Mutations in the α4-α5 allosteric lobe of RAS do not significantly impair RAS signaling or self-association

Mutations in one of the three RAS genes (HRAS, KRAS, and NRAS) are present in nearly 20% of all human cancers. These mutations shift RAS to the GTP-loaded active state due to impairment in the intrinsic GTPase activity and disruption of GAP-mediated GTP hydrolysis, resulting in constitutive activation of effectors such as RAF. Because activation of RAF involves dimerization, RAS dimerization has been proposed as an important step in RAS-mediated activation of effectors. The α4-α5 allosteric lobe of RAS has been proposed as a RAS dimerization interface. Indeed, the NS1 monobody, which binds the α4-α5 region within the RAS G domain, inhibits RAS-dependent signaling and transformation as well as RAS nanoclustering at the plasma membrane. Although these results are consistent with a model in which the G domain dimerizes through the α4-α5 region, the isolated G domain of RAS lacks intrinsic dimerization capacity. Furthermore, prior studies analyzing α4-α5 point mutations have reported mixed effects on RAS function. Here, we evaluated the activity of a panel of single amino acid substitutions in the α4-α5 region implicated in RAS dimerization. We found that these proposed “dimerization-disrupting” mutations do not significantly impair self-association, signaling, or transformation of oncogenic RAS. These results are consistent with a model in which activated RAS protomers cluster in close proximity to promote the dimerization of their associated effector proteins (e.g., RAF) without physically associating into dimers mediated by specific molecular interactions. Our findings suggest the need for a nonconventional approach to developing therapeutics targeting the α4-α5 region.

Mutations in one of the three RAS genes (HRAS, KRAS, and NRAS) are present in nearly 20% of all human cancers. These mutations shift RAS to the GTP-loaded active state due to impairment in the intrinsic GTPase activity and disruption of GAP-mediated GTP hydrolysis, resulting in constitutive activation of effectors such as RAF. Because activation of RAF involves dimerization, RAS dimerization has been proposed as an important step in RAS-mediated activation of effectors. The α4-α5 allosteric lobe of RAS has been proposed as a RAS dimerization interface. Indeed, the NS1 monobody, which binds the α4-α5 region within the RAS G domain, inhibits RAS-dependent signaling and transformation as well as RAS nanoclustering at the plasma membrane. Although these results are consistent with a model in which the G domain dimerizes through the α4-α5 region, the isolated G domain of RAS lacks intrinsic dimerization capacity. Furthermore, prior studies analyzing α4-α5 point mutations have reported mixed effects on RAS function. Here, we evaluated the activity of a panel of single amino acid substitutions in the α4-α5 region implicated in RAS dimerization. We found that these proposed "dimerization-disrupting" mutations do not significantly impair self-association, signaling, or transformation of oncogenic RAS. These results are consistent with a model in which activated RAS protomers cluster in close proximity to promote the dimerization of their associated effector proteins (e.g., RAF) without physically associating into dimers mediated by specific molecular interactions. Our findings suggest the need for a nonconventional approach to developing therapeutics targeting the α4-α5 region.
RAS GTPases are important mediators of intracellular signaling cascades that regulate cell proliferation and survival (1,2). The three RAS genes (HRAS, KRAS, and NRAS) encode four different protein isoforms: HRAS, splice variants KRAS4A and KRAS4B, and NRAS. Each of the RAS isoforms cycle through active (RAS-GTP) and inactive (RAS-GDP and apo-RAS) states that are tightly regulated by GTPase activating proteins (GAPs) and guanine nucleotide exchange factors (3). Prenylation and palmitoylation of cysteine residues of the C-terminal hypervariable region (HVR) of RAS localize it to the inner leaflet of the cell membrane where RAS is typically activated in response to growth factor receptor-mediated stimulation (4). However, activating mutations of RAS stabilize the active RAS-GTP state even in the absence of upstream stimulation (1). Notably, nearly 20% of all human cancers harbor such RAS mutations, making RAS the most frequently mutated oncogene as well as a valuable target for cancer therapy (5).
