Deregulation of Rho GTPases in cancer

ABSTRACT In vitro and in vivo studies and evidence from human tumors have long implicated Rho GTPase signaling in the formation and dissemination of a range of cancers. Recently next generation sequencing has identified direct mutations of Rho GTPases in human cancers. Moreover, the effects of ablating genes encoding Rho GTPases and their regulators in mouse models, or through pharmacological inhibition, strongly suggests that targeting Rho GTPase signaling could constitute an effective treatment. In this review we will explore the various ways in which Rho signaling can be deregulated in human cancers.


Introduction
Rho GTPases bind to a wide range of effector proteins and play central roles in the regulation of the actin and microtubule cytoskeletons and gene transcription. 1 Through these effects, Rho family proteins influence many normal cellular functions such as adhesion, polarity, motility and invasion, as well as cell cycle progression and survival. 2,3 Rho, Rac and Cdc42 were initially characterized as regulators of the actin cytoskeleton 1 with a typical pattern being Rho activation leading to the formation of contractile actin, Rac activation controlling peripheral actin structures such as lamellipodia and membrane ruffles, and Cdc42 actin structures such as filopodia. 1 However, it has long been clear that these proteins have roles far beyond direct regulation of the actin cytoskeleton. For instance, Cdc42 is a master regulator of polarity in organisms from yeast to mammals, while Rac regulates phagocytosis in the immune system, including production of reactive oxygen species. 1,4 They are involved in many essential physiological processes including embryonic development, neuronal differentiation and neurite formation and maintenance of stem cells in the bone marrow, skin and intestine. 2,3,5 Conversely, deregulation of Rho GTPases is linked to many of the "hallmarks of cancer," including oncogenic transformation, cell survival and tumor metabolism as well as metastasis (reviewed in ref. 2). While some consequences of deregulated Rho family signaling can be considered pro-tumorigenic, a number of cellular processes stimulated by Rho family proteinssuch as the role of Rac1 in apoptosis and maintenance of apicobasal polarity-can be considered to antagonize tumor formation and progression. 6 The anti-tumorigenic effects of Rho family proteins must be sufficiently differentiated from those pro-oncogenic functions to avoid undermining the therapeutic benefits to be achieved by pharmacologically antagonizing Rho GTPases.

The Rho GTPase cycle
Rho GTPases are molecular switches which cycle between an inactive GDP-bound form and an active GTP-bound form (see Fig. 1). The GTPase cycle is largely regulated by guanine nucleotide exchange factors (GEFs) and GTPase-activating proteins (GAPs). 7 GEFs displace the GDP bound in the active site allowing GTP binding. GTP binding alters the conformation of the GTPase, allowing it to interact with downstream effector molecules (Fig. 1). 7 GEFs have also been thought to contribute to signaling specificity through scaffolding upstream and downstream interactors; 8 this was recently demonstrated with the GEFs Tiam1 and P-Rex1 driving different behaviors via the same small GTPase, Rac1. 9 Conversely, GAPs activate the weak intrinsic GTPase activity of Rho proteins leading to the hydrolysis of bound GTP, switching the GTPase to an inactive conformation (Fig. 1). 7 The abundance of GEFs (at least 80) and GAPs (over 70) indicates the importance of tightly controlling Rho GTPase signaling.
Guanine nucleotide dissociation inhibitors (GDIs) are a third class of regulators of Rho proteins. They sequester inactive GTPases in the cytoplasm by masking their Cterminal lipid moieties that mediate plasma membrane localization, which can inhibit their activation 7 (Fig. 1). They can also protect GTPases from degradation 10 and also have more subtle effects, such as directing activation of Rho GTPases to specific membrane compartments. 11 Rho GTPases are also known to be modulated by a host of post-translational modifications, including phosphorylation, ubiquitylation, SUMOylation, ADP-ribosylation, glycosylation, adenylation, and transglutamination/deamidation. Given the wide variety of these modifications, detailed analysis is outside the scope of this review; for more details see refs. 12-14. At the simplest conceptual level, anything which increases the abundance of the active form should increase signaling, while anything decreasing the abundance of the active state, or actively stabilizing the inactive state, should decrease signaling. Disruption of this balance-by direct activation of Rho GTPases or indirectly through changes in regulators as described above-is increasingly being linked to oncogenesis (see Fig. 2). In this review we will focus on the variety of ways in which Rho signaling has been shown to be disrupted in cancer: alterations in protein levels of the GTPases, disruption to regulators of GTPases, changes in post-translational modifications of GTPases and Figure 1. The Rho GTPase cycle GTPase regulation occurs in a number of distinct stages. Guanine nucleotide exchange factors (GEFs) are able to bind to inactive GTPases, displacing the bound GDP, which is then replaced by GTP from the cytoplasm. In their active form Rho GTPases bind to a wide variety of effectors, mediating a large number of cellular processes, including migration, cell-cell adhesion, transcription and proliferation. GEFs also may act to direct signaling by scaffolding particular effectors. To end signaling, GTPase activating proteins (GAPs) bind to the GTPase and enhance their weak intrinsic GTPase activity. Bound GTP is converted to GDP, changing the conformation of the GTPase and rendering it unable to bind effector proteins. Inactive GTPases are mainly found in the cytoplasm, where they can be degraded, or stabilised by binding to Rho GDIs, which act as molecular chaperones and prevent activation by sequestering the GTPases away from GEFs.
finally we review the emerging literature on direct mutation of GTPases.

