How can same-gene mutations promote both cancer and developmental disorders?

The question of how same-gene mutations can drive both cancer and neurodevelopmental disorders has been puzzling. It has also been puzzling why those with neurodevelopmental disorders have a high risk of cancer. Ras, MEK, PI3K, PTEN, and SHP2 are among the oncogenic proteins that can harbor mutations that encode diseases other than cancer. Understanding why some of their mutations can promote cancer, whereas others promote neurodevelopmental diseases, and why even the same mutations may promote both phenotypes, has important clinical ramifications. Here, we review the literature and address these tantalizing questions. We propose that cell type–specific expression of the mutant protein, and of other proteins in the respective pathway, timing of activation (during embryonic development or sporadic emergence), and the absolute number of molecules that the mutations activate, alone or in combination, are pivotal in determining the pathological phenotypes—cancer and (or) developmental disorders.

INTRODUCTION: THE PUZZLES

Among the puzzling questions confronting basic and translational research is how certain oncogenic mutations can promote cancer, while other mutations of the same proteins, or even the same mutations, can provoke other syndromes. Among the syndromes are RASopathies (), a group of genetic neurodevelopmental disorders arising from mutations in mitogen-activated protein kinase (MAPK) pathway–related proteins (), recently shown to share traits with autism spectrum disorder (ASD) (), cerebral palsy (), and more. The question of why patients with neurodevelopmental disorders are more prone to cancer has also been enigmatic. Here, we suggest that these seemingly puzzling observations can be understood based on four key considerations: (i) the level of expression of the protein in the type-specific cell, (ii) the potency of the single or combined mutations, (iii) the time window (expression levels in a specific cell type vary, e.g., during embryonic development or in the adult differentiated state), and (iv) the expression levels of the nodes in the pathway that the signal flows through. Often, the mutations are considered independently of the mRNA levels. However, the number of activated molecules depends on protein expression, and signal propagation requires that the expression levels of all proteins in the pathway not be too low. Notably, even without the mutations, very high expression levels can drive cancer. Increased levels of signaling can also drive neurodevelopmental disorders ().

Recently, there has been a rapid hike in the number of reports relating to the linkage between neurodevelopmental disorders and cancer. These include many excellent reviews on the various aspects of neurodevelopmental disorders, such as RASopathy including, e.g., neurofibromatosis type 1 (NF1), Noonan syndrome (NS), Costello syndrome (CS) [e.g., (–)], PIK3CA-related overgrowth spectrum (PROS) [e.g., (–)], ASD [e.g., (–)], cerebral palsy, and more [e.g., (–)], and one uniquely also covering genetically engineered mouse models to study cancer or congenital disorders (). The linkage between developmental signaling cascades and aggressive central nervous system (CNS) tumors has also been reviewed [e.g., (, )]. Much progress has been made and vast amounts of experimental and clinical data have accumulated. Yet, some key enigmatic questions are still waiting to be deciphered.