Clinically available inhibitors of RAS have been elusive. However, the recent FDA approval of the KRAS G12C inhibitor, sotorasib, illustrates the feasibility of pharmacological inhibition of KRAS (6,7). Unfortunately, KRAS G12C represents only a fraction of oncogenic RAS mutations in human cancers (8,9), presenting a critical need for alternative RAS inhibitors. Due to the picomolar affinity of RAS for guanine nucleotides, targeting the nucleotide binding pocket has been considered pharmacologically problematic (10), although recent work provides support for the possibility of targeting the nucleotidefree state of RAS (11). In the absence of direct RAS inhibitors, interfering with RAS membrane localization has been explored as an alternative therapeutic approach (12). However, the ability of KRAS to be alternatively prenylated in response to farnesyl transferase inhibitor treatment rendered this approach ineffective for KRAS-mutant cancers (13), although these inhibitors have shown some efficacy in HRAS-mutant malignancies (14).
More recent studies have highlighted the possibility of inhibiting RAS nanoclustering as a potential approach to inhibit RAS-mediated signaling and biological activity both in vitro and in vivo (15)(16)(17)(18)(19). Utilizing monobody technology (20), we previously reported that the NS1 monobody specifically binds the α4-α5 allosteric lobe of HRAS and KRAS, but not NRAS, and inhibits RAS dimers/nanoclusters (15). Additionally, the K13 and K19 DARPins inhibit KRAS via binding of the α3-α4 allosteric lobe and disrupting RAS dimer/ nanoclusters (17). The importance of RAS dimers/nanoclusters was further highlighted by the endogenous RAS antagonist, DIRAS3, which inhibits RAS nanoclusters through binding of α5 region of RAS (21). These results are consistent with the reported presence of dimers of RAS in a number of crystal structures of RAS (15,(22)(23)(24), the GTP-dependent dimerization of the KRAS G domain (25), and the observation that artificial dimerization of the isolated G domain of RAS is sufficient to activate RAF in vitro (26) and increase mitogen-activated protein kinase (MAPK) signaling in cells (27). In contrast, disruption of these higher-order RAS assemblies inhibits downstream signaling and biological function in vivo (15)(16)(17)(18)(19).
Although these examples implicate RAS dimerization as an on-pathway event in RAS-mediated signaling and provide a rationale for drug discovery efforts aimed at development of RAS dimerization inhibitors, the importance of RAS dimerization remains a point of much debate. While active RAS stimulates the dimerization and activation of RAF kinases in the MAPK pathway (28), the G domain of HRAS lacks the propensity to dimerize in solution (23). Furthermore, the isolated HVR of KRAS, devoid of the G domain, is sufficient to drive dimerization of a fluorescent protein-HVR fusion protein in cells, suggesting that KRAS G domain is dispensable for dimerization (27). In light of these studies, many attempts have been made to identify binding interfaces that facilitate higherorder RAS assemblies. Three major RAS dimerization interfaces have been proposed: the α3-α4 interface, the α4-α5 interface, and the β-sheet interface (29). Experimental evidence supporting the validity of these dimerization interfaces, however, has been conflicting. Although RAS was inhibited by the NS1 monobody, point mutations within the NS1-binding interface of HRAS predicted to disrupt dimerization (e.g., R135A, D154Q, R161D) did not impair downstream MAPK pathway activation (15). Similarly, a recent study reported that the KRAS D154Q mutation did not affect RAS-RAS interaction in cells (30). In contrast, Ambrogio et al. provided evidence that mutation of D154Q or R161E decreased KRAS association (as measured by FRET), CRAF-BRAF association, and KRASmediated signaling and tumor formation in vivo (24). More recent work has suggested that RAF RBD engagement promotes RAS dimerization through the α4-α5 region although mutations in this region did not affect RAS dimerization (31).
Because of the contradictory results with these studies, we sought to characterize a panel of oncogenic HRAS, KRAS, and NRAS mutants to assess the effects of α4-α5 mutations on oncogenic RAS function. These mutations were selected based on previous reports of their importance in RAS dimer formation (15,22,25,(32)(33)(34). Here, we show that mutations in the α4-α5 allosteric lobe do not affect RAS association in cells. In addition, these mutants did not impair MAPK activation compared with the parental oncogenic RAS. Although downstream signaling was unaffected by these mutations, we observed isoform-specific differences in the transforming activity of selected mutants. Overall, our results are consistent with a model in which RAS protomers associate in close proximity to promote effector dimerization (e.g., RAF) without formation of molecularly defined RAS dimers. Further, we propose that the ability of selected biologics such as NS1 monobody and K13/K19 DARPins to perturb RAS function is due to steric hinderance of RAS rather than disruption of bona fide dimers. This study provides further insights into RAS oligomerization and should inform efforts in developing therapeutics directed at the α4-α5 allosteric lobe of RAS.

