|Home | About | Journals | Submit | Contact Us | Français|
There is now considerable and increasing evidence for a causal role of aberrant activity of the Ras superfamily of small GTPases in human cancers. These GTPases act as GDP-GTP-regulated binary switches that control many fundamental cellular processes. A common mechanism of GTPase deregulation in cancer is the deregulated expression and/or activity of their regulatory proteins, guanine nucleotide exchange factors (GEFs) that promote formation of the active GTP-bound state and GTPase activating proteins (GAPs) that return the GTPase to its GDP-bound inactive state. We assess the association of GEFs and GAPs with cancer and their druggability for cancer therapeutics.
Ras proteins (H-, N- and K-Ras) are the founding members of a large superfamily of monomeric small GTPases (20–25 kDa) that regulate diverse cellular processes that include cell cycle progression, cell survival, actin cytoskeletal organization, cell polarity and movement, and vesicular and nuclear transport1, 2. The Ras superfamily (>150 members in humans) is divided into five main families based on sequence identity and function: Ras, Rho, Rab, Arf, and Ran (BOX 1).
The human Ras superfamily comprised of over 150 members which is divided into five major branches on the basis of sequence and functional similarities1, 2. In addition to the three Ras isoforms, other members of the Ras family with important roles in cancer include Rheb and Ral proteins. The ~20 kDa core G domain (corresponding to Ras residues 4–166) is conserved among all Ras superfamily proteins and is involved in GTP binding and hydrolysis148. This domain is comprised of five conserved guanine nucleotide consensus sequence elements (Ras residue numbering) involved in binding phosphate/Mg2+ (PM) or the guanine base (G). The switch I (Ras residues 30–38) and II (59–76) regions change in conformation during GDP-GTP cycling and contribute to preferential effector binding to the GTP-bound state and the core effector domain (E; Ras residues 32–40). Ras and Rho family proteins have additional C-terminal hypervariable (HV) sequences that commonly terminate with a CAAX motif that signals for farnesyl or geranylgeranyl isoprenoid addition to the cysteine residue, proteolytic removal of the AAX residues and carboxylmethylation of the prenylated cysteine. Some are modified additionally by a palmitate fatty acid to cysteine residues in the HV sequence that contributes to membrane association. Rab proteins also contain a C-terminal HV region that terminates with cysteine-containing motifs that are modified by addition of geranylgeranyl lipids, with some undergoing carboxylmethylation. Arf family proteins are characterized by an N-terminal extension involved in membrane interaction, with some cotranslationally modified by addition of a myristate fatty acid. Ran is not lipid modified but contains a C-terminal extension that is essential for function. Rho proteins are characterized by an up to 13 amino acid “Rho insert” sequence positioned between Ras residues 122 and 123 involved in effector regulation.
Ras superfamily small GTPases, together with their two key classes of regulatory proteins constitute a three-protein machinery that functions as cellular GDP-GTP-regulated binary switches (BOX 2). Alternation between the active GTP-bound and inactive GDP-bound states of the small GTPase is controlled by guanine nucleotide exchange factors (GEFs), which stimulate the exchange of GDP for GTP, and by GTPase activating proteins (GAPs), which terminate the active state by stimulating GTP hydrolysis3, 4. In their GTP-bound state, small GTPases bind effectors to activate biochemical processes. Typically, each small GTPase mediates its functions through association with multiple and functionally distinct effectors, whose selection may depend on the identity of the activating GEF. This may be achieved by each GEF causing a spatially-distinct distribution of GTPase activation and by the function of the GEF as a scaffold that facilitates effector activation. Thus, small GTPases act as signaling nodes, with multiple input signals converging on GEFs and GAPs and upon GTPase activation, which initiates multiple output signals (FIG. 1). The Rho and Rab families possess a third class of regulatory proteins, guanine nucleotide dissociation inhibitors, which will not be discussed in this review.
Ras superfamily proteins possess intrinsic guanine nucleotide exchange and GTP hydrolysis activities. However, these activities are too low to allow efficient and rapid cycling between their active GTP-bound and inactive GDP-bound states. GEFs and GAPs accelerate and regulate these intrinsic activities. Members of the different branches of the superfamily are regulated by GEFs and GAPs with structurally distinct catalytic domains3, 4, 149–152. Here we have utilized the Rho family as an example to illustrate the complexity of this process, where multiple GEFs and GAPs may regulate one specific GTPase. For the 20 human Rho GTPases there are 83 GEFs and 67 GAPs and a subset of Rho GTPases are not likely regulated by GEFs and GAPs (e.g., Rnd3/RhoE). Rho GTPases are activated by distinct RhoGEF families. Dbl family RhoGEFs (68) possesses a tandem Dbl homology (DH) catalytic and pleckstrin homology (PH) regulatory domain topology. DOCK family RhoGEFs (11) are characterized by two regions of high sequence conservation that are designated Dock-homology region regulatory DHR-1 and catalytic DHR-2 domains. Two other RhoGEFs have been described (SWAP70 and SLAT) contain a PH but no DH domain (2) and smgGDS (1) is an unusual GEF in that it functions as a GEF for some Rho as well as non-Rho family GTPases. At least 24 Dbl RhoGEFs have been reported to activate RhoA151. Rho (and Rab) GTPases are also controlled by a third class of regulatory proteins, Rho dissociation inhibitors (RhoGDI) (of which there are 3) whose main function involves regulation of Rho GTPase membrane association by masking the isoprenoid group.
The best validated connection between small GTPases and cancer comprise the three Ras proteins5. Mutational activation of Ras is found in 33% of human cancers (collated from COSMIC database)6. Consequently, intensive efforts have been made to identify pharmacologic approaches to block Ras function for cancer treatment. To date, no successful “anti-Ras” strategies have reached the clinic. The low micromolar binding affinity of protein kinases for ATP, where potent nanomolar affinity ATP-competitive inhibitors have been developed (imatinib for example), has been a very successful avenue for anti-cancer drug development7. In contrast, the low picomolar binding affinity of small GTPases for GTP and milimolar cellular concentrations of GTP renders a similar strategy for Ras implausible8. Thus, past and current efforts have focused on indirect approaches for disruption of Ras function: inhibition of components that regulate Ras membrane association9 and inhibition of downstream effector signaling10 (BOX 3).
