Sf

BNa

Kb

a

ARRAA

KPESTCF

1

ftt[gcrobt

tv

h1

Journal of Molecular Graphics and Modelling 66 (2016) 133–142

Contents lists available at ScienceDirect

Journal of Molecular Graphics and Modelling

journa l h om epage: www.elsev ier .com/ locate /JMGM

tructure-based designing of sordarin derivative as potentialungicide with pan-fungal activity

iprashekhar Chakrabortya, Nikunjkumar Vinodray Sejpalb, Pavan V. Payghana,anda Ghoshala, Jayati Senguptaa,∗

Structural Biology & Bio-Informatics Division, Indian Institute of Chemical Biology (Council of Scientific & Industrial Research), 4, Raja S.C. Mullick Road,olkata 700 032, IndiaThe National Institute of Pharmaceutical Education and Research (NIPER), Indian Institute of Chemical Biology, Kolkata, India

r t i c l e i n f o

rticle history:eceived 26 August 2015eceived in revised form 23 March 2016ccepted 24 March 2016vailable online 30 March 2016

eywords:rotein synthesislongation factor-2ordarin-binding pocketranslation inhibition

a b s t r a c t

Fungal infections have become a significant problem for immunosuppressed patients. Sordarin, a promis-ing fungicidal agent, inhibits fungal protein synthesis by impairing elongation factor-2 (eEF2) function.Intriguingly, despite high sequence similarity among eEF2s from different species, sordarin has beenshown to inhibit translation specifically in certain fungi while unable to do so in some other fungalspecies (e.g. Candida parapsilosis and Candida lusitaniae).

The sordarin binding site on eEF2 as well as its mechanism of action is known. In a previous study, wehave detailed the interactions between sordarin and eEF2 cavities from different fungal species at themolecular level and predicted the probable cause of sordarin sensitivity.

Guided by our previous analysis, we aimed for computer-aided designing of sordarin derivatives aspotential fungicidal agents that still remain ineffective against human eEF2. We have performed struc-

omputational drug designungicide

tural knowledge-based designing of several sordarin derivatives and evaluated predicted interactionsof those derivatives with the sordarin-binding cavities of different eEF2s, against which sordarin showsno inhibitory action. Our analyses identify an amino-pyrrole derivative as a good template for furtherdesigning of promising broad-spectrum antifungal agents. The drug likeness and ADMET prediction onthis derivative also supports its suitability as a drug candidate.

© 2016 Elsevier Inc. All rights reserved.

. Introduction

Immunosuppressed patients are more prone to get systemicungal disease. Immunosuppression occurs due to cancer, AIDS,ransplantation, broad spectrum antibiotics and glucocorticoidherapy, in premature infants, peritoneal dialysis or haemodialysis1,2]. These fungal diseases are mainly treated by the azoles, poly-enes and echinocandins, but therapeutically satisfactory resultsannot always be achieved by these antifungal drugs due to theesistance in some fungal species, their toxicity, and limited waysf administration. These drugs act on the plasma membrane of celly binding or blocking the synthesis of ergosterol which will causehe inhibition in the fungal growth [3].

Sordarin, a known antifungal agent, may be a molecule with bet-er prospects compared to others because it targets one of the mostital cellular processes in fungi, protein synthesis, by impairing the

∗ Corresponding author.E-mail address: jayati@iicb.res.in (J. Sengupta).

ttp://dx.doi.org/10.1016/j.jmgm.2016.03.013093-3263/© 2016 Elsevier Inc. All rights reserved.

function of an essential translation factor, the eukaryotic elongationfactor 2 (eEF2). Eukaryotic elongation factor 2 (eEF2) is a member ofthe GTPase superfamily of proteins which is known to assist in thetRNA translocation step of protein synthesis. Sordarin is a knownantifungal agent which targets eEF2 so as to inhibit protein transla-tion either by blocking eEF2-mediated translocation of tRNAs fromA and P sites to P and E sites of the ribosome [4], or by interferingwith eEF2-dependent ribosome splitting [5]. Interestingly, sordarindoes not hinder GTP hydrolysis. Upon GTP hydrolysis, structuralreorientation of the C-terminal half (domains III, IV and V) relativeto the N-terminal half (domains G’, I and II) occurs within eEF2 pro-moting its release from the ribosome. Sordarin inhibits the domainrearrangement required for eEF2 release by binding to a pocketwithin domains III, IV and V [6,7] (Supplementary Fig. S1A). It isapparent that sordarin’s activity is directly related to its ability tobind inside the eEF2 cavity.

Sordarin has a remarkable selectivity of action which makesit ineffective against human eEF2. It is, however, also not equallyeffective against all fungal species. Certain fungal species are highlyresistant towards the drug (Supplementary Fig. S1B). This remark-

1 ar Gra

as‘eSc

sdadiddgtfsTtsomtdaiiwwr

ssaewcdopi5taiAf

2

2

g((itA

wcaa

34 B. Chakraborty et al. / Journal of Molecul

ble selectivity of the drug has been attributed to some speciespecific alterations within a stretch of amino acids (termed as thesordarin specificity region (SSR)’, residues 517–524 in yeast) at thentrance of the binding cavity of eEF2 [8]. In human, almost all theSR residues are found to be different with respect to their yeastounterparts (Supplementary Fig. S1B).

Our aim was to decipher the structural basis underlying thepecies selectivity of sordarin while targeting translation and toesign some modified versions of the drug with broader spectrumntifungal activity. In order to evaluate the compatibility of therug and the protein cavity, we have introduced several ‘determin-

stic parameters’ in a previous study [9] which could successfullyescribe the molecular basis of selectivity of sordarin binding toifferent fungal eEF2s. Our analysis revealed that although the fun-al eEF2s are globally highly conserved at the sequence level (morehan 80% conserved across fungal species) as well as in structuralolding, the local environment of the binding cavity interiors differsignificantly in some eEF2s altering drug-eEF2 interaction patterns.he effects collectively act as a key factor in determining the selec-ivity of the drug binding. The study inferred that subtle local leveltructural changes due to amino acid substitutions in the SSR regionf eEF2s from certain species (e.g. C. parapsilosis and C. lusitaniae)ake the eEF2 cavity incompatible both physically and chemically

oward sordarin and consequently render them insensitive to therug. We predicted that favourable interactions between sordarinnd an aromatic residue at position 521 of the SSR stabilize the drugn the cavity and that lack of this interaction results in sordarinnsensitivity. Further, we identified some nonconserved residues

