Medicinal and Computational Chemistry Dept., ChemDiv, Inc., 6605 Nancy Ridge Drive, San Diego, CA 92121 USA, Service: +1 877 ChemDiv, Tel: +1 858-794-4860, Fax: +1 858-794-4931, Email: ChemDiv@chemdiv.com
The small G-proteins KRas, HRas, and NRas function as GDP–GTP regulated binary switches in many signal transduction pathways that govern cell growth and differentiation. Alternation between the GDP-bound ‘off’ state and the GTP-bound ‘on’ state is controlled by the guanine nucleotide exchange factors (GEFs) SOS1 and SOS2, which catalyze the exchange of GDP for GTP, and by GTPase-activating proteins (GAPs), which terminate the ‘on’ state by catalyzing the hydrolysis of GTP to GDP (Fig. 1).
Regulation of the Ras GDP-GTP cycle in normal and neoplastic cells A. Normal Ras. Wild type Ras proteins cycle between inactive GDP-bound and active GTPbound states. Growth factors stimulate transient activation of Ras through activation of RasGEFs (e.g., Sos). Ras-GTP binds preferentially to downstream effectors (E). RasGAPs (e.g., neurofibromin) accelerate the intrinsic GTP hydrolysis activity, returning Ras to the inactive state. B. Tumorassociated Ras. Missense mutations primarily at glycine-12, glycine-13 or glutamine-61 impair intrinsic and GAPstimulated GTP hydrolysis activity, rendering Ras persistently active and GTP-bound.
Ras-GTP binds preferentially to a spectrum of functionally diverse downstream effectors. Of these, five have been validated in cell culture and/or mouse models for their requirement for mutant Ras-induced oncogenesis. In addition to direct mutational activation of Ras, Ras can also be activated indirectly, for example, by mutational inactivation of the neurofibromin RasGAP or by mutational activation of EGFR.
In the active state, Ras signals to a spectrum of functionally diverse downstream effector proteins, such as Raf, PI3K, and Ral-GDS (Fig. 2). The catalytic domain of Ras is highly homologous between the three Ras isoforms KRas, HRas, and NRas; however, major differences are found in the C-terminal hypervariable region (HVR) of these proteins. The latter part of these proteins is involved in membrane association, which is a critical requirement for downstream signaling.
Gain of function mutations in Ras proteins are frequently found in human cancers and represent poor prognosis markers for patients. These mutations occur mostly in codons 12, 13, and 61 and render these forms resistant to GAP-catalyzed hydrolysis and chronically active (Fig. 3). Mutations of the KRas isoform constitute some of the most common aberrations among all human cancers and occur in three of the top four neoplasms that cause cancer deaths in the USA: lung, colon, and pancreatic cancer.
RAS mutation in human cancers. A. Human Ras proteins. RAS genes encode 188 or 189 amino acid proteins that share the indicated amino acid identity. KRAS encodes K-Ras4A or K-Ras4B due to alternative exon four utilization, with KRAS4B the predominant transcript. B. Frequency of specific RAS mutations. KRAS mutations (17,342 unique samples with mutations in a total of 80,140 unique samples) comprise 86% of all RAS mutations documented in human tumor cells. Next most frequent are NRAS mutations (2,279 mutations in 28,521 samples) and HRAS is the least frequent (652 mutations in 19,589 samples). C. Genetic progression of pancreatic ductal adenocarcinoma. D. Genetic progression of colorectal carcinoma
Studies in animal models provided strong evidence for inhibiting Ras as a therapeutic strategy and, consequently, intensive drug discovery efforts have been directed toward targeting Ras. Yet, in 2012, three decades after the discovery of the Ras genes, this goal has not led to any effective solutions. Drug discovery efforts have mainly focused on inhibiting posttranslational modification steps, targeting the Ras protein directly, or utilizing indirect strategies, namely inhibition of downstream effectors. Phenotypic approaches, such as targeting synthetic lethality, have recently been explored as an alternative strategy (Table 1).
