BRET-based RAS biosensors that show a novel small molecule is an inhibitor of RAS-effector protein-protein interactions

The RAS family of proteins is amongst the most highly mutated in human cancers and has so far eluded drug therapy. Currently, much effort is being made to discover mutant RAS inhibitors and in vitro screening for RAS-binding drugs must be followed by cell-based assays. Here, we have developed a robust set of bioluminescence resonance energy transfer (BRET)-based RAS biosensors that enable monitoring of RAS-effector interaction inhibition in living cells. These include KRAS, HRAS and NRAS and a variety of different mutations that mirror those found in human cancers with the major RAS effectors such as CRAF, PI3K and RALGDS. We highlighted the utility of these RAS biosensors by showing a RAS-binding compound is a potent pan-RAS-effector interactions inhibitor in cells. The RAS biosensors represent a useful tool to investigate and characterize the potency of anti-RAS inhibitors in cells and more generally any RAS protein-protein interaction (PPI) in cells.


Introduction
RAS is the most prominent oncogene identified in cancer. Mutation in RAS proteins can be found in approximately 30% of all human tumors (Downward, 2003;Prior et al., 2012) (http://cancer. sanger.ac.uk/cosmic) prompting interest in the discovery of anti-RAS therapeutics. However, there are still no RAS-targeted drugs currently available in the clinic even though such molecules could prove widely efficacious in many human cancers as front-line drugs for therapy. Some forms of cancer, like pancreatic cancer, present late and are difficult therefore to treat (Kleeff et al., 2016) but these contain a high proportion of KRAS mutations and are thus potentially susceptible to RAS-binding drugs.
RAS has been regarded as undruggable partly because so far attempts to interfere with the protein have not been efficacious (Cox et al., 2014). RAS is a membrane-bound small GTPase switching between an inactive GDP-bound state and an active GTP-bound state. RAS signaling to the cell nucleus occurs after interaction of RAS-GTP with its effectors to trigger the activation of downstream signaling pathways. This activation thereby promotes cell survival and cell proliferation (Wennerberg et al., 2005) via gene modulation so that the blockade of mutant RAS signaling in tumors cells is an attractive therapeutic option. There are several ways in which this could be achieved (Athuluri-Divakar et al., 2016;Burns et al., 2014;Spiegel et al., 2014;Zimmermann et al., 2013) but methods such as implementing farnesylation inhibitors have limited success due to side effects (Berndt et al., 2011;James et al., 1995;Whyte et al., 1997). One avenue that has largely been avoided in inhibiting RAS is the interaction with its effectors, such as RAF, RALGDS and PI3K. However, the effectiveness of the orthosteric RAS-effector PPI inhibition was shown using intracellular antibodies (Tanaka and Rabbitts, 2003;Tanaka et al., 2007) (herein called macrodrugs (Tanaka and Rabbitts, 2008) to distinguish them from conventional small molecule drugs) and a single domain intracellular antibody that blocks effector interaction sites of RAS-GTP. This PPI inhibition can prevent tumor growth in xenograft models and tumor initiation in a transgenic mouse model Tanaka et al., 2007). Other macrodrugs, such as DAR-Pins (Guillard et al., 2017), have also been shown to be effective in interfering with RAS PPIs. Moreover, for many years, RAS was regarded as a protein without any pockets suitable for small molecule interactions (McCormick, 2016) but recent studies have described compounds that are able to bind RAS-associated pockets (Gentile et al., 2017;Lito et al., 2016;Maurer et al., 2012;Ostrem et al., 2013;Patricelli et al., 2016;Shima et al., 2013;Sun et al., 2012;Waldmann et al., 2004;Welsch et al., 2017).
Most of the current RAS inhibitors have been selected and identified through in vitro techniques (Ostrem et al., 2013;Trinh et al., 2016;Upadhyaya et al., 2015;Welsch et al., 2017) but cellbased assay technologies are needed to assess initial hits for efficacy before hit to lead development is undertaken. Indeed, a robust cell-based assay is a mandatory step in any drug discovery programme, as it provides insights into the behavior of compounds in physiological conditions, including cell permeability, stability and potency in the cellular complexity of a whole cell. We now describe a toolbox of mutant and wild-type RAS BRET-based biosensors that can be used to assess PPI between activated, GTP-bound RAS (KRAS, HRAS or NRAS) and effectors such as CRAF, RALGDS or PI3K in living cells. We validate the toolbox using a published anti-RAS intracellular domain antibody (hereafter named iDAb RAS) (Tanaka et al., 2007), which is an inhibitor of RAS PPI to establish the RAS biosensor resource. We have further used this methodology to test a RAS-binding compound (herein referred to as 3344) that we have derived from an in vitro medicinal chemistry programme starting with an intracellular antibody fragment. By monitoring the change in BRET2specific signal in transfected HEK293T cells expressing different RAS-effector donor-acceptor combinations, we have been able to characterize the pan-RAS-effector PPI inhibitor properties of 3344. This inhibitory mechanism shown using the BRET biosensor toolbox was supported by the crystal structure of KRAS with bound 3344, showing binding to a pocket close to the RAS switch. Therefore, the BRET2 toolbox we describe here is a critical resource and is available for all investigators in the international effort to produce anti-RAS drugs, that can be employed in the treatment of cancers with RAS mutations. eLife digest A group of proteins known as the RAS family plays a critical role in controlling animal cell growth and division. RAS proteins are normally active only some of the time, but genetic mutations can create permanently active forms of the proteins. These constantly interact with other proteins called effectors. In response, cells multiply uncontrollably and give rise to cancers.
In an attempt to find new cancer treatments, researchers across the globe are trying to develop inhibitor drugs that prevent RAS and effector proteins from interacting. New drugs are often tested in laboratory experiments that directly apply the drugs to the proteins that they are designed to work on. But in some cases a drug may work wellin the laboratory but fail to work when used in cells. Unfortunately, there are few ways to judge how well inhibitor drugs work inside living cells.
Bery et al. have now developed RAS biosensors -a collection of proteins that bind to RAS and produce light more brightly when RAS interacts with effector proteins in living cells. Tests on cells treated with an antibody that works inside cells and is known to prevent interactions between RAS and effector proteins confirmed that the RAS biosensors work well. Bery et al. then used the RAS biosensors to show that a new RAS inhibitor works in human cancer cells.
The RAS biosensors are available upon request to researchers across the globe. They should form an important tool for testing potential treatments for cancers that contain mutated RAS proteins.

