An Expanded Tool Kit for Modeling the Oncogenic Functions of KRAS

The first genetically engineered mouse model allowing for endogenous and conditional expression of oncogenic KRAS was generated in the laboratory of Tyler Jacks almost 20 years ago (1). This LSL-Kras^G12D model enabled tissue-specific expression dependent on Cre-mediated activation. At the time, it was known that distinct KRAS mutations were present at different frequencies in common tumor types, yet the biological significance of this observation was not fully appreciated. The LSL-Kras^G12D conditional model was quickly adopted by hundreds of laboratories and was used to model a variety of KRAS-driven cancers, most prominently lung, pancreatic, and colon cancers but also many others. Although it has become an indispensable tool for in vivo dissection of the mechanistic basis for KRAS-driven oncogenesis and has facilitated myriad important discoveries, the LSL-Kras^G12D mouse expressed only one of the many known KRAS-activating mutations, potentially obscuring important biological differences between these mutations that remain to be understood (2).

Over the past few years, several lines of evidence have accumulated indicating that KRAS mutations are not equivalent, but rather have unique biochemical properties that affect their ability to drive oncogenesis in different tissues. Epidemiologic studies first suggested that different KRAS alleles have distinct clinical consequences that are also tissue-dependent (3, 4). One key feature of the KRAS protein is that it has both intrinsic and GTPase-activating protein (GAP)–stimulated GTPase activity. A study led by Kenneth Westover compared the intrinsic and GAP-mediated GTP hydrolysis and RAF binding activities of some of the most common KRAS mutants (5). It proposed a classification scheme based on these two biochemical characteristics. For example, the G12C allele was observed to have intrinsic GTPase activity and RAF binding activity similar to that of wild-type KRAS, whereas the G12R and G12V mutants have low intrinsic GTPase and RAF binding activities. G13D, on the other hand, has a lower GTPase activity than wild-type and intermediate RAF binding activity, but a faster nucleotide exchange rate. These differences are supported by structural studies, which indicate that the differences in nucleotide exchange kinetics can be explained by the electrostatic potential of the protein surface.

Consistent with these mutation-specific biochemical properties, a study led by Channing Der found that the G12R allele, which is most commonly seen in pancreatic cancer, is defective in interaction with PI3Kα and thus impaired in stimulating macropinocytosis (6). Similarly, work from Kevin Haigis's laboratory directly compared the consequences of KRAS^A146T mutation (found primarily in colon cancer) with KRAS^G12D using genetically engineered mouse models expressing either mutation in the colonic epithelium (7). A notable observation was that in vivo these two mutations have distinct effects on MAPK pathway activation (WT < A146T < G12D). These and other studies have provided a deeper appreciation of the importance of specific KRAS mutations in cancer biology. Perhaps most dramatically, our awareness of the importance of this issue has increased because of the discovery of small molecules that specifically inhibit the G12C mutation (8). An important feature of the G12C mutant that allows the allele-specific inhibitors to function is that it retains the capacity for nucleotide exchange and thus can be “trapped” in the GDP-bound form by the inhibitor, thereby depleting the pool of GTP-bound KRAS protein.

Despite these advances, a key limitation in the field was the lack of conditional models for other KRAS alleles that could be directly compared with the LSL-Kras^G12D. This problem is now addressed to a significant extent in the work of Zafra and colleagues described in this issue of Cancer Discovery (9). Using CRISPR-mediated homology-directed repair in mouse embryonic stem cells, they generated a series of novel conditional mouse models carrying three KRAS mutations most prevalent in lung (G12C), pancreatic (G12R), and colorectal (G13D) cancers. They compare these new mutants with the “gold standard” KRAS^G12D model yielding important new insights into their allele-specific oncogenic function. An initial clue pointing to the unique function of each allele was provided by transcriptome analysis in mouse embryo fibroblasts, indicating that each allele had a distinct transcriptional signature. To determine how these different mutations affect tumor initiation in vivo, the authors expressed each allele in the epithelium of the colon or pancreas by using tissue-specific Cre driver strains. Expression of KRAS^G12C in the colon using FABP1-Cre induced widespread hyperplasia identical to the previously described phenotype of LSL-Kras^G12D mice. In contrast, activation of the G13D mutation, frequently seen in patients with colorectal cancer, resulted in only moderate hyperplasia, whereas the G12R mutation had no discernible phenotype and resembled the wild-type allele (an interesting observation given that some reports have suggested that G12R-mutant colon cancers have a worse prognosis). In contrast, when expressed in the pancreas, KRAS^G12D had the strongest phenotype, leading to the widespread development of pancreatic intraepithelial neoplasia (PanIN) lesions at 12 weeks of age. In mice expressing the G12C allele, only 50% of lesions progressed to PanINs while the remainder of the pancreas showed evidence of earlier stage acinar-to-ductal-metaplasia (ADM). The G13D and G12R mutants showed little progression to PanINs. The same was true in mice where pancreatitis was induced with cerulein treatment. Particularly surprising was the observation that mice carrying the G12R mutation (found in ∼20% of pancreatic cancers) showed almost no PanIN lesions even at 50 weeks of age. The authors suggest that this may be due to the observed low numbers of DCLK1-positive stem cells, previously implicated in disease progression in the pancreas. Together, these results demonstrate that in the pancreas, the different KRAS mutants do not follow the same disease progression as the G12D mutant (ADM > PanIN > PDAC), and G12C, G12R, and G13D have overall reduced disease-initiating capacity when compared with the G12D model (see Fig. 1).