In the last 15 years, major efforts have been directed toward developing inhibitors of Ras effectors. Driven by experimental validation for their critical role in Ras-driven oncogenesis, discovery of the mutant B-Raf, PTEN deficiency, and mutant PI3K in human cancers, the RafMEK-ERK and PI3K-AKT-mTOR signaling pathways have been the most extensively targeted (Fig. 4). Inhibitors of Raf kinase have shown clinical benefit in the treatment of B-Raf mutant metastatic melanoma, but failed to show benefit in Ras mutant tumors. It remains unclear to what extent Ras depends on Raf kinase for transforming activity, even though Raf proteins bind directly to Ras and are important Ras effectors in normal cells. Furthermore, Raf kinase inhibitors can lead to paradoxical activation, rather than inhibition, of the MAPK pathway in Ras mutant tumor cells, a clinically relevant finding. In addition, resistance has been observed upon prolonged treatment with B-Raf inhibitors, and multiple mechanisms have been identified, including activation through N-Ras mutations. Similarly, MEK inhibitors block the Ras-MAPK pathway, but often activate the phosphoinositide 3-kinase (PI3K) pathway, and have shown little clinical benefit in Ras mutant tumors as single agents. This activation is mediated by receptor tyrosine kinases through relief of a negative feedback loop from ERK to SOS.
Ras is recruited to the cytoplasmic membrane where it exerts signaling activity (Fig. 5). A cascade of post-translational modifications on Ras, including linkage of a farnesyl group, mediates Ras membrane localization. An early research focus was inhibiting the farnesyl transferase step with small molecules, and extensive efforts were directed both in industry and in academia toward identifying suitable small-molecule drugs.
For example, Tipifarnib (1) and Lonafarnib (2) proceeded to Phase II/III clinical trials, but lacked efficacy, as the most frequently mutated Ras isoform (K-Ras) can be modified by a compensatory mechanism of geranylgeranylation. Combination of farnesyl transferase and geranylgeranyl transferase inhibitors are associated with intolerable toxicity. Other efforts focused on the ‘postprenylation’ processing enzymes RCE1 and ICMT, but these enzymes, like farnesyl transferases, have many substrates beside Ras and are expected to provide low safety margins. The currently best-studied example of a selective ICMT inhibitor is the investigational compound cysmethynil (3), but neither this compound nor other ICMT inhibitors have proceeded into clinical trials. Farnesylcysteine mimetics, which compete with Ras for binding to the membrane associated escort proteins (galectins), have also been explored and an investigational compound, salirasib (4) proceeded to Phase I/II clinical trials for hematopoietic disease, pancreatic, and lung cancer. There is evidence that salirasib possesses multiple modes of action, and it is unclear if the therapeutic effects are predominantly driven by competitive galectin binding. Another competitive galectin binder, TLN-4601 (5), failed in clinical trials.
1. Structure of KRAS
The crystal structures of Ras proteins display a typical GTPase fold that comprises a sixstranded β-sheet in the center and five α-helices flanking on the sides (Fig. 6). The N-terminal domain (residues 1–86) contains the GTP binding site, the switch regions (switch 1 and 2), residues contacting GEF, GAP, and effector proteins. The presence of GTP or GDP alters the conformation in the switch regions: In the active GTP bound state, switch 1 and 2 both interact with the γ-phosphate moiety of GTP under considerable conformational strain This strain is released upon conversion to the GDP bound state.
The crystal structure of full length K-Ras4B28 shown as a ribbon diagram (left). Secondary structure elements are labeled. The switch 1 region is colored in green and the switch 2 region in orange. The stick model shows a GTP molecule bound to the nucleotide-binding site. The green sphere is a Mg2+ ion bound to GTP. The Cterminal segment is composed of amino acid residues displayed in letters and contains a poly-lysine stretch presumably interacting with the membrane. The C-terminus is disordered in the crystal structure and therefore depicted as a dotted line. Snapshots illustrating the configurations of SI (pink) and SII (yellow) together with key side chains (sticks) that undergo major reorientation upon nucleotide exchange.
GDP or GTP bound state represent conformational states with either high or low affinity to effectors; only in the GTP bound state Ras is able to signal downstream. The switch region of the protein sequence is conserved among H-, K-, and N-Ras. In contrast, the C-terminal domain (residues 87–172) of Ras contains a more variable sequence. This part of the protein is involved in interactions with the cytoplasmic membrane, but otherwise functionally less understood. The 17 C-terminal residues comprise a hyper-variable region (HVR). It harbors one or two prenylation sites that facilitate membrane localization. Uniquely to KRas4B, a polylysine sequence, is present in the HVR and promotes membrane localization, abrogating the requirement of an additional palmitoylation step prior to anchoring to the membrane (Fig. 6). The Ras structure has a relatively flat surface. Outside the nucleotide-binding grove there are no cavities obviously suitable for small molecule ligand binding. Although generally accepted, this view may neglect the conformational dynamics of the Ras protein and the transient opening of surface pockets not fully apparent in ‘ligand-free’ crystal structures. Furthermore, allosteric interactions between the effector-binding domain (N-terminal) and the membrane-binding domain (Cterminal) appear to play a critical role in regulating Ras activity. Ligands binding at the domain interface can potentially interfere with the Ras function in an allosteric manner.