Copy number alterations and misexpression of Rho GTPases in cancer
Before the finding of direct mutations of Rho GTPases, the main way they were thought to be misregulated in cancer was through changes in expression levels (see Fig. 2). Increased expression of Rho proteins is often associated with tumor formation, growth and progression, an indication of a positive contribution of increased Rho GTPase activity to tumorigenesis. 2 The interesting exception is RhoB which, as discussed below, appears to more commonly play a tumor suppressor role, and is accordingly found at reduced levels in tumor samples.
Rac1 has been found to be overexpressed in testicular, breast and prostate cancer, as well as gastric and lung cancers. [15][16][17][18][19] In recent studies, its overexpression in gastric cancer was correlated with the aggressiveness of the tumors, greater invasion and lymph node metastasis, as well as poor patient survival. 19 Rac1 is also overexpressed in acute myeloid leukemia cells, where it enhances migration and cell growth, and is linked to chemoresistance. 20 Animal experiments support a requirement for Rac1 in tumor formation and growth in many different tumor models. Mice with Rac1 deletion specifically from keratinocytes are resistant to developing Ras-induced skin cancer 21 while those with a Rac1 deletion in pancreatic progenitor cells are protected from development of pancreatic ductal adenocarcinoma (PDAC). 22 Rac1 is also required for K-Ras-induced lung tumors in mice, 23 and cooperates with APC loss in a mouse model of colorectal cancer, driving a stem-cell like signature in the developing cancer cells. 24 Another recent study showed that Rac1 affects stem cell behavior to drive oncogenic progression, by reducing the differentiation of tumor cells. 25 A splice variant of Rac1, Rac1b, was found at elevated levels in colon and breast tumors. 26 Rac1b includes an additional 14 amino acids compared with wild-type Rac1 and it is mainly found in its active GTP-bound state. Rac1b has reduced affinity for GDIs, meaning it is not sequestered in the cytoplasm, which could explain its increased activity and ability for cellular transformation. 27 Rac1b alone is insufficient to drive tumor formation in a non-small cell lung cancer mouse model, but it enhances the activity of K-ras mutations. 28 It is highly expressed in stages 1 and 2 of human lung adenocarcinoma, making it a candidate target for preventing progression to more aggressive stages. 29 Rac1b is also overexpressed in papillary thyroid carcinoma (PTC), where it is associated with the BRAF V600E mutations and subsequently with poor clinical outcomes. 30 Rac2, which shares a high degree of sequence conservation with Rac1, is restricted to expression in the haematapoetic cell lineage. Although no aberrations of Rac2 have been directly linked to oncogenesis, Rac2 is emerging as a therapeutic target, as abrogation of Rac2 signaling slows the growth of AML and CML tumors (reviewed in ref. 31). Rac3 activity was found to be increased in highly proliferative breast cancer cell lines, although this does not correspond to increases at the protein level 32 suggesting other mechanisms of activation.
RhoA and RhoC have been found overexpressed in a wide range of tumors, particularly those with epithelial origins, 2 and in some instances have been linked to oncogenic progression, such as in testicular cancer 15 and to poor prognosis, such as in esophageal squamous cell carcinoma. 33 In contrast to overexpression, loss or reduced expression of RhoB was observed in lung cancer and head and neck squamous cell carcinoma 34,35 suggesting that loss of function of RhoB can contribute to oncogenic progression. However, in a contradictory finding, RhoB is found overexpressed in breast cancer, 16 which suggests possible cell-or cancer-type specific roles for this GTPase which may result from differential expression of downstream effectors and/or upstream scaffolding proteins, or the balance between other Rho GTPases.
Analysis of gene expression data from the SAGE database reveals changes in Cdc42 levels in cancer tissue, both increased and decreased, compared to normal tissue. 36 Cdc42 is overexpressed in testicular and breast cancer, 15,16,37 in non-small cell lung cancer, 38 and in colorectal adenocarcinoma and cutaneous melanoma. 39,40 Finally a less well-studied GTPase, Rnd3/ RhoE is downregulated in HCC (hepatocellular carcinoma) and its downregulation is correlated with poor prognosis and tumor progression, 41,42 while it is upregulated in gastric cancer cells under hypoxic conditions promoting EMT, 43 again highlighting the signaling complexity of these GTPases and their downstream targets.
The evidence for altered expression of the above mentioned GTPases is indicative of a role in tumor initiation and/or progression. It should also be considered that lack of data for some of the lesser-studied members of the Rho GTPase family may in part be due to fewer reagents being available with which to look for alterna-tions in these proteins. More unbiased screening, and particularly genome-level sequencing for activating mutations (see below), may help to reverse some of this historical bias.

