A New View of Activating Mutations in Cancer

Abstract

A vast effort has been invested in the identification of driver mutations of cancer. However, recent studies and observations call into question whether the activating mutations or the signal strength are the major determinant of tumor development. The data argue that signal strength determines cell fate, not the mutation that initiated it. In addition to activating mutations, factors that can impact signaling strength include (i) homeostatic mechanisms that can block or enhance the signal, (ii) the types and locations of additional mutations, and (iii) the expression levels of specific isoforms of genes and regulators of proteins in the pathway. Because signal levels are largely decided by chromatin structure, they vary across cell types, states, and time windows. A strong activating mutation can be restricted by low expression, whereas a weaker mutation can be strengthened by high expression. Strong signals can be associated with cell proliferation, but too strong a signal may result in oncogene-induced senescence. Beyond cancer, moderate signal strength in embryonic neural cells may be associated with neurodevelopmental disorders, and moderate signals in aging may be associated with neurodegenerative diseases, like Alzheimer's disease. The challenge for improving patient outcomes therefore lies in determining signaling thresholds and predicting signal strength.

Introduction

Massive efforts have focused on searches, identification, and analyses of activating mutations, particularly those involved in cancer initiation and drug resistance, and multiple reviews have described them (1–10). Some are broad range; some involve specific cancers. Some focus on cancer hotspots; some on rare mutations (11–19). Some involve method development, particularly computational because they require large-scale searches and identification; some involve applications (20–22). Underlying these massive community efforts is the hypothesis that cancer treatment would be considerably better if guided by knowledge of the mutations that drive it. This conception has inspired major initiatives funded by government agencies that focus on whole-genome and/or exome sequencing (The Cancer Genome Atlas Program: https://www.cancer.gov/about-nci/organization/ccg/research/structural-genomics/tcga/), creation of large databases, and developing tools for their statistical analyses (21, 23–30). All strive to identify actionable alterations, and consequently targets, in patients harboring them (31–35). These couple with major structural initiatives to obtain the structures of the wild type and mutant proteins and unravel the activation mechanisms of the mutations (19, 36, 37). The compendium of observations over the last two decades, now topped with at least four significant recent publications (38–41) inspire a new view of activating driver mutations and signal transduction. They lead us to suggest that even though activating mutations are critical agents in cancer development, signal strength decides the cell fate, not the mutations that may initiate it. This innovative view articulates new goals and challenges in fundamental cancer research and in pharmacologic targets. This is further supported by replication repair deficient cancers, which are highly mutated in the Ras/MAPK pathway and with the respective proteins highly expressed. They have been shown sensitive to MEK inhibitors, underscoring pathways as therapeutic options for hypermutated, and expressed cancers (8).

A fundamental aim in cancer research is to decipher the hallmarks of cell transformation (42). Identification of all genes with mutations capable of driving tumor development is a vital component. A further challenge lies in decoding these mutations within the complex biological framework, accounting for their locations, quantity, and regulation of protein expression and signal propagation (43–53), and revealing targeting opportunities (54). Activating cancer drivers emit signals. The multiple steps along their propagation may enhance, dampen, or reroute them. The regulation of each can go awry. Mukherjee and colleagues's unearthing of the regulation of PTEN translation by PI3K signaling thereby maintaining pathway homeostasis (39), and Ingram and colleagues's unraveling of the role of NK2 homeobox 1 (NKX2–1) transcription factor in controlling lung cancer progression by inducing dual specificity phosphatase 6 (DUSP6) to dampen ERK activity (38), are remarkable examples of recent observations in different proteins and pathways. Notably, in both cases the homeostasis control is via the action of phosphatases. As it turns out, in a 2014 review in Oncogene, Stebbing and colleagues pointed to such regulatory roles of phosphatases in cancer (55). However, at the time, they were unable to provide the detailed mechanisms that Mukherjee and colleagues (39) and Ingram and colleagues (38) have now been able to work out. Chen and colleagues undertook different approach and scale of single-cell transcriptomics of the apparent opposing, anti- and protumorigenic roles of Src homology 2 (SH2) domain-containing phosphatase 2 (SHP2) in Myc-driven liver tumor cells and in the microenvironment (41). They revealed that the tumors develop selectively from rare SHP2-positive cells in a SHP2-deficient cell population. Finally, Guo and colleagues showed that alternating-current electric field activates ERK independently of growth factors (40). Although not shown for activating mutations, electrical synchronization, and modulation of the dynamics of ERK activation is likely to similarly regulate cell fate. This underscores signal modulation as the key, rather than specifically activation mutations.