ithin the human eEF2 cavity apart from those within the SSR,hich were also found to contribute substantially towards sordarin

esistance in human.Guided by the knowledge gained from this earlier study,

tructure-based design of sordarin derivatives with broad-pectrum antimycotic activity is the aim of the present study. Ournalyses showed that the C61 position of sordarin is the clos-st neighbour of residue 521 within the SSR region of eEF2 ande have chosen this position as the site for modification. We

omputationally modified the sordarin structure by introducingifferent functional groups at C61 position and assessed the effectn predicted drug-protein interactions using the ‘deterministicarameters’. Functional groups that might augment the possibil-

ty of sordarin’s favourable interaction with any residue at position21 irrespective of species were inserted at C61 position. Based onhe assessment we have proposed one derivative that might acts a potent fungicide with ‘pan-fungal’ inhibitory action. Interest-ngly, preliminary predictions of drug-likeness and computationalDMET predictions suggest the derivative could be a good template

or a drug candidate.

. Methods

.1. Modification of protein and ligand structures

The eEF2 models for all the fungal species used in this study wereenerated using a crystal structure of sordarin-bound yeast eEF2RCSB PDB code 1N0U) as the template for homology modellingdescribed in details in Ref. [9]). Substitution of conserved residuesn yeast eEF2 coordinate at different positions and modifications onhe sordarin structure with different substituent were done usingccelrys Discovery Studio 4.0 software.

Models of seven sordarin derivatives (Supplementary Fig. S2)

ere prepared by attaching different groups, e.g. long aliphatic

hain, aromatic/heterocyclic ring structure (good for stacking inter-ction), and others being part of known antifungal/antimicrobialgents (Fig. 2B).

phics and Modelling 66 (2016) 133–142

2.2. Approach for assessing the effects of the derivatives

For every derivative prepared, we first analyzed its predictedbinding with the S.cerevisiae eEF2 model with Tyr521Ser andSer523Asn substitutions. This was the first step to check whetherthe derivative is engaged in favourable contacts with Ser521 andAsn523 (residues present in eEF2s of sordarin-resistant fungalspecies C. parapsilosis and C. lusitaniae). Next and perhaps the mostimportant step was to examine the insensitivity of the drug towardshuman eEF2. Once a derivative successfully passed the two tests wemoved on to analyze its binding likeliness for eEF2 from differentsordarin-resistant fungal species.

2.3. Docking

The flexible docking procedure [10] within the “Affinity” mod-ule of Insight II was used to get an accurate description of ligand(sordarin and its derivatives) recognition by each of the eEF2structures (without imposing any initial bias from the crystal struc-ture). ‘Affinity’ is a suite of programs for automatically dockinga ligand to a receptor by a combination of Monte Carlo mini-mization followed by Simulated Annealing procedure (Affinity userguide. San Diego, USA: Accelrys Inc.; 1999) as described previously[9].

The resulting set of complexes following docking had the ligandbound to the cavity in various poses. To select the final structurefrom all the top-ranked poses for each case we have used the rootmean square deviation (rmsd) with respect to the sordarin pose inthe x-ray structure 1N0U as done earlier [9]. Additionally, we alsointroduced %ASA (percent accessible surface area) as another selec-tion criteria. The pose selection procedure is detailed in Section2.4.

2.4. Selection of the final ligand poses

As described previously, root-mean-square deviation (RMSD)of the different poses of sordarin and all its derivatives were calcu-lated by comparing with that of sordarin in crystal structure 1N0U.For the derivatives (since they are much bulkier), the extra groupwas excluded and sordarin backbone of the various poses was con-sidered while calculating the RMSD. Calculation of the rmsd wascarried out in the Discovery Studio version 4.0 (RMSD value of ≤1was considered to be acceptable).

For all those poses that met the criterion that the RMSD to thecrystal structure should be <1 Å, another selection criterion wasimposed. Sordarin-bound yeast eEF2 crystal structure is available(1N0U). We have compared the percentage of solvent exposed sur-face area of the molecules with that of yeast eEF2-bound sordarinin 1N0U. ASA for the ligands was obtained from ‘StrucTools’ server(http://helixweb.nih.gov/structbio/basic.html) and %ASA was cal-culated by using the following formula:

%ASA = (Total surface area of the ligand/

Accessible surface area of protein-bound ligand) × 100

The pose having the least %ASA value (preferably not exceedingthe %ASA value for sordarin in 1N0U) was considered reasonableover others. The ligand pose fulfilling both the criteria was selectedfor further analyses.

2.5. Analysis of ligand–protein contacts

As described in our previous study [9], the analysis ofligand–protein contacts was done using LPC/CSU server (http://ligin.weizmann.ac.il/cgibin/lpccsu/LpcCsu.cgi) [11,12]. This server

ar Gra

plcita(ltp

tvgi

2

sbesmsstsctcf

2

o‘ucdif

pta[chAimhfce

2

becea

B. Chakraborty et al. / Journal of Molecul

rovides the value of normalized complementarity (NC) forigand–protein assembly based upon the intermolecular atomicontact between them. This complementarity value provides andea regarding the possibility of the ligand binding to its recep-or. Normalized complementarity (NC) depends upon the surfacerea of the ligand engaged in contacts contributing in its binding“legitimate” atomic contacts between) and the surface area of theigand not contributing in its binding (“illegitimate” atomic con-acts). Again, “legitimacy” depends on the hydrophobic-hydrophilicroperties of the contacting atoms.

Evaluation of the binding site interaction of the protein withhe ligand in 2D was carried out by using the Discovery Studioersion 4.0. This tool provides information regarding the hydro-en bond, electrostatic interaction, van der Waals interactions, pi-pinteraction and pi-sigma interaction in a 2-D diagram.

.6. Estimation of cross-correlation values

The ‘positive casts’ of the cavities of the eEF2 models of all thepecies under consideration and simulated maps for all the cavity-ound ligands were generated using SPIDER [13] as describedarlier [9]. The ‘positive cast’ of the sordarin-binding cavity repre-ents the 3D shape and size of the cavity interior. Similarly, densityap of each ligand molecule was generated from the fitted atomic

tructure using SPIDER. The cross-correlation coefficient [14] (Pear-on’s correlation coefficient) between the simulated ligands andhe experimental (‘positive cast’) maps was used as a measure ofimilarity between them. All cross-correlation coefficients wereomputed considering only voxels inside the molecular envelope ofhe simulated map (‘local correlation’) [15]. The cross-correlationoefficient (cc c) between the ‘positive cast’ and the sordarin densityor each case was also calculated using SPIDER.