Results
Mutations in the α4-α5 region of KRAS do not impair oncogenic activity Given the contradictory reports on the importance of dimerization in RAS function, we tested whether mutations in the proposed dimerization interface (Fig. 1, A and B; termed α4-α5 mutants hereafter) impaired MAPK pathway activation relative to the parental oncogenic KRAS mutant. To reduce overexpression artifacts, we determined the appropriate amounts of K/H/NRAS DNA that yielded protein expression and ERK phosphorylation below the point of saturation for the signaling assays (Figs. S1-S4). Surprisingly, there were no significant differences in MAPK activation between the KRAS G12V α4-α5 mutants and the parental oncogenic KRAS G12V with the exception of KRAS G12V/K147D , a mutation that has been shown to decrease RAS-GTP levels (Figs. 2, A and B, and S5) (35,36).
Next, we analyzed CRAF-BRAF association in cells expressing KRAS G12V versus KRAS G12V harboring α4-α5 mutants, given the well-established role of RAF dimerization in MAPK signaling (37,38). In addition, a previous study demonstrated that D154Q and R161D KRAS mutants decreased CRAF-BRAF heterodimers (24). In agreement with the results from the MAPK signaling assays described above, there was no significant impairment in CRAF-BRAF interaction in cells expressing the KRAS α4-α5 mutants compared with parental KRAS G12V (Fig. 2, C and D). These data indicate that the α4-α5 mutations do not impair the activation of the canonical RAS/MAPK pathway mediated by KRAS G12V .
Lastly, we performed transformation assays in NIH/3T3 cells transfected with KRAS G12V or the KRAS G12V α4-α5 mutants (Fig. 2, E and F). Consistent with the signaling data, the KRAS G12V α4-α5 mutants retained the ability to transform cells as well as parental KRAS G12V . Taken together, these results suggest that mutations of amino acid residues proposed to be critical for KRAS-KRAS self-association do not significantly impair the signaling or transforming properties of oncogenic KRAS.
Mutations in the α4-α5 allosteric lobe do not affect KRAS association in cells Next, we addressed whether these α4-α5 mutations affected RAS-RAS association in cells. We employed Live-Cell Nano-Luc Binary Technology (NanoBiT), a protein-protein interaction (PPI) detection system where one protein partner is tagged with an 11-amino acid peptide (SmBiT), while the other protein partner is tagged with a 17.6 kDa NanoLuc fragment α4-α5 mutations do not impair oncogenic RAS signaling (LgBiT). When expressed in cells, PPIs between the two protein partners allows for complementation of the SmBiT and LgBiT tags to generate a luminescent signal upon substrate addition.
To avoid potential interference of endogenous RAS proteins with the SmBiT/LgBiT-tagged KRAS protein partners, we performed these assays in RAS-less MEFs transformed with BRAF V600E , a cell line that lacks all 3 RAS isoforms (39). To account for variations between transfections, cells were lysed after measuring the live-cell luminescence (Fig. S8A), and LgBiT-tagged proteins were quantified using HiBiT, an 11 amino acid peptide with high affinity (K D = 0.7 nM) for the LgBiT peptide. The HiBiT peptide out-competes the SmBiT peptide for LgBiT binding while still generating luminescent signal, allowing for a fast and sensitive method to quantify total LgBiT peptide levels in cells (Figs. 3A and S8B). When coexpressed in cells, SmBiT-KRAS G12V and LgBiT-KRAS G12V reconstituted luciferase activity (Fig. 3, A and B). In contrast, coexpression of SmBiT-KRAS G12V with EGFR-LgBiT, also a membrane-localized protein which served as a negative control, generated a weak luminescent signal (Fig. 3B). The introduction of the α4-α5 mutations to LgBiT-  50)). The Switch I and II regions are colored in tan. The α4 and α5 helices in protomer 1 are labeled, and those in protomer 2, that is, the symmetry-related copy, are labeled as α4 0 and α5 0 . The residues subjected to mutational studies are shown in cyan and labeled for protomer 1 and shown in blue for protomer 2. The side chains of E49, K128, and R135 are disordered in the 6VC8 model and thus not depicted. B, RAS α4-α5 mutants included in this study. . All experiments were repeated at least three times (n = 3) and results quantified using Welch's t test; error bars representing SEM (***p < 0.0005, **p < 0.005, and *p < 0.05). MAPK, mitogen-activated protein kinase.
