A New Rho(d) Map to Diffuse Gastric Cancer

Gastric cancer is the fifth most diagnosed cancer in the world, affecting more than a million people each year and causing nearly 800,000 deaths. There are two major types of gastric carcinomas, intestinal (IGC) and diffuse (DGC), which are characterized by different histologic features (1). DGCs lack strong intercellular adhesions and often include scattered, mucin filled, signet-ring cells. These cancers are often invasive and prone to metastasis, leading to a poor prognosis. In 2014, whole-genome sequencing of diffuse gastric tumors uncovered several new driver mutations (2, 3). Of note, mutations in RHOA, a gene encoding a well-studied small GTPase, were found in 15% to 25% of DGCs but not in intestinal tumors (2, 3). When present, RHOA mutations usually occur in tandem with loss-of-function mutations in CDH1 (encoding E-cadherin) or TP53, and a particular hotspot mutation, RHOA^Y42C, was shown to increase cell survival in intestinal organoid experiments (2, 3), consistent with a proposed driver role for this gene. Subsequent studies have tried to characterize the mechanism and signaling consequences of this mutation, but have come to conflicting conclusions as to whether it represents a gain or a loss of function (3, 4). In their article in this issue (5), Zhang and colleagues have used biochemical, cellular, and in vivo methods to demonstrate that RHOA^Y42C is in fact a gain-of-function mutation that, along with a loss of CDH1, can cause DGC in organoid and animal models, and that these models can be used to predict therapeutic vulnerabilities.

RHOA is a member of a larger family of RHO proteins of small GTPases, with structural and signaling similarities to RAS proteins (6). Although activating mutations in the RAS proteins have been recognized for decades, driver mutations that affect the related RHO family proteins, such as RAC1 and RHOA, were not identified until a few years ago. Instead, RHO pathway overactivation in cancer was more often seen in association with overexpression of the wild-type forms of these GTPases or activating mutations in guanine nucleotide exchange factors that stimulate RHO proteins, or inactivating mutations in GTPase-activating proteins (GAP) that normally deactivate RHO proteins by accelerating GTP hydrolysis. However, exon-sequencing efforts over the last few years have made it clear that recurrent missense mutations are also found in genes encoding RHO family proteins, albeit more rarely than in RAS and in a more restricted set of tumor types. For example, it is now well established that activating point mutations in the gene encoding the RHO family protein RAC1 are associated with sun-exposed cutaneous melanoma and testicular germ cell cancers, and, more rarely, with lung, pancreas, and head and neck cancers. Although the positions of the missense mutations in RAC1 are often atypical compared with those commonly found in RAS genes, they still fit the general paradigm that small GTPases such as RAS and RAC are positive signaling elements in cell proliferation, survival, cell motility, and invasiveness, and that, when activated by point mutation, they can drive cancer formation and evolution. One apparent outlier in this scheme has been the DGC-associated mutant RHOA^Y42C. The effects of the mutation on RHOA activity have been surprisingly hard to pin down. Previous attempts to correlate this mutation with RHOA activity have been hampered by conflicting biochemical data regarding the enzyme's GTPase activity and its ability to bind effectors. In their article in this issue, Zhang and colleagues convincingly put this question to rest, by a combination of rigorous biochemical analyses and the creation of the first organoid and mouse models of RHOA-driven DGC (5). Regarding the latter, most Cdh1^−/−;Rhoa^Y42C/+ mice developed tumors that resembled human DGC in terms of histology and patterns of spread, whereas no tumors developed in control mice bearing either mutation alone. In combination with the human sequencing data, these findings strongly support the idea that, whatever its biochemical effects, RHOA^Y42C is a bona fide oncogene whose driver activity requires concomitant loss of a tumor-suppressor gene such as CDH1.

