PTPN11 mutations

PTPN11, the predominant gene associated with Noonan syndrome, encodes a nonreceptor protein tyrosine phosphatase known as SHP2; missense mutations are found in approximately half of all clinically diagnosed Noonan syndrome patients examined (Ref. Reference Tartaglia32). The locus of PTPN11 is in chromosome band 12q24.1. The PTPN11 gene spans 90 kb and comprises 16 exons containing a coding sequence of 1779 bases (Ref. Reference Tartaglia39). The SHP2 protein consists of N-terminal and C-terminal SH2 domains and a catalytic domain (PTP). The catalytic function of the protein is autoinhibited through a blocking interaction of the N-SH2 domain and the catalytic PTP domain (Ref. Reference Hof40). The majority of Noonan-syndrome-causing mutations in PTPN11 are missense, and are restricted to specific regions within the gene. The mutations cluster in exons 3, 7, 8 and 13, which mainly affect residues involved in the interface between the N-SH2 and PTP domains, with the most common mutation being an A to G change at nucleotide 922 that results in an N308D substitution. This mutation, and others in this region, affect the stability of the catalytically inactive form, and therefore interfere with the protein's ability to transition from the active to the inactive conformation (Refs Reference Keilhack41, Reference Tartaglia42). Thermodynamic modelling (Ref. Reference Tartaglia32), and in vitro (Refs Reference Tartaglia39, Reference Fragale43, Reference Niihori44) and in vivo (Refs Reference Araki45, Reference O'Reilly46) studies, have demonstrated that these SHP2 mutants produce a gain of function, resulting in increased signalling down the Ras–MAPK pathway. Additionally, mutations that affect regions outside the N-SH2–PTP interaction region have also been shown to cause a gain of function of SHP2 through other mechanisms (Ref. Reference Martinelli47).

Individuals with Noonan syndrome are predisposed to juvenile myelomonocytic leukaemia. One Noonan-syndrome-causing mutation, which results in a T73I residue substitution, has been specifically associated with this predisposition (Ref. Reference Choong26). This same mutation is somatically present in individuals with juvenile myelomonocytic leukaemia who do not have Noonan syndrome (Ref. Reference Kratz48). Other PTPN11 somatic mutations are also present in leukaemia and in some tumours (Refs Reference Kratz48, Reference Tartaglia49, Reference Bentires-Alj50, Reference Chen51, Reference Loh52, Reference Miyamoto53, Reference Schubbert54, Reference Yoshida55), but the somatic PTPN11 mutations associated with cancer are typically distinct from those associated with Noonan syndrome, and usually confer a higher activation of SHP2 function (Refs Reference Tartaglia42, Reference Schubbert54). Not only do PTPN11 mutations demonstrate a paternal bias, but Noonan syndrome affects males approximately twice as often as it does females. This sex ratio bias may be due to sex-specific developmental effects of PTPN11 mutations that result in a higher mortality of female embryos (Ref. Reference Tartaglia56).

KRAS mutations

KRAS mutations have been implicated in a small percentage of individuals with Noonan syndrome (Table 3). The KRAS gene is located on 12p12.1 and consists of four coding exons. Two splice variants exist: KRASA and KRASB (Ref. Reference Friday and Adjei57). KRASA is expressed in a tissue-specific and developmentally restricted fashion, whereas KRASB is ubiquitously expressed. KRAS mutations were independently identified in a small number of individuals clinically diagnosed with Noonan and CFC syndromes (Refs Reference Schubbert37, Reference Niihori58).

When 175 PTPN11 mutation-negative individuals with the clinical diagnosis of Noonan syndrome and 12 individuals with the clinical diagnosis of CFC syndrome were examined for KRAS mutations (Ref. Reference Schubbert37), five Noonan syndrome and one of the CFC syndrome individuals were found to have novel missense mutations in KRAS. The mutations in Noonan syndrome individuals included T58I, V14I and D153V. The phenotype associated with the T58I substitution was reported to be more severe than with the V14I substitution. Functional studies revealed that these mutations activate the MAPK pathway (Refs Reference Schubbert37, Reference Schubbert59). Ras cycles between a GDP-bound inactive state and a GTP-bound active state. The transition between these two states is controlled by GTPase-stimulating proteins (GAPs) and guanine-nucleotide-exchange factors (GEFs). Biochemical analyses of mutations T58I and V14I showed a reduced intrinsic GTPase activity of Ras compared with the wild-type protein, resulting in a reduction in Ras inactivation and therefore increased signalling down the MAPK pathway (Ref. Reference Schubbert37). The one CFC syndrome individual in the study had a missense KRAS mutation resulting in a P34R residue substitution.

