Differential functions of the KRAS splice variants

Abstract

RAS proteins are small GTPases that transduce signals from membrane receptors to signaling pathways that regulate growth and differentiation. Four RAS proteins are encoded by three genes — HRAS, KRAS, NRAS. Among them, KRAS is mutated in human cancer more frequently than any other oncogene. The KRAS pre-mRNA is alternatively spliced to generate two transcripts, KRAS4A and KRAS4B, that encode distinct proto-oncoproteins that differ almost exclusively in their C-terminal hypervariable regions (HVRs) that controls subcellular trafficking and membrane association. The KRAS4A isoform arose 475 million years ago in jawed vertebrates and has persisted in all vertebrates ever since, strongly suggesting non-overlapping functions of the splice variants. Because KRAS4B is expressed at higher levels in most tissues, it has been considered the principal KRAS isoform. However, emerging evidence for KRAS4A expression in tumors and splice variant–specific interactions and functions have sparked interest in this gene product. Among these findings, the KRAS4A-specific regulation of hexokinase I is a stark example. The aim of this mini-review is to provide an overview of the origin and differential functions of the two splice variants of KRAS.

Introduction

The human genome contains three RAS genes (HRAS, KRAS, and NRAS) that encode four RAS proteins (HRAS, KRAS4A, KRAS4B, and NRAS) [1]. RAS proteins are small, membrane-associated GTPases that act as molecular switches that transduce signals from receptors to intracellular signaling cascades. RAS proteins cycle between ‘off’ and ‘on’ states in response to extracellular stimuli by binding to GDP or GTP, respectively [2,3]. When in the GTP-bound ‘on’ state, RAS proteins interact with more than 10 effectors that control a vast array of cellular processes [4,5]. Of these, the RAF kinases that regulate the MAPK pathway are the most ancient, best studied, and central to oncogenesis [3]. Though vital to normal cellular and organismal function, the biology of RAS proteins is most often studied in the context of cancer. Since their discovery in 1982, the RAS genes have been established as frequent drivers of tumorigenesis and are now understood to be the most frequently mutated oncogenes in human cancer [6,7]. Their central role in cancer biology makes RAS one of the most attractive targets for pharmacological inhibition [8–11]. Although RAS GTPases have been extensively studied for 40 years, until recently they were considered undruggable. Now there are FDA approved inhibitors of KRAS with G12C mutations and agents targeting other mutants are in clinical development [12].

Among RAS genes, KRAS is the most frequently mutated in cancer (16% of cancers) [6]. Oncogenic KRAS mutations are most frequently found in pancreatic, lung, and colon adenocarcinomas, as well as in multiple myeloma, which together represent the leading causes of cancer-related mortality in the United States [7]. Given its biological and public health impact, and the as of yet only modest success with direct inhibitors, a better understanding of KRAS biology will be essential to uncover novel therapeutic strategies that can translate into meaningful improvements in clinical outcomes.

KRAS transcripts are alternatively spliced to generate two mRNAs, KRAS4A and KRAS4B, which differ exclusively in the retention or removal of the first of two alternative fourth exons [13,14]. The most common oncogenic mutations (at codons 12, 13 and 61) occur in exons 1 and 2 such that each of the mRNAs encode gain-of-function mutant proteins. Indeed, each KRAS protein is capable of transforming cells and promoting tumors [15]. Despite their shared oncogenic potential, KRAS4B is by far the most studied splice variant (reviewed in [16]). Although KRAS4A was the transforming principle of the Kirsten sarcoma virus from which KRAS derives its name [17,18], it has been relatively ignored in part because of its low expression in commonly studied cell lines and normal adult tissues [19] and in part because of the early and wide distribution of cDNA for KRAS4B. While it is established that both splice variants engage canonical effectors like RAF kinases, little is known about the differential functions of the two oncoproteins. Currently, the differential functions are attributed to the different subcellular localizations of KRAS4A and KRAS4B, which provide privileged access to distinct pools of effectors. Emerging work demonstrates that KRAS4A expression is elevated in cancer [14] and performs functions that do not completely overlap with those of KRAS4B [20–22]. Understanding the unique biological properties of the KRAS splice variants is required to gain a more complete picture of KRAS biology, which will ultimately illuminate new vulnerabilities that can be exploited for drug discovery.