Since the identification and development of small molecule inhibitors that directly target Ras have not been successful, a majority of past and ongoing efforts have targeted Ras indirectly, to modulate the functions of proteins that influence or mediate Ras oncogenesis. Shown here are proteins that regulate Ras posttranslational processing, either signaled through the C-terminal CAAX tetrapeptide motif (farnesyltransferase, FTase; Rac converting enzyme1; Rce1; Isoprenylcysteine carboxyl methyltransferase; Icmt) or by protein kinase C alpha (PKCα)-dependent phosphorylation or ubiquitination. Similar to Rho GTPases, Ras proteins are also regulated by multiple GEFs and GAPs. GTP-bound Ras interacts with catalytically-diverse downstream effectors that possess Ras-binding (RBD) or Ras-association (RA) domains. Although shown here as interactions with Ras, some are interactions restricted to specific Ras isoforms. Question mark indicates that more interacting partners are yet to be discovered.
Considerable past efforts centered on the development of FTase inhibitors (FTIs), with many identified, and with two remaining in clinical trial analyses (lonafarnib and tipifarnib). The prenylation of KRAS and NRAS by a related enzyme, geranylgeranyltransferase-I, when farnesyltransferase activity is blocked by treatment with an FTI, proved to be the downfall of FTIs as effective Ras inhibitors. A second class of inhibitor of Ras membrane association is comprised of two small molecules with farnesyl lipid groups (salirasib and TLN-4601) and proposed to compete with Ras for membrane-associated docking proteins for the Ras isoprenoid group. Efforts to target Ras effector signaling first centered on the Raf-MEK-ERK MAPK cascade. Small molecule protein kinase inhibitors of MEK1/2 and later Raf have been developed, with many now in clinical evaluation. More recently, inhibitors of the p110 catalytic subunits of PI3K, AKT and mTOR have entered clinical trials and two mTOR inhibitors have been FDA approved for renal cell cancers. Compiled from information at http://www.clinicaltrials.gov.
Beyond Ras, the aberrant function of an expanding roster of Ras superfamily proteins has been implicated in human cancer growth and development. However, whereas mutational activation of Ras is seen commonly in human cancers, direct mutation of other Ras superfamily GTPases is not seen frequently. Instead, the deregulated gene expression and/or deregulated protein function of GEFs and GAPs, in particular for specific Ras and Rho family proteins but also Arf11, 12, have been found to play important roles in cancer (supplementary information S1 and S2 (tables) lists the mechanisms and roles of GEF and GAP deregulation in human cancers). Genome-wide sequence analyses of breast, colon, pancreatic and brain cancers have now been completed13–16 and a search of the COSMIC database reveals isolated mutations in numerous GEFs and GAPs from sequence analyses of 173 regulators of Ras superfamily GTPases. However, whether these mutated genes are passengers or drivers of oncogenesis, whether they encode proteins with altered function, are not known for most of these situations.
In this review, we summarize representative studies in which aberrant GEF or GAP function is observed in cancer cells and where sufficient validation has been done to show causal roles of individual GEFs or GAPs in the aberrant growth properties of human cancer cells or in mouse models of cancer. We will focus primarily on Ras and Rho family GTPases and summarize the current evidence validating a causal role for their regulators in causing aberrant small GTPase function in human cancer or cancer-related processes. We also discuss the issues surrounding pharmacologic manipulation of GEF or GAP function. Our conventional targets and approaches for anti-cancer drug discovery have been hampered by tradition and past success. While it is still early days in target validation, and our current success in therapeutic targeting of these regulators is more proof-of-concept than clinical reality, we believe that GEFs and GAPs hold exciting prospects for cancer therapy.
The potential involvement of GEFs in cancer was first suggested by the isolation of RhoGEFs17,18–20 and later RasGEFs21–23 as transforming proteins in expression library functional screens using genomic DNA or mRNA derived from human cancer cells. However, the transforming RhoGEFs identified were activated by genomic deletion of coding sequences during the process of experimental manipulation rather than due to genetic events that occurred in the cancer cells24. Nevertheless, these observations supported their potential role as oncogenes in cancer development. Since GEF activation is the most common mechanism for signal-mediated GTPase activation, the theme that has emerged is that aberrant signaling from growth factor receptors, in particular, transmembrane receptor tyrosine kinases (RTKs) and G protein-coupled receptors (GPCRs), leading to aberrant GEF regulation, contributes to small GTPase activation in cancer. Another common mechanism of aberrant GEF activation is upregulated gene expression, and to a lesser degree, missense mutations and the consequent expression of catalytically-altered GEFs (supplementary information S1). While it is possible that there is upregulation of Ras or Rho GTPase activity by multiple GEFs simultaneously or inactivity of multiple GAPs, since there are many family members that could regulate the same GTPases, we present examples where the roles of individual GEFs or GAPs are clear.
RasGEFs activate Ras and additionally may also act as GEFs for the related Rap or R-Ras, but not Ral, subfamily members of the Ras family. The most common mechanism by which RasGEFs are involved in cancer involves their activation by growth factor-activated cell surface RTKs or GPCRs. This is best represented by the “classical” Ras signaling pathway, where activation of the epidermal growth factor receptor (EGFR) causes activation of wild type Ras through GRB2-mediated activation of the two son of sevenless (Sos1 and Sos2) RasGEFs. EGFR overexpression, mutational activation or hyperactivation by autocrine mechanisms are commonly seen in many cancers, leading to persistent Ras activation25. RTK and GPCR activation can also cause Ras activation through downstream activation of phospholipase Cγ (PLCγ) and PLCβ, respectively. PLC activation and diacylglycerol production directly activates the Ras guanyl releasing protein (RasGRP) subfamily of RasGEFs26. Mutationally activated Ras may also still require RasGEF activity, perhaps to activate wild type Ras isoforms concurrently27.
Germline gain-of-function mutation of SOS1 RasGEF has been observed in Noonan syndrome (13%), a developmental disorder also associated with increased risk of cancer28, 29. This implies that SOS1 could be an oncoprotein. However, an extensive sequence analysis of samples from 810 primary malignancies found only three SOS1 mutations and concluded that SOS1 mutational activation is not common in human cancers30. Hence, similar to other mutations found in developmental syndromes that activate K-Ras, Raf-1, and MEK1/2, the SOS1 mutations are weakly activating and may not be potent enough to cause cancer31. Mutations in other RasGEFs are also rare in cancer (Supplementary information S1 and COSMIC).
