cytotoxicity Leu387 and His437) and residue Tyr432 (Fig. 7e,g), reduces the
was observed towards mammalian cells (EC B60mM; volumeavailableforcompound2,explainingtheinabilityof2to
50
Supplementary Fig. 6). The crystal structure of CaBdf1 BD2 inhibit the human BD. In contrast, the poor affinity for CaBdf1
boundto2revealsthatthecompoundformsahydrogenbondvia BD1
its amino group with the conserved Asn468 residue, as well as a is due to smaller side chains at signature positions 1 and 3
cation–pi interaction involving the phenolic ring and the partial (Val232andVal245),resultinginpoorcomplementaritybetween
positivechargeonAsn468(Fig.7d–f).However,unlikemostBET thebindingpocketandligand.Theaboveresultsestablishproof-
inhibitors, a water-mediated hydrogen bond with Tyr425 is not of-principle that CaBdf1 BDs can be selectively targeted by
observed for 2 (presumably because the ligand dislodges water small-molecule inhibitors in vitro without compromising the
molecule 1, consistent with the dislodgement of this water by a function of human BD-containing proteins.
glycerol molecule in the unbound BD structure; Supplementary
Fig.3d).Instead,thecompound’simidazolylandhydroxylgroups
form water-mediated hydrogen bonds with backbone atoms of ABdf1BD1inhibitorphenocopiesgeneticinactivationofBD1.
the Phe-Pro-Phe (FPF) shelf (Pro410) and helix B (Val460). Wenextassessedtheantifungalactivityofcompoundsidentified
Replacement of the hydroxyl group in 2 by a hydrogen (2a) or in our screen by evaluating their effect on C. albicans growth in
fluorine (2b) atom severely reduces inhibitory activity, under- liquid media. Not surprisingly, none of the compounds sig-
scoring the importance of this group for high-affinity binding nificantly inhibited growth of the WT strain, consistent with the
(SupplementaryFig.8).Themethylpyridinylandphenolicgroups fact that inactivating a single CaBdf1 BD only weakly affects
are sandwiched between the FPF shelf and gatekeeper residue growth (Fig. 2d,e) and that none of the compounds tested can
Val474 on one side, and residues Val415 and Leu420 on the individually inhibit both CaBdf1 BDs. Attempts to simulta-
other, while the tolyl group additionally contacts residues Ile422 neously inhibit both BDs by testing binary combinations of
and Phe467 (Fig. 7d–f). A structural alignment with Brd4 BD2 compounds also did not result in significant growth inhibition.
8 NATURECOMMUNICATIONS|8:15482|DOI:10.1038/ncomms15482|www.nature.com/naturecommunications
ARTICLE
NATURECOMMUNICATIONS|DOI:10.1038/ncomms15482
a b d
0.6
~80,000 compounds
1 + CaBdf1 BD1 ( )
N O 0.3
CaBdf1 BHDT1RF asCsaaBydf1 BD2 N ONN S μ–1cal s) 0.03
Hit compounds: 880 450 N 1 ΔP ( 0 1 + CaBdf1 BD2 ( )
0.3
IC50 determination c IC (μM) 1 + Brd4 BD1 ( )
50 0
CaBfd1BD1 4.3 ± 1.5 0 25 50 75 100
Confirmed hits: 498 227 " BD2 ND Time (min)
150 Brd4 BD1 ~40 0
Counter screen μJ)
Scoemlepcotivuends: Brd142 5BD1 Brd44 4BD2 HTRF signal 10500 of injection ( –––11505 KNd == 50..09 1± ±2 .05. 0μ6M
Additional criteria % eat –20
(IC , logD) 0 H
50 –25
Inhibitors further 0.1 1 10 100 103 104 105 0 1 2 3
characterized: 35 44 Compound 1 (nM) Molar ratio (compound 1 / protein)
e f
Brd4 BD1
CaBdf1 BD1 L94
CaBdf1 L243 V245
BD1 αZ’
Y248
αZαA αB V238 Y139
αC F290
75° 15° ZA loop W81 N291
VPF ZA V232 C287 αB
shelf loop P233
1 F234 I297
BC
αZ’ loop αC
g h
CaBdf1 BD1
L243
αC αB
αA (M256) H236 M256 M283
αZ (H236) O O O H2NO
VPF (V238) –1(M283) 3 –1
ShelfF234 4 3 2 0 N286αZ’ w OV232O4F234I297C2827 1 (N0) OHN286 V232 I297
V232 P233 1 w PN233 N O V2N3H82 Y248 Alignment with Brd4 BD1
O
Y248 N N S O N291 L94
N291
F290 *
1 L243 N L243 V245
BC loop
*
W81
*Steric clash
Figure6|IdentificationofaselectiveCaBdf1BD1inhibitor.(a)ScreeningstrategytoidentifyselectiveCaBdf1BD1andBD2inhibitors.Followingthe
counter-screenstep,mostcompoundsyieldedatleasta20-folddifferenceinIC valuerelativetothehumanBD,althoughafewexhibitedonlyathree-to
50
fourfolddifference.(b)Chemicalstructureof1.(c)HTRFassaysshowingselectiveinhibitionofCaBdf1BD1by1.(d)RepresentativeITCexperiments
showingselectivebindingof1toCaBdf1BD1.ThevaluesindicatedforK andNrepresentthemeanands.d.fromthreeindependentexperiments.Seealso
d
SupplementaryTable1.(e)CrystalstructureofCaBdf1BD1boundto1.Upperinset.Simulated-annealingomitF-F densityfor1contouredat3s.Lower
o c
inset.Conservedwaterstructureandhydrogenbondinginteractionsinthebindingsite.Residuesinteractingwith1throughdirectandwater-mediated
hydrogenbonds(dashedlines)areshowninstickrepresentation.Residuesinteractingthroughbackboneatomsarelabelledinparentheses.Water
moleculesarenumberedasinref.70.(f)AlignmentofCaBdf1BD1(violet)withBrd4BD1(green;PDBcode3UVW).SidechainsareshownforCaBdf1
residuesincontactwith1andforthecorrespondingBrd4residuesifdivergentfromCaBdf1.(g)Schematicsummaryofligandinteractions.Hydrogenbonds
areshownasdashedlines.ResiduesmediatingvanderWaalscontactswith1areindicatedbylabelswithinacartouche.Watermoleculesareindicatedin
red.(h)SurfacerepresentationshowingthebindingpocketofCaBdf1BD1boundtocompound1(top)andthatofBrd4BD1(bottom)superimposedonthe
fungalcomplex.Asterisksindicatestericoverlapbetweencompound1andBrd4BD1residues.
