Approach for targeting Ras with small molecules thatactivate SOS-mediated nucleotide exchangeMichael C. Burnsa, Qi Suna, R. Nathan Danielsa,1, DeMarco Campera, J. Phillip Kennedya, Jason Phana,Edward T. Olejniczaka, Taekyu Leea, Alex G. Watersonb,c, Olivia W. Rossanesea, and Stephen W. Fesika,b,c,2

Departments of aBiochemistry, bPharmacology, and cChemistry, Vanderbilt University School of Medicine, Nashville, TN 37232

Edited by John Kuriyan, University of California, Berkeley, CA, and approved January 24, 2014 (received for review August 21, 2013)

Aberrant activation of the small GTPase Ras by oncogenic muta-tion or constitutively active upstream receptor tyrosine kinasesresults in the deregulation of cellular signals governing growthand survival in ∼30% of all human cancers. However, the discov-ery of potent inhibitors of Ras has been difficult to achieve. Here,we report the identification of small molecules that bind toa unique pocket on the Ras:Son of Sevenless (SOS):Ras complex,increase the rate of SOS-catalyzed nucleotide exchange in vitro,and modulate Ras signaling pathways in cells. X-ray crystallogra-phy of Ras:SOS:Ras in complex with these molecules reveals thatthe compounds bind in a hydrophobic pocket in the CDC25 domainof SOS adjacent to the Switch II region of Ras. The structure–activity relationships exhibited by these compounds can be rational-ized on the basis of multiple X-ray cocrystal structures. Mutationalanalyses confirmed the functional relevance of this binding siteand showed it to be essential for compound activity. These mole-cules increase Ras-GTP levels and disrupt MAPK and PI3K signalingin cells at low micromolar concentrations. These small moleculesrepresent tools to study the acute activation of Ras and high-light a pocket on SOS that may be exploited to modulate Rassignaling.

The Ras family of small GTPases functions as molecularswitches, cycling between inactive (GDP-bound) and active

(GTP-bound) states, to relay cellular signals in response to ex-tracellular stimuli. Ras activation is tightly regulated by guaninenucleotide exchange factors (GEFs), which catalyze nucleotideexchange, and GTPase-activating proteins, which aid in GTPhydrolysis (1). On activation, Ras exerts its functions throughprotein–protein interactions with effectors, such as Raf kinaseand PI3K, to promote cell growth and survival.Aberrant activation of Ras by increased upstream signaling,

loss of GTPase-activating protein function, or oncogenic muta-tion results in the deregulation of cellular signals in cancer. In-deed, aberrant Ras signaling plays a role in up to 30% of allhuman cancers, with the highest incidence of Ras mutationsoccurring in carcinomas of the pancreas (63–90%), colon (36–50%),and lung (19–30%) (2, 3). Active Ras endows cells with capa-bilities that represent the hallmarks of cancer, including theability to proliferate, evade programmed cell death, alter me-tabolism, induce angiogenesis, increase invasion and metastasis,and evade immune destruction (4). Importantly, inactivation ofoncogenic Ras has been shown to be a promising therapeuticstrategy in in vitro and in vivo models of cancer (5, 6).Despite the clinical significance of targeting Ras, the discovery

of potent inhibitors has been challenging because of a lack ofsuitable binding pockets on the surface of the protein. Althougha number of small molecules have been reported to bind directlyto Ras (7–12), these compounds have relatively poor bindingaffinities, and none have advanced to the clinic to date.An alternate approach is to target the proteins that regulate

Ras activity. The GEF Son of Sevenless (SOS) catalyzes therate-limiting step in the activation of Ras by exchanging GDPfor GTP (13). During nucleotide exchange, Ras engages ina protein–protein interaction with SOS to form a complex con-taining one SOS and two Ras molecules (Ras:SOS:Ras) (14).

SOS is unique among Ras-specific GEFs in that it has an allo-steric Ras binding site that increases its catalytic activity (14, 15)and it can potentiate the oncogenic effects of mutant K-Rasthrough the activation of WT H- and N-Ras (16). Signaling fromthese WT isoforms of Ras can support the growth of cancer cellsharboring oncogenic Ras mutations (17), and inhibiting nucleo-tide exchange is a valid approach to abrogate signaling arisingfrom both mutant and WT Ras (11). As a key control point forthe activation of multiple Ras isoforms and propagation of RTK-Ras signaling, SOS represents a promising point of interventionfor Ras-driven cancers. Here, we describe the discovery and char-acterization of small molecules that bind to a functionally relevant,chemically tractable binding pocket on the Ras:SOS:Ras complexand disrupt signaling downstream of Ras.

