Contextual signaling in cancer

Abstract

The formation and maintenance of an organism is highly dependent on the orderly control of cell growth, differentiation, death, and migration. These processes are tightly regulated by signaling cascades in which a limited number of molecules dictate these cellular events. While these signaling pathways are highly conserved across species and cell types, the functional outcomes that result from their engagement are specified by the context in which they are activated. Using the Neurofibromatosis type-1 (NF1) cancer predisposition syndrome as an illustrative platform, we discuss how NF1/RAS signaling can create functional diversity at multiple levels (molecular, cellular, tissue, and genetic/genomic). As such, the ability of related molecules (e.g., K-RAS, H-RAS) to activate distinct effectors, as well as cell type- and tissue-specific differences in molecular composition and effector engagement, generate numerous unique functional effects. These variations, coupled with a multitude of extracellular cues and genomic/genetic changes that each modify the innate signaling properties of the cell, enable precise control of cellular physiology in both health and disease. Understanding these contextual influences is important when trying to dissect the underlying pathogenic mechanisms of cancer relevant to molecularly-targeted therapeutics.

Graphical abstract

1. Introduction

“The only man I know who behaves sensibly is my tailor; he takes my measurements anew each time he sees me. The rest go on with their old measurements and expect me to fit them.”

- George Bernard Shaw

Precise control of cell behavior involves continual sensing of a variety of physical and chemical cues from the external milieu, and interpreting those signals to respond in an appropriate manner. These extracellular cues take the form of cell-bound and diffusible ligands that bind specific receptors on cells to initiate a cascade of signaling events that culminate in changes in protein function through post-translational (e.g., phosphorylation) or transcriptional (e.g., mRNA or miRNA) alterations. In this regard, signaling represents the language of the cell, where molecules (words) and cellular context (syntax) serve as units of informational content. Creating meaningful sentences and paragraphs requires that a limited number of molecules can be used in combinatorial and situation-specific manners to produce a diversity of outputs and responses relevant to the specific environment at that particular time. Unfortunately, when we study signaling pathways in normal cells or in the setting of cancer, we often fail to consider how the cellular language conferred by these pathways is influenced by context, that is, the different extracellular signals present in the immediate milieu, the various adaptive responses that limit and promote intracellular signal transduction, the innate properties of distinct cell types responding to these cues, and the impact of epigenetic/genomic changes on the ultimate consequence of these informational signals. In this review, we use the Neurofibromatosis type 1 (NF1) predisposition syndrome as an illustrative platform to discuss how heterogeneity can be generated at the molecular, cellular, tissue, and genomic/genetic levels. Moreover, we suggest that the precision created by this “contextual signaling” enables diverse outcomes to arise from engagement of a limited number of key signaling pathways.

1.1 Neurofibromatosis type 1

Neurofibromatosis type 1 (NF1) is a monogenic syndrome affecting 1 in every 2,500 individuals worldwide. Caused by a germline mutation in the NF1 gene, affected children and adults are prone to the development of benign and malignant tumors. In addition, numerous other organ systems are affected, leading to skeletal, cardiovascular, dermatologic, ophthalmologic, endocrinologic, and neurological abnormalities. This latter group of clinical features is the most common, and includes peripheral nerve tumors (neurofibromas and malignant peripheral nerve sheath tumors) and brain tumors (optic pathway and brainstem gliomas, malignant glioma).

