Primary Prevention and Interception Studies in RAS-Mutated Tumor Models Employing Small Molecules or Vaccines

The RAS family was one of the earliest discovered oncogenes associated with carcinogenesis. The HRAS gene initially identified from the Harvey sarcoma virus, induces both lymphoid tumors and sarcomas in rodents (1). KRAS from the Kirsten sarcoma virus and causes erythroblastosis and sarcomas when injected into weanling rodents (2). Mutated forms of these genes were quickly identified in oncogene transformation assays in which human tumor DNA was transfected into embryonic rodent cells in culture (3). Most importantly, mutations in the RAS family of proteins are associated with cancers in various organs (Table 1), including epithelial cancers with the greatest mortality in humans (lung, colon, and pancreas). In addition, RAS mutations are associated with approximately 8% of bladder cancers (5% HRAS), 5% of head and neck cancers (HRAS), and 15% of melanomas (NRAS).

Mutations in HRAS or KRAS are clustered overwhelmingly in the 12th, 13th or, infrequently, 61st codons. In melanoma, the vast majority of the NRAS mutations are in the 61st codon. No single amino acid substitution predominates in any of the human tumor types (Table 1). Thus, no single mutation, including KRAS codon 12ASP, represents over 45% of the total pancreatic ductal carcinomas, while the most common single mutations for lung (codon 12CYS) and colon (codon 12ASP) represent only 15% to 16% of the total lung or colon adenocarcinomas. This problem of relatively low percentage of total tumors with a given KRAS mutation is even more problematic in lung cancer where lung adenocarcinomas represent roughly 60% of all lung cancers. These results have negative implications for developing treatments (e.g., small-molecule inhibitors or vaccines) which target a specific RAS amino acid substitution, particularly in a primary prevention setting where one will not have a lesion to sequence. This problem could be partially overcome in an interception protocol where there may be a lesion to analyze.

The three members of the RAS gene family have a key pathogenic role in cancer formation and a central role as mediators of cell signaling (4). RAS proteins are anchored on the cytoplasmic side of the cell membrane where they transduce signals from membrane-bound protein tyrosine kinase receptors such as EGFR and FGFR (4, 5). When stimulated by upstream signaling molecules, RAS proteins interact with guanine nucleotide exchange factors to replace GDP with GTP, resulting in an activated protein conformation that activates effector pathways and eventually nuclear transcription factors. The three main effectors of the RAS proteins, RAF/MEK/MAPK, PI3K/AKT, and RAL-GDS, each signal through parallel and sometimes overlapping pathways to influence cell proliferation, cell survival, cell adhesion, and cell motility (4–6).

The three most lethal epithelial cancers (lung, pancreas, colon) have a substantial incidence of KRAS mutations. Standard treatment until recently has been cytotoxic therapies. However, these are not particularly effective against RAS-mutated tumors. Thus, there have been substantial efforts to effectively target RAS for cancer treatment. Historically, this has been generally disappointing, as reviewed (7). Direct inhibition of the GTP-binding pocket proved difficult due to the high affinity of GTP in the picomolar range and relatively high cellular levels of GTP and GDP. Another potential target is the processing of RAS molecules to the cell membrane by targeting prenylation of a C-terminal CAAX-binding box. While inhibition of farnesylation proved effective for HRAS, it was ineffective for KRAS and NRAS due to alternate geranylation. Efforts to alter membrane processing or to develop inhibitors of specific mutated forms of KRAS continue (8, 9). The mutation-specific covalent inhibitors that have progressed clinically are directed against a glycine to cysteine substitution at codon 12 (8). This is the most common KRAS mutation in lung cancer, seen in approximately 40% of KRAS-mutant lung adenocarcinomas although representing <15% of all lung cancers, because fewer than 30% of lung adenocarcinomas have KRAS mutations (Table 1). This specific mutation is even less commonly observed in colon (8.6%) or pancreatic (3.8%) cancer, showing the difficulty of targeted agents in a prevention or interception setting without a lesion to sequence. A recent review clearly summarizes the state of the art of RAS therapy (10). Clinical trials showed encouraging results in lung cancer patients and moderate activity in patients with colorectal carcinoma (all with KRAS G12C), leading to the accelerated approval by the US FDA of the KRAS G12C inhibitors sotorasib and adagrasib for lung cancer with this specific mutation (11, 12). The differential sensitivity of the KRAS G12C inhibitors in KRAS-mutant lung versus colon tumors nevertheless demonstrates the complexity of using targeted agents even in a defined therapeutic setting. Finally, and not unexpectedly, patients appear to be developing resistance to these highly specific KRAS inhibitors due to the development of alternative KRAS mutations or other means (13). This is similar to EGFR inhibitors in EGFR-mutant lung adenocarcinoma (14, 15). Studies examining specific small-molecule inhibitors against a wide variety of other KRAS-mutated proteins continue preclinically but none have progressed to clinical trials (10). No prevention studies have been reported with this class of KRAS inhibitors so far. Another area of research has been the use of downstream inhibitors in the RAS pathway such as MAPK/ERK or MEK inhibitors (16–18). In general, these have proven to have limited efficacy clinically as monotherapies in KRAS-mutated tumors. Although combinations of targeted downstream and even targeting upstream proteins, e.g., various EGFR inhibitors, appear effective preclinically (10).