α4-α5 mutations do not impair oncogenic RAS signaling KRAS G12V resulted in no significant decreases in luciferase activity (Fig. 3B). These results are consistent with the results from signaling assays (Fig. 2), indicating that not only do the α4-α5 mutations have little to no effect on KRAS signaling and biology, they also do not impair RAS-RAS interactions in cells.
RAS α4-α5 mutants are still susceptible to inhibition by the NS1 monobody NS1 inhibits oncogenic RAS-mediated signaling, biological transformation, and tumor formation (15,18,19) and disrupts higher-order RAS associations at the plasma membrane (15). Through binding of the α4-α5 region of RAS, NS1 allosterically inhibits RAS function irrespective of its nucleotide state (15,16). We hypothesized that if the α4-α5 mutations decreased RAS-RAS interaction in cells, then NS1 would have less of an inhibitory effect on these mutants compared with the parental RAS oncogenic mutants. As illustrated in Figures S6  and S7, α4-α5 mutations in KRAS G12D did not impact ERK activation. Furthermore, coexpression of NS1 effectively inhibited downstream ERK phosphorylation in cells expressing KRAS G12D and KRAS G12D α4-α5 mutants (Figs. 4A and S9). The inhibitory effect of NS1 was highly specific as KRAS G12D/ R135E , which does not bind NS1 (15), was refractory to inhibition by NS1 (Figs. 4A and S9). These results further support the notion that these α4-α5 mutations are not sufficient to impair RAS association or oncogenic RAS-mediated signaling in cells.

HRAS and NRAS α4-α5 mutants exhibit isoform-specific biological properties
We then examined whether α4-α5 mutations in the other RAS isoforms had similar effects on their signaling and biological properties. In contrast to previously reported results with KRAS D154Q (24), we reported that mutations at D154 or R161 of HRAS G12V had no effect on ERK-MAPK activation (15), which might suggest isoform-specific effects of these mutations. Consistent with prior findings (15,30) and the results with KRAS G12V (Fig. 2), mutations at D154 or R161 had no effect on either HRAS G12V or NRAS G12V signaling (Figs. 5, A-D and S10).
Next, we assessed the impact of these mutations on the biological activity of H/NRAS G12V . Whereas the KRAS G12V mutants showed no significant differences in foci formation (Fig. 2, E and F), all of the HRAS G12V α4-α5 mutants (Fig. 5, E and F) and one of the NRAS G12V α4-α5 mutants (R161D) (Fig. 5, G and H) resulted in significantly fewer foci than those of the parental control. Together, these data show that mutations in the α4-α5 allosteric lobe of all three RAS isoforms have no effect on ERK-MAPK signaling but impair the transforming activity of HRAS G12V and to a lesser extent NRAS G12V .

Discussion
The link between RAS dimerization and RAS-mediated signaling was first observed by Santos et al. in 1998 (40) and was revisited at the turn of the 21st century (26). Indeed, artificial dimerization of RAS at the plasma membrane α4-α5 mutations do not impair oncogenic RAS signaling activated the MAPK pathway (27), while inhibition of RAS clusters at the plasma membrane is associated with inhibition of the MAPK pathway (15,21). Although there is still debate surrounding the exact mechanism of RAS nanoclustering at the plasma membrane, the α4-α5 region of RAS has been proposed in several studies to be an important interface contributing several stabilizing interactions to facilitate formation of RAS dimers (15,22,25,(32)(33)(34)41). Based on these  . RAS α4-α5 mutants are still susceptible to inhibition by the NS1 monobody. A, ERK/MAPK signaling assay in HEK 293 cells cotransfected with indicated HA-tagged KRAS G12D mutants and FLAG-tagged NS1 or a negative control monobody (Mb (neg)). Results are representative of one of three biological replicates. Graph represents the relative pERK levels. The mean and sd (n = 3) for the normalized pERK/ERK in the NS1 compared to Mb (neg) sample is shown. Dotted line at 1 represent pERK levels in Mb (neg) samples. All p values were generated using an unpaired Student's t test (***p < 0.0005, **p < 0.005, and *p < 0.05). MAPK, mitogen-activated protein kinase.
α4-α5 mutations do not impair oncogenic RAS signaling studies, mutations within the α4-α5 allosteric lobe (e.g., D154Q) have been proposed to disrupt the interactions necessary to form higher-order RAS assemblies. However, there has been conflicting data surrounding the ability of these RAS mutants to impact signaling and biology.