Neurofibromas and optic pathway gliomas (OPGs) form following bi-allelic NF1 loss, which is the consequence of somatic inactivation of the remaining normal NF1 gene and total loss of NF1 protein (neurofibromin) expression. Individuals with NF1 start life with a germline NF1 gene mutation, which increases their chance of cancer, since only one additional genetic event is required (somatic NF1 gene loss). As neurofibromin functions as a GTPase-activating protein (GAP) to accelerate the conversion of active GTP-bound rat sarcoma (RAS) to inactive RAS-GDP, neurofibromin loss in neoplastic cells leads to RAS hyperactivation and engagement of multiple RAS downstream effectors, culminating in increased cell growth. In addition, tumorigenesis is further controlled by non-neoplastic cells in the local microenvironment that elaborate key stromal growth factors and chemokines. Moreover, the impact of NF1 loss is highly dependent on the specific cell type and tissue, as well as by modifying genomic loci and cooperating genetic mutations. The influence of all of these factors may partly explain why individuals with NF1 exhibit such a wide range of clinical variability. In this regard, children and adults from the same family with an identical germline NF1 gene mutation can manifest different medical features of NF1 and have markedly distinct clinical courses and severity. Understanding these factors is vitally important for providing risk assessments to affected individuals, as well as for designing personalized therapies for these clinical abnormalities when they develop.

2. Molecular Level

While neurofibromin functions as a RAS-GAP, the consequence of bi-allelic NF1 loss on signaling pathway activation and cell growth is not identical in all cell types. This reflects the fact that a signaling molecule (e.g., RAS) is not a single molecule, but rather comprises a family of highly homologous proteins with slightly different functions. Moreover, RAS can signal by engaging a variety of downstream effector proteins to control cell biology. Lastly, some signaling effector proteins function as part of a multi-molecular complex whose composition determines which signaling intermediates are activated and what cellular output is controlled (Figure 1).

Figure 1.

Neurofibromin loss activates RAS signaling to potentially activate a large number of downstream effector proteins.

2.1 Signaling molecules comprise a group of functionally-distinct proteins

One mechanism by which RAS pathway activation can generate distinct cellular responses is through differential engagement of RAS/RAS effector proteins. In this regard, RAS, rapidly accelerated fibrosarcoma (RAF) and protein kinase B (AKT) each comprise families of molecularly-similar, but functionally-distinct, proteins.

RAS exists as at least four highly homologous RAS proteins (H-RAS, N-RAS, K-RAS4A, K-RAS4B), which differ in their potency to activate downstream effectors [1]. While these molecules share 85% amino acid sequence similarity, the specific functions of each RAS protein are dictated by distinct 25-residue hypervariable regions (HVR) at the carboxyl terminus. Differential lipid modifications within this HVR direct H-RAS and N-RAS to the plasma membrane, while K-RAS is trafficked to different domains within the plasma membrane [2, 3]. In addition, oncogenic K-RAS activation (Noonan syndrome) increases cellular proliferation and differentiation, while hyperactive N-RAS (Noonan syndrome) promotes cell survival [4]. Similarly, mutational H-RAS activation (Costello syndrome) stimulates tumor angiogenesis [5] and DNA synthesis [6], while mutational K-RAS activation has no effect [7]. Moreover, homozygous inactivation of K- RAS is embryonically lethal, whereas N-RAS or H-RAS knockout mice are viable [8, 9]. Lastly, only mutationally-activated K-RAS, but not H-RAS or N-RAS, increased neural stem cell proliferation in vitro and astroglial differentiation in vitro and in vivo [10].

One of the direct targets of RAS is the RAF serine/threonine-specific protein kinase, which, like RAS, comprises three distinct molecules (A-RAF, B-RAF, C-RAF [RAF-1]). These proteins are differentially expressed and regulated, and are non-redundant in their ability to activate their downstream effectors. RAF molecules, though structurally similar, regulate MEK1/2 with varied affinity through the formation of multiple RAF heterodimers. In this regard, the B-RAF/C-RAF heterodimer stimulates MEK much more efficiently than B-RAF or C-RAF activation alone [11]. Additionally, B-RAF and C-RAF require additional post-translational modifications for full activity [12, 13]. Moreover, this family of protein kinases interacts with numerous adapters, kinases, G-proteins, and chaperones to create signaling diversity [14].