Modulation of Rho family regulators
An alternative mechanism by which many tumors upregulate Rho GTPase signaling is by changing the levels or activities of GTPase regulators, including GEFs, GAPs and Rho GDIs (Fig. 2). 44,45 While the general trend is toward overexpression of GEFs, and reduced expression of GAPs and GDIs (indicative of a positive contribution of Rho GTPase signaling to tumorigenesis) the detailed picture emerging is of much more complex regulation, seemingly dependent on tumor type and level of progression.
The GEFs Ect2, MyoGEF, P-Rex1, Tiam1, LARG, Dock180, Vav1, Vav2, Vav3 and b-PIX are overexpressed in a variety of human tumors. 46 Ect2, which has activity for multiple members of the Rho GTPase family including RhoA, Rac1 and Cdc42, has been recognized as an oncogene in human cancer since 2010, being aberrantly overexpressed and mislocalised in various types of tumors. 47 Activation of MyoGEF-a RhoA and RhoC GEF-regulates the invasion of breast cancer cells. 48 Overexpression of the Rac1 GEF P-Rex1 promotes metastasis of prostate cancer 49 and mutations have been identified in PREX2 (a Rac GEF) in melanoma. 50 Tiam1, another Rac1 GEF, was initially identified as being important for invasion in T-cell lymphoma. 51 Tiam1 displays high levels of expression in breast cancer where it is associated with grade and metastatic potential 52 and is a marker for poor prognosis. 53 Overexpression of Tiam1 has also been observed in prostate cancer. 17 Furthermore, overexpression of Tiam1 in lung adenocarcinomas as well as in squamous-cell carcinoma of the head and neck (SCCHN) is associated with disease progression and poor patient survival. 54 In lung cancer, levels of Tiam1 inversely correlate with expression of the E3 ubiquitin ligase HUWE1, which degrades Tiam1 specifically from cell-cell adhesions, indicating that localized regulation of GEF abundance may play a role in cancer. 55 The Tiam1 ortholog STEF/Tiam2 was found to promote proliferation and invasion in liver cancer when overexpressed, and is therefore implicated in the pathogenesis of HCC. 56 b-PIX has also been found overexpressed in many breast cancers. 57 The haematopoietic specific GEF Vav1 is ectopically expressed in pancreatic adenocarcinoma with a positive correlation to reduced patient survival 58 and its presence in a subset of neuroblastoma tumors indicates its involvement in the tumorigenesis process. 59 Moreover, high levels of expression of Vav1 are a marker for poor prognosis in breast cancer. 53 The Vav1 orthologues Vav2 and Vav3 have also been shown to be deregulated in human tumors. Vav3 is overexpressed in gastric cancer 60 as well as in prostate cancer where a novel nuclear function was found to be responsible for its role in malignant progression. 61 Moreover, both Vav2 and Vav3 regulate a lung-metastasis specific transcriptome that leads to breast cancer progression. 62 Finally, the bipartite Rac1 GEF composed of Dock180 and ELMO1 is overexpressed in malignant gliomas, where it contributes to invasion, 63 whereas LARG (leukemia-associated Rho GEF) is found fused with the MLL locus in acute myeloid leukemia (AML) 64 leading to aberrant expression. While not exhaustive, this list is highly indicative of an oncogenic function for upregulated Rho GTPase signaling.
This data from human tumors is supported by evidence from transgenic mouse models highlighting the importance of a number of GEFs in oncogenic progression. Tiam1 has been shown to be important for Ras-mediated skin 65 and intestinal tumorigenesis. 66 Interestingly, Tiam1 deficient mice develop fewer tumors, but those which do grow are more invasive, suggesting both positive and negative roles for Tiam1 in oncogenesis. Loss of P-Rex1 leads to a reduction in the invasive potential of melanoma cells in a mouse model of the disease, consistent with work in vitro showing that P-Rex1 can regulate invasion. 67 P-Rex2 is also frequently mutated in melanoma, and a truncating mutant, E824 Ã , has recently been shown to cooperates with NRAS to accelerate melanoma development in a mouse model. 68 Mice deficient for the Rac1/Cdc42 GEFs Asef1 and Asef2, which are downstream of APC and are overexpressed in colorectal tumors, show reduced spontaneous formation of intestinal adenomas. 69 Mice transplanted with leukemic B-cell progenitors expressing the p190-BCR-ABL transgene develop tumors at high frequency; however if these cells are deficient for Vav3 then tumor formation is significantly decreased, and survival time increased. 70 Both Vav2 and Vav3 are required for initiation and promotion of skin tumorigenesis. 71 The GAP DLC1 (deleted in liver cancer) is a tumor suppressor frequently downregulated in many cancer types either by deletion or epigenetic silencing. Loss of DLC1 leads to an activation of RhoA, and cooperates with oncogenic Myc in a mouse model of liver cancer. 72 DLC2 was also found downregulated in hepatocellular carcinoma, 73 and more recently was shown to be required to regulate Cdc42 activity for faithful chromosome segregation during mitosis. 74 P190RhoGAP is another RhoGAP thought to act as a tumor suppressor; it is frequently deleted in gliomas, and its overexpression is able to suppress tumor formation in a mouse model of the disease. 75 However not all GAPs are found downregulated in human tumors; ARHGAP8 is found overexpressed in colon cancer. 76 The picture for Rho GDIs is relatively complex, possibly due to their ability to target multiple Rho GTPases and their roles in regulating Rho GTPase activity, stability and trafficking. 11 For instance, Rho GDI1 is found downregulated in some breast cancer studies, 77 but overexpressed in others. 78 Downregulation of Rho GDI2 in bladder cancer is associated with decreased patient survival 79 whereas overexpression in pancreatic cancer is associated with invasion. 80