Cancer and Activating Driver Mutations

Cancer has been described as abnormal and uncontrolled cellular growth resulting primarily from genetic mutations that confer selective advantages on their harboring cells (42, 56). Oncogenes act in a dominant way (57). A single altered copy leads to unregulated growth. In contrast, in the case of tumor suppressor genes, both copies are defective in uncontrolled cell division, as shown early on for retinoblastoma tumor suppressor RB1 (a gene encoding pRb), where both alleles are altered (58). Solid tumors can contain millions of mutations (59, 60). A few frequent ones are classified as strong drivers (61–63). Driver mutations have a large impact on fitness, thus are not expected to occur in germ-line DNA (64) although genetic ancestry have recently been correlated with driver mutations (65). Some strong drivers are statistically rare, as are “latent” drivers (66). Rare and latent drivers can be challenging to identify in protein sequences (11, 15, 67–74). Efforts have focused on analyses of three-dimensional (3D) structures (75), with the premise that activating driver mutations are likely to be clustered (18, 27, 76, 77), which would enhance their actions. Most mutations are passengers, although they can also act as latent drivers whose mild or moderate actions are difficult to discern. In addition to single residue substitutions, chromosomal alterations including gene duplication and deletions impact transcription and protein expression levels (Overview of Chromosomal Mutations, Types & Examples: https://www.bioexplorer.net/chromosomal-mutations.html/; Chromosomal Mutation- Definition, Causes, Mechanism, Types, Examples: https://thebiologynotes.com/chromosomal-mutation/). Although cancer has been associated with accumulation of mutations in oncogenes and tumor suppressor genes, the number of mutations that are required for cell transformation has been unclear. In 2015, Tomasetti and colleagues postulated that only three driver gene mutations are required for the development of lung and colorectal cancers (59), a number that is lower than that thought to be required for the formation of cancers of these and other organs. Below we propose a definition for mutation strength and that signal strength is the pivotal element, not the number of mutations. The influence on the emitted signal strength may also suggest why one mutated allele is sufficient for oncogenes, but two for tumor suppressors with subtle dosage effects (78). The challenge is to measure signal strength, quantitate the elements deciding it and determine the threshold required for transformation (79).

Activating Mutations: What Makes a Mutation a Strong Driver?

Some activating mutations are strong, others relatively weak. What makes a driver mutation strong? In our view, activating mutations work by mimicking the protein physiological activation mechanism (36, 80). Mutants harboring strong activating mutations shift the ensemble of the inactive state to sample more frequently the conformation populated by the protein's physiological active state. According to our definition, mutations that strongly bias the ensemble toward the active state are strong drivers. Among the mutations, those with stronger population bias, are stronger activating mutations. Below we provide examples of proteins from the Ras oncogenic signaling network.