Consistent with our previous results (15), we found that single point mutations within the α4-α5 allosteric lobe of RAS predicted to disrupt dimer formation (Fig. 1) did not decrease downstream MAPK pathway activation compared with parental oncogenic RAS, with the exception of the KRAS G12V/ K147D mutant. K147 is an important site for posttranslational modifications that regulate RAS activity. Monoubiquitylation of K147 impedes the RAS-GAP interaction and hence GAPstimulated GTP hydrolysis, thereby favoring the RAS-GTP state (35,42,43). Furthermore, K147 acetylation regulates nucleotide binding and is associated with increased KRAS activity and tumor growth in vivo (44). Thus, the reduction in downstream MAPK signaling from KRAS G12V/K147D , especially since it did not impair self-association of KRAS (Fig. 3), could be a consequence of the K147D mutation preventing these tumorpromoting posttranslational modifications. Consistent with the MAPK signaling activity, there was no significant impairment in foci formation from the KRAS G12V α4-α5 mutants compared with KRAS G12V . Together, these results demonstrate that mutations at residues in the α4-α5 region that have been proposed to mediate RAS dimerization do not impair the oncogenic signaling or biological activity of KRAS. This suggests that if these mutations truly disrupted dimerization, then RAS dimerization, per se, may not be a necessary, on-pathway step for oncogenic RAS signaling. Conversely, if RAS dimerization is necessary for its activity, then our data suggest that these mutations do not affect RAS-RAS interactions to the extent necessary to inhibit oncogenic signaling.
This study provides evidence that not only do the KRAS α4-α5 mutants retain the ability to activate MAPK signaling, but they also do not have impaired PPIs with other RAS monomers in cells. While NS1 does not disrupt the association of RAF with HRAS, it decreases CRAF-BRAF association in cells, reflecting the ability of NS1 to sterically interfere with RAS clustering at the plasma membrane (15). In contrast, the KRAS G12V α4-α5 mutants were not significantly impaired in their ability to induce CRAF-BRAF interaction compared with KRAS G12V , which was also reflected in the MAPK signaling assays. These results were further corroborated by the Nano-BiT assays showing that the α4-α5 mutations did not affect KRAS-KRAS association in cells. Furthermore, these mutants remained sensitive to inhibition by NS1. Overall, we have shown that oncogenic KRAS is essentially unimpaired by mutations in the α4-α5 allosteric lobe.
The isoform-specific differences in transformation observed with the D154 and R161 mutations in HRAS and the R161 mutation in NRAS were unexpected but raise important questions. First, could these mutations disrupt differential effector activation from the RAS isoforms? For instance, Yan et al. (1998) reported that KRAS potently activated RAF but was less efficient at activating PI3K when compared with HRAS (45). While there is no evidence for PI3K dimerization, targeting RAS with NS1 monobody decreased downstream phospho-AKT levels (15) suggesting that disrupting RAS nanoclusters may affect multiple pathways. Second, it is possible that each isoform may utilize distinct mechanisms of nanoclustering (41). Nevertheless, given the inability of these α4-α5 mutations to impair oncogenic RAS signaling (i.e., RAF-MAPK activation), we conclude that either these mutations do not affect RAS dimerization or that RAS dimerization is not needed for RAS-induced MAPK activation.
Our results are consistent with a model in which RAS protomers rely on proximity, but not direct association with one another to form a signaling-competent complex (Fig. 6). The required proximity may be in the form of loosely associated nanoclusters where RAS protomers are close enough to promote RAF dimerization but do not require well-defined interactions between amino acid side chains of residues within the α4-α5 allosteric lobe. Disruption of these nanoclusters may require larger molecules, such as NS1 or DIRAS3, which may reduce the density of RAS on the membrane surface and/or distort the RAS-RAF complex into an inactive conformation (46). In contrast, point mutations of specific amino acid residues on the α4-α5 lobe of RAS do not appear sufficient to disrupt the interactions necessary for downstream pathway activation. The implications of these results for drug design are that smaller molecules targeting the α4-α5 allosteric lobe may be insufficient to impair downstream pathway activation. Instead, an approach to bring a large molecule to the α4-α5 region utilizing "glue" compounds (47) may be required to exploit this vulnerability of RAS.