Another potential modality directed against KRAS-mutant tumors is the immune response, employing either immune checkpoint inhibitors (ICI) or vaccines targeting KRAS-mutant tumors. The ICIs anti—programmed death-ligand 1 (PD-L1) or anti—programmed cell death protein 1 (PD-1) are now part of standard treatment for many cancers, including those with KRAS mutations. Such use of anti–PD-1/PD-L1 antibodies is supported by PD-1 overexpression in KRAS-mutant tumors (19). Trials in metastatic NSCLC showed significant activity of ICIs in squamous cell carcinomas and in lung adenocarcinomas without EGFR mutations or ALK alterations. These would include lung adenocarcinomas with or without mutant KRAS. However, minimal to no activity is seen in colon and pancreatic cancers that commonly harbor KRAS mutations, arguing against KRAS-specific targeting by immunotherapy. The ICIs, due to their expense and side effects, seem unlikely to progress to a prevention setting. Furthermore, their activity in therapy is premised on the expectation that the growth of the tumor and the killing of tumor cells by conventional cytotoxic or targeted therapies will have exposed the individual to various tumor-related antigens (neoantigens or tumor-associated antigens). Such mechanisms seem irrelevant to a prevention setting, where minimal or no exposure to the potentially antigenic lesions has occurred.

This review discusses agents that prevent the development of KRAS- or HRAS-mutant tumors in rodent models. While the molecular driver may be the same, therapy targets an established cancer where a biopsy or tumor samples allows for molecular characterization of the lesion. In prevention (no detectable cancer or precancer lesions) or interception (early treatment of precancerous lesions) settings, the genomic features of the potential tumor or precancer are not likely to be established. Descriptions of the various animal models and the specific RAS mutations for the models are presented in Tables 2 and 3. Methods to reduce toxicities of NSAIDs and EGFR inhibitors are discussed in the section on Prevention of KRAS tumors in pancreatic models (see below).

Lung adenomas or adenocarcinomas induced by carcinogens in A/J mice or certain F1 crosses commonly have KRAS mutations (Table 3). A given carcinogen routinely causes mutations at specific base pairs (Table 3). MNU and the nicotine-derived nitrosamine ketone (NNK) cause methylation of guanine residues, resulting in G to A transitions at codon 12, yielding an amino acid change from glycine to aspartate. In contrast, vinyl carbamate (VC) yields mutations in codon 61 that are not observed in human lung tumors (Table 2). Interestingly, the polycyclic aromatic hydrocarbon benzo(a)pyrene (B(a)P) causes multiple mutations in codon 12, resulting in substitutions of cysteine, valine, or aspartate for glycine, the most common substitutions observed in human lung adenocarcinomas (Table 1; refs. 20, 21). MNU causes a specific 12th codon mutation in KRAS-mutant tumors in the mouse lung model, and it induces tumors with the same 12th codon amino acid change in HRAS in mouse skin (22) and rat breast tumor models (23, 24). Thus, while the basic chemistry and amino acid substitutions (glycine to aspartate) are the same, the specific mutated RAS gene selected is organ dependent (22). Carcinogen treatment of A/J mice leads to formation of predominantly lung adenomas, with few progressing to adenocarcinomas. To develop an adenocarcinoma model, a transgene with a dominant-negative P53 mutation is inserted into the germline of A/J mice (25). The resulting tumors following carcinogen treatment grow rapidly, appear by histopathology to be adenocarcinomas, and have various genomic amplifications and deletions. Adenocarcinomas induced by carcinogens in the P53 mutant-A/J transgenic mice have the same KRAS carcinogen-specific mutations observed in the standard A/J mice (Table 3). Although a wide variety of agents have been studied in these models, we discuss agents from five classes [glucocorticoids, retinoid X receptor (RXR) agonists, the antioxidant response element (ARE) agonist 5MeCDDO, the MEK inhibitor AZD 6244, and the EGFR inhibitor afatinib].