NATURECOMMUNICATIONS|8:15482|DOI:10.1038/ncomms15482|www.nature.com/naturecommunications 9
ARTICLE
NATURECOMMUNICATIONS|DOI:10.1038/ncomms15482
a d
CaBdf1
N
OH BD2 αB
N αA αB αZ αA (M433)
NH αZ (A413)
αC αC
ZA loop (V415) (V460)
FPF
2 Shelf F411 4 3–1 0 N463
2
αZ’
b
IC50 (μM) FshPeFlf αZ’ P410 5
CaBDF1 BD1 >20 2 F409 Y425
" BD2 1.1 ± 0.4 N468
hBRD4 BD2 >20 2
L420
100
BC loop
al
gn 65°
si
F 50
R g e
T
% H CaBdf1 BD2 Brd4-BD2 LI432827
0 CaBdf1-BD2 V415
I422
0.1 1 10 100 103 104 105 L420 L420
Y425
Compound 2 (nM)
Y432
F467
F467
c
N463
0.8
N468
2 + CaBdf1 BD1 ( ) F409 V474 P410
–1s) 1.06 WF430794 F41V1474 N46H8437 αB
μcal 2 + CaBdf1 BD2 ( ) Alignment with Brd4 BD2 αC T472
P ( 0.8
Δ L387
*
0
* f M433
0 25 50 75 100 Y432 A413 V460
Time (min) O
O O O
0 W374 3 H2N
ction (µJ) ––2100 * StericH h4in3d7rance P4O10 54F411 OH2 0 O H N463
e
Heat of inj ––4300 KNd == 21..12 1± ±0 .02. 1μ1M VF447049 NN NH Nδ+HV2415(N) –1 Y425
O
0 1 2 L420 N468
Molar ratio (compound 2 / protein) F467
I422
Figure7|IdentificationofaselectiveCaBdf1BD2inhibitor.(a)Chemicalstructureofcompound2.(b)HTRFassaysshowingselectiveinhibitionof
CaBdf1BD2by2.(c)Bindingof2toCaBdf1BDsmeasuredbyITC.ThevaluesindicatedforK andNrepresentthemeanands.d.fromthreeindependent
d
experiments.SeealsoSupplementaryTable1.(d)CrystalstructureofCaBdf1BD2boundto2.Lowerinset.Simulated-annealingomitF-F densityfor
o c
2contouredat3s.Rightinset.Detailsofthebindingsite.Residuesinteractingwith2throughdirectandwater-mediatedhydrogenbonds(dashedlines)are
showninstickrepresentation.Residuesinteractingthroughbackboneatomsarelabelledinparentheses.Watermoleculesarenumberedasinref.70.
(e)AlignmentofCaBdf1BD2(violet)withBrd4BD2(green;PDBcode4Z93).SidechainsareshownforCaBdf1residuesincontactwith2andforthe
correspondingBrd4residuesifdivergentfromCaBdf1.(f)Schematicsummaryofinteractions.Hydrogenbondsareshownasdashedlines.Watermolecules
areinred.Thecation–pitypeinteractionbetweenthepartiallychargedN468aminogroupandthephenolringisshownasathickdashedline.Residues
mediatingvanderWaalscontactswith2areindicatedbylabelswithinacartouche.(g)SurfacerepresentationshowingthebindingpocketofCaBdf1BD2
boundtocompound2(top)andthatofBrd4BD2(bottom)superimposedonthefungalcomplex.Asterisksindicateclosecontactspredictedtosterically
inhibittherecognitionof2byBrd4BD2.Doublearrowsindicateshortdistanceswhichmaycontributeadditionalsterichindrance.
We therefore tested individual compounds on mutant strains in against these strains. We speculate that the compounds failed to
which either BD1 or BD2 was inactivated. Most inhibitors, enterthefungalcellbecauseofcellularpermeabilitybarriers37,38,
including compounds 1 and 2, showed little antifungal activity or were extruded by efflux pumps39,40 or metabolized before a
10 NATURECOMMUNICATIONS|8:15482|DOI:10.1038/ncomms15482|www.nature.com/naturecommunications