KRAS hypervariable regions confer unique subcellular distributions and dictate isoform effector specificity

The KRAS transcripts encode 21 kDa protein isoforms that, like all RAS proteins, consist of a guanine nucleotide binding domain (G domain) and a KRAS hypervariable region (HVR). The G domain is a 166 amino acid (aa) globular GTPase that catalyzes the hydrolytic conversion of GTP to GDP. The HVR is an unstructured domain of 22–23 aa that becomes post-translationally modified and mediates the affinity of the GTPase for membranes [23,24]. The two KRAS transcripts encode nearly identical G domains that differ only in 3 aa that are encoded by the 5′ sequences of the alternative fourth exons (residues 151, 153, and 165). The HVRs of KRAS4A and KRAS4B direct distinct subcellular trafficking pathways and localizations [14,24]. Thus, the function of alternative splicing is to provide KRAS with alternate membrane targeting sequences. All RAS HVRs contain CaaX motifs that are processed sequentially by farnesylation, aaX proteolysis, and carboxylmethylation by farnesyltransferase, RAS converting enzyme 1, and isoprenylcysteine carboxylmethyltransferase, respectively [23,24]. However, RAS proteins require a second membrane targeting signal within the HVR in order to stably associate with membranes and access the plasma membrane (PM) [24].

The second signal in the KRAS4B HVR is a string of lysine residues referred to as the polybasic region (PBR) that interacts electrostatically with the negative charge of the phosphatidylserine (PS) rich inner leaflet of the PM [25,26] (Figure 1). This interaction can be modulated by the phosphorylation of Ser181 by PKC where, because it is situated within the poly-lysine rich PBR, the negative charge of the phosphorylated residue partially neutralizes the positive charge and thereby weakens the electrostatic relationship between PBR and PM [27,28] (Figure 1). Unique to KRAS4B, phosphorylation at Ser181 leads to a significant redistribution of the protein from the PM to endomembrane compartments, including the endoplasmic reticulum (ER) where it can modulate the interaction of BCL-XL with the IP₃ receptor and thereby limit cell survival [29]. Binding of calmodulin (CALM1) to KRAS4B prevents Ser181 phosphorylation and leads to suppression of noncanonical Wnt signaling and may enhance tumorigenesis in pancreatic cancers [30,31]. Additionally, KRAS4B is a client of the prenyl-binding protein PDE6δ [32]. PDE6δ acts as a cytosolic chaperone for prenylated proteins and shuttles them between membrane compartments by concealing their hydrophobic lipid moiety from the cytosol [32,33]. Although KRAS4A is also prenylated, it is not a binding partner of PDE6δ [14] (Figure 1).

Figure 1. Differential posttranslational processing regulates KRAS isoform subcellular trafficking and interactome.

The trafficking and interactome of KRAS isoforms KRAS4A (blue) and KRAS4B (green) are schematized. For both KRAS4A and KRAS4B, (+) signs indicate polybasic residues that confer electrostatic affinity for negatively charged phospholipid headgroups. KRAS4B is phosphorylated by PKC (dark green) leading to a significant redistribution to endomembrane compartments, including the ER [27]. KRAS4B interacts with v-ATPase 2 on the cytosolic face of lysosomal membranes [35]. Additionally, KRAS4B is a client of the prenyl-binding protein PDE6δ (pink) that shuttles prenylated proteins between membrane compartments [32]. On the other hand, KRAS4A undergoes a cycle of palmitoylation/depalmitoylation on cysteine 180 that drives association with the plasma membrane or outer mitochondrial membrane (OMM), respectively. KRAS4A interacts with HK1 on the OMM in a GTP-dependent fashion and stimulates HK1 activity [21]. The subcellular distribution of KRAS4A also depends on SIRT2 deacylase activity leading to redistribution of KRAS4A onto endomembranes increasing the affinity to ARAF (green) [20]. Farnesylation, palmitoylation and transport of the proteins are depicted in the figure legend. Created with BioRender.com.