Wattenberg and coworkers showed that dexamethasone strongly inhibited adenoma development when administered after B(a)P (26). Subsequent studies showed that multiple glucocorticoids (dexamethasone, beclomethasone, and budesonide) were all extremely active (26, 27). Striking activity was independent of the carcinogens employed which induce different KRAS mutations. The glucocorticoid agents were quite effective when treatment was initiated 1 or 8 weeks after carcinogen (27); and glucocorticoids were highly effective in blocking adenocarcinoma formation and progression in mice bearing a dominant-negative P53 mutation (27) with toxicities that depend on dose, route of administration, and duration of use. Systemic administration in human subjects is associated with hyperglycemia, hypertension, osteoporosis, adrenal suppression, risk of infections. While inhaled steroids carry lower risks, toxicities are not eliminated (28, 29). Supplementary Table S1 lists the varying classes of effective preventive agents and their toxicities.

RXR agonists have also proven highly effective. These agents bind to the RXR receptors and then form active heterodimers with a vast group of nuclear receptors (PPAR α, δ, γ; CAR; VDR; TXR; and LXR). Similar to the glucocorticoids, multiple RXR agonists (bexarotene and LGD 268) have shown activity independent of the initiating carcinogen (30, 31), and activity is observed in mice with or without a dominant-negative P53 mutation (Table 4). Interestingly, there is some limited clinical evidence that bexarotene may preferentially affect KRAS-mutated tumors in late-stage human lung cancer (32). Common clinical toxicities of bexarotene include hypertriglyceridemia, requiring treatment, hypothyroidism, dermatitis, and headaches (33). Some of these side effects might be reduced via aerosol administration of an RXR agonist (34, 35).

The triterpenoid 5MeCDDO, which acts as a potent agonist of the ARE genes by activating nuclear factor erythroid 2–related factor 2 (NRF2) activity, strongly inhibits VC-induced tumor development in A/J mice (31). This agent is structurally complex and: (i) strongly stimulates both ARE and NRF2, and (ii) alters the nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) pathway. It is hypothesized that the effects of 5MeCDDO on NF-κB where it apparently inhibits this pathway may be driving its anticarcinogenic effects in lung cancer. One major hurdle to the development of this agent is the fact that mutations in NRF2 or the Kelch-like ECH-associated protein (KEAP)-binding protein, both of which play a role in activating the ARE (similarly to 5MeCDDO), are found in up to 25% of lung adenocarcinomas and squamous cell lung cancers (36, 37) and are associated with a poorer prognosis clinically. Furthermore, mutations in the KEAP gene are associated with enhanced tumor development in a transgenic mouse (38). This potential mixed effect on lung tumorigenesis (39) probably precludes its clinical use.

Another effective compound is the MEK inhibitor AZD6244. Inhibition of this pathway is related to KRAS because it is part of a major downstream effector pathway (MEK/ERK; ref. 40). AZD6244 is a competitive inhibitor of the active ATP site of MEK, is relatively short-lived (t_½ in humans is 8 hours, t_½ in mice is 2–3 hours) and is administered twice daily in humans. Preclinical studies show that KRAS-mutant tumor cell lines are preferentially sensitive to MEK/ERK inhibitors (40). Daily AZD6244 dosing in the A/J mouse lung model achieved a 75% decrease in tumor volume following daily dosing (Table 4). However, the MEK/ERK inhibitors do not appear to be highly effective as monotherapy or in combination with chemotherapy in human KRAS-mutated lung cancer (16).

One agent that appears to have some activity is the suicide substrate afatinib, which inhibits EGFR/HER1, HER2, and HER4 (41). The relatively selective EGFR inhibitors (gefitinib and erlotinib) are used clinically in the treatment of EGFR-mutant but not KRAS-mutant lung cancers. Studies found afatinib (42) inhibits tumor growth in KRAS-mutated lung tumor xenografts and in a transgenic model that develops KRAS-mutated lung adenocarcinomas (42). The rationale is human lung cancers and lung tumors from transgenic mice with KRAS mutations overexpress the EGFR/HER2 pathway, including HER2,3,4 and related ligands (EGF, epiregulin, and amphiregulin; ref. 42). Various agents are ineffective in the chemically induced lung tumor model, including the COX inhibitors (NSAIDs or Coxibs) when administered during the progression stage.