Engineering and validation of mutant RAS biosensors
RAS biosensors were developed for use in the BRET2 method (Bacart et al., 2008) as a real-time system allowing the monitoring of protein-protein interactions and their inhibition in live cells. The scheme used is outlined in Figure 1A. The intracellular localization of BRET donor RAS proteins was recapitulated by expressing the full-length proteins including the CAAX box, which is the farnesylation site for trafficking to the plasma membrane. The CAAX sequences were fused to the carboxy terminal end of the Renilla Luciferase variant 8 (RLuc8) to act as the donor molecule in BRET2 (De et al., 2007) (for simplicity of the nomenclature, CAAX has been omitted from the RAS construct names). We used available structural data for RAS/effector and RAS/iDAb complexes to optimize the proximity of donor and acceptor moieties. Hence, RLuc8 was fused to the amino termini of fulllength RAS family proteins and the GFP 2 (Ramsay et al., 2002) fused to the C-termini of the effectors (RALGDS, CRAF, PI3K) or of the iDAbs. Other parameters can influence the BRET2 signal such as the linker length between RLuc8/RAS and effector-iDAb/GFP 2 . For our study, we observed a higher BRET signal with a (GGGS) 3 linker between RLuc8-KRAS G12D construct, a (GGGS) 3 linker between the CRAF RBD-GFP 2 molecule and a (GGGS) 2 linker between iDAb RAS-GFP 2 construct ( Figure 1-figure supplement 1A). Therefore, we implemented these observations to all our BRET biosensors (Supplementary file 1). When donor and acceptor plasmids are transfected into HEK293T cells (although any cell line of choice would be suitable), the resultant cells are fluorescent and bioluminescent if treated with the luciferase substrate (coelenterazine 400a). If an interaction occurs between RAS and a partner-GFP 2 fusion, bringing the RLuc8 and GFP 2 within 100 Å , an energy transfer occurs from the RLuc8-RAS donor to the GFP 2 acceptor and a BRET2 signal is achieved ( Figure 1A, middle panel). Inhibitors of the donor-acceptor molecule interaction will decrease the BRET signal whilst maintaining the RLuc8 bioluminescence and GFP 2 fluorescence signals ( Figure 1A, right hand panel). The BRET signal (or BRET ratio) is calculated as the light emitted by the GFP 2 acceptor constructs (at 515 nm) upon addition of coelenterazine 400a, divided by the light emitted by the RLuc8 donor constructs (at 410 nm) (Pfleger et al., 2006). A background BRET signal is only observed with the donor-only construct where the RLuc8 plasmid is transfected alone into the cells (Figure 1-figure supplement 1B) and this signal is therefore subtracted from that BRET ratio. As shown in Figure 1-figure supplement 1B, un-transfected cells and those transfected with GFP 2 -only construct have a negligible auto-luminescence and emission at 515 nm upon addition of the BRET substrate and are not considered in the calculation of the BRET ratio.
BRET donor saturation assessments were first carried out with the RAS effector RAS binding domains (RBDs) to evaluate the optimal levels of expression plasmid transfection for the competition experiments ( Figure 1B). All of the effector domains were found to interact specifically with KRAS G12D since the BRET signal reached a donor saturation level ( Figure 1B). Further, all the transfected plasmids expressed the proteins at equivalent levels as indicated by western blot analysis ( Figure 1C) and their expression does not modify KRAS G12D expression (Figure 1-figure supplement 2A shows the increase of acceptor protein level has little effect of donor protein levels). To further characterize this BRET2 system, we used the dominant negative mutant KRAS S17N , which does not interact with the effectors (Cool et al., 1999;Nassar et al., 2010;van den Berghe et al., 1997), as a donor. We found that the BRET signal increased linearly with the concentration of acceptor for all the RAS binding domains. This result is typical of non-specific interactions (Mercier et al., 2002), confirming the S17N mutant does not interact with the effectors and supports the sensitivity of this system (Figure 1-figure supplement 2B).
We initially characterized the biosensor pairs with the iDAb RAS that is known to interact with mutant KRAS on the switch regions (Tanaka et al., 2007), compared with a non-relevant anti-LMO2 iDAb (Sewell et al., 2014;Tanaka et al., 2011) that was designated as iDAb control in this study (herein called iDAb Ctl). Introduction of mutations in the three CDRs of the iDAb RAS to generate a dematured iDAb RAS (iDAb dm RAS), was shown to reduce its affinity towards RAS-GTP from 6.2 nM to~1 mM affinity (Assi et al., 2010). While this did not alter the protein expression (Figure 1-figure supplement 2C,D), there was an expected BRET signal reduction (Figure 1-figure supplement 2C). Indeed, it significantly increased the BRET 50 (an approximation of the relative affinity of the acceptor fusion for the donor fusion proteins, corresponding to the acceptor/donor ratio necessary to reach 50% of the BRET max ) and significantly reduced the BRET max (an approximation for the total Figure 1. RAS-effector BRET biosensors and interference of KRAS-effector interactions by a RAS-binding compound. An outline of the BRET2-based RAS biosensor system is shown in A. RAS bound to the plasma membrane (PM) is fused at its amino terminal end to the RLuc8 moiety (donor). When a protein fused to the GFP 2 moiety (acceptor) does not bind to RAS, it only produces a background BRET signal. However, when an acceptor binds to RAS, it induces a BRET signal, if the luciferase and GFP domains are within 100 Å . The BRET signal can be decreased by addition of a competitor (either by a macrodrug or a small molecule inhibitor). The interaction titration of full-length KRAS G12D -CAAX (for simplicity, the CAAX motif is omitted in all the RAS constructs described hereafter) with the four effector acceptor proteins and the effect on intracellular protein levels are shown in B and C. Competition assays show the specificity of the RAS biosensors in D (iDAb) and E (RAS-binding compounds). In D, the non-relevant anti-LMO2 iDAb (called hereafter iDAb control, Ctl) serves as a negative control and anti-RAS iDAb (herein named iDAb RAS) serves as a positive control. In E, 3344 (black bars) decreases KRAS G12D /effector domain interactions in a dose-dependent manner showing its broad range of inhibition. Cells were treated with 5, 10 and 20 mM of 3344 (black bars), Abd-2 (grey bars) or DMSO alone (white bars) as the negative control. Statistical analysis was performed with a one-way ANOVA followed by Dunnett's post-hoc tests (*p<0.05, ***p<0.001, ****p<0.0001). Each experiment was repeated three (B, D) or four times (E). Where error bars are presented, these correspond to mean values ± SD of biological repeats (B, D-E). See also  number of complex RAS/iDAb and the distance between the donor and the acceptor within the dimer), which together are consistent with a decreased affinity of this mutant iDAb toward RAS. Therefore, the results obtained with the iDAb RAS confirmed the sensitivity and accuracy of the RAS biosensors.
Finally, we tested the inhibition of interaction between RAS and its effector partners using BRET in a competition assay. HEK293T cells were transiently transfected with KRAS G12D , each of the RASeffector domain and a competitor (non-GFP 2 ) version of the iDAb RAS or iDAb control. This competition showed that iDAb RAS, but not the control, drastically decreased the BRET ratio of all the interactions tested ( Figure 1D). These results confirmed that the BRET2 biosensors enable monitoring of PPI inhibition of KRAS G12D with each of the four effectors tested by the anti-RAS single domain antibody.