.7. Deterministic parameters to estimate drug compatibility

The compatibility of the drug within its cavity was assessedn the basis of certain parameters (previously termed by us as

deterministic parameters’ [9]). The deterministic parameters’ val-es obtained for sordarin binding in yeast eEF2 cavity [9] wereonsidered as minimum requirements for drug binding and in turnrug sensitivity of a species. Human eEF2, on the other hand, is

ncluded as negative control (a good drug molecule should be inef-ective towards human).

To assess the possibility of drug binding we have analysed thehysical compatibility as well as chemical compatibility betweenhe drug and the cavity. The ASA (In the ‘StrucTools’ server ([16])ccessible surface is calculated using the ‘calc-surface’ program17]) and the cross-correlation value between the drug and theavity interior assess the physical compatibility. The ASA estimatesow well the drug gets buried within the binding pocket. HigherSA value indicates that the ligand is not well-accommodated

nside the cavity and the cross correlation value estimates the shapeatching between the cavity interior with the drug. On the other

and, the complementarity parameter (along with the details ofavourable and unfavourable contacts) offers an assessment for thehemical compatibility. The parameters were collectively consid-red as a function.

.8. Assessment of drug sensitivity

To put it in a quantitative manner, a 10% rule as threshold haseen followed for each of the deterministic parameters. It has been

xperimentally shown that among various fungal species, while S.erevisiae eEF2 is sensitive to sordarin, C. parapsilosis C. lusitaniaeEF2s along with Human eEF2 are insensitive to it. As mentionedbove, the values of those parameters obtained for eEF2-sordarin

phics and Modelling 66 (2016) 133–142 135

binding for S. cerevisiae have been considered to be necessary valuessuggesting proper binding (‘standard values’, Table 4). Comparingthe values obtained for insensitive species with the standard val-ues, 10% rule has been introduced to identify sensitivity of the eEF2stowards a ligand. Thus, a complementarity or a cross correlationcoefficient value below 10% (complementary and cross-correlationcoefficient values <0.495 and 0.612 respectively) and an accessiblesurface area (ASA) value above 10% (>180.4) of the ‘standard val-ues’ are considered as ‘out of range’. We have considered a speciesto be ‘sensitive’ only when the values of at least two out of thethree parameters obtained in a particular case are within the 10%range. On the other hand, if any two of the values are ‘out of range’,the species is considered ‘insensitive’ towards that particular drugmolecule.

2.9. Estimation of electrostatic environment inside the cavity andof drug

Estimation of electrostatic potential are carried out by usingAdaptive Poisson Boltzmann Solver (APBS) software availablewithin VMD [18]. In these evaluations, we have analyzed electro-static environment for drug-bound sordarin cavity of eEF2.

2.10. Characterizing drug-likeness and ADMET parameters

Early stage preclinical evaluation can save both time andmoney with increased positive outcome, needed for success-ful drug discoveries. Accordingly, to characterize our designedderivatives for drug-likeness and ADMET parameters, we per-formed in silico prediction calculations. We used different methodssuch as: ADMET descriptor from Discovery studio 4.0 (AccelrysSoftware Inc., Discovery Studio Modeling Environment, Release4.0 (2013) San Diego: Accelrys Inc. San Diego, CA), ADMEscreen from TSAR 3.1(TSAR, Version 3.1 (2007) Accelrys Inc.,San Diego, CA), DruLiTo module (http://www.niper.gov.in/pi devtools/DruLiToWeb/DruLiTo index.html), T.E.S.T. (version 4.1) [19]and PreADMET server (http://preadmet.bmdrc.org/).

The ADMET predictions based on different developmentalmodels included descriptors on aqueous solubility, intestinalabsorption, CYP2D6 binding, blood-brain barrier (BBB) penetration,mutagenicity and Permeability-glycoprotein (P-gp) inhibition.

The compound set for testing contained different compoundsalong with our designed derivatives. To ensure unbiased screen-ing we included other known compounds that carry different PKPD profiles. A multi-method approach was used to cross validatethe output from different methods and to remove the model spe-cific biasness. Overall, pragmatic screening approach was used forstringent drug-likeness and ADMET screening.

3. Results and discussion

3.1. Tyr521Ser and Ser523Asn together confer sordarin resistance inC. parapsilosis and C. lusitaniae

In this study we have focused on five fungal species (having sub-stitutions at important positions within SSR) which show varyingdegree of insensitivity towards sordarin (Supplementary Fig. S1B).In order to design a broad spectrum antifungal agent, it is impor-tant to identify the molecular basis of sordarin-resistance shownby the most resistant fungal species.

Out of all the fungal species under consideration, C. parapsilosisand C. lusitaniae were of special interest as they showed complete

resistance towards sordarin. Interestingly, eEF2s of both the specieshave a Tyr to Ser and Ser to Asn substitution at 521 and 523 posi-tions of the SSR respectively implying the significance of thesetwo residues in sordarin binding. Previous biochemical studies also

136 B. Chakraborty et al. / Journal of Molecular Graphics and Modelling 66 (2016) 133–142

Fig. 1. Changes within the SSR and non-SSR residues in the sordarin-binding cavityof human eEF2 structure. (A) The SSR and non-SSR set of mutations found in humaneEF2 cavity which were introduced in the cavity of S.cerevisiae eEF2 (pdb code 1N0U),both separately and together. (B) Sordarin binding cavity (pale cyan) of S. cerevisiaewith sordarin (brown) within it where SSR and nonSSR set of residues shown in sticks(coloured in green and yellow respectively). (C) The cavity in human eEF2 with thedct

ds

wrt(tdaaseri

dai(

3f

hnhAtinter

maciwT

Fig. 2. Sordarin-binding pocket and modification in the sordarin structure. (A) Sor-darin binding cavity of S. cerevisiae eEF2 (pale cyan) and C. parapsilosis eEF2 (cyan)superimposed over one another with SSR of each species coloured in smudge greenand lime green respectively. The nonconserved residues of C. parapsilosis SSR (Ser521and Asn523) have been coloured in red. Tyr521 of S. cerevisiae comes very close tothe C61 position of sordarin (brown) while Ser521is placed quite far apart from thesame. (B) Sordarin structure (2D) and different modifying groups attached at C61position of the drug in order to create the derivatives. (For interpretation of the ref-

rug (brown) where the same set of corresponding residues shown in sticks (sameolour codes). (For interpretation of the references to colour in this figure legend,he reader is referred to the web version of this article.)

emonstrated the importance of the 521 residue for sordarin sen-itivity [8].