A major focus of in silico analyses has been to identify potential binding sites on Ras and subsequently predict small-molecule ligands suitable for binding at those sites (Fig. 7a). Interestingly, despite substantial differences in the methods used, there is strong agreement in the prediction of certain ligand binding sites. A comparison of the active and inactive form of Ras showed clear differences in the location of some predicted binding sites, raising the possibility of specifically targeting activated Ras. Protein–ligand interactions usually involve energetic ‘hotspots’ which are certain regions of a protein surface that contribute the majority of binding free energy, disproportional to its surface area. Targeting the ‘hotspots’ are particularly important for inhibitors of protein–protein interactions (PPI)—fragment screening is thought to be an effective method of searching for hotspots. Thus, using multiple solvent crystal structures (MSCS) and Fourier transform correlation mapping algorithm (FTmap), in parallel, the Ras surface was mapped out, and derived complementary sets of ligand binding sites (Fig. 7b, Table 2). Among the high consensus regions, sites 1 through 4 are clustered around the switch regions, whereas site 5 is located at the interface between the catalytic (residues 1–86) and allosteric (residues 87–171) domains. As shown later, these sites are indeed associated with a higher frequency of experimentally observed ligand binding.
From Figure 7 shows Ras ligand binding sites. In (A), the ligand binding sites identified experimentally or computationally are mapped onto the full length KRas crystal structure (shown in two views, front and back, for convenience). The areas of the KRas surface involved in ligand binding are color-coded. The sites are classified based on the level of consensus among different predictions; green: low, yellow: medium, red: high. Ligands that were found to bind at individual sites are listed in the text boxes. Solvent-like probes from MSCS are shown in black text. Abbreviations: ETF, 2,2,2-trifluoroethanol; GOL, glycerol; RSF, R,S,R-bisfuranol; RSG, S,R,S-bisfuranol; HEZ, 1,6-hexanediol; YEG, cyclopentanol; HEX, hexane; DMF, dimethylformamide; ACT, acetate; BDZN, benzamidine; BZIM, benzimidazole. In (B), the crystallographically determined ligands are shown as sticks. The representative contact residues for Raf and SOS1 are depicted as colored patches. Left, blue: Raf-RBD binding site; right, orange: SOS1 binding site. Site 2 is in the vicinity of a region that interacts with both RBD and SOS1.
From Figure 8: the most structurally flexible regions, ‘switch 1’ and ‘switch 2’, give rise to the transient formation of surface pockets. Some of these transient pockets are partially revealed in crystal structures while some others can be derived from molecular dynamics simulations. Using multiple mapping methods including FTmap, the entire ensemble of Ras conformations derived both experimentally and computationally was mapped. Analysis of the union of consensus sites identified four main regions of potential ligand-binding. A virtual screen using compounds from public libraries against these sites indicated the highest propensity of ligand binding at a site located between loop 7, loop 9, and helix 5 (site 5). They followed up with functional cellular assays and showed that three of the hit compounds imparted a reduction in Ras-GTP levels and decreased downstream signaling. However, no experimental evidence of direct binding to Ras was provided, and the exact inhibitory mechanism remains unclear. Cellular activities of these compounds were observed at concentrations above 10 μM, and the possibility of off-target effects cannot be ruled out.
The quality and accuracy of the computational predictions are variable given the limitations of each method. However, consensus among orthogonal methods tends to be more meaningful. Fig. 7a depicts the level of consensus of all predicted ligand-binding sites. Among the high consensus regions, sites 1 through 4 are clustered around the switch regions, whereas site 5 is located at the interface between the catalytic (residues 1–86) and allosteric (residues 87– 171) domains. As shown later, these sites are indeed associated with a higher frequency of experimentally observed ligand binding. Several groups recently succeeded with the experimental identification of small-molecule binders to Ras. Encouragingly, the identified binding sites agree with the sites of highest prediction consensus and many of the binding sites appear to be functionally relevant (Table 2, Fig. 7a). An overview of the known Ras surface binders and their mechanistic effects is shown in Table 2. The binders with more accurately characterized binding sites were found to have a functional effect on nucleotide exchange. Curiously, the compounds with a proven effect on disrupting effector protein (Raf) interaction, MCP110 and sulindac derivatives, lack such stringent biophysical characterization.