The BRET2 biosensors show that 3344 is an inhibitor of KRAS-effector interactions
Our major purpose in the development of the RAS BRET2 biosensors was to create a validation tool for compounds that bind to RAS and interfere with its PPI in living cells. We have identified compounds that bind to KRAS using in vitro screening and one compound 3344 (chemical structure and 1-D NMR characterization shown in  Figure 1D), 3344 was used for validation of the BRET2 toolbox for RAS-effector PPI inhibitors. In the subsequent experiments reported here, we compare 3344 with an initial compound (Abd-2) obtained through a SPR in vitro screening, which binds HRAS/KRAS with low affinity. It is the precursor of the 3344 compound and both share the same benzodioxane group (the structures of 3344 and Abd-2 are shown in Figure 1-figure supplement 3A,F). These compounds have been selected from a medicinal chemistry programme in order to validate the BRET-based RAS biosensors.
HEK293T cells were transiently transfected with BRET pairs and, after 24 hr to allow protein expression, the cells were seeded in 96-well plates. The compounds were added at different concentrations (5, 10 and 20 mM) and incubated on cells for a further 20 hr before the BRET reading. For each assay, the donor protein was RLuc8-KRAS G12D and the acceptor proteins were PI3Ka RBD-GFP 2 , PI3Kg RBD-GFP 2 , CRAF RBD-GFP 2 or RALGDS RA-GFP 2 . We observed a dose response reduction in BRET signal for the assays with compound 3344 but not with the Abd-2 indicating that only 3344 interferes with the RAS-effector PPI ( Figure 1E). To rule out the possibility of false positive compounds (for instance, that might interfere directly with the BRET signal), we included control BRET-based biosensors. We tested the RAS compounds with the iDAbs RAS biosensors, either with RLuc8-LMO2 donor and iDAb dm LMO2 (a dematured anti-LMO2 iDAb (Sewell et al., 2014) Abd-2 has no effect on any of these assays while 3344 only interferes, in a dose response, with KRAS G12D /iDAb dm RAS-induced BRET without affecting the expression of the biosensors (Figure 1-figure supplement 3J). Hence, the inhibitory effects of 3344 on KRAS G12D -effectors interactions are not simply due to interference with the BRET assay.
BRET2 reporter and associated RAS-CRAF signaling are affected by compound 3344 The RAS binding domain of the effector molecules lack some regulatory domains, which impedes a direct study of RAS inhibitors on pathways downstream of RAS. To reduce this limitation, we developed an optimized RAS biosensor of the full-length CRAF S257L mutant (herein named CRAF FL ) since the S257L mutation increases ERK phosphorylation (Razzaque et al., 2007) and because we found that CRAF FL interacts with KRAS G12D but not with KRAS S17N (Figure 2-figure supplement 1A). We performed a competition assay with the iDAb RAS confirming that it impedes the BRET2 signal due to the binding of CRAF FL with KRAS G12D , in a dose response mode, whereas the iDAb control had no effect (Figure 2A). There was no alteration in CRAF FL and KRAS G12D protein expression due to the transfection of the iDAbs, shown by western analysis (Figure 2-figure supplement 1B). In addition, iDAb RAS inhibition significantly decreased the phosphorylation of MEK1/2 and ERK1/2 kinases ( Figure 2B shows western blot data, quantitated in Figure 2C), confirming results affecting endogenous ERK phosphorylation by iDAb RAS interaction with RAS .
We further tested the ability of the small molecule 3344 to inhibit the KRAS G12D /CRAF FL biosensor and the downstream biomarker pathways with either a long incubation (20 hr, Figure 2D-F) or a short incubation (3 hr, Figure 2-figure supplement 1D-F) to further validate the specificity of  Figure 2-figure supplement 1. Quantification of the relative levels of pMEK1/ 2 and pERK1/2, normalized to total MEK1/2 and ERK1/2 respectively, are shown in F. The RAS biosensor toolkit includes KRAS G12A, G12C, G12V and G12R, in addition to KRAS G12D. In G, each was expressed with CRAF FL and BRET ratios determined at 0, 5, 10 and 20 mM Abd-2 or 3344. Statistical analyses in C were performed using a one-way ANOVA followed by Sidak's post-hoc tests and in A, D, F and G using a one-way ANOVA followed by Dunnett's post-tests (*p<0.05, **p<0.01, ***p<0.001, ****p<0.0001). Each experiment was repeated twice (E-F), three times (B-D), four times (A) or five times (G). Where error bars are presented, they correspond to mean values ± SD of biological repeats (A, D, G) or correspond to mean ±SEM of biological repeats (C, F). See also inhibition. Indeed, long-term incubation with the compound may indirectly inhibit RAS downstream pathways by affecting autocrine mechanisms involved in secondary activation of RAS pathways (Arthur and Ley, 2013;Zhang et al., 2011). We compared the effect of Abd-2 and 3344 on the BRET pair and found a significant decrease in BRET signal with 3344 that occurred in a dose-dependent manner ( Figure 2D and Figure 2-figure supplement 1D) without modifying RAS or CRAF expression (as shown by western analysis, Figure 2-figure supplement 1C,G). Western blots using anti-pMEK and anti-pERK showed that 3344 also significantly inhibited MEK1/2 and ERK1/2 phosphorylation whilst Abd-2 did not ( Figure 2E, quantified in Figure 2F and Figure 2-figure supplement 1E-F). Therefore, these observations show a specific and functional effect of the inhibition of interaction between RAS and CRAF FL by the 3344 with a long and short incubation.
Some compounds have been previously characterized that bind selectively on the cysteine of KRAS G12C mutant (Lito et al., 2016;Ostrem et al., 2013;Patricelli et al., 2016). We assessed whether our compound 3344 was able to interfere with binding of a range of mutant KRAS Gly12 proteins, including G12C, with CRAF in BRET assays. Analysis of the BRET2 signals from interaction of KRAS G12A , KRAS G12C , KRAS G12V and KRAS G12R with CRAF FL showed a dose response effect of compound 3344 but not Abd-2 ( Figure 2G). The corresponding BRET biosensor acceptor and donor proteins are equally expressed after transfection as judged by western blot analysis ( Therefore, using this new set of validated RAS biosensors, we show that the compound disrupts mutant KRAS/CRAF FL interaction in cells. In turn, this leads to inhibition of the RAF/MEK/ERK downstream signaling pathway (that emanates from the transfected protein expression).