ResultsSmall Molecules Increase a Catalytically Active Form of Human SOS1-Mediated Nucleotide Exchange on Ras. Our laboratory has recentlyreported small molecules that bind to Ras and inhibit a catalyti-cally active form of human SOS1 (SOScat) -catalyzed nucleotideexchange (8). During these studies, we also identified moleculesfrom a related chemical series that have the opposite effect andincrease the rate of nucleotide exchange in vitro (Fig. 1). Com-pound 1, a 3-(4-aminopiperidinyl)methyl-indole with an attachedglycine, weakly increased SOScat-catalyzed nucleotide exchange,which was indicated by an increase in the exchange of BODIPY-GDP for unlabeled GTP. To improve the activity of these mol-ecules, we synthesized additional compounds based on the ami-nopiperidine indole core. The addition of a tryptophan resultedin compound 2, which activated nucleotide exchange in a con-centration-dependent manner (Fig. 1 B and E) and was more

Significance

Ras is one of the most highly validated targets in cancer;however, the discovery of potent inhibitors of Ras has beendifficult to achieve. We report the discovery of small moleculesthat bind to a pocket on the Ras:Son of Sevenless:Ras complexand alter Ras activity in biochemical and cell-based experi-ments. High-resolution cocrystal structures define the protein–ligand interactions, and the lead compounds provide a startingpoint for the discovery of potent inhibitors of Ras signaling.

Author contributions: M.C.B., Q.S., O.W.R., and S.W.F. designed research; M.C.B., Q.S.,R.N.D., and D.C. performed research; M.C.B., R.N.D., D.C., J.P.K., and A.G.W. contributednew reagents/analytic tools; M.C.B., Q.S., J.P., E.T.O., T.L., A.G.W., and O.W.R. analyzeddata; and M.C.B., A.G.W., O.W.R., and S.W.F. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Data deposition: The atomic coordinates and reflections have been deposited in theProtein Data Bank, www.pdb.org (PDB ID codes 4NYI, 4NYJ, and 4NYM).1Present address: Department of Pharmaceutical Sciences, Lipscomb University College ofPharmacy and Health Sciences, Nashville, TN 37204.

2To whom correspondence should be addressed. E-mail: stephen.fesik@vanderbilt.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1315798111/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1315798111 PNAS | March 4, 2014 | vol. 111 | no. 9 | 3401–3406

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Oct

ober

18,

202

0

potent than 1 (Table S1). The addition of a methyl or halidegroup to position 5 of the indole (compounds 3 or 4, respectively)produced an additional increase in nucleotide exchange activa-tion and lower EC50 values (Table S1). Compound 4 increasedSOScat-catalyzed nucleotide exchange with an EC50 of 14 μM(Fig. 1 C and E). Replacement of the methylene linker betweenthe indole and piperidine ring with a carbonyl resulted in a com-plete loss of activity (compound 5) (Fig. 1 D and E). Unlike ourpreviously reported inhibitors of nucleotide exchange (8), thestructure–activity relationship of this series did not correlate withdirect binding to Ras (Table S1). It was, therefore, important tounderstand how these molecules function at the molecular level.

Compounds Activate Nucleotide Exchange in an SOS-DependentManner That Does Not Involve Ras Binding to the Allosteric Siteof SOS. To determine how the compounds activate nucleotideexchange, we examined previously reported GEF-independentand -dependent mechanisms. No increase in intrinsic nucleo-tide exchange on Ras was observed on addition of up to 400 μMcompound 4 (Fig. 2A), indicating that chelation of magnesiumor destabilization of bound nucleotide is not responsible for theactivity (18, 19). Comparison of the exchange rates between in-trinsic and SOScat-catalyzed exchange revealed that compound 4

activates nucleotide exchange in an SOS-dependent manner(Fig. 2B).Crystal structures and biochemical experiments have identified

an allosteric Ras binding site on SOS that increases its catalyticactivity (14, 15). Indeed, titration of RasY64A, a mutant form ofRas that binds to the allosteric site of SOS but does not undergonucleotide exchange (15), resulted in a concentration-dependentincrease in the rate of SOScat-catalyzed nucleotide exchange.GTP-bound RasY64A was more effective than GDP-boundRasY64A at stimulating nucleotide exchange (EC50 = 0.74 μM),consistent with its role as the preferred binding partner for theallosteric site of SOS (Fig. 2 C and D) (20). Under the sameconditions, addition of 100 μM compound 4 produced an in-termediate rate of nucleotide exchange, suggesting that thecompound effect on nucleotide exchange is physiologically rel-evant compared with the activation resulting from Ras bindingat the allosteric site (Fig. 2E).To test whether binding of Ras to the allosteric site on SOS

is required for the compound-mediated increase in nucleo-tide exchange, we used two previously reported mutants,SOScat-W729E and SOScat-L687E/R688A, as well as the longerSOSDbl homology–pleckstrin homology (DH-PH)-cat construct to preventRas binding to the allosteric site (20). Consistent with reporteddata, SOScat-W729E, SOScat-L687E/R688A, and SOSDH-PH-cat hadslightly slower basal nucleotide exchange rates than WT SOScat (Fig.S1). The addition of 100 μM compound 4 to either SOScat-W729E