SCH-54292 and derivatives: SCH-54292 (6) was discovered in a program to identify compounds that bind to Ras at a site different from the active site and inhibit exchange from Ras-GDP to Ras-GTP. The compound was confirmed to bind to Ras and exhibited inhibitory activities against GEF-independent intrinsic nucleotide exchange with IC50’s ranging from 0.5 to 0.7 μM. It is unclear whether it also inhibits GEF-catalyzed nucleotide exchange. Guided by NMR mapping, 6 was docked onto the crystal structure of Ras. The docking model postulates that 6 binds in a cleft adjacent to the switch 2 region (site 3 in Fig. 7a) in a binding mode consistent with 6 being selective for the Ras-GDP state. The phenylhydroxylamine moiety of 6 is situated near the Mg2+ ion and potentially serves as a chelator, thereby enhancing its binding affinity. A close analogue of 6, SCH-53870 (7), displayed moderate inhibitory activity (EC50 ~ 10–20 μM) in a cellular assay while other compounds were inactive in this assay, presumably due to poor cell permeability.
The compound binding and activity appear to be highly dependent on the metal-chelating phenylhydroxylamine moiety, an undesirable functional group. The experience from other medicinal chemistry programs suggests that replacing metal-chelating groups or overcoming the dependency on them altogether can be highly challenging. Such optimization efforts would greatly benefit from high-resolution crystal structures, which to date have not been reported with compounds from this class.
Several modifications of compound 6 with a novel bicyclic scaffold derived from Darabinose—while maintaining the benzyl and the phenylhydroxylamine moieties essential for Ras binding—were recently developed using 3D-molecular modeling docking approach. Analogues from this series showed improved activities compared to 6. For example, compound 8 inhibits GDP dissociation with greater strength and the amide linker imparts higher cell permeability. Efforts to increase water solubility of 6 led to the identification of compound 9 containing more substantial structural modifications, an O-benzyl-N-(3,4-dihydroxybenzyl) hydroxylamine moiety that is Nglycosylated with D-glucose (Fig. 8). The hydroxylamine moiety is benzylated in compound 9 and no longer available for coordination with the Mg2+ ion. Compound 9 only weakly inhibits GEFcatalyzed nucleotide exchange (IC50 = 100 μM). NMR mapping suggests that 9 interacts with site 2, while the molecular dynamics and docking results favor the cleft at site 3. Removal of the phenylhydoxyamine moiety is a significant structural change from the original SCH compounds. It is possible that 9 interacts with Ras in a different way and at a different site; further characterization of the molecular interactions will be required to gain more detailed insights.
DCAI: 4,6-dichloro-2-methyl-3-minoethyl-indole (10), devloped by Genentech, represents a significant advancement in the pursuit of Ras inhibitors. DCAI was discovered through an NMR-based fragment screening effort. Unlike previously reported compounds with scarce structural information, binding of 10 and its analogs to Ras has been characterized by X-ray crystallography (Fig. 10a). The high-resolution co-crystal structures delineated a unique ligand-binding pocket on the Ras surface overlapping with site 2 (Fig. 7a), which can be expanded by compound binding. Structure analysis predicted that compound binding would interfere with the Ras/SOS interactions. Indeed, selected compounds inhibited SOS-mediated nucleotide exchange and prevented Ras activation by blocking the formation of intermediates of the exchange reaction. While Ras binders with effects on nucleotide exchange had previously been reported, the study provides for the first time high resolution structural information on functionally active ligands bound to the full-length KRas. The combination of co-crystal structures and detailed biochemical studies appeared to be particularly effective in elucidating the inhibitory mechanism for fragments with weak binding affinity.
Indole analogues: an independent NMR fragment screening study was conducted and revealed compounds that bind to site 2 (Table 2, compounds 11–14) with an affinity of 1.3–2 mM. Expanded analogues were shown to extend to site 1 affording significantly improved binding affinities compared to unsubstituted counterparts. For example, the indole analogues 15b and 15c bind to KRasG12D-GDP with a KD of 190 and 340 μM, respectively, compared to a KD of ~ 1300 μM of the unsubstituted analogue 15a. This represents the first reported example of successful affinity improvement of Ras binding molecules using structure-guided design. Similar to 10 described above, these compounds block the interaction between Ras and SOS, thereby inhibiting SOS-mediated nucleotide exchange (Table 2 and Fig. 9b). Compared to 15b and c, compound 10 does not reach into site 1 indicating that this is not a requirement for disrupting the Ras–SOS interaction. However, bexpanding into the direction of site 1imparts greater overlap with the Raf binding epitope improving the chance to orthosterically inhibit the interaction with Raf (Fig. 7b). No data were reported for 15 to support this hypothesis; nevertheless, this series represents an interesting starting point for further medicinal chemistry efforts aimed at enhancing binding affinity and functional activity.