NEURODEVELOPMENTAL DISORDERS

Most of the disorders arise from changes in the development of the nervous system (). Transcriptomics of the developing brain revealed that autism risk genes are highly expressed in prefrontal and primary motor cortices, striatum, cerebellum, and medial dorsal nucleus of the thalamus, particularly in early and midfetal development. Connective tissue dysplasia is common in NF1 microdeletions with a potential for associated cardiac manifestation. NF1 patients have higher chances of CNS tumors (gliomas and astrocytomas), neurofibrosarcomas, and leukemias than the normal population, suggesting dysregulated expression in these tissues (). Cerebral palsy is associated with connective tissue in the gastrocnemius muscle (plural gastrocnemii), a superficial two-headed muscle that is in the back part of the lower leg (). Germline mutations in Ras signaling pathways that relate to different familial, developmental syndromes frequently share phenotypic cardiofaciocutaneous features. Proteins involved in neurodevelopmental disorders relate to common pathways, including synaptic plasticity/function, chromatin remodelers, and the phosphoinositide 3-kinase (PI3K)/protein kinase B (AKT)/mechanistic target of rapamycin (mTOR) () and MAPK pathways, revealing RNA enrichment of genes involved in neuronal development, chromatin dynamics, and cell cycle regulation. RASopathies include NF1, NS, NSML (NS with multiple lentigines, formerly known as LEOPARD syndrome), CS, LS (Legius syndrome), CFC (cardiofaciocutaneous) syndrome, CM-AVM (capillary malformation–arteriovenous malformation) syndrome, and autosomal dominant intellectual disability type 5. They also present an increased risk of cancer (). The mutations occur in Ras proteins, their effectors, e.g., B-Raf, regulators, e.g., guanine nucleotide exchange factors (GEFs), guanosine triphosphatase (GTPase)–activating proteins (GAPs), associated scaffolding proteins, phosphatases, e.g., Src homology 2 (SH2) domain–containing phosphatase 2 (SHP2), ubiquitin ligases, and pathway inhibitors. They are also observed in epidermal growth factor receptor (EGFR) (), extracellular signal–regulated kinase (ERK) (), and other members of the Ras superfamily, e.g., Rho in cerebral palsy (–), Rac3 (), RalA (), and Rab (), and their regulators (), e.g., RhoGAP in schizophrenia risk (–). The disorders emerge from germline mutations during embryonic development, with their timing and location critically influencing the resulting disease phenotype. Mutations associated with neurodevelopmental disorders also include oncogenic proteins in other pathways, such as overgrowth disorder involving PI3Kα (), autism, a neurological developmental disorder involving phosphatase and tensin homolog (PTEN) (), and recently observed in cyclin-dependent kinases (CDKs) and their activating cyclin subunits in the cell cycle (). Structural studies of the proteins suggest that the conformational and dynamic changes promoted by the mutations, including at the protein-protein interaction sites, can be relatively minor and apparently inconsequential as in case of Ras, or substantial as in EGFR, where the extracellular domain becomes unstable, negatively affecting EGF binding (). The patterns of direct interactions of the mutated protein with other proteins may, as in SHP2 (Fig. 1) (, ), or may not be affected by the mutations, leaving the seemingly baffling question of same protein with—mostly different but also the same—mutations, different disease () unanswered. Why patients with neurodevelopmental disorders have higher chances of developing cancer has also been enigmatic. Here, we focus on these questions.

Fig. 1.

SHP2-associated Noonan syndrome in RASopathy.

(Top) Domain structure of SHP2 and its activation by disease-associated mutations. SHP2 contains nSH2 and cSH2 domains in the N-terminal region, and a phosphatase [protein tyrosine phosphatase (PTP)] and an unstructured tail in the C-terminal region. The C-terminal tail contains two tyrosine phosphorylation sites (Tyr542 and Tyr580) and a proline-rich domain with the PxxP motif. Upon upstream stimulation, the phosphorylated tyrosine (pY) motif of receptor tyrosine kinase (RTK) targets the SH2 domains of SHP2, recruiting it to the plasma membrane and releasing the autoinhibition of SHP2. Adapter proteins, such as Grb2-associated binding protein 1 (GAB1), also provide the pY motif, leading to recruitment of SHP2. Inactive SHP2 exhibits low basal activity owing to the autoinhibitory interaction between the nSH2 and PTP domains. Mutations in SHP2 causing human diseases, such as cancers or neurodevelopmental disorders, lead to gain of function. (Middle) These mutations mainly cluster in the nSH2-PTP interface, abolishing the inhibitory interaction. The locations and types of the germline mutations related to the NS and LEOPARD (LPRD) syndrome (also known as NS with multiple lentigines) are not the same as those responsible for the cancers, juvenile myelomonocytic leukemia (JMML), and acute myeloid leukemia (AML), although the same residues are mutated. (Bottom left) Inactive SHP2 does not contribute to the MAPK pathway. (Bottom right) Active SHP2 mutants with open conformation can bind to RTK or GAB1, promoting catalytic activity. SHP2 dephosphorylates the Ras GTPase-activating protein (RasGAP) binding sites in RTK, preventing RasGAP recruitment to the plasma membrane (, ). This negative regulation of RasGAP leads to increase of Ras activity, and thus MAPK signaling. NS and leukemia-associated SHP2 mutants increase MAPK signaling, while LPRD syndrome–associated mutants decrease PTP activity, with both enzymatically activating and inactivating mutations in PTPN11 unexpectedly resulting in neurodevelopmental disorders with overlapping clinical phenotypes (). PDB, Protein Data Bank.

