A review of the biological and clinical implications of RAS-MAPK pathway alterations in neuroblastoma

QIAGEN’s Ingenuity Pathway Analysis software (IPA, QIAGEN, Redwood City, USA) was used to plot the established RAS-MAPK pathway and retrieve the references describing its interactions. The genes associated with NB were extracted from publications by Tolbert et al. [17] and Tonini et al. [18]. Next, IPA software (QIAGEN, Redwood City, USA) was used again to screen for references describing the associations between those genes and the RAS-MAPK pathway (Table 1). Figures 1 and 2 were created using BioRender.com.

The COSMIC database was used to collate the mutation data on seven different tumor types: blood, brain, breast, colorectal, lung, neuroblastoma, and skin cancers. Data were extracted for thoracic NB and all NB, and then they were pooled by adding the number of analyses and identified mutations.

The clinicaltrials.gov database was used to screen for all active, recruiting, or completed phase I to III clinical trials using molecules targeting the RAS-MAPK pathway in NB patients (Table 1, Figs. 1 and 2).

Many different studies have reported ALK mutations in approximately 6–10% of sporadic primary NB tumors and around 12–14% of the high-risk category [9‐11, 15]. To understand the relative importance of ALK in the development of NB we compared the frequency of ALK mutations in NB with the frequencies of ALK mutations in other tumor types. ALK is found to be mutated more frequently in NB than in brain, colorectal, lung, and blood cancers. Only breast and skin cancers have higher ALK mutation rates than NB, suggesting that ALK mutations play a more significant role in NB development than in other common cancer types (Table 1).

ALK mutations in NB are mainly single amino acid substitutions; in 90% of cases they are located in the kinase domain, with hot spots at amino acids F1174, R1275, and F1245 [9‐11, 15]. Interestingly, in approximately 50% of NB cases, ALK mutation is found together with MYCN amplification and correlated to poor prognosis [11].

Interestingly, in an additional 2–3% of NB samples, ALK was found to be activated by gene amplification, increasing both protein expression and activity [19‐21]. ALK amplification, which is mutually exclusive of point mutation, is also almost always associated with MYCN amplification, with both genes located in the same locus (2p23–24, Table 2) and associated with poor prognosis [11, 13, 15, 22].

Although ALK mutations have been observed in all clinical risk groups, a recent study analyzing a large cohort of more than 1500 NB patients demonstrated that the presence of an ALK mutation was independently correlated with survival: patients with an ALK mutation showed a 1.4-fold greater risk of an event within 5 years than patients without one. Statistically significant correlations were independently observed for all the groups of ALK mutations: ALK aberration, ALK copy number gain, and ALK amplification [15]. For a more general review of ALK and NB, we recommend the recent review by Trigg et al. [23].

Different studies have also reported somatic RAS-MAPK pathway gene alterations (mutation, amplification, or deletion) leading to the activation of this pathway in large cohorts of sporadic NB patients, with a frequency of around 4–10% [21, 24, 25]; the most frequently mutated genes were PTPN11 (2.9%), NRAS (0.8%), and NF1 [14, 24]. Typically, the first mutations observed in NRAS were the known activators Q61K and C181A, which were initially discovered in SK-N-SH and SK-N-AS cell lines, respectively [16, 26‐28]. Also, HRAS and KRAS can be mutated, notably by G12D and Q61K substitution, for instance [16, 29].

Most recently, Ackerman et al. sequenced the genomes of 418 pre-treated NB. In line with a previous study, they found that 9% had ALK mutations, but also that 6% had alterations in the RAS-MAPK pathway genes (NF1: 1.2%; HRAS: 1%; PTPN11: 1%: NRAS: 0.5%; BRAF: 0.5% KRAS: 0.2%; and FGFR1: 0.2%). ALK-RAS pathway alterations were detected in all NB risk categories and were found to be strongly correlated to poor outcomes, even in the high-risk category [25]. Interestingly, patients harboring RAS pathway mutations were found to have a worse prognosis than those with ALK mutations. Furthermore, Ackermann et al. also investigated the relationship between ALK-RAS alterations and telomere maintenance mechanisms (defined by MYCN, TERT, and APB alterations). Although the presence of that telomere maintenance mechanism was associated with worse outcomes compared to those without those mechanisms, an additional mutation in the ALK-RAS pathway increased NB aggressiveness, resulting in a higher mortality rate. In contrast, in NB patients without a telomere maintenance mechanism, the presence of an ALK-RAS alteration did not affect patient outcomes [25]. Eleveld et al. were recently able to define an mRNA signature, consisting of six genes (ETV4, ETV5, DUSP6, MAFF, ETV1, and DUSP4) that was correlated with an increase in RAS-MAPK pathway activity. A high expression of these signature genes was associated with poor survival in primary NB but also with MEK1/2 and ERK1/2 phosphorylation and sensitivity to different MEK1/2 inhibitors in cell lines [30].