HBS3 peptide: The synthetic HBS3 peptide (16) was designed as an orthosteric inhibitor that disrupts the SOS1–Ras protein interaction and inhibits SOS1-mediated nucleotide exchange. The αH helix of SOS1 is known to be the key structural element interacting with Ras. Compound 16, a peptidyl mimetic of SOS1 αH, was shown to bind to nucleotide-free Ras with a KD of ~ 28 μM and to GDP-bound Ras with a KD of ~ 158 μM. An NMR mapping study indicated that 16, similar to 15b and c, bind between site 1 and site 2 and spans the nucleotidebinding pocket, supporting the prediction that 16 can act as a direct mimic of the αH helix in SOS1. Compound 16 inhibited SOS1-mediated nucleotide exchange with an IC50 of ~ 25 μM. Compound 16 may prove useful as a tool to discover other compounds targeting this site on Ras.
Cyclen derivatives: Ras function is tightly correlated with its conformation. The effectors bind to the active conformation with significantly higher affinity than to the inactive conformations. Spoerner and co-workers used 31P NMR as a sensitive means of monitoring the Ras conformations in solution and showed that multiple Ras conformations co-exist in equilibrium. The authors discovered that binding to Ras by Zn2+-cyclen (17) shifted the Ras conformational equilibrium towards an inactive conformation. Rosnizeck et al. later confirmed that both 17 and Cu2+-cyclen (18) stabilize a conformational state of Ras-GTP with a low affinity to effector proteins. NMR chemical shifts indicated that there are two binding sites for 17 (site 4 and site 5 in Fig. 7b) on Ras that can be saturated at 2 mM and 6 mM compound concentrations, respectively. Binding of 17 at the distal site 5 was demonstrated by crystallography. While metal-cyclens are not drug-like, these studies revealed sites on Ras that are distinct from the ones characterized by other approaches. In an independent study, calcium acetate was found to bind at the site 5, but no functional studies were performed (Fig. 11). Further investigation will be required to assess the general utility of this binding site for allosteric modulation.
Sulindac derivatives: Sulindac, a non-steroidal anti-inflammatory drug, has been used in cancer preventative therapy. Under physiological conditions sulindac is metabolized into sulindac sulfide (19). It has recently been reported that 19 strongly inhibits Ras induced malignant transformation and Ras-dependent activation of Raf. Using biophysical methods, it was demonstrated that 19 binds reversibly to Ras, and the 15N HSQC spectra suggested a binding site in proximity to the switch regions of Ras. Biochemical assays revealed impairment of the interaction of Ras with the partner proteins SOS, Raf, and other effectors in the potency range of 100–200 μM. The primary liabilities of the sulindac derivatives are their poor potency and promiscuous pharmacological effects, confounding the interpretation of cellular data. Besides the NMR-based binding site mapping, which confirmed direct binding to Ras and provided insights into the mechanism of action, high-resolution crystal structures, currently unavailable, would be highly desirable for rationally designing inhibitors with improved potency, selectivity, and drug-likeness.