Post-translational modifications
As discussed earlier, Rho GTPases are regulated by a whole host of post-translational modifications, many of which are now being linked to inappropriate Rho GTPase function in human cancers and a few of which we will discuss here as illustrative examples. Ubiquitylation of Rac1, RhoA and Cdc42 can be deregulated in cancer cell lines, a fact that could indicate a link between Rho GTPase protein ubiquitylation and cancer. 14 For instance, the E3 ligase SMURF1 targets RhoA for degradation at the leading edge of migrating cells, affecting tumor cell migration. 81 PIAS3 SUMOylates Rac1 stabilizing the active form of the protein following HGF stimulation and therefore promoting cell migration and invasion, suggesting a possible role in cancer progression. 13 Conversely, Rac1 can be ubiquitylated by the E3 ligase HACE1, resulting in its proteasomal degradation, reducing Rac1 mediated migration. 82 Ubiquitylation of RhoA has also been reported to be impaired following FBXL19 downregulation in lung cancer epithelial cells. 83 FBXL19 ligase also ubiquitylates Rac1 and Rac3, with degradation impairing esophageal cancer cell EMT. 84 Finally, phosphorylation of Rho GTPases has also been shown to regulate their transforming ability; for instance phosphorylation of Cdc42 by the Src tyrosine kinase modulates its interaction with Rho GDI which is necessary for cellular transformation. 85 These examples from the literature demonstrate some of the great diversity of mechanisms by which cancer cells can indirectly disrupt upstream signals which lead to Rho GTPase activation.