Finally, another association between RasGEFs and cancer involves their roles as downstream effectors of Ras (BOX 3). PLCε, a downstream effector of Ras32–34, contributes to mutant HRAS-mediated skin tumor formation; whether the RasGEF function is relevant for this role is not known. However, caution in interpreting these experiments is warranted, as followup studies found that PLCε loss reduced a stromal tissue inflammatory response and that isolated PLCε-deficient keratinocytes displayed no reduction in proliferative capacity35. Hence, whether PLCε loss caused reduced tumorigenesis in its role as a critical downstream Ras effector in cancer cells, or serves a tumor cell autonomous function, is unclear. Additionally, while there is evidence that PLCε can activate Ras, most evidence supports its role as a Rap activator36.
Other CDC25 domain-containing RasGEFs that are not activators of Ras and instead, are activators of the RalA and RalB small GTPases (also members of the Ras GTPase family) include RalGDS, Rgl2(Rlf), Rgl2 and Rgl337 (Supplementary FIG. 1) Mice deficient in RalGDS show impaired tumor formation in mutationally-activated HRAS-driven skin tumor formation38. Rgl2 overexpression was described in pancreatic tumors and cell lines and suppression of Rgl2 expression impaired tumor cell anchorage-independent growth and Matrigel invasion39. Moreover, Ral is activated in human tumors and promotes the growth of bladder, pancreatic, prostate and other cancers40–43. Ral GTPases function as GDP/GTP-regulated binary switches that are regulated by distinct GEFs and GAPs and activate distinct downstream effectors that regulate endocytosis, exocytosis and actin organization. Thus, targeting GTPase activation by GEFs or GTPase activation of GEF effectors are two potential applications of GEF inhibitors (Supplementary FIG. 2).
It is now clear that Rho GTPases play a major role in many different aspects of tumorigenesis44, 45. Most Rho GTPases promote tumorigenesis, and thus hyperactivation of their GEFs would likewise be oncogenic. However, there are examples, such as RhoB, that exert tumor suppressor properties, and thus activation of their GEFs would likewise be considered tumor suppressive. Unlike Ras, which is mutated in a large percentage of human cancers, mutations in Rho GTPases are rare. Instead, Rho GTPase hyperactivation occurs through overexpression, loss of GAP-mediated inactivation, and upstream activation (FIG. 2) or overexpression of the RhoGEFs. Below we highlight some examples, with others summarized in Table S1.
Vav RhoGEFs have been implicated in the growth of several cancers. First, the normally haematopoietic cell-specific VAV1 was overexpressed in pancreatic carcinoma cells as a consequence of promoter demethylation, leading to Rac activation and signalling46. VAV1 was activated by Src-dependent phosphorylation, which in turn was activated by EGFR, and led to activation of a Rac-Pak-NF-κB signalling pathway and cyclin D1 upregulation. RNAi depletion of VAV1 abrogated anchorage-independent growth in vitro and tumour growth in mouse xenografts. Moreover, VAV1 expression in pancreatic carcinomas was associated with decreased survival.
The related RhoGEF VAV2 is hyperactivated in head and neck squamous cell carcinoma (HNSCC) through an autocrine loop dependent on EGFR. Knockdown of VAV2 inhibited RAC1 activation and EGFR-stimulated invasion through Matrigel47. Another member, VAV3, was overexpressed at the mRNA and protein level in human glioblastomas compared to unmatched normal brain samples, and knockdown in cell lines decreased migration in vitro and in an ex vivo organotypic brain slice invasion assay48. Finally, Vav2−/−Vav3−/− double knockout mice had reduced xenograft tumor growth when transplanted with lung or melanoma cells, in part due to deficient angiogenesis, largely due to a defect in tumor-induced endothelial cell migration49. This suggests a role for RhoGEF signaling in the host microenvironment, highlighting the many ways in which RhoGEFs may affect tumorigenesis.
A Rac-specific GEF, phosphatidylinositol-3,4,5-trisphosphate-dependent Rac exchange factor 1 (PREX1), has been implicated in prostate cancer cell invasion50. The three Rac isoforms RAC1-3 are known to be important in many cancers by through a variety of ways, including stimulating migration and invasion through induction of lamellipodia as well as growth51. PREX1 gene and protein expression was highest in metastatic prostate cancer cell lines and protein expression was highest in metastatic prostate tumor tissue. Suppression of endogenous PREX1 expression in the PC-3 metastatic prostate cancer cell line inhibited Rac activity and reduced ligand-stimulated cell migration and invasion in vitro and ectopic PREX1 overexpression in PC-3 cell xenografts did not enhance tumourigenic growth but did promote metastasis to lymph nodes. In addition, PREX1 overexpression was associated with activation of ERK-MAPK signalling in melanomas52. Finally, a recent study identified the related PREX2 protein as a binding partner for the PTEN tumor suppressor53. PTEN is a lipid phosphatase that converts phosphatidylinositol-3,4,5-trisphosphate to phosphatidylinositol-4,5-disphosphate and thus antagonizes PI3K activity. PREX2 mRNA was overexpressed in PTEN wild type breast cancers and RNAi depletion reduced the levels of activated AKT and impaired the growth of PTEN wild type tumours. Taken together, these studies with Rac-selective GEFs underscore their importance in migration, invasion and metastasis.
ECT2, an activator primarily of RhoA, but also of Rac and Cdc4254, 55, mRNA or protein has been found to be overexpressed in a variety of human tumor cell lines and tissues, including lung and esophageal squamous cell carcinomas48, 56–59, and correlated with poor prognosis57, 58. ECT2 overexpression at the mRNA and protein level was found in patient glioblastoma samples compared to non-matched normal brains and RNAi-mediated suppression of ECT2 expression in glioblastoma cells reduced migration and growth rates in vitro and invasion in an ex vivo organotypic rat brain slice model48. Finally, a recent study found ECT2 mRNA and protein overexpression in non-small cell lung carcinomas (NSCLCs)60. ECT2 expression was mislocalized to the cytoplasm and was associated with Rac, and surprisingly not RhoA, activation. RNAi-mediated knockdown of ECT2 blocked the anchorage-independent growth and Matrigel invasion in vitro, and tumor xenograft growth in vivo, of NSCLC cell lines.
