Ultrapotent vinblastines in which added molecularcomplexity further disrupts the target tubulindimer–dimer interfaceDaniel W. Carneya,b, John C. Lukesh IIIa,b, Daniel M. Brodya,b, Manuela M. Brütscha,b, and Dale L. Bogera,b,1

aDepartment of Chemistry, The Scripps Research Institute, La Jolla, CA 92037; and bThe Skaggs Institute for Chemical Biology, The Scripps Research Institute,La Jolla, CA 92037

This contribution is part of the special series of Inaugural Articles by members of the National Academy of Sciences elected in 2014.

Contributed by Dale L. Boger, July 14, 2016 (sent for review June 1, 2016; reviewed by Erick Carreira and Mark Searcey)

Approaches to improving the biological properties of naturalproducts typically strive to modify their structures to identify theessential pharmacophore, or make functional group changes toimprove biological target affinity or functional activity, changephysical properties, enhance stability, or introduce conformationalconstraints. Aside from accessible semisynthetic modifications ofexisting functional groups, rarely does one consider using chemicalsynthesis to add molecular complexity to the natural product. Inpart, this may be attributed to the added challenge intrinsic in thesynthesis of an even more complex compound. Herein, we reportsynthetically derived, structurally more complex vinblastines in-accessible from the natural product itself that are a stunning 100-fold more active (IC50 values, 50–75 pM vs. 7 nM; HCT116), and thatare now accessible because of advances in the total synthesis ofthe natural product. The newly discovered ultrapotent vinblas-tines, which may look highly unusual upon first inspection, bindtubulin with much higher affinity and likely further disrupt thetubulin head-to-tail α/β dimer–dimer interaction by virtue of thestrategic placement of an added conformationally well-defined,rigid, and extended C20′ urea along the adjacent continuing pro-tein–protein interface. In this case, the added molecular complex-ity was used to markedly enhance target binding and functionalbiological activity (100-fold), and likely represents a general ap-proach to improving the properties of other natural products tar-geting a protein–protein interaction.

vinblastine | drug design | chemical synthesis | tubulin | protein–proteininteraction

Vinblastine and vincristine are superb clinical drugs success-fully used in combination therapies for treatment of cancer

(Fig. 1) (1–4). Vinblastine is used in frontline therapies fortreatment of Hodgkin’s disease, testicular cancer, ovarian cancer,breast cancer, head and neck cancer, and non-Hodgkin’s lym-phoma, whereas vincristine is used in the curative treatment reg-imens for childhood lymphocytic leukemia and Hodgkin’s disease.Originally isolated from the leaves of Catharanthus roseus (L)G. Don (periwinkle) (5–8), vinblastine and vincristine were amongthe initial small molecules shown to bind tubulin and to inhibitmicrotubule formation and mitosis, defining an oncology drugtarget central to one of the most successful mechanisms of actionstill pursued today (9). As a result, they continue to be extensivelystudied due to interest in their complex dimeric alkaloid struc-tures, their role in the discovery of tubulin as an effective oncologydrug target, and their clinical importance (10–13).In the development of a total synthesis of vinblastine and

vincristine, we introduced an Fe(III)/NaBH4-mediated free-radical oxidation of the anhydrovinblastine trisubstituted alkenefor penultimate installation of the C20′ tertiary alcohol found inthe natural products (14–16). This now-powerful hydrogen atomtransfer (HAT)-initiated free-radical reaction was subsequentlydeveloped to provide a general method for functionalization of

alkenes through use of a wider range of free-radical traps (17,18) beyond O2 (air) and was explored specifically for the purpose ofproviding the late-stage, divergent (19) preparation of vinblastineanalogs that bear alternative C20′ functionality at a site previouslyinaccessible for systematic exploration (Fig. 2) (17). In addition tothe alternative free-radical traps that were introduced, the broadalkene substrate scope was defined, the addition regioselectivitywas established, the outstanding functional group tolerance wasdemonstrated, a range of Fe(III) salts and initiating hydride sourceswere shown to support the reaction, its underlying free-radical re-action mechanism was refined, and mild reaction conditions (0–25 °C,5–30 min, H2O/cosolvent) were developed that are remarkablyforgiving of the reaction parameters (17, 18).Although the vinblastine C20′ site and its hydroxyl substituent

were known to be important, the prior exploration of C20′substituent effects had been limited to a handful of alcohol ac-ylation reactions, the removal of the C20′ hydroxyl group, and aspecialized set of superacid-mediated functionalizations (3). Ourstudies permitted systematic changes at C20′ where we initiallydemonstrated that incorporation of a C20′ azide (5) or its re-duced amine (6) provided compounds 100-fold less potent thanvinblastine, but that the conversion of the amine 6 to a C20′ urea(7) provided a compound with cell growth inhibition activity