or SOScat-L687E/R688A resulted in activation of nucleotide ex-change, suggesting that compound-mediated activation doesnot require Ras binding to the allosteric site (Fig. 2F and Fig.S1). Nucleotide exchange reactions catalyzed by the longerconstruct of SOS containing both the autoinhibitory DH-PHand catalytic domains, SOSDH-PH-cat, revealed that compound 4is also capable of activating nucleotide exchange catalyzed byautoinhibited SOS (Fig. 2F and Fig. S1). Probing allosteric Rasbinding and compound in combination revealed that compound 4can further activate SOS-catalyzed nucleotide, even in the pres-ence of saturating levels of GTP-bound RasY64A (Fig. 2G).These data strongly support the hypothesis that these compoundsactivate nucleotide exchange through a distinct mechanism, whichcan be elicited regardless of the presence or absence of Ras boundat the allosteric site.Addition of 100 μM compound 2 to nucleotide exchange

reactions catalyzed by murine Ras-GRF1, an alternate Ras-GEF,resulted in no activation, suggesting that these compounds main-tain a degree of specificity for SOS (Fig. S2A). This finding isconsistent with a sequence alignment of the CDC25 domains ofSOS1 and Ras-GRF1, which have a 30% overall identity (Fig. S2B).Although not an activator, Brefeldin A inhibits GEF-catalyzed

nucleotide exchange by acting as an interfacial inhibitor of aGEF–GTPase interaction (21). Under our conditions, the de-crease in fluorescence observed on nucleotide release from Raswould not preclude the formation of a dead end Ras:SOS com-plex. To examine this possibility, we tested the ability of com-pound 4 to activate nucleotide exchange using a range of Ras(50 nM to 8 μM) and SOScat (25 nM to 3.5 μM) concentrations.A similar activation was observed at high Ras:SOS ratios, whichwould require multiple catalytic turnovers (Fig. S3 A and B).EC50 values remained consistent irrespective of Ras or SOScat

concentrations (Fig. S3 C and D). Compound 4 also activatednucleotide exchange using unlabeled Ras followed by the addi-tion of a mixture of SOScat and BODIPY-GTP (Fig. S3E). GDPrelease, intermediate complex formation, and BODIPY-GTP as-sociation must occur to observe an increase in fluorescence here,which was described previously (7). Activation of nucleotide ex-change under these conditions supports the conclusion that thesecompounds activate the full process of nucleotide exchange, unlikeinterfacial GEF:GTPase inhibitors.

Fig. 1. Aminopiperidine indole compounds increase SOScat-catalyzed nu-cleotide exchange on Ras. (A) Chemical structures of compounds 1–5. SOScat-catalyzed nucleotide exchange assays conducted with increasing concen-trations of compounds (B) 2, (C) 4, and (D) 5. Compound was added (at 10 s)to BODIPY-GDP–loaded Ras followed by a second addition of excess GTP ± GEF(at 120 s). Kinetics of nucleotide exchange were monitored as a decrease inrelative fluorescence units (RFU) with time. Ras alone (blue) and Ras + SOScat

(red) DMSO-matched controls are shown. Compound was added in a 10-point,twofold dilution series with a top concentration of 100 μM (black curves). Ex-periments shown in B–D were conducted in triplicate. (E) Mean rate was cal-culated and is plotted (±SD) for each compound as a function of concentration.

3402 | www.pnas.org/cgi/doi/10.1073/pnas.1315798111 Burns et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

18,

202

0

Compounds That Activate Nucleotide Exchange Bind to a PocketIdentified in the Ras:SOS:Ras Complex. We obtained X-ray structuresof multiple compounds bound to the H-Ras:SOScat:H-RasY64A

(GppNHp) ternary complex (Ras:SOS:Ras) (Fig. 3 and TableS2). The ligand-bound Ras:SOS:Ras complex structures wereobtained using both soaking and cocrystallization methods undermultiple conditions and in different crystal packing lattices. TheH-Ras isoform was found to crystallize more readily in thiscomplex than K-Ras. K-Ras and H-Ras have no residue changeswithin close proximity to the binding pocket, suggesting that thecompounds are not likely to be specific for activating oneisoform of Ras over another.The crystal structures revealed that the compounds bind to the

Ras:SOS:Ras complex in a hydrophobic pocket that is formedby the CDC25 domain of SOS adjacent to the Switch II (SwII)region of Ras (Fig. 3 A and B). The pocket is formed principallyby the αE and αF helices of the SOS catalytic domain, which areconnected by an exposed helical turn involving P894 (22). Theremainder of the pocket is formed by residues from the coiledregion and helical turn connecting the αD and αE helices of SOS.Some of the residues of SOS that form the pocket (e.g., N879,Y884, and H905) have previously been reported to directly in-teract with Ras (22). Notably, R73, located in the SwII region ofRas at the catalytic site of SOS, forms a stacking interaction withY884 and interacts with the backbone carbonyl of N879 (Fig. 3B).Importantly, N879 and Y884 form the anterior wall of the bindingpocket (Fig. 3B) and provide a direct link from the compound tothe SwII region of Ras, which is critical for binding to the catalyticdomain of SOS (23).The structure–activity relationships of aminopiperidine indole-

based compounds can be rationalized from the X-ray cocrystalstructures. All compounds bind in a similar fashion, with the coreindole occupying the most hydrophobic portion of the pocket(Fig. 3 C–E). The NH of the core indole forms a hydrogen bondat the bottom of the pocket with the backbone carbonyl of M878,whereas the aminopiperidine moiety is surface-exposed and ro-tated to D887. For compound 1, the terminal amine is orientedto the solvent (Fig. 3C). The tryptophan moiety of compound 2folds back and occupies a hydrophobic pocket located at thehelical turn formed by P894 (Fig. 3D). Compound 1, which lacksthis tryptophan moiety, cannot access this pocket (Fig. 3C, arrow).