Three different RhoGEFs are structurally-mutated in human cancers by chromosome rearrangement and formation of chimeric fusion proteins. One involves the ARHGEF12 (also known as LARG) RhoA-specific GEF, which was identified initially in tumor cells from a patient with acute myelogenous leukemia61. The rearrangement encodes a MLL-ARHGEF12 fusion protein that retains the DH-PH catalytic domains of ARHGEF12. Whether the fusion protein represents a constitutively activated variant of ARHGEF12 has not been determined. LARG and the related PDZ domain-containing RhoGEFs (p115-RhoGEF and PDX-RhoGEF) may also be activated by GPCRs that are coupled to Gα12/13 or by Gα12/13 overexpression (FIG. 2)62, 63. The second example is the BCR-ABL1 fusion protein encoded by the translocation associated with the Philadelphia chromosome found in 90% of chronic myelogenous leukemias. BCR possesses a RhoGEF and a RhoGAP domain and ABL1 is a protein tyrosine kinase. In the resulting BCR-ABL1 chimera, the RhoGEF but not RhoGAP domain is retained and fused to a truncated ABL1, resulting in constitutive activation of the kinase activity critical for BCR-ABL1-mediated oncogenesis. BCR-ABL1 transforming activity, as measured by anchorage-independent growth, is also dependent, in part, on the RhoGEF activity, which results in activation of RhoA64. Finally, a third RhoGEF, TRIO, is activated in adult T-cell leukaemias by alternative splicing, which results in a truncated protein with the second catalytic DH domain attached to a unique 15-residue peptide (designated TGAT)65. The TGAT transcript was detected in peripheral blood mononuclear cells of 14 of 21 T-cell leukaemia patients, but not in four control subjects. Ectopic expression of TGAT caused tumorigenic transformation of NIH 3T3 mouse fibroblasts, although no evidence for T-cell leukemia growth was determined.
TIAM1, a Rac-specific GEF, is associated with a variety of cancer types. First, it can function as a downstream effector of Ras66. Tiam1−/− mice had impaired carcinogen-induced HRAS activation and squamous cell skin carcinoma formation, including fewer tumors and smaller tumor size, although the tumors that did form metastasized more readily67. Second, mouse models of APC-induced colon cancer and ERBB2 (also known as Neu)-induced mammary cancer also show impaired tumor formation in the absence of Tiam1, although in the case of the mammary cancer model the tumors were more invasive68, 69. Third, there are several reports of altered TIAM1 (mutation and overexpression) in various human cancers (supplementary information S1). Importantly, although TIAM1 may be important in tumor initiation, the increased malignancy and invasion in the skin and mammary models upon loss of TIAM1 and the observation that TIAM1 protein expression is lower during breast cancer progression70 suggests that TIAM1 may act as a metastasis suppressor, and thus inhibiting its activity in some settings may not be beneficial.
In addition to the DH-PH family of RhoGEFs, there is evidence for the aberrant function of DOCK family RhoGEFs in cancer. This family is comprised of 11 members in humans and possess a structurally-distinct RhoGEF catalytic domain71. Interestingly, to date, DOCK family proteins activate Rac or Cdc42 but not RhoA72, although the structure of the RhoGEF catalytic DHR-2 domain bound to Cdc42 suggests that they may activate a broader spectrum of Rho GTPases73.
DOCK1 (also known as DOCK180) is a RacGEF and its overexpression together with its activator, ELMO, was found to promote glioblastoma cell invasion in vitro and in vivo74. DOCK RhoGEFs are also implicated in distinct facets of melanoma cell migration. DOCK10, a Cdc42GEF, was identified as a key regulator of protease-independent amoeboid melanoma cell migration75. In contrast, protease-dependent mesenchymal-type movement was driven by DOCK3, a RacGEF76. These results suggest distinct roles for Cdc42 and Rac in promotion of tumor cell migration.
ARF1 and ARF6, the most studied isoforms of the Arf GTPase subfamily, are active regulators of proliferative and/or invasive properties of cancer cells, notably in melanoma and breast cancer cell lines77, 78, and have also been linked to resistance to apoptosis79. Their functions in invasion may stem from their role at the crossroad between membrane trafficking and Rho GTPase-controlled actin remodelling, notably in the formation of invadopodia12, 80. Several subfamilies of ArfGEFs have recently emerged as candidate regulators that support invasion of cancer cells. GEP100 (also known as BRAG2), a GEF for Arf6, has been implicated in breast cancer invasion81. GEP100 was overexpressed in primary ductal breast carcinomas, commonly with EGFR overexpression, with overexpression correlating with higher grade tumours81. RNAi knockdown of GEP100, but not other ArfGEFs, reduced breast cancer cell invasion through Matrigel in vitro and reduced metastasis to the lung in a mouse model of breast cancer82. The expression of EFA6, an Arf6 GEF, is increased glioma tissue samples, and its expression in a human gliobastoma cell line enhanced ERK-dependent invasion83. Overexpression of ARF6 has been reported in highly invasive breast cancer cell lines82. This may result from loss of expression of FBX8, an unconventional ArfGEF that mediates the ubiquitination of Arf6 and suppresses its activity84
GAPs are the flip-side of the coin to GEFs, and although less is known about them in general, many studies have demonstrated their crucial roles in curtailing GTPase activity in cancer. Since activation of GEFs for the Ras superfamily GTPases has many roles in cancer, it is perhaps not surprising that loss of GAP activity allows uncontrolled GTPase activity and can promote cancer. We discuss some pertinent examples that demonstrate their importance and ways in which their activities are regulated (Supplementary information S2). Even though in general GAPs are tumor supressors, there are also examples of oncogenic GAPs.
The mutants of Ras are missense mutations (primarily at residues 12, 13, or 61) that impair the intrinsic and GAP-stimulated ability to hydrolyse GTP, rendering Ras constitutively GTP-bound and active in the absence of extracellular stimuli. Although the intrinsic activity of the GAPs is not altered in these cancers, the fact that they can no longer deactivate Ras indirectly implicates them in the oncogenic process. Although the RasGAPs in this case would not be considered drivers of this process, one of the earliest unsuccessful efforts made to develop anti-Ras drugs was to develop small molecules that restored GAP sensitivity to mutant Ras.