In summary, A/J mouse lung tumors respond during in situ development to agents (glucocorticoids, RXR agonists, and 5MeCDDO) that may not specifically alter the RAS pathways. Furthermore, their efficacy appears to be independent of specific KRAS mutations. The MEK inhibitor selumetinib and the pan EGFR inhibitor afatinib are also relatively effective. These inhibitors appear more directly related to KRAS mutation - selumetinib inhibits the downstream target MEK, while afatinib inhibits proteins in the EGFR pathway that appear to be activated in tumors with KRAS mutations (42). The ability to predictably induce tumors with specific mutations in the 12th, 13th, or even 61st codon in the A/J mouse model (Tables 3 and 4) may prove useful in testing small-molecule inhibitors or vaccines directed against specific amino acid. Thus, protocols which specifically affect an aspartate substitution induced by MNU or NNK should not work against tumors induced by VC.

Colon/Intestinal cancers are the second most common cause of cancer death. Mutations in the Wnt pathway are found in >80% of colon cancers and are overwhelmingly truncations of the APC gene in humans. However, >45% of human colon cancers and sporadic advanced adenomas have mutations in KRAS. Interestingly, although KRAS mutations are fairly common, in colon adenomas and even more common in aberrant crypt foci, mechanistically they appear to contribute to later stages in tumor development, including invasion and metastasis (43). The two common rodent models of chemically induced colon cancer use 1, 2 dimethylhydrazine or its activated metabolite AOM, resulting in methylation-induced mutations in DNA. Similar to humans, virtually all of the tumors in the rodent models develop Wnt-related mutations, although unlike humans, mutations are in the downstream protein β-catenin (44). However, APC truncations or missense mutations of β-catenin activate the Wnt pathway by increasing levels of the transcription factor TCF4. The carcinogen-induced rat models have KRAS mutations in approximately 30% to 50% of tumors (44, 45). Most studies have reported a high incidence of G to A transitions in codon 12 and less commonly codon 13, resulting in an amino acid change from glycine to aspartate. In humans, the aspartate substitution is observed in approximately 35% of KRAS-mutated colon adenomas or cancers. Multiple agents are strikingly effective in preventing these tumors, including various COX1/2 inhibitors (46, 47) and the ornithine decarboxylase inhibitor difluoromethylornithine (DFMO; ref. 47). The combination of a COX inhibitor/NSAID plus DFMO is profoundly effective in animal models, reducing tumor multiplicity >85% (47).The combination of the NSAID (sulindac) plus DFMO inhibited the development of new adenomas and advanced adenomas by >70% in a human phase II prevention trial (48). Because KRAS mutations are present in >40% of adenomas, this implies that the combination was effective in reducing KRAS-mutant lesions in humans, although this was not formally addressed by sequencing. This question was directly addressed in the AOM colon model where lesions were examined for the presence of specific KRAS mutations. It was shown that an NSAID/Cox1/2 inhibitor (piroxicam) alone, DFMO alone, or the combination of piroxicam and DFMO decreased the multiplicity of both KRAS-mutant tumors, as well as tumors without KRAS mutations (ref. 46; Table 5). Treatment with sulindac in a transgenic model showed that mouse colon tumors with relatively high expression of mutant KRAS are less sensitive to NSAIDs than tumors without KRAS mutations (ref. 49; Table 5 and Supplementary Table S1). The interpretation of these studies is complicated because transgenic mice with the KRAS mutations had a striking increase in tumor size, total tumor burden, and advanced histology compared with mice with tumors lacking mutant KRAS. Furthermore, mice with mutant KRAS tumors were sensitive to sulindac albeit less sensitive than transgenic mice without a KRAS mutation (Table 5). The combination of NSAIDs and DFMO are profoundly effective in a model of intestinal carcinogenesis (Min mice with APC germline mutations) that does not spontaneously develop RAS mutations (50). It appears intermittent dosing of COX2/1 inhibitors strikingly reduces gastric toxicity with no decrease in efficacy (51). While a variety of NSAIDs are effective in both colon and pancreatic models (Table 5), lower doses of aspirin which minimally inhibits COX 2 but does inhibit COX-1 is ineffective in most animal models. The RXR agonist bexarotene (52) and the ARE Agonist 5MeCDDO (53) are ineffective in reducing colon cancer development in the AOM model. These results show that the efficacy of these agents in lung tumors but not colon tumors with KRAS mutations is not dependent on RAS mutations per se, but rather appears associated with organ-specific prevention (lung vs. colon; refs. 31, 53).

