The importance of Ras in drug resistance in cancer

In this review, we analyse the impact of oncogenic Ras mutations in mediating cancer drug resistance and the progress made in the abrogation of this resistance, through pharmacological targeting. At a physiological level, Ras is implicated in many cellular proliferation and survival pathways. However, mutations within this small GTPase can be responsible for the initiation of cancer, therapeutic resistance and failure, and ultimately disease relapse. Often termed “undruggable,” Ras is notoriously difficult to target directly, due to its structure and intrinsic activity. Thus, Ras‐mediated drug resistance remains a considerable pharmacological problem. However, with advances in both analytical techniques and novel drug classes, the therapeutic landscape against Ras is changing. Allele‐specific, direct Ras‐targeting agents have reached clinical trials for the first time, indicating there may, at last, be hope of targeting such an elusive but significant protein for better more effective cancer therapy.


| INTRODUCTION
Resistance to conventional therapeutic agents is an increasingly concerning issue across all areas of disease, including cancer. Whilst the heterogeneous nature of cancer means there are many mechanisms resulting in drug resistance, Ras mutations underpin resistance to a variety of therapies (Hobbs et al., 2016;Prior et al., 2012Prior et al., , 2020. Oncogenic mutations in this small GTPase, which occur in approximately 19% of all cancers, cause constitutive activation of proliferative and survival pathways (Prior et al., 2020). This can abrogate the effects of standard chemotherapy and newer receptor-targeted therapies. Examples of such resistance is seen across a wide range of cancers, including metastatic colorectal cancer (mCRC), non-small cell lung cancer (NSCLC), pancreatic cancer, acute myeloid leukaemia (AML) and basal cell carcinoma (Li et al., 2004;McMahon et al., 2019;Misale et al., 2012;Tao et al., 2014). Thus, there is a distinct clinical need to target Ras pharmacologically. Whilst this has been particularly challenging due to structural difficulties and very high levels of intrinsic activity, significant developments have been made within the last decade. Allele-specific, direct Ras-targeting reaching clinical trials present a new era of potential Ras therapeutics and increase the likelihood of overcoming this resistance mechanism.

| WHAT IS DRUG RESISTANCE?
There are two general classes of resistance, intrinsic and acquired.
Intrinsic resistance occurs due to overt pre-existing factors, including variations in protein expression levels (such as increased expression of the P-gp [MDR1] transporter), epigenetic modifications (including by the long non-coding RNA (lncRNA) HAND2-AS1 and chromatin modifier Jarid1A (lysine demethylase 5A)) and somatic mutations (such as in Ras) (Burrell et al., 2013;Gruber et al., 2012;Marusyk et al., 2012;Sharma et al., 2010). Changes conferring drug resistance often coexist, resulting in a resistance heterogeneity similar to that seen in the original disease itself (Gerlinger et al., 2012;Ramirez et al., 2016).
Acquired resistance is thought to occur for a number of reasons, including through pre-existing (but initially undetectable) and de novo mutations (Bhaduri et al., 2020;Russo et al., 2019). This is often identified weeks to months after treatment has commenced (Santoni-Rugiu et al., 2019). In recent times, the concept of disease clonal heterogeneity has provided greater insight into the causes of acquired resistance, suggesting that chemoresistance and disease relapse occur as a result of minor subclonal populations. Given that these likely contribute to the heterogeneous nature of cancer, it seems likely that these also contribute to disease re-emergence and relapse (Bonnet & Dick, 1997;Gerlinger et al., 2012;Pattabiraman & Weinberg, 2014;Roy & Cowden Dahl, 2018;Seth et al., 2019). Within these minor clonal populations, mutations conferring drug resistance may exist at the early stages of disease but remain dormant and undetectable upon first presentation (Pietrantonio et al., 2017;Russo et al., 2018). When the bulk of the cancer is eliminated as a result of initial chemotherapy targeted at overt mutations, cells from this minor subclone proliferate and become dominant in the tumour bulk (Jones et al., 2019;McMahon et al., 2019).
A second, related, resistance mechanism involves a very rare subpopulation of cells, known as cancer stem cells. Cancer stem cells were first described in acute myeloid leukaemia but have since been applied to many other cancers including (but not limited to) breast cancer, colorectal cancer, myeloma and pancreatic adenocarcinoma (Bonnet & Dick, 1997;Koury et al., 2017;Lapidot et al., 1994).
Cancer stem cells have unique properties compared to bulk tumour cells. They are undifferentiated and have strong self-renewal and proliferative capabilities and typically remain in a quiescent state, thus avoiding chemotherapy targeted at rapidly dividing cells (as in bulk tumour cells). However, they do have the proliferative capacity to maintain and expand the tumour burden, as bulk tumour cells are eliminated (Jordan et al., 2006). These "stemness characteristics" render this subset of cells intrinsically resistant to chemotherapy, with distinct immunophenotypic and molecular signatures (Bonnet & Dick, 1997). This includes increased expression of efflux transporter P-gp (MDR1) and the ability to repair damaged DNA, a common method of inducing cancer cell death (Dean et al., 2005;Pattabiraman & Weinberg, 2014).
