Ras history

Abstract

Although the roots of Ras sprouted from the rich history of retrovirus research, it was the discovery of mutationally activated RAS genes in human cancer in 1982 that stimulated an intensive research effort to understand Ras protein structure, biochemistry and biology. While the ultimate goal has been developing anti-Ras drugs for cancer treatment, discoveries from Ras have laid the foundation for three broad areas of science. First, they focused studies on the origins of cancer to the molecular level, with the subsequent discovery of genes mutated in cancer that now number in the thousands. Second, elucidation of the biochemical mechanisms by which Ras facilitates signal transduction established many of our fundamental concepts of how a normal cell orchestrates responses to extracellular cues. Third, Ras proteins are also founding members of a large superfamily of small GTPases that regulate all key cellular processes and established the versatile role of small GTP-binding proteins in biology. We highlight some of the key findings of the last 28 years.

Introduction

RAS genes were identified initially as viral genes transduced from the rodent genome and responsible for the highly oncogenic properties of RNA tumor viruses.¹ The study of these viral genes, their cellular counterparts, and the 21 kDa (p21) proteins encoded by these genes established key and fundamental properties of cellular Ras proteins. The identification in 1982 of mutationally activated and potently transforming human RAS genes in human cancer cell lines began an intensive focus on studying Ras structure, biochemistry and biology that continues to this day.² Fueled by the high frequency of RAS mutations in a wide spectrum of human cancers, including three of the four most deadly cancers in the U.S. (lung, colon and pancreatic cancer),³ there has been an intensive search for the still elusive “anti-Ras” therapy for cancer treatment.⁴ Dissection of the function of Ras proteins as simple on-off binary switches for diverse extracellular signaling cascades has established many of our current fundamental paradigms of signal transduction. The recent identification of germline RAS mutations in a class of developmental syndromes (RASopathies) expands the contribution of aberrant Ras signaling into other human disorders.⁵ Finally, that Ras proteins represent simply the tip of the iceberg, and are the founding members of a superfamily of Ras-related small GTPases that now number more than 150,⁶ has established that small GTPases are versatile and key regulators of virtually all fundamental cellular processes. In this review, we take a retrospective view of the rich history of Ras research, provide a snapshot of the current state-of-the art, and speculate on what is in store for the future. We provide a chronology (Fig. 1 and Suppl. Table 1) of representative key discoveries regarding the biochemistry and structure of Ras proteins, the mechanisms of Ras signal transduction, the Ras superfamily, and the implication of aberrant Ras activation in cancer and developmental syndromes.

Figure 1.

Timeline of representative key discoveries in Ras research. See Suppl. Table 1 for references.

The Retrovirus Years

The identification of Ras emerged during the extensive study of acutely transforming retroviruses isolated from mice, rats, cats, monkeys, chickens and turkeys. These oncogenic viruses cause rapid formation of sarcomas in infected animals and potently transform cells in culture. Discoveries of the potently oncogenic Harvey murine sarcoma virus in 1964⁷ and the Kirsten murine sarcoma virus in 1967⁸ provided our first glimpses of oncogenic genetic elements that only many years later were established as comprising the human HRAS and KRAS oncogenes, respectively. That the transforming properties of these sarcoma viruses resulted from transduction of normal cellular rat sequences into their own genomes was predicted by Scolnick and colleagues in 1973,¹¹ a time when the tools to prove this were lacking. Initially the genes that we now know as RAS were named as variants of src. Later their ability to cause rat sarcomas became the basis for their current gene names, and their discoverers' names became the basis for distinguishing each of them from the other: Harvey and Kirsten viral ras genes, or H-ras and K-ras, with their protein names cited as Ha-Ras or H-Ras and Ki-Ras or K-Ras. In a seminal discovery published in 1976, Varmus, Bishop, Vogt and colleagues determined that the potent viral src oncogene is a normal chicken gene transduced by the virus into its own genome, thereby converting a normal gene into a potent oncogenic agent.⁹ Viral oncogenes of several other acutely transforming retroviruses were found to encode proteins that only much later were determined to be key components of Ras signaling networks, including the Raf and Akt serine/threonine kinases, the epidermal growth factor receptor (EGFR) tyrosine kinase, the catalytic subunit of class I phosphatidylinositol 3-kinase (PI3-K) (p110α) and the Ets transcription factor.

Scolnick and colleagues in the late 1970s to the early 1980s performed an extensive series of pioneering studies that determined the cellular origin of the viral H-ras and K-ras genes,^10,11 and that these genes encode 21 kDa proteins¹² that bind GDP and GTP¹³ and are associated with the plasma membrane.¹⁴ They also identified the cellular counterparts of the viral genes;^15,16 that Ras proteins expressed in vertebrate cells¹⁷ also bind GDP and GTP and are membrane-associated;¹⁸ that, when overexpressed, these proteins can also transform cells,¹⁹ and that preferential binding to GTP is key for transformation.^13,20 These studies established many of the fundamental biochemical and cellular properties of Ras as membrane-associated GTP-binding proteins and served as an important basis for the many subsequent studies that would expand on their findings (Fig. 2).

Figure 2.

Ras is a GDP/GTP-regulated binary switch. (A) The three RAS genes encode four 188–189 amino acid proteins that share 82–90% overall sequence identity; KRAS encodes two splice variants due to alternative exon 4 utilization, leading to divergent C-terminal sequences. Exons 4A and 4B encode 39 and 38 amino acids, respectively, with 19 identical and 4 conserved substitutions. K-Ras4A is most similar to viral K-ras while K-Ras4B is the predominant isoform expressed in human cells. Residues 1–164 comprise the G domain that contains six conserved sequence motifs shared with other Ras superfamily and GTP-binding proteins. These motifs are involved in binding either phosphate/Mg²⁺ (PM) or the guanine base (G) of GDP and GTP. Residues in Switch I (aa 30–38) and II (aa 60–76) change in conformation during GDP/GTP cycling. The core effector binding domain (E; residues 32–40) and flanking sequences are involved in effector binding specificity. (B) Regulators of the Ras GDP/GTP cycle. RasGEFs stimulate GDP/GTP exchange. With the 10-fold higher cellular concentrations of GTP over GDP, the net result of RasGEF stimulation is formation of active Ras-GTP. Ras-GTP binds preferentially to downstream effectors. RasGAPs accelerate the intrinsic GTP hydrolysis activity of Ras to promote formation of inactive Ras-GDP. Shown are “classic” missense mutants of Ras proteins that have been useful for dissection of Ras function. The Ras(S17N) dominant negative sequesters and blocks RasGEF activity, preventing Ras activation. The G12V and Q61L mutations, found in human cancers, impair GAP-stimulated GTP hydrolysis. The T35S, E37G and Y40C effector domain mutants (EDMs) differentially impair effector binding. The T35S mutant retains efficient binding to Raf but not PI3K or RalGEF, whereas the E37G mutant retains efficient binding to RalGEF but not Raf or PI3K, and the Y40C mutant retains efficient binding to PI3K but not Raf or RalGEF.

The Human Cancer Years

Despite recognition of the highly oncogenic potency of acutely transforming retroviruses, the realization that human cancers are not initiated by such infectious agents dampened enthusiasm for studying them as the basis of human cancer development. Instead, a new direction arose from the finding that biologically active eukaryotic DNA could be transferred into mammalian cells by transfection, following precipitation of the DNA using calcium phosphate.^21,22 A critical key to the success of these experiments in demonstrating whether the transfected DNA had transforming properties was the use of NIH/3T3 mouse fibroblasts as the recipient cells.²³ Although immortalized, NIH/3T3 cells nevertheless retained some normal growth properties (density-dependent growth inhibition, high dependence on serum growth factors), and failed to propagate when deprived of substratum (e.g., did not form colonies in soft agar) or to form tumors when inoculated into immunocompromised mice.²³ However, these cells did display high sensitivity to retrovirus-induced “focus formation”, such that morphologically altered cells that were no longer contact-inhibited grew as easily visualized foci of transformed cells over the background monolayer of untransfected “normal” cells (Fig. 3). The ability of NIH/3T3 cells to become morphologically and growth-transformed by a single viral oncogene provided a sensitive one-hit biological assay for the activated oncogenes that were speculated to be present in DNA obtained from tumor but not from normal cells. NIH/3T3 cells (also referred to as NIH 3T3 or NIH3T3) have therefore been the longtime workhorse cell culture model for these studies, and were instrumental in characterizing RAS and many other oncogenes.

Figure 3.

Detection of activated and mutated HRAS in human EJ/T24 bladder carcinoma cells. The NIH/3T3 focus formation assay was used to detect activated oncogenes present in human tumor but not normal genomic DNA. High molecular weight DNA was isolated from the EJ/T24 human bladder carcinoma cell line, converted to a calcium phosphate precipitant, and added to the growth medium of a monolayer of NIH/3T3 cells. After 14 days, foci of morphologically and growth transformed cells can be detected in cultures treated with DNA from tumor cells but not in parallel cultures treated with DNA from normal human cells. The active HRAS fragment from EJ/T24 bladder cells lies within a 4.6 kDa XhoI-SphI fragment. Human H-Ras protein is encoded by sequences spanning four exons. Exon 1 encodes amino acids 1–37. Sequence comparison of the bladder carcinoma-derived HRAS DNA identified a single base substitution at codon 12, resulting in a single missense mutation (G12V).