inhibits the wild type KRAS-CRAF biosensor and its downstream signaling pathway
We extended the repertoire of biosensors by analyzing wild-type KRAS (KRAS WT ) donor molecule and also assessed if epidermal growth factor (EGF)-stimulated MEK/ERK phosphorylation (Burgering et al., 1993;Lange-Carter and Johnson, 1994) could be altered through the interaction of a KRAS WT /CRAF FL BRET2 biosensor protein pair. Although the iDAb RAS binds weakly to RAS WT in transfected mammalian two-hybrid reporter cells (Tanaka et al., 2007), we first established if the BRET2 signal from RLuc8-KRAS WT and GFP 2 -CRAF FL PPI could be inhibited by the iDAb RAS in the BRET transfection assay. HEK293T cells were transfected with the BRET pair and serum was removed for 24 hr, stimulated for 5 min with EGF and the BRET ratio directly determined after the stimulation. EGF treatment brings KRAS WT and CRAF FL fusion proteins in a closer proximity and enhances the number of KRAS WT /CRAF FL dimers because the BRET max value increases from 4.02 to 10.01 (Figure 3-figure supplement 1A). A dose response inhibition of the BRET2 signal was observed with iDAb RAS, but not iDAb control ( Figure 3A), which correlated with the reduction of pMEK1/2 and pERK1/2 detected by western blots ( Figure 3B and quantified in Figure 3C). This shows that the RAS BRET2 biosensors can be used to couple PPI effects and signaling effects.
We conducted parallel BRET2 dose response experiments with the 3344, compound compared to Abd-2, implementing EGF stimulation and using the KRAS WT /CRAF FL biosensor with short and long incubation times (3 hr and 20 hr, respectively). Compound 3344 inhibits this interaction in a dose-response manner ( Figure 3D and Figure 3-figure supplement 1D) and prevents the phosphorylation of MEK1/2 and ERK1/2 kinases ( Figure 3E, quantified in Figure 3F and