Before proceeding with modification in sordarin structure, weanted to categorically estimate the effect of each of the two

esidues towards conferring sordarin resistance. We introduced thewo above mentioned substitutions in S. cerevisiae eEF2 structurepdb code 1N0U) both separately as well as together. Local struc-ure optimization of sordarin into the modified eEF2 cavity wasone using flexible docking procedure (see Methods). The physicalnd chemical compatibility of the drug and the eEF2 cavity weressessed based on the ‘deterministic parameters’ [9]. The resultsuggested that dual substitution at positions 521 and 523 is moreffective than either single substitution alone in causing sordarinesistance implying that sordarin insensitivity of both the speciess a cumulative effect of both the residues (Table 1).

Based on this finding, we analyzed the binding of each and everyerivative prepared with S. cerevisiae eEF2 model with Tyr521Ser

nd Ser523Asn substitutions as the first step to check the credibil-ty of a derivative for being a broad spectrum antifungal candidateTable 2).

.2. Accumulating effects of SSR and non-SSR residues responsibleor sordarin resistance of human eEF2

An effective antifungal agent should be insensitive towardsuman which made it necessary to analyze the effects of all theonconserved residues of human eEF2 in detail. Human eEF2 cavityave been found to have some additional changes (Gln490–Arg506;la562–Ser578; Phe729–Tyr745; Asn603–Lys619) apart from

hose within the SSR (Fig. 1), which were also found to be contribut-ng substantially towards sordarin resistance in human [9]. Theseonconserved residues will be collectively called as ‘non-SSR muta-ions’ (Fig. 1A and C) from now onwards. Next, we evaluated theffects of SSR and non-SSR mutations towards conferring sordarinesistance.

In order to understand the contribution of SSR and nonSSRutations in human, we introduced the two sets of mutations (SSR

nd nonSSR) separately as well as together into S. cerevisiae eEF2

avity (Fig. 1). Although the SSR mutations are found to have greatermpact than the nonSSR ones, complete resistance is achieved only

hen both the set of mutations were introduced together (Table 3).his result implies that the nonSSR mutations, although not much

erences to colour in this figure legend, the reader is referred to the web version ofthis article.)

effective on their own, significantly add up to the effect of the SSRmutations.

This observation is crucial for developing a broad spectrum sor-darin derivative. Since we were aiming at increasing favourablecontacts of the drug with SSR of different fungal species so that itbinds well in all the fungal eEF2 cavities, it is quite possible that thedrug might even bind into the human eEF2 cavity which is strictlyundesirable. Our observations showing that the nonSSR muta-tions make the human cavity significantly different from its fungalcounterparts, ensuring, at least to certain extent, some favourablecontacts established between the derivatives and human SSR uponmodification of sordarin would not contribute to human eEF2 sen-sitivity towards those derivatives.

3.3. Generation of sordarin derivatives guided by structuralknowledge and rationale behind choosing sites for modification

Unlike most of the previous attempts, our approach to developa broad spectrum antifungal agent was by structural knowledge-based modification of sordarin. We aimed to generate a derivativeof sordarin that is effective for all the fungal species while remainsineffective towards human.

To this end, we introduced different modifications at a singlesite (C61) of the drug (Fig. 2A and B) and analyzed for physical andchemical compatibility following docking of each derivative withinthe sordarin binding cavity of different fungal as well as humaneEF2s (as described in Section 2). The ‘deterministic parameters’were used to assess the drug-protein compatibility.

The logic behind choosing the site for modification within sor-darin is as follows. Our observations pointed out that favourableinteraction of the drug with the residue at position 521 is criticalfor drug binding. Our goal was to modify sordarin so as to increase

B. Chakraborty et al. / Journal of Molecular Graphics and Modelling 66 (2016) 133–142 137

Table 1Values for the ‘Deterministic Parameters’ showing the effects of Tyr 521Ser and Ser 523Asn mutations (incorporated separately as well as collectively) in yeast eEF2 towardsconferring sordarin resistance.

eEF2 vs. Sordarin Complementarity Cross correlation coefficient Accessible Surface Area (Å2)

C. parapsilosis 0.56a 0.56b ∼190b

C. lusitaniae 0.45b 0.51b ∼198b

EF2 1N0U (Tyr 521Ser ) 0.59a 0.61b ∼172a

EF2 1N0U (Ser 523Asn) 0.58a 0.66a ∼176a

EF2 1N0U (Tyr 521Ser ; Ser 523Asn) 0.50a 0.54b ∼182b

a Within range.b Out of range.

Table 2Values of the ‘Deterministic Parameters’ generated for all the sordarin derivatives (Der1-7) bound to yeast eEF2 (pdb code: 1N0U) with Tyr521Ser; Ser523Asn substitutions.

Sordarin derivatives against eEF2 1N0U(Tyr521Ser;Ser523Asn) Complementarity Cross correlation co-efficient Accessible surface area

Der1 0.42b 0.33b 184b

Der2 0.49b 0.50b 206b

Der3 0.58a 0.47b 198b

Der4 0.38b 0.57b 179a

Der5(Triazole) 0.53a 0.62a 155a

Der6(CHNH) 0.63a 0.68a 146a

Der7(Aminopyrrole) 0.63a 0.65a 151a

a Within range.b Out of range.