KRAS4A relies on two second signals for membrane tethering: (i) its HVR is reversibly palmitoylated at Cys180 and (ii) it contains two short PBRs (PBR1: aa167–170, PBR2: aa182–185) that interact with the PM in a similar fashion to the stronger PBR of KRAS4B [14,24]. Both palmitoylation and PBRs are required to achieve maximal localization of KRAS4A to the PM such that loss of either of these leads to accumulation on other membrane compartments, including the Golgi apparatus, ER, and the outer membrane of the mitochondria, where it can interact with effectors that are restricted to those compartments [14]. The ability of KRAS4A to signal to ERK is greatly diminished if either of its second signals are disrupted, suggesting its access to RAF kinases is diminished when KRAS4A is not on the PM [14]. Disruption of second signals also diminishes the ability of oncogenic KRAS4A to induce leukemia in mice [34]. The Lin group has demonstrated that KRAS4A, but not KRAS4B, can be modified by fatty acylation of lysine 182 of PBR2, a modification that is reversed by Sirtuin 2 (SIRT2), an NAD-dependent lysine deacylase [20]. KRAS4A’s subcellular distribution depends on SIRT2 deacylase activity and loss of PBR2 leads to redistribution onto endomembranes, enhancing its interaction with ARAF and thereby promoting cellular transformation [20] (Figure 1). Thus, the differential biology of the KRAS isoforms lies in their ability to gain access to effectors on distinct membrane compartments.

To identify KRAS isoform-specific interactions, the Lin group performed stable isotope labeling with amino acids in cell culture (SILAC) followed by affinity purification and mass spectrometric identification of proteins [35]. These screens uncovered v-ATPase 2 as a KRAS4B-specific interactor. However, lack of dependance on GTP-binding suggested that v-ATPase2 is not an effector. The interaction was found to occur on the cytosolic face of lysosomal membranes, a compartment accessible to KRAS4B but not KRAS4A [35]. These authors also found that the δ subunit of eukaryotic initiation factor 2B (eIF2Bδ) interacted with KRAS4B but not KRAS4A, also in a nucleotide-independent and HVR-dependent fashion [35]. Differences were also observed with canonical effectors. In line with previous observations [15], the authors showed that RAF1 interacts more efficiently with KRAS4A than with KRAS4B leading to stronger RAF-MEK-ERK pathway activation [35]. Altogether, these results suggest that differential access to classical RAS effectors may also contribute to the differential role of KRAS transcripts variants. Moreover, another isoform-specific interaction involves Stress-activated MAP kinase–interacting protein 1 (Sin1). Among the RAS proteins, Sin1 preferentially interacts with KRAS4A through an atypical RAS-binding domain. However, despite the essential role of Sin1 in the assembly and activity of mTORC2, the interaction with KRAS4A is not required for mTORC2 kinase activity [22]. Another example of compartment-based specificity is the interaction between KRAS4A and hexokinase I (HK1) [21]. When depalmitoylated, KRAS4A redistributes onto endomembrane structures including the outer mitochondrial membrane, where it can directly interact with and stimulate the kinase activity of HK1[21] (Figure 1). Relative to parental cell lines, KRAS4A knockout cell lines demonstrated diminished glycolytic flux in vitro and in xenograft tumors in vivo, as measured by [¹⁸F]-Deoxyglucose postitron emission tomography (PET) [21]. These findings suggest that cancers with high expression of KRAS4A will exhibit a unique metabolic phenotype that could potentially be exploited for novel drug discovery.

Evolution of the RAS genes

The RAS proto-oncoproteins are a highly evolutionarily conserved group of proteins that belongs to a family of 40 closely related proteins that are in turn part of a larger RAS superfamily of small GTPases that also includes ARF, RAB, RAN, and RHO proteins [36–39]. The recent availability of the genomes of basal vertebrates has allowed the study of the evolution of the RAS proto-oncoproteins [1]. Phylogenetic classification of RAS HVRs has shed new light on the origins of RAS isoforms and revealed persistence of KRAS4A isoform since the emergence of jawed vertebrates over 450 million years ago [1] (Figure 2). This work establishes KRAS4B as the vertebrate ortholog of the primordial RAS gene and suggests that the sequential appearance of HRAS, NRAS, and finally KRAS4A resulted from the two whole genome duplication events that characterize vertebrate speciation. Whereas HRAS and NRAS appeared as a direct result of gene duplications, sequence similarity with NRAS suggests that KRAS4A arose from exon shuffling that resulted in insertion of exon 4 of NRAS into the third intron of the KRAS locus [1] (Figure 2).

