Solution structure and function of YndB, an AHSA1 - Powers Group

13 Pages·2010·1.61 MB·English

proteins
STRUCTUREOFUNCTIONOBIOINFORMATICS
Solution structure and function of YndB, an
AHSA1 protein from Bacillus subtilis
Jaime L. Stark,1 Kelly A. Mercier,1,2 Geoffrey A. Mueller,2 Thomas B. Acton,3 Rong Xiao,3
Gaetano T. Montelione,3,4 and Robert Powers1*
1
DepartmentofChemistry,UniversityofNebraskaLincoln,Lincoln,Nebraska68588-0304
2
LaboratoryofStructuralBiology,NationalInstituteofEnvironmentalHealthSciences,Durham,NorthCarolina27709
3
DepartmentofMolecularBiologyandBiochemistry,CenterforAdvancedBiotechnologyandMedicine,Northeast
StructuralGenomicsConsortium,Rutgers,TheStateUniversityofNewJersey,Piscataway,NewJersey08854
4
RobertWoodJohnsonMedicalSchool,UniversityofMedicineandDentistryofNewJersey,Piscataway,NewJersey09954
INTRODUCTION
ABSTRACT
The Bet v 1 protein from birch is a major allergen with high
ThesolutionstructureoftheBacillussubtilisprotein
sequencesimilaritytotheplantPR-10pathogenesis-relatedproteins,
YndBhasbeensolvedusingNMRtoinvestigatepro-
whichareinvolvedintheresponseofplantstowardmicrobialinfec-
posed biological functions. The YndB structure
1 2
exhibits the helix-grip fold, which consists of a b- tion. AstheBetv1proteinstructurewassolved, numerousother
sheetwithtwosmallandonelonga-helix,forminga proteins from among eukaryotes, archaea, and bacteria have been
hydrophobic cavity that preferentially binds lipid- identified as having the same characteristic fold.3 The Bet v 1-like
likemolecules.Sequenceandstructurecomparisons superfamily of proteins now contains (cid:1)10,135 sequences and con-
with proteins from eukaryotes, prokaryotes, and sists of 13 unique families. The four largest families in the Bet v
archaea suggest that YndB is very similar to the 1-like superfamily are the polyketide cyclases (3475 sequences), the
eukaryote protein Aha1, which binds to the middle ring hydroxylases a-chain (2022 sequences), the activator of Hsp90
domain of Hsp90 and induces ATPase activity. On
ATPase homolog 1-like protein (AHSA1) family (1762 sequences),
thebasisofthesesimilarities,YndBhasbeenclassi-
and the StAR-related lipid transfer (START) family (1026 sequen-
fied as a member of the activator of Hsp90 ATPase
ces). The sequence similarity among the different Bet v 1-like fami-
homolog1-likeprotein(AHSA1)familywithafunc-
lies tends to be relatively low (0–38%), but all contain the same
tionthatappearstoberelatedtostressresponse.An
in silico screen of a compound library of (cid:1)18,500 helix-grip fold that forms a hydrophobiccavity in between the long
lipidswasusedtoidentifyclassesoflipidsthatpref- C-terminal a-helix and the antiparallel b-sheet.3 This hydrophobic
erentiallybindYndB.Theinsilicoscreenidentified, cavity has been shown to preferentially bind to lipids, sterols, poly-
in order of affinity, the chalcone/hydroxychalcone, ketideantibiotics,andotherhydrophobicmolecules.3
flavanone, and flavone/flavonol classes of lipids, Although the Bet v 1-like superfamily members share a similar
which was further verified by 2D 1H-15N HSQC fold, the biological functions vary across the different families. The
NMR titration experiments with trans-chalcone,
ring hydroxylases degrade polycyclic aromatic hydrocarbons into
flavanone, flavone, and flavonol. All of these com- nonaromatic cis-diols,4 the START family appears to be involved
poundsaretypicallyfoundinplantsasprecursorsto in steroidogenesis,5,6 whereas the polyketides cyclase family is
various flavonoid antibiotics and signaling mole-
involved with the biosynthesis of polyketide-based antibiotics and
cules. The sum of the data suggests an involvement
of YndB with the stress response of B. subtilis to
chalcone-like flavonoids released by plants due to a Thecontentof this article issolely the responsibility of the authors anddoes not necessarily
pathogen infection. The observed binding of chal- representtheofficialviewsoftheNationalInstituteofAllergyandInfectiousDiseases.
cone-like molecules by YndB is likely related to Grantsponsor:ProteinStructureInitiativeoftheNationalInstitutesofHealth;Grantnumber:
the symbiotic relationship between B. subtilis and U54GM074958;Grantsponsor:NationalInstituteofAllergyandInfectiousDiseasesNebraska;
Grantnumber:R21AI081154;Grantsponsor:NIH;Grantnumber:RR015468-01;Grantspon-
plants. sor: Battelle (U.S. Department of Energy Office of Biological and Environmental Research);
Grantnumber:KP130103; Grantsponsors: Tobacco Settlement Biomedical ResearchDevelop-
Proteins2010;00:000–000. mentFund,NIH (Intramural ResearchProgram), National Institute of EnvironmentalHealth
VVC 2010Wiley-Liss,Inc. Sciences
JaimeL.StarkandKellyA.Merciercontributedequallytothiswork.
Keywords: NMR structure; YndB Bacillus subtilis; *Correspondenceto:RobertPowers,DepartmentofChemistry,722HamiltonHall,Universityof
Nebraska-Lincoln,Lincoln,NE68588-0304.E-mail:[email protected].
NMR ligand affinity screen; in silico screen; chal- Received12March2010;Revised7July2010;Accepted11July2010
cones;stress response; symbiotic relationship. Publishedonline30July2010inWileyOnlineLibrary(wileyonlinelibrary.com).