K-Ras4B, a splice variant of the K-Ras isoform, works by switching between its active and inactive states (Fig. 1A; refs. 80–92). It is frequently mutated in epithelial cancers (1, 93–96), resulting in elevated cell survival and proliferation. Like all Ras superfamily proteins, K-Ras4B is inactivated by GTPase-activating proteins (GAP) that hydrolyze the GTP to GDP (91, 97). Strong drivers, such as G12D, G12V, and G12C, as well as the Q61L substitution, abolish GAP's action (Fig. 1B). The consequent GTP-bound active K-Ras4B molecules stimulate oncogenic signaling (98–102). Intrinsic hydrolysis and especially exchange of GDP by GTP to maintain Ras activity are stimulated by weaker drivers such as K-Ras4B^A146T (103). Mechanistically, GTP hydrolysis involves proton transfer from an attacking water molecule to another. A different proton is then transferred to the GTP. With extended side-chains, the G12 substitutions sterically block the penetration of the arginine finger of GAP and thereby hinder hydrolysis (104). The action of Q61L differs. Q61 stabilizes the transient OH⁻ and H₃O⁺ molecules, which lower the transition state barrier. L61 cannot perform this action (105). K-Ras4B^G12D was shown to have a higher rate of GTP hydrolysis versus K-Ras4B^G12V (106), which can explain why G12D is the strongest among the Ras hotspots, followed by G12V. In line with expectation, despite the global similarity among the crystal structures of the mutant proteins and wild-type Ras (106), their populated conformations vary (107, 108). Solution studies suggest not one conformer as in the crystal but differing distributions of the conformational ensembles of the wild type, K-Ras^G12V, and K-Ras^G12D (109). The altered conformational bias may clarify their distinct rates of GTP hydrolysis, nucleotide exchange, and phospholipid selectivity, as well as the different favored effectors; K-Ras^G12V interacts with Raf-1 and signals through MAPK and Ral guanine nucleotide dissociation stimulator (RalGDS), whereas K-Ras^G12D activates PI3K, c-Jun N-terminal kinase (JNK), p38, and focal adhesion kinase (FAK) pathways (110, 111). Their sensitivity to drugs also differs (Fig. 1C; refs. 112–114). G12D/V/C mutants, but not Q61X mutants, stabilize a conformation resembling the active state through interaction with the Switch II (SII). Long-timescale molecular dynamics (MD) simulations of active (GTP-bound) and inactive (GDP-bound) wild-type and K-Ras^G12D mutant show that the mutation shifts the local population, especially in the SII and α3-helix regions, favoring structural changes, resulting in a catalytically impaired conformation. It also causes SII motions to anticorrelate with other regions (109).

Figure 1.

Regulation of the Ras GTP–GDP cycle. A, Wild-type Ras is activated by the nucleotide exchange factor (GEF) via the GDP-to-GTP exchange and deactivated by the GTPase-activating protein (GAP) via the GTP-to-GDP hydrolysis. Highly populated conformational ensembles represent the active and inactive states for the GTP-bound and GDP-bound Ras, respectively. B, Oncogenic driver mutations at the position 12, such as G12D, G12C, and G12V, impair the GAP-mediated hydrolysis, retaining Ras in a constitutively active GTP-bound state. These strong activating mutations shift the populations, strongly biasing the ensembles of the conformations toward the active state. C, Examples of Ras inhibitors embedded in the crystal structures of K-Ras4B. These are MRTX1133, a noncovalent inhibitor for K-Ras4B^G12D (PDB ID: 7RPZ); AMG510 (sotorasib), a covalent inhibitor for K-Ras4B^G12C (PDB ID: 6OIM); and H-REV107, a peptide inhibitor for K-Ras4B^G12C (PDB ID: 7C41). These inhibitors target the Switch II (SII) pocket in the GDP-bound form.

Protein and lipid kinases in the Ras signaling network provide additional examples. Inactive kinases populate the αC-helix-out conformation (80). Activation involves switching to the αC-helix-in conformation. The switch is coupled with formation of a salt bridge between the β3-Lys and the αC-Glu, and formation of an R-spine. Oncogenic mutations mimic the physiologic activation by stabilizing the active αC-helix-in conformation or destabilizing the αC-helix-out, as exemplified by the L858R mutation in EGFR, which stabilizes the αC-helix-in conformation by heterodimerization. Being in the hydrophobic core, substituting Leu by Arg also destabilizes the αC-helix-out conformation. The T790M mutation in EGFR, T315I in Bcr-Abl, T334I in c-Abl, T341I in Src, and T670I in Kit act by enhancing the stability of the hydrophobic R-spine, thus the active state. In another example, PI3Kα lipid kinase is an obligate heterodimer of the p85α and the catalytic p110α subunits. Physiologic activation involves nSH2 release by a receptor tyrosine kinase (RTK), such as an insulin receptor, which triggers conformational changes in p110α, resulting in exposed kinase domain for membrane binding. The interaction of the nSH2 with the phosphorylated tyrosine pYXXM motif at the C-terminal of an RTK relieves PI3Kα’s autoinhibition (115–117). Structural rearrangement of the kinase domain reduces the ATP-substrate distance for phosphoryl transfer to phosphatidylinositol 4,5-bisphosphate (PIP₂) producing phosphatidylinositol 3,4,5-trisphosphate (PIP₃). Single activating mutations in PIK3CA (a gene encoding the catalytic p110α subunit of PI3Kα) are common in tumors (118), including colon, ovary, breast, brain, liver, stomach, and lung (119). Strong mutations (120–124) include E542K and E545K that mimic RTK's action in relieving the autoinhibition, lowering the barrier height of the transition state (k_a) by reorganizing the active site. H1047R can substitute for the action of Ras at the membrane. It increases the population time of the PIP₂ in the active site (k_m; refs. 37, 125–127). Weak mutations (E453K/Q, E726K, and M1043V/I) often collaborate with the hotspots (E542K, E545K, and H1047R; refs. 16, 37, 123).