Significance

Vinblastine is a clinical drug used in frontline combinationtherapies for treatment of cancer. It acts by inhibition of mi-tosis through binding tubulin and disrupting microtubule for-mation. Because of advances in its total synthesis, we reportpreviously inaccessible and unusual modifications to vinblas-tine that improve potency a remarkable 100-fold. These ultra-potent vinblastines display much higher tubulin binding affinitiesand likely further disrupt the tubulin head-to-tail dimer–dimerinteraction by strategic placement of an added rigid, extendedgroup along the adjacent continuing protein–protein interface.Significantly, the ultrapotent vinblastines are accessible by chem-ical synthesis in three steps from commercially available materials(catharanthine, $16/g; vindoline, $36/g) based on newly in-troduced synthetic methodology and are inaccessible by nat-ural product derivatization, late-stage functionalization, orbiosynthetic methods.

Author contributions: D.W.C., J.C.L., and D.L.B. designed research; D.W.C., J.C.L., D.M.B.,and M.M.B. performed research; D.W.C., J.C.L., D.M.B., M.M.B., and D.L.B. analyzed data;and D.W.C. and D.L.B. wrote the paper.

Reviewers: E.C., Eidgenössische Technische Hochschule; and M.S., University of EastAnglia.

The authors declare no conflict of interest.1To whom correspondence should be addressed. Email: Boger@scripps.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1611405113/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1611405113 PNAS | August 30, 2016 | vol. 113 | no. 35 | 9691–9698

CHEM

ISTR

YINAUGURA

LART

ICLE

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 24

, 202

0

equal to vinblastine (Fig. 1) (17). In subsequent studies, weidentified the key structural features of such ureas that contrib-ute to their activity, including the importance of the H-bonddonor site on the C20′ nitrogen substituent (20). We additionallydefined a trend in activity where substitution of the urea terminalnitrogen improves the differential in activity of the derivativesagainst matched sensitive and resistant tumor cell lines (NR2 >NHR > NH2), discovered a series of potent disubstituted C20′ureas (e.g., 8 and 9) that displayed further improved activityagainst resistant tumor cell lines, and established that stericallydemanding C20′ ureas were surprisingly well tolerated (20, 21).The target of vinblastine is the tubulin α/β dimer–dimer in-

terface where its binding destabilizes microtubulin assemblyderived from the repetitive head-to-tail tubulin binding (9, 22).This disruption of a protein–protein interaction by vinblastineis often overlooked in discussions of such targets as candidate,but challenging, biological targets to address with small mole-cules perhaps because the target identification preceded thecontemporary interest (23–27). Herein, we report the discovery

of compounds modified at C20′ that are now a stunning 100-fold more potent than vinblastine and that may initially lookunusual in their structure. We also show that this increase inpotency correlates directly with enhanced target tubulin bindingaffinity. Significantly, the remarkable potency of the com-pounds (IC50 values as low as 50–75 pM) suggest that it is notlikely, or even possible, that their cellular functional activity isderived from stoichiometric occupancy of the relatively largenumber of intracellular tubulin binding sites, but rather impli-cates effective substoichiometric or catalytic occupancy ofcandidate binding sites sufficient to disrupt tubulin dynamics orassembly during mitosis. We suggest that the newly added lin-ear and rigid C20′ urea substituents on vinblastine extend intoand create a narrow channel adjacent to the vinblastine bindingsite, bind across this adjacent region of the protein–proteininteraction defined by the tubulin dimer–dimer interface, furtherdisrupt the tubulin α/β head-to-tail dimer–dimer interaction, andextend across the full protein–protein interface contacting solventon the distal side of the binding site. This further disruption of thetarget protein–protein interaction by an added structural elementto an already complex natural product (added molecular com-plexity) represents a unique but rational approach to substantiallyimprove the functional properties of such molecules and likelyrepresents a general approach to improving the properties ofother natural products that target protein–protein interactions.

ResultsSynthesis and Activity of Vinblastine C20′ Ureas. Two C20′ ureas 8and 9 were prepared in our earlier studies that served as ourstarting points. The first of these is the pyrrolidine urea 8, whichexhibited improved cell growth inhibition relative to vinblastine.Thus, a series of functionalized or substituted pyrrolidines wereincorporated into the C20′ urea through reaction of 20′-amino-vinblastine (6), derived from reduction of the hydroazidationproduct 5, with either p-nitrophenyl chloroformate (1.5 equiv4-NO2PhOCOCl, 10 equiv Et3N, THF, 23 °C, 4 h) or triphosgene(0.4 equiv, 2.4 equiv i-Pr2NEt, CH2Cl2, 0 °C, 15 min) followed bytreatment with a secondary amine to afford the product C20′ureas in good yields (Fig. 3). The latter procedure that uses tri-phosgene represents an improvement in our original conditions(20), avoiding the purification challenges of removing residualp-nitrophenol from the reaction products. Both set of conditionsgenerate the same intermediate C20′ isocyanate that in turn reactswith the added secondary amine to provide the product ureas.For the comparisons reported herein, the compounds were

examined for cell growth inhibition against L1210 (mouse leu-kemia), HCT116 (human colon cancer), and HCT116/VM46

Fig. 1. Natural product structures and earlier results.