The increased activity of compound 2 could be caused by theadditional interactions made by the tryptophan moiety. Methylsubstitution at position 5 of the core indole (compound 3) furtherimproved compound activity (Table S1). The methyl group pointsto a space unoccupied by the unsubstituted indole of compound 2(Fig. 3 D, arrow and E). Compound 4, the most active compound(Table S1), was unable to be crystallized because of limited

Fig. 2. Nucleotide exchange activation by aminopiperidine-indole compounds is SOS-dependent and does not require the allosteric Ras binding site. (A)Intrinsic nucleotide exchange in the presence of compound 4 (10-point, twofold dilution; 400 μM top concentration). Intrinsic and SOScat-catalyzed controlsare shown in blue and red, respectively. (B) SOScat-catalyzed and intrinsic nucleotide exchange displayed as a function of compound concentration (n = 3 ±SD). Nucleotide exchange with RasY64A loaded with GDP and GTP is shown in C and D, respectively (10-point, twofold dilution; 16 μM top concentration). (E )Quantification of SOScat-catalyzed nucleotide exchange with the indicated activator present (n = 3 ± SD). (F) Nucleotide exchange in the presence orabsence of 100 μM compound 4 catalyzed by SOScatW729E, SOScatL687E/R688A, or SOSDH-PH-cat. (G) SOScat-catalyzed nucleotide exchange rates displayed as a functionGTP-loaded RasY64A concentration in the presence or absence of 100 μM compound 4.

Fig. 3. Aminopiperidine indole compounds bind to the Ras:SOS:Ras ternarycomplex. (A) X-ray cocrystal structure of compound 2 bound to the H-Ras:SOScat:H-RasY64A(GppNHp) ternary complex. SOScat (orange) is bound byRasY64A-GppNHp (gray; switch regions shown in blue) at the allosteric siteand nucleotide-free Ras (gray; switch regions shown in red) at the catalyticsite. (B) The hydrophobic pocket is formed by the CDC25 domain of SOSadjacent to the SwII region of Ras. Important residues forming the pocketare labeled. (C–E) Surface depictions with aminopiperidine indole com-pounds 1, 2, and 3.

Burns et al. PNAS | March 4, 2014 | vol. 111 | no. 9 | 3403

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Oct

ober

18,

202

0

compound solubility. However, it is hypothesized to bind sim-ilar to compound 3, with the chloro substitution occupying thesame space as the methyl group of 3. Based on the crystalstructures, FITC-conjugated derivatives of compounds 2 and 4were designed. Saturation binding and competition experimentsconducted with these probes indicate that improved compoundsbind SOS with a higher affinity (Fig. S4 A and B and Table S1).These crystallographic and biochemical data suggests that theactivity of the compounds is determined by their ability to optimallyfill the pocket.

Amino Acid Substitution of Residues Within the Pocket PreventsCompound-Induced Activation of Nucleotide Exchange. Residues inthis pocket have been previously identified as being mutated indevelopmental RASopathy disorders. Two mutations in the CDC25domain of SOS, E846K and P894R, cause Noonan Syndrome (24).E846K has been shown to profoundly perturb intracellular sig-naling and P894R slightly activates nucleotide exchange on Rascompared with WT SOS (25, 26). Of particular note, P894 formsthe helical turn that defines the pocket occupied by the tryptophanmoiety of compounds 2 and 3 (Fig. 3B), further supporting thehypothesis that the binding pocket occupied by the compoundsin the cocrystal structures is important for the activation of Rasby SOS.We used the crystal structures to design mutations that would

be predicted to perturb compound binding. Nine mutants ofSOScat (D887A, D887E, D887H, D887N, D887V, F890L, L901M,L901K, and H905M) were cloned, expressed, and purified. Muta-tions of F890, L901, and H905 were designed to reduce the spaceavailable at the bottom of the hydrophobic pocket, whereasmutations of D887 were used to determine the importance ofthis residue for binding (Fig. 3B). Nucleotide exchange rateswere determined for each mutant form of SOScat from experi-ments conducted in the presence of DMSO or 100 μM com-pound 2 (Fig. 4B and Fig. S5). All mutant forms of SOScat

catalyzed nucleotide exchange, confirming the proper foldingand function of the mutant proteins (Fig. S5). Mutation of F890,L901, and H905 prevented compound-induced activation of nu-cleotide exchange, suggesting that compound activity is mediated

predominantly by hydrophobic interactions in the pocket (Fig. 4).In contrast, mutation of D887 did not prevent the ability ofcompound 2 to activate nucleotide exchange (Fig. 4 B and C).These data strongly support the conclusion that this bindingpocket is functionally important for the activation of Ras bySOS and responsible for the compound-mediated activationof SOScat-catalyzed nucleotide exchange.