DOI:10.1002/prot.22840
VVC 2010WILEY-LISS,INC. PROTEINS 1
J.L.Starketal.
pigments.7 Members of the AHSA1 family are named af- N-terminal b-strands and an a-helix, which also makes
ter the human activator of Hsp90 ATPase protein the proteins larger. The structure of BC4709 and BH1534
(Aha1). Although the proteins of this family have similar do not have these additional structural components further
structures, the functions for most of the AHSA1 family supporting their AHSA1 classification.
members, except for its namesake, are ambiguous and Assigning a function to an uncharacterized protein like
8
are currently classified by UniProtKB as either a general YndBcanbeadauntingtaskthatinvolvesobtainingahigh-
18
stress protein or a conserved putative protein of resolution structure combined with detailed studies that
unknown function. The eukaryotic protein Aha1 is pro- may include generating knockout libraries to analyze cell
posed to interact with the middle domain of heat shock phenotypes, monitoring gene expression levels, or perform-
protein 90, which stimulates its ATPase activity.9,10 The ingpull-downassays,allofwhichrequirein-depthbioinfor-
19–23
domain organization of many homologous eukaryotic maticsanalyses. Asthebiologicalfunctionofaprotein
proteins in the AHSA1 family also suggests a function is,bydefinition,derivedfromitsinteractionswithotherbio-
that is similar to Aha1. Conversely, homologous prokary- molecules or small molecules, identifying interacting part-
otic proteins have a much more diverse domain organi- nersisanalternativeroutetoobtainingafunctionalannota-
3 24,25
zation suggesting a wide range of possible functions. tion. One such technique, FAST-NMR, utilizes a small
Of the 80 total structures solved for 59 members of the biologicallyfocusedcompoundlibrarycombinedwithNMR
Betv1-likesuperfamily,32haveligandsbound.Thetypesof high-throughput screening (HTS), rapid protein-ligand co-
ligands that have been experimentally determined to bind structures using AutoDock26 and chemical shift perturba-
27
Bet v 1-like proteins include membrane lipids, plant hor- tions (CSPs), and a comparison of protein active site
28
mones,secondarymetabolites,polycyclicaromatichydrocar- structures to assist the functional annotation of proteins.
3
bons, and DNA/RNA. There are 12 total proteins in the However, the utility of FAST-NMR relies on structural
AHSA1 family with known structures. The only protein in homologs being found within the diverse functionalchemi-
the AHSA1 family with a solved structure of its protein– callibrary.InthecaseofYndB,theknownBetv1-likesuper-
ligandcomplexistheself-sacrificingresistanceproteinCalC familyligandscombinedwiththeexpectedhydrophobiccav-
fromMicromonosporaechinosporato,11whereCalCisshown ity for YndB already suggests the protein is likely to bind
bound to calicheamicin g1,12 a potent antitumor antibiotic lipid-likemolecules.Thiseliminatestheneedforscreeninga
13 14
compound. Both Pfam and SCOP databases classify diverse arrayofcompounds found in the FAST-NMR com-
CalC as belonging to the AHSA1 family due to its 43–55% pound library and instead requires an extensive screen
sequencesimilarity toother uncharacterizedbacterialmem- againstafocusedlipid-likelibrary.Becauseofthelargenum-
29
bersofAHSA1.However,CalCcontainsabreakintheC-ter- ber of biologically relevant lipid-like compounds and the
minalhelixthatisuncharacteristicofmostBetv1-likepro- correspondinglimitedcommercialavailability,anHTSassay
teinsandwouldlikelyindicateanewCalC-likefamilywithin is not practical or cost effective. Instead, an in silico
the Bet v 1-like superfamily. This leaves only the human screen30,31 provides an attractive alternative method to
Aha1withaproposedfunctionwithintheAHSA1family. identifyspecificclassesofcompoundsthatmayinteractwith
The Bacillus subtilis YndB protein is a protein of YndBandtofocusfollow-upinvitroefforts.
unknown biological function targeted for structuralanalysis To better understand the general biological role of
by the Northeast Structural Genomics Consortium (NESG; AHSA1 proteins, the structure and putative biological
http://www.nesg.org; NESG target: SR211). We previously function of the B. subtilis YndB protein was determined
reported the near complete nuclear magnetic resonance using NMR spectroscopy and the in silico ligand-binding
15
(NMR) assignments for B. subtilis YndB, where the pro- screen. The three-dimensional solution structure of YndB
tein was originally identified as being a member of the (PDB ID: 2kte) is reported herein and is consistent with
START15,16 domain due to the similar helix-grip fold other AHSA1 proteins. As most Bet v 1-like and AHSA1
found in the structure of two homologous proteins and proteins contain a hydrophobic ligand-binding pocket,
based on CATH comparisons.17 The NMR structures the in silico screen of a (cid:1)18,500 lipid compound
29
reported for Bacillus cereus protein BC4709 [Protein Data library was performed to identify a particular class of
Bank (PDB) ID: 1xn6] and Bacillus halodurans protein lipids that preferentially bind YndB and to provide
BH1534 (PDB ID: 1xn5) led totheir START domain classi- insight into its biological function. The B. subtilis YndB
16
fication. These two proteins are 64% and 57% homolo- protein was shown to experimentally bind trans-chalcone,
gous to YndB, respectively, inferring a similar annotation a member of an important class of antibiotics and an im-
for YndB. However, the SCOP and Pfam databases have portant plant metabolite produced by chalcone synthase
32
suggested that YndB, BC4709, and BH1534 belong to the (CHS). Three other compounds similar in structure to
AHSA1 family. Sequence similarity searches with YndB chalcones (flavanone, flavone, and flavonol) and part of
only identify proteins annotated as either AHSA1 or pro- the same metabolic pathway were also shown to bind
teins of an unknown function. The primary difference YndB, albeit weaker binders than trans-chalcone. These
between the START domain and AHSA1 structures is that chalcone-like molecules are often found as precursors to
START domain proteins typically contain two additional flavonoids that play a key role in plant-microbe signaling
2 PROTEINS
SolutionStructureandFunctionofYndB
TableI ard protocols used in the NESG consortium.34 The
StructuralStatisticsandAtomicrmsDifferences recombinant vector containing the gene for YndB was
transformed into BL21 (DE3) pMGK cells. The soluble
Structuralstatisticsforthe18lowestenergyconformations
fraction of the lysed cells was collected and purified with
NOEdistancerestraintsa
a Ni-NTA affinity column (Qiagen) and gel filtration col-
All 1669
Inter-residuesequential(|i2j|51) 466 umn (HiLoad 26/60 Superdex 75 pg, Amersham Bio-
Inter-residueshortrange(1<|i2j|<5) 369 sciences) chromatography, described in full previously.15
Inter-residuelong-range(|i2j|>5) 509 The NMR sample was stored in 20 mM 2-(4-morpho-
Intraresidue 325
lino)ethanesulfonic acid (MES) buffer, pH 6.5 (uncor-
H-bonds 53
Dihedralrestraints 126 rected) with 5% D2O, 0.02% NaN3, 10 mM dithiothreitol
EnsembleRMSDb((cid:1)) (DTT), 5 mM CaCl , and 100 mM NaCl. The sample
2
Secondarystructurebackbone 0.8 was stored in a sealed Shigemi tube (Shigemi Inc, Allison
PSSVeScZo-nsdcaorryessbtructureheavy 1.3 Park, PA) at 48C when not actively collecting NMR data.