Fig. 2. Hydrogen atom transfer (HAT) free-radical functionalization ofunactivated alkenes.

9692 | www.pnas.org/cgi/doi/10.1073/pnas.1611405113 Carney et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 24

, 202

0

(a matched resistant human colon cancer) tumor cell lines, thelatter of which exhibits resistance (100-fold) to vinblastinethrough the clinically observed overexpression of P-glycoprotein(Pgp). In initial studies (20), we had established that the elec-tron-rich nature of the urea was responsible in part for theirenhanced activity relative to their corresponding carbamate oramide counterparts. As a consequence, we systematically probedthe pyrrolidine urea 8 to establish whether β-substituents on thepyrrolidine (10–21) or systematic changes in its structure (22–27)might impact its binding to tubulin and cell growth inhibitionproperties much like their impact on the pyrrolidine basicityitself. Nearly all such substituted pyrrolidine ureas were lesspotent than 8 itself, most showed little preference for the sub-stituent stereochemistry (R vs. S), and there did not appear to bea direct correlation of the substituent electronic properties withits impact on the cell growth inhibition activity (Fig. 4). Theexceptions to these generalizations are the behavior of 14,bearing the (R)-3-methoxypyrrolidine, and the two isomers of the

3-fluoropyrrolidine urea (17 and 18). As interesting as theselatter observations are, these compounds only approach (14)or slightly surpass (17 and 18) the activity of 8, which bears nosubstituent. As a consequence, and rather than ascribing pro-ductive roles to the substituents, it appears that most detractfrom the properties of the parent compound 8 and a few (14,17, and 18) constitute benign substitutions.More significant was the impact of adding unsaturation (com-

pound 22). Whereas incremental decreases in cell growth activitywere seen with increasing ring size of the urea terminal cyclicamine (8 vs. 23–25, 5 < 6 or 7 < 8) or with the incorporation of thepyrrolidine in an azabicyclo[2,2,1]heptane (26), the introduction ofπ-unsaturation in either the form of a 3,4-double bond (22) or a3,4-fused phenyl ring (27 and 9) provided compounds that en-hanced cell growth inhibition and improved the differential in ac-tivity between the sensitive and resistant HCT116 cell lines (Fig. 5).It is possible that this improved activity is derived in part fromalterations in the basicity of the pyrrolidine, reductions in desta-bilizing steric interactions surrounding the pyrrolidine 3,4 position,or through acquisition of additional stabilizing interactions withtubulin, and it is most likely the result of a different combination ofsuch features for 22/9 vs. 27. Finally, replacement of the fusedbenzene in isoindoline 9 with fused pyridines (28 and 29) was notonly permitted, but the change provided a further small enhance-ment in the cell growth inhibition potency. It is especially notablethat 22, 9, 28, and 29 all exhibit cell growth inhibition activity in thevinblastine-sensitive tumor cell lines with IC50 values of 400–600pM and that their activity against the vinblastine-resistant tumorcell line (HCT116/VM46) now matches the activity that vin-blastine displays against the matched sensitive HCT116 cell line(IC50 value, 6.8 nM).Of these, further modifications of the isoindoline urea 9 were

pursued for systematic examination, recognizing the potential forfurther disruption of the tubulin dimer–dimer interface (Fig. 6).Substitution of the isoindoline was well tolerated and the seriesdisplayed a trend where electron-donating substituents (30, 35–39) proved slightly more potent than compounds bearing elec-tron-withdrawing substituents (31–34). Within the former series,further alkylation (37–39) or acylation (40–53) of the aniline 36was well tolerated. Simple acetylation to provide the acetamide40 provided a derivative with IC50 values of 470–500 pM in thevinblastine-sensitive tumor cell lines and N-benzoylation main-tained this activity with 50 displaying IC50 values of 400 pM.The site proved remarkably tolerant of steric bulk with eventhe pivoloyl amide 44 displaying potent activity, and it provedcapable of accommodating extended rigid acyl groups (e.g., 49).

Fig. 3. Three-step synthesis of vinblastine C20′ ureas.

Fig. 4. Substituted pyrrolidine ureas.