Compounds Increase Ras-GTP and Perturb Ras Signaling in Cells. HeLacells treated with FITC-conjugated compound 4 and subsequentlywashed with PBS showed a strong intracellular fluorescence signal,confirming that these compounds are suitable for use in cell-basedexperiments (Fig. S6A). HeLa cells were treated for 15 min withDMSO, the inactive compound 5, or the active compounds 2 and 4to assess the ability of these compounds to activate endogenousRas. Ras-GTP levels were determined using a Ras binding domainpull-down assay. No increase in Ras-GTP levels was observed incells treated with the inactive compound 5. In contrast, treatmentwith compound 4 resulted in a threefold increase in Ras-GTPlevels (Fig. 5A), whereas compound 2 resulted in a smaller in-crease, consistent with the relative in vitro nucleotide exchangeactivity (Table S1). Treatment of HeLa cells with 100 μM com-pound 4 led to elevated Ras-GTP levels within 5 min that remainedelevated for the entirety of a 30-min time course (Fig. 5B). Theseexperiments show that the compounds activate nucleotide ex-change in the cellular setting containing full-length endogenousSOS and Ras proteins.We determined the effect of compounds 2–5 on Ras-mediated

signaling in the MAPK and PI3K pathways. Treatment withcompounds 2–4 causes a biphasic response in the MAPK path-way that is characterized by inhibition of extracellular signal-regulated kinase (ERK) phosphorylation at high compoundconcentration followed by a peak of increased ERK phos-phorylation as compound concentration decreases (Fig. 5C).This signaling pattern is most evident with compounds 3 and 4.Because of the decreased potency of compound 2, only the in-creased ERK phosphorylation is visible in this concentrationrange. Compounds 2–4 also inhibit PI3K pathway signaling, whichwas evidenced by a decrease in phosphorylation of the proteinkinase AKT. Importantly, the peak in ERK phosphorylationcorrelates with the IC50 for inhibition of phosphorylation of AKT(Fig. 5C), suggesting that the two are regulated by the sameunderlying mechanism. As expected, the inactive compound 5had no effect on ERK or AKT phosphorylation.The biphasic response in ERK phosphorylation closely resem-

bles the signaling induced by inhibitors of the B-Raf kinase in cellscontaining WT Raf (27). To investigate a similar paradoxical acti-vation mechanism, we examined compound effect on Ras signalingin melanoma lines harboring well-characterized mutations in theRas pathway. In the context of WT Ras (CHL-1) or N-RasQ61L

(SK-MEL-2), the B-Raf inhibitor dabrafenib and compound 4elicited a biphasic response in both mitogen-activated protein/extracellular signal-regulated kinase kinase (MEK) and ERKphosphorylation (Fig. 5D). In MALME-3M cancer cells, whichharbor a B-RafV600E mutation, dabrafenib was able to potentlyinhibit MEK and ERK phosphorylation, which was expected(Fig. 5D). Compound 4, however, had no effect on MEK orERK phosphorylation, suggesting that this compound acts bya unique mechanism of action at the level of the Ras–SOSinteraction, upstream of Raf kinase.To further test the hypothesis that these compounds act at the

level of the Ras–SOS interaction, serum-starved HeLa cells werepretreated with DMSO or 100 μM compound 4 and then stim-ulated with EGF. Compound 4 prevented EGF-induced activa-tion of MEK and ERK; however, it had no effect on the activationof EGF receptor upstream of Ras, which was shown by an in-crease in tyrosine 1068 phosphorylation (Fig. 5E). These data

Fig. 4. Mutation of the aminopiperidine indole binding site prevents acti-vation of nucleotide exchange. (A) Nucleotide exchange was conducted witheach mutant form of SOScat in the presence of DMSO or 100 μM compound 2as shown for SOScat-L901K. (B) Nucleotide exchange rates in the presence ofDMSO or 100 μM compound 2 (n = 3 ± SD). (C) Percent increase in nucleotideexchange rate after the addition of compound to each mutant.

3404 | www.pnas.org/cgi/doi/10.1073/pnas.1315798111 Burns et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

18,

202

0

support the conclusion that the compounds act at the level ofthe Ras–SOS interaction, downstream of EGF receptor and up-stream of Raf, and establish a means to study acute Ras-mediatedsignaling using an approach that is distinct from other smallmolecules targeting this pathway.Compounds 4 and 5 were assessed for their ability to affect cell

growth and transformation. Consistent with the signaling ob-served in these cell lines, both WT (HeLa and CHL-1) andmutant (SK-MEL-2 and PANC-1) Ras harboring cancer cellsshowed a decrease in cell proliferation and anchorage-independentgrowth after treatment with compound 4 (Fig. 5 F and G andFig. S6B). In contrast, the inactive compound 5 had little or noeffect at concentrations up to 100 μM. This evidence suggeststhat compound binding to the Ras:SOS:Ras complex does notenhance cell growth but instead, may represent a mechanismto inhibit cell proliferation and transformation.