inhibits the RAS-PI3K-AKT signaling pathway
We have also explored the second best-characterized RAS effector family, the RAS-PI3Ka-AKT pathway (Castellano and Downward, 2011) by establishing a KRAS G12D /full-length PI3Ka (herein PI3Ka FL ) biosensor. In this case, we required a tripartite system as we observed that co-expression of the p85a regulatory subunit with PI3Ka FL -GFP 2 was required to obtain detectable, specific and optimized BRET signal from interaction of KRAS G12D and PI3Ka FL (Figure 4-figure supplement  1A). KRAS S17N mutant showed no specific interaction with PI3Ka FL further confirming the accuracy of this biosensor (Figure 4-figure supplement 1A). We validated the BRET biosensor by showing that the iDAb RAS impaired that interaction in a dose-dependent manner, whereas the iDAb control did not ( Figure 4A). Western blot analysis showed some reduction in PI3K and RAS proteins, specifically concordant with expression of the iDAb RAS ( Figure 4-figure supplement 1B) and there was also a dose response reduction of phosphorylation of the downstream biomarker AKT at Ser473 ( Figure 4B and quantified in Figure 4C).
Implementing the same biosensor assay treated with the compound 3344 for 3 or 20 hr, we confirmed this compound interferes with the KRAS G12D /PI3Ka FL interaction ( Figure  Abd-2 has no effect on the phosphorylation of AKT that results from KRAS G12D /PI3Ka FL interaction. Conversely, 3344 does affect RAS-PI3K interaction and AKT phosphorylation. When increasing doses of either Abd-2 or 3344 were used in the BRET-transfected cells, we observed dose response reduction of BRET signal with 3344 but not Abd-2 ( Figure 4D and  The BRET2 biosensor toolbox includes NRAS and HRAS and shows 3344 inhibits PPI of the RAS family The KRAS, NRAS and HRAS family members are conserved proteins that have an almost identical amino-acid domain (G domain) from residues 1-166 but a C-terminal hypervariable domain (Wennerberg et al., 2005). We have extended the RAS biosensor toolbox to include NRAS and HRAS. We used full-length NRAS Q61H and HRAS G12V mutants to build these new RAS biosensors for use with the various effector RBDs. These mutants were used at the positions Q61 and G12, for NRAS and HRAS respectively, as these are the positions most frequently mutated in human cancer involving NRAS and HRAS mutants (Cox et al., 2014). Titration of the RAS donor and CRAF FL acceptor proteins show that the RLuc8-NRAS Q61H and RLuc8-HRAS G12V proteins interact and reach plateau BRET signals with GFP 2 -CRAF FL ( Figure 5-figure supplement 1A). Furthermore, the BRET2 signal is diminished by increasing levels of the iDAb RAS but not the iDAb control ( Figure 5-figure supplement 1B-D) as expected from the analysis of the effects of the anti-RAS intracellular antibody Tanaka et al., 2007).
We further evaluated the efficacy of the RAS-binding compounds Abd-2 and 3344 in binding to NRAS and HRAS using a BRET assay in which the RAS protein donors were co-expressed with either PI3K, CRAF or RALGDS acceptors ( Figure 5A-D). While the low-affinity Abd-2 compound does not interfere with the BRET signal in any of the NRAS and HRAS BRET assays using either effector RBDs ( Figure 5A,B) or full-length CRAF ( Figure 5C,D), the compound 3344 disturbs the BRET2 signal in Compound 3344 binds to a pocket close to the switch regions of mutant KRAS The implementation of our RAS BRET2 toolbox showed that the compound 3344 is able to bind the transfected RAS protein products at the plasma membrane and interfere with their effector interaction. In addition, the downstream signaling was impeded. The mechanism of the interaction inhibition was corroborated by X-ray crystallography of KRAS Q61H soaked with compound 3344. Figure 6A shows that 3344 binds to KRAS in a previously identified pocket (Maurer et al., 2012;Sun et al., 2012) close to the switch regions where the effectors interact with RAS ( Table 1 has the refinement statistics for the X-ray data). The superimposition of the structures of three RAS-effector protein complexes with the structure of KRAS-3344 complex shows that parts of 3344 would overlap with the bound effector structures, suggesting that the competition effect of 3344 can be explained by straightforward steric hindrance ( Figure 6B). We further confirmed that 3344 could interfere with the endogenous RAS-effector PPI in two human cancer cell lines (viz. colorectal adenocarcinoma DLD-1 cells expressing KRAS G13D and non-small cell lung carcinoma H358 cells expressing  Figure 6 continued on next page KRAS G12C ). The cells were serum starved 24 hr and stimulated 10 min with EGF in the presence of increasing amounts of 3344, followed by western blot protein analysis to detect phosphorylated AKT Ser473 or phosphorylated ERK ( Figure 6C,D). 3344 decreases EGF-induced pAKT and pERK1/2 abundance in both cell types with an observed IC 50 of~5-10 mM without any change in the total levels of AKT or ERK1/2. Therefore, 3344 can interfere with endogenous RAS signaling in human cancer cell lines. As our BRET2 results show direct interference of RAS-effector PPI by 3344, we conclude that this is the mechanism of inhibition of the biomarkers in the tumor cell assay.

Discussion
BRET-based biosensors have been successfully used to discover and characterize small molecules inhibitors (Beautrait et al., 2017;Corbel et al., 2011;Lavoie et al., 2013;Mazars and Fåhraeus, 2010;Robinson et al., 2014). The development of such biosensors involves the optimization of multiple parameters such as the fusion position of the RLuc8 and GFP 2 moieties on their respective protein N-or C-terminus and the determination of the appropriate quantity of donor and acceptor plasmids for intracellular expression. Notably, the latest parameter has to be optimized in order to avoid the titration of active compounds if transient protein expression is used (Couturier and Deprez, 2012). In this study, we have engineered and optimized a complete set of RAS biosensors that includes several different mutant forms of KRAS and other family members (viz. mutant NRAS and HRAS). This toolbox allows the monitoring of RAS-effector interactions and the assessment of RAS PPI inhibition by a macrodrug (iDAb RAS) and 3344, a new anti-RAS small molecule derived from an intracellular antibody fragment, in living cells. Furthermore, when the full-length biosensors were used, we could couple the RAS PPI inhibition to the signaling effects, thereby providing additional insights into the behavior of RAS inhibitors.
The inhibition of RAS PPI by 3344 in cells was demonstrated by the RAS biosensors toolbox and validated by X-ray crystallography. 3344 binds to a hydrophobic pocket near to the effector-binding switch regions of RAS ( Figure 6). Whereas 3344 does not make direct contact with the switch regions, the BRET data show that the binding geometry and potency of 3344 is sufficient to interfere with the interaction of RAS-effector molecules that bind close to the 3344 site.
While the RAS biosensors rely on transfection and expression of RAS with one of its partner proteins rather than observations of endogenous proteins, it nevertheless offers several advantages for the study of RAS-effector interactions inhibition. It provides a direct and quantitative measurement of the PPI interference with inhibitors (i.e. small molecules or macrodrugs), which could allow the comparison of different compounds (e.g. for structure-activity-relationship studies) or macrodrugs and therefore the selection of more potent inhibitors. It is also sensitive and consequently requires a small quantity of cells to study the inhibition of the interaction. Nonetheless, 3344 prevents endogenous RAS-dependent signaling in two different human tumor cell lines at a lower concentration (IC 50 around 5 mM) ( Figure 6C,D) than in the BRET assay with observed IC 50 around 20 mM. This difference probably reflects the expression levels of the target proteins in the two assays, where the BRET2 assay relies on transient transfection. Indeed, the overexpression in HEK293T cells probably produces higher amount of mutant RAS/effector proteins than the endogenous counterparts in cancer cells. Therefore, it might be more difficult to quantitatively inhibit the exogenous RAS/effector interaction than the endogenous one with 3344 compound. Generating stable BRET2 cell lines could minimize this difference. were serum-starved for 24 hr and treated 3 hr with different concentrations of 3344 (2, 5, 10 and 20 mM) before stimulation with EGF (50 ng/mL) for 10 min. Cells were harvested, proteins extracted and separated by SDS-PAGE for western blot analysis. Western membranes were treated with anti-pAKT S473; anti-pan AKT; anti-pERK1/2 and anti-ERK1/2 as indicated. Statistical analyses of pERK/ ERK and pAKT/AKT quantifications were performed using a one-way ANOVA followed by Dunnett's post-tests (*p<0.05, **p<0.01, ***p<0.001, ****p<0.0001). Where error bars are presented, they correspond to mean values ± SEM of biological repeats (C-D). Each experiment was performed twice (C-D). DOI: https://doi.org/10.7554/eLife.37122.015 Another advantage of this toolbox has been shown by using the iDAb RAS as an acceptor within the RAS biosensors allowing a recapitulation of the published features of this intracellular single domain antibody. Therefore, the biosensors are also important tools to study RAS protein interactions in living cells and their effect on the RAS downstream pathways before being tested in cancer cell lines. RAS biosensors use should not be limited to the discovery and characterization of RAS inhibitors. Indeed, studies suggested that isoform and residue-or codon-specific RAS mutants show differences in their ability to engage effectors and signaling properties (Hunter et al., 2015;Nakhaeizadeh et al., 2016;Yan et al., 1998). Accordingly, RAS biosensors could also be a methodology to decipher RAS isoform/mutant properties in cells. Our toolbox is an available resource for RAS-drug development programmes, and more generally for the RAS community, since our results demonstrate the possibility of using these RAS biosensors as a generic method to characterize cellpotent RAS-binding compounds or RAS-binding macrodrugs. The BRET2 biosensor system could also be used for direct screens of PPI inhibitors with libraries of compounds. However, because initial compounds from a library are not expected to have high affinity for their target, relatively weak interactions between donor and acceptors should be involved in the generation of BRET2 signal. This provides a further use of intracellular domain antibodies where reduction of affinity (dematuration) from a tool initially used for target validation, can be achieved to make a screening tool. Thus, the method is an approach that is transferable to other PPI situations required for drug development programmes in cancer or any other clinical indication. Cell culture HEK293T human embryonic kidney cells, DLD-1 cells and H358 cells were grown in DMEM medium (Life Technologies) supplemented with 10% FBS (Sigma) and 1% Penicillin/Streptomycin (Life Technologies). Cells were grown at 37˚C with 5% CO 2 and were tested using a MycoAlert Mycoplasma Detection Kit (Lonza) and found to be mycoplasma-free before use.