Cell culture and cloning
Freshly thawed HEK 293 (MUSC Tissue Culture Facility) and NIH/3T3 (National Institutes of Health) cells were maintained in Dulbecco's modified Eagle's medium (DMEM) (Corning) supplemented with 10% fetal bovine serum (FBS) or 10% calf serum, respectively. RAS-less MEFs (BRAF V600E ) were obtained from the National Cancer Institute and maintained in DMEM supplemented with 10% FBS and 4 μg/ml blasticidin. RAS α4-α5 mutants were generated via site-directed mutagenesis using pCGN-HA-RAS G12V or pCGN-HA-RAS G12D as templates for each isoform. Primers used to generate RAS α4-α5 mutants are listed in Table 1. All monobodies were subcloned into CMVdriven expression vectors containing an mCherry-tag followed by a FLAG-tag on the N-terminus. KRAS G12V was subcloned into a high-expression, CMV-driven vector downstream of the SmBiT cDNA sequence. Similarly, KRAS G12V and the α4-α5 mutants were cloned downstream of LgBiT-containing vectors; however, these vectors were low-expression, HSV-driven vectors. Lastly, EGFR-LgBiT clone was in a high-expression, CMV-driven vector. All NanoBiT vectors were provided by Matt Robers and Dr Jim Vasta (Promega).

Transfections and cell signaling assays
Transfections and cell signaling assays were performed as previously described (11,48). Briefly, HEK 293 cells were α4-α5 mutations do not impair oncogenic RAS signaling transfected with HA-tagged RAS using polyethylenimine (PEI). Typically, we transfected cells using 3 μl of PEI for every 1 μg of DNA. When indicated, HA-tagged RAS was cotransfected with MYC-tagged ERK for signaling assays. Transfected cells were incubated for 30 h in complete media (DMEM with 10% FBS), then serum-starved overnight. MYC-tagged ERK was immunoprecipitated from the cell lysates using α-MYC antibody (Millipore-Sigma), then ERK and pERK levels were analyzed via Western Blot using α-ERK (Cell Signaling Technology) and α-pERK (Cell Signaling Technology) antibodies. ERK and pERK protein levels were quantified using Image Studio Lite (Ver 5.2) software. pERK/ERK ratio was determined for each mutant and normalized to the parental oncogenic RAS mutant. Each experiment was performed three times (n = 3).
To analyze CRAF-BRAF association, HEK 293 cells were transfected with the indicated RAS mutants using the same conditions as described above. After cell lysates were collected, a coimmunoprecipitation was done by pulling down endogenous CRAF using α-CRAF (BD Biosciences) antibody and probing for CRAF and BRAF via Western Blot with α-CRAF (BD Biosciences) and α-BRAF (Santa Cruz) antibodies. BRAF/ CRAF ratio was determined for each mutant and all values were normalized to the parental oncogenic RAS mutant. Each experiment was performed three times (n = 3).
For signaling assays performed with KRAS G12D and the NS1 monobody, 1 × 10 6 HEK293T cells were cultured 24 h before transfection on a 6-well plate using DMEM supplemented with 10% FBS. Cells that were between 70 and 90% confluent were then serum starved and transfected with either KRAS variants Figure 6. Model for the role of the α4-α5 region of oncogenic RAS higher-order assembly. GTP loading of RAS results in recruitment of RAF through binding of the CRD-RBD region of RAF resulting in unmasking of the dimerization interface on RAF. This results in RAF-mediated clustering of activated RAS. NS1 disrupts RAS clusters and RAS signaling through steric hinderance which prevents RAF from forming productive dimers. Mutations within the α4-α5 region are insufficient to disrupt this clustering due to a lack of electrostatic interactions between requisite amino acid side chains. Created with BioRender. com Table 1 Primers used to generate the RAS G12V α4-α5 mutants analyzed in this study.  