Similarly, AKT is encoded by three separate genes, PKBα (AKT1) PKBβ (AKT2) and PKBγ (AKT3), whose activation also has different consequences. As such, Akt1 knockout mice display defects in overall growth, while Akt2 null mice are insulin intolerant and demonstrate a diabetes-like syndrome [15]. In sharp contrast, Akt3 knockout mice have a selective reduction in brain size, reflecting its robust expression in brain tissues [16].

2.2 Individual signaling molecules activate different effectors

GTP-bound RAS can directly activate at least three different proteins (RAF, phosphoinositide 3-kinase [PI3K], and protein kinase C-zeta [PKCζ]), which transduce growth-promoting messages through distinct signaling pathways [17–19].

RAS-dependent RAF activation leads to the sequential phosphorylation of mitogen-activated protein kinase kinase (MEK) and p44/p42 extracellular signal-related kinase (ERK) [20]. While neurofibromin loss and activation of MEK/ERK can lead to increased proliferation through ERK activation of transcriptional factors [21], it can also increase cell proliferation through a 90 kDa ribosomal S6 kinase (p90-RSK)/mechanistic target of rapamycin (mTOR)-dependent manner [22]. In addition, neurofibromin controls glial and neuronal differentiation in a RAF/MEK-dependent, but mTOR-independent, manner in brain neural stem cells (NSCs) by activating the Jagged1/Notch pathway [23]. Finally, oncogenic KRAS increases NSC growth by negatively regulating the retinoblastoma protein in a RAF-dependent, but MEK-independent, fashion [10].

As another RAS effector, PI3K signaling is required for a wide variety of critical cell processes [19, 24]. RAS activity increases PI3K activation, which allows phosphatidylinositol (4,5)-biphosphate (PIP₂) to convert into phosphatidylinositol (3,4,5)-triphosphate (PIP₃). PIP₃ then recruits protein kinase B (AKT) to the plasma membrane, promoting PI3K-mediated phosphoinositide dependent protein kinase-1 (PDPK1/PDK1) phosphorylation and activation of AKT. However, AKT-mediated mTOR activation can occur through a variety of mechanisms, including direct mTOR activation [25], phosphorylation of the proline-rich in AKT substrate of 40 kDa (PRAS40) [26], or through inactivation of the tuberous sclerosis complex (TSC) [27].

Lastly, RAS can transmit its growth promoting signaling through atypical Protein Kinase C molecules, like PKCζ, either involving PI3K [28] or independent of PI3K [29]. In neurons, reduced neurofibromin function leads to RAS-mediated PKCζ activation, resulting in G protein-coupled receptor kinase 2 (GRK2) suppression of G protein-coupled receptor function and reduced cyclic adenosine monophosphate (cAMP) production [17]. These distinct neurofibromin/RAS downstream pathways (RAF/MEK, PI3K/AKT, and PKCζ) create a diversity of signaling pathway activation following NF1 loss in specific cell types.

2.3 Individual molecules form protein complexes necessary for signaling transduction

Beyond RAS/RAS effector families and differential engagement of RAS downstream effectors, another mechanism for generating functional diversity is through the formation of multi-protein signaling complexes. Critical for neurofibromin/RAS growth control is the mTOR complex. mTOR operates in as at least two molecularly- and functionally-distinct protein complexes. mTOR complex 1 (mTORC1) consists of regulatory-associated protein of mTOR (Raptor) and PRAS40. Raptor is required for mTOR kinase activity by directly binding and activating mTORC1 effectors, such as the translational regulators p70 S6 kinase (S6K) and 4E (elF4E) binding protein (4EBP1) [30]. In contrast, mTOR complex 2 (mTORC2) contains rapamycin-insensitive companion of mTOR (Rictor), protein observed with rictor (Protor), and mammalian stress-activated MAP kinase interacting protein-1 (mSIN1), which are all required for the activation of mTORC2-specific effectors, AKT [31], protein kinase Cα,β, γ (PKCα, β, γ) [32], and serum- and glucocorticoid-induced protein kinase 1 (SGK1) [33]. While mTORC1 is critical for mediating ribosomal biogenesis and translational control, mTORC2 is essential for controlling cell survival, migration, and cytoskeletal dynamics [34].