Figure 2. Scheme of oncoprotein expansion from cephalochordates to mammals.

Oncogene exon–intron structures are depicted with exons as boxes (1–4 coding, 0 and 0’ non-coding) and introns as lines connecting the exons (a–d). Modified from [1]. Created with BioRender.com.

For most of the forty years of their study, RAS proteins were considered largely interchangeable. The persistence of the four RAS isoforms across hundreds of millions of years of evolution strongly suggests nonoverlapping functions. This conclusion is supported by knockout studies in mice [40–42]. Though earlier studies suggested that Kras is necessary and sufficient for development [40–42], recent experiments in genetically engineered mouse models (GEMMs) with pure genetic background demonstrate a requirement of Hras and Nras in neonatal development (see below). With regard to the KRAS splice variants, the Balmain group has showed that both are required for oncogenesis in one mouse model of lung tumors [43]. Overall, these studies suggest unique functions for each RAS isoform, explaining the persistence through vertebrate evolution.

Expression of the KRAS isoforms in homeostasis, development, and cancer

Alternative splicing of the KRAS transcript was discovered in 1983 [44]. Despite knowledge of this for four decades, remarkably little is known about the biological cues or molecular mechanisms that determine whether immature message is spliced into mature KRAS4A or KRAS4B. The majority of studies have focused on measuring the ratios of KRAS4A/KRAS4B mRNA in a variety of cellular contexts both in tissues and during development, but principally, in the context of cancer (reviewed in [16]).

Prior and colleagues have determined the relative quantities of Ras mRNA and protein in mouse tissues and cell lines [45]. RT-qPCR quantification of Ras message in a panel of mouse tissues derived over a time course from embryogenesis to adulthood demonstrate a common pattern of expression where Kras4b >> Nras ≥ Kras4a > Hras, relative to total Ras message, with 4b message representing 60–99% of all Ras transcripts. Interestingly, stomach, intestine, kidney, and heart, demonstrate dynamic expression of Kras4a over the course of development, suggesting a unique role for this isoform in these tissues during embryogenesis [45]. When Prior and colleagues employed a protein standard absolute quantitation (PSAQ) proteomic approach they observed that, unlike their RT-qPCR analysis, the proportion of Ras protein isoforms in adult mouse tissue is Kras4b > Hras > Nras, suggesting that Ras message may not correlate well with Ras protein expression [19]. Other approaches utilizing northern blotting to study Kras in mouse tissue or RT-PCR with [³³P]dATP-based PAGE autoradiography quantification to analyze human tissue demonstrate similar findings: KRAS4B represents the vast majority of KRAS transcripts, while KRAS4A expression represents a small fraction of transcripts and is generally expressed best in tissues of the gastrointestinal tract [46]. However, other investigators performing RT-PCR analysis of adult mouse tissues have found that Kras4a is the predominant Kras isoform in subsets of tissues, such as the cecum, and expression equivalent to Kras4b in tissues such as the seminal vesicles [47]. This suggests that the Kras4a/Kras4b ratio may not be homogenous across tissues and that subsets of cells within tissues may express different proportions of the Kras isoforms. Datasets generated with single cell transcriptomic platforms will allow more accurate determination of the relative expression the Kras isoforms and, by relating these data to the cell states determined with transcriptomics, provide insight into the biological processes and molecular mechanisms that control the Kras4a/Kras4b ratio and conversely the cell states associated with high or low expression of Kras4A.