Germline mutational loss of the NF1 tumor suppressor, which encodes the RasGAP neurofibromin, is found in patients with neurofibromatosis type 1 (NF1)85, 86. Two recent sequencing studies established frequent somatic mutation of NF1 in glioblastoma (15–23%), representing the fifth most frequently mutated gene in this cancer15, 87. Although some of the mutations are null mutations or truncations resulting in loss of RasGAP catalytic function, consistent with its role as a tumor suppressor, the function of the several point mutations found remains to be determined. Post-translational loss of neurofibromin - induced by protein kinase C-mediated proteasomal degradation – has also been observed in sporadic glioblastomas88. Since the only known catalytic function of neurofibromin is its RasGAP activity, the functional consequences of neurofibromin loss is attributed to the observed hyperperactivation of wild type Ras. However, since the RasGAP domain comprise but a small portion of the total protein, non-Ras functions associated have been speculated.
The loss of other RasGAPs, Ras homolog enriched in brain (RHEB1) and RHEB (also known as RHEB2)89, is associated with tuberous sclerosis complex (TSC), which is a syndrome characterized by the formation of tumor-like lesions, hamartomas, in kidney, lung, brain and skin90. This autosomal dominant disease is caused by germline and somatic mutational loss of either TSC1 (harmartin) or TSC2 (tuberin). Tuberin contains the RhebGAP catalytic domain whereas harmartin stabilizes tuberin and prevents its degradation; hence, the harmartin:tuberin complex is required for RhebGAP activity91. Although the tumor phenotype is distinct, the increased incidence of renal cell and other cancers in Eker rats, which contain germline heterozygous Tsc2 mutations that inactivate the RhebGAP activity, supports the role of TSC2 as a tumor suppressor92–94. Loss of TSC1/2 RhebGAP function results in Rheb activation and persistent activation of its downstream effector, mTOR. The functions of the harmartin-tuberin complex as a RhebGAP are also regulated by phosphorylation, in particular AKT phosphorylates and thereby inactivates tuberin. Thus, genetic and biochemical activation of the PI3K signaling pathway (e.g., PIK3CA gain-of-function or PTEN loss-of-function mutations) in cancer cells can also cause Rheb-mediated activation of the rapamycin-sensitive mTOR complex 1 (mTORC1). mTOR regulates mRNA translation and ribosome biogenesis, regulating cell cycle progression, cellular proliferation and growth, autophagy and angiogenesis.
One RhoGAP in particular has stood out recently as having a central role as a tumor suppressor in several different cancer types: deleted in liver cancer 1 (DLC1, also known as ARHGAP7)95, 96. DLC1 was first discovered as a gene which is under-represented in a human hepatocellular carcinoma (HCC) specimen and is deleted in HCC cell lines and tumors97. Subsequent studies found DLC1 was deleted or transcriptionally silenced by promoter methylation in many cancer types (Supplementary Table S2). A comprehensive analysis of the genomic loss of DLC1 showed that heterozygous loss in tumors happens at a rate that approaches that of TP53 (which encodes p53) mutation or loss in breast, lung, liver, colon and pancreatic tumors98. Additional studies identified two genes that encode DLC1 related proteins, DLC2 (also known as STARD13)99, 100 and DLC3 (also known as STARD8)101, and the expression of both genes lost in a variety of human cancers, although it is unknown whether they are lost separately from or concurrently with DLC1. Finally, protein-protein interactions with 14-3-3 isoforms and another GAP, p120RasGAP, may also cause loss of DLC1 function102, 103.
Together with loss of expression in cancer, genetic and biochemical analyses in cell culture and mice provide functional evidence for DLC1 as a tumor suppressor. Ectopic re-expression of DLC1 in DLC1-deficient human tumor cell lines suppressed proliferation, anchorage-independent growth, invasion through Matrigel and tumor formation in xenograft mouse models of a variety of cancer types104–107 and re-expression in breast cancer cell lines reduced metastasis in a mammary fat pad orthotopic injection model108. In an ex vivo mouse model of Myc-induced tumorigenesis, knockdown of endogenous DLC1 accelerated the onset of tumorigenesis and resulted in more aggressive tumors that resembled aggressive human HCC, providing strong evidence for the role of DLC1 as a tumor suppressor98. Similarly, ectopic expression of DLC2 or DLC3 in expression-deficient human tumor cell lines caused impairment in tumor cell growth99, 101.
Although DLC proteins are multi-domain proteins comprised of sterile alpha motif (SAM), RhoGAP and StAR-related lipid transfer (START) domains, evidence supports the crucial role of the RhoGAP domain in DLC1 tumor suppression. The substrates of DLC1 are RhoA, RhoB, RhoC, and to a lesser degree CDC42, but not Rac107 and cell-based studies suggest that RhoA activation is a major consequence of DLC1 loss of function. In the ex vivo mouse model of Myc-induced liver tumorigenesis, activated RHOA phenocopied loss of DLC98. Because of the high frequency of reduced DLC expression in many different types of tumors and the functional evidence that DLC1-3 are tumour suppressors, inactivation of which primarily inactivates Rho GTPases, alterations in this RhoGAP protein family represent the most common mechanism of altering Rho GTPase activity in human cancer.
Surprisingly, in contrast to the many RhoGEFs that are altered in cancer, aside from the DLC family, there is limited evidence for the role of other RhoGAPs in cancer. However, putative tumor suppressors, such as GRAF, ARHGAP25, ARHGAP5 and ARHGAP8 may exist (Supplementary Table S2), but more work is required to validate these and to determine whether their RhoGAP activity is crucial. It could be that in many cancers, GAP activity is normal, but that the excessive activation through GEFs or GTPase overexpression overrides normal GAP-mediated inactivation.