DiscussionWe have discovered small molecules that increase the rate ofSOS-catalyzed nucleotide exchange in a GEF-dependent manner

that does not involve chelation of magnesium or destabilizationof bound nucleotide. Compounds activate nucleotide exchangeregardless of mutation or Ras occupancy at the allosteric site onSOS, suggesting that activation occurs through a distinct mech-anism. Consistent with this hypothesis, the compounds bind to ahydrophobic pocket on the CDC25 domain of SOS, which isadjacent to the SwII region of Ras at the catalytic site of SOS.Importantly, mutations, both naturally occurring and designed,support the conclusion that this pocket is functionally importantfor regulating the activation of Ras by SOS.The structure of Ras:SOS:Ras cocomplexed with compound

3 was superimposed on the known structures of the CDC25domain core of SOS1 (amino acids 780–1049, excluding thehelical hairpin amino acids 929–976; Research Collaboratory forStructural Bioinformatics Protein Data Bank ID codes 1BKD,1NVU, 1NVV, 1NVW, 1NVX, 1XD2, 1XD4, 2II0, and 3KSY).The structure of the ligand-bound CDC25 domain closelyresembles the CDC25 domain in the ligand-free structures (rmsdof 0.24 ± 0.08 Å in the Cα positions). The pocket is available forcompound binding in 77% of the known structures. Two struc-tures contain the side chain of H905 occupying the indolebinding pocket (Protein Data Bank ID codes 1XD4 and 3KSY).Although the evidence presented here suggests that the pocket isavailable when SOS is in either the active or autoinhibited state,how ligand binding is affected by membrane localization of SOS,which has been shown to be important for the activation of Ras,remains to be determined (28).Residues forming this pocket in SOS1 have a 30% identity with

the Ras-specific GEF Ras-GRF1. No activation of Ras-GRF1–catalyzed nucleotide exchange was observed on compound ad-dition. Although this result suggests that these compounds main-tain a degree of specificity for SOS1 over Ras-GRF1, sequenceand structural alignments of the CDC25 domain of SOS1 withother GEF proteins suggest that a similar pocket may exist in otherGEFs. Targeting this conserved pocket may represent a uniqueapproach to alter the function of these closely related proteins.Based on our in vitro biochemical studies, we hypothesized

that treatment of cells with these nucleotide exchange activatorswould result in an increase in downstream signaling of the MAPKand PI3K pathways. Indeed, Ras-GTP levels increase after treat-ment of HeLa cells with the compounds, consistent with theincrease in nucleotide exchange activity. However, we did notexpect the observed biphasic response in MAPK signaling or theinhibition of PI3K signaling downstream of Ras.A similar pattern of biphasic MAPK signaling has been ob-

served in other instances. Notably, B-Raf inhibitors inducea paradoxical activation of MAPK signaling in cells with WTB-Raf, and this effect is intensified by the presence of a mu-tant Ras (27). In this same setting, compound 4 elicitedsignaling similar to Raf inhibitor-induced paradoxical activation.However, in contrast to dabrafenib, no effect was seen aftertreatment with compound 4 in MALME-3M cancer cells, whichharbor a V600E mutation in B-Raf. This data suggests that thecommon biphasic signaling pattern elicited by these two compoundclasses is brought about through distinct mechanisms. Raf di-merization has been shown to underlie paradoxical activation inthe case of B-Raf inhibitors, and although Ras has been impli-cated in this dimerization event, the biochemical and structuralroles of Ras in this process remain to be elucidated (29, 30).Based on the importance of Ras in Raf inhibitor-induced par-adoxical activation and the data presented here, it is temptingto hypothesize that the signaling observed after treatment withcompound 4 is regulated at the level of the Ras–Raf interaction.Additional investigation of how these compounds alter otherinteractions, such as the Ras–Raf interaction, how they affect Rasand SOS localization, and how they influence negative feedbackloops governing signal output will be required to fully understandtheir effects.

Fig. 5. Aminopiperidine indole compounds perturb Ras signaling by actingat the level of the Ras–SOS interaction. (A) Endogenous Ras-GTP levels fromHeLa cells treated for 15 min with DMSO or 100 μM compound 5, 2, or 4. (B)Endogenous Ras-GTP levels from HeLa cells treated with 100 μM compound 4for 0–30 min. (C) HeLa cells treated for 30 min with compounds 2–5 andanalyzed by Western blot. EGF (50 ng/mL; 10 min) was used as a positivecontrol. (D) Lysates from CHL-1, SK-MEL-2, and MALME-3M cells treated for30 min with compound 4 or dabrafenib were analyzed by Western blot. (E)HeLa cells were serum-starved overnight, preteated for 5 min with DMSO or100 μM compound 4, and stimulated with EGF (50 ng/mL) for 0–15 min. (F)IC50 values for cell proliferation and (G) anchorage-independent growthafter treatment with compound 4 or 5.