Carney et al. PNAS | August 30, 2016 | vol. 113 | no. 35 | 9693

CHEM

ISTR

YINAUGURA

LART

ICLE

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 24

, 202

0

Only bulky or rigid extended substitution on the para position ofan added N-benzoyl group (51–53) proved to diminish the ac-tivity of the now potent series. Even appending a 4-biphenyl acylgroup to the isoindoline aniline provided a compound (51) that,although less potent than either 40 or 50, was still slightly morepotent than vinblastine despite being buried in a region of thetubulin binding site thought to be sensitive to steric interactionsbefore our studies. Finally, disubstitution of the isoindoline wasfound to be well tolerated and, in the case examined (54), pro-vided a compound equipotent with 35 and slightly more potentthan the parent isoindoline 9.Even more impressive levels of activity and further improve-

ments in the cell growth inhibition were observed when thebenzoyl amide of 50 was replaced with heteroaromatic amides(Fig. 7). A clear trend in activity emerged in the series examined(55–67) to confidently indicate the functionality responsible forand needed to provide the additional enhancements in activity.Although all such compounds were exceptionally potent, thosewith six-membered heteroaromatic groups with nitrogen in the2-position relative to the acyl carbonyl routinely displayed themore potent activity (e.g., 55 vs. 50, 56, and 57; 58, 60, and 61 vs.59 and 62), and those with two such nitrogens where at least oneof which is in the 2-position displayed stunning potency (58, 60,and 61). A similar trend was observed with the five-memberedheteroaromatic substituents. It is likely this reflects an inter-nal hydrogen bond between the heteroaromatic nitrogen at the2-position and the adjacent amide NH that is observed in solutionin the 1H NMR spectra of the compounds. Complementary to thisconformational constraint, the C20′ urea also preferentially adoptsthe carbonyl eclipsed conformation that fixes a preferred orienta-tion of the C20′ urea (Fig. 7). Compounds 58, 60, and 61 exhibitedIC50 values of 50–75 pM against the vinblastine-sensitive tumor celllines (L1210 and HCT116), representing increases of 100-foldrelative to vinblastine. They also displayed subnanomolar activityagainst the vinblastine-resistant tumor cell line (HCT116/VM46;IC50 values, 830–880 pM), representing increases of 700-fold rel-ative to vinblastine. They are also 10-fold more potent than theC20′ urea 9 that lacks a substituent.Because of their extraordinary potency, two heterocyclic am-

ides (55 and 60) were further functionalized with an aryl amine,which would permit covalent conjugation with targeting modal-ities that require this ultrapotent activity (e.g., antibodies, folate)for selective delivery to tumors (Fig. 8). In each case, the anilinederivative maintained the activity of the parent analog (69 and 70vs. 60) or even further improved its potency (68 vs. 55). Clearly,varied opportunities with subtle impacts on activity are availablefor both the site and appropriate functionalization of such deriv-atives for the preparation of tumor-targeting conjugates for whichthe released free drug payload remains extraordinarily potent.

Tubulin Binding. With compounds displaying remarkable func-tional cellular potency, we sought to determine whether the im-provements in activity correlated with enhanced tubulin bindingaffinity. The most common method for assessing tubulin binding

entails inhibition of tubulin polymerization upon exposure to thecandidate compounds. However, and as noted by others, thismethod is insensitive, rarely correlates with the functional activity,and lacks the resolution to discriminate among a series of closely

Fig. 5. Pyrrolidine replacements.

Fig. 6. Substituted isoindoline ureas.

9694 | www.pnas.org/cgi/doi/10.1073/pnas.1611405113 Carney et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 24

, 202

0

related compounds. Methods that measure competitive bindingat the vinblastine tubulin binding site are more predictive, andthese typically involve measurement of the competitive dis-placement of radiolabeled vinblastine or fluorescently labeledvinblastine probes. We have previously used the competitive dis-placement of 3H-vinblastine to assess the tubulin binding of selectedmembers of the vinblastine C20′ ureas (20, 21). This assay involvesfiltering coincubation solutions through DE-81 filter disks, whichsequester tubulin and its bound ligands from solution (28, 29).However, the assay is not ideal due to both the cost associated withusing 3H-vinblastine and the technical challenges of quantita-tively reproducing the results of the assay, which uses repeatedrinses of the sequestered proteins containing reversibly boundcompounds. This is further complicated by the fact that themanufacture of DE-81 filter paper has been discontinued. As aconsequence, we elected to examine a commercially availableboron-dipyrromethene (BODIPY)–vinblastine fluorescent probeto assess tubulin binding and to develop and generalize its use ina direct competitive binding assay (30–34). Although introducedand used principally to visualize protein driven efflux from cells,BODIPY–vinblastine exhibits a significant increase in fluores-cence intensity (FI) upon binding to tubulin. This behavior ofBODIPY–vinblastine allows fluorescent measurements of coin-cubation solutions to establish the percentage displacement ofprobe by a competitive binding site ligand. Tubulin was in-cubated with BODIPY–vinblastine and a representative range ofcompetitive ligands including vinblastine, 10′-fluorovinblastine[71, a previously reported vinblastine analog with 10-fold im-proved functional activity (35)], the potent compound 28, andthree ultrapotent C20′ urea analogs (58, 60, and 61) reportedhere. At the concentration tested, vinblastine displaced 31% ofthe BODIPY–vinblastine bound to tubulin (Fig. 9). 10′-Fluo-rovinblastine and 28, which are roughly 10-fold more potentthan vinblastine, displayed a higher tubulin affinity, displacing48–49% of the BODIPY–vinblastine. Remarkably, the three