Table 3‘Deterministic Parameters’ used to evaluate the effect of SSR and NonSSR mutations in eEF2 cavity towards sordarin binding. (In EF2 1NOU SSR HUMAN, yeastSSR residues were substituted by Human SSR residues, in EF2 1N0U NonSSR HUMAN, human nonSSR residues were introduced into yeast eEF2 cavity, and inEF2 1N0U SSR NonSSR HUMAN, both SSR and nonSSR human residues were introduced in yeast eEF2 cavity (1N0U: S.cerevisiae eEF2 PDB code).

eEF2 vs. Sordarin Complementarity Cross correlation coefficient Accessible Surface Area (Å2)

HUMAN 0.44b 0.54b ∼177a

EF2 1N0U SSR HUMAN 0.50a 0.54b ∼152a

EF2 1N0U NonSSR HUMAN 0.53a 0.68a ∼148a

EF2 1N0U SSR NonSSR HUMAN 0.39b 0.57b ∼160a

tipwdgt5

3c

satittw(swhCw

bl

a Within range.b Out of range.

he possibility of its interaction with any residue at position 521rrespective of species. With that perspective we have chosen C61osition of sordarin (which is the closest neighbour of residue 521ithin the SSR region) as the site for modification and generatederivatives by attaching various types of groups at that position. Theroups were selected such a way that additional favourable interac-ions may occur in case of sordarin resistant eEF2s where position21 is occupied by Ser residue (C. parapsilosis and C. lusitaniae).

.4. 2-aminopyrrole derivative of sordarin found to be a potentialandidate as a pan-fungal antimycotic agent

Upon primary computational screening (Table 2) of all theeven derivatives we had generated (Supplementary Fig. S2), theminopyrrole derivative was finally selected as the one most likelyo be effective (Table 4). The derivative was prepared by introduc-ng a 2-aminopyrrole group at C61 position by replacing one of thehree hydrogen of the methyl group at that position. Pyrrole deriva-ives have been used as part of different known antimicrobial drugshich prompted us to consider this group [20,21]. Further, amino

NH2) group was incorporated at position 2 (2-aminopyrrole is thetable tautomeric form of the molecule as reported in: Ref. [22])ith the speculation that it might form hydrogen bond with theydroxyl (OH) group of tyrosine (in S. cerevisiae) and serine (in. parapsilosis and C. lusitaniae) residues present at position 521

ithin SSR.

Interaction of the predicted orientation of the drug with itsinding site is analysed in 2D diagrams (Fig. 3). By comparing the

igand-protein interactions in case of sordarin and aminopyrrole

derivative with yeast (Fig. 3A, B) and other two resistant varietiesof fungal eEF2s (Fig. 3C–F), it appears that favourable interactionswith amino acid side/main chains of sordarin-resistant eEF2 speciesare increased for aminopyrrole derivative. Surface potentials insidethe cavity show marked differences in different eEF2s because oflocal changes (Fig. 4). It is evident that the antipolar juxtaposition ofelectrostatic surface potentials lends stabilization when aminopyr-role binds to the eEF2 of sordarin-resistant fungal species (Fig. 4A).In contrast, this derivative shows unfavourable interactions (samepolarity interactions) with human eEF2 (Fig. 4B). Based on our find-ings we suggest that aminopyrrole derivative would be effective onall the fungal species under consideration, and importantly, wouldremain ineffective towards human (Table 4).

It should be mentioned here that two derivatives (termed asCHNH and Triazole) other than aminopyrrole exhibited effectivebinding likeliness from primary screening (Table 2). However,Triazole is not effective toward one of the fungal species (Sup-plementary Table S1) whereas human eEF2 is sensitive to CHNH(Supplementary Table S2) according to our analysis.

3.5. Early stage R&D on sordarin derivatives for drug-likenessand ADMET properties

The structure-based evaluation indicates aminopyrole deriva-tive of sordarin has increased efficiency compared to the parent

molecule sordarin (Table 4). Structure-based mandate furtherneeds to be supported with in silico screening for drug-likenessand ADMET to complete the evaluation. Along with the aminopy-rrole derivative we included CHNH and Triazole derivatives and

138 B. Chakraborty et al. / Journal of Molecular Graphics and Modelling 66 (2016) 133–142

Fig. 3. Interactions occurred between sordarin and its aminopyrrole derivative with different eEF2 cavities. Different types of contacts (represented in 2D) made by sordarin(A, C, E) and its aminopyrrole derivative (B, D, F) with their binding sites within S. cerevisiae, C. parapsilosis, C. lusitaniae eEF2s respectively. Blue lines indicate hydrogen bondswith amino acid side-chains and green lines indicate hydrogen bonds with amino acid main chains. (For interpretation of the references to colour in this figure legend, thereader is referred to the web version of this article.)

Fig. 4. Electrostatic environment of the aminopyrrole derivative-bound cavities within eEF2s. (A) Close-up view of the electrostatics surface potential of sordarin’s aminopy-rrole derivative-bound C. parapsilosis eEF2 cavity. For clarity, the surface potentials inside the cavity (top panel) and on the cavity-interacting side of the drug (bottom panel)are shown in the inset. (B) Surface charge distribution for aminopyrrole derivative-bound human eEF2 cavity is viewed. Similar to (A), the cavity (top panel) and the drug(bottom panel) interface charge patterns are shown in the inset. Blue represents positive charge, while red represents negative charge in the surface electrostatic potentialrepresentation. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)The interacting regions of the drug and cavity are marked with asterisks (*). While qualitative electrostatic complementarity is seen in (A), the drug faces same polarity onhuman eEF2 (B).

B. Chakraborty et al. / Journal of Molecular Graphics and Modelling 66 (2016) 133–142 139

Table 4Parameters used to evaluate the binding affinity of 2-aminopyrrole (aminopyrrole) derivative of sordarin to eEF2 across different fungal species.[“s” stand for “sensitive” (at least values of two parameters ‘within range’; and “is” stands for “insensitive” (at least values of two parameters ‘out of range’); ‘standard values’are given in the bottom of the Table].

eEF2 vs. Aminopyrrole derivative of sordarin Complementarity Cross correlation coefficient Accessible Surface Area Sensitivity

S. cerevisiae eEF2 [pdb code: 1N0U] 0.60a 0.61b ∼170a sEF2 1N0U (Tyr521Ser; Ser523Asn) 0.63a 0.65a ∼151a sHuman 0.46b 0.53b ∼156a isCandida glabrata (b) 0.59a 0.67a ∼163a sCryptococcus neoformans (f) 0.68a 0.66a ∼175a sSchizosaccharomyces pombe (g) 0.57a 0.70a ∼140a sCandida parapsilosis (c) 0.51a 0.75a ∼146a sCandida lusitaniae (i) 0.63a 0.59b ∼173a s

eEF2 vs. Sordarin Complementarity Cross correlation coefficient Accessible Surface Area Sensitivity

S. cerevisiae eEF2 [pdb code: 1N0U] 0.55 0.68 ∼164 s

tafl(

apipltaAdftSlwtpd

t

a Within range.b Out of range.

wo other known sordarin derivatives as well, namely, GW479821nd GW471558 which are reported to have antifungal activity [23]or screening so as to act as control compounds. As reported ear-ier our binding analysis also suggested considerable effectivenessTable 7A) of these two known azasordarin derivatives.