Certain pathways up-regulated in cancer stem cells have been linked to the increased self-renewal capacity seen in this subset of cells. Key examples include the Wnt/β-catenin and Hedgehog (Hh) pathways, as well as the Ras-dependent MAPK and PI3K/AKT pathways. Of these, the Wnt/β-catenin pathway is perhaps the most associated with cancer stem cells. Briefly, at a physiological level, Wnt signalling is not active and β-catenin is ubiquitinated and sent for proteasomal degradation following interaction with GSK3β, APC and Axin. In cancer stem cells however, activation of Wnt signalling inhibits formation of the β-catenin-GSK3β-APC-Axin complex, stabilising β-catenin. This translocates to the nucleus whereby it stimulates transcription of proliferative genes, including c-Myc. This promotes proliferative signalling, increasing the self-renewal capacity of cancer stem cells (Krausova & Korinek, 2014;Moon et al., 2014).
Although still ambiguous, it has been reported that β-catenin stabilisation can promote Ras stabilisation and protects Ras against proteasomal degradation Lee et al., 2018). This in turn promotes MAPK and PI3K pathway signalling, further contributing to the self-renewal capacity of cancer stem cells.
Alternatively, up-regulation of the hedgehog (Hh) signalling pathway has been shown in cancer stem cells. Whilst there are many facets to this pathway, its activation can stimulate up-regulation of stemness-associated transcription factors including NANOG, OCT4 and SOX2, thus promoting self-renewal (Boyer et al., 2005;Po et al., 2010;Takahashi & Yamanaka, 2006;Zhu et al., 2019). Increased activation and stabilisation of these genes (through either Hh signalling or biochemical stresses) have been implicated in the dedifferentiation of bulk tumour cells to cancer stem cells (Herreros-Villanueva et al., 2013;Kumar et al., 2012).
Interest in epigenetic remodelling has grown considerably in the last decade and it has considerable effects in promoting drug resistance. lncRNAs are also implicated in survival of cancer stem cells. In hepatocellular carcinoma, the lncRNA HAND2-AS1 is up-regulated and stimulates the self-renewal capacity of cancer stem cells, through activation of the bone morphogenetic protein (BMP) signalling pathway. BMP activation has in turn been shown to induce chemoresistance in other cancer models, such as lung cancer (Gruber et al., 2012;Wang et al., 2015). Indeed, this pathway is also regulated by the fusion gene CBFA2T3-GLIS2 and the Hh and JAK-STAT signalling pathways (Gruber et al., 2012;Okada et al., 2014).
Taken together, it is evident that many signalling pathways and associated factors play a critical role in mediating drug resistance.
Identifying commonalities between these pathways could present an opportunity to reduce the incidence and impact of drug resistance.
Although the incidence of Ras mutations in cancer stem cells is relatively low, Ras-mediated pathways have been implicated (Corces-Zimmerman et al., 2014). In addition to the aforementioned β-cateninmediated stabilisation of Ras, stabilised OCT4 and SOX2 expression has also been shown in response to AKT activation, by extracellular biochemical and radiation stresses (Maiuthed et al., 2018;Park et al., 2021). Indeed, activation of the MAPK pathway through the Ras-Raf interaction promotes increased expression of SOX2 and NANOG (Chan et al., 2018;Du et al., 2019). Furthermore, correlation has been seen between up-regulation of the MAPK, PI3K and BMP pathways in drug-resistant lung cancer (Wang et al., 2015). Thus, not only can these pathways contribute to an increased cellular proliferation rate in bulk cancer, but they can also assist with the selfrenewal and stemness properties evident in cancer stem cells. Perhaps unsurprisingly, inhibition of the Hh signalling pathway, when combined with inhibition of mTOR (downstream of AKT), has been seen to eliminate cancer stem cells in pancreatic cancer (Mueller et al., 2009;Prieur et al., 2017). Furthermore, MAPK kinase (MEK) inhibition has been implicated in reduced pancreatic cancer stem cell survival (Walter et al., 2019).

| KEY EXAMPLES OF DRUG RESISTANCE IN CANCER
In recent years, resistance has been documented amongst many cancer therapeutics, including traditional chemotherapeutic agents, more novel small molecule inhibitors and monoclonal antibodies (Caiola et al., 2015;McMahon et al., 2019;Pietrantonio et al., 2017). The origin of this resistance (intrinsic or acquired) varies considerably between drugs and so advances have been made in treatment stratification, based on a patient's mutational status, for some therapies. For example, vemurafenib, the BRAF inhibitor, is restricted to melanoma patients with the BRAF V600E mutation (Hopkins et al., 2019). Likewise, the monoclonal antibody panitumumab is only recommended for patients with wild-type Ras (Amado et al., 2008). However, as mentioned previously, some mutations are only detectable when patients relapse, such as the emergence of fms-like TK 3 (FLT3) or NRAS mutations in acute myeloid leukaemia patients (Man et al., 2012;McMahon et al., 2019;Piloto et al., 2007;Smith et al., 2017). Many of these mutations occur downstream of the site targeted by the drug. For example, since Ras operates downstream of the EGF receptor (EGFR), Ras mutations play a role in rendering cetuximab and panitumumab (which bind and inhibit EGFR) ineffective as the constitutive activation is not inhibited by the drug (De Roock et al., 2010;Li et al., 2015;Zhao et al., 2017). However, some of these putative resistance-causing mutations can only be detected at relapse, once they have expanded from only existing in a minor, undetectable subclone (as they did at diagnosis). Thus, it is difficult to predict which patients will develop these resistance mutations, meaning initial treatment stratification is difficult. Thus, combination therapy between these established agents and novel agents targeting these common mutational sites may be a way forward in preventing resistance before it occurs, an example of which is the combination of BRAF and MEK inhibitors, vemurafenib and cobimetinib (Hopkins et al., 2019).