Mutations in the genes coding for the RAF family of proteins are very rare in NB. BRAF aberrations have been observed in 1% of NB cases, whereas the study by Shukla et al. reported no alterations in ARAF and CRAF [26]. However, there was a greater expression of all the RAF proteins in the high-risk NB category [31]. The main mutation found in BRAF is the V600E substitution [26]. This mutation is located in BRAF’s kinase domain and leads to a constitutively active form of BRAF [32]. Interestingly, a tandem duplication in the BRAF gene was also detected in a relapse NB. This duplication resulted in the expression of a BRAF transcript that encodes a protein with two kinase domains. Accordingly, the protein was twice as active, which made the RAS-MAPK signaling pathway more active [16].

A MEK1/2 mutation is a very rare event in NB, with a frequency of 0.5% [26]. In one NB sample, a MEK1 somatic activating mutation (K57N) was observed, located between the protein’s nuclear export signal and its kinase domain. This genetic lesion leads to constitutive activation of the RAS-MAPK pathway in vitro [33]. On the contrary, no reports of ERK mutations have been found in NB.

Comparisons of NB with six other major cancer types showed significant differences in the frequency of mutations in the major RAS-MAPK pathway genes. NB, brain and breast tumors do not demonstrate high mutation frequencies along this pathway (Table 1), while colorectal, blood, lung, and skin cancers have high frequencies in at least one of the RAS-MAPK pathway genes. Interestingly, two genes, RAF1 (3p25.2) which encodes for the CRAF protein, and GRB2 (2p23-2p23.1), were found to be located in regions that sustain frequent chromosomal abnormalities in NB (Table 2).

Finally, different studies have also reported alterations in genes associated with the RAS-MAPK pathway, such as CDK4, LIN28B, CCND1, SMO, SOS1, CIC, DMD, and DUSP5 [24, 25, 30, 34, 35].

Different studies have reported that ALK mutations were more frequently detected in NB relapse samples, with a frequency of 15–40% [16, 24, 36]. In 2015, Eleveld et al. analyzed 23 paired diagnostic and relapsed NB samples. They were able to show not only a clonal evolution from diagnosis to relapse sample but also that 78% (18/23) of the relapsed NB samples harbored a genetic aberration predicted to activate the RAS-MAPK pathway (ALK (10/23); NF1 (2/23); PTPN11 (1/23); FGFR1 (1/23); NRAS (1/23); KRAS (1/23); HRAS (1/23); and BRAF (1/23)) [16]. In line with these results, Padovan-Merhar et al. reported that suspected ALK driver mutations were present in 7% (3/43) of samples at diagnosis, 17% (7/41) of the post-treatment samples, and 20% (11/54) of the samples at relapse. Furthermore, they observed more suspected oncogenic ALK mutations in relapsed disease than at diagnosis, as well as enrichment of RAS-MAPK pathway mutations at relapse [24]. Also, single-nucleotide variants (SNVs) in RAS proteins are more frequent in relapsed NB, notably, for instance, in HRAS with the somatic Q61K mutation [16, 24, 26], resulting in the constitutively active form of RAS due to inactivation of the GTPase domain [37].

Familial cases of NB are very rare, representing 1–2% of all NB cases, and they are inherited in an autosomal-dominant manner. In recent years, many advances have been made in our genetic comprehension of these cases. PHOX2B (4p12), a gene known to be an essential regulator of the development of a normal autonomic nervous system and frequently associated with neurocristopathies, was the first to be identified in this setting in 2004 [38, 39]. However, it was only detected in a small subset of familial NB (approximately 10%). In 2008, two different, independent studies reported ALK (2p23) as a gene involved in more than half of familial NB cases. Additional studies finally reported ALK germline mutations in almost 80% of these families [15, 17, 40]. The majority of these mutations are located in the ALK tyrosine kinase domain, resulting in constitutive activation of the kinase [9, 10]. The penetrance is variable, highly related to the activating effect of the mutation [15], but estimated at around 50% overall [41]. As yet, no other genes have been identified to explain the 10% of familial cases without the PHOX2B or ALK germline mutations.