Parent molecule sordarin and these reported compounds acteds cross referral for understanding the drug-likeness and ADMETroperties of newly designed derivatives. Drug-likeness character-

zation (Table 5) indicates that all new derivatives carry molecularroperties that are very near to the criteria of druggability. Molecu-

ar weight of new derivatives is >500 (Table 5) which lies very nearo known antifungal GW479821 (Table 7B). As per Vebers’ rule [24]ll compounds have shown number of rotatable bonds below 10.part from that slight deviation from both the cut offs of H-bondonor (<5) & acceptor (<10) was seen for Triazole derivative androm H-bond donor (<5) cutoff for 2-aminopyrole derivative. For allhe compounds we observed no violations for MDDR-like rule [25].imilarly, logP values for all new derivatives were found within theimits of RO5. Properties of all new derivatives resided within the

indow, set by parent molecule sordarin on the lower side and byhe other known antifungal compounds on the upper side. Till this

oint we do not have any finding that supports which one of theesigned derivatives is more likely to be a drug candidate.

Next, we conducted ADMET property predictions to understandhe pharmacokinetics and pharmacodynamics of each of the com-

pounds (Table 6). ADMET calculations clearly estimate that parentmolecule sordarin holds good PK PD profile than all other com-pounds in the set. All the compounds show aqueous solubilityat 25 ◦C. Intestinal absorption and apparent Caco-2 permeability(nm/sec) provides a combo prediction method to fully under-stand the oral absorption of any drug-like compound. Amongstthree potential derivatives designed, 2-aminopyrole has shownpromising oral absorption. Metabolic enzyme cytochrome P4502D6 (CYP2D6) inhibition contributes for drug-drug interactions.Hence, as a part of regulatory procedure it is advised to checkthe compounds for any activity at this target. It is only the 2-aminopyrole derivative which has shown no CYP2D6 activity, thuscarry no drug-drug interactions unlike other designed derivativestested (CHNH and Triazole). So, CHNH and Triazole derivatives ofsordarin were not found to be suitable drug candidates from ADMETanalysis as well (Table 6).

Permeability-glycoprotein (P-gp) modulates cellular permeabil-ity by acting as efflux transporter to check the inflow of exogenoustoxic substances [26]. They are known for multidrug resistance andany affinity for P-gp can be a limiting factor for oral absorption.None of our designed derivatives showed P-gp inhibition, it was

seen only for the known antifungal GW479821. To know centralnervous system (CNS) activity of xenobiotics, blood-brain barrier(BBB) permeability prediction is a well established method. All thedesigned derivatives carry undefined BBB permeability indicating

140 B. Chakraborty et al. / Journal of Molecular Graphics and Modelling 66 (2016) 133–142

Table 5Drug likeliness prediction.

Parent Ligand& Derivatives

Mol. Weight logP H-bond Rotatable Bonds Mol. SA Mol. Vol MDDR-likerule

Acceptor Donor

Sordarin 493.63 3.21 8 2 6 428.88 378.01 No violationsTriazole 580.81 2.23 11 6 9 483.78 436.63 No violationsAminopyrrole 576.81 3.51 8 6 9 493.37 442.46 No violationsCHNH 539.79 3.10 9 5 10 475.80 419.56 No violations

Table 6ADMET predictions.

Parent Ligand &Derivatives

AqueousSolubility at 25 ◦C

IntestinalAbsorption

Apparent Caco-2Permeability(nm/sec)

CYP2D6inhibition

P-gpinhibition

BBB permeabilityfor CNS activity

Mutagenicity

Sordarin Yes Good Good Medium non non Undefined (−)veTriazole Yes Low Very low Poor inhibitor non Undefined (−)ve

nt

eiAdbiGbai7towt

tdgb

TP

Aminopyrrole Yes Low Very low Medium

CHNH Yes Good Very low Poor

o CNS activity. Mutagenicity predictions were also negative for allhe compounds under consideration.

A previous study [23] reported two azasordarin derivatives asffective antifungal agents, one (GW479821) having some safetyssues while another (GW471558) was shown to be safer as drug.mongst these two compounds (GW471558, GW479821) our pre-ictions, in line with previous studies [23], suggest GW471558 toe the safer one (Table 7C). However, medium level CNS activ-

ty is predicted (Table 7C) for this molecule also. The moleculeW479821, on the other hand, was found to have problems ofoth P-gp inhibition and CYP2D6 (Table 7C) that can be associ-ted with drug resistance and drug-drug interactions respectively,mposing doubts on its safety profile. Our assessments (TablesA–7C) on these two derivatives support the previous observa-ions [23], and thus validate our scrutiny process. Furthermore,ur proposed drug candidate 2-aminopyrrole derivative of sordarinas found to be safer than both the above mentioned deriva-

ives.Overall, our screening result suggests 2-aminopyrole as a good

emplate for prospective antifungal drug amongst the newly

esigned derivatives. From the predictions of other known antifun-al compounds the results in our hand for the designed derivativesecome much more meaningful and robust.

able 7Aarameters used to evaluate the binding affinity of two known azasordarin derivatives of

eEF2 vs. known sordarin derivatives Derivatives Complementa

EF2 1N0U (Tyr521Ser; Ser523Asn) GW479821 0.61

EF2 1N0U (Tyr521Ser; Ser523Asn) GW471558 0.5

non non Undefined (−)veinhibitor non Undefined (−)ve

4. Conclusion

One avenue of computational drug designing (e.g. High-throughput screening) relies on chance-based discoveries of activedrug molecules. The other avenue is the structure-based method(by modeling and analyzing protein-inhibitor structures) [27],which we have followed in this study. However, our approachtowards designing a more effective sordarin derivative is somewhatunconventional as compared to other commonly used structure-based drug designing pathways.