CANCER EVOLUTION AND NEURODEVELOPMENTAL PATHOLOGIES

Cancer is expressed by uncontrolled cell proliferation. Cancer evolution requires (i) that the cell type–specific expression level of the associated protein be above a certain threshold, and the additive contributions of the oncogenic mutations be potent, together leading to a high population of activated mutant molecules, and (ii) that the activation signal propagates downstream. This requires that the cell type–specific expression levels of all nodes (proteins) in the pathway should not be low. If low, signaling in that pathway would not go through and another protein from a parallel pathway might be recruited (), leading to network rewiring. Neurodevelopmental pathologies require (i) a high expression level of the protein in the specific embryonic brain cell type () and potent mutations, single or in combination, and (ii) propagation of the signal downstream (Fig. 2). For this, all nodes in the pathway should be sufficiently highly expressed in that cell. The expression levels of the proteins are driven by chromatin accessibility and signaling in the specific cell (). Pathological and nonpathological brain cell type–specific profiles of the mRNA are critical (), particularly in the striatum, an essential relay of motor, cognitive, and limbic processes, suggesting highly specific roles in cell subpopulations (). This scenario is supported by analysis of the mRNA of brain cells, which observed gene expression changes linked to ASD (, ) and other neurodevelopmental disorders, such as bipolar disorders (). The phenotypes were associated with neurons in the upper layers of the cortex and microglial activation genes, and the developmental transcription factor genes were most affected (). It is also supported by SHP2-associated NS in RASopathy (Fig. 1), where binding partners mediating Ras signaling are abundant in excitatory but not inhibitory neurons, also pointing to cell type–specific changes in gene expression and the signaling network (). Transcriptional dysregulation of trophic signaling pathways was documented in cerebral palsy patient-derived cell lines from an unselected cohort of 182 affected individuals (). High expression levels require chromatin accessibility of the associated genes (). The involvement of chromatin remodeling is supported by improved viability by histone deacetylase inhibitors in most RASopathy cell lines as compared to several MAPK inhibitors (). The mutations should be sufficiently strong to obtain a high population of the activated proteins. Mutation strength relates to its mechanism of protein activation (–). In cancer, hyperactivation leads to cell proliferation; extremely vigorous activation signal can lead to oncogene-induced senescence (OIS). In development, it can lead to premature senescence (Fig. 2) (). Cells with aborted cell cycle can promote developmental disorders.

Fig. 2.

The cellular network and cell cycle.

Schematic diagrams of the networks and cell cycles of a wild-type cell, a cancer cell, and a mutant embryonic brain cell. Constitutive potent activation of the proteins is via single or combination of mutations (red circle and explosion shape). Mutations in proteins associated with the transcription machinery and chromatin remodeling can lead to high expression levels of these (red circle) proteins in the type-specific cell. The level of expression of the protein, coupled with the potency of activation by the mutations, results in a large number of active molecules, leading to cancer or neurodevelopmental disorder phenotypes, respectively. The time window—germline (in neurodevelopmental disorders) or somatic (in cancer)—of the mutation and cell type (embryonic brain cell or dedifferentiated cell) distinguish between them. In addition, the expression levels of the nodes in the pathway that the signal flows through, controlled by the chromatin remodeling proteins, should not be low to permit the signal to transmit. In cancer, the resulting hyperactivated signaling can lead to OIS. In development, they can lead to premature senescence. In embryonic cells, cell cycle arrest due to the premature senescence alters phenotypic traits toward the developmental disorders, including RASopathies. Germline embryonic neurodevelopmental disorder mutations can increase the risk of cancer by coupling with emerging somatic mutations. G1, S, G2, and M are stages in the cell cycle. In both cancer and neurodevelopmental disorders, signaling levels are critical in determining the disease consequences, and these levels are the outcome of (i) mutation strengths, (ii) expression levels, (iii) cell type, and (iv) timing window.