Verify3D 22.89 Initial structures of YndB were calculated using CYANA
ProsaII(2ve) 21.12 2.1.35TheprogramassignedthenuclearOverhausereffects
Procheck(phi–psi) 22.83
(NOEs) utilizing a homology model in the first iteration,
Procheck(all) 23.90
which was based on BC4709 and BH1534, the two protein
MolProbityclashscore 21.73
RPFscoresb structures with high-sequence identity fromB.cereus(PDB
Recall 0.647 ID: 1xn6) and B. halodurans (PDB ID: 1xn5), respectively.
Precision 0.668
CYANAassignedatotal of1669distanceconstraints.Addi-
F-measure 0.657
tional data included 126 dihedral restraints from TALOS
DP-score 0.552
Energy(kcal/mol)c and 53 hydrogen bonds based on the secondary structure
Total 25504.8(cid:2)137.9 from TALOS.36 The 20 final structures of CYANA were
Bond 79.8(cid:2)4.7
manually inspected and two were removed due to an un-
Angle 272.2(cid:2)19.7
Dihedral 714.0(cid:2)5.4 usual conformation of the N-terminus. The final 18 struc-
Impropers 108.3(cid:2)8.4 tures were then further refined in explicit solvent using the
vanderWaals 2604.0(cid:2)21.4 RECOORD protocols in Crystallography and NMR System
Violationsc 37,38
(CNS). The final structures agree well with the NMR
NOE>0.5(cid:1) 0
Dihedral>58 2.6 data, which is apparent from the low root-mean square
RMSD(experimental)c deviations (rmsd) from the experimental distances and di-
NOE((cid:1)) 0.0220(cid:2)0.001 hedral angles. To demonstrate that the final NOE assign-
H-bonds((cid:1)) 0.023(cid:2)0.001
Dihedralangles(8) 1.18(cid:2)0.14 ments do not result in unreasonable atomic clashes, the
RMSD(covalentgeometry)c MOLPROBITY39clash score of21.73is extremely low for
Bonds((cid:1)) 0.012(cid:2)0.0004 NMR structures.40 Table I contains other important struc-
Angles(8) 1.3(cid:2)0.04
turalstatisticsrelatedtothequalityofthestructure.Accord-
Impropers(8) 1.6(cid:2)0.08
Ramachandranspaced ing to MOLPROBITY, 85% of the torsion angles of YndB
Mostfavored 85.0% are inthe most favorable region of the RamachandranPlot
Allowed 11.3% with 11.3% in the allowed regions. According to PSVS,40
Disallowed 3.7%
average rmsd for the ordered residues of 18 structures
aCalculatedwithCYANA. refined with explicit solvent was 0.8 A˚ for backbone atoms
bCalculatedwithPSVS. and 1.3A˚ forall heavyatoms. Theensemble was deposited
cCalculatedwithCNS. in the PDB (PDB ID: 2kte) [Fig. 1(A)]. Putative binding
dCalculatedwithMOLPROBITY.
sites of YndB and homologous proteins BC4709 and
41
BH1534 were investigated and compared using CASTp,
and defense, where Bacillus strains have been shown to
which attempts to identify protein ligand binding sites and
have a beneficial impact on plant health by protecting
33 activesitesbydefiningthemolecularsurfaceanddetermin-
against fungal and bacterial pathogens. This suggests B.
ingsurfaceaccessiblepockets. Thethree-dimensionalstruc-
subtilis YndB may respond to a plant infection signal and
tures of the proteins are represented here using the UCSF
induce a stress response.
ChimerapackagefromtheResourceforBiocomputing,Vis-
ualization, and Informatics at the University of California,
MATERIALS AND METHODS 42
SanFrancisco(http://www.cgl.ucsf.edu/chimera).
SolutionstructureofB.subtilisYndB
SequenceandstructuresimilaritytoYndB
Uniformly 13C, 15N-enriched YndB (152 amino acids
with eight non-native residues LEHHHHHH at the C- To identify homologous proteins and elucidate a possi-
terminus for purification) was purified following stand- ble function, multiple similarity comparisons were per-
PROTEINS 3
J.L.Starketal.