Mutation detection of RAS mutations using RT-PCR
RNA was extracted from 5 Â 10 6 DLD-1 or H358 cells using the RNeasy Plus Mini Kit (Qiagen) according to the manufacturer's instructions. cDNA was synthesized from 1.5 to 2 mg RNA using SuperScript II Reverse Transcriptase (Invitrogen). Primers were designed to amplify KRAS DNA and incorporate HindIII and BamHI restriction sites for subcloning: 5'-TAAGCAAAGCTTATGACTGAATATAAACTTGTGGTAG-3' and 3'-GAAAATTAAAAAATGCATTATAATGTAAGGATCCTAAGCA-5' DNA was amplified using Phusion High-Fidelity DNA Polymerase (New England Biolabs) and, following digestion with HindIII and BamHI, the DNA was cloned into pBlueScript II SK (+) (Stratagene). Plasmid DNA was prepared from indivudial DH5a transformants using a QIAprep Spin Miniprep Kit (QIAGEN). KRAS mutations were verified by Sanger sequencing (Source Bioscience) of at least six clones from each cell line. The KRAS mutations in the two human cancer cell lines were confirmed as KRAS G13D in DLD-1 and KRAS G12C in H358.

Cell treatment
For dose response experiments (BRET and western blot), drugs were prepared in 100% DMSO at 10 mM. Cells were treated with Abd-2 or 3344 compounds at concentration of 5, 10 or 20 mM for 3 hr (short-term incubation) or 20 hr (long-term incubation). The compounds were diluted in the BRET medium: OptiMEM no phenol red (Life Technologies) supplemented with 4% FBS and with a final concentration of 0.2% DMSO.
For serum starvation studies with the BRET assay, cells were grown 24 hr in the presence of Opti-MEM no phenol red supplemented with 1% FBS and stimulated with 50 ng/mL EGF (Life Technologies) for 5 min at 37˚C. For serum-starvation studies of cancer cell lines, cells were grown 24 hr in the presence of DMEM without FBS and stimulated 10 min with 50 ng/mL EGF. The compound was incubated for 3 hr before the EGF stimulation at 2, 5, 10 and 20 mM.
LMO2 was amplified by PCR and cloned between NcoI/XhoI sites of the pEF-MCS-RLuc8 plasmid.
All RAS and effectors are human sequences except RALGDS RA (mouse). All the RAS BRET constructs DNA and protein sequences have been listed in the supplementary file 1.

BRET2 titration curves and competition assays
The BRET experiment protocols have been adapted from previous studies (Lavoie et al., 2013;Pfleger et al., 2006). For all BRET experiments (titration curves and competition assays) 650,000 HEK293T were seeded in each well of a six well plates. After 24 hr at 37˚C, cells were transfected with a total of 1.6 mg of DNA mix, containing the donor + acceptor ± competitor plasmids, using Lipofectamine 2000 transfection reagent (Thermo-Fisher). Cells were detached 24 hr later, washed with PBS and seeded in a white 96 well plate (clear bottom, PerkinElmer) in OptiMEM no phenol red medium complemented with 4% FBS. Cells were incubated for an additional 20-24 hr at 37˚C before the BRET assay reading.
The BRET signal or BRET ratio corresponds to the light emitted by the GFP 2 acceptor constructs (515 nm -30 nm) upon addition of coelenterazine 400a divided by the light emitted by the RLuc8 donor constructs (410 nm -80 nm). The background signal is subtracted from that BRET ratio using the donor-only negative control where only the RLuc8 plasmid is transfected into the cells. The normalized BRET ratio is the BRET ratio normalized to a negative control (DMSO, no competitor or iDAb control) during a competition assay. Total GFP 2 and RLuc8 signals were used to control the protein expression from each plasmid.