Multiple proteins are involved in cell cycle progress. Perturbations of their expression levels in specific cell types coupled with the signal strength, as can be measured by the absolute numbers of molecules (products) that the respective mutant protein(s) activate and propagation of the signal in the pathway, make them challenging to understand and especially to forecast. The different proteins in cancer and OIS, and in premature senescence in neurodevelopmental disorders such as RASopathy, suggest rewired signaling.

OIS AND PREMATURE SENESCENCE

Precursor cells divide until they reach a fully differentiated state; differentiation arrests proliferation, permanently exiting from the division cycle (). Strong activity of CDK-cyclin in the G1 cell cycle phase supports the undifferentiated state; however, hyperactivated OIS can emerge as protection against genomic instability and DNA damage, blocking cell division cycles (, ). PI3K lipid kinase, which phosphorylates phosphatidylinositol 4,5-bisphosphate (PIP2) to phosphatidylinositol 3,4,5-trisphosphate (PIP3), a signaling lipid required for signal transduction via the PI3K/AKT/mTOR pathway, can provide an example (Fig. 3). The number of PIP3 molecules produced by PI3K reflects its expression level and the potency of the mutations. A very high number of PIP3 and a consequent vigorous signaling can lead to cell proliferation. However, too high numbers by hyperactivated PI3K—generated by multiple PI3K mutations, or with collaborating mutant proteins—can lead to short signaling duration due to OIS. Mutations in both PI3K and PTEN, a tumor suppressor that dephosphorylates PIP3 to PIP2, can result in extremely high levels of PIP3 and such an outcome.

Fig. 3.

Mutations and expression levels of PI3K and PTEN in different cell lines.

PI3K and PTEN signaling in cancer (top) and PI3K (middle) and PTEN (bottom) signaling in neurodevelopmental disorders. Oncogenic and neurodevelopmental driver mutations powerfully activate PI3K, increasing the production of signaling lipid PIP3. PTEN dephosphorylates PIP3 back to PIP2, impairing its tumor suppressor activity and leading to the same outcome. High expression levels of PI3K with embryonic postzygotic mutations in specific brain (nerve) cells can lead to PROS, which encompasses a range of unique clinical entities, but with a continuum and overlapping diagnoses (). High expression levels of PTEN with embryonic germline mutations in different brain cell types can lead to ASD. The tumor suppressor mutants can present the PTEN hamartoma tumor syndrome (PHTS), with variable phenotypes, such as those diagnosed with autism or cancer.

OIS is a robust physiological antitumor response. Developmental senescence commonly takes place in aging and can also be recruited as a contributor to neurodevelopmental disorders (, , , ). A disorder regime requires high expression level of the mutant protein in the specific neuron cell and that the expression of nodes in the signaling pathway not be too low. If too low, or the level of a node in a parallel pathway is high, cell type–specific pathway branching or rewiring can take place (). An activated ERK can regulate targets in the cytosol or translocate to the nucleus to phosphorylate a range of transcription factors altering gene expression. SHP2 branching of the Ras signaling pathway involves proteins that are populated in excitatory but not inhibitory neurons, rewiring the pathway that can result in the NS phenotype (Fig. 1) ().