In 1979 Weinberg and colleagues showed that DNA isolated from chemically transformed rodent fibroblasts caused morphologic transformation of NIH/3T3 mouse fibroblasts.²⁴ In 1981, using the same approach, Krontiris and Cooper identified transforming activity with DNA isolated from the human EJ bladder carcinoma cell line²⁵ (also referred to as T24 cells²⁶), and Weinberg and colleagues also detected transforming activity in carcinoma and leukemia cell lines.^27,28 Wigler and colleagues then made similar observations using DNA from lung and colon carcinoma cell lines,²⁹ followed by similar discoveries in human sarcoma cell lines.³⁰ The hunt was then on to identify and isolate the human oncogenes that transformed NIH/3T3 cells.

In 1982, this seemingly distinct direction of study unexpectedly collided with the retrovirus oncogene studies. Three groups made the discovery that the transforming genes identified in the NIH/3T3 DNA transfection assays were the same RAS genes identified earlier in Kirsten and Harvey sarcoma viruses.^31–33 By the end of 1982, three groups had established the molecular basis of HRAS gene activation in the EJ/T24 bladder carcinoma cell line,²⁶ unexpectedly and remarkably by a single missense mutation in codon 12, which was also found in the viral H-ras and K-ras genes (Fig. 3).^34–36 Mutation of codon 12 was also established as the mechanism of activation of KRAS from lung and colon tumor cells.³⁷ Additionally, a third transforming human RAS gene, not identified previously in any retroviruses, was discovered in neuroblastoma-derived DNA and was designated NRAS (Fig. 2A).^38,39 The identification of mutant RAS genes in patient tumors but not in normal tissue was an important validation that the RAS mutations identified in tumor cell lines were not simply artifacts of in vitro cell passage.^40,41 This surprisingly simple mechanism of Ras protein activation then focused attention on the issue of what one amino acid substitution would do to the as yet to be determined biochemical and biological functions of Ras proteins.

The discovery of RAS oncogene activation by DNA and protein mutation also stimulated an intense focus on pursuing the origins of cancer at the molecular level. This led to the subsequent identification of oncogene activation, and of tumor suppressor gene inactivation, by alterations in protein structure and function. This bias and emphasis continues to this day, such that genome-wide DNA sequencing of the cancer genome is now a major mandate in the “war against cancer.”

Subsequent studies using the NIH/3T3 transfection assay, and later analyses using DNA sequencing methods, detected activated RAS genes in a wide spectrum of human tumor cell lines as well as in primary patient tumor material.⁴² Most striking was the high frequency of RAS mutations found in colon,^43,44 lung⁴⁵ and pancreatic^46,47 cancers. In addition to mutations at codon 12, RAS mutations were later identified at codons 13 and 61. Although mutations affecting other regions of the proteins have been found at very low frequencies, these three codons are the sites of 97–99% of all RAS mutations in cancer (Fig. 4) and thus comprise the three hot spots of Ras activation. Also strikingly, not all Ras isoforms are equally likely to be mutated. Among the RAS isoforms, missense mutations are found most commonly in KRAS (85%), less commonly in NRAS (12%), and rarely in HRAS (3%) (Fig. 4D). Interestingly, the missense mutation frequency at each position varies widely between isoforms, with G12 mutations the most common in KRAS and HRAS, while Q61 mutations are the most common in NRAS (Fig. 4D).

Figure 4.

Ras mutations in cancer and developmental disorders. Missense mutations in (A) H-Ras, (B) K-Ras and (C) N-Ras in human cancers were compiled from COSMIC (www.sanger.ac.uk/genetics/CGP/cosmic/). Each specific substitution seen at residue 12, 13 or 61 is indicated separately (pink boxes placed above the Ras protein ribbon), followed in parentheses by the number of cancers identified to have that mutation. The numbers and types of missense mutations in each Ras isoform found in developmental syndromes (RASopathies) were compiled from The Ras/MAPK Syndrome Homepage (www.medgen.med.tohoku.ac.jp/RasMapk%20syndromes.html). Specific substitutions are indicated (green boxes) below the Ras protein ribbon and numbers are given in parentheses after each mutation. (D) Distribution of Ras missense mutations in cancer. The distribution of mutations in H-Ras, K-Ras and N-Ras was calculated from data in COSMIC depicted in (A–C). The percentages of missense mutations at 12, 13, 61 and all other positions were determined for H-Ras (629 total mutations), K-Ras (15,594 total) and N-Ras (2,189 total).

Analysis of the current data available at the COSMIC database (www.sanger.ac.uk/genetics/CGP/cosmic/) reveals that HRAS, the first activated RAS gene detected and characterized, is the least frequently mutated in human cancers (3%), whereas mutation of KRAS is the most prevalent (21%), followed by mutations in NRAS (8%) (Fig. 4D). Activated RAS genes were also identified in tumors that arose from mutagenic treatment of rodents with chemical carcinogens.^48,49

The potent ability of mutant RAS genes alone to convert immortalized NIH/3T3 cells to invasive and tumorigenic cells gave an initially simple and misleading perception of the genetic basis of cancer. This notion was revised subsequently by the finding that mutant H-Ras alone was unable to transform primary rodent fibroblasts, but instead, required concurrent activation of an oncogene (e.g., Myc) or inactivation of a tumor suppressor (e.g., p53) for Ras-mediated transformation of these cells.^50,51 This requirement for a cooperating second hit was later extended by cell culture studies of primary human fibroblasts and epithelial cells, in which hTERT-mediated immortalization, inactivation of the p53 and Rb tumor suppressors, and protein phosphatase 2A inactivation were all found to make required contributions to Ras-dependent transformation.^{52 ,53} In 1987, studies in transgenic mouse models supported both the causal role of Ras activation in cancer development as well as the need for cooperating genetic events. Similarly, such studies showed the cooperation of viral H-ras and myc in mammary tumor formation.^54,55

The requirement for complementing genetic events is consistent with the accumulation of genetic mutations in human colon and pancreatic cancer (Figs. 5A and B, respectively), and that cancer incidence increases with age. In pancreatic cancer, KRAS mutation is an early and initiating event, leading some to speculate wrongly that KRAS is not required for tumor maintenance and is not a suitable drug target in this disease. However, the importance of subsequent additional mutations can clearly be seen in mouse models, where endogenous KRAS activation alone induces tumor progression,⁵⁵ but where concurrent inactivation of INK4A/ARF, TP53 or SMAD4 causes greatly accelerated and more advanced tumor development.^56–60

Figure 5.

Colorectal and pancreatic cancer progression. (A) Colorectal cancer progression and gene mutations. Colonic epithelial cells undergo a histologic transition from normal to malignant state that is driven by specific genetic events including inactivation of tumor suppressors (APC, SMAD4 and TP53) and activation of the KRAS oncogene.³²³ The three stages of adenomas represent tumors of increasing size, dysplasia, and villous content. (B) Pancreatic cancer progression and gene mutations. Multiple tumor types arise from the exocrine pancreas, of which greater than 95% are pancreatic ductal adenocarcinoma (PDAC). Normal ductal epithelium progression to infiltrating cancer (left to right) is illustrated through a series of histologically defined precursor lesions (PanINs) that show increasing degrees of disruption of cellular morphology, nuclear atypia and dysplastic growth.³²⁴ High grade PanIN-3 progresses to invasive PDAC. Activating point mutations in the KRAS gene occur early, inactivation of the p16/INK4A gene occurs at an intermediate stage, and inactivation of the TP53, SMAD4/DPC4 and BRCA2 genes occurs relatively late. Oncogenes are indicated in green text and tumor suppressors in red text.

With advances in sequencing technology, it has become feasible to take an unbiased approach to begin to establish the full genetic complexity of the cancer cell genome. With the genomewide sequencing of breast and colon cancers,^61,62 pancreatic cancer⁶³ and glioblastoma,^64,65 a picture has emerged showing that cancers arise through a combination of the frequent mutation of a small common set of genes (mountains) and the infrequent mutation of many other genes (hills) specific to the individual cancers. Although the cancer cell genome harbors many gene mutations, the vast majority of these are passenger mutations that do not serve a causal role, whereas approximately 15–20 driver mutations contribute to cancer progression. A key finding that has arisen from these studies, in particular for colon and pancreatic cancer, is that mutations in KRAS are the biggest oncogene mountains in these two deadly cancers.

One issue that has been puzzling is the preferential mutational activation of specific RAS isoforms in different cancers. For example, there is near-exclusive mutational activation of KRAS in pancreatic, colon and lung cancers. Does this simply reflect distinct carcinogenic assaults that preferentially favor mutation of a certain isoform, or does tumor formation occur only through mutation of certain isoforms? While the evidence is still very limited, one provocative study that addressed this question using mouse models found that activation of KRAS but not NRAS promoted colonic tumorigenesis.⁶⁶ More such studies will be needed to determine if indeed mutation of a specific RAS isoform is critical to initiate oncogenesis in specific tissue types.

The Family Years

Evolutionary conservation of Ras.