Direct mutations of GTPases in human cancers
Early studies had identified mutations in RhoH such as the rearrangement of RhoH/TTF gene and the mutation of the 5 0 -UTR of RhoH gene in some haematopoietic malignancies. 86,87 However, mutations within Rho GTPases, except for RhoH, were believed to be rare in cancer until recently. This led to the speculation that Rho GTPases were not direct drivers of oncogenic progression, but merely downstream players in a disease more directly modulated by upstream signaling pathways. With the development of faster and cheaper deep sequencing technology this idea has been challenged, as Rho GTPases have now been found mutated in a wide variety of cancer types (see Table 1). 88 In particular, the discovery of a recurrent Rac1 mutation in melanoma has significantly altered the perception of the role of Rho GTPases as drivers of oncogenic progression. For this review, we gathered data on published mutations in the Rho GTPases Rac1, Rac2, Rac3, Cdc42, RhoA, RhoB, RhoC, RhoH and RhoT1 using the cBio portal (http:// www.cbioportal.org/), a database that collects cancer genomics data sets from tumor samples across cancer studies, 89,90 and IntOGen (https://www.intogen.org/ search), which assesses mutational data across multiple tumor types to identify potential driver mutations. 91 Both databases are user-friendly, regularly updated, and include additional information such as expression levels, amplifications and deletions (see Table 1). While any table of this kind becomes quickly outdated, it nonetheless serves to highlight the remarkable impact of sequencing technology on the discovery of mutations in human cancers in recent years, as well as the range of cancer types harboring mutations in Rho GTPases. The following section will focus on the emerging literature around these newly-identified mutations and other identified deregulations of Rho GTPases in human cancers.

Rac1 mutations
One early study aiming to detect Rac1 mutations in human brain tumors identified deletions, frame shift and point mutations in 12 out of 45 samples from human patients with brain tumors, 92 suggestive of a role for Rac1 in brain tumor development. Now, next generation sequencing has identified a number of cancer-associated mutations along the length of the Rac1 protein, with Rac1 being identified as a driver mutation in head and neck squamous cell carcinoma and cutaneous melanoma (see Table). Among these, P29 is a hot-spot for Rac1 mutation. It was first identified by 2 groups in 2012. 93,94 Whole-exome sequencing was performed in melanoma samples and 5% of them were found to harbour missense mutations in the Rac1 gene, making Rac1 the third most highly mutated gene in melanoma (after BRaf and NRas). 93 The functional effect of the P29S recurrent mutation is to induce a 'fast cycling' form of Rac1, as opposed to the more common gain-of-function mutations used in a laboratory setting which are modeled on activating Ras mutations found at high frequency in human cancers. These mutations, found at G12 and Q61, block GTPase activity and so trap the GTPase in its active, GTP-bound form. In contrast, the P29 residue lies in a hydrophobic pocket in the switch I region of the Rac1 GDP-bound form, and the substitution of the proline residue for a serine enhances the exchange of GDP for GTP, 95 while still maintaining the ability to hydrolyse GTP back to GDP. Overall this enhances the interaction of Rac1 with effectors, such as the Pak family of kinases. P29S is therefore considered a gain-of-function mutation that likely promotes oncogenic events during melanoma through mechanisms thought to include altered cell proliferation, adhesion, migration and invasion. 93 Expression of the mutant form of Rac1 in melanocytes leads to enhanced cell proliferation and migration, 94 and the Rac1 P29S mutant form is able to transform mouse fibroblasts and immortalised breast epithelial MCF10A cells. 96 Subsequently 2 other fast-cycling mutants of Rac1 have been identified, N92I and C157Y. 96 The ability to cycle from the off-state to the on-state may render these fast-cycling mutants more efficient at driving transformation than the constitutively active mutants, possibly because they more closely mimic normal signaling by being able to associate and dissociate from effectors, or potentially by still associating with GEFs acting also as scaffolding proteins. Rac1 N92I was able to efficiently transform mouse fibroblasts and MCF10A cells, whereas the C157Y mutation was less effective. 96 Interestingly, Rac1 P29S (which has also been found as a somatic mutation in a breast cancer cell line) transformed MCF10A cells more efficiently than fibroblasts, whereas the opposite was true for the Rac1 N92I mutation (known as a somatic mutation in a fibrosarcoma cell line), 96 suggesting that there are further subtleties to the effects of these different activating mutations still to be uncovered.
A serious clinical problem in the treatment of melanoma is the swift development of resistance to the front line treatments of RAF and MEK inhibition. A 2014 study revealed that Rac1 P29S expression in melanoma cell lines and in mouse tumor models conferred resistance to RAF and MEK inhibitors 97 with overexpression of Rac1 P29S decreasing apoptosis after RAF and MEK inhibitor treatment. A further clinical study suggested the potential of Rac1 P29S as a predictive biomarker for resistance to therapy in melanoma patients under treatment with these inhibitors. 98 Further histological and clinical evidence showed that this hot spot mutation may be responsible for the early metastatic progression of BRAF mutant and BRAF wild-type melanoma. 99 A more recent biological insight into the P29S mutation showed   Notes. * = Stop codon; fs = Frame shift; Mutations in bold = Hotspot mutation; Solid boxes = Identified as a driver mutation by IntOGen algorithms. Numbers in brackets indicate number of times a mutation identified across data sets.
increased expression of PD-L1 in Rac1 P29S melanoma patients compared to Rac1 wild type or other Rac1 mutants. 100 PD-L1 is a suppressor of the immune system thus its upregulation may allow cancers to evade the host immune system and therefore oncogenic Rac1 P29S may promote the reduction of anti-tumor immune response. As PD-L1 is a candidate biomarker for increased benefit from treatment with anti-PD1 or anti-PD-L1 antibodies, this finding could also have implications in the clinic.