ultrapotent vinblastines (58, 60, and 61), each 100-fold more potentthan vinblastine, displaced 100% of the BODIPY–-vinblastine, in-dicating that these ultrapotent C20′ ureas possess a much strongerbinding affinity for tubulin than vinblastine, 10′-fluorovinblastine,and 28. Although it is not possible to rule out the impact of otherfeatures or even introduction of an unrecognized second mecha-nism of action, this direct correlation of functional cell growth

Fig. 7. Ultrapotent heterocyclic amides of the C20′ isoindoline urea.

Fig. 8. Further functionalized derivatives.

Fig. 9. Tubulin (0.1 mg/mL, 0.91 μM) was incubated with BODIPY–vinblastine(BODIPY–VBL) (1.8 μM) and a vinblastine analog (18 μM) at 37 °C in PEM (80 mMPIPES-K pH 6.8, 2 mM MgSO4, 0.5 mM EGTA pH 6.8) buffer containing 850 μMGTP. The BODIPY–VBL fluorescence intensity (FI) [excitation (ex), 480 nm; emis-sion (em), 514 nm] of 100-μL aliquots from each incubation were measured in afluorescence microplate reader at 37 °C. Control experiments were performedwith BODIPY–VBL (control 1) in the absence of a competitive ligand (maximum FIenhancement due to tubulin binding) and (control 2) in the absence of tubulin(no FI enhancement due to tubulin binding). Percentage BODIPY–VBL displace-ment was calculated by the formula (control 1 FI – experiment FI)/(control 1 FI –control 2 FI) × 100. Reported values are the average of four measurements ± SD.

Carney et al. PNAS | August 30, 2016 | vol. 113 | no. 35 | 9695

CHEM

ISTR

YINAUGURA

LART

ICLE

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 24

, 202

0

inhibition activity with target tubulin binding affinity and therelative magnitude of the effects suggest that the properties ofthe potent and ultrapotent C20′ ureas are derived predominately, ifnot exclusively, from on-target effects on tubulin.

Binding Model. In contrast to expectations based on the stericconstraints of the tubulin binding site surrounding the vinblastineC20′ center depicted in the X-ray cocrystal structure of a tubulin-bound complex (22), large C20′ urea derivatives such as thosedetailed herein are accommodated, with some exhibiting ultra-potent functional activity in cell-based proliferation assays andremarkable tubulin binding affinity. This raises several key ques-tions that include whether there is room for such substituents in thepurported vinblastine binding site. The binding site for vinblastineis created by and lies at the head-to-tail tubulin α/β dimer–dimer

interface (Fig. 10). Vinblastine is essentially completely buried inthe protein binding site, adopting a T-shaped bound conforma-tion with C3/C4 (bottom of T) lying at the solvent interface andthe C20′ site (top corner of T) extending deepest into the bindingpocket lying at one corner. This binding of vinblastine destabilizesthe protein–protein interaction integral to microtubulin assembly.Inspection of the vinblastine–tubulin X-ray structure reveals thatthe C20′ alcohol extends toward a narrow channel that leadsfrom the buried C20′ site to the opposite face of the protein,representing the continuation of the protein–protein interactiondefined by the tubulin dimer–dimer interface. Even withoutadjusting the proteins found in the vinblastine-bound X-raystructure, the modeled C20′ isoindoline group of 9 extends be-hind a key peptide loop in β-tubulin that constitutes the top ofthe vinblastine binding site into this narrow channel, continuingalong the tubulin protein–protein interface and presumably isresponsible for further disruption of the dimer tubulin–tubulininteraction (Fig. 10).It is likely that the rigid linear C20′ urea of the ultrapotent

vinblastines extends into and expands this narrow channel, evenfurther disrupts the tubulin α/β head-to-tail dimer–dimer interface,and extends across the full protein–protein interface, contactingsolvent on the distal side (Fig. 11). Thus, and although the struc-ture of the potent C20′ ureas and their optimization may appearunusual on a first inspection, they represent rational structuraladditions to a specific site on the natural product with appropriateconformational constraints needed to further enhance disruptionof the target protein–protein interaction.More provocatively, the activity of the ultrapotent analogs (IC50

values, 50–75 pM) suggest that it is not likely or even possible thattheir cellular functional activity is derived from stoichiometricoccupancy of the relatively large number of intracellular tubulinbinding sites. Rather, their potency suggests substoichiometric or