The impact of the oncogenic mutations on cancer development is due to the strength of the activation signal that they transmit. The key then is signal strength. Strong drivers will emit strong signals. However, as two of the recent papers show (38, 39), under certain circumstances and time window, negative regulation by phosphatases, rendered at the chromatin level, can practically quell the signal, making activation mutations largely moot. This argues that a focus merely on activating mutations in the target protein may not accurately portray the cancer landscape and thus pharmacology in a patient. That the key is the emitted signal, rather than the activating mutation has been elegantly shown by the third paper (40), and can explain the apparent perplexing observation emphasized in fourth paper (41). To capture the composite signal strength, we have recently suggested a signaling by-the-numbers paradigm of activating driver mutations and signal transduction (79). In our view, activating mutations are critical elements in cancer development; however, it is signal strength that determines the cell fate, not the mutations that may initiate it.

Activating Mutations Are Critical Elements in Cancer Development, but It Is Signal Strength That Determines the Cell Fate

PTEN and DUSP6 are homeostatic guardians of proper cell function (Fig. 2). They are negative regulators of both physiologic and oncogenic signaling. They can act against growth factor ligands-stimulated RTKs and activating mutations by reducing the signal output. However, when the pathways are hyperactivated in cells hijacked by tumor growth both repressors are lost. Protein phosphatase 2 (PP2A) and SHP2 are also tumor suppressor phosphatases with data at different levels.

Figure 2.

Phosphatases in cell homeostasis. In the PI3K/AKT/mTOR pathway, PI3K phosphorylates PIP₂ to PIP₃, and PTEN dephosphorylates PIP₃ back to PIP₂. Mutations in PTEN impair the tumor suppressor activity. PDK1 and mTORC2 phosphorylate and activate PIP₃-bound AKT. Then the active AKT regulates the activation of mTORC1, leading to cell growth. Upon activation, mTORC1 phosphorylates S6K1 and 4E-BP1. Phosphorylated S6K1 phosphorylates eIF4B and S6, leading to increased protein translation. The phosphorylation of 4E-BP1 causes its release from eIF4E, initiating PTEN translation. In physiologic or oncogenic activation of PI3K/AKT/mTOR signaling, the PTEN expression is increased, controlled by mTORC1/4E-BP1-dependent translation. However, tumor treatment with PI3K inhibitors causes inhibition of the PTEN translation, reactivating the pathway. The MAPK pathway involves Ras and kinase cascades, including Raf, MEK, and ERK. Kinase suppressor of Ras (KSR) is a scaffolding protein, facilitating the MAPK activation. Oncogenic Ras hyperactivates MAPK signaling, leading to cell proliferation. DUSP6, a member of the MAPK phosphatases (MKP), dephosphorylates ERK via negative feedback. In lung adenocarcinoma (LUAD), NKX2–1 induces DUSP6 transcription, which is delivered to the ribosome for DUSP6 translation. The cytoplasmic phosphatase DUSP6 deactivates ERK in the cytosol, whereas its nuclear counterpart DUSP5 deactivates ERK in the nucleus. In the cartoon, red explosion shapes denote mutations.