MCP compounds: Using a yeast two-hybrid system, Kato-Stankiewicz et al. discovered a series of compounds that inhibit the formation of the Ras–Raf complex. In a recent study, this effect was more precisely defined as an interference of the interaction between Ras and the Ras binding domain of Raf (Raf-RBD). A selected compound, MCP110 (20) at 3–30 μM concentration, blocked the physical interaction between Ras and Raf-RBD and functionally impaired the Ras-Raf-MEK-ERK pathway both in vivo and in cell-based systems, but it is not known whether this compound or analogs directly bind to Ras or Raf. A subsequent chemical analoging study revealed no conclusive SAR trends, and it is unclear if the observed functional activities originate from a common mechanism. The Ras superfamily comprises over 150 members that function as molecular switches in diverse cellular processes. Their functional diversity is augmented by differential sub-cellular localization and tissue distribution. Identifying means of inhibition is a major challenge not limited to Ras, but small G-proteins in general. To date, only a few small molecule inhibitors were found to directly bind to this family of proteins. We propose five theoretical mechanisms to inhibit Ras via binding to its surface:
- by GTP-competitive binding (less preferable)
- by restoring the GTPase activity compromised in mutant Ras
- by inhibiting the Ras nucleotide exchange factors, such as SOS
- by blocking the Ras-effector interactions
- by trapping Ras in an inactive conformation incompatible with effector binding and signaling
Ras and other small G-proteins possess a natural nucleotide binding pocket that accommodates GTP and GDP. Presence of the γ-phosphate in GTP-bound Ras forms the basis for maintaining the protein in a state capable of high-affinity binding to effector proteins. The idea of developing GTP mimics, similar to ATP mimics in the field of kinase inhibitors, had been initially considered but rapidly discarded. Guanine nucleotides bind with picomolar affinity (KD = 10-12 M) to Ras and occur in sub-millimolar cytosolic GTP concentrations, making it extremely challenging to develop inhibitors capable of competing off nucleotides. Conversely, the affinity of ATP for kinases is typically >1000-fold lower. Furthermore, ATP in kinases is used for rapid catalytic phosphoryl transfer, while the nucleotides in Ras stabilize inactive or active states, and it is questionable if the concept of GTP competitive inhibition could be leveraged in the same way as ATP-competitive inhibition for kinases. It is therefore unlikely that the concept of GTP-competitive inhibition is applicable for Ras. A hallmark feature of oncongenic Ras mutants is their insensitivity to the action of GTPase activating proteins (GAPs). GAPs are critical for stimulating the intrinsically low GTPase activity of Ras. One of the earliest efforts made to inhibit Ras was therefore to identify small molecules that restored GAP sensitivity to mutant Ras, an approach that will yield inhibitors specific to mutant Ras. Ideally, such compounds would mimic the effects of GAP and insert an arginine-like residue that would facilitate GTP hydrolysis in the mutant protein. Unfortunately, small molecule screening campaigns failed to find compounds that would bind and interact at the site of action. A closer look at the structural features of mutant Ras provides some clues. The most common mutation in K-Ras is a glycine to aspartate mutation at position 12 in the GTPase catalytic domain. The lack of a side chain in glycine gives sufficient room for GAP to insert an arginine residue into the catalytic site, stabilizing the negative charge of the phosphate intermediate formed during the reaction. Substitution of glycine for any amino acid with a bulkier side chain occludes the access by this arginine residue, explaining the insensitivity of this mutant to hydrolysis (Fig. 12). This occlusion poses an obstacle not only to GAP but also to any small molecule potentially mimicking GAP.
The oncogenic mutation G12D blocks GAP activity. The crystal structure of HRas and GAP (PDB code: 1WQ154) suggests that a bulky residue replacement at the G12 position will clash with the finger loop and hinder GAP from catalyzing the GTPase reaction. GAP is colored in red. HRas is colored in white. The structure captures a GTP hydrolysis transition state using a corresponding arrangement of GDP and AlF3 (colored in yellow)
Thus, unless opened through a conformational change, this site is likely not accessible to any molecule aiming to stabilize the GTPase transition state. The second most frequent Ras mutation is substitution of Gln-61. The Gln-61 side chain constitutes the heart of the catalytic center, and restoration of the catalytic competence of Ras through a small molecule binder at this area is not conceivable. Many of the Ras binders identified to date bind near or overlap with the surface involved in the association with the nucleotide exchange factor and most of these compounds were shown to inhibit nucleotide exchange. Some of these compounds were identified from fragment screens, which provide an excellent experimental approach to determine the druggability of challenging therapeutic targets. An impressive testimony to the power of this approach was the identification of inhibitors of Bcl2, a target that previously had been considered undruggable. Even with better tools in hand, increasing the binding affinity of the identified Ras binders will be extremely challenging and will require well-conceived strategies beyond simple expansion of first generation binders. Inhibition of protein-to-protein interactions is thermodynamically challenging and stabilizing an abortive complex rather than breaking of a PPI might represent a more viable strategy. In fact, this has been demonstrated by the small molecule Brefeldin A (23) in the Arf-Arno system. As 23 shows no affinity to either protein alone, it will clearly be necessary to devise screens that are specifically designed to identify interfacial binders.