Figure 1
TheNMRsolutionstructureofB.subtilisproteinYndB(A)abackbonetraceofthe18lowestenergyconformationmodelsand(B)aribbon
diagramwherethea-helicesarecoloredred,theb-strandsarecoloredyellow,andtheloopsarecoloredgreen.
formed. The pair-wise sequence alignment of YndB to ther divided into a total of 538 distinct subclasses of lipid
protein sequences in a nonredundant database was per- compounds. The two-dimensional structure files pro-
43–45
formed using BLASTP and the default BLOSUM62 vided by the Nature Lipidomics Gateway were converted
46
scoring matrix. DaliLite v.3 was used to perform the into three-dimensional conformers using the program
50
structural similarity comparisons of YndB (model #10) OMEGA 2.3.2 (OpenEye Scientific Software, Santa Fe,
47
with proteins from the RCSB PDB. ClustalW was used NM). OMEGA generates a database of multiple three-
to align the sequences of YndB and the two homologous dimensional conformers for each ligand in the com-
proteins, BC4709 and BH1534, for a detailed analysis of pound library using fragment assembly, ring conforma-
conserved amino acid residues that make up functionally tion enumeration, and torsion driving. In this study,
relevant components of each protein. The ClustalW OMEGAwas used to generate a maximum of 600 unique
sequence alignments used the default settings. (>0.5 A˚ rmsd) conformers for each lipid molecule for a
total searchable database consisting of (cid:1)10,000,000 con-
formers. OMEGA failed to generate conformers for 3306
Virtualscreeningofalipidcompoundlibrary
of the lipid structure files, leaving a chemical library of
The in silico screen of YndB against a lipid library was 18,518 compounds for the in silico screen. Most of these
performed to identify classes of lipid molecules that are failures occurred during the processing of the sphingoli-
favored to bind the protein. The lipid library used in this pid category of lipids (3196 out of 3376 failed) due
study was obtained from the Nature Lipidomics Gateway largely to the large size and number of branches/rotatable
(www.lipidmaps.org), which contains two-dimensional bonds of the molecules in this category.
51
structures of 21,824 lipid molecules (as of January 2010) The docking program FRED 2.2.5 (OpenEye Scientific
found in mammalian species.29,48,49 Clearly, the lipid Software, Santa Fe, NM) was used for the virtual screen of
library is not exhaustive and many lipid molecules found YndB against the lipid library. FRED is a rigid docking
in nonmammalian organisms are not represented, but program, which uses the multiple conformers of each
the goal of the virtual screening effort is to identify a ligand created in OMEGA and generates 100 docked poses
structural homolog to the natural ligand or to identify a within the defined binding site by rotating and translating
particular class of lipid that preferentially binds YndB. the rigid molecule to optimize shape complementarity.
Eight major categories of lipids are represented in the The poses of each ligand conformer are then ranked using
Nature Lipidomics Gateway library: fatty acyls (3476 the built-in consensus scoring method, where only the top
structures), glycerolipids (3012 structures), glycerophos- scoring pose is kept. As the conformers are rigid during
pholipids (1958 structures), sphingolipids (3376 struc- this docking process, FRED has been shown to be very
tures), sterol lipids (2125 structures), prenol lipids (1156 fast as compared with other docking programs that allow
52
structures), saccharolipids (13 structures), and polyketi- for ligand flexibility. This speed is necessary in order to
des (6708 structures). The eight major categories are fur- screen the large lipid-like library in a reasonable amount
4 PROTEINS
SolutionStructureandFunctionofYndB
of time. Although some accuracy may be lost due to rigid DMSO-d (Sigma-Aldrich, St. Louis, MO) before titra-
6
docking and a lack of a biologically relevant conformation tion. The titration analysis was performed with an 80
for the ligand, FRED was primarily used to rapidly filter lM 15N-labeled YndB sample (20 mM MES buffer, pH
out compounds that could not fit into the YndB ligand- 6.5 with 10% D O, 0.02% NaN , 10 mM DTT, 5 mM
2 3
binding pocket. Before initiation of the docking, model 10 CaCl , 100 mM NaCl, and 50 lM 4,4-dimethyl-4-silapen-
2
from the YndB PDB file (PDB ID: 2kte) was prepared tane-1-sulfonic acid (DSS)) and increasing concentrations
using FRED Receptor 2.2.5 (OpenEye Scientific Software, (ranging from 0 to 600 lM) of each ligand. The NMR
53
Santa Fe, NM), where a high-quality shape potential grid data were processed with NMRpipe and the spectra
of 3403 A˚3 was generated that encompassed the proposed viewed using NMRViewJ.54 Kaleidagraph 3.5 (Synergy
binding cavity. Model 10 was selected as the target recep- Software) was used to fit the NMR data to Eq. (2),55,56
tor for the virtual screen as it had the lowest violation qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
energies during the structure calculations. The lipid library ðK þ½L(cid:4)þ½P(cid:4)Þ(cid:3) ðK þ½L(cid:4)þ½P(cid:4)Þ2(cid:3)ð4½L(cid:4)½P(cid:4)Þ
compounds were ranked using the default Chemgauss3 CSP ¼ D D
obs 2½P(cid:4)
scoring function that includes descriptors of shape and
molecular chemical properties. Chemgauss3 incorporates ð2Þ
steric and hydrogen bond interactions, and protein and
where CSP is the 2D 1H-15N HSQC CSPs, [P] is the
ligand desolvation parameters that are smooth using a obs
protein concentrations, [L] is the ligand concentration,
Gaussian function. The relative enrichment for each lipid
and K is the dissociation constant.