WaterLOGSY NMR
The waterLOGSY NMR method (Dalvit et al., 2001) was used to measure RAS ligand interaction . WaterLOGSY experiments were conducted at a 1 H frequency of 600 MHz using a Bruker Avance spectrometer equipped with a BBI probe. All experiments were conducted at RT, 25˚C. 3 mm diameter NMR tubes with a sample volume of 200 mL in all experiments. Solutions were buffered using an H 2 O PBS buffer corrected to pH 7.4. The sample preparation is exemplified as follows; the compound (10 mL of a 10 mM solution in DMSO-d 6 ) was added to an Eppendorf tube before sequential addition of the H 2 O PBS buffer (163.6 mL), D 2 O (20 mL), and protein (6.4 mL, 311.8 mM). The resulting solution was vortexed to mix and transferred to a 3 mm NMR tube prior to the NMR analysis.
For competition experiments using anti-RAS scFv, protein preparation for NMR was carried out in a similar manner; the compound (10 mL of a 10 nM solution in DMSO-d 6 ) was added to an Eppendorf tube before sequential addition of the H 2 O PBS buffer (146.4 mL), D 2 O (20 mL), protein (6.4 mL, 311.8 mM) and anti-RAS scFv (17.2 mL, 116.6 mM). The resulting solution was vortexed to mix and transferred to a 3 mm NMR tube prior to the NMR analysis.
Negative controls (compound alone) were prepared in a similar manner, in order to obtain an end volume of 200 mL.

Chemical synthesis procedures
All reactions involving moisture-sensitive reagents were carried out under a nitrogen atmosphere using standard vacuum line techniques and glassware that was flame-dried before use. Anhydrous solvents were prepared following the procedure outlined (Pangborn et al., 1996). Water was purified by an Elix UV-10 system. All other solvents and reagents were used as supplied (analytical or HPLC grade) without prior purification. Brine refers to a sat. aq. solution of NaCl. In vacuo refers to the removal of solvent by the use of a rotary evaporator attached to a diaphragm pump.
Thin layer chromatography was performed on normal phase Merck silical gel 60 F254 aluminumsupported thin layer chromatography sheets. Visualization of spots was either by absorption of ultra violet light (lmax 254 nm), or by thermal development after staining with 1% aq. KMnO4. Flash column chromatography was performed on Kieselgel 60 silica in a glass column, under a positive pressure.
NMR spectra were recorded on Bruker Avance spectrometer (AVIII 600) in the deuterated solvent stated. The field was locked by external referencing to the relevant deuteron resonance. Chemical shifts (d) are reported in parts per million (ppm). The multiplicity of each signal is indicated by: app. (apparent), s (singlet), br s (broad singlet), d (doublet), t (triplet), q (quartet), dd (doublet of doublets) or m (multiplet). Coupling constants (J) are quoted in Hz and are reported to the nearest 0.1 Hz.
Low-resolution mass spectra were recorded on an Agilent 6120 spectrometer operating in positive or negative mode, from solutions of MeOH. Accurate mass measurements were run on either a Bruker MicroTOF internally calibrated with polyalanine, or a Micromass GCT instrument fitted with a Scientific Glass Instruments BPX5 column (15 m x 0.25 mm) using amyl acetate as a lock mass, by the mass spectrometry department of the Chemistry Research Laboratory, University of Oxford, UK. m/z values are reported in Daltons.