3. Cellular Level

In addition to signaling diversity generated at the molecular level, there are cell type-specific constraints that operate to alter the way neurofibromin regulates RAS pathway activation and cell function. These include differences between cell types, as well as differences between the same cell type within a given tissue (Figure 2).

Figure 2.

The consequence of neurofibromin loss and RAS activation is dictated by cell type- and brain region-specific differences.

3.1 Cell type-specific control of RAS effector engagement

While neurofibromin suppresses RAS activity to control cell growth and survival in all cell types examined to date, how neurofibromin signals to its downstream effectors to mediate this effect depends entirely on the individual cell type. As such, central nervous system (CNS) neuronal growth and survival relies on RAS/PKCζ regulation of cAMP [17], rather than through MEK/ERK or PI3K [22]. In contrast, brain astroglial and NSCs use neurofibromin to control cell growth through the RAS/mTOR pathway [22, 35, 36]. However, Nf1-deficient astrocyte proliferation is dependent on mTOR/Rac1 activation [37], whereas NSCs use RAS/mTORC2/AKT/p27 signaling [36]. Additionally, neurofibromin control of mast cell function operates through RAS/PI3K/Rac1 [38], whereas Nf1-deficient osteoblast growth is dependent on RAS/ERK/RSK/activating transcription factor 4 (ATF4) signaling [39]. In microglia, neurofibromin controls cell proliferation and activation through the Rac1/mixed-lineage protein kinase 3 (MLK3)/mitogen-activated protein kinase kinase 4 (MKK4)/c-Jun N-terminal kinase (JNK) pathway [40]. In addition, the signature genetic alteration in low-grade gliomas involves the fusion of the BRAF gene to the amino terminal of the KIAA1549 protein product (f-BRAF). Similar to neurofibromin, ectopic f-BRAF expression increases RAS/MEK signaling in both astrocytes and NSCs, but only increases proliferation in NSCs and eventual tumor formation as a result of ERK-mediated mTOR activation [36, 41].

3.2 Regional constraints dictate RAS pathway function

Another contextual determinant that impacts RAS pathway signaling is the regional identity of the cell. In this respect, isolated astroglial cells have different levels of neurofibromin expression, with significantly higher Nf1 protein expression in the optic nerve, brainstem and cerebellar astrocytes relative to those from the neocortex [42]. For example, reduced neurofibromin expression in CNS neurons results in attenuated survival and neurite outgrowth, which depends on neurofibromin/RAS-controlled cAMP production [43], whereas, Nf1-deficient peripheral nervous system (PNS) neurons have increased survival and longer neuritic processes, reflecting RAS-mediated AKT hyperactivation [44]. Finally, Nf1 loss in NSCs has differential effects depending on the region of the brain in which they reside. For example, Nf1-deficient NSCs from the third ventricular and brainstem exhibit increased proliferation and gliogenesis, whereas those from the lateral ventricle or neocortex of the same mouse do not [36].

4. Tissue Level

While it is clear that cell type and tissue location can differentially influence RAS signaling and functional outcome, additional factors operate at the tissue level to affect RAS pathway signaling (Figure 3). In the case of low-grade tumors arising in children and adults with NF1 (neurofibromas, OPGs), evidence exists for an obligate role for cellular and acellular determinants in tumor formation and maintenance. While less is known about the role of acellular factors in NF1-associated tumors (extracellular matrix [ECM] components), expression of the ECM components, laminin-2α and integrin α6β1, potentiates high-grade glioma cell growth [45, 46].

Figure 3.

Non-neoplastic cells and acellular signaling in the local microenvironment control RAS-dependent neoplastic cell growth.