Rac2 and Rac3 mutations
In the 2012 study conducted by Hodis et al., a homolog to the Rac1 P29 residue was found to be mutated in Rac2, substituting Proline (P) with Leucine (L) (P29L mutation). 93 Two Rac2 mutations -P29L and P29Qsimilar to the P29S mutation, were confirmed as transforming mutations of Rac2. 96 Additional mutations have since been identified in Rac2 (see Table 1) and among these the R102Q was found as a hot spot mutation. Mutations in the Rac3 gene have been identified from a range of cancers, including melanoma, stomach and prostate, but none has yet been studied functionally. Rac2-KO and Rac3-KO mice showed slightly increased survival in a CML and ALL background, respectively, 101,102 suggesting a possible oncogenic role for these genes; further experimentation will be required to determine the functional significance of these cancer-associated mutations.

RhoA mutations
As with Rac1, no mutations in RhoA had been detected in human cancers until very recently. RhoA mutations were identified by several groups in 2014, 103 with an IntOGen search indicating a driver role for RhoA in stomach adenocarcinoma (see Table 1), as well as a general pan-cancer driver role. RhoA mutations have been identified in 25% of diffuse-type gastric carcinoma cases studied. 104 Recurrent mutations were R5Q, G17E and Y42C. Expression of both RhoA G17E and Y42C were able to rescue growth defects of SW948 colon cancer cells grown in 3D culture following knockdown of endogenous RhoA in contrast to re-expression of wildtype RhoA which was unable to rescue. 104 Several groups have found frequent RhoA mutations, specifically the G17V mutation, in angioimmunoblastic T cell lymphoma and peripheral T cell lymphomas. [105][106][107] Interestingly this mutation appears to act similarly to well-characterized dominant negative mutations of RhoA, rather than as an activating mutation. Expression of this mutant form of RhoA increases proliferation in Jurkat cells, an effect also observed with expression of dominant negative RhoA. This fits well with work showing that inactivation of RhoA promotes tumor formation in colorectal cancer models. 108 Silencing of RhoA in colon cancer cell lines promoted proliferation, largely through activation of the Wnt/b-catenin pathway and subsequent upregulation of Myc signaling, and this led to increased metastasis. In a mouse model of colorectal cancer, metastatic sites were found to have lower RhoA signaling than the primary tumors, and this held for samples from human tumors as well. 108 Another example of inactivating RhoA mutations are those found recurrently in Burkitt Lymphoma, the most common type of childhood B-cell lymphoma. Translocations of the MYC locus leading to deregulated Myc signaling are necessary but not sufficient to drive disease progression, and both whole genome studies 109 and exome sequencing 110 identified RhoA mutations as additional drivers of the disease. 8.5% of cases had RhoA mutations, and molecular modeling of these mutations suggested that they would reduce RhoA activity, or reduce binding to RhoA effectors. 110 Another study conducted with gastric adenocarcinoma samples 103 added a number of additional mutations including Y34C, F39C, E40K, N41K, Y42S/C/I, L57V, D59Y, T60K, A61D and G62R/E (and see Table 1). These were accumulated in regions that participate in the interaction of RhoA protein with effector molecules; for instance mutations at Y42 reduce downstream activation of PKN but not mDia or ROCK1. 111 This indicates that distinct mutants may have different alterations in effector binding/ activation with some of them leading to reduced interaction of RhoA with specific effector proteins. Depending on the target affected, this altered RhoA activity could account for the increased cell spread and the absence of cell cohesion observed in this kind of tumors. These studies suggest either that wild-type RhoA, in the cells of origin for these cancer types, is acting in a tumor suppressive capacity, or that inactivation of RhoA in some way leads to hyperactivation of an oncogenic pathway. C3 toxin-mediated inactivation of RhoA, B and C causes the development of aggressive malignant thymic lymphomas in mice. 112 Such findings support a tumor suppressor role for these members of the Rho family. It will require further experiments to reconcile data from these mutational studies with earlier work showing that overexpression of RhoA promotes tumorigenesis. This could be due to differences in the expression of downstream effectors in different tissue types, or different requirements for RhoA throughout the life-cycle of a tumor.