The conservation of Ras in invertebrate species amenable to genetic analyses has contributed significantly to delineation of the roles and mechanisms of Ras in vertebrate signal transduction (Fig. 6). Two functionally redundant Ras proteins (Ras1 and Ras2) were identified as required for spore viability in the budding yeast S. cerevisiae.^67,68 Interestingly, while Ras1 and Ras2 do show strong sequence identity with human Ras proteins, they possess additional divergent C-terminal sequences and, at ∼40 kDa, are much larger proteins. In the fission yeast S. pombe, one ∼21 kDa Ras protein (Ras1) has been identified as required for sporulation and mating.^69,70 Two Drosophila RAS genes encoding ∼21 kDa proteins were identified initially (Dras1 and Dras2),⁷¹ and DRas1 was shown later to be the authentic Ras ortholog and a regulator of eye development. The one RAS gene (encoding LET-60) in C. elegans controls vulval development.^72,73 Finally, Ras proteins have also been studied in other model organisms, including the slime mold Dictyostelium⁷⁴ and zebrafish (N-Ras and K-Ras orthologs).^75,76 However, no RAS genes are present in plant genomes.

Figure 6.

Conservation of Ras proteins in evolution. (A) Domain architecture of Ras proteins and the boundaries of the constant G domains were determined using SMART (http://smart.embl-heidelberg.de/). (B) Clustal/W was then used to align the G domain sequences, and the percent of amino acid identity was determined and used to generate the dendrogram.

The Ras superfamily: GTPases are versatile regulators.

In addition to their conservation in evolution, the RAS genes are the founding members and prototypes for a large family of RAS-related genes found in both invertebrates and vertebrates.^77,6 Sequence identity and/or functional relationships subdivide the human family into at least five distinct branches: Ras, Rho, Rab, Arf and Ran (Fig. 7A and Suppl. Table 2). Much of the confusing nomenclature of small GTPases reflects the common practice of naming of each newly identified gene product in a manner that acknowledges its relationship to Ras, the founding member of the family, or to the tissue or cell line in which it was initially identified.

Figure 7.

Human and invertebrate Ras superfamily proteins. (A) The numbers of members of each species and of each Ras superfamily branch were obtained from the references cited in Suppl. Table 2, and then used to generate the graphs. Numbers indicate total members per family. The graph of human family proteins was generated using the numbers in reference ⁶. (B) Human Ras family proteins. Dendrogram generated by Clustal/W and adapted from reference ³²⁵. (C) Human Rho family proteins. Adapted from reference ³²⁶.

In 1983, Ypt1 became the first Ras-related protein to be added to the family, when it was found fortuitously in S. cerevisiae as the gene product of an open reading frame located between the genes encoding tubulin and actin, and was speculated to share nucleotide-binding properties with the “strikingly homologous” p21 Ras proteins.⁷⁸ Similarly, in 1985 came the fortuitous discovery of Ras homologous (Rho) GTPases in the snail Aplysia, in which Rho was identified inadvertently during a search for genes homologous to the alpha subunit of human chorionic gonadotropin.⁷⁹ Using Aplysia Rho as a DNA probe, this group then identified Rho orthologs in S. cerevisiae (Rho1 and Rho2),⁸⁰ Drosophila, rat and humans (RhoA, RhoB and RhoC).^81,82

The fortuitous discovery of evolutionarily-conserved Ras and Ras-related genes prompted deliberate efforts to identify additional Ras-related genes. In particular, Tavitian and colleagues performed DNA oligonucleotide hybridization using probes corresponding to the GTP-binding motif 57-DTAGQEE/D-63. They identified sequences encoding a simian Ras-like protein (RalA) in 1986⁸³ and mammalian orthologs of Ypt1 from a rat brain cDNA library (Rab1–4) in 1987.⁸⁴ Utilizing Drosophila Dras3 as a probe, they identified two Ras proximate proteins (Rap1 and Rap2) in 1988.⁸⁵ In 1987, another group performing low-stringency hybridization to viral H-ras identified both human and mouse Ras-related (R-ras) genes.⁸⁶ Takai and colleagues utilized biochemical approaches to identify six membrane-bound small molecular weight GTP-binding proteins (designated smg),⁸⁷ including K-Ras and RhoB, and independently discovered Rab3A⁸⁸ and Rap1A.⁸⁹

First discovered in 1980 as an ADP ribosylation factor (Arf) in cholera toxin-catalyzed ADP-ribosylation of adenylate cyclase, Arf was determined in 1986 to be a GTP-binding protein^90–92 and four years later to function in Golgi transport.⁹³ Similarly, the related gene SAR1 (secretion-associated and Ras-related) was identified in a genomic library functional screen for suppressors of the ER-Golgi transport mutant of SEC12 in S. cerevisiae⁹⁴ and determined later to be a GTP-binding protein.⁹⁵ Finally, Ran was initially designated TC4 upon its identification in a hybridization screen for RAS-related genes using teratocarcinoma cells.⁹⁶ The later determination that the gene encodes a Ras-related nuclear protein became the basis for the name Ran.^97–99

More recently, with the sequencing of human and other genomes, in silico database searches have identified the complete repertoire of genes encoding Ras-related small GTPases.^77,6 These searches revealed 56 members in C. elegans¹⁰⁰ and 90 in Drosophila¹⁰¹ (Suppl. Table 2). Interestingly, while the genomes of the flowering plant Arabidopsis thaliana and of the rice Oryza sativa were reported to contain Ras-related small GTPases of the Rab, Rho, Arf and Ran families, no Ras GTPases have been found in any plants.^102,101

As the Ras-related gene families have expanded, an obvious speculation in the field has been whether other members are also involved in cancer and other human disorders. This has led to research showing that numerous Ras superfamily proteins, while not mutated in human cancer, nevertheless do serve critical roles in oncogenic growth. In particular, numerous members of the Rho family and their regulators have been implicated in cancer as well as in other human diseases. This role was foreshadowed by the discovery of activated and transforming RhoGEFs (e.g., Dbl, Vav, Ect2) in the same NIH/3T3 focus formation assays used to identify mutant Ras genes.¹⁰³

Several members of the Ras branch have been implicated in cancer (Fig. 7B). In particular, Ras homologue enriched in brain (Rheb) was identified by differential cloning techniques applied to identify genes whose transcription is rapidly induced in rat brain neurons upon synaptic activity.¹⁰⁴ Rheb was determined to be an activator of mTOR and to be activated downstream of the PI3K signaling pathway.¹⁰⁵ Rheb is also activated by loss of function of the tuberous sclerosis complex proteins (Tsc1/Tsc2), which function as a Rheb-specific GAP. Tuberous sclerosis is a rare genetic disorder associated with tumor formation in many different organs, primarily in the skin, brain, eyes, heart, kidney and lungs.

Many of the findings of Ras structure and biochemistry have provided invaluable clues to dissecting the function and regulation of these Ras superfamily members. In turn, the other branches of the Ras superfamily (Fig. 7A) have provided key foundations for our understanding of how many normal cellular processes are regulated. A survey of several invertebrate genomes finds that the Rab family is the largest branch in all species characterized (Fig. 7A and Suppl. Table 2). All species also have significantly large Arf and Rho families. In contrast, as mentioned above, while C. elegans, Drosophila and S. cerevisiae all possess Ras families, the two plant genomes lack any Ras family proteins. At 64 human members, Rab family proteins comprise the largest branch. They control vesicular transport to regulate membrane and protein traffic in the secretory and endocytic pathways.¹⁰⁶ Members of the Arf family (Arf, Arl and Sar), comprised of 29 human members,¹⁰⁷ are also involved in regulation of vesicular trafficking and of endocytosis and exocytosis. The Ran branch, unique in being composed of only a single human member, is involved in regulation of nucleocytoplasmic transport in interphase cells and in organization of the spindle apparatus during mitosis. With some variations and exceptions, Ras superfamily proteins function as GDP/GTP-regulated binary switches. Regardless of regulatory mechanism, they all transduce information through signaling cascades.

The Signal Transduction Years

1993 was a milestone year and “signaling achievement” for the delineation of the “classic” Ras signaling pathway, in which the EGF receptor tyrosine kinase was shown to activate Ras, with activated Ras in turn then stimulating the ERK mitogen-activated protein kinase (MAPK) cascade. While it is now well-appreciated that this single linear pathway represents a gross oversimplification of the complex Ras signaling network, these early findings provided key foundations for understanding the biochemical regulation of small GTPases and heterotrimeric G protein alpha subunits, as well as the mechanistic basis for maintenance of cellular signaling integrity. Here we highlight the key findings that established the mode of regulation of the Ras GDP/GTP cycle, how Ras is activated by extracellular stimuli and how activated Ras stimulates cytoplasmic signaling to relay the signal to the nucleus to cause changes in gene expression.

Regulation of the Ras GDP/GTP cycle: GAPs and GEFs.