CCAGAACAGTAGACACAGAACAGGCTCAGGACTTAGCA  TGCTAAGTCCTGAGCCTGTTCTGTGTCTACTGTTCTGG  Q131E  CAGTAGACACAAAACAGGCTGAGGACTTAGCAAGAAGT  ACTTCTTGCTAAGTCCTCAGCCTGTTTTGTGTCTACTG  R135E  CAGGCTCAGGACTTAGCAGAAAGTTATGGAATTCCTTT  AAAGGAATTCCATAACTTTCTGCTAAGTCCTGAGCCTG  K147D  TTTATTGAAACATCAGCAGACACAAGACAGGGTGTTGA  TCAACACCCTGTCTTGTGTCTGCTGATGTTTCAATAAA  D154Q  ACAAGACAGGGTGTTGATCAAGCCTTCTATACATTAGTT  AACTAATGTATAGAAGGCTTGATCAACACCCTGTCTTGT  D154R  ACAAGACAGGGTGTTGATAGAGCCTTCTATACATTAGTT  AACTAATGTATAGAAGGCTCTATCAACACCCTGTCTTGT  R161D  GCCTTCTATACATTAGTTGATGAAATTCGAAAACATAAA  TTTATGTTTTCGAATTTCATCAACTAATGTATAGAAGGC  HRAS G12V D154Q  ACCCGGCAGGGAGTGGAGCAGGCCTTCTACACGTTGGTG  CACCAACGTGTAGAAGGCCTGCTCCACTCCCTGCCGGGT  D154R  ACCCGGCAGGGAGTGGAGCGGGCCTTCTACACGTTGGTG  CACCAACGTGTAGAAGGCCCGCTCCACTCCCTGCCGGGT  R161D  GATGCCTTCTACACGTTGGTGGACGAGATCCGGCAGCAC  GTGCTGCCGGATCTCGTCCACCAACGTGTAGAAGGCATC  NRAS G12V D154Q  ACCAGACAGGGTGTTGAACAAGCTTTTTACACACTGGTA  TACCAGTGTGTAAAAAGCTTGTTCAACACCCTGTCTGGT  D154R  ACCAGACAGGGTGTTGAAAGAGCTTTTTACACACTGGTA  TACCAGTGTGTAAAAAGCTCTTTCAACACCCTGTCTGGT  R161D GCTTTTTACACACTGGTAGATGAAATACGCCAGTACCGA TCGGTACTGGCGTATTTCATCTACCAGTGTGTAAAAAGC α4-α5 mutations do not impair oncogenic RAS signaling or in a 1:1 ratio with monobodies using lipofectamine according to the manufacturer's protocol. Raw band intensity was evaluated using Image Studio Lite Version 5.2. For calculating the normalized (pERK/ERK)/HA, each band was first normalized to vinculin. For calculating the NS1/Mb (neg) ratio, pERK/ERK ratio was first calculated then used to generate the NS1/Mb (neg) ratio. Statistical analysis was performed using GraphPad Prism 9.

NIH/3T3 transformation assays
For NIH/3T3 transformation assays, 2.5 × 10 5 cells were split into 60 mm tissue culture plates and seeded overnight. The following day, cells were transfected with the indicated RAS mutants using PEI transfection method (48). Media on cells was changed every 2 days following transfections. Oncogenic RAS induced foci formation approximately 2 to 3 weeks following transfections, and cells were fixed and stained with 0.1% crystal violet before quantification of foci. Assays were performed three times each (n = 3).

NanoBiT protein-protein interaction assays
For NanoBiT PPI assays, 3.0 × 10 4 cells per well (RAS-less MEFs) were plated in a white-wall, clear-bottom 96-well plate (Thermo Scientific 165306) and incubated at 37 C overnight. The next day, all wells were transfected with SmBiT-KRAS G12V , and selected wells were transfected (technical replicates per experiment = 6) with either EGFR-LgBiT, LgBiT-KRAS G12V , or LgBiT-KRAS G12V α4-α5 mutants using PEI transfection method. Twenty four hours after transfection, media was aspirated from wells, and luminescence was measured using NanoGlo Live-Cell Substrate (Promega; Cat # N2012) suspended in Opti-MEM reduced serum media (Gibco; cat # 31985070). After the live-cell luminescence measurement, cells were lysed with 1.0% Triton X-100 and incubated with HiBiT peptide (0.1 μM) for 10 min on orbital shaker. Then, luminescence was measured to quantify LgBiT peptide levels. Live-cell luminescence was normalized to luminescence after HiBiT peptide addition, and all samples were normalized to SmBiT-KRAS G12V /LgBiT-KRAS G12V . Assays were performed three times each (n = 3).

Statistical analysis
All statistical analyses were performed using GraphPad Prism 9 software.

Data availability
All reagents in this article are available upon request and completion of an MTA with the Medical University of South Carolina and/or New York University.
Supporting information-This article contains supporting information.
Puretech Health, and Argenx BVBA. The other authors declare no competing interests.