class within the top 1000, the top 500, the top 200, the D
top 100, and the top 50 ranked compounds were calcu-
lated according to the following equation: B.subtilisYndB-ligandco-structures
Co-structures of YndB bound to each compound used
%Ab (cid:3)%Ab
%RE ¼ Lib FRED 3 100 ð1Þ in the NMR titration experiment (trans-chalcone, flava-
%Ab
Lib none, flavone, and flavonol) were generated to analyze
the ligand-binding pocket. Although a definitive identifi-
where %RE is percent relative enrichment, %Ab is the cation of the binding pose of these ligands to YndB
Lib
percent abundance of a lipid class in the Nature Lipido- would require extensive NMR experiments and data anal-
mics Gateway library, and %Ab is the percent abun- ysis similar to the original effort to solve the apo-YndB
FRED
dance of a lipid class observed in either the top 1000, structure, molecular docking can provide a rapid and
500, 200, 100, or 50 ranked compounds by FRED. reliable NMR-based model to examine the details of the
binding interactions.
AutoDock 4.0126,57 with the AutoDockTools 1.5.2
NMRtitrationexperiment
(http://mgltools.scripps.edu) graphical interface was used
Based on the results of the virtual screen, three classes
to simulate 100 different binding poses for each YndB-
of lipid molecules were identified as possible binders: fla-
ligand complex. AutoDock was used instead of FRED
vones/flavonols, flavanones, and chalcones/hydroxychal-
due to the accuracy gained from flexible ligand docking
cones. Experimental validation of these possible binders
and because it is one of the most highly cited docking
was performed using CSPs in 2D 1H-15N HSQC NMR 58
programs available. The grid map was generated with
spectra collected on a Bruker 500 MHz Avance spectrom- 0.375 A˚ spacing with xyz grid point dimensions of 50 3
eter equipped with a triple-resonance, Z-axis gradient 58 3 48, which is of sufficient size to encompass the
cryoprobe. The 2D 1H-15N HSQC NMR experiment was proposed binding pocket previously identified in CASTp.
collected at 298 K with 32 scans, 1024 data points and a
The docking calculations were performed using the
sweep width of 15 ppm centered on the water peak at
Lamarckian genetic algorithm default settings with a
4.693 ppm in the direct 1H-dimension and 128 data
population size of 300 and 5,000,000 energy evaluations.
points with a sweep width of 36 ppm in the indirect
15N-dimension. The ligands selected to represent each of
RESULTS
the three potential binding lipid classes were trans-chal-
cone, flavanone, flavone, and flavonol (Sigma-Aldrich, St.
SolutionstructureofB.subtilisYndB
Louis, MO). Flavone and flavonol belong to the same
class of lipids, but there was interest in how the binding The observed secondary structure and fold for B. subti-
would be affected with the addition of a polar functional lis YndB are characteristic of the helix-grip fold found in
group. The fatty acyl oleic acid (Sigma-Aldrich, St. Louis, the Bet v 1-like superfamily. The helix-grip fold consists
MO) was also selected as a negative control. These com- of a b sheet with two small and one long a-helix. The b
pounds were selected based on availability, a simple scaf- sheet is comprised of five strands instead of the normal
fold that clearly represented the lipid class, and cost. six. The missing short strand, which is normally b2,
Each compound was dissolved in ‘‘100%’’ deuterated forms sheet like interactions in only two of the 18 struc-
PROTEINS 5
J.L.Starketal.
Figure 2
AnoverlayoftheNMRsolutionstructureofB.subtilisproteinYndB(yellow),with(A)BacilluscereusproteinBC4709(blue)(PDBID:1xn6)and
(B)BacillushaloduransproteinBH1534(red)(PDBID:13n5).C:ThemultiplesequencealignmentfromClustalWofYndB,BC4709andBH1534
withthe14activesiteresidues(<5A˚ fromboundligand)indicatedinblackrectangles.
tures in the ensemble and appears to protect the edge of ously been identified as being crucial to the function of
b3; unprotected edges can be adventitious interaction the structurally related START domains.60,61 In both of
sites for aggregation.59 However, the strands are anno- these earlier studies, the removal of part of the a3-helix
tated 1, 3, 4, 5, and 6 to facilitate comparisons with other eliminated ligand binding. Based on the NMR solution
family members: residues 12–18 (b1), 63–69 (b3), 73–78 structure for YndB, removing the C-terminal residues
(b4), 83–91 (b5), and 96–104 (b6). The three a-helices would result in a3 no longer being associated with the
are comprised of residues 22–28 (a1), 33–36 (a2), and b-sheet, and therefore, the protein would probably not
120–143 (a3) [Fig. 1(B)]. There is significant variability be folded properly. Hence, we surmise that the previous
in the loop regions of the protein corresponding to resi- results were likely due to protein instability and not the
dues 37–63 and 105–120. These loops appear to be im- activity of specific residues to ligand binding.