Interestingly, NB susceptibility genes were recently identified in the germline DNA of a patient with sporadic NB, suggesting that common germline polymorphisms with low penetrance (i.e., NBAT-1, CASC15, BARD1, LMO1, HSD17B12, DUSP12, LIN28B, HACE1, SPAG16, NEFL, MLF1/RSRC1, CPZ, CDKN1B, SLC16A1, MSX1, MMP20, KIF15) or more rare variants with higher penetrance (i.e., ALK, BARD1, CHEK2, AXIN2, TP53, APC, BRCA2, SDHB, SMARCA4, LZTR1, BRCA1) may also have a relevant role in NB carcinogenesis [17, 18]. Remarkably, when investigating the associations between these genes and the RAS-MAPK pathway, many of them were reported to interact with this pathway either in NB or other cancers (Table 3). This further supports the view that the RAS-MAPK pathway is one of the central targets in the process of tumorigenesis.

Additionally, NB has been reported to be associated with diseases well-known to predispose cancer, such as Li Fraumeni syndrome (TP53 mutations) [42], familial paraganglioma/pheochromocytoma (SDHB mutations) [43], and Beckwith–Wiedemann syndrome (CDKN1C mutations or loss of expression) [44]. It has also been associated with other rare syndromic diseases, more particularly with those characterized by the presence of germline mutations located along the RAS-MAPK pathway. This group of diseases, also known as RASopathies, includes neurofibromatosis 1 (NF1 mutations), Costello syndrome (HRAS mutations), Noonan syndrome, and such Noonan-like syndromes as Noonan syndrome-like disorder with loose anagen hair and the LEOPARD syndrome (mutations in PTPN11, SOS1, KRAS, NRAS, RAF1, BRAF, MEK1, SHOC2, MEK2, RIT1, and CBL) [41, 45‐47].

The RAS-MAPK pathway is usually activated by receptor tyrosine kinase (RTK) (Fig. 1). As discussed above, ALK is the most studied RTK in NB. However, other RTKs are known to activate the RAS-MAPK pathway and have sometimes been found to be mutated in relapsed NB. These include FGFR1 (fibroblast growth factor receptor 1), EGFR (epidermal growth factor receptor) [24], and Trk A and B [73]. The main aberration found in FGFR1 [50] is the somatic N546K mutation [16], which is located in the kinase domain and alters auto-phosphorylation, leading to an increase in its kinase activity [74]. Because the present review focuses on the RAS-MAPK pathway, we refer the reader to other excellent general reviews of RTKs [75] and their involvement in NB [23, 76].

Activated RTKs’ first targets are the three RAS proteins—KRAS, NRAS, and HRAS (Fig. 1). The activation is a complex sequence of events, involving the ultimately inactive RAS being converted into its active form by the addition of GTP. At least three adaptor proteins—SHP2 (PTPase encoded by PTPN11), Shc, and Grb2—are required for the first step in the signalization reaction [77]. This complex forms after RTK’s phosphorylation by the recruitment of PTPases which contain the SH2 motifs responsible for specific binding between PTPase and RTK. Next, PTPase becomes tyrosine-phosphorylated, which induces a significant change in its conformation and increases phosphatase activity [78, 79]. The change in conformation appears to be essential for the recruitment of the other two factors, Shc and Grb2 [80]. Nevertheless, SHP2 also appears to serve other functions, as some studies suggest that it dephosphorylates STAT3, an important transcription factor activated by terminal ERK1/2 kinase. After being recruited, Shc is activated by phosphorylation at Tyr317 [81]. Adaptor proteins are assembled on RTKs’ intracellular sites, and many known mutations lead to the constitutive association of both Shc and Grb2 proteins with RTK [77]. As already mentioned, mutations in the PTPN11 gene are found in 3% of high-risk NBs [82]. The major mutation is the heterozygous A72T substitution located in the first SH2 domain of SHP2, converting SHP2 into a constitutively active phosphatase. After the assembly of Shc and Grb2 adaptor proteins, Sos1 protein is recruited to catalyze the addition of GTP to RAS [83]. Sos1 is a mammalian nucleotide exchange factor that, once recruited to the Shc/Grb2 complex by translocation to the plasma membrane, activates RAS through the exchange of GDP for GTP [84].