PTEN dephosphorylates signaling lipid PIP₃, produced by PI3K lipid kinase, back to PIP₂ (128). Physiologic activation of PI3K, for example, by insulin receptor, homeostatically increases the expression of PTEN to safeguard emission of a too-strong signal through the PI3K/protein kinase B (AKT)/mTOR pathway, which leads to cell growth, a prerequisite for cell division and proliferation. Recently, Mukherjee and colleagues made the observation that not only physiologic stimulation, but mutations can similarly unleash PTEN expression (39). PTEN can then dampen the physiologic and oncogenic signals, setting a threshold on the duration and output of the pathway. Suppressing PI3K activation with inhibitors reduces its signal, thus PTEN expression, leading to the tumor bouncing back. The authors further show that PTEN expression is controlled by its mTOR-dependent translation through eukaryotic translation initiation factor 4E (eIF4E)-binding protein 1 (4E-BP1). mTOR complex 1 (mTORC1) phosphorylates S6 kinase 1 (S6K1, a.k.a p70S6) and 4E-BP1 to stimulate protein synthesis (Fig. 2, left), whereas mTOR complex 2 (mTORC2) phosphorylates AKT to promote cell survival (129).

In a similar vein, despite the expectation that the emergence of strong activating mutations in Ras would lead to an increase in ERK signaling through the MAPK phosphorylation cascade, data indicated otherwise. Early in oncogenesis the changes are minor. Why ectopic expression of K-Ras mutants results in strong ERK activation (130, 131), whereas introduction of a strong GTPase-defective activation mutation such as K-Ras^G12C in tumor cell lines can reduce dual-phosphorylated ERK levels (132) has also been puzzling. In addition, experiments involving introduction of the mutation in a single K-Ras allele failed to detect an increase in ERK activity, and even observed a decrease (131–135). Further, the levels of ERK activity across cells harboring the mutant K-Ras^G12C isoforms were similar. As in the case of PTEN, however, it has been noticed that the level of DUSP6, an ERK phosphatase, is consistently elevated in the presence of ERK inhibitor, and that DUSP6 expression is suppressed not only in K-Ras^G12C cells but also by physiologic EGF stimulation of cells harboring wild-type K-Ras (133). Although the expectation has been that homeostatic feedback mechanisms that limit ERK activity are at play, this behavior has been perplexing, and the mechanism has also been unclear. Why a strong Ras activation mutation fails to increase ERK signaling output, and exactly how DUSP6 is involved, have been clarified by Ingram and colleagues (38). They showed that the lung lineage transcription factor NKX2–1 suppresses ERK signaling. NKX2–1 induces the ERK phosphatase DUSP6, which inactivates ERK (Fig. 2, right; ref. 136). However, silenced NKX2–1 in late-stage cancer cells promoted DUSP6 transcription and suppressed metastasis. They further showed that DUSP6 is a requirement for NKX2–1-mediated inhibition of tumor progression. Thus, suppression of NKX2–1 and consequently DUSP6 can promote strong ERK activation. How NKX2–1 works in lung cells has been explored albeit in different contexts (137). The pulmonary alveolar epithelium is mainly composed of alveolar type I (AT1) and type II (AT2) cells. NKX2–1 promotes their opposite fates. NKX2–1 cell-type-specific functions result from distinct chromatin binding scenarios during development in the presence and absence of YAP/TAZ (yes-associated protein/transcriptional coactivator with PDZ-binding motif).

PP2A tumor suppressor (138) acts to dephosphorylate Myc at S62, leading to Myc's degradation. Its suppression results in a high overall pS62/pT58-Myc ratio (RAS and MYC: Co-conspirators in Cancer: https://www.cancer.gov/research/key-initiatives/ras/ras-central/blog/2017/myc-ras/; refs. 139–141). Phosphorylation provides a pool of Myc that can readily bind its target genes to drive proliferation (139). Interaction of the oncoprotein Myc, a potent transcription factor, with its chromatin cofactor WD repeat-containing protein 5 (WDR5), colocalizing on chromatin at genes involved in protein synthesis, is essential for tumor maintenance (142).