In a separate study, 43 individuals with the clinical diagnosis of CFC syndrome were also screened for KRAS mutations (Ref. Reference Niihori58). Three individuals had missense KRAS mutations that resulted in either a G60R or D153V residue substitution, with the latter mutation also being reported in clinically diagnosed Noonan syndrome individuals (Ref. Reference Schubbert37). Subsequently, KRAS mutations have been identified in cohorts of individuals with severe Noonan syndrome (Refs Reference Zenker38, Reference Carta60).

These studies demonstrate that a variable phenotype is associated with KRAS mutations, and reinforce the difficulty of clinically distinguishing between Noonan syndrome and CFC syndrome, even given knowledge of underlying molecular defects. Further evaluation of the patient populations and functional studies will resolve these issues.

SOS1 mutations

SOS1 missense mutations were reported independently by two groups in individuals with Noonan syndrome and without mutations in either PTPN11 or KRAS (Refs Reference Roberts33, Reference Tartaglia34). At present, SOS1 mutations are the second most common cause of Noonan syndrome after PTPN11 mutations, and account for ~13% of individuals with Noonan syndrome (Table 3). SOS1, located on chromosome 2p22.1, consists of 23 exons and encodes a Ras-GEF. Ras-GEF proteins are responsible for stimulating the conversion of Ras from the inactive GDP-bound form to the GTP-bound active form. The majority of SOS1 missense mutations reported were located in codons encoding residues that are responsible for stabilising the SOS1 protein in an inhibited conformation. Therefore, alteration of these residues disrupts the autoinhibition of SOS1 Ras-GEF activity, resulting in a gain of function of SOS1 and a subsequent increase in the active form of Ras and increased Ras–MAPK pathway signalling (Refs Reference Roberts33, Reference Tartaglia34).

The phenotype resulting from SOS1 mutations, although it falls within the spectrum of characteristics encompassed by Noonan syndrome, appears to be distinctive. There is a higher prevalence of ectodermal abnormalities and a lower prevalence of cardiac abnormalities than observed with PTPN11 mutations (atrial septal defects were present in individuals with SOS1 mutations that produced the highest degree of SOS1 protein activation). Although Noonan and CFC syndromes may phenotypically overlap, one initial study did not identify SOS1 mutations in a cohort of individuals clinically diagnosed with CFC syndrome (Ref. Reference Zenker61). It is unclear if individuals with germline mutations in SOS1 are predisposed to cancer. In a survey of 810 primary human malignancies, somatic mutations in SOS1 were very rare, which may hold significant implications as to the predisposition for developing cancer in Noonan syndrome individuals who have SOS1 mutations (Ref. Reference Swanson62).

RAF1 mutations

Two groups simultaneously reported mutations in RAF1 in Noonan syndrome individuals who did not have mutations in PTPN11, SOS1 or KRAS (Refs Reference Pandit35, Reference Razzaque36). The majority of missense RAF1 mutations associated with Noonan syndrome were clustered in two regions: in conserved region 2 (CR2) flanking S259; and in conserved region 3 (CR3), in and surrounding the activation segment (Refs Reference Pandit35, Reference Razzaque36). In addition, two mutations in the C-terminus were also reported (Ref. Reference Pandit35). Mutations flanking S259 all resulted in a gain of function, since this region is responsible for autoinhibition of kinase activity (Refs Reference Pandit35, Reference Razzaque36). Mutations in the CR3 and the C-terminus also resulted in increased kinase activation (Refs Reference Pandit35, Reference Razzaque36). By contrast, mutations within the activation segment were shown to have lower kinase activity than wild-type CRAF (the protein product of RAF1) (Ref. Reference Pandit35). All the gain-of-function mutations in RAF1 were associated with a much higher prevalence of hypertrophic cardiomyopathy compared with other Noonan syndrome individuals who have other known gene mutations (~95% compared with ~18%, respectively) (Ref. Reference Pandit35).