Ductal pancreatic cancer (Table 1) has the highest incidence of KRAS-mutations, including preinvasive lesions. Two models of pancreatic cancer are routinely used in prevention studies. The carcinogen-induced (N-nitroso-bis-(2oxopropyl) amine) model in hamsters creates ductal tubular tumors similar to most human pancreatic tumors (54, 55). Virtually 100% of the tumors have KRAS mutations that are observed even in preinvasive lesions. The tumors have G to A transitions, resulting in substitution of aspartate for glycine, the most common amino acid substitution in human pancreatic cancer. The prevalence and early development of the KRAS mutations, similar to humans, would argue that KRAS is a gatekeeper in pancreatic cancer and is necessary for cancer development. These hamster cancers lose p16 function, which is similarly lost in approximately 85% of human ductal pancreatic cancer (54, 55). There is a transgenic mouse model of pancreatic cancer with a knock-in of a mutation of KRAS in codon 12 (56). The model develops adenocarcinomas over a relatively long latency period, which can be markedly shortened with additional alterations in INK4A [p16] and/or P53. Both models have proven sensitive to the preventive activity of COX inhibitors. Thus, the COX inhibitors celecoxib, nimesulide, high dose aspirin, indomethacin and licofelone were effective in either the hamster or transgenic mouse models (57, 58) and appear significantly effective in intercepting and blocking tumor progression even after the development of initial PanIN lesions which routinely have KRAS mutations. The combination of licofelone (a combined COX/LOX inhibitor) and the EGFR inhibitor gefitinib blocked development of all pancreatic adenocarcinomas in the mouse transgenic model (57). A similar combination of an NSAID (sulindac) and an EGFR inhibitor (erlotinib) has proven highly effective in the colon/duodenum in a clinical familial adenomatous polyposis (FAP) trial (59). Awareness of the potential human toxicities of each component allowed modulation of these side effects in the combination either by administering NSAIDs intermittently and or administering an EGFR inhibitor in a pulsatile manner (51). The NSAIDs are probably the best known group of agents to work in a wide variety of organs (colon, pancreas, bladder, head and neck, esophagus, etc.) in the preclinical setting (46), where a wide variety of NSAIDs are similarly effective (Table 5) arguing at least indirectly that they all work by a similar mechanism. Up until 10 years ago, the mechanism of NSAID efficacy was open to some variety of interpretations often based on in vitro studies at very high doses. However, it now appears that the primary mechanism of the NSAIDs in vivo is altering the immune response which facilitates the natural immunity of the host to respond to the tumor cells. Thus, treatment with COX inhibitors (naproxen, sulindac, celecoxib, and higher doses of aspirin) have been shown to have major effects on a variety of immune parameters in cells infiltrating tumors, including increases in CD4 and CD8 lymphocytes in lesions and the expression of cell surface biomarkers associated with lymphocyte activation (60). Furthermore, NSAIDs have been associated with reduced expression of PD-L1 in lesions in mice (61). Finally and perhaps most importantly, the number of CD4 and CD8 lymphocytes is increased in adenomas of persons taking NSAIDs (62).

Estrogen receptor–positive (ER⁺) mammary cancers develop in adolescent Sprague–Dawley rats following treatment with the polycyclic hydrocarbon DMBA, or following MNU treatment (63). HRAS mutations are carcinogen-specific: MNU, mutations only at codon 12 in at least 50% of tumors; DMBA, mutations only at codon 61 in approximately 1/3 of tumors (23, 64). These rapidly arising adenocarcinomas are similar to highly differentiated ER⁺ breast adenocarcinomas in women both histologically and by array analysis (65). RAS mutations are rare in human ER⁺ breast cancers, although mutations in the downstream effector protein PI3K are relatively common. Data with the MNU-induced model are included because the model has been extensively employed and the HRAS mutated tumors illustrate many of the characteristics observed with prevention and interception of RAS mutated tumors: multiple classes of agents are effective; various agents of a given class are effective; a generalized anti-RAS agent (e.g., tipifarnib for HRAS) can effectively and preferentially block RAS mutated tumors; agents which are highly effective in the model are effective against both RAS mutated tumors and RAS wild-type tumors.