H CPMG NMR experiments for compound Kd calculation
Typical experimental parameters for Carr-Purcell-Meiboom-Gill (CPMG) NMR spectroscopy were the following: total echo time, 40 ms; relaxation delay, 2 s; and number of transients, 264 (Abboud et al., 2016). The PROJECT-CPMG sequence (90˚x-[TÀ180˚y-T-90˚y-TÀ180˚y-T] n -acq) was applied. Water suppression was achieved by presaturation. Prior to Fourier transformation, the data were multiplied with an exponential function with 3 Hz line broadening. The CPMG experiments were conducted at a 1 H frequency of 700 MHz using a Bruker Avance with 5 mm inverse TCI 1 hr/ 13C/15N cryoprobe. All experiments were conducted at RT and lapsed 128 scans. 3 mm diameter NMR tubes with a sample volume of 200 mL were used in all experiments. Solutions were buffered using a D 2 O PBS buffer corrected to pH 7.4. The sample preparation is exemplified as follows: for a 5 mM GST-KRAS G12V sample: 55 mM of the 3344 compound (1.1 mL of a 10 mM solution in DMSOd 6 ) was added to an Eppendorf before sequential addition of the D 2 O PBS buffer (194.0 mL) and GST-KRAS G12V (4.9 mL of a 205 mM solution, the protein is in an H 2 O buffer for stability reason). The resulting solution was vortexed to be fully mixed and transferred to a 3 mm NMR tube before the run. Negative controls (compound alone, without the KRAS protein) were prepared in a similar manner, in order to obtain an end volume of 200 mL. CPMG experiments were carried out at a fixed 3344 concentration (55 mM, optimal concentration for these CPMG NMR experiments) and a variable GST-KRAS G12V concentration. The amount of GST-KRAS G12V was increased from 0 mM until the signals of the compound completely disappear in the proton NMR at 20 mM. Seven measurements were done in total with 0 mM, 2.5 mM, 5 mM, 7.5 mM, 10 mM, 15 mM and 20 mM of GST-KRAS G12V . The integrations of the protons acquired were all compared to the compound alone (with no KRAS) in order to obtain a percentage decrease for each concentration of KRAS. Three different proton signals were used and a mean was calculated for each run. KRAS concentration experiments were run in triplicate and a mean was also calculated for each concentration. Concentration and percentage of decrease were plotted and Kd fitting was run on the generated curve using Origin 2017 software with the following function: A*(1/ (2*C))* ((B + x + C)-sqrt(((B + x + C)2)-(4*x*C))) where A is the maximum % of inhibition (i.e. 100), B is the Kd, C is the concentration of compound and x the concentration of KRAS protein necessary to reach 100% of signal reduction of the compound.
Recombinant protein expression for crystallography and NMR: KRAS G12V , KRAS Q61H and scFv KRAS G12V cDNA was cloned into the pGEX vector in-frame with an N-terminal Glutathione-S transferase (GST) tag. pGEX-GST-KRAS G12V was transformed into E.coli BL21 (DE3) cells. Bacterial cells were cultured at 37˚C to an OD 600 of 0.5 and induced with IPTG (isopropyl 1-thio-beta-D-galactopyranoside, final concentration 0.1 mM) at 16˚C overnight. The bacteria cultures were harvested by centrifugation and the cell pellets re-suspended in 50 mM Tris-HCl pH8.0, 140 mM NaCl, 1 mM mercaptoethanol supplemented with complete protease inhibitor (Roche). The GST-fusion proteins were purified by glutathione-sepharose column chromatography (GE Healthcare) and eluted with 50 mM Tris-HCl pH8.0, 10 mM reduced glutathione, 1 mM mercaptoethanol, 5 mM MgCl 2 .
KRAS Q61H cDNA was cloned into the pRK-172 vector in-frame with an N-terminal 6xHis-tag and TEV protease recognition site. The plasmid containing KRAS Q61H sequence was transformed into E. coli B834(DE3) pLysS cells, which were grown in 25 mL LB medium with 50 mg/mL Carbenicillin and 34 mg/mL Chloramphenicol for 16 hr, prior to inoculation of 1L LB medium. Protein expression was induced at OD 600 = 0.6 by addition of IPTG to a final concentration of 0.5 mM and cells grown overnight at 16˚C. Bacteria were harvested by centrifugation and sonicated in 50 mM Tris-HCl, pH 7.5, 500 mM NaCl, 5 mM MgCl 2 and 10 mM imidazole and EDTA-free protease inhibitor cocktail (Roche Diagnostics). Proteins were purified using nickel agarose beads (Invitrogen) and bound proteins were eluted batch-wise in 50 mM Tris-HCl, pH 7.5, 500 mM NaCl, 5 mM MgCl 2 and 300 mM imidazole. RAS protein samples were concentrated using Vivapore 10/20 mL concentrator (7.5 kDa molecular weight cut-off; Sartorius Vivapore) to a final volume of approximately 1 mL. Nucleotide exchange for crystallographic samples was carried out following published procedures (Herrmann et al., 1996). RAS proteins were further purified by gel filtration on a HiLoad Superdex 75 10/300 GL column (GE Healthcare) in a buffer containing 20 mM HEPES pH 8.0, 150 mM NaCl, 5 mM MgCl 2 and 1 mM DTT at a flow rate of 0.5 mL/min. Fractions corresponding to the protein were pooled and concentrated to 45-75 mg/mL for crystallization trials. Protein concentration was determined by extinction coefficient (e 280 = 12045 L/mol/cm). Protein purity was analyzed by SDS-PAGE stained with Coomassie Brilliant Blue. scFv recombinant protein was expressed and purified as described elsewhere (Tanaka et al., 2007).

Crystal structure and 3344 soaking
For X-ray diffraction experiments, KRAS Q61H -GppNHp crystals were grown by vapour diffusion at 4˚C by mixing 1.5 + 1.5 volumes of KRAS solution at a concentration of 75 mg/mL KRAS Q61H , with 8-15% w/v Polyethylene Glycol 3350 and 0.2 M lithium citrate pH 5.5. The resulting crystals are termed crystal form I hereafter. Prior to X-ray data collection, crystals were cryo-protected by addition of 20% glycerol to the crystallization buffer and flash-cooled in liquid nitrogen. 3344 was initially dissolved at 200 mM in 100% DMSO and sequentially mixed in a ratio of 1:1 with crystallization buffer (8-15% w/v Polyethylene Glycol 3350, 0.2 M lithium citrate 7.0 and 20 mM Tris-HCl pH 7.0) to give a final concentration of compound of 50 mM and 25% DMSO in a 5 mL drop. Soaked crystals were flash-cooled in liquid nitrogen prior to data collection using the final DMSO concentration on the soaking drop as cryo-protectant. X-ray diffraction data were collected at beamline ID30A-1 (Bowler et al., 2015;Bowler et al., 2016;Nurizzo et al., 2016;Svensson et al., 2015) at The European Synchrotron Radiation Facility (ESRF, Grenoble, France). The structure of KRAS Q61H GppNHp-3344 was solved by molecular replacement using a KRAS169 Q61H GPPNHP-Abd-2, (PDB ID 5OCO) as a search model within the program Phaser (McCoy, 2007;McCoy et al., 2007). Structures were manually adjusted using COOT (Emsley et al., 2010) and refined using REFMAC (Murshudov et al., 1997). Crystal Form I (KRAS Q61H ) has six KRAS molecules in the asymmetric unit, assembled as a hexamer. Electron density maps averaged with six-fold non-crystallographic symmetry (NCS) were used to improve the definition of the bound compounds. Refinements were also performed with the six fold NCS applied. The refined models were validated using PROCHECK (Laskowski et al., 1993a), MolProbity  and Phenix software packages (Adams et al., 2010;Laskowski et al., 1993b). Figures were created using PyMOL (Schrodinger). Data collection and refinement statistics are summarized in Table 1.

Quantification and statistical analysis
All quantifications were performed using ImageJ or Prism 7.0 c (GraphPad Software), BRET titration curves and statistical analysis were performed using Prism 7.0 c (GraphPad Software). Data are typically presented as mean ± SD or SEM as specified in the figure legends. Statistical analyses were performed with a one-way ANOVA followed by Dunnett's post-hoc tests or Sidak's post-hoc tests unless otherwise indicated in the figure legends. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

Data and software availability
Structure files and coordinates have been deposited to PDB under this accession number: 6F76.