Fig. 10. X-ray crystal structure of tubulin-bound vinblastine (PDB ID code1Z2B) (22). (A) Binding site at the tubulin α/β head-to-tail dimer–dimer in-terface. (B) Binding site comparison of vinblastine (left column, X-ray) andmodeled vinblastine C20′ urea 9 (right column); ribbon structure of bindingsite (Top), corresponding space-filling model (Middle), and side view withthe adjacent α-tubulin protein (green) at the dimer–dimer interface re-moved. A peptide loop (highlighted in orange) on the β-tubulin protein(light blue) covers the top side of the vinblastine velbanamine subunit in-cluding C20′, creating a deep and largely hydrophobic pocket for ligandbinding. Without adjusting the proteins found in the vinblastine-boundX-ray structure, the isoindoline group of modeled 9 extends behind the looppeptide into a channel continuing along the tubulin protein–protein in-terface. Aryl amide substituents on the isoindoline such as those found in 50and 55–67 continue to extend along the tubulin dimer protein–protein in-terface (Fig. 11).

Fig. 11. Models of the ultrapotent vinblastine 60 in the vinblastine bindingsite of tubulin constructed without altering the proteins found the X-raycrystal structure of tubulin-bound vinblastine (PDB ID code 1Z2B) (22). Twopossible binding conformations are shown representing a 180° rotationabout the C20′ isoindoline nitrogen bond. In one conformation (Top), thepyrazinoyl group extends over the top of the β-tubulin peptide loop (high-lighted in orange) into a region where the α-tubulin (tan) and β-tubulin(orange/blue) subunits are in close contact, thereby further disrupting theprotein–protein interaction. In the second conformation (Bottom), the pyr-azinoyl group continues to bind along the extended dimer–dimer interface,establishing additional new contacts with α-tubulin.

9696 | www.pnas.org/cgi/doi/10.1073/pnas.1611405113 Carney et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 24

, 202

0

catalytic occupancy of candidate binding sites is sufficient to dis-rupt tubulin dynamics or assembly during mitosis. The questionthese ultrapotent vinblastines pose is whether substoichiometricbinding site occupancy is sufficient to either trap/cap polymerizingmicrotubulin in nonproductive conformations or disrupt microtubulindynamics, or whether such compounds serve catalytically to ac-tively disassemble microtubulin. Although we cannot yet answerthis question, it is a focus of current efforts.

DiscussionIn recent studies, we reported concise 12-step total syntheses ofvinblastine and related natural products (14–16) based in parton introduction of a powerful single-pot two-step diastereoselectiveFe(III)-promoted vindoline/catharanthine coupling and sub-sequent Fe(III)-NaBH4/air–mediated free-radical C20′ oxidationthat was extended for use with alternative free-radical traps (17,18). This methodology, inspired by the structure of vinblastine, haspermitted systematic studies of the effects of deep-seatedstructural changes within either the lower vindoline or uppervelbanamine subunits (13). One general observation made inthese studies is that, although modifications to the key struc-tural features of vinblastine typically result in reductions inbiological activity, addition of structural features can sub-stantially improve them (20, 21, 35, 36). This includes not onlyour demonstration of the impact of the addition of an indole C10′fluorine substituent (35) and the incorporation of the C20′ ethylgroup into a benign more complex cis-fused C15′-C20′ six-mem-bered ring (36), but also the discovery of the improved activity ofC20′ urea replacements for the tertiary alcohol (20, 21). Herein, wereported the discovery of compounds in this latter series that arenow a stunning 100-fold more potent than vinblastine. Such com-pounds including 60 exhibited IC50 values of 50–75 pM against thevinblastine sensitive tumor cell lines (e.g., HCT116), representingincreases of 100-fold relative to vinblastine, and even subnanomolaractivity against a resistant tumor cell line (HCT116/VM46; IC50values, 830–880 pM) insensitive to vinblastine by virtue of theclinically observed Pgp overexpression, representing increasesof 700-fold relative to vinblastine (Fig. 12).Approaches to improving the biological properties of natural

products typically involve efforts to simplify their structures or tomodify their structures through semisynthetic transformations,diverted total synthesis (37), or late-stage functionalizations (38,39). Rarely does one consider adding molecular complexity to

the underlying structure in a way that requires chemical synthesisof a modified natural product. In part, this may be attributed tothe added challenge that might accompany the more complexcompound synthesis. The synthetically derived ultrapotent vin-blastines that we detailed, which are 100-fold more active than thenatural product, represent vinblastine analogs accessible by chem-ical synthesis in three steps from commercially available and rela-tively inexpensive materials (catharanthine, $16/g; and vindoline,$36/g) based on methodology we introduced and are presentlyinaccessible by classical natural product derivatization, late-stagefunctionalization, or biosynthetic methods. The newly discoveredultrapotent vinblastines, which display much higher tubulin bind-ing affinities, likely further disrupt the target tubulin head-to-tailα/β dimer–dimer interaction by virtue of the strategic placement ofthe added rigid and extended C20′ urea along the adjacent con-tinuing protein–protein interface (Fig. 12). The stunning successwith vinblastine suggests the approach of adding molecular com-plexity to enhance target binding and functional biological activityfor other natural products targeting a protein–protein interactionmay represent an appealing and general approach to achievingsuch objectives.