The ability of viral Ras proteins to bind both GDP and GTP provided the first clues to the biochemical function of cellular Ras proteins.^13,108 This finding prompted speculation that, like G alpha subunits, normal Ras proteins may be GDP/GTP-regulated binary switches involved in signal transduction.¹⁰⁹ The later finding that mutant Ras proteins are impaired ∼10-fold in their intrinsic GTP hydrolysis (GTPase) activity^{110,20,111,112} provided the first clue that their aberrant function in cancer cells involves preferential binding to GTP. However, impairment of intrinsic GTPase activity alone was not sufficient to explain Ras transforming activity in vivo.¹¹³ A major milestone in understanding Ras function was the discovery by Trahey and McCormick in 1987 that a cytosolic GTPase activating protein (GAP) activity was responsible for a 300-fold acceleration of the hydrolysis of GTP bound to normal Ras but not to tumor-associated mutant Ras proteins.¹¹⁴ The first RasGAP protein, p120 RasGAP, was identified and characterized the next year.^115–117 In 1990, pursuit of the tumor suppressor protein(s) involved in neurofibromatosis type I (NF1) unexpectedly revealed the existence of a second RasGAP, neurofibromin (Nf1).^118–120 Subsequently, additional RasGAPs have been identified.¹²¹ These observations, together with the findings that the GTP-bound form of Ras is the activated state, established that the key biochemical defect of mutant Ras proteins is GAP insensitivity, resulting in impaired GAP-stimulated GTP hydrolysis that in turn leads to persistent accumulation of the active, GTP-bound protein.

Determination of the crystal structure of H-Ras also contributed significant insight into how Ras functions as a GDP/GTP-regulated switch. The first structure of the “G domain” (lacking the C-terminal hypervariable sequence, which is disordered and interferes with crystallization) was reported in 1988,¹²² although a second report by Wittinghofer and colleagues was more accurate¹²³ and the original structure was later revised accordingly.¹²⁴ Structural studies identified two key regions of conformational differences between GDP- and GTP-bound H-Ras.^125,126 Referred to as switch I and II, these regions coincide with sequences critical for Ras interaction with its regulators and effectors. Subsequent structural determinations of Ras in complex with GAP, GEF and Ras-interacting domains of effectors revealed the critical role of switch I and II in these interactions. The structure of H-Ras in complex with the RasGAP catalytic domain of p120RasGAP then identified a molecular basis for the biochemical consequences of the activating substitutions at G12 and Q61: mimicry of the GTP-bound state and GAP insensitivity.¹²⁷

Concurrent with the discovery of GAPs, the race was also on to determine how normal Ras proteins are activated. The search for regulators of the other side of the GDP/GTP cycle began with clues from studies in yeast and flies. In 1987, genetic studies in S. cerevisiae were instrumental in identifying CDC25 as an upstream activator of Ras proteins,^128,129 although it was not yet known that the biochemical basis for this activation was the ability of CDC25 to act as a guanine nucleotide exchange factor (GEF) that enhanced removal of GDP from Ras, thereby preferentially allowing the more abundant GTP to bind in its place. Ironically, the first evidence that CDC25 may function as a RasGEF came from biochemical analysis of a related gene product, SDC25,¹³⁰ yet SDC25 is a dispensable Ras regulator in S. cerevisiae.¹³¹ Subsequently, CDC25 itself was shown to function biochemically as a RasGEF¹³² (Fig. 2B and Fig. 8). Similarly, genetic studies in Drosophila identified Son of Sevenless (SOS) function downstream of the Sevenless and EGFR tyrosine kinases¹³³ and later analyses determined that SOS shares homology with yeast CDC25.¹³⁴ In 1992, utilizing the information from the yeast and fly CDC25 genes, several groups identified mammalian CDC25-homologous RasGRF and Sos proteins.^135–138 In 1996, structural elucidation of the Sos1-Ras interaction showed that Sos1 disrupts nucleotide binding by displacing Switch 1 and distorting Switch 2,¹³⁹ providing a mechanistic basis for GEF-enhanced GDP release and rebinding of GTP. Members of a third class of RasGEFs, Ras guanine nucleotide releasing proteins (RasGRPs), were identified initially in functional screens for activated oncogenes^140–142 (Fig. 8).

Figure 8.

Ras interactome. Proteins that regulate Ras GDP/GTP cycling, catalyze posttranslational modification, or serve as immediate downstream effectors are indicated. Compiled in part from Table 1 in reference ¹²¹.

Ras activation by diverse extracellular stimuli: mechanisms of signaling convergence.

Another important step in dissecting the role of Ras signal transduction was the positioning of Ras downstream of the EGFR and other cell surface receptor tyrosine kinases. A first suggestion came from the observation that EGFR stimulation increased Ras-GTP binding.¹⁴³ Further evidence came from seminal studies by Stacey and colleagues using the Y13-259 Ras neutralizing antibody.^144,145 Microinjection of Y13-259 blocked serum stimulation of cell cycle progression in untransformed NIH/3T3 cells and the ability of membrane-associated viral tyrosine kinases (Fms/CSF-1 receptor, Fes and Src) but not cytoplasmic serine/threonine kinases (Raf, Mos), to transform these cells. The resulting model, in which Ras functions downstream of membrane-associated kinases and upstream of cytoplasmic kinases, would take another eight years to be validated experimentally.

Biochemical and genetic studies also positioned Ras downstream of receptor tyrosine kinases. For example, C. elegans Ras (LET-60) could overcome a defect in EGFR (LET-23) function in vulval development,^72,146,73 Drosophila Ras1 was found to function downstream of the Sevenless receptor tyrosine kinase to regulate eye development,¹⁴⁷ and ligand stimulation of EGFR or of the platelet-derived growth factor receptor (PDGFR) caused transient increases in GTP-bound Ras in mammalian cells.^148–150

Led by further genetic studies in C. elegans and Drosophila, the linkage between the EGFR and Ras was completed with the discovery of the Grb2 adaptor protein (Fig. 9A). Grb2, comprised of a central Src homology 2 (SH2) domain that associates with tyrosine phosphorylated peptide sequences, and that is flanked by two SH3 domains that recognize proline-rich sequences, lacks any catalytic function and instead functions solely as an adaptor protein that facilitates the formation of protein complexes. Identified initially as a gene (Sem-5) involved in C. elegans vulval development, a mammalian counterpart was identified independently as a protein that bound to activated EGFR (Grb2).¹⁵¹ These observations prompted a flurry of eight studies showing that Grb2, through its SH3 domains, associates with Sos1, thus completing the connection between EGFR and Ras.^152–159 That Grb2:Sos1 association with the activated EGFR promotes Ras activation in part by relocalization of the normally cytosolic Sos1 to the plasma membrane where Ras is situated was supported by studies showing that membrane-targeted Sos1 could activate Ras.^160,161 This observation established an important concept in signal transduction: that regulation of subcellular location is an important mechanism to regulate the function of signaling proteins. Subsequent studies revealed further variations on the Grb2-Sos-Ras signaling mechanism, whereby Grb2 recognizes other adaptor proteins (e.g., Shc, IRS-1, Gab2), creating links with yet other receptor tyrosine kinases.¹⁶²

Figure 9.

Ras signaling. (A) Conservation of Raf-MEK-ERK cascade downstream of Ras. (B) Ras effector pathways implicated in Ras-mediated oncogenesis. Compiled in part from literature cited in reference ³²⁷.

Other classes of RasGEFs are regulated by distinct signaling mechanisms as a consequence of the divergent sequences flanking their conserved CDC25 homology RasGEF catalytic domains¹⁶³ (Fig. 8). In particular, the RasGRPs possess a C1 domain that binds diacylglycerol (DAG). Therefore, phospholipase C (PLC)-catalyzed production of the membrane-associated DAG lipid recruits RasGRP to the plasma membrane to activate Ras. Differential activation of PLC isoforms downstream of distinct classes of signaling modulators such as receptor tyrosine kinases (PLCγ), G protein-coupled receptors (PLCβ) or Ras/Rho small GTPases (PLCε)¹⁶⁴ thus provides another mechanistic basis for an amazing diversity of extracellular stimuli to converge on Ras activation.

Ras downstream effector signaling: mechanisms of signaling divergence.

Identification of the first downstream effector of Ras also came from studies in S. cerevisiae, in which adenylyl cyclase (Cyr1) was identified as an effector of RAS1 and RAS2.¹⁶⁵ Adenylyl cyclase stimulates the production of cAMP, which then initiates protein phosphorylation cascades and yeast cell growth. Given the ability of mammalian Ras to substitute for S. cerevisiae RAS proteins to activate yeast adenylate cyclase and conversely of activated yeast RAS to transform mouse NIH/3T3 cells,^166–168 it was assumed that effector utilization would also be interchangeable between S. cerevisiae and mammals. Disappointingly, and perhaps surprisingly in light of the conservation of regulatory RasGEFs and RasGAPs identified subsequently, adenylate cyclase was firmly excluded as the effector of mammalian Ras.^169,165 Another eight years would pass before the final link in the canonical Ras pathway was discovered.

In 1993, a remarkable convergence of information from the genetic dissection of Ras signaling in Drosophila¹⁷⁰ and in C. elegans¹⁷¹ with information on the biochemistry of Ras proteins determined in mammalian cells identified the Raf serine/threonine kinase as a protein that bound preferentially to activated Ras-GTP (Fig. 9A), suggesting that it might play a role in selective transmission of signals from the active form of Ras.^{172–175,171} Since Raf had been identified previously as a viral oncogene,^176–178 much was already known about its function. This association, together with other observations of Ras- and Raf-dependent activation of the ERK1 and ERK2 MAPKs and the ability of Raf to activate MEK1 and MEK2,^179–183 defined the Raf-MEK-ERK protein kinase cascade downstream of Ras. Interestingly, whereas Ras activates adenylyl cyclase in budding yeast (S. cerevisiae), Ras binding to Byr2 in fission yeast (S. pombe) activates a protein kinase cascade analogous to that seen in mammals.¹⁸⁴

The striking conservation of the Raf-MEK-ERK cascade downstream of Ras in both Drosophila and C. elegans (Fig. 9A), where activation of this pathway alone could phenocopy the roles of Ras to regulate eye development and vulval cell fate, respectively, suggested that perhaps a full delineation of Ras effector signaling had been achieved. However, even with the discovery of Raf, there were already indications that other Ras effectors would be found.¹⁸⁵ By 1994, the p110 catalytic subunits of class I PI3-K were recognized as the second class of validated Ras effectors¹⁸⁶ (Fig. 9B). That p110α and a key downstream target of PI3K (the AKT serine/threonine kinase) were also identified independently as retrovirus oncogenes and that mutationally activated alleles of the gene that encodes p110. (PIK3CA) have been found in human cancers further support an important role for this effector in Ras-mediated oncogenesis.