Two subfamilies of ArfGAPS, AGAPs and ASAPs, have been implicated in oncogenesis11, 109, although whether this is through their GAP activity towards Arf GTPases is not yet established. AGAP2 (also called PIKE, Centaurin γ1 or GGAP2) is amplified and overexpressed in glioblastoma, prostate carcinoma and other cancers79, 110, 111. Cancer cells with AGAP2 overexpression resist apoptosis more strongly than those with normal levels, and ectopic expression of AGAP2 activates the AKT pathway and inhibits apoptosis in human glioblastoma cells, suggesting that the oncogenic properties of AGAP2 are achieved through the AKT pathway79, 110–112, AGAP2 is a multi-domain protein, which includes a domain remotely related to small GTPases113 in addition to its ArfGAP domain. Whether these domains and/or the GAP domain are involved in the oncogenic effect remain unclear.
ASAP1 (also called AMAP1, DDEF1 or Centaurin β4) overexpression is associated with invasive phenotypes in melanoma, prostate cancer and breast cancer cells114–116. ASAP1 has been best studied in breast cancer cells, where it co-localizes with Arf6 to invadopodia, and it is associated with proteins involved in actin remodeling116. A peptide derived from the C-terminal SH3 domain [G] of ASAP1 was able to block breast cancer cell invasion and metastasis117. A related ArfGAP, ASAP3 (also called UPCL1, DDEFL1 or ACAP4), was identified by its up-regulation in hepatocellular carcinomas118 and is involved in migration and invasion in a mammary carcinoma cell line, although it is not involved in invadopodia formation its localization is very different than that of ASAP, thus the two likely play nonredundant roles119.
In summary, as with GEFs, there are a diversity of genetic and biochemical mechanisms by which GAP function, most commonly as tumor suppressors, is deregulated in cancer. However, to date, despite the large numbers of GAPs for Ras and Rho GTPases, those that have been implicated in cancer remain limited. Perhaps this reflects the fact that there has traditionally been a greater focus on GEFs or perhaps there is greater functional redundancy with GAPs, making it unlikely that loss of function of any one GAP will be sufficient to cause significant deregulation of GTPase activity.
As with most proteins propagating information by intracellular protein-protein interactions, with large contact surfaces that lack the grooves and pockets for small molecule interactions, GEFs and GAPs are not classically considered as “druggable” targets120. However, it is important to remember that the development of ATP-competitive inhibitors of protein kinases, which were once considered undruggable, now comprise the major class of clinically-useful signal transduction anti-cancer drugs. Hence, druggability is defined primarily on current success and not a static concept. Instead, the strength of target validation, rather than conventional wisdom, should prioritize efforts to establish target druggability.
With increasing evidence for aberrant GEF or GTPase activity in cancer, a logical issue is whether these regulatory proteins are attractive targets for anti-cancer drug discovery, particularly those GEFs that exhibit gain-of-function mutations or are overexpressed. Additionally, GEF activation defines where and when a GTPase is activated and probably what the downstream events are, and are thus likely to convey high signaling specificity. This may limit off-target effects when inhibited. The structures of representative GTPase-GEF complexes have been determined121–123: all feature a very large protein-protein interface resulting from the structural remodeling of the small GTPase upon binding. The shape, structural dynamics and chemistry of GEF-GTPase interaction surfaces are thus very different from those of catalytic sites of enzymes, such as the ATP-binding site of signaling kinases, and may therefore appear inappropriate for small molecule binding. However, despite this perception, below we summarize experimental evidence indicates that it may be feasible to develop small molecule inhibitors of GEFs.
Brefeldin A (BFA) is a natural product isolated from the fungus Eupenicillium brefeldianum and is the first known inhibitor of a GEF. BFA was discovered in the late 1950’s124 and some 30 years later demonstrated to inhibit trafficking at the Golgi network by blocking the activation of Arf GTPases by Sec7 domain containing ArfGEFs, specifically Arf1 and Arf5125, 126. The molecular basis for this activity took another decade to be resolved by a combination of yeast genetics, biochemistry and structural biology127–129. BFA targets the complex between Arf-GDP and the catalytic domain of the ArfGEF (the Sec7 domain) at the beginning of the exchange reaction and freezes the complex in an abortive conformation that cannot proceed to nucleotide exchange (FIG. 3)127, 130, 131. Despite a modest apparent inhibition constant of 15 μM, and a stabilization of the Arf-GDP-Sec7 complex by only a factor of 10, BFA is remarkably efficient in live cells due to the nature of its inhibition mechanism. The inhibitor contact both Arf-GDP and the ArfGEF in the abortive complex, k which allows it to have a restricted specificity for a subset of both ArfGEFs and Arf proteins. On the ArfGEF side, BFA-sensitivity depends on a small number of residues in the BFA-binding site that differ, either alone or combined, between BFA-sensitive and BFA-insensitive ArfGEFs. A remarkable property of BFA is that is also discriminates between Arf1-GDP and Arf6-GDP, the major cellular Arf isoforms, although the two proteins have the same sequence in the BFA-binding site - yet probably not the same structure and/or structural dynamics. BFA has also demonstrated a number of anti-cancer effects in cells, which in light of these mechanistic studies, are thus likely to result from its impairment of ArfGEF functions.
The extensive analysis of the mechanism of action of BFA led to the general concept of ’interfacial inhibition’, which refers to inhibitors that act by stabilization of protein complexes and target regions in or near interfaces128 (FIG. 3b). Some inhibitors of natural original origin already used in the clinic have been recognized as interfacial inhibitors, such as the anti-cancer drugs vinblastin or camptothecin, suggesting a novel avenue to therapeutic intervention that has started to be explored132.
LM11 was discovered by an in silico screen based on this concept, and was shown to target an interfacial depression at the surface of the complex between Arf1-GDP and BFA-insensitive GEFs such as ARNO and to block ARNO-dependent cellular migration131. A few other promising examples of cell-active small molecule ArfGEF inhibitors have been selected by in vitro133, 134 and phenotypic screens135. These studies demonstrate that despite the high homologies that are found within a given GEF family, GEF-specific inhibitors can be developed. Therefore, the specific flexibility and conformational changes that characterize small GTPase-GEF complexes are likely to be advantageous to drug development, notably for interfacial inhibitors. However, the design of high throughput biochemical assays to screen effectively for such inhibitors remains a challenge132.