The mammalian genome contains three Raf genes: ARAF, BRAF and RAF1. RAF1 is located on chromosome 3p25, contains 17 exons and spans 80 kb. Its protein product CRAF is a serine/threonine kinase that is one of the direct downstream effectors of Ras and is, in turn, responsible for the activation of MEK1 and MEK2. CRAF has a much lower basal kinase activity than BRAF, with somatic RAF1 mutations being very rare in cancer (Ref. Reference Emuss63). The few RAF1 mutations associated with cancer are not the same as those associated with Noonan syndrome.

HRAS mutations

Costello syndrome was hypothesised to be caused by dysregulation of the Ras–MAPK pathway because of the high prevalence of cancer seen in this syndrome and phenotypic overlap with Noonan syndrome and neurofibromatosis 1 (Ref. Reference Tartaglia18). In the first report of mutations in individuals with a clinical diagnosis of Costello syndrome, heterogeneous germline mutations in HRAS were identified in 12 of 13 patients (Ref. Reference Aoki81). In this initial cohort, over half of the patients (7/13) had a nucleotide c.34G > A transition in codon 12 of coding exon 1; other mutations found at a lesser frequency included c.35G > C and c.35G > T in codon 12 and a c.38G > A substitution in codon 13. Subsequent confirmatory reports on larger cohorts of patients with the clinical diagnosis of Costello syndrome have shown that the c.34G > A substitution in codon 12 – resulting in a G12S missense substitution in HRAS – accounts for the vast majority of HRAS mutations associated with Costello syndrome (Refs Reference Estep74, Reference Gripp82, Reference Kerr83, Reference Lo84, Reference Schulz85, Reference van der Burgt86, Reference van Steensel87, Reference Zampino88); the distribution frequency of mutations reveals that more than 80% of individuals have a G12S (c.34G > A) substitution, with the second most common substitution being G12A. Rare mutations, which may result in an atypical phenotype, have also been reported that result in amino acid substitutions at residues T58, L117 and A146 (Refs Reference Kerr83, Reference Zampino88, Reference Denayer89, Reference Gripp90). Because phenotypic features in Costello syndrome are heterogeneous and most mutations affect codon 12, additional investigation will be needed on larger patient cohorts to establish if a genotype–phenotype correlation exists. In addition, the phenotypic overlap between Costello and CFC syndromes may make a clinical diagnosis challenging. However, individuals with HRAS mutations are considered to have Costello syndrome (Refs Reference Rauen91, Reference Kerr92).

The origin of constitutional germline mutations causing Costello syndrome reflects a paternal bias. One study examined 42 Costello syndrome mutation-positive individuals. Sixteen families were informative when four polymorphic markers were used to identify the parental origin (Ref. Reference Sol-Church93). Of these, 14 Costello syndrome patients had the causal mutation on the paternal chromosome, and in two the mutation was of maternal origin. Further support for the paternal bias of de novo mutations has been demonstrated (Refs Reference Schulz85, Reference Zampino88). In addition to paternal bias, somatic mosaicism for an HRAS G12S substitution has been identified in an individual with Costello syndrome (Ref. Reference Gripp94).

HRAS is a highly conserved gene located on 11p15.5 (Ref. Reference Midgley and Kerr3). Substitutions of glycine at codon 12 and 13 are gain-of-function somatic mutations that occur in cancer. These substitutions disrupt guanine-nucleotide binding, which results in a reduction in intrinsic and GAP-induced GTPase activity. This causes Ras to remain in the active state and leads to increased MAPK activity (Refs Reference Gibbs95, Reference McGrath96, Reference Sweet97). In fact, substitution of any amino acid except for proline at this location may cause activation with varying transformation efficiencies (Refs Reference Seeburg98, Reference Fasano99). However, activated Ras alone does not cause transformation of primary cells in culture, but rather transformation requires additional co-operating oncogenes (Refs Reference Newbold and Overell100, Reference Land, Parada and Weinberg101, Reference Hahn102).

Biochemical investigation of a novel HRAS mutant protein carrying the substitution L117R that causes Costello syndrome has demonstrated normal intrinsic GTP hydrolysis and responsiveness to GTPase-activating proteins; however, the nucleotide dissociation rate is increased, resulting in an effective 80-fold increase in guanine-nucleotide exchange (Ref. Reference Denayer89). The crystal structure of this mutant showed alteration of the side chain, predicting unfavourable nucleotide-binding properties. Therefore, this novel L117R mutant protein produces a phenotype similar to that arising from mutations in codons 12 and 13 that impair GTP hydrolysis, albeit by a different mechanism.