Studies of GEMMs have allowed for a greater understanding of the requirement of the Ras isoforms for proper mammalian development. Initial studies demonstrated that Kras is necessary and sufficient for development, and that the Nras and Hras loci are dispensable, though Hras^−/−;Nras^−/− double knockout pups are born at ratios below the expected mendelian frequencies [40–42]. Recent data revealed that the altered mendelian ratios of mice lacking both Nras and Hras occured due to improper pulmonary development which results in pulmonary failure in neonatal mice [48], suggesting the Kras locus alone is not sufficient for development in all cases. Furthermore, gene swap experiments from the Balmain lab demonstrated that mice that have an Hras cDNA knocked into the Kras locus (KI) develop properly, although they develop arterial hypertension and dilated cardiomyopathy in adulthood [49]. These data suggest that it is not Kras proteins but rather the Kras locus itself that is important for development, which might be at least in part due to the relatively high protein expression levels generated by the locus. Further contributing to this notion, the Balmain group also developed Kras4a^−/− (Kras4b KI to Kras locus) and Kras4b^−/− (Kras4a KI to Kras locus) animals and demonstrated that, similar to Hras KI, both animals develop properly into adulthood [49,50]. Similar conclusions have been drawn by other groups studying Kras4a^−/− animals: Kras4a is dispensable for development as long as Kras4b is functional. The Barbacid group has developed a novel GEMM that inserts a premature STOP codon into Kras4b’s unique exon, which generates an unstable Kras4b protein while leaving Kras4a transcription and translation untouched [51]. Kras4b null animals do not demonstrate any postnatal development, implying that Kras4b is required [51]. These observations are in line with previous observations that the Kras locus generates the majority of Ras transcripts, and they are overwhelmingly spliced as Kras4b. Unable to generate Kras4b protein, the mice lack sufficient Ras protein, and this gene dosage disparity may explain the lack of proper development.

GEMMs have allowed investigators to determine the requirement of the KRAS isoforms for oncogenesis. Balmain and colleagues employed a urethane-induced lung cancer protocol in their Kras4a^−/− (Kras4b KI to Kras locus) and Kras4b^−/− (Kras4a KI to Kras locus) animal models [50]. Surprisingly, compared with control animals with intact Kras loci, both Kras4a^−/− and Kras4b^−/− animals developed significantly fewer lung tumors in response to urethane challenge. Though both Kras4a^−/− and Kras4b^−/− animals are capable of generating tumors, these data suggest that the isoforms act synergistically to promote oncogenesis in lung tumors [50]. However, the Barbacid group demonstrated that expression of endogenous KRAS4A G12V was sufficient to drive spontaneous lung adenocarcinomas in vivo, although tumors appeared significantly later and with lower incidence than those expressing both oncoproteins [51]. Patek et al. have made similar observations to Balmain whereby wild-type animals develop more tumors than Kras4a^−/− mice in a N-methyl-N-nitrosourea-induced lung cancer model [52]. Notably, the rate of spontaneous colonic tumor formation in Kras4a^−/−;Apc^Min/+was no different from that of Kras4a^+/+;Apc^Min/+mice [53]. However, Luo et al. observed an increased number of colon tumors in Kras4a^−/− mice in response to 1,2-dimethylhydrazine [54]. Collectively, these data imply that the requirement for Kras4a for oncogenesis varies with tissue of origin as well as specific driver and passenger mutations.

KRAS4A also plays a role in stem cell biology where is has been shown to be expressed in embryonic stem cells and in subsets of adult tissues [55]. Recently, Chen et al. have shown that loss of KRAS4A leads to a decrease in the proportion of cancer stem cells in cancer cell lines [50]. These authors also found that hypoxia is capable of up-regulating expression of KRAS4A but not KRAS4B, suggesting that stem cells might depend on isoform-specific expression to adapt to metabolic stress associated with hypoxic conditions [50].

Little is known about the mechanisms that govern splicing of the KRAS pre-mRNA. Recently, Balmain and colleagues reported that splicing of the KRAS transcript is regulated by DCAF15/RBM39 and demonstrated that KRAS4A protein expression is diminished in RBM39^−/− cells or in cells treated with indisulam, a molecular glue that selectively targets RBM39 for degradation by DCAF15 [50].