What is the mechanism behind this mutant? Like most other small GTPases, the RHOA GTPase operates via a switch-like, GTPase cycle, in which the protein toggles between an inactive GDP-bound form and an active GTP-bound form. In its active form, RHOA recruits a variety of effector proteins such as RHO kinase (ROCK), PKN, LIMK, Rhotekin, and mDia (7). It is therefore intuitively attractive to suppose that the RHOA^Y42C mutation activates this protein by interfering with GTP hydrolysis or by increasing the on–off cycling rate, but, given the position of the altered amino acid, one might at first glance predict the opposite to be true. That is because the Y42 site resides in switch 1, a substrate-binding region of RHOA that is analogous to Y40 in RAS and RAC proteins, respectively. Mutation of this tyrosine residue in RAS and RAC does not affect GTPase activity but uncouples these latter two proteins from key effectors such as RAF and PAK, respectively. Therefore, it is easy to imagine that the Y42C mutation in RHOA might uncouple it from its essential effectors. However, long before the RHOA^Y42C mutation was found in cancer, this mutant was purposely constructed as part of structure/function studies of RHOA signaling mechanisms (8, 9). Interestingly, one of these studies showed that the Y42C mutant exhibited reduced binding to Rhotekin, but maintained normal binding to ROCK and mDia2 (8). In addition, both studies showed that this mutant did not affect the ability of a GTPase defective form of RHOA (RHOA^G14V) to induce stress fiber formation. Thus, despite its intriguing location, the mutation at Y42 seems not to be analogous to the equivalent mutations in RAS or RAC1. Moreover, as these older experiments were performed in the context of a preexisting RHOA^G14V activating mutation, it was not clear what effect the Y42C mutation might have in isolation, as actually found in DGC and other tumors. Indeed, until this work of Zhang and colleagues (5), it has remained an open question whether this mutation causes an increase, decrease, or perhaps even a neomorphic change in RHOA's activity by altering its effector repertoire. Here, the authors show that the mutation causes an incomplete loss of GTPase activity and, consistent with the prior studies noted above, a reduced ability to bind Rhotekin, but, somewhat surprisingly, an increased ability to bind ROCK. As Rhotekin impedes actin stress fiber formation whereas ROCK promotes it, these biochemical findings correlate well with the observed increase in stress fibers and focal adhesions induced by this mutant form of RHOA.

As one of the most characteristic outcomes of RHOA action is the production of focal adhesions, RHOA signaling strongly promotes the activation of focal adhesion kinase (FAK), and this phenomenon was observed in RHOA^Y42C-expressing gastric organoids. The authors also observed PI3K-dependent AKT activation, with subsequent inactivation of GSK3β and stabilization and nuclear localization of β-catenin. In addition, they found that expression of RHOA^Y42C led to increased expression of the transcriptional coactivators YAP and TAZ. Overexpression of β-catenin and YAP together could reconstitute transformation activity in Cdh1-null gastric organoids lacking the RHOA^Y42C transgene, suggesting that the WNT and Hippo pathways represent key elements in the oncogenic activity of RHOA^Y42C (Fig. 1A) Although these two pathways are challenging to target, there are a number of anti-FAK clinical agents now in trials, and the authors show that such compounds were effective in vitro and also in their new mouse model. Whether ROCK would also represent a useful target is not known, but it might also be a reasonable candidate to pursue, alone or in tandem with FAK inhibitors.

Though not discussed by Zhang and colleagues, one unexpected and potentially important additional clue to the role of this mutation was recently reported by Liu and colleagues, who found that Y42 in RHOA was a target site for phosphorylation by the receptor tyrosine protein kinase MET (10). Phosphorylation of this site promoted ubiquitination and subsequent degradation of the RHOA protein. In this scenario, mutation of Y42 to a nonphosphorylatable residue might likewise impede ubiquitination and thereby stabilize RHOA, increasing its signaling intensity and/or duration (Fig. 1B). It would be interesting to test this idea by comparing the half-lives of wild-type RHOA and RHOA^Y42C in a suitable cell type or animal model.

This study definitively shows, as some but not all had suspected, that RHOA^Y42C is a gain-of-function mutation after all. In fact, like the conventional but artificial G14V and Q63L mutations, the RHOA^Y42C protein has impaired GTPase activity, although not to the same degree as the lab-made RHOA^Q63L mutant. The authors posit that this limited GTPase dysfunction sets up a necessary Goldilocks scenario, in which RHOA enzymatic activity in neither too low nor too high, but just right to support a level of stress fiber formation that increases motility. More than that, the authors sketch an outline for its route(s) to transformation, and present a plausible case for why this mutation is often found in concert with loss of CDH1. In addition, they provide hints toward rational therapeutics. What remains to be clarified is the relationship between this mutation and RHOA stability, the precise effector pathways engaged by the mutant, and whether the findings presented here also pertain to the other RHOA mutations found in DGC and other malignancies because some of these other mutations, in particular RHOA^G17E, would seem to represent dominant negative forms. The answers to these questions should help to define the molecular pathophysiology of this deadly cancer and allow the formulation of targeted therapies where none currently exist.

No potential conflicts of interest were disclosed.

The Chernoff laboratory is supported by grants from the NIH (R01CA142928, R01CA148805, and R01CA227184) and the Melanoma Research Foundation. Fox Chase Cancer Center is supported by NCI Core Grant P30 CA06927.