portant for the structure of the hydrophobic ligand-bind- The Bet v 1-like superfamily classification for YndB
ing cavity [Fig. 1(B)]. and the reliability of the NMR structure is further sup-
Like other proteins with a helix-grip fold, YndB has an ported by the structural similarities to two homologous
exposed hydrophobic core, likely used in the binding of proteins, BC4709 and BH1534 [Fig. 2(A,B)]. The YndB
lipid-like molecules. Analysis with CASTp shows that the protein exhibits a backbone rmsd of 1.1 A˚ and 1.2 A˚ to
volume of this putative binding cavity is 790 A˚3. The BC4709 and BH1534, respectively, when only secondary
core of YndB consists primarily of aromatic side chains. structural elements are included in the alignment. The
One element of the YndB binding pocket is the long a3- main difference among the structures lies in the loop
helix. This helix is anchored to the b sheet by residues regions, where there appears to be a significant difference
W130, V134, and L138, which show NOE interactions to in the loop conformation of residues 37–63 and 105–120
the sheet residues S15, T17, and L18. Helix a3 has previ- for YndB. This difference affects the size of the hydro-
6 PROTEINS
SolutionStructureandFunctionofYndB
Figure 3
SummaryoftheFREDinsilicoscreeningresultsoftheNatureLipidomicsGatewaylipidlibraryandB.subtilisproteinYndB.(cid:1)18,500lipid
structurescorrespondingto(cid:1)10,000,000conformersweredockedintoYndB.ThecompoundswererankedusingtheFREDChemgauss3scoring
function.A:Aplotoftherelativeenrichment[Eq.(1)]foreachoftheeightmajorlipidcategorieswithinthetop1000hits(red),thetop500hits
(green),thetop200hits(purple),thetop100hits(cyan),andthetop50hits(orange).Onlythepolyketideswerepositivelyenrichedinthevirtual
screenrelativetotheirrepresentationintheoriginallipidlibrary.B:Thechemicalstructuresofthefourflavonoidcompoundschosentorepresent
thethreemostenrichedsubclassesoflipids(chalcones/hydroxychalcones,flavanones,andflavones/flavonols)identifiedfromtheinsilicoscreen.
phobic cavity for YndB, where BC4709 and BH1534 have E-valueof1.0310221orlowerthatincludedBC4709and
much smaller volumes (199 A˚3 and 106 A˚3, respectively) BH1534. All of these proteins belong to the Gram positive
relative to YndB. organismsoftheorderBacillalesandhavesequenceidenti-
As expected from the high-sequence identity, the ties>39%andsequencesimilarities>58%.Thereisaclear
sequence compositions of the ligand binding sites are division in sequence similarity between the 58 Bacillales
also similar [Fig. 2(C)]. Nine of the 14 residues that line proteinsandotherGrampositivebacteriaproteinshomol-
the binding cavity are identical and predominately hydro- ogoustoYndB.Thesequencesimilarityscoredropssignif-
phobic (V29, G34, F76, W78, W83, V85, F87, W130, and icantly from E-valuesof 10221for Bacillalesproteins to E-
L138) and two others show high similarity. Although values of (cid:5)1029 for other Gram positive proteins. This
none of these residues exist within the loop regions, the sequence distinction may indicate a function specific to
large loop from residues 37–63 is very well conserved theBacillalesorganisms.
with 16 residues being identical among the three pro- ThestructuralsimilaritysearchusingDaliLiteidentified
teins, once again indicating the importance of these loops 590 proteins with a Z-score over 2.0. Once again, the two
for ligand binding. The loop regions corresponding to proteins with the greatest structural similarity are BC4709
residues 37–63 and 105–120 are predicted to be confor- and BH1534 with Z-scores of 14.0 and 14.2, respectively.
mationally flexible based on the lack of NMR assign- The top 100 proteins with the highest structural similar-
ments.15 Of the 43 amino acids that comprise the two ities have Z-scores ranging from 9.0 to 14.2 with sequence
loop regions, a total of 21 residues are unassigned. Corre- identities <25%, except for BC4709 and BH1534. All of
spondingly, these loop regions have a limited number of theproteinsidentifiedinthisrangeareeitheruncharacter-
structural constraints resulting in the observed conforma- ized or members of the Bet v 1-like superfamily, which
tional variability in the ensemble of calculated structures includesBetv1-likeproteinsinplants.
[Fig. 1(A)]. Although the lack of NMR assignments and
NOEs suggests conformational flexibility, these observa- Virtualscreeningofalipidcompoundlibrary
tions are not sufficient to define the loops as dynamic
and requires further experimental evidence for verifica- The in silico screen of YndB with the entire Nature
tion.62 Nevertheless, it is anticipated that a bound sub- Lipidomics Gateway lipid library took (cid:1)44 h with the
computation dispersed across 16 nodes of a Linux Beo-
strate would restrict the loop conformation near the
wulf cluster. Of the 18,518 structures in the library,
YndB ligand binding site.
FRED successfully docked 17,475 compounds to YndB.
The relative enrichment [Eq. (2)] of each lipid class from
SequenceandstructuresimilaritytoYndB
the FRED docking is plotted in [Fig. 3(A)]. Only one
The BLASTP search of YndB against a nonredundant lipid category, the polyketides, had a positive relative
protein sequence database identified 58 proteins with an enrichment among the top 1000 docked lipid molecules.
PROTEINS 7
J.L.Starketal.
Figure 4
A:Overlayofthe2D1H-15NHSQCspectraofB.subtilisproteinYndBtitratedwithchalcone,wherethechalconeconcentrationwasincreasedfrom
0lM(blue)to160lM(cyan).ThesignificantCSPsofthenineassignedaminoacidresiduesusedtodeterminethedissociationconstant(K )are
D
highlightedwithablackovalandlabeledaccordingly.Notalloftheperturbedpeakswereassigned;theseresiduesarelikelyfromtheloopregions.
B:NMRtitrationdatafortrans-chalcone(blue),flavanone(green),flavone(purple),andflavonol(orange).ThenormalizedCSPsfortheninemost
perturbedaminoacidresiduesareplottedversustheprotein–ligandconcentrationratios.Thetitrationcurveswerefittoabindingisotherm[Eq.
(2)]usingKaleidagraph3.5(SynergySoftware).Thebest-fitcurvesareshownasasolidline.Thetheoreticalcurvedisplayedfortrans-chalcone
correspondstoaK of1lMandrepresentstheupper-limitfortheK .ThemeasuredK valuesare(cid:6)1lM(trans-chalcone),32(cid:2)3lM
D D D
(flavanone),62(cid:2)9lM(flavone),and86(cid:2)16lM(flavonol).