A structural feature common to the majority of Ras effectors is the presence of a Ras-binding domain (RBD) or Ras association (RA) domain that promotes association of the effector with GTP-bound Ras. Because most researchers screening for Ras binding partners utilized 3′-extended cDNA libraries, their screens favored the isolation of potential effectors such as RalGEFs that contain a C-terminal RA domain.^187–189 In contrast, the Raf RBD is positioned at the N-terminus of the protein, and hence, was better represented in the random-primed library used by Vojtek and Cooper.¹⁸⁵ Initial studies of the role of RalGEFs (Fig. 9B) and of their substrates, the Ras family RalA and RalB small GTPases (Fig. 7B), suggested only a minor role for RalGEF-Ral cascades in Ras transformation of NIH3T3 cells.^190,191 Only when subsequent studies were done in human cells did a significant role emerge for RalGEF-Ral signaling in oncogenesis.^192,193 The importance of Ral GTPases in pancreatic, prostate, bladder and other human cancers has now been established¹⁹⁴. That RalGDS, a RalGEF, is required for development of skin carcinomas induced by mutant HRAS also supports the importance of this effector pathway in Ras-mediated oncogenesis.¹⁹⁵ Thus, although no mutations in RalGEF-Ral signaling components have been found in human cancer, the RalGEF-Ral pathway has emerged as perhaps the third best validated effector in Ras oncogenicity.¹⁹⁴

Two other classes of Ras effectors have been implicated as positive mediators of Ras oncogenesis (Fig. 9B). First, in silico searches for novel proteins with homology to the Raf-RBD identified Tiam1, a RacGEF (Fig. 7C), as an effector of Ras. Binding to Ras promotes Tiam1 RacGEF activity, and Tiam1 deficiency impairs development of mutant HRAS-driven skin carcinomas.^196,197 Second, a novel RA domain-containing isoform of phospholipase C (PLC210) was identified initially in a yeast two-hybrid screen for LET-60-interacting proteins.¹⁹⁸ Subsequently, three groups searching BLAST databases independently identified a human ortholog of PLC210 (PLCε) and showed that it is an effector for human H-Ras stimulation of DAG and calcium release.^199–201 Using the same chemical carcinogen approach as was done with RalGDS and Tiam1, it was later shown that mice deficient in PLCε were viable but impaired in HRAS-induced skin carcinoma development.²⁰²

Finally, members of the RA domain-containing RASSF family have been implicated as effectors and mediators of Ras-induced apoptosis (Fig. 8 and Fig. 9B), with the strongest evidence for Nore1/RASFF5.²⁰³ Although RASSF1 to RASSF6 harbor a C-terminal RA domain and RASSF7 to RASSF10 are characterized by an N-terminal RA domain, only a subset of these proteins has been validated to serve as Ras effectors and to initiate apoptosis. Promoter methylation commonly extinguishes the expression of RASSF family members in cancer. The first linkage with Ras came from the discovery of Nore1 (now also called RASSF5) in a yeast two-hybrid library screen for H-Ras effectors.²⁰⁴ Association of activated Ras with Nore1, in complex with the proapoptotic protein kinase MST1. was shown to induce apoptosis.²⁰⁵ Subsequently, evaluation of other RASSF proteins for their ability to associate with Ras revealed limited evidence that RASSF1, RASSF2, RASSF4 and RASSF6 can also serve as Ras effectors in apoptosis.^206–209 A full determination of their physiological role as effectors of Ras, whether in apoptosis or oncogenesis, awaits analyses of endogenous RASSF protein function in cell culture and mouse models driven by endogenous mutant Ras.

The first structure of a Ras effector complex was determined using the isolated RBD of Raf-1 in complex with Rap1A, a Ras family protein highly related to Ras (Fig. 7B),²¹⁰ which ironically does not itself activate Raf-1. The Raf-RBD exhibits a 1000-fold preferential affinity for Ras-GTP over Ras-GDP.²¹¹ This property was exploited later for the development of the now-classic pulldown assay to measure Ras activation in cell lysates;²¹² analogous pulldown assays were developed later for other Ras and Rho family small GTPases. Subsequent studies determined the structure of the isolated RA domains of RalGDS,^213,214 PLC epsilon,²¹³ and RASSF1²¹⁵ as well as the RBD of p110γ.²¹⁶ Overall, these studies revealed the similar ubiquitin-fold structure of the different Ras-interacting domains and both similar and distinct features of their interactions with the switch I and II domains of H-Ras.

Posttranslational mechanisms of regulation: location, location, location in space and time.

Nearly all Ras superfamily proteins have very well-conserved and highly related isoforms that differ from each other almost exclusively in their “variable” membrane targeting domains rather than in their “constant” G domains (Fig. 10A). Thus, Ras activity is not controlled simply by its GDP/GTP cycle, a concept that has been clear since the 1984 studies^217,218 showing that the C-terminus of Ras is required for proper membrane association, lipid binding and biological activity.

Figure 10.

Ras posttranslational processing and membrane association. (A) The C-terminal 24–25 residues comprise the membrane targeting sequence which is comprised of the C-terminal CAAX box, required for posttranslational lipid modification, and the hypervariable (HV) domain that includes a second membrane targeting sequence element (palmitoylated cysteine(s) or polybasic stretches). (B) CAAX motif-signaled posttranslational processing. The cysteine residue of the CAAX box (red) is modified posttranslationally by cytosolic farnesyltransferase (FTase)-catalyzed covalent addition of a C15 farnesyl isoprenoid. Processing is completed on the cytosolic leaflet of the endoplasmic reticulum (ER) by Rce1 (Ras and a-factor converting enzyme-1)-catalyzed proteolytic removal of the AAX residues and then by Icmt (isoprenylcysteine carboxyl methyltransferase)-catalyzed carboxylmethylation of the now terminal farnesylated cysteine residue. FTase inhibitor (FTI) treatment blocks farnesylation and all subsequent CAAX modifications, rendering Ras cytosolic and inactive. Similarly, a serine substitution of the cysteine residue of the CAAX motif (“SAAX”) also prevents farnesylation and all CAAX-signaled modifications. The X residue of the CAAX motif dictates prenyltransferase specificity. Substitution of the X residue with leucine generates mutants that undergo modification by GGTase-I and addition of a C20 geranylgeranyl lipid instead of a C15 farnesyl isoprenoid. (C) Ras plasma membrane association is dependent on a second targeting element. The CAAX-signaled modifications alone are not sufficient to promote plasma membrane association. Additional sequence elements in the HV provide a second membrane targeting signal. In H-Ras, N-Ras and K-Ras4A, cysteine residues in the HV region are modified posttranslationally by ER-associated protein acyltransferases (PATs) that promote covalent addition of a C16 palmitate fatty acid. In K-Ras4B, lysine-rich polybasic amino acids comprise the second signal, which facilitate association with the negatively charged head groups of phosphatidylserine and phosphatidylinositol in the cytosolic face of the plasma membrane. Whereas H-Ras and N-Ras (and presumably K-Ras4A) traffic through the classical secretory pathway through the Golgi to the plasma membrane, K-Ras4B bypasses the Golgi and transits to the plasma membrane by a poorly characterized mechanism. Additionally, the differing second signals of K-Ras4B and H-Ras dictate localization to distinct plasma membrane subdomains, that may lead to distinct effector signaling.

The general concept that Ras proteins must be localized correctly in order to become activated and to be biologically active is well accepted, but a complete picture of what “localized correctly” looks like has yet to be filled in. The simplified localization paradigm that has evolved over the past two decades is that Ras proteins are synthesized as soluble precursors which then undergo several steps of post-translational modification (Fig. 10B) that facilitate their required transit to and tenure at the plasma membrane, where they are activated by GEFs and inactivated by GAPs, and interact with effectors to transmit their signals downstream (Fig. 9B). The full story, both for the family as a whole and for individual family members, is more complex and interesting, especially in terms of dynamic spatiotemporal regulation rather than static positioning, and is still emerging as we write.