Like human ER⁺ breast cancers, the carcinogen-induced rat cancer model responds both in preventive and interception settings to selective estrogen receptor modulators (SERM) and aromatase inhibitors (66, 67). The preventive and interceptive activity of either of these classes of agents is virtually 100%, implying efficacy in tumors with or without HRAS mutations. MNU-induced tumors are highly responsive to the preventive and therapeutic effects of HER1/EGFR inhibitors (erlotinib and gefitinib) (68, 69). This parallels the finding that gefitinib is relatively effective in the therapy of ER⁺/PR⁺ breast cancers in women in a neoadjuvant setting (70) and strongly inhibits cell proliferation (Ki67) in a presurgical setting (71). Thus, the rat MNU model is responding similarly to agents known to be effective in human ER⁺ breast cancer.

Gottardis and coworkers found that the RXR agonist bexarotene was strikingly effective in the MNU model (66). Subsequently, investigations showed that multiple RXR agonists are highly effective (72). The RXR agonists are also effective in treatment of palpable breast cancers in the MNU model (73, 74). The mechanism by which these agonists work is complex because they can form heterodimers with various nuclear receptors and their coactivators/corepressors in an organ- and cell type–specific manner. This potentially modulates how they affect gene expression (75, 76) and the physiologic/pharmacologic processes that mediate their cancer preventive activities in breast and lung, as shown below. Tipifarnib, which inhibits the farnesylation of a wide range of proteins and especially HRAS, was tested for efficacy in the MNU-induced model (23, 77). Although tipifarnib could block development of all the tumors at higher doses, tumors with HRAS mutations could be prevented or treated with lower doses (23). When palpable lesions are treated with tipifarnib, there was greater inhibition of proliferation and higher levels of apoptosis in HRAS mutated versus wild-type tumors (23). Gene expression analyses of treated tumors (77) at higher doses required for prevention of most tumors showed striking inhibition of multiple proliferation-related pathways in HRAS-mutated tumors. While in HRAS–wild-type tumors at these higher doses there were clear alterations in certain G protein–related pathways, but minimal effects on proliferative pathways (77).

These results show that multiple classes of agents have profound efficacy on both HRAS-mutated and wild-type breast tumors. Furthermore, their efficacies reflect class-specific effects on major targets: (i) The highly effective SERMs tamoxifen (78), toremifene, basodoxifene, arzoxifene, and the aromatase inhibitors vorozole or letrozole; (ii) RXR nuclear receptor agonists bexarotene, UAB-30, 4-Me UAB-30, or LGD 268, or (iii) EGFR/HER2 receptor antagonists gefitinib, erlotinib or lapatinib. These agents prevent the development of virtually all tumors, implying their strong efficacy is independent of HRAS mutations, and indicating that HRAS-mutated tumors can be prevented/intercepted by agents that do not directly target RAS. In contrast, the COX1/2 inhibitors, which are highly effective in the prevention of KRAS-mediated cancers of the colon or pancreas or HRAS-mutant tumors in a mouse skin model (22), are relatively ineffective against HRAS-mutated breast cancer in rats or KRAS-mutated lung cancer in mice. Characteristics (specific agents, mechanisms of action, known human toxicities) of classes of agents that show consistent preventive or interceptive activity are presented in Supplementary Table S1.

Another approach to prevention involves vaccines specifically directed against KRAS-mutant tumors. The rationale is that mutated KRAS is a neoantigen, not expressed in normal cells, and therefore immunogenic. Mutant RAS-specific peptides can be processed and presented as foreign antigens by both MHC class I or II molecules (79). In previous preclinical and clinical studies, KRAS-mutant–specific CD4 and/or CD8 immunogenic peptide epitopes were identified. For example, a 9-mer peptide covering the G12V mutation activates both CD4⁺ and CD8⁺ T-cell responses in mice (80). There have been various preclinical reports on vaccines against KRAS. Investigators have used a variety of peptide or DNA vaccines to achieve mutation specific immune responses. The problem with comparing the results is that investigators use different immunologic tests, different cancer models for determining efficacy, different adjuvants, etc. We present briefly two such studies as illustrative, not arguing for or against these approaches. Wan and colleagues (81) employed a short peptide coding for G12D (glycine to arginine) linked to diphtheria toxin. Mice (Balb/c) were immunized 3X in alum and challenged in the flank with the syngeneic lung tumor (CT26). 87.5% and 50% of vaccine-treated mice in the preventive and therapeutic models were tumor free, which was associated with the development of TH-1 immunity against the immunizing peptide. Thus, this syngeneic graft allows one to fully vaccinate prior to challenge with the syngeneic tumor. Jaffee and coworkers (82) immunized against pancreatic cancer in a transgenic model employing attenuated Listeria monocytogenes engineered to express the KRAS G12D mutation. The transgenic knock-in mouse pancreatic cancer model employed was similar to what we described above (see pancreas section), although the authors introduced a P53 knockout to enhance development of tumors. The vaccine alone elicited an immune response. However, when given to weanling transgenic mice alone, it failed to inhibit tumor development. In this case, vaccination is initiated against lesions growing in situ. But when administered immediately following low dose cyclophosphamide to reduce T regulatory cells and combined with an immune modulatory antibody, it conferred a significant survival advantage when given at the time of early PanIN 1 development but not at the time of mid or late PanINs. These two studies illustrate the variety of approaches: (i) vaccination prior to challenge with a lung syngraft with a KRAS mutation given subcutaneously versus vaccination of a tumor developing in situ in the pancreas; (ii) in the syngraft, vaccination is complete prior to challenge versus vaccination of animals developing early lesions; and (iii) in the case of the transgenic mice, the use of low dose cyclophosphamide to reduce suppressor T cells. With such wide variety of differences between the models, one cannot compare the efficacy of the two vaccines.