| INTRODUCTION TO Ras
As eluded, Ras mutations underpin resistance to a variety of therapies.
KRAS is most commonly mutated of the three Ras isoforms and is particularly frequently mutated in lung and pancreatic cancer (Moore, Rosenberg, et al., 2020;Prior et al., 2020). However, patterns of Ras mutation differ between cancers and NRAS is the most frequently mutated RAS gene in acute myeloid leukaemia. Also, HRAS accounts for most Ras mutations in head and neck squamous cell carcinoma (Prior et al., 2020). Indeed, this variation carries varying prognoses of Ras-mutated cancer, with KRAS mutations conferring the lowest overall survival 5 years post diagnosis (43%) and HRAS mutations conferring the highest (63%) (Figure 1) (Cerami et al., 2012;Gao et al., 2013).
Most oncogenic mutations in Ras occur in the residues G12, G13 and Q61, which are located in a region conserved between HRAS, KRAS and NRAS (Figure 2a). They occur within the effector lobe, which comprises the first 85 amino acids and is a region of complete sequence identity between the different RAS isoforms. Most key interaction sites occur within the effector lobe, including nucleotide and effector interaction sites, as well as the switch regions that mediate effector interactions (Figure 2b). There is 90% similarity in the next 80 amino acids (allosteric lobe). The only considerable sequence differences occur at the C terminal end of the protein, known as the hypervariable region. This is the region in which post-translational modifications occur and membrane-targeting sequences are found (Hobbs et al., 2016). Initial attempts at direct pharmacological targeting of Ras centred around inhibiting these post-translational modifications, including farnesylation (Cox & Der, 2002;Whyte et al., 1997). However, more recent attempts have considered the effector lobe and structures within this to be better therapeutic targets, as discussed later (Canon et al., 2019;Janes et al., 2018;Ostrem et al., 2013).
Ras is activated or inactivated when bound to GTP or GDP, respectively. Oncogenic mutations increase the frequency of GTPbound (active) Ras, through two key mechanisms. This is either through decreasing the affinity of Ras for GTPase-activating proteins (GAPs), which stimulate the intrinsic GTPase activity of Ras and therefore facilitate hydrolysis of GTP to GDP (off/inactivating mechanism) or by decreasing the need for guanine nucleotide exchange factors (GEFs), which mediate the on/activating mechanism (Smith et al., 2013;Wittinghofer & Waldmann, 2000). These differences depend on multiple factors including the specific amino acid mutation and resulting protein conformation (Hunter et al., 2015;Miller & Miller, 2012;Poulin et al., 2019;Prior et al., 2012;Smith et al., 2013).
Some mutations (including G12C) promote rapid cycling between the inactive and active states, an area which has been exploited by a range of novel Ras-targeting drugs (Fell et al., 2020;Hunter et al., 2015;Patricelli et al., 2016).

| Ras SIGNALLING-REGULAR AND ONCOGENIC
Ras is involved in key pathways regulating cell proliferation, differentiation and sensitivity to apoptosis, perhaps the most notable being the MAPK pathway and the PI3K/AKT pathway. KRAS has also been implicated (albeit in a more indirect way) in a range of other pathways, including the Wnt/β-catenin pathway and the nuclear factor erythroid 2-related factor 2 (NRF2) pathway (DeNicola et al., 2011;Ferino et al., 2020;Moon et al., 2014;Park et al., 2019;Tao et al., 2014). These pathways are detailed briefly below and are summarised in Figure 3.
Ligand binding to receptor tyrosine kininases (RTKs), such as EGFR or FLT3, causes autophosphorylation of key tyrosine residues on the RTK, after which adaptor proteins such as GRB2 bind to these phosphorylated tyrosines via SH2 domains. Son-of-sevenless (SOS), a Ras-guanine nucleotide exchange factor, then associates with GRB2 via two SH3 domains and ultimately facilitates the exchange of GDP for GTP on RAS, thereby activating it (Freeman, 2000).
Following this, Ras stimulates recruitment of its serine-threonine kinase effector, RAF, to the membrane, which binds through its Ras binding domain and then proceeds to phosphorylate downstream kinases including MEK and ERK (Marais et al., 1995;Molina & Adjei, 2006). ERK then translocates from the cytoplasm to the nucleus, causing the activation of many transcription factors, which go on to regulate a host of processes, including proliferation and differentiation (Guo et al., 2020;Zhang & Liu, 2002).
Alternatively, activated Ras can promote the PI3K-AKT pathway.
This pathway also plays a key role in controlling survival, division and metabolism and can be activated in many different ways (Castellano & Downward, 2011;Mendoza et al., 2011). Ras binds and activates the p110α catalytic subunit of PI3K, which in turn promotes transformation of the plasma membrane-bound lipid PIP 2 to PIP 3 and activation of AKT by binding through its pleckstrin homology (PH) domain (Cantley, 2002;Castellano & Downward, 2011). AKT interacts with many downstream effectors, including MDM2, BAX BAD and NF-κB. These interactions regulate apoptosis (Cantley, 2002;Castellano & Downward, 2011;Chang et al., 2003;Duronio, 2008). Furthermore, activation of FOXO by AKT promotes cellular metabolism (Engelman et al., 2006). Given that the MAPK and PI3K pathways are both strongly involved in controlling cell survival and death, it is clear that aberrant signalling within these pathways will lead to cancer, of which key hallmarks include sustaining proliferative signalling, enabling replicative immortality and resisting cell death (Hanahan & Weinberg, 2011).