Ras mutations in codons 12, 13 or 61 resulting in aberrant protein activation are present in ~20% of all tumours (Ref. Reference Bos2). Although germline mutations in codon 12 and 13 have been identified in Costello syndrome, no mutations in codon 61 have been reported to date. Somatic point mutations in HRAS account for only a small fraction (~1%) of Ras mutations observed in sporadic cancers, with this mutation appearing most commonly in bladder cancer (Ref. Reference Levesque103). Costello syndrome patients have a predisposition to neoplasms including rhabdomyosarcoma. Interestingly, HRAS mutations are rarely seen in sporadic rhabdomyosarcoma (Refs Reference Takahashi104, Reference Yoo and Robinson105), implicating a possible alternative mechanism of rhabdomyosarcogenesis in Costello syndrome. Loss of heterozygosity (LOH) in Costello syndrome malignancies (Ref. Reference Kerr106) is supported by the observation that mRNA isolated from a ganglioneuroblastoma obtained from a Costello syndrome individual with a G12A germline mutation had monoallelic expression of the mutated allele, although there was no LOH detected in the tumour genomic DNA (Ref. Reference Aoki81). In addition, the loss of wild-type HRAS with retention of the activated HRAS allele has been demonstrated in a patient with recurrent biphasic embyronal rhabdomyosarcoma/fibrosarcoma who had a germline c.34G > A mutation (G12S) (Ref. Reference Estep74).

BRAF mutations

Two research groups independently identified mutations in BRAF as causing CFC syndrome (Refs Reference Niihori58, Reference Rodriguez-Viciana121). When 23 unrelated individuals with the clinical diagnosis of CFC syndrome who lacked mutations in HRAS (Ref. Reference Estep74) and PTPN11 were screened for BRAF mutations (Ref. Reference Rodriguez-Viciana121), 18 (78%) were found to have heterogeneous missense mutations in BRAF. In a separate study, 43 CFC-syndrome-diagnosed individuals were screened for mutations, which identified eight unique BRAF mutations among 16 patients (37%) (Ref. Reference Niihori58). BRAF is located on chromosome 7q34, contains 18 exons and spans ~190 kb. The protein product BRAF is a serine/threonine protein kinase and is one of the many direct downstream effectors of Ras (for review see Ref. Reference Wellbrock, Karasarides and Marais122). The only known downstream effectors of BRAF are MEK1 and MEK2. There are three conserved regions in BRAF: CR1, part of the regulatory N-terminal region contains the Ras-binding domain and the cysteine-rich domain, both of which are required for recruitment of BRAF to the cell membrane; CR2, also part of the N-terminal region, is the smallest of the conserved regions; and CR3 encompasses the kinase domain and contains a glycine-rich loop (exon 11) and the activation segment (exon 15) of the catalytic domain. Mutations identified in CFC syndrome cluster in two regions: in the cysteine-rich domain of CR1 and in the protein kinase domain.

Examination of published CFC syndrome cohorts reveals that mutations in BRAF account for ~73% of mutation-positive individuals (Refs Reference Niihori58, Reference Schulz85, Reference Rauen91, Reference Al-Rahawan120, Reference Rodriguez-Viciana121, Reference Narumi123, Reference Nava124, Reference Gripp125). There is a wide distribution of BRAF mutations; 32 novel CFC syndrome BRAF mutations in 23 different codons have been identified. Of the 18 exons in BRAF, CFC syndrome mutations have been found in exon 6 and exons 11–16, with the majority of missense mutations in exon 6 (41%) and exon 12 (21%). The most frequent amino acid substitutions include exon 6 Q257R (29%), exon 12 E501 (12%) and exon 11 G469E (6%).

Functional analysis of expressed proteins from BRAF genes harbouring mutations reveal that the BRAF mutations associated with CFC syndrome are similar to the mutations identified in cancer (Ref. Reference Wan126). The majority of CFC syndrome mutations result in proteins with an increased kinase activity; however, a few have kinase-impaired activities (Refs Reference Niihori58, Reference Rodriguez-Viciana121). Increased BRAF activity results in increased signalling down the MAPK pathway. Curiously, kinase impairment of BRAF has also been shown to increase signalling down the MAPK pathway, possibly through CRAF interaction (Refs Reference Wan126, Reference Garnett127).