Materials and MethodsSynthesis of Vinblastine C20′ Ureas. Full details of the synthesis, purification,and characterization of all compounds reported herein, including copies ofthe 1H NMR of all tested compounds, are provided in SI Appendix. All re-agents were obtained from commercial sources unless noted otherwise.

Cell Growth Inhibition Assays. Full details of the cell growth inhibition assays(L1210, HCT116, and HCT116/VM46) are provided in SI Appendix. Compoundswere tested in duplicate (n = 2−18 times) at six concentrations between0–1,000 nM, or 0 and 10,000 nM. The IC50 values reported represent of theaverage of 4−36 determinations (SD ± 10%).

Tubulin Binding Assay. Full details of the tubulin binding assay are provided inSI Appendix. Reported values are the average of four measurements ± SD.

Supplementary Information. Full experimental details and copies of 1H NMRspectra are provided. The supplementary data associated with this articlecan be found in SI Appendix.

ACKNOWLEDGMENTS. We especially thank Katharine K. Duncan for thebiological assessments of early samples in the series reported in refs. 20 and21. We gratefully acknowledge the financial support of National Institutesof Health Grants CA115526 and CA042056.

1. Neuss N, Neuss MN (1990) Therapeutic use of bisindole alkaloids from catharanthus.

The Alkaloids, eds Brossi A, Suffness M (Academic, San Diego), Vol 37, pp 229–240.

2. Kuehne ME, Marko I (1990) Syntheses of vinblastine-type alkaloids. The Alkaloids, eds

Brossi A, Suffness M (Academic, San Diego), Vol 37, pp 77–131.

Fig. 12. Structure of 60, the contributing conformational features of the added molecular complexity, and modeled interaction with tubulin.

Carney et al. PNAS | August 30, 2016 | vol. 113 | no. 35 | 9697

CHEM

ISTR

YINAUGURA

LART

ICLE

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 24

, 202

0

3. Borman LS, Kuehne ME (1990) Functional hot spot at the C20′ position of vinblastine.The Alkaloids, eds Brossi A, Suffness M (Academic, San Diego), Vol 37, pp 133–144.

4. Pearce HL (1990) Medicinal chemistry of bisindole alkaloids from catharanthus. TheAlkaloids, eds Brossi A, Suffness M (Academic, San Diego), Vol 37, pp 145–204.

5. Noble RL, Beer CT, Cutts JH (1958) Role of chance observations in chemotherapy:Vinca rosea. Ann N Y Acad Sci 76(3):882–894.

6. Noble RL (1964) Catharanthus roseus (Vinca rosea)—importance and value of achance observation. Lloydia 27(4):280–281.

7. Noble RL (1990) The discovery of the vinca alkaloids—chemotherapeutic agentsagainst cancer. Biochem Cell Biol 68(12):1344–1351.

8. Svoboda GH, Neuss N, Gorman M (1959) Alkaloids of Vinca rosea Linn. (Catharanthusroseus G. Don.). V. Preparation and characterization of alkaloids. J Am Pharm Assoc(Sci Ed) 48(11):659–666.

9. Jordan MA, Wilson L (2004) Microtubules as a target for anticancer drugs. Nat RevCancer 4(4):253–265.

10. Potier P (1980) Synthesis of the antitumor dimeric indole alkaloids from Catharanthusspecies (vinblastine group). J Nat Prod 43(1):72–86.

11. Kutney JP (1993) Plant cell culture combined with chemistry: A powerful route tocomplex natural products. Acc Chem Res 26(10):559–566.

12. Fahy J (2001) Modifications in the upper velbenamine part of the Vinca alkaloids havemajor implications for tubulin interacting activities. Curr Pharm Des 7(13):1181–1197.

13. Sears JE, Boger DL (2015) Total synthesis of vinblastine, related natural products, keyanalogues and development of inspired methodology suitable for the systematicstudy of their structure–function properties. Acc Chem Res 48(3):653–662.

14. Ishikawa H, Colby DA, Boger DL (2008) Direct coupling of catharanthine and vindolineto provide vinblastine: Total synthesis of (+)- and ent-(–)-vinblastine. J Am Chem Soc130(2):420–421.

15. Ishikawa H, et al. (2009) Total synthesis of vinblastine, vincristine, related naturalproducts, and key structural analogues. J Am Chem Soc 131(13):4904–4916.

16. Gotoh H, Sears JE, Eschenmoser A, Boger DL (2012) New insights into the mechanismand an expanded scope of the Fe(III)-mediated vinblastine coupling reaction. J AmChem Soc 134(32):13240–13243.