Aside from these two classic Ras-dependent pathways, increasing evidence suggests the implication of normal and mutant Ras function in alternative mechanisms, such as redox homeostasis and stem cell survival (Mukhopadhyay et al., 2020;Wu et al., 2019). The role of Ras in cancer stem cells is yet to be fully elucidated. However, its crosstalk with the Wnt/β-catenin pathway in tumorigenesis may suggest its involvement in the maintenance of cancer stem cells (Moon et al., 2014). It is understood that, in colorectal cancer, there is overactivation of both the Wnt/β-catenin and Ras pathways, through mutations within APC (a tumour suppressor) and KRAS, respectively . There is a synergistic effect seen with these mutations (D'Abaco et al., 1996;Janssen et al., 2006;Jeong et al., 2018;Margetis et al., 2017). Mutations within APC cause loss of function of the tumour suppressor gene, whilst KRAS mutations lead to phosphorylation of key tyrosine residues within β-catenin, causing its accumulation within the cytoplasm. This ultimately increases activation of downstream Wnt pathway target genes, including REG4, a marker of cancer stem cells (Hwang et al., 2020;Janssen et al., 2006).
Up-regulation of this pathway is highly associated with the increased survival and plasticity properties seen in cancer stem cells and has been seen in many cancers (Al-Hajj et al., 2003;Koury et al., 2017;Lapidot et al., 1994). Taken together, it seems plausible that these KRAS mutations may play a role in the protection of the minor subset of cells, which likely have a key role in relapse.
Furthermore, Ras mutations can also be implicated in ROS   Table 1 Hence, the DNA of the healthy cells remains undamaged by ROS (Basak et al., 2017;Gorrini et al., 2013). However, in cancer, the NRF2 pathway can be up-regulated to constitutively degrade ROS (induced by chemotherapeutics), thereby conferring a protective effect to the cancer cells, resulting in chemoresistance (Basak et al., 2017;Gorrini et al., 2013). Therefore, the balance of NRF2 activation level is crucial.
In the last decade, it has been shown that KRAS G12D mutations can increase NRF2 transcription, through activation of the TRE (12-O-Tetradecanoylphorbol-13-acetate (TPA) response element) via the MAPK pathway (DeNicola et al., 2011;Mukhopadhyay et al., 2020;Shirazi et al., 2020;Tao et al., 2014). Thus, these KRAS mutations can render the cancer cell more capable of coping with chemotherapyinduced ROS, thereby mediating drug resistance.

| THE INVOLVEMENT OF Ras IN CHEMOTHERAPY RESISTANCE
Should mutations occur as described above, it follows that one or many of these pathways can be perturbed, leading to increased cell proliferation, decreased cell death and promotion of cancer stem cells, amongst other effects. The following scenarios illustrate some of these key resistance mechanisms, highlighting the necessity for better Furthermore, KRAS mutations can lead to hyperactivation of the PI3K-AKT pathway, which is starting to be implicated as a cisplatin resistance mechanism. As mentioned, up-regulation of the PI3K-AKT-mTOR pathway can have multiple effects, including inhibition of apoptosis and increased cell proliferation (de Vries et al., 2020).
Whilst there are other reasons for cisplatin resistance, Ras pathway mutations are heavily implicated in the key mechanisms. Thus, pharmacologically targeting Ras would provide an opportunity to overcome many causes of this resistance.
Up-regulation of Ras-mediated pathways as a means of chemoresistance is by no means restricted to cisplatin resistance and is a common mechanism of resistance to TK inhibitors (TKIs), such as those targeting RTKs including FLT3 and EGFR (Eberlein et al., 2015;Massarelli et al., 2007;McMahon et al., 2019;Ortiz-Cuaran et al., 2016;Piloto et al., 2007;Van Emburgh et al., 2016). TKIs are used in a variety of cancers, including renal cell carcinoma (RCC), colorectal cancer, acute myeloid leukaemia and nonsmall cell lung cancer, to name a few examples. The mechanism of action of TKIs involves inhibition of phosphorylation sites within the protein, thereby preventing it exerting kinase activity on downstream effectors (Ciardiello & Tortora, 2008;Yamaoka et al., 2018). However, resistance to these can occur through two predominant mechanisms,

| OPTIONS FOR TARGETING THE Ras PATHWAY
With improved capability of detecting minor cancer subclones, alongside the greater understanding of the impact of Ras mutations in various cancers, the next considerable challenge is improved pharmacological targeting of Ras. However, drugging Ras has proven exceptionally difficult. A key drawback is the lack of available binding sites for small molecule inhibitors. The nucleotide binding site (where GTP or GDP binds) seems a desirable pocket to target, however the picomolar affinity with which both GDP and GTP bind, as well as their high intracellular concentrations, effectively outcompetes the binding of any drug at this site (Cox et al., 2014;McCormick, 2018). Therefore, targeting alternative proteins within Ras-regulated pathways has been strongly investigated, with positive results seen, for example with trametinib, the MEK inhibitor. Currently approved for patients with BRAF V600-mutant metastatic melanoma or BRAFmutated (V600) non-small cell lung cancer, trametinib is well tolerated in patients (Lugowska et al., 2015;Odogwu et al., 2018) and is also being assessed in Ras-mutant myeloid malignancies (Borthakur et al., 2016). This compound binds to phosphorylated MEK and inhibits its downstream effectors (e.g. ERK), despite the presence of constitutive Ras signalling. Subsequently, aberrant growth signalling and apoptosis inhibition is reduced (Hofmann et al., 2012). Whilst this has shown promising results (Borthakur et al., 2016;Lugowska et al., 2015;Odogwu et al., 2018), this compound only inhibits the MAPK pathway downstream of MEK, so constitutive activation of other Ras-dependent pathways (e.g. PI3K-AKT) will still occur in the presence of Ras mutations even when treated with this drug. In this way, cancer can persist (Jones et al., 2019;Stinchcombe & Johnson, 2014). However, if Ras were to be targeted directly, signalling of both of these pathways would be inhibited, leading to cell death.