Finally, at the population level, the apparent paradoxical SHP2 phosphatase observations could be clarified by considering the population of cells in the microenvironment. SHP2 acts in several pathways including Ras/Raf/MEK/ERK, PI3K/AKT, JAK/STAT, and programmed death receptor 1/programmed cell death ligand 1 (PD-1/PD-L1; ref. 143). The SH2-containing tyrosine phosphatase transmits RTKs and cytokine receptors signals under physiologic conditions and in cancer (41, 144, 145). Mutant PTPN11 (a gene encoding SHP2) is common in leukemia and solid tumors (146, 147). Its inhibition can block oncogenic RTK signaling (148–155). At the same time, its deficiency worsened a carcinogen-elicited hepatocellular carcinoma (156) and accelerated cancer progression (157). This apparent paradoxical anti- and protumorigenic effects of SHP2 (156–160), can reflect parallel pathways, or a more likely explanation would involve existence a minor population of SHP2-positive cells in the microenvironment. Such an explanation is in line with single-cell transcriptomics that showed that Myc-induced tumors only resulted from the rare SHP2-positive cells in SHP2-deficient liver (41).

Hypomorph Mutations in Repressor Parallel-Activating Mutations

Hypomorph mutations that reduce the activity of tumor repressor, parallel-activating mutations in activating oncoproteins. A hypomorph mutation in tumor suppressors acts to hinder its cancer guardian activity. Mutations in PTEN tumor suppressor provide examples (161, 162). An additional example in our context here is provided by suppressor of mothers against decapentaplegic 4 (SMAD4)-suppressed TGFβ signaling, where loss-of-function mutation abolishes the interaction of SMAD4^R361H with SMAD3, recently shown to be restored by Ro-31–8220, a bisindolylmaleimide derivative (163). SMAD4 is a tumor suppressor. Deletion of 18q21 chromosome, which harbors the SMAD4 gene, has been observed in pancreatic and colorectal cancers (164). Biallelic mutations in SMAD2 and SMAD3 are mutually exclusive to SMAD4 mutations, which are common in colorectal cancers (165), emphasizing their critical combined outcomes, possibly resulting in oncogene-induced senescence (OIS).

Signaling By-the-Numbers Scenarios

The emerging conclusion is that the signal should be sufficiently strong to drive tumor growth and cell proliferation, and that activating mutations are decisive components. At the same time, the signal that the mutation emits should not be too strong since homeostatic mechanisms can kick in. This view is supported by other observations: (i) strong activating (or repressing, in a repressor) mutations rarely co-occur in PIK3CA and PTEN. Both result in high PIP₃ levels; (ii) double/multiple strong driver mutations are infrequent in the same oncoprotein, as observed for PI3Kα (16, 123, 124), and broadly, in other oncogenes as well (Ruken Yavuz et al., unpublished data; ref. 36); (iii) the frequency of Ras driver mutations does not correlate with Ras GTPase activity (133), suggesting many mutations have moderate strength; (iv) mutations of intermediate strength are common, unlike hotspot drivers (166); and (v) the mutation load of Ras is only weakly correlated with dually-phosphorylated ERK (ppERK; refs. 167, 168). Finally, (vi) signaling exerts control over the cell cycle, which controls cell growth and replication. Hyperactivation can drive cells to senescence, eliciting phenotypic changes (169). OIS is a powerful intrinsic tumor-suppressive mechanism (169, 170). It arrests cell-cycle progression, with senescence phenotypes depending on the oncogenic stimulus. Forceful oncogenic signaling of both Ras/MAPK and Ras/PI3K/AKT pathways lead to this outcome (Fig. 2). Most studies have focused on the MAPK, but recently PI3K/AKT as well (171). Hyperactivation of the PI3K/AKT pathway triggers senescence through p53 synthesis via mTORC1 (172, 173). In the embryo, senescence is a normal programmed, and instructive developmental mechanism (174). Moderate signaling during embryo development can lead to premature developmental senescence, resulting in neurodevelopmental disorders. Cellular senescence is also the cause of tissue and organismal aging. It is accompanied by chronic, low‐degree inflammation, termed immunosenescence (175). Cell division rates decrease with age, providing a potential explanation for the age-dependent deceleration in cancer incidence (176) with mild activation related to aging (177).