Recruited downstream factors can rewire the network. In contrast to replicative senescence (RS) and OIS, developmentally programmed senescence occurs with undetectable γH2AX (phosphorylated histone H2AX) and p53 signaling. OIS is mostly mediated by the higher expression of p16INK4A (CDK inhibitor 2A), p19ARF (ARF tumor suppressor, mice; p14ARF, human), p21WAF1 (CDK inhibitor 1 or CDK-interacting protein 1, p21CIP1), and p53 (). Senescence during embryonic development does not appear to be dependent on p16INK4A, but on p21WAF1, p15INK4B (CDK4 inhibitor B), and TGF-β (transforming growth factor–β)/SMAD (mothers against decapentaplegic homolog) and PI3K/FOXO (forkhead box protein O) pathways, suggesting modified expression levels and resultant rewiring (, –). The altered patterns of expression associated with senescence are linked to changes in the genome architecture. In OIS, the heterochromatin undergoes extensive three-dimensional (3D) genome reorganization, resulting in active genes that are positioned in regions bordering senescence-associated domains to be in spatial proximity, promoting their expression (). In neural stem cells (NSCs), all genes linked to cancer can affect proliferation; however, only those that are expressed during a certain time window may elicit neuronal developmental disorders (). Overexpression of cyclin D1 and CDK4 in NSCs can shorten the G1 phase and delay neurogenesis (, , ). Changes in expression levels and consequent rewiring are mediated by expression levels of cell type–specific transcription factor isoforms, generated by, e.g., alternative splicing (), or by different factors. Data support the involvement of transcription factors in rewiring [e.g., (, )].

Recent data have provided mechanistic level insight into the role of p21 in developmental senescence (). The inhibition of NSC senescence underlies corticogenesis—that is, the formation of the cerebral cortex, which is the outer layer of the brain, during the development of the nervous system—and neurogenesis, the formation of new neurons that later differentiate into specific types of neurons. The receptor for activated C kinase (RACK1) is abundant in NSCs. Depletion of RACK1 as the cerebral cortex is forming during the development of the nervous system can lead to developmental senescence in NSCs; its presence prevents senescence during corticogenesis. p21WAF1 and TGF-β/SMAD are among the critical mediators of developmental senescence during embryonic development (). RACK1 interacts with SMAD, blocking the TGF-β/SMAD pathway, which suppresses p21 expression and cellular senescence.

p19ARF and p16INK4A are critical mediators of cellular senescence (, ). Recent data suggest that p19ARF can also be among the proteins taking part in senescence during embryonic development, albeit through a different scenario (). Valproic acid induces p19ARF-mediated cellular senescence causing neurodevelopmental defects including cognitive defects and autism. Valproic acid has been used to treat epilepsy and bipolar disorder, but not in pregnancy, as it can elicit birth defects, cognitive impairment, and autism. p19ARF-mediated cellular senescence and microcephaly was detected in valproic acid neurogenesis-defective neuroepithelial tissues. Considering that p21WAF1 acts in both OIS () and developmental senescence, the expression levels and consequent network wiring of senescence mediators may control both developmental senescence and OIS.

CHROMATIN REMODELING IN CANCER AND IN NEURODEVELOPMENTAL DISORDERS

Proteins with mutations associated with both cancer and neurodevelopmental disorders occur broadly, commonly in those associated with expression. Leucine-zipper transcription regulator 1 (LZTR1), a highly mutated tumor suppressor gene involved in several cancer types and developmental disorders, is one example (, ). Cohesin, a chromosome-associated multi-subunit protein complex (), is another, possibly due to chromatin association. Chromatin regulators appear to be among the most frequently mutated genes in individuals with autism, and single-cell transcriptomics supports a role of CHD8, a DNA helicase that acts as a chromatin remodeling factor and regulates transcription in autism (, ). Genome-wide analyses pointed to candidate genes encoding nuclear factors implicated in chromatin remodeling. These include histone demethylation, and the recognition of DNA methylation (, ) and promoter architecture reorganization during neuronal cell differentiation implicates target genes for neurodevelopmental disorders (). Remodeling can affect regulatory regions of distal genes via long-range chromatin interactions, in a cell type–, developmental stage–, and disease-specific manner (). Variants of mediator MED12 subunit, with a central role in RNA polymerase II transcription and regulation of cell growth, development, and differentiation, have also been linked to neurodevelopmental disorders (). Last, Pol III mutations linked to neurodevelopmental disorder defects were identified and up-regulation of Pol III transcription was observed in cancer (). Cancer and neurodevelopmental disorders are linked, with an increased cancer risk among patients with certain developmental syndromes, and cancer-driving genes making up more than a third of the risk genes for developmental disorders (). As in cancer, unless the expression level is very high, single mutations are unlikely to be sufficient for neurodevelopmental disorders. A single cancer driver confers a selective growth advantage during cancer development but is insufficient for cancer initiation or maintenance (, ). This may be the case for developmental disorders as well (, ). Thus, preexisting germline mutations may couple with emerging sporadic mutations and high expression, resulting in higher incidence of cancer in neurodevelopmental disorders. Chromatin may also be involved in the apparent absence (e.g., in PIK3CA) or scarcity (e.g., in Ras) of germline driver hotspots in humans. The relative sparseness of highly transforming driver mutations in the germline has been attributed to embryonic lethality resulting from severe phenotype that includes cardiomegaly and abnormal brain development (). However, single mutations may be insufficient for emergence of cancer. These mutant alleles could couple either with mutations in chromatin remodelers that are likely to be rare in embryonic cells or, more likely, with high expression levels during transient exposure of the corresponding genes taking place during chromatin reorganization in the course of embryo development. Here, we focus on proteins in the major Ras signaling pathways, MAPK and PI3K/AKT/mTOR.