RhoB, RhoC and RhoT1 mutations
RhoB has been found to be mutated in 5% of bladder cancer cases from a sample of 131 high grade tumors not treated with chemotherapy (with more than 200 additional samples still to be sequenced at the time of writing) making it one of 9 genes mutated in this disease. 113 Our cBio search for published RhoB mutations (see Table 1) indicates that P75S/T/L is a hot spot mutation, though it has not yet been studied functionally. In a model of Ras-driven skin cancer, Liu and colleagues showed that the RhoB-null mice had increased skin tumors compared to the heterozygote mice and that RhoB-deficient MEFs transformed with E1A and Ras showed greater resistance to DNA-damage induced apoptosis, 114 which suggests that, if functional, these might be inactivating mutations.
Two other family members, RhoC and RhoT1, present a number of published mutations in cancer samples and cell lines, with the S73 residue a hotspot in RhoC, while mutations in RhoT1 include a P30L mutation, which by homology may have similar effects to the Rac1 P29S mutation. Deletion of RhoC from mice has been observed to reduce the frequency and growth of tumors, 115 which might suggest that activating mutations might promote tumor formation, but further analysis of mutations in these family members is required to determine their functional relevance.

Cdc42 mutations
The classical activating mutation G12D (equivalent to the G12V activation mutation of Ras) has been found in Cdc42 in melanoma cells in the same study which identified the Rac1 P29S mutation, 93 although this mutation was present in only a single patient sample, and has not been functionally characterized. Table lists 14 different published mutations in Cdc42, although no function has yet been ascribed to them. However, given the evidence for a role for Cdc42 in cellular transformation, 36 we conclude that it is highly likely that at least some of these mutations will be functionally active. It is also possible that some of these may be inactivating mutations, as in vivo evidence, such as deletion of Cdc42 from hepatocytes which lead to spontaneous tumor formation, 116 suggests that Cdc42 might also play a role as a tumor suppressor.