That the decisive measurement determining cell fate is the signaling strength has been shown directly by the remarkable work of Guo and colleagues (40). The team's innovative approach explored the modulation of the amplitude, frequency, and duration of MAP kinase activation. They observed that the alternating electrical current stimulation induced synchronized ERK activation, and that the amplitude and duration of ERK activation were controlled by the varying stimulation strength and duration. Further, the frequencies of ERK activation were modulated by short alternating currents. Thus, alternating current can decide cell fate independent of activating mutations or growth factors, which it mimics.

That signaling strength and duration are the deciding element is further substantiated by large-scale analysis (178). Even though there are a large number of mutations in a tumor across patient populations, their complexity argues that occurrence of single driver oncogene is uncommon. Mutations map into pathways. Multiple mutations in nodes of a pathway, and certain pathway combinations collaborate in cancer, suggesting that signaling should be considered not only in terms of specific activation mutations, but the pathways that harbor them. Among the newly identified interactions of cancer drivers in head and neck squamous cell carcinomas is one between PI3K and the HER3, which mediates protein interactions (179). HER3 lacks or has little intrinsic tyrosine kinase activity (180, 181). Frequent activating mutations in the helical domain of p110α, retain binding to HER3, but remain sensitive to HER3 mAbs, which is not the case with a mutation in the catalytic domain. These may suggest complementary signaling and translation output. The proteins and pathways can be targeted by drug combinations.

Our signaling by-the-numbers paradigm suggests that the number of activated molecules is a more accurate assessment and prediction of the outcome than a measure of the mutation strength, or the number of mutations. That number includes the number of the protein molecules activated by the mutation and the expression level of the respective protein. Strong drivers lead to a higher population of active molecules. The expression level is largely determined by chromatin remodeling, which influences the interaction of the transcription machinery with the regulatory regions of the relevant genes. Notably, apart from the expression level of the mutant protein, the expression levels of other proteins in the pathway should also be sufficiently high for the signal to go through, and of factors regulating the protein output, such as the DUSP6, PTEN (Fig. 2), and SHP2 repressors discussed above and their regulators and transcription factors, for example, NKX2–1 and Myc1. Additional factors controlling expression level include gene duplication, exonic distribution, and proteolysis. That a potent cell survival strategy is to increase the number of the activated proteins by mutations and higher expression levels has recently been observed to be the case in hypermutated replication repair-deficient (RRD) cancers. Hypermutated Ras/MAPK pathway proteins are also highly expressed in RRD tumors (8), resulting in a large population of active proteins. The pathway is sensitive to MEK inhibitors. It is reasonable to expect that the population also harbors rare drug-resistant mutations, which will grow upon decimation of the sensitive cells (60, 182).

Conclusions

Signal strength measured in specific cell types, at certain time windows is the decisive factor. Our view articulates new goals and challenges in fundamental cancer research and in pharmacologic targets. It underscores that in addition to identification of activating driver mutations in a patient's genome, the cell-type specific isoform expression levels of the mutant protein, its regulators, their transcription factors, and their chromatin modulators, are critical in assessing tumor progression and potential for drug resistance. As to the question of whether specific activating mutations couple with the mutant protein's hot spot driver and can serve as a signature, we believe that even though it is possible, it is unlikely. Many of these mutations are likely to preexist, however the cells that host certain background load can be a minor population. Pharmacology decimating the cell hosting the driver mutations is likely to promote the proliferation of those rare resistant cells. These could be uncovered by single-cell transcriptomics, albeit doing so with minor populations, which present scant and noisy data, is challenging.

To date, in the vast majority of models mutations are treated as binary events: the gene either harbors them or does not. Just as the consensus thinking around many oncogenes has moved beyond wildtype versus mutated, here we propose the notion that in different contexts (e.g., lineage, copy number, negative feedback loops, or any other influencer of cell transcriptional state) the same mutation can have varying degrees of effects, which can be measured by the resulting number of activated molecules. In terms of how the number of activated protein molecules could be experimentally determined, apart from the challenging direct measurement it could also be approximated by network analysis of proteomic or transcriptomic data, such as that performed by the VIPER algorithm from Califano's lab, which accurately assesses protein activity from gene expression data (183).

Authors' Disclosures

No author disclosures were reported.