We have used a set of ‘deterministic parameters’ to evaluate thechemical as well as physical compatibility of the drug with the drug-binding cavity and the ‘binding likeliness’ predicted based on theseparameters correlates well with sordarin sensitivity. Although val-ues of the deterministic parameters vary only marginally becauseonly subtle structural differences observed among highly homolo-gous eEF2 structures, our analysis shows differences in the cavityinteriors both at physical and chemical levels which affect col-lectively in determining the ‘binding likeliness’ of the drug. Thesordarin binding to yeast eEF2 has been established by different

biochemical studies [8,28–31]. So we have compared the valuesof the ‘deterministic parameters’ for all the eEF2-sordarin (and itsderivatives) complexes with the values derived for the yeast eEF2-sordarin complex.

sordarin to yeast eEF2 mutated at 521 and 523 positions.

rity Cross correlation coefficient Accessible Surface Area

0.47 215.640.65 122.68

B. Chakraborty et al. / Journal of Molecular Graphics and Modelling 66 (2016) 133–142 141

Table 7BDrug-likeness predictions for the two known azasordarin derivatives.

Derivatives Mol. weight logP H-bond Rotatable Bonds Mol. SA Mol. Vol MDDR-likerule

Acceptor Donor

GW479821 555.83 5.2 7 2 9 507.19 439.69 No violationsGW471558 470.69 6.73 5 0 7 434.86 371.57 No violations

Table 7CADMET predictions for the two known azasordarin derivatives.

Derivatives Aqueoussolubility at 25 ◦C

IntestinalAbsorption

Apparent Caco-2Permeability(nm/sec)

CYP2D6inhibition

P-gpInhibition

BBB Permeabilityfor CNS activity

Mutagenicity

maTtsunp

asnnbid

A

IobNIf

A

i0

R

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

GW479821 Yes Low Good Good

GW471558 Yes Low Good Medium

Following this approach we identify a sordarin derivative whichight bind to all fungal eEF2 with higher affinity and thus should

ct as a better template for broad spectrum fungicide than sordarin.his derivative shows better physical and chemical complemen-arity not only to yeast eEF2 but also in sordarin-resistant fungalpecies like C. parapsilosis and C. lusitaniae. Importantly, it showsnfavourable interactions with human eEF2 cavity and simulta-eously passes early stage R&D for drug likeness and ADMETroperties, which enhances its prospect as a drug candidate.

A number of studies have been done aiming at developingnti-fungal agents with broad spectrum activity by modifying thetructure of sordarin [23,32–39]. However, while some came outot to be equally effective for all fungal species, some others wereot found to be absolutely safe. We suggest that our method woulde a useful alternative approach in structure-based drug design-

ng endeavour where the structure of drug-bound target protein isefined well.

cknowledgements

This work was supported in part by funds from the CSIR-Indiannstitute of Chemical Biology, and NIPER, Kolkata, India and Councilf Scientific and Industrial Research (CSIR) Network project. BC haseen awarded Senior Research Fellowship from CSIR, Govt. of India.. Ghoshal (Emeritus scientist, CSIR) thanks Council of Scientific &

ndustrial Research (CSIR), New Delhi for funding. P.V.P thanks CSIRor senior research fellowship under ES grant of N. Ghoshal.

ppendix A. Supplementary data

Supplementary data associated with this article can be found,n the online version, at http://dx.doi.org/10.1016/j.jmgm.2016.03.13.

eferences

[1] M.E. Klepser, Candida resistance and its clinical relevance, Pharmacotherapy26 (2006) 68S–75S.

[2] F.C. Odds, A.J. Brown, N.A. Gow, Antifungal agents: mechanisms of action,Trends Microbiol. 11 (2003) 272–279.

[3] M. Liu, M.D. Healy, B.A. Dougherty, K.M. Esposito, T.C. Maurice, C.E. Mazzucco,et al., Conserved fungal genes as potential targets for broad-spectrumantifungal drug discovery, Eukaryot. Cell 5 (2006) 638–649.

[4] G.R. Andersen, P. Nissen, J. Nyborg, Elongation factors in protein biosynthesis,Trends Biochem. Sci. 28 (2003) 434–441.

[5] S. Kurata, B. Shen, J.O. Liu, N. Takeuchi, A. Kaji, H. Kaji, Possible steps ofcomplete disassembly of post-termination complex by yeast eEF3 deduced

from inhibition by translocation inhibitors, Nucleic Acids Res. 41 (2013)264–276.

[6] R. Jorgensen, P.A. Ortiz, A. Carr-Schmid, P. Nissen, T.G. Kinzy, G.R. Andersen,Two crystal structures demonstrate large conformational changes in theeukaryotic ribosomal translocase, Nat. Struct. Biol. 10 (2003) 379–385.

[

Inhibitor Inhibitor Undefined (−)veNon Non Medium (−)ve

[7] R. Jorgensen, S.P. Yates, D.J. Teal, J. Nilsson, G.A. Prentice, A.R. Merrill, et al.,Crystal structure of ADP-ribosylated ribosomal translocase fromSaccharomyces cerevisiae, J. Biol. Chem. 279 (2004) 45919–45925.

[8] M. Shastry, J. Nielsen, T. Ku, M.J. Hsu, P. Liberator, J. Anderson, et al.,Species-specific inhibition of fungal protein synthesis by sordarin:identification of a sordarin-specificity region in eukaryotic elongation factor2, Microbiology 147 (2001) 383–390.

[9] B. Chakraborty, R. Mukherjee, J. Sengupta, Structural insights into themechanism of translational inhibition by the fungicide sordarin, J. Comput.Aided Mol. Des. 27 (2013) 173–184.

10] H.A. Carlson, J.A. McCammon, Accommodating protein flexibility incomputational drug design, Mol. Pharmacol. 57 (2000) 213–218.

11] V. Sobolev, E. Eyal, S. Gerzon, V. Potapov, M. Babor, J. Prilusky, et al., SPACE: asuite of tools for protein structure prediction and analysis based oncomplementarity and environment, Nucleic Acids Res. 33 (2005) W39–43.

12] V. Sobolev, A. Sorokine, J. Prilusky, E.E. Abola, M. Edelman, Automated analysisof interatomic contacts in proteins, Bioinformatics 15 (1999) 327–332.

13] J. Frank, M. Radermacher, P. Penczek, J. Zhu, Y. Li, M. Ladjadj, et al., SPIDER andWEB: processing and visualization of images in 3D electron microscopy andrelated fields, J. Struct. Biol. 116 (1996) 190–199.

14] J. Frank, Three-Dimensional Electron Microscopy of MacromolecularAssemblies, Oxford University Press, New York, 2006.

15] A.M. Roseman, Docking structures of domains into maps from cryo-electronmicroscopy using local correlation, Acta Crystallogr. D Biol. Crystallogr. 56(2000) 1332–1340.