There has also been a recent increase in discovery of inhibitors of Rho GTPase activation. Inhibitors that target specific RhoGEFs have been discovered by high throughput screens. The first example was an aptamer screen, in which peptides coupled to thioredoxin were selected in yeast for their binding to the catalytic DH2 domain of TRIO136. This identified a potent inhibitor of TRIO, which was subsequently optimized to inhibit its oncogenic splice variant TGAT137. The corresponding optimized peptide was active in cells in vitro and in reducing TGAT-induced tumour formation in nude mice xenograft models. Another assay screened a small chemical compound library by monitoring the interaction of the GTPase with an effector in the presence of a co-expressed GEF138. This ’yeast 3-hybrid assay’ identified several inhibitors of RhoG activation by TRIO. One of these, ITX3, was specific and active in cell-based assays139. Screening using a fluorescence polarization guanine nucleotide-binding assay also identified small molecule inhibitors of ARHGEF12 (LARG)-stimulated RhoA nucleotide binding in vitro140. Although the inhibitors and aptamers discovered in these screens were of low potency, they support the potential for identifying GEF-targeted inhibitors.
Another related example of a way to target GTPase activity is through targeting the surface of GTPases that is required for GEF activation. Through computational screening of the surface of Rac1 known to interact with GEFs, the small molecule NSC23766 was discovered, which inhibited activation of Rac1 by the Rac-specific GEFs Trio and Tiam1, but not GEF activation of RhoA or Cdc42 in vitro and in cells141. Using a similar strategy, and utilizing structural information from NSC23766 in complex with Rac1, five additional small molecules structurally unrelated to NSC23766 were discovered that could specifically block Rac activation by GEFs142. These molecules do not directly target GEFs, and are likely to lack GEF specificity since they would block the surface of GTPases and thus activation by a variety of GEFs. They could nonetheless provide an interesting approach to block GEF activation of Rho or other small GTPases important in cancer.
RasGAPs stimulate the intrinsic GTPase activity of Ras by up to 105-fold, but have virtually no effect oncogenic Ras mutants143. Therefore, one strategy has been to identify small molecules that restore the ability of RasGAPs to work on mutant Ras. However, despite great effort, this was unsuccessful, likely because oncogenic mutations disturb the active site of Ras, preventing the proper transition state that is needed for GAP-mediated hydrolysis144. Thus, even if the GAP activity of RasGAPs was increased by small molecules, Ras will likely still be refractory to the higher activity. The involvement of GAPs in cancer is most commonly associated with loss-of-function and hence they exhibit properties of tumor suppressors, although as listed in Supplementary Table S2, some GAPs may have oncogenic properties and could thus be drug targets. Since it is traditionally easier to develop small molecule antagonists rather than agonists, GAPs are less attractive targets. Instead, since loss of GAP function leads to GTPase activation, most efforts are focused on blocking the persistent GTPase effector signaling that occurs.
There is limited but promising evidence that small molecule modulators of Ras superfamily GAPs may be possible to develop. High throughput screening identified small molecule inhibitors of RGS domains, which are GAPs for heterotrimeric G proteins145. Despite their low structural homology to RasGAPs, they share a similar enzymatic transition state144, suggesting that this could be a starting point for the design of Ras superfamily GAP inhibitors. One class of RhoGAPs, the Rac-selective chimaerins (CHN) possess C1 zinc finger domains that bind diacyglycerol, a cofactor for their activity26. Therefore, small molecules that bind C1 domains may activate their GAP activities, causing downregulation of Rac GTPase activity26. While such a therapeutic approach will be complicated by the existence of other proteins with C1 domains (e.g., RasGRP), there is evidence that C1 binding molecules can have some degree of selectivity for a subset of C1-containing proteins. This approach may be a therapeutic option for cancer where there is RacGEF-mediated activation of Rac.
We have highlighted key evidence for the role of aberrant expression and function of GEFs and GAPs of Ras superfamily small GTPases in cancer, with an emphasis on the two key early steps in cancer drug discovery, target validation and druggability. With the continued application of genome-wide analyses of cancer cells, additional correlative evidence for aberrant GEF and GAP expression and function is expected to continue at a rapid rate although validation of their functional importance in cancer will be a rate-limiting factor. Even the current body of experimental evidence validating GEFs and GAPs will require more rigorous validation. While RNAi-based analyses have contributed critical validation, the multi-domain and multi-functional nature of GEFs and GAPs emphasizes that caution must be exercised in simply concluding that any phenotypic alterations are due solely to their roles in regulating small GTPase GDP-GTP cycling. For example, RalGDS can activate AKT independent of its RalGEF function146. Is the impaired HRAS-driven skin tumor formation due to ablation of RalGDS expression due to loss of Ral or AKT activation, or both? Rescue experiments with carefully designed GEF or GAP domain-impaired mutants are needed to access possible GEF/GAP-independent functions of these regulatory proteins.
Furthermore, while mouse model analyses where a deficiency in a GEF or GAP is achieved at the onset of tumor formation provide important validation, these studies validate more the preventative value rather than the therapeutic value with a pre-existing tumor. For example, TIAM1 was shown to be necessary for initial growth of HRAS-induced skin tumors, but mice lacking it had more aggressive tumors when they did arise67. Finally, genetic ablation of a target is not equivalent to pharmacologic inhibition of a target. This is demonstrated dramatically with studies that showed that preventing Ras binding to PI3K but not pharmacologic inhibition of PI3K was effective in preventing mutant KRAS-induced lung tumor formation147.
Regarding druggability, it is still very early days in this phase of drug discovery, with the current body of evidence more proof-of-concept and less one of identifying promising leads for clinical evaluation. The lessons learned from BFA currently provide the best evidence for the tractability of GEFs. Perhaps chemical libraries based on such natural products will be a more fruitful direction than the traditional use of libraries based on chemical structures based on past success with enzymes and GPCRs. As the processes and paradigms of drug discovery continue to evolve, so will the definition of druggability. With advances in the use of structural information in virtual screening, structure-based design, fragment-based library screening, coupled with functional screens focused on protein complexes rather than isolated proteins, perhaps GEFs and GAPs can be rendered druggable. That protein kinases, currently the “low hanging fruit” of anti-cancer drug discovery, may serve as key regulators of GEFs and GAPs and their downstream signalling pathways, suggests that more conventional directions for GEF and GAP drug discovery are also promising directions (BOX 4).