It was known as early as 1979¹² that viral Ras proteins were cytosolic when newly synthesized, but did not remain so. In 1980, an important study used electron microscopy to illustrate the presence of H-Ras at the inner leaflet of the plasma membrane, and that characterization remained the default for many years.¹⁴ To determine the nature of the alteration to newly synthesized Ras that allowed it to become membrane-bound, labeling studies were done by analogy to Src proteins, which were known to be modified by the saturated fatty acid myristate. Therefore Ras was evaluated for modification by fatty acid acylation, and both myristic acid and its slightly longer cousin, the C16 palmitic acid, were tested. In 1982, it was clear that v-H-Ras could incorporate 3H-palmitic acid²¹⁹ but not 3H-myristic acid, thereby firmly establishing that these two classes of proteins were modified in a distinct manner. Two years later, Willumsen, Lowy and colleagues used palmitic acid labeling as well as deletion and structural mutants to characterize the processing that led to Ras becoming membrane-associated. They observed correctly that the C-terminus of v-H-Ras is necessary for its plasma membrane localization,²¹⁷ and specifically that residue Cys186²¹⁸ is required for Ras to bind lipid as well as to associate with membranes and exert transforming activity. They also correctly speculated that “no more than 3 [C-terminal] residues are cleaved during processing” (Fig. 10B), and validated that v-H-Ras can incorporate 3H-palmitate. The latter finding led them and others to speculate or declare that Cys186 is the site of palmitoylation, as that lipid was the only one known at the time to be relevant to Ras.²²⁰ In 1987, pulse-chase studies showed that the acylation of N-Ras is highly dynamic, with a half-life turnover of about 20 min.²²¹ This result suggested another possible mechanism for control of Ras activity.

However, also in 1987, Tamanoi and colleagues identified a mutant in S. cerevisiae, dpr1, that was unable to modify Ras despite retention of fatty acid acylation.²²² This finding suggested that acylation cannot be the first processing step of Ras, but other modifications were unknown at the time. Over the next couple of years, several studies emerged indicating that Ras is carboxyl-methylated in species from yeast to man,^223–225 and that this modification does not proceed until after C-terminal proteolysis. Indeed, sequencing demonstrated that the 1984 prediction²¹⁸ of removal of three amino acids was correct.²²⁶ The significance of this number would become apparent when the Ras converting enzyme Rce1 was identified. Meanwhile, each of these steps was shown to positively influence Ras membrane binding, in studies using techniques as varied as aqueous-detergent fractionation, differential centrifugation and immunofluorescence with anti-Ras antibodies. However, the puzzle was still missing an important piece.

Only in 1989, when it was recognized after much comparison of sequence alignments that Ras proteins share a C-terminal CAAX motif (C = cysteine, A = aliphatic amino acid, X = terminal amino acid) with yeast a-factor, a protein known to be modified by a farnesyl isoprenoid lipid, was it determined that Ras proteins are also modified by farnesylation^227–229 and that farnesylation is an obligate first step to allow palmitoylation to then occur on additional cysteine(s) upstream of Cys186. Without farnesylation at Cys186 (for example, by replacement with a Cys>Ser “SAAX” mutation), Ras proteins become cytosolic. In contrast, Ras proteins that are farnesylated but do not undergo subsequent processing steps are localized to endomembranes, as has been clarified over many years of experimentation.

Beginning in 1989, Hancock, Marshall and colleagues demonstrated that H-Ras and N-Ras, but not K-Ras, are modified both by farnesylation at Cys186 and by palmitoylation at one (N-Ras) or two (H-Ras) cysteine residues almost immediately upstream of the farnesylated cysteine, and that palmitoylation enhances membrane association and biological activity.²²⁸ The same group then established or contributed to several guiding principles that are relevant for many signaling proteins whose membrane targeting is modulated by lipidation: (1) a “second signal” found in the hypervariable domain is required in addition to farnesylation,²²⁸ (2) the second signal consists of either a polybasic domain (K-Ras4B) or palmitoylation (H-Ras, N-Ras and K-Ras4A),²³⁰ thereby immediately highlighting the distinctive regulation of K-Ras4B, and (3) the minimal plasma membrane targeting signal of Ras is the CAAX or CAAL (CAAL specifies geranylgeranylation rather than farnesylation) plus a hypervariable domain-derived second signal. Together, these two motifs are sufficient to target either Ras or a cytosolic heterologous protein, such as GFP, to membranes.²³¹ Once these principles were established, it was demonstrated that a fusion protein of the serine/threonine kinase and Ras effector, Raf, with the C-terminal ∼20 amino acids of Ras [designated “Raf-CAAX”], was activated simply by virtue of being tethered to the plasma membrane without being brought there specifically by full length Ras.^232,233 Such a result suggested that a key role of Ras is to bring Raf to the plasma membrane for activation.

Exactly where Ras proteins are localized when they are not yet fully processed and at the plasma membrane has been clarified gradually. In 1999 Philips and colleagues showed that H-Ras and N-Ras, which can both be palmitoylated, trafficked through the Golgi on their way to the plasma membrane, whereas K-Ras, a nonpalmitoylated protein, did not.²³⁴ This study and others²³⁵ also confirmed, consistent with the principles laid down nearly a decade before, that the CAAX-containing motif was necessary and sufficient to direct the proteins to endomembranes for carboxymethylation.

Whether Ras could signal from internal membranes was answered more definitively with the development of new tools such as the use of GFP-Raf-RBD for “bystander FRET” to detect Ras-GTP, FRET-based Raichu probes^236,237 and others. A 2002 study showed not only that Ras proteins can signal when associated with the endoplasmic reticulum and Golgi, but that they preferentially use different effectors depending on the endomembrane from which they are signaling.²³⁸ The next year, monitoring localization of the Raf-RBD probe for activated Ras revealed highly selective activation of Ras in response to growth factor signaling via PLCgamma and the RasGEF RasGRP1, rapidly at the plasma membrane and then later at the Golgi.²³⁹ Biochemical studies could also be performed using Ras proteins physically tethered to sites of interest to show site-specific activation and deactivation, depending on the available regulatory proteins at those sites.²⁴⁰ Regardless of the specific details, Ras activation and deactivation has been shown repeatedly to be tightly regulated, both spatially and temporally, and this tight regulation is due to modulating the localization of both Ras and its interacting partners.

The question of how Ras proteins “get around” was also beginning to be addressed. In 2005, time-lapse microscopy and photobleaching experiments showed that palmitoylation traps Ras proteins on membranes where they can undergo vesicular transport, and that depalmitoylation allows their release for recycling from the cytosol back to endomembranes.²⁴¹ A similar study also emphasized more clearly the importance of deacylation for cytosolic transit.²⁴² Collectively, these results supported the importance of the dynamic nature of palmitoylation, as shown nearly 10 years earlier,²²¹ to the turnover and transit of H-Ras and N-Ras between the cytosol, endomembranes including especially the Golgi complex, and the plasma membrane.

Other posttranslational modifications have been shown to influence Ras subcellular localization and differential engagement of effectors. For example, Bar-Sagi and colleagues showed that H-Ras modification by mono- and di-ubiquitination of H-Ras, but not of K-Ras, stabilizes its association with endosomes and modulates its ability to activate ERK MAPK signaling.²⁴³

C-terminal phosphorylation has also emerged as an important additional regulator of Ras localization and function. Although the viral Ras proteins were identified as threonine phosphoproteins,¹² surprisingly, the cellular proteins were poorly serine phosphorylated.¹⁸ This biochemical difference was found later to be due to a missense mutation in residue 59 in the viral Ras proteins, producing an autophosphorylation mechanism whereby the substituted threonine served as a phosphate acceptor from the bound GTP.²⁴⁴ However, in 1987 Rosen and colleagues²⁴⁵ showed that K-Ras4B became phosphorylated upon PKC activation by phorbol ester. The site was not identified, and the modification was apparently not investigated further until 2006, when Philips, Cox and colleagues showed that PKC. phosphorylated K-Ras4B on Ser181 in the hypervariable membrane targeting domain, relocalizing it from the plasma membrane to internal membranes including Golgi, ER and mitochondria, and converting it at the latter site to an apoptosis-inducing protein.²⁴⁶ Related mechanisms of Ras family GTPase regulation of subcellular location and function upon phosphorylation of the C-terminal membrane targeting domain have also been described for the Rnd3 and RalA small GTPases,^247–251 as well as for Rap1²⁵² and RhoA.^253,254

Originally, because of the strong sequence identity and shared effector utilization, there was a general acceptance that the different Ras isoforms were essentially identical in function. Thus, while KRAS and NRAS mutations are seen more frequently in human cancers, the study of H-Ras has dominated much of the early history of Ras research. This bias towards H-Ras was encouraged by the ready availability of H-Ras expression vectors and mutant proteins, recombinant protein and antibodies. However, in recent years, a shift has begun towards a focus on K-Ras. The essential requirement for mouse development of KRAS^255,256 but not NRAS²⁵⁷ or HRAS,²⁵⁸ or both,²⁵⁹ or KRAS4A,²⁶⁰ has certainly helped to fuel this shift. While HRAS was able to substitute for KRAS to support mouse development, the occurrence of cardiovascular defects in adult mice nevertheless supported functional differences.²⁶¹

There has been a steady accumulation of evidence that the Ras isoforms are functionally distinct. First, as discussed above, trafficking is distinct for the different isoforms, exposing them to different pools of regulators and effectors.^262,263 Second, palmitoylation also has been shown to contribute greatly to the differential trafficking of H-Ras and N-Ras compared to K-Ras by directing them into specific nanoclusters, or “rasosomes.”^264–267