As often seen with many diseases, combination of trametinib with other therapeutics to inhibit multiple pathways together may reduce the likelihood of continued cancer signalling and potential development of resistance to this and other drugs (Infante & Swanton, 2014;Planchard et al., 2016;Zhou et al., 2020). However, even with the approved combination regimen of dabrafenib (BRAF inhibitor) with trametinib (Lugowska et al., 2015), only the MAPK pathway is inhibited, thereby maintaining the potential for aberrant PI3K pathway signalling, which can in itself cause resistance to MEK inhibitors (Jaiswal et al., 2009;Sos et al., 2009;Vitiello et al., 2019).
Alternatively, positive effects have been shown in vitro and in vivo of co-administering AKT inhibitors with dabrafenib to inhibit two Ras-mediated pathways (Lassen et al., 2014). However, combination of the two classes of drugs in patients did not yield significant clinical activity in reducing resistance seen in trametinib monotherapy, with 25% of patients exhibiting grade 3-4 toxicity (Algazi et al., 2018).
Taken together, whilst downstream pathway inhibition has proved successful, this treatment method does have considerable disadvantages and may not be a long-term solution for many patients.
Therefore, there is a distinct clinical need for novel means of treating Ras-mutant cancers, which could include the targeting of Ras itself.
Given the difficulties with targeting Ras, inhibition of elements upstream of Ras has been investigated. This includes inhibition of GEFs including SOS Ras/Rac guanine nucleotide exchange factor 1 ( SOS1), to reduce the likelihood of Ras maintaining its GTP-bound state and therefore inhibiting constitutive signalling (Evelyn et al., 2014;Hillig et al., 2019). For example, BAY293 is a first-in-class compound with the ability to bind directly to SOS1 and inhibit the Ras-SOS interaction and thereby downstream signalling of the PI3K and MAPK pathways (Hillig et al., 2019). Although the in vivo bioavailability for this compound was poor, the concept of Ras-SOS inhibition could prove useful in the future, with promising high throughput in silico and in vitro screening results serving as a proof of concept for inhibition of Ras via this mechanism (Evelyn et al., 2014(Evelyn et al., , 2015Hillig et al., 2019). More recently, BI-1701963, a SOS1-pan-Ras interaction inhibitor, has reached Phase I clinical trials, the first of its kind to do so. Modified from the structure of BI-3406, a quinazoline-derived compound, this novel inhibitor binds to the catalytic site of SOS1, preventing its interaction with inactive KRAS, thus inhibiting activation Hofmann et al., 2020).
In addition, alternative, more indirect pathways are also being targeted as a means of inhibiting Ras, which may also be able to eliminate the cancer stem cell. For example, the small molecule KYA1797K has been shown to be effective against oncogenic Ras in colorectal cancer and erlotinib-resistant non-small cell lung cancer (Park et al., 2019). This compound indirectly targets Ras, through inhibition of the Wnt/β-catenin pathway, which usually stabilises Ras (Jeong et al., 2012;Moon et al., 2014). This pathway is also up-regulated in cancer stem cells (Malanchi et al., 2008). KYA1797K has initiated anti- promoted apoptosis of both KRAS wild-type and mutant cells, questioning the specificity of the drug for cancer cells whilst sparing healthy cells. This study found both KRAS and β-catenin were overexpressed in tumour regions compared to non-tumour regions (Park et al., 2019). This may explain the limited toxicity seen during KYA1797K studies, with low doses of KYA1797K having a more substantial effect on tumour cells, compared to healthy cells. Indeed, only limited toxicity was seen during in vivo studies (Park et al., 2019).
However, protein expression varies considerably between tissues (with KRAS particularly highly expressed in the brain) and so targeting based on comparative expression between cancer and non-cancer regions in one tissue may not represent overall toxicity potential (Newlaczyl et al., 2017). Taken together, there may be a role for KYA1797K as a concomitant therapy in non-small cell lung cancer (as well as other cancers such as colorectal cancer). It presents a means of eliminating both the primary cause of the disease (if KRASmutated) and also minor subclones and cancer stem cells that could give rise to resistance . Nevertheless, better understanding of the drug's effects on other tissues with high Ras expression must be gained.

| OPTIONS FOR TARGETING Ras-DIRECT Ras TARGETING
Ras post-translational modifications were targeted as a means of preventing Ras trafficking to the membrane and therefore inhibiting downstream signalling. This included generation of farnesyltransferase inhibitors, including lonafarnib and tipifarnib (Van Cutsem et al., 2004). However, the effectiveness of these was questionable, with most patients with KRAS-mutated diseases (such as pancreatic cancer and leukaemia patients) receiving no clinical benefit from these farnesyltransferase inhibitors (Borthakur et al., 2006;Burnett et al., 2012;Harousseau et al., 2009;Van Cutsem et al., 2004).