Burns et al. PNAS | March 4, 2014 | vol. 111 | no. 9 | 3405

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Oct

ober

18,

202

0

X-ray crystallographic studies provided a detailed understandingof how these compounds bind and can be used to rationalize thestructure–activity relationships. The observation of additionalbinding pockets not exploited by the current compounds leadsus to believe that additional improvements in activity will beobtained by the design and synthesis of new analogs. In addition,the close proximity of the compound binding site to the SwIIregion of Ras suggests that it may be used as a starting point forthe design of interfacial inhibitors. An example of an interfacialGTPase:GEF inhibitor is provided by Brefeldin A, which targetsthe Arf1:Sec7 domain complex (21). Analogous interfacial inhib-itors, anchored in this newly identified pocket on SOS, could renderRas incapable of engaging effector proteins by forming a dead endGEF:GTPase complex.Despite being considered one of the most validated targets in

cancer, the inhibition of oncogenic Ras remains a significant chal-lenge. The scientific community has sought unique, functionallyactive small molecules to provide a path forward for the dis-covery of Ras-targeted therapeutics (31), and recent work hasaimed at validating new strategies to achieve this goal (32). Theidentification and characterization of a functionally importantsmall molecule binding site on the Ras:SOS:Ras complex provideanother innovative approach to target Ras signaling. Additionalelucidation of how this pocket regulates Ras activity and in-vestigation of this approach as a way to inhibit Ras function in cellsmay enable the discovery of therapeutics for the treatment of Ras-driven tumors.

Materials and MethodsProtein Purification and X-Ray Crystallography. For nucleotide exchange assays,recombinantly purified K-RasG12D (referred to as Ras; amino acids 1–169) and

SOScat (amino acids 564–1049) containing the Ras exchanger motif and CDC25domain were purified as described previously (8).

The H-Ras:SOScat:H-RasY64A(GppNHp) complex was prepared as describedpreviously (14). Protein:ligand complexes were prepared by adding a con-centrated DMSO stock solution of the ligand to a final concentration of 2–5mM. Additional detailed methods are in SI Materials and Methods.

Nucleotide Exchange Assays. The rate of nucleotide dissociation was de-termined using Ras preloaded with BODIPY-GDP (excitation: 485, emission:510; Life Technologies). Reactions were performed on a Hamamatsu Func-tional Drug Screening System 6000 with sequential additions of compoundfollowed by GTP ± SOScat at 10 and 120 s to a well containing BODIPY-GDP–loaded Ras. Additional details are in SI Materials and Methods.

NMR, X-Ray Crystallography, and Cell-Based Experiments. Details are in SIMaterials and Methods.

ACKNOWLEDGMENTS. Use of the Advanced Photon Source, an Office ofScience User Facility operated for the US Department of Energy Office ofScience by the Argonne National Laboratory, was supported by US De-partment of Energy Contract No. DE-AC02-06CH11357. Use of the LifeSciences Collaborative Access Team Sector 21 was supported by the Michi-gan Economic Development Corporation and Michigan Technology Tri-Corridor Grant 085P1000817. Use of the Vanderbilt NMR facility was sup-ported, in part, by Major Research Instrumentation Grant 0922862 fromthe National Science Foundation (acquisition of a 900-MHz ultra high-fieldNMR spectrometer), Shared Instrumentation Grant 1S-10RR025677-01from the National Institutes of Health (NIH; console upgrades for biolog-ical NMR spectrometers), and Vanderbilt University matching funds. Thiswork was supported by the Ann Melly Scholarship in Oncology (to M.C.B.),Public Health Service Award T32 GM07347 from the National Institute ofGeneral Medical Studies for the Vanderbilt Medical-Scientist Training Pro-gram (to M.C.B.), US NIH Grants 5DP1OD006933 (NIH Director’s PioneerAward; to S.W.F) and 5P50A095103-09 (National Cancer Institute Special-ized Program of Research Excellence in gastrointestinal cancer; to R. J. Coffey),and a Lustgarten Foundation grant (to S.W.F.).

1. Vigil D, Cherfils J, Rossman KL, Der CJ (2010) Ras superfamily GEFs and GAPs: Vali-

dated and tractable targets for cancer therapy? Nat Rev Cancer 10(12):842–857.2. Bos JL (1989) ras oncogenes in human cancer: A review. Cancer Res 49(17):4682–4689.3. Prior IA, Lewis PD, Mattos C (2012) A comprehensive survey of Ras mutations in

cancer. Cancer Res 72(10):2457–2467.4. Pylayeva-Gupta Y, Grabocka E, Bar-Sagi D (2011) RAS oncogenes: Weaving a tumori-

genic web. Nat Rev Cancer 11(11):761–774.5. Chin L, et al. (1999) Essential role for oncogenic Ras in tumour maintenance. Nature

400(6743):468–472.6. Podsypanina K, Politi K, Beverly LJ, Varmus HE (2008) Oncogene cooperation in tumor

maintenance and tumor recurrence in mouse mammary tumors induced by Myc and

mutant Kras. Proc Natl Acad Sci USA 105(13):5242–5247.7. Maurer T, et al. (2012) Small-molecule ligands bind to a distinct pocket in Ras and

inhibit SOS-mediated nucleotide exchange activity. Proc Natl Acad Sci USA 109(14):