To evaluate the utility of the new mutant strains as tools for preclinical studies, the authors generated ductal pancreatic organoids from KRAS-mutant mice, in which the CRISPR/Cas9 system was also used to disrupt p53. In general, the G12R organoids had a weaker KRAS transcriptional signature. When transplanted orthotopically, KRAS G12D, G12C, and G12R organoids readily formed tumors, whereas G13D organoids did not grow, perhaps suggesting that G13D allele is more reliant on upstream signals which may be disrupted in this xenograft model. The fact that the G12R allele was a potent oncogene in this setting is consistent with the relatively high frequency of this mutation in pancreatic cancer. This was in stark contrast to the observed delay in premalignant progression in the pancreas, as noted above, suggesting that the impact of this activating mutation is largely context-dependent. The authors subsequently tested the response of the KRAS-mutant pancreatic organoids to EGFR and KRAS^G12C inhibitors, which are shown to elicit a therapeutic response in G13D-driven colorectal cancer and G12C-driven lung cancer, respectively. Consistent with previous reports in colon cancer, only G13D organoids showed a profound and durable response to the EGFR inhibitor gefitinib. Surprisingly, G12C organoids did not respond to the G12C inhibitor ARS-1620. A recent study demonstrated that basal high EGFR activity is the main mechanism of resistance to KRAS^G12C inhibitors in colorectal cancer (10). In their work, Zafra and colleagues show that the same is true for pancreatic organoids and that only combined treatment with ARS-1620 and gefitinib leads to complete eradication of KRAS^G12C organoids. The additional insight into the therapeutic effect of targeting relatively rare mutations, such as G13D, is especially useful. Although these mutations receive less attention, review of the GENIE database (genie.cbiportal.org) identified 1,350 of approximately 90,000 patients with KRAS codon 13 mutations (including several hundred in diseases other than colorectal cancer), suggesting a significant population of patients in which therapy with gefitinib or possibly other upstream inhibitors of KRAS signaling should be further explored.

Overall, Zafra and colleagues present convincing evidence that different KRAS codon 12 and 13 mutations have distinct influence on disease initiation and progression in different tissues. A key aspect of their biology, which remains incompletely understood, is how their biochemical properties alter their effects on downstream effectors. Given that effector-based therapy is likely to be needed for many KRAS-activating mutations, understanding the link between these mutations and their effector pathways is critical. Although the unique capabilities of these mutations have been well described in vitro in recent years, tools to rigorously evaluate their contribution to in vivo oncogenic phenotypes have been lacking. Although it is likely that the core biochemical properties of KRAS contribute significantly to their oncogenic capacity, there is also significant evidence that this is influenced by the genetic context in which these mutations occur. Further investigation of the effect of co-occurring mutations on the output of specific oncogenic KRAS mutations is a major question in the field. Another important question is the role of wild-type KRAS, which has been previously shown to function as a tumor suppressor under certain conditions (11). The mouse models generated by the authors retain the expression of wild-type KRAS, and as such will certainly be an invaluable resource to investigate its impact on disease progression. Thus, the new tools developed by Zafra and colleagues should provide tractable new avenues of investigation that will bring the full power of mouse modeling to bear on our ability to fully understand how KRAS drives oncogenesis in human cancer.

No potential conflicts of interest were disclosed.