There are published clinical data using a KRAS-mutant peptide as a vaccine. The preponderance of patients with pancreatic cancer demonstrated peptide-specific immunity to longer peptides (15 amino acids) which included the mutant region (codon 12) of KRAS (79, 83). Patients with advanced disease who developed an immune response against the vaccine experienced a longer survival compared with nonresponders (141 vs. 61 days; P = 0.0002; ref. 83). After 10 years of follow up of 23 patients who were vaccinated against mutated KRAS in two phase I/II studies, the presence of persistent immune memory and a 20% survival at 10 years were reported (83). However, most of the clinical KRAS vaccine trials to date have employed short HLA-restricted peptides and were examined for safety and immunologic response and not efficacy response per se. As described above, any specific KRAS amino acid changes represent <10% of the total cancers for most cancers (lung, colon) but not pancreatic cancer. This is not an impossible hurdle in therapy where one has tumor to sequence but is unsurmountable in a prevention setting. Both the preclinical and clinical data demonstrate that one can raise an immune response against KRAS and observe some antitumor effect. Nevertheless, the fact that different investigators employ different models, different vaccines and different adjuvants makes any rational choice of one specific vaccine as contrasted with another difficult.

Pan and colleagues evaluated both the immunogenicity and preventive efficacy of a multipeptide vaccine targeting four epitopes of the KRAS molecule in a mouse model of a KRAS-driven lung tumors (84). Four longer (15–20 amino acid) peptides were used that were not HLA-restricted. Two of the four peptides employed were independent of the mutation site in codon 12, while the other peptides were a pair of 17-amino acid peptides overlapping the codon 12 mutation site coding for the wild-type or the mutated peptides. This approach was based on prior preclinical efficacy studies with multivalent long peptide vaccines (85, 86). A long (15–17 amino acids) multipeptide vaccine not directed against the EGFR mutant epitope was profoundly effective in an EGFR-mutant model of lung cancer in mice (86). In the KRAS vaccine (84), long peptides with 100% homology between human and mouse peptides were developed. The rationale for peptides with 100% homology includes: (i) that any question of tolerance may be similar for both humans and mice, and (ii) that using the same peptides in both species makes performing FDA toxicology studies easier. A few important findings arose from this study. First, the vaccine was strikingly effective in preventing tumor development when administered prior to doxycycline in the Varmus model of KRAS-induced lung cancer (87) where most of the cells in the lung rapidly overexpress the immunosuppressive mutated KRAS oncogene following doxycycline initiation, and the vaccine reduced total tumor load >80%. Second, the peptides (15–17 amino acids) coding for both the mutant region (peptide 1 AA 5–21 wild-type; peptide 2 AA 5–21 mutated), and the nonmutant region (peptide 3 AA17–31; peptide 4 AA78–92) elicited a strong immune response, as monitored by the production of interferon gamma. These specific peptides were chosen on the basis of a combination of computer algorithm which predicts likely MHCI and II reactive peptides, and testing of immunologic response to individual peptides in naïve mice. Third, in the 15–17 amino acid peptides where 100% homology between human and mouse peptides was achieved, this is an advantage because these would be the peptides that would go directly into any clinical studies. Fourth, the peptides employed elicited primarily a Th1 and Th17 T-cell response, and not a Th2 T-cell response, which is thought to be potentially immunosuppressive. Thus, it appears that KRAS-peptide-based vaccines targeting not only mutant KRAS G12D, but also that various non-mutated epitopes, are capable of inducing robust Th1 immune responses that lead to significant inhibition of mutant KRAS-driven lung tumorigenesis. This vaccine has also proven to be significantly effective in intercepting tumor development in the VC driven A/J model with a different KRAS mutation (Table 3) when vaccination was initiated after administration of VC (88).