DEVELOPMENTAL DISORDERS MAY OR MAY NOT INVOLVE THE SAME MUTATIONS AS CANCER

Exactly how to link same-gene mutations to neurodevelopmental disorders versus cancer has been unclear. Some publications focused on developmental syndromes and the cell cycle, whereas others explored the mutational properties and performed network analysis (, ). A decade-old attractive hypothesis argued that the primary feature distinguishing between the two types of mutations is signaling strength and duration (). If strong and short, it can elicit developmental syndromes; if weak and sustained, the likely outcome is cancer. This hypothesis has been tested by measuring signal strength, for example, phosphorylation levels for MAPK-associated RASopathies. However, the linkage between a mutation and signal strength and duration, and the differential clinical outcomes have been elusive. Signal strength and duration are also unlikely to be independent parameters. Strength may tune duration via senescence.

Mutations in RAS and related pathway genes such as NF1 can lead to cancer and developmental disorders (). Apart from CS, RASopathy-associated mutations involve germline variants, whereas cancer emerges mostly from somatic mutations. KRas and NRas germline RASopathy variants are rarely at G12, G13, and Q61, which are frequent in cancers. KRas mutations are especially frequent at G12, particularly in pancreatic ductal adenocarcinoma (PDAC), colorectal adenocarcinoma (CRC), and non–small cell lung cancer (NSCLC) (, ). G13X, A146T, and K117N are common in CRC (). G12/13 block GAP-catalyzed and intrinsic hydrolysis. A146T enhances nucleotide exchange, with a 12-fold higher guanosine diphosphate (GDP) dissociation rate than wild-type KRas. K117N increases guanosine triphosphate (GTP) binding up to five- to sixfold (). On the other hand, germline KRas mutations were identified at, e.g., K5, V14, Q22, P34, I36, T58, G60, V152, D153, and F156 (Fig. 4). They are associated with NS and CFC syndrome (). Most increase the levels of active KRas, but with a milder phenotype than the oncogenic mutations (). Only a moderate increase in the phosphorylation levels of downstream signaling proteins was observed.

Fig. 4.

Ras mutations and their associated diseases.

(Left) Locations of the somatic mutations of Ras proteins in cancer. The somatic mutations can be observed in all Ras isoforms including HRas, NRas, and KRas, but expressed with different strengths. For example, G12X is populated in KRas and HRas, while Q61X is populated in NRas. (Right) Locations of the embryonic mutations of KRas in RASopathy. The white cartoons represent the catalytic domain of Ras protein. The locations and types of the mutations with associated diseases are highlighted in the boxes.

In a B-Raf example, phosphoproteomic analyses showed mutations in septo-optic dysplasia (SOD) including hypopituitarism and CFC syndrome. Activation of the MAPK pathway by B-RafV600E, or B-RafQ241R, leads to abnormal cell lineage determination and terminal differentiation of hormone-producing cells, causing hypopituitarism. Expression of the B-RafV600E in embryonic pituitary progenitors promotes expression of cell cycle inhibitors, cell growth arrest, and apoptosis, but not cancer (). B-RafV600E is also a common strong driver of cancer. Thus, there appears to be no rule: Different diseases may arise from different or the same mutations. The mutations are not differentiated by their strengths. Neither type seems to be consistently stronger than the other. Computations observed a trend that energy changes upon residue substitutions are higher for cancer compared to RASopathy mutations ().