17. Leggans EK, Barker TJ, Duncan KK, Boger DL (2012) Iron(III)/NaBH4-mediated addi-tions to unactivated alkenes: Synthesis of novel 20′-vinblastine analogues. Org Lett14(6):1428–1431.

18. Barker TJ, Boger DL (2012) Fe(III)/NaBH4-mediated free radical hydrofluorination ofunactivated alkenes. J Am Chem Soc 134(33):13588–13591.

19. Boger DL, Brotherton CE (1984) Total synthesis of azafluoranthine alkaloids: Ru-fescine and imelutine. J Org Chem 49(21):4050–4055.

20. Leggans EK, Duncan KK, Barker TJ, Schleicher KD, Boger DL (2013) A remarkableseries of vinblastine analogues displaying enhanced activity and an unprecedentedtubulin binding steric tolerance: C20′ urea derivatives. J Med Chem 56(3):628–639.

21. Barker TJ, Duncan KK, Otrubova K, Boger DL (2013) Potent vinblastine C20′ ureasdisplaying additionally improved activity against a vinblastine-resistant cancer cellline. ACS Med Chem Lett 4(10):985–988.

22. Gigant B, et al. (2005) Structural basis for the regulation of tublin by vinblastine.Nature 435(7041):519–522.

23. Arkin MR, Wells JA (2004) Small-molecule inhibitors of protein–protein interactions:Progressing towards the dream. Nat Rev Drug Discov 3(4):301–317.

24. Wells JA, McClendon CL (2007) Reaching for high-hanging fruit in drug discovery atprotein–protein interfaces. Nature 450(7172):1001–1009.

25. Yin H, Hamilton AD (2005) Strategies for targeting protein–protein interactions withsynthetic agents. Angew Chem Int Ed 44(27):4130–4163.

26. Meireles LMC, Mustata G (2011) Discovery of modulators of protein–protein inter-actions: Current approaches and limitations. Curr Top Med Chem 11(3):248–257.

27. Boger DL, Desharnais J, Capps K (2003) Solution-phase combinatorial libraries: Mod-ulating cellular signaling by targeting protein–protein or protein–DNA interactions.Angew Chem Int Ed 42(35):4138–4176.

28. Owellen R, Donigian D, Hartke C, Hains F (1977) Correlation of biologic data withphysico-chemical properties among the Vinca alkaloids and their congeners. BiochemPharmacol 26(13):1213–1219.

29. Bhattacharyya B, Wolff J (1976) Tubulin aggregation and disaggregation: Mediationby two distinct vinblastine-binding sites. Proc Natl Acad Sci USA 73(7):2375–2378.

30. Jiang JD, et al. (1998) 3-(Iodoacetamido)-benzoylurea: A novel cancericidal tubulinligand that inhibits microtubule polymerization, phosphorylates bcl-2, and inducesapoptosis in tumor cells. Cancer Res 58(23):5389–5395.

31. Zhang C, et al. (2009) S9, a novel anticancer agent, exerts its anti-proliferative activityby interfering with both PI3K-Akt-mTOR signaling and microtubule cytoskeleton.PLoS One 4(3):e4881.

32. Rai A, Surolia A, Panda D (2012) An antitubulin agent BCFMT inhibits proliferationof cancer cells and induces cell death by inhibiting microtubule dynamics. PLoS One7(8):e44311.

33. Ibbeson BM, et al. (2014) Diversity-oriented synthesis as a tool for identifying newmodulators of mitosis. Nat Commun 5:3155.

34. Rashid A, Kuppa A, Kunwar A, Panda D (2015) Thalidomide (5HPP-33) suppressesmicrotubule dynamics and depolymerizes the microtubule network by binding at thevinblastine binding site on tubulin. Biochemistry 54(12):2149–2159.

35. Gotoh H, Duncan KK, Robertson WM, Boger DL (2011) 10′-Fluorovinblastine and 10′-fluorovincristine: Synthesis of a key series of modified Vinca alkaloids. ACS Med ChemLett 2(12):948–952.

36. Allemann O, Brutsch M, Lukesh JC, III, Brody DM, Boger DL (2016) Synthesis of apotent vinblastine: Rationally designed added benign complexity. J Am Chem Soc138(27):8376–8379.

37. Wilson RM, Danishefsky SJ (2006) Small molecule natural products in the discovery oftherapeutic agents: The synthesis connection. J Org Chem 71(22):8329–8351.

38. Szpilman AM, Carreira EM (2010) Probing the biology of natural products: Molecularediting by diverted total synthesis. Angew Chem Int Ed 49(50):9592–9628.

39. Hong J (2014) Natural products at the interface of chemistry and biology. Chemistry20(33):10204–10212.

9698 | www.pnas.org/cgi/doi/10.1073/pnas.1611405113 Carney et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 24

, 202

0