The differential properties of KRAS isoforms in tumor biology has raised the question of whether cancer cells have altered expression ratios of KRAS4A/KRAS4B (reviewed in [16]). The expression of KRAS isoforms has been shown to vary widely across cancer cell lines and generally correlate with that of their tissue of origin [14]. Although a number of studies have measured KRAS transcripts (see below), it is worth noting that KRAS message may not correlate well with KRAS protein expression, as shown in previous reports [19]. Early studies using RNA hybridization on cancer cell lines demonstrated that expression of KRAS4A was 20-fold lower than KRAS4B [56,57]. Other investigators employing a polymerase colony method found that the cancer cell lines SW480 and PANC-1 had 9.6- and 47-fold, respectively, higher expression of KRAS4B relative to KRAS4A. They also reported that normal colon contained only 1.4-fold more KRAS4B mRNA than KRAS4A, suggesting that tumorigenesis may suppress the expression of KRAS4A [58]. Tsai et al. employed RT-qPCR-based standard curve quantification of KRAS4A and KRAS4B transcripts in a panel of 30 human cells lines utilizing primers that span the isoform-specific splice junctions [14]. Although KRAS4B transcripts were more abundant than KRAS4A in all of the examined cell lines, KRAS4A represented over 20% of RAS transcripts in 40% of the cells lines and was expressed at levels similar KRAS4B in SKMEL192, SKMEL187, Caco2, and HT29 cells. Increased expression in these lines was confirmed via western blot utilizing a KRAS4A specific antibody developed by the authors [14]. Importantly, Tsai et al. also quantified 4A/4B transcript ratios in fresh frozen colorectal adenocarcinoma samples and demonstrated that KRAS4A is expressed at levels equivalent to KRAS4B in the vast majority of these samples, suggesting that expression patterns of KRAS isoforms in vivo is not well recapitulated in commonly utilized cell lines [14]. A similar observation was made by Aran and colleagues who used RT-qPCR analysis of non-small cell lung adenocarcinomas and found KRAS4B expression to be ~2-fold greater than that of KRAS4A in tumor samples. Relative to normal lung controls, over 75% of tumor samples overexpressed KRAS4A, whereas KRAS4B expression remained unchanged [59]. Validation of KRAS4A overexpression in tumors was obtained by immunohistochemistry (IHC) [59], although a lack of controls as well as a lack validated RAS antibodies for IHC make these data difficult to interpret. The relative expression of the KRAS isoforms in human cancers may also be important clinically, as it correlates with survival outcomes in certain tumor types. Yang and Kim analyzed RNA expression and somatic mutation of lung adenocarcinomas datasets in The Cancer Genome Atlas (n = 516) and correlated these parameters to overall survival (OS) and disease-free survival (DFS). KRAS4A transcript abundance was positively correlated with the presence of KRAS mutations as well as with poor OS and DFS [60]. Eilertsen et al. utilized splicing-sensitive microarrays and RNA-seq to investigate KRAS isoform expression in microsatellite stable colorectal cancer patient samples [61]. The authors found diminished KRAS4A expression in tumor relative to normal colonic mucosa independent of KRAS mutational status. Interestingly, in the KRAS wild-type group decreased KRAS4A levels correlated with increased KRAS signaling and poor outcomes, suggesting a prognostic value for KRAS4A expression in this tumor type [61]. In summary, contradictory results have been reported regarding the KRAS4A/KRAS4B mRNA ratios in cancer and their prognostic value. Further studies with standardized methods will be required to settle this question.

Perspectives.

• KRAS-driven cancers represent a leading cause of cancer-related morbidity and mortality worldwide. Despite recent advances in direct inhibitors that target specific mutations, the intractability of other mutations and the universal acquired resistance to targeted therapy makes drugging KRAS an ongoing challenge. Uncovering the various functions of KRAS that underlie its impact as an oncogene will be critical to identifying novel vulnerabilities that can be exploited for drug discovery.

• Despite the knowledge that KRAS is alternatively spliced for over 40 years, the majority of KRAS research has focused on KRAS4B. Evolutionary conservation for over 450 million years, multiple lines of evidence of a distinct protein interactome, and potential prognostic roles across multiple tumor types suggests that KRAS4A plays fundamental roles in both homeostasis and cancer.

• The discovery of isoform specific regulation of hexokinase I by KRAS4A highlights a novel metabolic feature that could be exploited to treat tumors with oncogenic KRAS4A expression. Looking forward, elucidation of the unique biological functions of the KRAS splice variants has the potential to uncover new vulnerabilities of KRAS-driven cancers.

Abbreviations

DFS

disease-free survival

ER

endoplasmic reticulum

GEMMs

genetically engineered mouse models

HK1

hexokinase

HVRs

hypervariable regions

IHC

immunohistochemistry

KI

Kras locus

OMM

outer mitochondrial membrane

OS

overall survival

PBR

polybasic region

PM

plasma membrane

SIRT2

Sirtuin 2