Polyketides represent 86.8% of the molecules in the top ligand interactions, to measure dissociation constants
1000, whereas they only make up 37.9% of the entire (K ) and to identify ligand-binding sites through the ob-
D
27,63,64
compound library for a relative enrichment of 129%. servation of CSPs. CSPs were calculated by com-
The polyketide representation increases significantly as paring the average 1H and 15N resonance changes
the cutoff for the FRED scoring energy for molecules between ligand-free and ligand-bound YndB 2D 1H-15N
accepted in the top rankings is decreased. If only the top HSQC NMR spectra. The advantage of this approach is
50 docked compounds are considered, 98.0% of these the speed and minimal amount of protein and ligand
compounds are polyketides, with only one hit being a required. Unfortunately, access to the specific lipid com-
member of the fatty acyl category. All of the polyketides pounds predicted to bind YndB is very limited due to
identified belong to the flavonoid class of lipids. Within low commercial availability and/or high cost. Based upon
the flavonoids, three subclasses emerge as favorable hits the in silico screening of YndB with a lipid library, chal-
from the virtual screen: chalcones/hydroxychalcones, fla- cones/hydroxychalcones, flavanones, and flavones/flavo-
vanones, and flavones/flavonols. nols were identified to be the most likely to bind YndB.
The chalcone/hydroxychalcone subclass turns out to be Therefore, representative molecules were sought for each
the most significant hit as 44.9% of the flavonoids in the class containing the basic structural scaffold that would
top 50 hits were chalcones, whereas they only make up likely have characteristic binding properties. A member
9.4% of the library of flavonoids. The remaining flavo- of the fatty acyl category of lipids was also sought for use
noids in the top 50 hits belong to the flavanone (28.6%) as a negative control.
and flavone/flavonol (14.3%) subclasses. The molecules For the chalcone/hydroxychalcone subclass of lipids,
from these three subclasses all have very similar chemical trans-chalcone was selected to represent the basic struc-
structures, which consist of at least two benzene rings tural scaffold for this class. The titration of YndB with
and contain only a few rotatable bonds [Fig. 3(B)]. trans-chalcone resulted in significant CSPs [Fig. 4(A)].
Nine YndB residues with the most significant CSPs
(greater than two standard deviation from the mean)
NMRtitrationexperiment
were identified: Thr80, Trp105, Val111, Ile112, Val122,
Although virtual screening appears to be a useful tool Met126, Trp130, Thr131, and Ile133. These residues line
for identifying particular classes of lipids that have struc- the opening of the proposed binding pocket [Fig. 5(A)]
tural and chemical properties amenable to binding to identified by CASTp. Six more residues with significant
YndB, these results require validation by experimental perturbations (greater than one standard deviation from
methods. NMR is routinely used to evaluate protein– the mean) were also identified: Glu110, Val121, Arg123,
8 PROTEINS
SolutionStructureandFunctionofYndB
Figure 5
A:ArepresentationoftheB.subtilisYndBproteinsurfaceusingtheNMRsolutionstructure,whereaminoacidresiduesthatexhibitedNMRCSPs
causedbythetitrationoftrans-chalcone,flavanone,flavone,andflavonolarecoloredred((cid:5)2standarddeviationsfrommean)andblue((cid:5)1
standarddeviationfromthemean).TheresidueswiththelargestCSPscanbefoundneartheentrancetotheligandbindingcavity,whereasthe
remainingresiduesareassociatedwithhelixa3.Shownwithintheligandbindingcavityarethedockedconformationsofthefourligands
experimentallydeterminedtobindYndB:chalcone(yellow),flavanone(green),flavone(purple),andflavonol(red).B:TheNMRsolutionstructure
ofYndBdockedwithtrans-chalcone(green).Thesidechainsforthe14aminoacidresidueswithin5A˚ oftheligandareshownandlabeled.Five
aromaticsidechainssurroundthetrans-chalconemoleculeandformahydrophobicpocket.
Asp127, Gly128, and Asn135. These amino acids reside chalcone with dissociation constants of 32 (cid:2) 3 lM, 62 (cid:2)
in the long a3-helix and contribute to a portion of the 9 lM, and 86 (cid:2) 16 lM, respectively [Fig. 4(B)]. The
ligand binding pocket. Their perturbation may indicate a range of the dissociation constants mirrors the represen-
structural change in a3-helix upon binding. The remain- tation of each subclass in the virtual screen, where chal-
ing loop residues that define the binding pocket are cones were the most abundantly ranked compounds, fol-
unassigned in apo-YndB. lowed by flavanones and then the flavones/flavonols. Ti-
The normalized CSPs for each of the nine amino acid tration of oleic acid to YndB, which was used to
residues were plotted as a function of protein–ligand represent the fatty acyl category of the lipids, showed no
concentration ratios and fit to a binding isotherm [Eq. significant CSPs and therefore no evidence of binding
(2)] to determine a dissociation constant. trans-Chalcone (data not shown).
binds tightly to YndB with K of (cid:6)1 lM and a stoichi-
D
ometry of 1:1. The binding stoichiometry is based on the
B.subtilisYndB-ligandco-structures
observation that a two-site model does not fit the data as
evidenced by the fact that the CSPs reaches a maximum Using AutoDock, the three-dimensional structures of
at (cid:1)1:1 protein–trans-chalcone concentration ratio [Fig. trans-chalcone, flavanone, flavone, and flavonol were
4(B)]. Calculating an exact K for trans-chalcone was each docked into the YndB binding pocket identified by
D
not possible given the YndB concentration (80 lM) used CASTp and supported by NMR CSPs. The docking of
for the 2D 1H-15N HSQC titration experiments, signifi- each compound did not result in much variation between
cantly lowering the YndB concentration was not feasible. the poses for each of the ligands. Out of 100 docked
Superimposed on the trans-chalcone NMR titration data poses for each ligand, at least 80 were within a 2.0 A˚
[Fig. 4(B)] is a theoretical curve for a K of 1 lM, rmsd of each other. A comparison of the most energeti-
D
implying an upper-limit for the trans-chalcone dissocia- cally favorable poses for each ligand shows that the com-
tion constant. pounds essentially bind with the same orientation [Fig.