Inefficacy was largely due to the redundancy mechanism of geranylgeranyltransferase, which sufficiently modifies KRAS in the absence of farnesyltransferase to permit its trafficking to the membrane (Basso et al., 2006;Whyte et al., 1997). Interest in this strategy has recently been revived with new personalised medicine approaches now capable of identifying patients harbouring HRAS-or NRAS-driven cancers that are more likely to respond (Gilardi et al., 2020;Lee et al., 2020).
In recent years however, more promising steps have been made regarding direct inhibition of oncogenic Ras. A key feature of this has been the discovery of novel potential binding pockets for small molecule inhibitors, to inhibit GEF activity or effector binding (Cruz-Migoni et al., 2019;Maurer et al., 2012;Ostrem et al., 2013). Fragment-based screening identified a previously undiscovered hydrophobic pocket located between the Switch I and II regions (termed S-IIP), which was successfully targeted by Ostrem et al. (2013) (Figure 4a). Binding of peptide fragments was specific to G12C-mutated KRAS since the compounds functioned through irreversible cysteine binding in this particular pocket (and not with other cysteines found in wild-type KRAS). Other key residues within this pocket include, but are not limited to, V7, V9, M72, F78, Q99 and I100. In vitro models of KRAS G12C-mutated lung cancer treated showed decreased survival upon treatment with these compounds, with inactive Ras (Ras-GDP) levels considerably greater than Ras-GTP. Further analysis showed that the conformational disruption caused by binding of these fragments reduced interactions with both SOS and effector molecules and pathways, including B-RAF, C-RAF and the PI3K pathway (Gentile et al., 2017;Ostrem et al., 2013).
The binding of these fragments to Ras also reduce SOS-catalysed nucleotide exchange, a method of Ras inhibition which had been previously explored. Compounds acting in this way either inhibited conversion of Ras-GDP to Ras-GTP (Patgiri et al., 2011) or increased the amount of Ras-GTP to such a level that it inhibited ERK phosphorylation, since overactivation of Ras can be cytotoxic (the Ras "sweetspot model") . Either way, these compounds were shown to inhibit the MAPK pathway but were largely tested against  (Lito et al., 2016;Patricelli et al., 2016). This compound has subsequently been fully characterised in silico and these studies revealed a dynamic nature of the SII-P pocket, a feature which could be utilised in further study (Khrenova et al., 2020).
Based on this work, alternative iterations of KRAS G12C inhibitors have been produced. This was required since the probability of ARS-853 locking KRAS in its inactive state in vivo was debatable, given a lack of understanding regarding the cycling efficiency of Ras between its inactive and active states. It had been deduced in vitro that the G12C mutation permits rapid cycling of KRAS between these states, hence permitting the binding of ARS-853 (Figure 4b). However, the possibility of finding the correct therapeutic window to translate this compound to in vivo work proved complex (Janes et al., 2018;Lito et al., 2016;Patricelli et al., 2016). Thus, alternatives including ARS-1620 were developed to improve in vivo capability ( Figure 4c). Modifications to the ARS-853 structure resulted in favourable pharmacokinetic (PK) properties, permitting a greater understanding of KRAS activation status and dependency in vivo (Janes et al., 2018). ARS-1620 is an orally bioavailable quinazolinebased compound with limited side effects witnessed in preclinical animal models (Janes et al., 2018;Li et al., 2018). Optimisation from ARS-853 by inclusion of a fluorophenol, hydrophobic binding group permitted stronger covalent (irreversible) binding within the SII-P pocket (more specifically, interaction with H95 in this pocket), thus  (Figure 4). The effects of this compound remained mutation specific, thereby eliminating the risk of binding to KRAS wild type in non-tumour cells and thus reduced toxicity potential. This compound was well tolerated in patient-derived xenograft mice, where a reduction in tumour burden through decreased Ras-mediated downstream signalling was evident. This was the first example of an in vivo trial using a compound targeting the SII-P pocket (Janes et al., 2018).
From this, AMG-510 was developed, following further modifications to the ARS-1620 structure, including addition of more aromatic entities (Figure 4d). This enables AMG-510 to bind within a slightly different groove of SII-P, enhancing potency and selectivity (Canon et al., 2019). Promising preclinical in vitro and in vivo experiments indicated arrest of the MAPK pathway and induction of a pro-inflammatory tumour micro-environment. This compound has since progressed into clinical trials, the first KRAS G12C-selective inhibitor to do so. Early data indicate that out of 29 patients evaluable for response at the time of publication, 5 exhibited partial response, 18 had stable disease and 6 had progressive disease. As with the previous ARS-1620 animal studies, AMG-510 was generally well tolerated, with no dose limiting toxicities recorded (Govindan, 2019;Romero, 2020). However, larger cohorts and longer trials are imperative in determining the true impact of this compound. Such compounds could help combat the KRAS-mediated resistance to monoclonal antibodies seen in non-small cell lung cancer and colorectal cancer (Lièvre & Laurent-Puig, 2009;Park et al., 2019). However, success of this compound and indeed this targeting mechanism, is likely restricted to KRAS G12C-mutant cancer since some key residues for binding of this compound are unique to KRAS and not conserved between the different isoforms. Indeed, in vitro work completed by Ostrem et al. (2013) illustrated that transduction of lung cancer cell lines with KRAS G12V rescued the cancerous phenotype (resistance to cell death, increased proliferation), thereby illustrating how this mutation renders resistance to KRAS G12C inhibitors, as expected.