In addition, each Ras isoform is chaperoned by its own preferred binding partner. In 2001, the proposal was made that galectins could act as chaperones for Ras by virtue of their extensive structural similarity with RhoGDI.²⁶⁸ Galectin-1 has a farnesyl-binding pocket that is strikingly similar to the geranylgeranyl-binding pocket of the RhoGDI/Cdc42 interaction,²⁶⁸ and is a preferred binding partner for and enhances the activities of H-Ras^269,267 whereas Galectin-3 is a preferred binding partner for K-Ras.^270,271 Collectively these studies have shown that the interactions between Ras proteins and galectins, which themselves modulate transformation and are often overexpressed in cancers, are independent of galectin carbohydrate-binding interactions, and that galectins modulate both the spatial distribution of Ras proteins in nanoclusters and their effector utilization. Most recently, mathematical modeling and FRET/FLIM microscopy have been used to identify a role for the “constant” or G domain of Ras proteins for orientation with respect to the plasma membrane, suggesting synergy between this type of localization recognition and the specificity conferred by the hypervariable domain.²⁶²

Ras-Targeted Therapies: The Search Continues

Early efforts to target Ras for cancer treatment focused on the fundamental defect of mutated Ras, which is the refractory response to downregulation by GAPs and consequently the persistent binding of GTP (Fig. 2B). However, efforts to develop a small molecule “GAP” that is active on mutant Ras met with failure. Additionally, efforts to develop GTP-competitive antagonists of Ras proteins analogous to the ATP-competitive inhibitors of protein kinases have been precluded by the picomolar binding affinity of Ras for GTP. While exploration of possible approaches to direct targeting of mutant Ras continue, most current efforts to develop small molecule anti-Ras strategies are focused on indirect approaches.

Farnesyltransferase inhibitors—“What a long strange trip it's been.”

One of the most promising early directions for the development of anti-Ras inhibitors centered on ways to prevent Ras posttranslational lipid modification and plasma membrane association (Fig. 10B). As mentioned above, the discovery in 1989 that Ras proteins are farnesylated prompted a huge effort to target this modification.²⁷² Because the farnesyl pyrophosphate that contributes this lipid group to proteins is an obligate intermediate component of the mevalonate-cholesterol biosynthetic pathway, whose synthesis can be blocked by cholesterol-lowering drugs already in clinical use (e.g., lovastatin), the determination that Ras is farnesylated stimulated considerable excitement that an approach for pharmacologic inhibition of Ras function might soon be at hand.

The HMG-CoA reductase inhibitor lovastatin was the first statin approved by the FDA in 1987 for lowering cholesterol to prevent cardiovascular disease in patients with hypercholesterolemia. However, because the clinically effective concentration for lowering cholesterol biosynthesis is much lower than that needed to block Ras farnesylation,²⁷³ the use of statins as specific Ras inhibitors would likely be ineffective. This in turn prompted the search for the enzyme required for the addition of the farnesyl group to Ras, which culminated in the isolation, by Goldstein, Brown and colleagues, of farnesyltransferase (FTase) in 1990.²⁷⁴ The finding that the Ras CAAX tetrapeptide sequence alone was effective in blocking FTase activity initiated intensive efforts by the pharmaceutical industry to develop cell-permeable CAAX peptidomimetics as possible anti-Ras inhibitors.²⁷² The development of CAAX mimetics, as well as the use of high throughput screens to identify small molecule inhibitors of FTase from libraries of previously synthesized as well as natural compounds resulted in numerous potent and selective cell-penetrant FTase inhibitors (FTIs).

The early dramatic success of FTIs in cell-based^275,276 and mouse model studies, when evaluated in H-Ras-driven models, fueled considerable excitement that an anti-Ras inhibitor would soon reach the clinic. One of the most dramatic observations was the ability of FTI treatment to cause regression of H-Ras-driven mouse mammary tumors.²⁷⁷ However, this excitement was soon tempered by a key study showing that tumor cell line sensitivity to FTI growth inhibition in vitro did not correlate with Ras mutation status,²⁷⁸ and dashed by the unexpected findings that K-Ras4B and N-Ras proteins not only are resistant to FTI treatment,^279–281 but in the presence of FTI also undergo alternative prenylation (modification by a related geranylgeranyl isoprenoid lipid), whose addition is catalyzed by a related, FTI-insensitive enzyme, geranylgeranyltransferase-I (GGTase-I), under conditions of FTI treatment^282,283 (Fig. 10B). That the two Ras proteins most commonly mutated in human cancers could escape the inhibitory activities of FTIs provided an explanation for the lack of clinical efficacy of FTIs in clinical trials. This was seen perhaps most dramatically and disappointingly in trials for pancreatic cancer, where essentially all cancers are KRAS mutant.²⁸⁴

The disappointing demise of FTIs as anti-Ras inhibitors also stimulated an appreciation that the three Ras proteins are not identical in function. As mentioned, the early bias in studying H-Ras was due largely to the ready availability of antibodies, recombinant protein and molecular constructs for H-Ras, and to the general perception that the Ras proteins were essentially identical in function as oncoproteins. With additional observations, in particular the ability of activated KRAS but not NRAS to drive colon tumorigenesis, recent studies have shifted to the study of KRAS.

One very unexpected outcome of the development of FTIs is the fortuitous development of a drug that may be useful for the treatment of Hutchinson-Gilford Progeria Syndrome (also called Progeria).²⁸⁵ Progeria is a very rare, fatal genetic condition characterized by an appearance of accelerated aging in children. Children with Progeria die of atherosclerosis at an average age of 13 years. With fewer than 50 patients worldwide, commercial development of drug specifically for this premature aging disorder would never occur otherwise. Progeria is caused by the incomplete posttranslational processing of lamin A, another FTase substrate. However, after farnesylation, the farnesylated C-terminus is removed by proteolytic cleavage to generate the normal final protein. In progeria, this normal modification does not occur, leaving lamin A farnesylated. Lamin A and other nuclear lamins are the structural proteins of the nuclear lamina, an intermediate filament network that provides scaffolding for the cell nucleus. Therefore, FTI treatment may be one approach to prevent accumulation of this mutant lamin A protein, and three clinical trials to evaluate this possibility have been initiated with the FTI lonafarnib that has already been evaluated extensively in clinical trials with cancer patients (www.progeriaresearch.org). While the high expectations that FTIs would be effective anti-Ras inhibitors has not come to pass, if FTIs do find clinical usefulness for this disease, or perhaps targeted to FTase enzymes important in parasitic diseases such as malaria or sleeping sickness, it would provide a positive outcome to what has been a very disappointing era of anti-Ras drug discovery.

Finally, the genetic validation in cell culture and mouse models of their importance to Ras-mediated growth transformation has stimulated interest in targeting both Rce1 and Icmt, the two CAAX-signaled processing steps after farnesylation (Fig. 10B).^286–289 The results of these studies may also reflect the fact that Ras transformation can be dependent on multiple other CAAX-terminating small GTPases (e.g., Ral, RhoA, Rac1) whose functions also depend on these processing steps (Fig. 7B and C). Efforts to develop inhibitors of these enzymes are ongoing. ²⁹⁰

Competitors of Ras-chaperone interactions.

An entirely different approach to blocking modifications has been taken by Kloog and colleagues, who have developed the small molecule farnesylthiosalicyclic acid (FTS) into a clinical candidate, salirasib, that has gone to Phase I/II trials in hematopoietic disease as well as cancers of the pancreas and lung. The basis for their enthusiasm for salirasib includes the proposed mechanism of action, which is competition for Ras membrane-binding sites and/or galectin binding,²⁹¹ as well as the apparent efficacy without toxicity in models of a wide range of pathological conditions including those dependent on oncogenic K-Ras.²⁹⁰

Inhibitors of Ras effector signaling.

Currently, the greatest ongoing efforts in anti-Ras drug discovery have focused on blocking Ras effector signaling (Fig. 9B). Driven by considerable experimental validation for their critical role in Ras-driven oncogenesis, and by the finding of mutated BRAF and PIK3CA in human cancers, the Raf-MEK-ERK and PI3K-AKT-mTOR signaling pathways have been the most intensely targeted. A survey of www.clinicaltrials.gov finds at least 38 inhibitors of these pathways in phase I/II clinical evaluation, which have been reviewed recently.⁴

Driven by the conservation of the Raf-MEK-ERK cascade in evolution, by the finding that BRAF and RAS mutations occur in nonoverlapping frequency in the same cancers (e.g., melanomas, colon cancer),²⁹² and finally, by the success of protein kinase inhibitors for cancer treatment, inhibitors of MEK and Raf have been the most intensively pursued. CI-1040 (PD 184352) was the first small-molecule MEK1/2 inhibitor that proceeded to clinical testing. Although it was well-tolerated, CI-1040 exerted insufficient anti-tumor activity in phase II evaluation and was discontinued.²⁹³ Subsequently, other MEK inhibitors with improved potency and pharmacologic properties have entered clinical evaluation. Although developed originally as a Raf inhibitor,²⁹⁴ sorafenib was subsequently found to have broad activity,²⁹⁵ in particular, for tyrosine kinase receptors involved in tumor angiogenesis such as those for VEGF and PDGF. FDA approval of sorafenib in 2005 for advanced renal cell cancer, where neither RAS nor RAF mutations are seen, is most likely based on its anti-angiogenic rather than anti-Raf activity. More recently, a mutant BRAF-selective inhibitor (PLX4032) has shown promising activity in melanomas expressing mutant BRAF.²⁹⁶ However, three recent studies reported that this and other Raf inhibitors displayed the paradoxical ability to activate, rather than block, ERK activation in RAS mutant tumor cells,^297–299 a finding foreshadowed by two 1999 studies reporting that multiple Raf kinase inhibitors unexpectedly activated Raf, possibly due to depleting a negative feedback mechanism.^300,301 In another approach, small molecules have been identified that inhibit the Ras-Raf interaction,^{302, 303} although it is unclear whether such molecules can be developed into effective drugs.