One of the most important aspects of RAS phosphorylation is its location. It has been demonstrated that the entire complex’s location with respect to cellular membranes facilitates the rate of exchange. The diffusion of proteins was shown to be insufficient for the reaction to happen [85]. Active transport is necessary for the subsequent transmission of the signal towards its targets. Although the sequences of all three RAS proteins are almost identical, they are located at different cellular compartments. KRAS is more abundant in cytoplasmic membranes, whereas HRAS is more common in intracellular vesicular membranes. These differences are presumably due to differential processing in the Golgi apparatus. Nevertheless, it has been demonstrated that Grb2a associates with both KRAS and HRAS, irrespective of their location [86].

In a physiologically normal state, cells have several means of controlling the activation and deactivation of RAS proteins. The first negative regulation to be described was the activation of the hydrolysis of RAS-bound GTP to GDP by p120GAP, neurofibromin, and GAP1. Neurofibromin, coded by the NF1 gene, exerts its inhibitory action by stimulating the hydrolysis of RAS-GTP [87]. NF1 seems particularly important in the development of NB as NF1 aberrations have been found in primary NB and at an even higher frequency in relapsed NB samples [24, 88]. The NF1 mutations observed in primary NB are typically SNVs, promoting the loss of neurofibromin’s activity. In relapsed tumors, however, we mostly observe NF1 deletion. Depending on the NB cell lines, we observe either homozygous deletion of various sizes or heterozygous deletion with a splice site mutation in one of the alleles [16, 88]. In all of these situations, loss of neurofibromin results in the constitutively active form of RAS and thus over-signalization of the RAS-MAPK pathway.

The second negative regulation to be described was the feedback loop through the phosphorylation of Sos1 and the subsequent inactivation of RTK/Shc/Grb2/Sos1 by ERK1. The phosphorylation of Sos1 does not activate the dissociation of Sos1 from Grb2 but rather the dissociation of the entire Grb2/Sos1 complex from phosphorylated RTK [89]. It is currently unknown how much the deregulation of this feedback contributes to the activation of the RAS-MAPK pathway and to the development of NB or other cancer types.

In normal cells, RAF proteins (ARAF, BRAF, and CRAF) represent the next step in the signaling cascade (Fig. 1). Although BRAF is one of the most frequently mutated genes in many cancers, it is interesting that BRAF does not directly interact with activated RAS. Signal transduction to the BRAF protein is accomplished by the heterodimerization between CRAF and BRAF that is dependent on the presence of RAS. That heterodimerization is also dependent on the phosphorylation of BRAF at the serine 621 and of the CRAF’s C terminus, whereas CRAF’s N terminus is not important for this interaction. On the other hand, CRAF’s association with RAS is independent of RAS activity as assessed by the dominant-negative form of RAS (G12V) or by the truncated RAS mutant lacking the COOH-terminal serine residues necessary for membrane localization [90, 91]. Interestingly, MEK1/2 phosphorylation can be achieved by CRAF alone, but it is significantly stronger in the presence of the BRAF/CRAF heterodimer. In cells, this complex is located at the membrane, which is to be expected as RAS is a membrane-bound protein. Nevertheless, the evidence suggests that membrane localization is not responsible for the dimerization of BRAF/CRAF. 14–3-3 is an additional protein necessary for the interaction between BRAF and CRAF. It is responsible for maintaining the complex structure and RAS-dependent RAF activation through the binding of CRAF’s S621 amino acid residue [92‐94]. Although the three RAF isoforms are very similar, their ability to activate downstream targets is markedly different. As discussed above, the heterodimerization of different RAF isomers is important to their activity, with the most sensitive being CRAF’s dependence on the presence of BRAF. Likewise, BRAF’s activity is significantly lower in the absence of CRAF. ARAF, however, the least active isoform, is also the most dependent on BRAF but is independent of CRAF. It is of note that BRAF is capable of spontaneous homodimerization but that CRAF and ARAF are not. Mutations in BRAF and CRAF play significant roles by increasing homo and heterodimerization and catalytic activity [95]. Although ARAF is the least activate isoform of RAF, it is very significantly activated by the oncogenic Src protein [96].