Alternative KRAS G12C inhibitors are also in development and
showing considerable promise, including MRTX849 (Fell et al., 2020;Hallin et al., 2020a). In a similar way to AMG-510 and the ARS compounds discussed above, MRTX849 covalently binds to the inactive form of KRAS, in SII-P (Figure 4e). This induces apoptosis through down-regulation of the MAPK pathway. Interestingly, the PI3K-AKT pathway remained relatively unaffected by MRTX849 (Hallin et al., 2020b). In vivo trials with MRTX849 exhibited favourable pharmacokinetic/pharmacodynamic properties (Fell et al., 2020) and both cell line-and patient-derived xenograft modelling of pancreatic and lung cancers also indicated up to a 30% reduction in tumour burden (Fell et al., 2020;Hallin et al., 2020b). Individual patient case studies from Phase I trials have also shown MRTX849 to be effective in reducing tumour burden in both lung cancer and colorectal cancer, although these data are largely incomplete (Hallin et al., 2020b). Taken together, it is clear that KRAS G12C inhibitors have promise as a means of abrogating Ras-mediated resistance, although there remain drawbacks which need assessing, most notably, the lack of efficacy against other Ras mutations, which are prevalent across Ras-mutated disease.
Clearer understanding of the structure of Ras has permitted not only elucidation of the SII-P pocket but also alternative binding sites, including a pocket between the Switch I and II regions of Ras (termed pocket I). Key residues available for interaction with small molecule compounds include K5, L6, V7, V8, S39, D54, I55, L56, Y71, T74, G75 and E76 (Maurer et al., 2012;Quevedo et al., 2018). Antibody-fragment-directed site exploration can be used to explore and exploit previously unconsidered drug interaction sites. In the case of Ras, this mechanism has been used to analyse potential compound binding sites within the previously identified pocket I that could be targeted using small molecule inhibitors to interrupt effector proteins (such as c-RAF and p110α) from binding to Ras via the Ras binding domain (Maurer et al., 2012;Quevedo et al., 2018). This would provide an alternative mechanism of abrogating the effects of oncogenic Ras activation to those discussed previously and early results showed effectiveness of antibody-derived compounds against a range of Ras mutations and isoforms (Quevedo et al., 2018). However, given high affinity binding of certain effectors to Ras (such as PI3K and B-RAF, with 3.2-and 0.04-μM affinity, respectively) (Erijman & Shifman, 2016), the high EC 50 s of the compounds identified in these in vitro assays mean that many further modifications would be required to convert these putative compounds into usable therapeutics. Nevertheless, such antibody-derived fragments have the potential to be fused with small molecule protein-protein interaction inhibitors to improve efficacy (Cruz-Migoni et al., 2019). Whilst many Ras-effector interaction inhibitors have been trialled preclinically, none have been implemented in the clinic in the context of Rasmutant cancer, owing to lack of efficacy or toxicity potential (Canon et al., 2019;Keeton et al., 2017). However, crystal structure determination showed that fusion of compounds developed through antibody-derived fragment screening with known small molecule protein-protein interaction inhibitors results in better binding within pocket I, thus inhibiting Ras-effector interactions with a lower EC 50 .
Nevertheless, the therapeutic use of pocket I may be restricted since such a pocket has also been detected in wild-type Ras (Cruz-Migoni et al., 2019), thus increasing the risk for on-target toxicity.
In recent times, inhibition of the Ras-effector interaction has been seen through competitive binding of rigosertib at the Ras binding domain. This compound elicits effects against MAPK, PI3K and RAL pathway activation, in both wild-type and mutant Ras situations (Athuluri-Divakar et al., 2016). Inhibition of multiple Ras-mediated diseases have been seen in response to rigosertib, including pancreatic cancer and leukaemia (Athuluri-Divakar et al., 2016;Baker et al., 2019). Rigosertib is moderately to well tolerated in clinical trials thus far, although is yet to be specifically tested in the context of Ras-mutant cancer (Bowles et al., 2014;Navada et al., 2020). As with the antibody-derived, small molecule proteinprotein interaction inhibitors described above, toxicity may be as a result of rigosertib's ability to target both wild-type and mutant KRAS and thus, further studies are needed to fully assess the impact of such a drug on healthy cells. Antibody therapy is also currently being explored as means of direct Ras targeting. However, a major drawback has been the capability of the antibody to cross the cell membrane, as with any proteinbased therapy (Bolhassani et al., 2017). Therefore, the development of inRas37, a pan-Ras-targeting antibody, is a considerable step forward in targeting Ras. Although the cellular uptake remains low (approximately 4%), in vitro and in vivo work has shown promise in the potential of inRas37 to inhibit both the MAPK and PI3K-AKT pathways, in a dose-dependent manner (Shin et al., 2020). Briefly, this drug binds to integrins αVβ3 and αVβ5 on the cancer cell surface which then undergo endocytosis. The antibody "escapes" the endosome as a result of pH-determined cleavage (the antibody is cleaved from the integrins better at pH 7, the cytoplasmic pH, compared to pH 6.5, the pH of early endosomes) (Podinovskaia & Spang, 2018;Putnam, 2012;Shin et al., 2020). The antibody then co-localises with Ras to block the effector binding site, in a manner similar to the protein-protein interaction inhibitors described above. Mutations introduced into the general antibody structure render it specific for mutant Ras binding, with little activity against Ras wild type, thereby limiting on-target toxicity.
inRas37 has greater effect on cells with a greater dependency on Ras signalling, which can include some tumour cells (Weinstein, 2002;Yi et al., 2020). Nevertheless, when tested on large tumour models (spheroids and cell-derived xenograft mice), efficacy decreased considerably (Shin et al., 2020). This therefore shows that treatment with this drug may be suitable for patients at earlier stages of Ras-mutated cancer, before tumour burden is too great and reduces drug efficacy.