Synthetic lethal partners of mutant KRAS—New targets for anti-Ras inhibitor development?

Recently, large-scale interfering RNA screencs have been applied to take a functional and unbiased approach to identify therapeutic targets for anti-Ras inhibition.^304–306 These screens are based on the concept of synthetic lethality, in which two genes are defined as synthetically lethal if mutation of either gene alone is compatible with viability but the simultaneous mutation of both genes leads to death.³⁰⁷ Mutationally-activated RAS genes thus represent one gene and RNA interference-mediated ablation in cancer cells of the expression of a second gene provides the second hit. Since normal cells lack mutant RAS, genes identified in this manner should in principle be selectively lethal for tumors but not normal cells.

One study of a limited RNAi library targeting 1011 genes with a focus on protein kinases identified the STK33 serine/threonine kinase as a synthetic lethal partner of mutated KRAS.³⁰⁵ A diverse spectrum of KRAS mutant tumor cell lines showed STK33 dependency. The identification of STK33 illustrates an important strength of this type of functional screen, since no alteration in its expression, no mutations, and no transforming activity of STK33 was detected. Hence, STK33 would not have been identified by the more classical criteria for cancer-causing genes. In a second study, a broader genome-wide screen was done targeting 32,293 unique human transcripts.³⁰⁴ The genes identified encode functionally diverse proteins that regulate several biological processes, especially mitosis. For example, one such gene encodes Polo-like kinase 1 (Plk1), a serine-threonine kinase with a key role in mitosis. Inhibitors of Plk1 have been in development for cancer treatment, and RAS mutant cells showed increased sensitivity to a Plk1-selective inhibitor. A third limited screen identified TANK-binding kinase 1 (TBK1) as synthetic lethal partner of mutant KRAS.³⁰⁵ The TBK1 serine/threonine kinase can activate the NF-kappaB transcription factor and support cell survival. This role is likely to be associated with TBK1 function downstream of RalB, a critical effector of Ras-mediated survival signaling in human cells.³⁰⁸ Finally, in a different approach, a fourth study identified a gene signature for KRAS dependency, which included genes encoding the Syk and Ron tyrosine kinases.³⁰⁹ To validate this screen, they showed that KRAS mutant tumor cell lines were more sensitive to induction of apoptosis by treatment with a small molecule inhibitor of Syk.

These studies identified several protein kinases as synthetic lethal partners of mutant KRAS, and as such suggest highly tractable targets that may accelerate the development of these new leads for “anti-Ras” drugs. Overall, the diversity of genes identified in these screens will surely provide many additional intriguing candidates for anti-Ras therapeutic development.

Ras and Developmental Syndromes

An unexpected recent discovery has been that germline mutations both in RAS and in components upstream and downstream of Ras signaling are associated with a class of developmental syndromes now referred to as RASopathies (Fig. 11 and Suppl. Table 3).³¹⁰ Germline HRAS mutations were first identified in Costello syndrome,³¹¹ followed by KRAS mutations in cardio-facio-cutaneous (CFC)³¹² and Noonan syndromes.^313,314 These syndromes each exhibit unique features, but additionally, because of a common genetic basis in Ras signaling, also share overlapping characteristics. These include craniofacial dysmorphology, cardiac malformations, cutaneous, musculokeletal and ocular abnormalities. Finally, whereas persons with Costello syndrome are at increased risk for malignant tumors,³¹⁵ persons with Noonan or CFC syndrome have no or only a slightly increased risk of cancer, as discussed below.

Figure 11.

Ras signaling components mutated in RASopathies. Generated based on information summarized in Suppl. Table 3 and from references cited in reference ⁵.

In addition to mutational activation of Ras, loss-of-function mutations in the p120 or neurofibromin RasGAPs or gain-of-function mutations in Sos1 also cause Ras hyperactivation in RASopathies.⁵ Mutations in the Shp2 protein tyrosine phosphatase can also lead to downstream activation of Ras. The occurrence of mutations in RAS and RAF1, BRAF, MEK1 and MEK2 suggests that hyperactivation of the ERK MAPK cascade is the key effector mechanism by which Ras activation promotes these disorders. Thus, inhibitors of Raf, MEK and other inhibitors of this pathway developed for cancer treatment may fortuitously be useful for the treatment of these disorders.³¹⁶

An interesting feature of these diseases that distinguishes them from Ras- and Ras pathway-driven cancers is that the mutations seen in KRAS and BRAF in RASopathies are largely atypical of those seen in human cancers, and that mutations in Ras signaling components seen in RASopathies are not typically found in cancers at all. The KRAS mutations occur not at the three hot spots of codons 12, 13 or 61, but instead tend to be those that result in less potently activated Ras proteins (Fig. 4B). The V600E mutation that represents ∼80% of the BRAF mutations seen in cancer is not seen in the ∼75% of individuals with CFC syndrome who have mutant BRAF. Mutational activation of Sos1,³¹⁰ or of MEK1 or MEK2 is also uncommon.^317,61–63 One possible explanation is that germline mutations that potently activate K-Ras signaling, which has been proposed to preferentially activate the Raf kinase cascade, compared to H-Ras activation of the PI3K cascade, are deleterious for human development and therefore individuals with those mutations would not survive to birth. While some analyses in mouse models have supported this premise,³¹⁸ others found that potently activated RAS alleles are well-tolerated.³¹⁹ However, the situation may be different for the frequent (∼85%) germline HRAS mutations in patients with Costello syndrome, where ∼90% of the mutations (G12S and G12A) also occur in human tumors (Fig. 4A). Mice with a germline HRAS(G12V) mutation exhibited abnormalities observed in Costello syndrome patients and displayed mammary hyperplasia but developed tumors rarely,³²⁰ despite expressing a strongly activating and tumor-associated mutant form of H-Ras. One possible explanation for this surprising result comes from a study in which zebrafish with germline expression of HRAS(G12V) displayed Costello-like symptoms and oncogene-induced senescence (OIS), whereas overexpression of the same protein driven in larvae off a heat-shock promoter resulted in hyperproliferation that required p53 to drive OIS.³²¹ Perhaps HRAS-driven tumorigenicity in Costello patients requires loss of p53. A few NRAS mutations have recently been described to occur in Noonan syndrome³²² and are also distinct from the hotspot mutations seen most commonly in cancer (Fig. 4C). It will be of great interest to determine the basis for isoform- and mutation-selective dependence on aberrant Ras signaling in cancers versus RASopathies.

The Future

When describing the rich history of Ras, we have often found it appropriate to cite the wise sayings of Yogi Berra. The finding that Ras utilizes a multitude of effectors beyond Raf, with more likely to be found, emphasized that “it ain't over 'til it's over”. The disappointing failure of efforts to develop FTIs as Ras inhibitors reflected the mindset that “when you come to a fork in the road, take it”. That we have reached numerous stages where we felt our knowledge of Ras was complete, only to find yet new wrinkles, has repeatedly prompted the sentiment that “the future is too hard to predict.” The finding that Raf inhibitors paradoxically activate Raf is an example of unanticipated signaling whose mechanistic basis is only now being worked out (“we made too many wrong mistakes”). What is in store for the future? Certainly Ras will teach us new ways in which the cell signaling circuitry is regulated. We anticipate yet more striking differences in the regulation and roles of the three Ras isoforms in normal biology and disease, at the subcellular to organismal levels. Recently, the structure of human K-Ras4B has been determined, and although not published it is available online (www.thesgc.org/structures/structure_description/3GFT/). Unexpectedly, despite the complete sequence identity of H-Ras and K-Ras4B switch II regions, the side chain orientation in this loop is described to be strikingly different in the two proteins. As essentially all previous structural studies have focused on H-Ras, future structural studies of K-Ras complexes will likely identify molecular interactions, in addition to subcellular localization, that reflect biological differences. Functional genetic screens in cell culture and in vivo model systems will continue to identify genes that unexpectedly influence Ras-dependent biology. Although Ras is presently considered an “intractable” drug target, it is still possible that new thinking and technology may produce effective approaches to directly target Ras for cancer treatment. We anticipate that development of better mouse models of Ras-driven oncogenesis will provide more accurate preclinical identification of therapeutic approaches that will prove effective in the cancer patient, compared to current mouse models in which genetic ablation of candidate targets for Ras inhibition evaluates prevention rather than “treatment” of a pre-existing cancer. Additionally, genetic ablation of a given target is clearly not equivalent to pharmacologic ablation of that target. As we head into the future, one Yogi-ism to avoid is, “You've got to be very careful if you don't know where you are going, because you might not get there”. Anticipating that unexpected findings may ultimately prove at least as informative as expected ones will give us the courage to take both logical and perhaps seemingly illogical approaches and directions, as we try to solve the remaining riddles of Ras.