The RAF protein kinase family is the major activator of the mitogen-activated protein kinases, MEK1 and MEK2 (MEK1/2) (Fig. 1). CRAF activates both MEK1 and MEK2, whereas ARAF and BRAF predominantly activate MEK1 [97]. To be fully active, MEK1/2 must be phosphorylated on two serine residues, at positions 218 and 222, located in the activation loop. Although RAF proteins are the predominant activators, MEK1/2 can also be phosphorylated at the same residues by MEK Kinase (MEKK), c-Mos, and growth factor-stimulated MEK1/2 activator [98‐100]. Once activated, MEK1/2 can, in turn, activate its main substrates—ERK1 and ERK2—by phosphorylation. The ERK family is composed of several isoforms, from ERK1 to ERK8, but only ERK1 and ERK2 (also known as the p44 and p42 kinases, respectively) are involved in the RAS/MAPK signaling pathway. To learn more about other isoforms, see Bogoyevitch et al. [101]. In contrast to the activation of MEK1/2, no other ERK1/2 activators are currently known. MEK1/2 phosphorylates ERK1/2 at threonine 183 and tyrosine 185 in the Thr-Glu-Tyr motif [102]. ERK1/2 activation is facilitated by its spatial proximity to MEK1/2 within a protein complex including KSR and MP-1 [103]. The enzymatic reaction is enhanced by electrostatic interactions between the basic amino acids (lysine) present in the ERK2-docking site in the N-terminal region of MEK1 and the acidic amino acids (aspartate) located in the MEK1-docking site in the C-terminal portion of ERK2 [104, 105]. The specificity of the RAF-MEK1/2-ERK1/2 reactions is reinforced by the fact that MAPKs are proline-directed kinases, meaning that they strongly recognize their specific phosphorylation sites through the proline at position + 1 of them. This proline is required for efficient activation of MEK1/2 from RAF and of ERK1/2 from MEK1/2 [97, 106].

Upon activation, ERK dimerizes to form either ERK1 or ERK2 homodimers, although the functional implication of this dimerization remains unclear. Khokhlatchev et al. showed that this dimerization is involved in nuclear translocation, whereas most recently, Casar et al. demonstrated it to be essential for the activation of cytoplasmic targets [107, 108]. Via phosphorylation, active-ERK1/2 regulates a large number of cytoplasmic and nuclear targets which mainly regulate cell proliferation, differentiation, survival, and apoptosis (Figs. 1 and 2). Many of those targets contain a D domain and a DEF motif that improves the efficacy and specificity of the interaction with ERK1/2’s CD/CRS (common docking motif/cytosolic retention sequence) domain [109]. In 2017, based on various phospho-proteomic analyses, Ünal et al. reviewed 2507 ERK1/2-targets, of which 659 were direct and 1848 were indirect targets [110]. Direct targets are mainly involved in cell-cycle regulation and cell-signaling, for example, the well-known ERK1/2 targets, the Ets family of transcription factors—ELK-1, ELK-4, Ets-2—but also MYC, phospholipase A2, p90^RSK, paxillin, and calnexin [67, 111‐113]. Under stress conditions, ERK1/2 can become involved in cell death by phosphorylating p53 and thus inducing apoptosis [114]. However, ERK1/2 is also available for the activation of substrates in a phosphorylation-independent way through protein–protein interaction, such as with topoisomerase II alpha [115].

Importantly, the signal duration must be finely regulated to avoid the loss of cell control which can promote the development of diseases such as cancer. Negative feedback and phosphatase activity regulate the signal induced by the ERK1/2 / MAPK pathway. ERK1/2 can rapidly inhibit the pathway’s main components (e.g., MEK1/2, RAF, Sos1, RTK, KSR) through phosphorylation and regulate the expression of phosphatases such as dual-specificity MAPK phosphatase (MKPs) and Sprouty proteins, both of which inhibit the pathway. To learn more about this negative regulation, see the review by Lake et al. [116].