Pharmacological inhibition of Rho GTPases
Given the long-standing in vitro and in vivo data showing Rho protein involvement in malignant transformation, observed changes in Rho protein expression levels or changes in their regulators and post-translational modifications, and now direct mutation of Rho GTPases, in human cancers, targeting Rho protein signaling is an increasingly attractive target for new cancer therapeutics.
Small molecule inhibitors of many Rho proteins are currently being developed and tested.
Two different small molecule inhibitors of Rac are currently in use, utilizing 2 different strategies for inhibition. NSC23766 works by inhibiting the interaction between Rac1 and its GEF Tiam1, reducing the activation of Rac1. 117 EHT 1864 is a pan-Rac inhibitor which directly targets the Rac GTPase itself, by displacing GTP from the active site. 118 NSC23766 can halt the proliferation, anchorageindependent growth and invasion of prostate cancer cells. 117 Rac1 inhibition can additionally reduce growth of non-small-cell lung cancer tumors in a mouse model that present resistance to (EGFR)-tyrosine kinase inhibitors such as gefitinib, making it attractive as a potential combination therapy to help circumvent the resistance mechanisms. 119 Moreover, Rac1 inhibition impedes the growth, invasion and metastasis of gastric tumors. 19 However, while both these inhibitors do indeed target Rac activity, they also have significant off-target effects, as demonstrated by assays using wild-type and Rac1-deficient mouse platelets. 120 This emphasizes the need to develop better versions of these drugs, or find other ways of targeting Rac, and other small GTPases. One approach is to use in silico screening to predict potential binding partners which might block GTPase-GEF interactions. 121 It is worth noting that this strategy of targeting the interaction between GEFs and GTPases is predicated on the function of the GEFs regulating tumorigenesis via their ability to activate the GTPases. However, this is not always the case. For instance, the activation of the PI3K/Akt pathway by the GEF P-Rex2 does not depend on the GEF activity of the protein. 122 Also, a recent paper from our lab demonstrates that different GEFs can have differential effects on cell behavior, despite activating the same GTPase to similar levels, 9 most likely by scaffolding different downstream effectors of the GTPase; therefore it will be important to target the correct GEF-GTPase activity for the specific cancer type.
Given that Rac1 and Cdc42 are highly expressed and active in ovarian cancer, 123 inhibitors of these 2 GTPases have been tested in immortalized and primary human ovarian cancer cells. 124 The R enantiomer of ketorolac, (ketorolac is given as an anti-inflammatory drug), can inhibit Rac1 and Cdc42 and was shown to improve patient outcomes in treatment for ovarian cancer. 124 Another Rac1 and Cdc42 dual-inhibitor, AZA1, identified from a screen of molecules based on modifying the structure of NSC23766, has been used in in vitro studies to target prostate cancer cells. 125 This synthetic compound reduced cancer cell migration and proliferation and succeeded in increasing the survival of xenograft mouse models of prostate cancer by targeting Rac1 and Cdc42 but not RhoA. 125 An additional 3 inhibitors of Cdc42 have been developed. Secramine has been identified as a small molecule inhibitor that perturbs Cdc42 activity in a RhoGDI1dependent manner, 126 although is likely to affect other GTPases in the same manner. ZCL278 is a small molecule inhibitor of Cdc42, designed to block the interaction of Cdc42 with the GEF Intersectin. It is thought to disrupt both GEF interactions and GTP binding 127 and was shown to inhibit actin-based motility and migration in a metastatic prostate cancer cell line. 127 Finally, AZA197, another recently developed Cdc42 inhibitor which appears not to inhibit Rac1 activity has shown some efficacy in reducing tumor size in a xenograft model of colon cancer. 128 Reducing signaling through the Rho pathway is often achieved by targeting the Rho target ROCK. 129 The ROCK inhibitor Y-27632 retards migration of human prostate cancer cells in mice 130 and blocks the invasive activity of cultured rat hepatoma cells. 131 Moreover, inhibiting the Rho/ROCK signaling pathway in NSCLC using the ROCK inhibitor fasudil, when combined with inhibition of the proteasome, effectively reduced the viability of mutant K-Ras cells compared with wild-type cells. 132 It is likely that further structural modification of these compounds, or further high-throughput compoundscreening, will lead to more specific inhibitors, and that as we further our understanding of both normal and abnormal Rho GTPase signaling we will be better placed to deploy them therapeutically.

Conclusion
In conclusion, Rho GTPase signaling is frequently seen to be modified in human cancers through a variety of mechanisms, and work is continuing to understand the consequences of this aberrant signaling. Understanding the wider landscape of Rho GTPase signaling in a tumor type is likely to be important for making the correct, clinically-relevant interventions. Modifications occur from the level of mutation of the GTPases to under or overexpression of their regulating proteins, which both generates a highly complex signaling network that needs further work to be untangled and also suggests many fertile avenues for therapeutic intervention.

Disclosure of potential conflicts of interest
No potential conflicts of interest were disclosed.

Funding
This work was supported by Cancer Research UK (grant number C5759/A12328) and Worldwide Cancer Research (grant numbers 12-0037 and 16-0379). AP was supported by a scholarship from "IKY Fellowships of Excellence for Postgraduate Studies in Greece -SIEMENS Program."