There are a number of clear findings in preclinical models:

• 1. Various agents exhibit organ-specific efficacy and prevent and or intercept the development of RAS-mutated tumors in situ. Multiple members of effective classes are all efficacious in a given organ site. This latter finding is an indirect measure of the reproducibility of the assays (84) and implies that another member of the same mechanistic class will similarly show efficacy. Thus, various RXR agonists are effective in the MNU-induced rat mammary model (HRAS mutations in 12th codon; Supplementary Table S2) and mouse lung (KRAS mutations at multiple sites (Table 4) dependent on the carcinogen employed) tumors. A great variety of COX inhibitors are effective in the AOM-induced colon cancer model (Table 5) in which roughly 40% of the tumors have KRAS mutations. The efficacy of multiple NSAIDs in prevention or interception of non–RAS-mutant tumors in bladder and UV-induced squamous cell cancers of the skin has also been shown.

• 2. Most agents are effective against both RAS mutated and RAS–wild-type tumors in a given organ (colon and breast) and against different RAS mutations in a given organ (lung). Preventive activity appears to be organ-dependent, most agents appear to be effective against both wild-type and RAS-mutant tumors (colon and breast). In the MNU breast model, hormonal agents, RXR agonists, EGFR inhibitors are profoundly effective irrespective of whether tumors are HRAS-mutated or wild-type. NSAIDs and DFMO and their combination are effective against both wild-type and KRAS-mutated colonic tumors. The combination of an NSAID and DFMO or the combination of an NSAID and an EGFR inhibitor could be tested in FAP and those lesions that failed to regress or grew could be sequenced for RAS mutations and compared with lesions in placebo treated group. One problem if one samples most adenomas at the beginning of the study, there might not be enough tissue to follow that lesion to the end of the study.

• 3. In the lung, agents are effective in prevention or interception of tumors with different RAS mutations (glucocorticoids or RXR agonists; Table 4). Because lung cancer is a mixture of subtypes, e.g., adenocarcinoma (WT, KRAS-mutant, EGFR, or ALK alterations), squamous cell carcinoma, it is hard to imagine a true prevention study without agents expected to be effective against this wide variety of lesions. If there is a clear lesion to examine, then an interception protocol could be applied. One exception to this might be CT scan–detected lung lesions in individuals from Eastern Asia where EGFR mutated adenocarcinomas are so common that you might employ an EGFR inhibitor for all (this is not KRAS-driven cancer).

• 4. In both the hamster and mouse transgenic models of pancreatic cancer, COX inhibitors are effective. Furthermore, the combination of licofelone (a combined COX/LOX inhibitor) and the EGFR inhibitor gefitinib appear to prevent virtually every pancreatic tumor in a transgenic model. This combination has proven to be highly effective in colon both in animal models (51) and in human FAP trials (59, 89). These data imply that this combination may translate in other organs as well. It is necessary to establish preclinically whether this works with a more standard NSAID (not licofelone) and whether interception could be initiated when preinvasive lesions already exist. Initiating studies when pancreatic lesions are relatively early is important because the EGFR pathway appears quite significant in early stages of pancreatic tumor development. Because therapy of pancreatic cancer is difficult, an initial study could be a presurgical study in early pancreatic cancer examining alterations of genomic expression.

• 5. KRAS vaccines in a prevention setting are appealing. One question remaining regarding the long peptide vaccines discussed above is whether they are effective against various mutant KRAS proteins or they are specific for a single amino acid change. If they are mutation specific, then a combined vaccine would be required that covers the various mutations associated with a given form of cancer (Table 1). If they are effective against multiple KRAS mutants, this makes the approach more feasible. The most likely initial clinical trial would determine whether a Th1 mediated peptide specific immune response can be elicited.

K.H. Dragnev reports other support from Eli Lilly, Merck, Daiichi Sankyo, Molecular Templates, Io Therapeutics, G1 Therapeutics, Novartis; and other support from Roche/Genentech outside the submitted work. No disclosures were reported by the other authors.

The opinions expressed by the authors are their own and this material should not be interpreted as representing the official viewpoint of the US Department of Health and Human Services, the NIH, or the NCI.