Somatic mutations in BRAF have been reported at a high frequency in melanoma and thyroid cancer and at a lower frequency in colorectal and ovarian cancer. Approximately 70 missense mutations affecting 34 codons have been reported (http://www.sanger.ac.uk/genetics/CGP/cosmic). The majority of somatic BRAF mutations result in missense substitutions found in, but not limited to, exon 11 (the glycine-rich loop) and exon 15 (the activation segment) in the kinase domain (for review see (Ref. Reference Wellbrock, Karasarides and Marais122). By contrast, only 18% of CFC syndrome mutations are found in exon 11 and 7% in exon 15. Mutations in these two regions are believed to convert BRAF into its active conformation (Ref. Reference Wan126). One somatic mutation, ^V600EBRAF, results from a T → A transversion at c.1796, substituting glutamic acid for valine at position 600. ^V600EBRAF, which has increased kinase activity, accounts for over 90% of somatic BRAF mutations identified in human cancer (Ref. Reference Davies128). This mutant BRAF protein may exert its oncological effect in a similar way to the oncogenic effect of an activated Ras. However, somatic ^V600EBRAF mutations are also found in benign skin lesions and premalignant colon polyps as well as a majority of benign naevi (Refs Reference Yuen129, Reference Rajagopalan130, Reference Pollock131). This suggests that the ^V600EBRAF mutation activation of MAPK signalling is important in melanocytic neoplasia, but in isolation is insufficient for tumourigenesis (Ref. Reference Pollock131).

MAP2K1 and MAP2K2 mutations

Missense mutations in MAP2K1 and MAP2K2 also cause CFC syndrome (Ref. Reference Rodriguez-Viciana121). The MAP2K1 gene is located on chromosome 15q22.31 and spans ~104 kb; MAP2K2 is located on chromosome 19p13.3 and spans ~34 kb (Ref. Reference Pearson132). Each gene contains 11 exons with intervening sequences. The MAP2K1 gene encodes the mitogen-activated protein kinase kinase 1 (MEK1); likewise MAP2K2 encodes the protein MEK2. MEK1 and MEK2 are threonine/tyrosine kinases activated by Raf, with both isoforms having the ability in turn to activate ERK1 and ERK2. The MEK1/2 proteins have about 85% amino acid identity (Ref. Reference Wu133) but do not serve redundant purposes (Refs Reference Brott134, Reference Alessandrini, Brott and Erikson135).

Missense MEK mutations were initially identified in three of five CFC syndrome individuals who were BRAF-mutation negative (Ref. Reference Rodriguez-Viciana121). MAP2K1 and MAP2K2 mutations make up ~27% of mutations in CFC syndrome individuals in whom a gene mutation has been identified (Refs Reference Niihori58, Reference Schulz85, Reference Rauen91, Reference Al-Rahawan120, Reference Rodriguez-Viciana121, Reference Narumi123, Reference Nava124, Reference Gripp125). The vast majority of these are missense substitutions located in exons 2 and 3. The most common mutation is MEK1 Y130C (~41% of all the MEK mutations). Rare in-frame deletions have been identified (Refs Reference Yoon117, Reference Nava124). Functional studies of the proteins encoded by the novel CFC syndrome MEK mutations have shown that all the mutant proteins studied are more active than wild-type MEK in stimulating phosphorylation of ERK (Ref. Reference Rodriguez-Viciana121; K. Rauen, unpublished).

MEK, like Raf, has been well studied in the context of cancer. Cellular transformation due to activation of the MAPK cascade ultimately is the result of MEK activation (Ref. Reference Cowley136). Constitutively active MEK mutants, produced in the laboratory by deleting the N-terminus of the protein, or by altering key regulatory residues, promote transformation of mammalian cells both in vitro and in vivo (Ref. Reference Mansour137). Several primary human tumour cell lines have increased MEK1 and MEK2 activities (Refs Reference Sivaraman138, Reference Hoshino139), but until recently no somatic kinase domain MAP2K1 and MAP2K2 mutations had been reported in cancer (Refs Reference Bignell140, Reference Davies141, Reference Hunter142, Reference Stephens143; http://www.sanger.ac.uk/genetics/CGP/cosmic). The first functional MEK1 mutation, D67N, was identified in an ovarian cancer cell line, with functional studies determining that this mutant protein has increased activity as measured by an increase in ERK phosphorylation (Ref. Reference Estep144). Interestingly, this mutation was also identified as causal for CFC syndrome (Ref. Reference Nava124).