In experimental mouse models of plexiform neurofibroma, both macrophages and bone marrow-derived mast cells control tumor formation and growth through the activation of RAS signaling. Using a combination of bone marrow transplantation and pharmacological inhibitor approaches, mast cells are required for murine plexiform neurofibroma formation and continued growth [47]. In these experiments, mast cells recruited by Nf1-deficient Schwann cell-produced stem cell factor (SCF or KIT-ligand) secrete cytokines that increase neurofibroma growth [48]. Moreover, SCF can also stimulate Nf1+/− mast cells to increase fibroblast proliferation and collagen deposition, thereby forming a permissive tumor microenvironment [49]. Interruption of this paracrine circuit with the c-KIT inhibitor, Imatinib, reduced plexiform neurofibroma growth [47].

In human low-grade gliomas, resident brain tissue macrophages (microglia) comprise 30–50% of the cells [50], where they are critical for murine Nf1 OPG formation and progression [51]. As such, pharmacologic or genetic reduction of microglial function results in reduced tumor proliferation in vivo [40, 50, 52]. Since these tumor-associated microglia are the likely source of stromal chemokines and growth factors, one large-scale RNA-sequencing effort identified several potential candidates. The most promising molecule, the (C-C motif) ligand 5 (CCL5) chemokine, increased Nf1-deficient optic nerve astrocyte proliferation in vitro. Importantly, inhibition of CCL5 using a neutralizing antibody resulted in reduced tumor growth in vivo [53], which operates in a RAS/AKT-dependent fashion [54].

NF1-associated CNS/PNS tumors arise in close proximity to central and peripheral nerves, raising the intriguing possibility that neuronal activity might influence tumorigenesis or tumor growth. Elegant studies in high-grade glioma (HGG) model systems have shown that neuronal activity increases HGG growth and proliferation in a PI3K/mTOR-dependent manner [55]. As such, increased cortical neuronal activity, controlled optogenetically in a mouse xenograft model, resulted in greater glioma growth and proliferation in vivo. Using a proteomic strategy, several secreted factors were identified from stimulated cortical brain slides, culminating in the discovery of neuroligin-3 (NLGN3) as the responsible neuron-derived mitogen [55].

5. Genomic/Genetic Level

Beyond the influences of the individual signaling molecules in different cell types and the impact of factors from the local tissue microenvironment, another mechanism by which diversity can be created is at the level of genetic and genomic changes. In the context of NF1, these include the specific germline NF1 gene mutation, cooperating genetic mutations, and genomic modifier loci (Figure 4).

Figure 4.

Genetic and genomic factors differentially impact on neurofibromin/RAS signaling and tumor formation.

5.1. The Nf1 germline mutation

Every individual with NF1 is born with a germline mutation in the NF1 gene; however, >98% of individuals harbor a unique mutation. While there are few clear relationships between the specific germline NF1 gene mutation (genotype) and the clinical features of the disease (phenotype), converging evidence from epidemiologic, human NF1-patient iPSCs, and genetically-engineered mouse studies suggest that the specific germline NF1 gene mutation may be one predictive risk factor. As such, germline NF1 gene mutations have been reported in families who exhibit other hallmarks of the disease, but do not develop cutaneous neurofibromas (c2979-2972 delAAT and Arg1809 missense mutations) [56–58]. These early-phase observations are bolstered by studies employing NF1-patient derived iPSCs and derivative neural progenitor cells. In these experiments, different NF1 germline mutations have differential effects on neurofibromin expression and function, with some mutations leading to minor reductions in neurofibromin levels and others with >70% decreases [59]. Using genetically-engineered mice designed to harbor the R681X mutation observed in a patient with an OPG or a missense mutation (G848R) found in individuals with spinal neurofibromatosis, differential effects of these mutations were found. Mice with the R681X mutation had in >70% reductions in neurofibromin levels, whereas those with the G848R mutation had <25% reductions [54]. Importantly, only those mice with the R681X mutation developed OPGs.