5299–5304.8. Sun Q, et al. (2012) Discovery of small molecules that bind to K-Ras and inhibit Sos-

mediated activation. Angew Chem Int Ed Engl 51(25):6140–6143.9. Waldmann H, et al. (2004) Sulindac-derived Ras pathway inhibitors target the Ras-Raf

interaction and downstream effectors in the Ras pathway. Angew Chem Int Ed Engl

43(4):454–458.10. Rosnizeck IC, et al. (2010) Stabilizing a weak binding state for effectors in the human

ras protein by cyclen complexes. Angew Chem Int Ed Engl 49(22):3830–3833.11. Hocker HJ, et al. (2013) Andrographolide derivatives inhibit guanine nucleotide ex-

change and abrogate oncogenic Ras function. Proc Natl Acad Sci USA 110(25):

10201–10206.12. Shima F, et al. (2013) In silico discovery of small-molecule Ras inhibitors that display

antitumor activity by blocking the Ras-effector interaction. Proc Natl Acad Sci USA

110(20):8182–8187.13. Konstantinopoulos PA, Karamouzis MV, Papavassiliou AG (2007) Post-translational

modifications and regulation of the RAS superfamily of GTPases as anticancer targets.

Nat Rev Drug Discov 6(7):541–555.14. Margarit SM, et al. (2003) Structural evidence for feedback activation by Ras.GTP of

the Ras-specific nucleotide exchange factor SOS. Cell 112(5):685–695.15. Freedman TS, et al. (2006) A Ras-induced conformational switch in the Ras activator

Son of sevenless. Proc Natl Acad Sci USA 103(45):16692–16697.16. Jeng HH, Taylor LJ, Bar-Sagi D (2012) Sos-mediated cross-activation of wild-type Ras

by oncogenic Ras is essential for tumorigenesis. Nat Commun 3:1168.

17. Young A, Lou D, McCormick F (2013) Oncogenic and wild-type Ras play divergentroles in the regulation of mitogen-activated protein kinase signaling. Cancer Discov3(1):112–123.

18. Lenzen C, Cool RH, Wittinghofer A (1995) Analysis of intrinsic and CDC25-stimulatedguanine nucleotide exchange of p21ras-nucleotide complexes by fluorescencemeasurements. Methods Enzymol 255:95–109.

19. Shutes A, et al. (2007) Specificity and mechanism of action of EHT 1864, a novel smallmolecule inhibitor of Rac family small GTPases. J Biol Chem 282(49):35666–35678.

20. Sondermann H, et al. (2004) Structural analysis of autoinhibition in the Ras activatorSon of sevenless. Cell 119(3):393–405.

21. Mossessova E, Corpina RA, Goldberg J (2003) Crystal structure of ARF1*Sec7 com-plexed with Brefeldin A and its implications for the guanine nucleotide exchangemechanism. Mol Cell 12(6):1403–1411.

22. Boriack-Sjodin PA, Margarit SM, Bar-Sagi D, Kuriyan J (1998) The structural basis ofthe activation of Ras by Sos. Nature 394(6691):337–343.

23. Hall BE, Yang SS, Boriack-Sjodin PA, Kuriyan J, Bar-Sagi D (2001) Structure-basedmutagenesis reveals distinct functions for Ras switch 1 and switch 2 in Sos-catalyzedguanine nucleotide exchange. J Biol Chem 276(29):27629–27637.

24. Lepri F, et al. (2011) SOS1 mutations in Noonan syndrome: Molecular spectrum,structural insights on pathogenic effects, and genotype-phenotype correlations. HumMutat 32(7):760–772.

25. Chen PC, et al. (2010) Activation of multiple signaling pathways causes developmentaldefects in mice with a Noonan syndrome–associated Sos1 mutation. J Clin Invest120(12):4353–4365.

26. Smith MJ, Neel BG, Ikura M (2013) NMR-based functional profiling of RASopathiesand oncogenic RAS mutations. Proc Natl Acad Sci USA 110(12):4574–4579.

27. Poulikakos PI, Zhang C, Bollag G, Shokat KM, Rosen N (2010) RAF inhibitorstransactivate RAF dimers and ERK signalling in cells with wild-type BRAF. Nature464(7287):427–430.

28. Gureasko J, et al. (2008) Membrane-dependent signal integration by the Ras activatorSon of sevenless. Nat Struct Mol Biol 15(5):452–461.

29. Heidorn SJ, et al. (2010) Kinase-dead BRAF and oncogenic RAS cooperate to drivetumor progression through CRAF. Cell 140(2):209–221.

30. Hatzivassiliou G, et al. (2010) RAF inhibitors prime wild-type RAF to activate theMAPK pathway and enhance growth. Nature 464(7287):431–435.

31. Wang W, Fang G, Rudolph J (2012) Ras inhibition via direct Ras binding—is therea path forward? Bioorg Med Chem Lett 22(18):5766–5776.

32. Ostrem JM, Peters U, Sos ML, Wells JA, Shokat KM (2013) K-Ras(G12C) inhibitorsallosterically control GTP affinity and effector interactions. Nature 503(7477):548–551.

3406 | www.pnas.org/cgi/doi/10.1073/pnas.1315798111 Burns et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

18,

202

0