Apart from the challenge of druggability, it is critical to assess the therapeutic potential of inhibiting the nucleotide exchange reaction. Mutant Ras is locked in the active GTP-bound state and is expected to be less dependent on nucleotide exchange factors for its activation. Hence, blocking the interaction with nucleotide exchange factors may have a greater effect on the activation of wild-type Ras than the mutant Ras, an undesirable therapeutic outcome. However, there is accumulating evidence that mutant and wild-type Ras isoforms signal to different downstream pathways and are both required for oncogenic transformation. Furthermore, the activation of wild-type Ras isoforms may be potentiated by mutant Ras. This potentiation mechanistically occurs through allosteric activation of SOS by abundant mutant Ras-GTP. The implications and generality of this positive feedback mechanism are not fully understood, but it might suggest the utility of a SOS inhibitor even in a Ras mutant context. Nevertheless, the question of a sufficient therapeutic window with inhibition of SOS remains. While SOS inhibitors could be efficacious in wild-type Ras tumors, particularly in combination with Raf pathway inhibitors, the importance of mutant Ras and resilience to available treatments are the primary reasons for targeting this class of enzymes.
Given that Ras exerts all its biological functions through binding to downstream effectors, inhibiting Ras–effector complexes is likely to provide the best efficacy in Ras mutant tumors. Ras signals to over a dozen different downstream effector pathways, and it remains to be determined which of these effector pathways need to be inhibited to elicit therapeutic effects. The effector-binding sites on Ras partly overlap the SOS binding site (Fig. 7b), and targeting former sites with small molecules is believed to face similar, if not greater, challenges as discussed for the disruption of the Ras–SOS interaction. The effector interaction surfaces consist of helix 1, switch 1, and the edge of the b-sheet. Apart from the flat surface, the challenge of identifying small molecule ligands binding at this site also arises from the dynamic nature of the switch 1 loop. Nevertheless, two small molecules, sulindac sulfide (19) and MCP110 (20) were found to disrupt the interaction between Ras and Raf. While there is some knowledge of the binding site of sulindac sulfide (19) (see above), there is no information on the binding site of MCP110 (20). Interestingly, this parallels the situation for EHT1864 (22) in the related GTPaseeffector system Rac1-PAK1. More detailed biophysical characterization of these compounds would be valuable. The concept of interfacial binding is likely not applicable for Ras–effector complexes unless the interfacial binders can trap these complexes in a state that is non-competent for signaling. Well-devised screens should in principle be able to identify such compounds. We expressed reservations against targeting the nucleotide exchange as an approach for Ras inhibition since it would preferentially inhibit wild-type Ras. With compounds that disrupt the effector interaction we expect to target mutant and wild-type Ras at least equally. The sites of the Ras mutations are at the catalytic domain, distal from the effector-binding site. It is therefore unlikely that differentiation between mutant and wild-type Ras can be achieved using this approach. However, cells with mutant Ras are more dependent on signaling to effectors for their survival and proliferation than cells with wild-type Ras. This differential dependence may provide an adequate therapeutic window for targeting Ras-effector interactions.
In conclusion, 30 years after the identification of the Ras oncogene no effective strategies to target Ras mutant tumors are available. It is becoming increasingly evident that inhibition of effector pathways, even with inhibitors used in combination, might not provide a satisfactory solution. Hence, the search for direct Ras inhibitors must continue. Targeting protein-to-protein interactions with small molecules has, despite the challenges, emerged as an accepted direction in drug discovery, and the Ras oncoproteins are unlikely to be overlooked. In fact, several groups recently reported small molecule Ras binders with functional effects. The best-characterized binding data resulted from NMR fragment screens conducted by two independent research groups at Vanderbilt University and at Genentech. These studies provide the first ever reported examples of high-resolution Ras small-molecule co-crystal structures and reveal clear opportunities for identifying higher affinity binders. Both compound series act as inhibitors of nucleotide exchange. For the reasons elucidated above, we are doubtful about the benefit and safety profile associated with this MOA. It remains to be seen if these binding sites can be utilized for developing binders that can block the interaction of Ras with its effector molecules. Ras has a variety of partner proteins, the nucleotide exchange factors, GAPs and many effector proteins with significant overlap in their respective binding surfaces, and it may be challenging to selectively address one set of interactions without affecting others. For this reason it will be important to complement the biophysical studies with appropriate biochemical/mechanistic assays that provide information on immediate mechanistic consequences as well as physiological consequences further downstream. An alternative to using fragment screens for lead discovery could be function-based HTS screens that utilize well-defined mechanistic readouts—for example, Ras association with a selected effector. Such efforts could be particularly powerful if the setup of the screen provides selectivity information, such as differential effects in mutant versus nonmutant Ras. While in theory compelling, such strategies have been tried by us and others, they were found to be prone to high false-positive hit rates hampering the identification of truly valuable starting points. Nevertheless, compounds have been identified from these efforts, such as MCP110 (20), a compound proven to disrupt Ras-to-Raf association and eliciting cellular responses consistent with this MOA. We suggest that biophysical characterization should be the next logical step following the identification of compounds with interesting mechanistic profiles. We believe that the complexities of the system are too high to attempt optimization in the absence of this information. Clearly, biophysical characterization may not always be straightforward; for example, Brefeldin A (19), which acts as an interfacial inhibitor of a GTPase and its nucleotide exchange factor, has no affinity to either protein and would have been missed by a fragment screen identifying Arf1 binders. Elucidation of its mechanism and unique binding characteristics required the study of the dynamic process of the nucleotide exchange mechanism. Specific mutant forms of Ras that conformationally resemble known catalytic transition states could prove highly valuable for biophysical studies; we think that this is an interesting direction worth further exploration.