PIK3CA, encoding the catalytic subunit of PI3Kα lipid kinase in the PI3K/AKT/mTOR pathway, provides an example for another developmental disorder (Fig. 3). It is the second (third) most highly mutated protein in cancer. Its mutations are also observed in benign PROS (). Although PROS are not inherited germline line mutations, they emerge during embryonic development, with timing and location (skin, vasculature, bones, fat, and brain tissue) depending on the specific disease and determining the disease phenotype. Germline mutations in PIK3CD can lead to the activated PI3Kδ syndrome (APDS), with neurodevelopmental abnormalities such as speech and developmental retardation (). APDS, particularly APDS2, also increases the risk of Hodgkin’s and non-Hodgkin’s lymphomas. PI3Kδ isoform mostly expresses in lymphocytes. APDS can also promote T cell senescence and a rare primary immunodeficiency disorder (). PIK3CD mutations elevate basal activity through enhanced membrane binding of p110δ.

Somatic mutations in PTEN occur in many cancers, most frequently in prostate, and endometrial cancers, as well as melanoma. They are also common in brain glioblastomas and astrocytomas. Preexisting germline autism-associated mutations could couple with somatic mutations contributing to the evolution of brain cancers ().

As observed in RASopathies, mutations can express unique developmental syndromes in specific cell types; however, there can also be certain overlapping phenotypic features, including craniofacial abnormalities, congenital heart defects, and neurocognitive delay (). These may arise from contributions from other cell types if the expression levels exceed the threshold and expression of downstream nodes is at sufficiently high levels for the potent signaling to go through.

THE MECHANISM OF ACTIVATION AND SIGNALING STRENGTH

Despite the absence of a clear pattern of relative strength, the mutation should be sufficiently strong for a high percentage of the molecules to be activated. Molecular activation mechanisms can be a key factor in deciding signal strength and disease outcome. One example is KRasG12D predominating in pancreatic cancer but not KRasA146T, which is almost exclusively in cancers of the intestinal tract particularly colon, and blood (). The overall frequency of the G12D mutation is 23.4%, and that of the G12V mutation is 21.1% (); that of A146T is approximately 4% in colorectal cancers (). The statistics correlates with the strength. Prognosis of KRasG12V mutant tumors is poorer than that of other subtypes (), and in metastatic pancreatic cancer, it is correlated with poor overall survival (). Tissue specificity and varied strengths point to differences in signaling downstream (). Ras is activated by exchange of GDP by GTP; it is deactivated by GTP hydrolysis. KRasG12X mutations interfere with GTP hydrolysis, retaining KRas in an active, GTP-bound state. On the other hand, KRasA146T promotes an extension of Switch I away from the KRas catalytic domain and the nucleotide binding site, and nucleotide exchange. The differences in the strengths of the mutations derive from the location of the mutations and their consequent activation mechanism. GAP-induced hydrolysis is impaired by Gly12 but not by A146T mutations, which act by accelerating nucleotide exchange. On the other hand, mutations in tumor suppressor PTEN abolish its activation mechanism at the membrane ().

Mutations in kinases and phosphatases may lead to divergent phenotypes by phosphorylating/dephosphorylating different substrates in different tissue types (). Recent data suggest that HRasG12V is involved in hypotonia in CS. A mouse model showed an inverse correlation of Ras/MAPK and p38, which is essential for skeletal muscle growth and homeostasis through the dual-specificity phosphatase (DUSP). DUSP6 dephosphorylates ERK1/2, as well as p38 in the presence of pERK2, with DUSP6, pERK2, and p-p38 forming a stable complex.