16] E.F. Pettersen, T.D. Goddard, C.C. Huang, G.S. Couch, D.M. Greenblatt, E.C.Meng, et al., UCSF Chimera—a visualization system for exploratory researchand analysis, J. Comput. Chem. 25 (2004) 1605–1612.

17] M. Gerstein, R.M. Lynden-Bell, What is the natural boundary of a protein insolution? J. Mol. Biol. 230 (1993) 641–650.

18] W. Humphrey, A. Dalke, K. Schulten, VMD: visual molecular dynamics, J. Mol.Graph. 14 (1996) 33–38 (27-8).

19] T.M. Martin, P. Harten, R. Venkatapathy, S. Das, D.M. Young, A hierarchicalclustering methodology for the estimation of toxicity, Toxicol. Mech. Methods18 (2008) 251–266.

20] M.S. Mohamed, R.A. El-Domany, R.H. Abd El-Hameed, Synthesis of certainpyrrole derivatives as antimicrobial agents, Acta Pharm. 59 (2009) 145–158.

21] K.R.A. Idhayadhulla, A.J.A. Nasser, A. Manila, Synthesis and antimicrobialactivity of some new pyrrole derivatives, Bull. Chem. Soc. Ethiop. 26 (2012)429–435.

22] M. De Rosa, R.P. Issac, M. Marquez, M. Orozco, F.J. Luque, M.D. Timken,2-Aminopyrrole and simple 1-substituted 2-aminopyrroles: preparation andab initio study on the effect of solvent on the amino-imino tautomericequilibrium, J. Chem. Soc. Perkin Trans. 2 (1999) 1433–1437.

23] E. Herreros, M.J. Almela, S. Lozano, F. Gomez de las Heras, D. Gargallo-Viola,Antifungal activities and cytotoxicity studies of six new azasordarins,Antimicrob. Agents Chemother. 45 (2001) 3132–3139.

24] D.F. Veber, S.R. Johnson, H.Y. Cheng, B.R. Smith, K.W. Ward, K.D. Kopple,Molecular properties that influence the oral bioavailability of drugcandidates, J. Med. Chem. 45 (2002) 2615–2623.

25] T.I. Oprea, Property distribution of drug-related chemical databases, J.Comput. Aided Mol. Des. 14 (2000) 251–264.

26] P. Gupta, T. Garg, M. Tanmay, S. Arora, Polymeric drug-delivery systems: rolein P-gp efflux system inhibition, Crit. Rev. Ther. Drug Carrier Syst. 32 (2015)247–275.

27] G. Sliwoski, S. Kothiwale, J. Meiler, E.W. Lowe Jr., Computational methods indrug discovery, Pharmacol. Rev. 66 (2014) 334–395.

28] J.M. Dominguez, V.A. Kelly, O.S. Kinsman, M.S. Marriott, F. Gomez de las Heras,J.J. Martin, Sordarins: a new class of antifungals with selective inhibition ofthe protein synthesis elongation cycle in yeasts, Antimicrob. Agents

Chemother. 42 (1998) 2274–2278.

29] J.M. Dominguez, M.G. Gomez-Lorenzo, J.J. Martin, Sordarin inhibits fungalprotein synthesis by blocking translocation differently to fusidic acid, J. Biol.Chem. 274 (1999) 22423–22427.

1 ar Gra

[

[

[

[

[

[

[

[

[

42 B. Chakraborty et al. / Journal of Molecul

30] H. Liang, Sordarin, an antifungal agent with a unique mode of action, BeilsteinJ. Org. Chem. 4 (2008) 31.

31] M.C. Justice, M.J. Hsu, B. Tse, T. Ku, J. Balkovec, D. Schmatz, et al., Elongationfactor 2 as a novel target for selective inhibition of fungal protein synthesis, J.Biol. Chem. 273 (1998) 3148–3151.

32] P. Aviles, A. Pateman, R. San Roman, M.J. Guillen, F. Gomez De Las Heras, D.Gargallo-Viola, Animal pharmacokinetics and interspecies scaling of sordarinderivatives following intravenous administration, Antimicrob. AgentsChemother. 45 (2001) 2787–2792.

33] E. Herreros, C.M. Martinez, M.J. Almela, M.S. Marriott, F.G. De Las Heras, D.Gargallo-Viola, Sordarins: in vitro activities of new antifungal derivativesagainst pathogenic yeasts, Pneumocystis carinii, and filamentous fungi,

Antimicrob. Agents Chemother. 42 (1998)2863–2869.

34] T. Hanadate, M. Tomishima, N. Shiraishi, D. Tanabe, H. Morikawa, D. Barrett,et al., FR290581, a novel sordarin derivative: synthesis and antifungal activity,Bioorg. Med. Chem. Lett. 19 (2009) 1465–1468.

[

phics and Modelling 66 (2016) 133–142

35] A. Basilio, M. Justice, G. Harris, G. Bills, J. Collado, M. de la Cruz, et al., Thediscovery of moriniafungin, a novel sordarin derivative produced by Moriniapestalozzioides, Bioorg. Med. Chem. 14 (2006) 560–566.

36] M.H. Serrano-Wu, D.R. Laurent, T.M. Carroll, M. Dodier, Q. Gao, P. Gill, et al.,Identification of a broad-spectrum azasordarin with improvedpharmacokinetic properties, Bioorg. Med. Chem. Lett. 13 (2003) 1419–1423.

37] M.H. Serrano-Wu, D.R. St Laurent, Y. Chen, S. Huang, K.R. Lam, J.A. Matson,et al., Sordarin oxazepine derivatives as potent antifungal agents, Bioorg.Med. Chem. Lett. 12 (2002) 2757–2760.

38] J.M. Bueno, J. Chicharro, J.M. Fiandor, F. Gomez de las Heras, S. Huss,Antifungal Sordarins. Part 4: synthesis and structure–activity relationships of3′ ,4′-fused alkyl-tetrahydrofuran derivatives, Bioorg. Med. Chem. Lett. 12

(2002) 1697–1700.

39] T.C. Kennedy, G. Webb, R.J. Cannell, O.S. Kinsman, R.F. Middleton, P.J.Sidebottom, et al., Novel inhibitors of fungal protein synthesis produced by astrain of Graphium putredinis. Isolation, characterisation and biologicalproperties, J. Antibiot. 51 (1998) 1012–1018.