Thus, it could be used to treat Ras-initiated relapse as soon as it occurs. However, its utility in eliminating minor subclones prior to relapse is limited, given the low cellular uptake and the need for a high Ras dependency in the cell.

| RESISTANCE TO Ras-TARGETING AGENTS
Of course, there is potential for resistance to any therapeutic and Ras-targeting drugs are no different. Some of these have been previously discussed here, such as the use of geranylgeranyltransferase to overcome the effects of farnesyltransferase inhibitors (Whyte et al., 1997) or the up-regulation of alternative Ras-mediated pathways, as seen when patients are treated with MEK inhibitors (Vitiello et al., 2019). Other mechanisms of resistance are also possible, such as the reactivation of ERK, which may be a potential resistance mechanism in the case of a Ras-targeting agent (Bruner et al., 2017;Ercan et al., 2012;Ochi et al., 2014). This has already been seen in the case of EGFR-inhibitor resistance, whereby negative regulators of ERK are down-regulated, so pro-apoptotic BIM is not fully upregulated and so cannot fully induce apoptosis. Alternatively, the gene encoding ERK1, MAPK1, is amplified. These scenarios resulted in in vitro and in vivo resistance to the putative EGFR inhibitor WZ4002 (Ercan et al., 2012). ERK reactivation has also been found to contribute to gefitinib resistance but, in this case, was found to be mediated by Src (Ochi et al., 2014). Given these kinases are either side of Ras, their co-operation could result in resistance to a Ras inhibitor. However, Src-mediated ERK reactivation is avoidable through treatment with Src inhibitors (Ochi et al., 2014), which may present a means of overcoming this potential Ras-inhibitor resistance mechanism. Taken together, these studies imply that whilst resistance to Ras-targeting drugs is possible, there are already means of overcoming this resistance, just as a Ras-targeting drug would provide the means of overcoming Ras-mediated resistance.
Whilst other elements of pathways contributing to resistance can be targeted relatively easily, acquisition of secondary mutations in Ras present a more pressing problem. For example, at present, the G12Cspecific inhibitor is perhaps the most developed means of inhibiting Ras; however, a subsequent mutation in Ras would most likely render this inhibitor ineffective. This has already been evidenced in studies into KRAS G12C inhibitors, whereby rescue experiments with the G12V mutation restored the cancerous phenotype (Ostrem et al., 2013). Therefore, a pan-mutation-targeting drug, such as a derivative of inRas37, may be favoured.

| DISCUSSION
Ras mutations in cancer and chemoresistance are important when considering patient prognosis. Whilst it seems inevitable that resistance will be an issue for a long time to come, better targeting of potential causes is imperative. Advances in Ras inhibition could help reduce the risk of resistance and relapse for a wide range of cancers, given its mutational frequency. At present, some success has been seen when multiple drugs are used to target different pathways implicated in Ras-mutated disease. However, other studies have shown lack of long-term efficacy when combining multiple therapies. Therefore, single agent, multi-pathway targeting agents, including direct Ras inhibitors, are becoming more heavily researched. It will be interesting to see the effects of these in vivo, since this type of therapy may reduce the potential for future development of drug resistance by up-regulation of an alternative pathway. Targeting a common factor at the centre of multiple pathways may target more cancer cell types and thus reduce the heterogeneity of the tumour, eliminating potential resistance causes before they become dominant.
Previous failures of Ras-targeting agents, including farnesyltransferase inhibitors, as well as perceived unfavourable protein dynamics, have resulted in Ras being largely considered undruggable.
However, development of structural analysis techniques and a clearer understanding of the key residues in Ras has altered this thinking, with new compounds with novel, allele-specific Ras binding mechanisms showing great promise. Phase I trials of AMG-510, a first-in-class direct Ras-targeting agent, suggest a turning point has been reached in this field of study. Whilst there is much more to be done, these preliminary data indicate a solution for Ras-mediated resistance may be possible.
Nevertheless, a greater understanding of resistance must be gained before relapse risk can be eliminated. Despite evidence supporting the cancer stem cell theory, limited standard-of-care detection sensitivities for initial diagnostic samples prevent identification of the minor subclones present at diagnosis (McMahon et al., 2019). It can therefore be difficult to determine likely causes of relapse upon initial diagnosis and so constant monitoring for changes in expression of genes commonly implicated in drug resistance, such as Ras, may provide a useful tool for predicting disease trajectory.
Nevertheless, this is only useful if the effects of the acquired/emergent mutations can be abrogated. In the cases discussed here, improved analytical tools would ideally be combined with Rastargeting agents to prevent resistance taking hold.
Ultimately, chemoresistance, either intrinsic or acquired, due to Ras mutations, whether primary or secondary, remains a considerable problem. The concepts presented here, amongst many other examples, illustrate the necessity for Ras-targeting drugs. There is a distinct clinical requirement for the improved targeting of Ras in cancer, with Ras implicated in both initial disease presentation and relapse. Although no universal, direct Ras inhibitor has yet been achieved, considerable progress has been made in the last decade with the advent of allele-specific inhibitors. This brings promise to the field, with the potential for better treatment of Ras-initiated resistance a real prospect.

| Nomenclature of targets and ligands
Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org and are permanently archived in the Concise Guide to PHARMACOLOGY 2019/20 Alexander, Fabbro, Kelly, Marrion, et al., 2019;.

CONFLICT OF INTEREST
The authors declare no conflicts of interest.