In contrast to Ras, the specific downstream effectors that mediate the cancer cell phenotype, proliferation and survival, invasion and metastasis of other Ras and Rho family GTPases remain poorly understood. In this figure, we highlight protein kinases as effectors or regulators of Ras and Rho family GTPase oncogenesis. First, analogous to the role of Raf in Ras function, protein kinases have been implicated as downstream effectors of GTPase-mediated oncogenesis. In particular, there is evidence that the ROCK153–155, MRCK156, PAK157–160 and ACK161, 162 protein kinase effectors can promote oncogenesis. Much of the evidence for ROCK involvement in cancer is based on studies with ROCK inhibitors163. However, since these inhibitors have considerable off-target activities, it is unclear if ROCK inhibition alone accounts for the anti-tumor activities of ROCK inhibitors. There is emerging evidence that protein phosphorylation is an important mechanism for regulation of small GTPase function, often by controlling subcellular localization and interaction with other proteins. PKCα phosphorylation causes K-Ras4B translocation from the plasma membrane to the mitochondria, where K-Ras4B association with Bad results in apoptosis164, suggesting that agonists of PKCα may act as K-Ras-directed therapies. Similarly, Aurora-A phosphorylation of RalA is essential for RalA promotion of pancreatic cancer cell line tumorigenic growth165. Additional effectors of Rho GTPases that regulate actin organization (e.g., mDIA) may influence cell motility, and hence, be important mediators of Rho GTPase induction of tumor cell invasion and metastasis166. A second theme is the signaling crosstalk that can occur between different members of the Ras and Rho families. For example, the RalBP1/RLIP76 effector of Ral functions as a RhoGAP for Rac and Cdc42 inactivation associated with transformation165. Ras activation of mTOR can involve AKT activation, leading to inactivation of Tsc2, causing Rheb activation.
Supplementary Figure S1. CDC25 homology domain-containing Ras family GEFs There are up to thirty CDC25 homology domain-containing proteins encoded in the human genome177. The majority also possess an upstream REM domain. The RasGEFs typically activate all three Ras isoforms. In addition to Ras, some are activators of Rap, R-Ras and Ral members of the Ras family149, 178. Some are promiscuous and activate multiple Ras family proteins whereas some are very selective and activate only a subset of Ras family proteins (e.g., the RalGEFs). The strength and extent of data defining the substrate specificity is highly variable, with some poorly characterized and some with conflicting observations in the literature. The complete domain topology is not shown (e.g., REM), with only key regulatory domains/motifs included: C1, Protein kinase C conserved region 1 (C1) domains (also called cysteine-rich domain); CDC25, CDC25 homology RasGEF domain; cNMP, cyclic nucleotide-monophosphate binding domain; EF, EF-hand, calcium binding motif; DH, Dbl homology RhoGEF domain; PDZ, Domain present in PSD-95, Dlg, and ZO-1/2; PH, pleckstrin homology domain; RA, Ras-association; SH2, Src homology 2 domain. An arrow from a GEF to a GTPase indicates GEF activation of the GTPase. In contrast, an arrow from the GTPase to a GEF indicates that the GEF acts as an effector of the active GTPase.
Supplementary Figure S2. RAS interaction with GEF effectors. There are two ways small GTPases have been reported to interact with GEFs, either as substrates of GEFs for the GDP/GTP exchange reaction (pictured in the bottom structures), or with GEFs acting as effectors of their GTP-bound form. This dual situation is illustrated here. The top views depict modeled structures of Ras-GTP in complex with the RA domain of RalGDS, a GEF for Ral, and with the RBD domain of Tiam1, a Rac1 GEF. These interactions may regulate effector function, in part, by release of autoinhibitory interactions in the GEFs and/or promotion of their association with the plasma membrane, and hence stimulate the activation of Ral by the CDC25 domain of RalGDS, or Rac by the DH-PH domains of Tiam1. These pathways are used by Ras to promote tumorigenesis, which could be explored as novel sites of pharmacological intervention. For illustration, the structures of RAS•RBD (top right) and RAL•RALGDS (bottom left) are represented by the related structures of RAP1•RAF (PBD id. 1CY1) and RAS•SOS (PBD id. 1BKD), respectively. The determined structures of RAS•RALGDS RA domain and TIAM1•RAC1 are PDB id. 1LFD and 1FOE, respectively. These pathways are used by Ras to promote tumorigenesis, which could be explored as novel sites of pharmacological intervention.
The authors’ research on GEFs and GAPs is supported in part by grants number CA042978, CA129610, CA127152, CA67771 and CA106991 from the US National Institutes of Health (C.J.D), from the American Cancer Society (D.V. and K.L.R.), and from the Centre National de la Recherche Scientifique (J.C.), and by grants from the Association pour la Recherche Contre le Cancer (J.C.) and the Agence National de la Recherche (J.C.). We thank L. DeGraffenreid for outstanding assistance in the preparation of this manuscript.
Dominico Vigil completed his Ph.D. at the University of California, San Diego, studying protein kinase A structure, biochemistry and regulation. He is currently a postdoctoral research fellow at at the Lineberger Comprehensive Cancer Center at the University of North Carolina at Chapel Hill. His research has focused on GEFs and GAPs involved in cancer.
Jacqueline Cherfils is Head of the Enzymology and Structural Biochemistry Laboratory, Centre National de la Recherche Scientifique (CNRS), Gif-sur-Yvette, France. Her main research interest is structural biology of small GTP-binding proteins and their activation by GEFs, and the discovery of small chemicals that inhibit members of these proteins families involved in human diseases.
Kent Rossman completed his Ph.D. and postdoctoral research at the University of North Carolina at Chapel Hill. He is currently a Research Assistant Professor of Pharmacology at the Lineberger Comprehensive Cancer Center at the University of North Carolina at Chapel Hill. His research has focused on the biochemistry, regulation and structure of Dbl family Rho GEFs.
Channing J. Der
Channing J. Der is Sarah G. Kenan Professor of Pharmacology at the Lineberger Comprehensive Cancer Center at the University of North Carolina at Chapel Hill. His research has focused on dissection of the signaling mechanisms of Ras superfamily small GTPases involved in oncogenesis and on molecularly-targeted therapies based on these mechanisms for cancer treatment.
Links to websites:
The Cancer Genome Project: http://www.sanger.ac.uk/genetics/CGP/
Channing Der’s homepage: http://cancer.med.unc.edu/derlab/
Competing interests statement
The authors declare no competing financial interests.