Recently, SHP2 liquid-liquid phase separation (LLPS) helped address the perplexing question of why both enzymatically activating and inactivating mutations in PTPN11 result in developmental disorder with overlapping clinical manifestations (). The authors showed that the inactivating mutation of SHP2 activates wild-type SHP2 through inducing LLPS, thereby accounting for the activation mechanism of wild-type SHP2. The activation of SHP2 is required for some receptor tyrosine kinase (RTK) signaling pathways ().

In addition to the location, the biophysical and biochemical attributes of the mutations in cancer and neurodevelopmental disorders matter. They can alter the stabilities, conformational distributions, dynamics, and interactions of the respective protein. SHP2 is an example of the potential therapeutic consequences of the conformational properties of the mutations in specific proteins (); Ras is an example of the mutational properties and activation mechanism (). At the same time, the observations that even the same mutations in the same proteins can lead to different phenotypic disease manifestations, and the lack of consistency of the mutational strengths in cancer versus neurodevelopmental disorders, argue that mutational properties per se are unable to resolve the enigma of their different clinical outcomes.

CONCLUDING REMARKS: CANCER, NEURODEVELOPMENTAL DISORDERS, PHARMACOLOGY, AND THE ULTIMATE AIM OF MEASUREMENTS

Cancer depends on the number of the corresponding activated proteins, which is determined by their expression levels and oncogenic mutations. Neurodevelopmental disorders embody the same principles. The mutations can occur in the same proteins and may or may not overlap. No uniform pattern is observed in their relative strengths. In both, propagation of a potent signal requires large-enough numbers of available active molecules in each node in the respective pathways. These numbers depend on protein expression levels, which rest on the 3D structures of the genome and chromatin accessibility of the corresponding genes, which vary across cell subtypes and timing (embryonic development or differentiated) windows. If the number of molecules in a specific node is too low, pathway rewiring through involvement of transcription factor isoforms (or other factors) can take place, resulting in premature developmental senescence. One way for this blueprint to be tested is by the mRNA levels of the nodes in the specific cell ().

From the pharmacological standpoint, deciphering the same gene mutations but different disease enigma is vastly important. Recent evidence indicates that PI3K pathway inhibitors undergoing trials in cancer can provide a therapy for PROS. Since both mutation types are in the same protein, this can be understood. However, considering the dependence of signaling on expression of every single node in the pathway and rewiring, combinatorial drug regimens adopted in cancer, such as canonical Ras pathway inhibitors, may or may not work in neurodevelopmental disorders. Network analysis suggested that RASopathy proteins participate in 33 pathways (). Inclusion of drugs acting on chromatin modifiers, such as statins and histone deacetylase inhibitors, may have a better chance (); however, contending with toxicity remains a challenge (). Last, with the advent of the impact of artificial intelligence (AI), the question has been raised as to whether AI can offer a breakthrough in predicting risks and prognosis for neurodevelopmental disorders () and drug repurposing (). Within this framework, the challenging aim is an integrative systems biology platform that merges in vitro and tissue-specific in vivo models, allelic series, transcriptomics and proteomics, and patient samples to start to measure the differences and gauge the effects of these on neurodevelopmental disorders and cancer phenotypes and on therapeutics.

Within this ambitious framework, the first hurdle is measurement of the mutational activation strength and signal potency. We consider it in terms of signaling by the numbers, that is, aiming to measure the absolute number of activated molecules, and asking what thresholds should be considered for a signal to propagate through the pathway, reach the cell cycle, and activate (repress) a transcription factor. In line with our discussion here, the strength needs to account for the mutation potency and cell type–specific transcription level. The sparse measurements of the single-cell mRNA levels, especially across time windows during development and consequent likely dropout events, transcription bursts, and noisy data, are big hurdles to overcome. Deep learning, low-dimensional representation for scRNA-seq (single-cell RNA sequencing) data visualization and analysis is expected to assist ().

The premise of measurements of signals induced by driver hotspots (and drug resistance mutations in cancer), leading to signal transduction from an upstream to a downstream node, to activate (or repress) function in a specific cell at a specific developmental state, with respect to an established threshold set by experiments, and exploited in deep learning predictions, is a powerful aim. If (or when) attained, it could provide a compelling starting point for biologists to investigate.