Representing the flavanone and flavone/flavonol sub- 5(A)]. The docked structures are also consistent with the
classes, flavanone, flavone, and flavonol all showed CSPs 1:1 binding stoichiometry predicted by the NMR titra-
of the same residues found to be perturbed in the trans- tion experiments and CSPs. Binding two or more com-
chalcone titration, indicating that all the molecules bind pounds in the YndB binding pocket is sterically prohibi-
in a similar manner. However, flavanone, flavone, and tive and the NMR CSPs do not identify a secondary
flavonol bound YndB significantly weaker than trans- ligand binding site.
PROTEINS 9
J.L.Starketal.
human phosphatidylcholine transfer protein (PC-TP)
complexed with dilinoleoylphosphatidylcholine (PDB ID:
66
1ln1), a protein structure in the related START domain
family (Fig. 6). Although the sequence identity between
YndB and PC-TP is low (5%), both proteins have struc-
tural similarities (3.6 A˚ rmsd) with binding pockets
located in the same region of the protein. However, the
binding pocket of YndB is significantly smaller than
the pocket found in PC-TP due to the tighter packing of
the b-sheet with the loop regions and the long a3-helix.
Nevertheless, the overlay of the YndB-chalcone model
with the PC-TP complex indicates that chalcone and the
other flavonoids bind within the large dilinoleoylphos-
phatidylcholine-binding pocket.
DISCUSSION
The NMR structure for B. subtilis protein YndB indi-
catesthattheproteinadoptsahelix-gripfoldandisclearly
Figure 6 amemberoftheBetv1-likesuperfamily.TheYndBstruc-
ture contains an apparent hydrophobic cavity between the
AstructuralalignmentofthehumanPC-TPcomplexedwith
dilinoleoylphosphatidylcholine(PDBID:1ln1)withtheB.subtilis long C-terminal a-helix and the antiparallel b-sheet. Like
YndB-trans-chalconeNMR-basedmodel.Onlytrans-chalconeisshown other members of the Bet v 1-like superfamily, the cavity
fromtheYndB-trans-chalconestructure.Thestructuralalignment
suggests YndB binds to lipids, sterols, polyketide antibiot-
indicatesthelocationofthedockedtrans-chalcone(red)relativetothe
PC-TPproteinstructure(blue)andatransparentmolecularsurface ics,orotherhydrophobicmoleculesaspartofitsbiological
(cyan)representationofthebounddilinoleoylphosphatidylcholine.The function. The YndB protein was originally assigned as a
bindingpocketsofthetwoproteinsareinthesameregion,indicatinga START domain protein based on the high-sequence simi-
reasonabledockingoftrans-chalconetoYndB.
larity to B. cereus BC4709 and B. halodurans BH1534,
whichwereassignedtoSTARTdomainsbasedoncommon
15,16
The AutoDock predicted free energy of binding was structural features. Instead, SCOP and Pfam data-
essentially identical for each compound, averaging 27.2 bases have suggested that YndB belongs to the closely
kcal/mol and correlates with a dissociation constant of (cid:1)5 related AHSA1 subfamily. Also, YndB, BC4709, and
lM. With the exception of trans-chalcone, this is a stronger BH1534 do not have the additional N-terminal b-strands
binding affinity than observed for the three compounds in and the additional a-helix that are characteristics of a
3
the NMR titration experiments. Predicting the actual free START domain structure. Likewise, a BLASTP sequence
energy of binding using AutoDock has an estimated error alignment search indicates YndB is more appropriately
65
of 2.2 kcal/mol. In the YndB-trans-chalcone modeled assignedasamemberofAHSA1.TheBLASTPsearchiden-
structure, there are 14 residues that reside within 5 A˚ of tified 58 proteins from organisms belonging to the Gram
the docked trans-chalcone, where five of these residues are positiveBacillalesorderthatarehomologoustoYndBwith
aromatic [Fig. 5(B)]. These aromatic residues presumably sequence identities >39%. The functions for prokaryotic
have a strong influence on ligand binding and selectivity, AHSA1 family members are typically classified as either a
consistent with the hydrophobic and aromatic nature of general stress protein or a conserved putative protein of
trans-chalcone and the other flavonoids. Conversely, the unknown function. Likewise, the Dali search identified a
binding of trans-chalcone to YndB does not appear to large number of structural homologs to Bet v1-like pro-
involve any hydrogen bonding interactions. teins,AHSA1family membersandanabundanceofhypo-
Most of the difficulty in generating an accurate pro- theticalproteinsorproteinsofunknownfunction.
tein-ligand co-structure for YndB stems from the sus- To furtherexplore the potential functional annotation of
pected flexibility of the two loop regions that define the YndB,theinsilicoscreenagainsta(cid:1)18,500lipid-likechemi-
hydrophobic cavity. Any variation in the orientation of cal library was conducted. The best binders identified from
the loop sidechains directly results in changes in the the in silico screen were from the three general lipid classes
binding site conformation that may be required to of flavones/flavonols, flavanones, and chalcones/hydroxy-
accommodate a ligand. This effect can be seen in some chalcones. Representative compounds from all three classes
of the YndB structures found in the NMR ensemble. The were screened by NMR, where trans-chalcone, flavanone,
YndB-chalcone model and the relative trans-chalcone ori- flavone, and flavonol were all shown to bind in the YndB
entation does correlate well with the binding site for the hydrophobic cavity with K values of (cid:6) 1, 32, 62, and
D
10 PROTEINS