In essence, we believe that both biophysical screens and functional screens can be starting points, but mechanistic characterization of biophysical screening hits and biophysical characterization of functional hits are necessary as immediate follow-up steps prior to investigating functional effects in cells, which are more prone to confounding off-target activities. Such efforts will be resource-intensive and require well-integrated and experienced biophysics, biochemistry, and medicinal chemistry teams.
The above comments addressed mainly the druggability of Ras. Another important consideration, which is at least equally challenging, is selectivity. An ideal therapeutic would selectively inhibit mutant KRas while sparing other GTPases, including the close family members H- and NRas, and wild-type KRas. It is unlikely such a profile can be achieved with a Ras-binding molecule. Mutant versus wildtype selectivity in particular will be a major challenge since the mutation is distal from the effector-binding site. Given the central role of Ras in coupling extracellular receptor signals to multiple intracellular signaling pathways governing cell growth, differentiation, and other vital cellular processes, we believe that stringent selectivity requirements are necessary in order to achieve an adequate therapeutic window. However, at this point we can only speculate, and an accurate assessment of the selectivity requirements will need to be revisited once sufficiently potent tool compounds are available. The identification of therapeutically useful Ras binders will require well-staffed teams with strong biology, biophysical, and medicinal chemistry expertise, and such efforts require robust funding and longterm commitment. Given the significant challenges and extensive failures encountered in decades of Ras research, companies are not inclined to taking such risks in the current economic and business climate. Nevertheless, the large unmet medical need and profound therapeutic benefit in developing Ras inhibitors mandate continued efforts in this field. Alliances between companies and academic labs offer perhaps the best way to tackle this challenge by bridging different perspectives (exploratory research vs therapeutic development, long term commitment vs short term deliverables), leveraging core expertise (e.g. Ras biology, fragment-based drug discovery) and sharing risks; we are hopeful to see such partnerships evolving in the near future.
2. CONCEPT AND APPLICATIONS
A combined profiling methodology that provides a consensus score and decision based on various advanced computational tools and theoretical knowledge: (A) Front-line “in silico” approaches (were applied herein)
- 2D-structure similarity to the reported K/HRas ligands (Tanimoto metric)
- Isosteric morphing procedure
- Privileged structures/fragments
- A set of compounds with high diversity in structure
- A set of compounds from PPI-targeted library (10%)
- The common “MedChem” filters.
The fundamentals for these applications were described in a wide series of our publications on the design of exploratory small molecule chemistry for bio-screening [for related data visit ChemDiv. Inc. online source: www.chemdiv.com].
- Synthesis, biological evaluation and SAR study for the selected structures:
- High-throughput synthesis with multiple parallel library validation. Synthetic protocols, building blocks and chemical strategies are available
- Library activity validation via bio-screening with related SAR study is available
- Optimization of active compounds through the second-line in silico approaches (H2L optimization)
Bioisosteric/isosteric morphing procedure
Biosteric and isosteric transformations are one of the tools that allow to balance different lead-like papameters including specificity, physicochemical, and PKPD properties in the SAR studies. In addition, bioisosteric morphing provides insight into patentability of lead candidates. Several Ras binders have been designed using this technique. We have applied this approach to design of our Ras-targeted library. Typical examples of biosteric and isosteric modifications for this class of molecules are shown in Fig. 13. Representative examples of isosteric modifications within Ras-targeted library are shown in Fig. 14
To continue reading, click the download button below.