5.2. Cooperating genetic mutations

In addition to NF1 gene inactivation, additional molecular changes have been identified in human NF1-associated OPG that converge on the same signaling pathway regulated by neurofibromin [60, 61]. For example, a heterozygous PTEN deletion was identified in one NF1-OPG, raising the possibility that these coincident genetic changes cooperate with NF1 loss to increase glioma growth and lead to clinically more aggressive neoplasms. Consistent with this idea, Nf1 genetically-engineered mice that also harbored a heterozygous PTEN mutation exhibited larger tumors with greater proliferation as a result of increased AKT activation [62].

5.3. Genomic modifier loci

Another way neurofibromin/RAS function can be modulated involves genomic modifier genes. While there is evidence for differences in racial and ethnic group risks for brain tumors [63], the most compelling data to support the existence of modifier loci derives from Nf1 genetically-engineered mouse models [64]. In these studies, NPcis mice, which carry mutations in the Nf1 and Trp53 genes on the same chromosome, develop HGG. However, NPcis mice on a C57BL/6 background developed brain tumors with high penetrance, whereas those on other genetic backgrounds do not. Further analyses of NPcis mice on different genetic backgrounds revealed the presence of modifier loci that influence spinal tumor development and brain astrocytoma formation [65], and high-grade peripheral nerve and brain tumor resistance in a sex-specific manner [66]. While it is not known how these modifiers operate to control tumorigenesis, one mechanism might involve differential Nf1 gene expression [67].

6. Conclusions

Creating functional diversity from one protein and/or pathway through contextual signaling allows for a great deal of outcome specificity without having to increase the amount of genetic material, and thus is evolutionarily efficient. Using NF1 as a model, we have outlined how RAS signaling can be contextually modified at the molecular, cellular, tissue, and genomic/genetic levels (Figure 5). In this regard, each specific genetic mutation, signaling effector family member, cell type, and tissue work in a combination with one another, as well as with genomic constraints, to encode a curated milieu that determines an explicit functional outcome and tumor pathology. As such, each RAS downstream effector can function in several states depending on the context in which it is activated.

Figure 5.

Contextual signaling in cancer operates at the genetic/genomic, molecular, cellular and tissue levels.

The assumption that canonical signaling pathways function in a linear and static manner across all cell types and tissues may not fully represent the manner in which they truly operate. As detailed in this review, converging evidence from several types of experiments suggest that understanding the most accurate context in which mitogenic signaling drives cell growth yields a clearer picture of the mechanisms underlying NF1 heterogeneity at the cellular, tissue, and organismal levels. A deeper appreciation of contextual signaling may improve our understanding of the basic principles that govern development and is also likely to lead to the design of more effective therapies for diseases characterized by inappropriate RAS/RAS pathway activation, such as seen in NF1-related tumors. Defining and encoding these variables relative to disease pathogenesis will hopefully result in better risk assessment strategies and the individualized therapies.

Highlights.

• Cellular responses are dictated by context-dependent signaling pathway engagement

• A limited number of molecules can activate multiple downstream effectors

• Specific effector pathways are activated in a cell type and tissue-specific manner

• Extracellular cues and genomic/genetic factors modify signaling pathway output

Abbreviations

AKT

protein kinase B

ATF4

activating transcription factor 4

cAMP

cyclic adenosine monophosphate

CCL5

(C-C motif) ligand 5

CNS

central nervous system

4EBP1

4E (elF4E) binding protein

ECM

extracellular matrix

ERK

p44/p42 extracellular signal-related kinase

GAP

GTPase-activating protein

GRK2

G protein-coupled receptor kinase 2

HGG

high grade glioma

HVR

hypervariable region

iPSCs

induced pluripotent stem cells

JNK

c-Jun N-terminal kinases

MEK

mitogen-activated protein kinase kinase

MKK4

mitogen-activated protein kinase kinase 4

MLK3

mixed-lineage protein kinase 3

mSIN1

mammalian stress-activated MAP kinase interacting protein-1

mTOR

mechanistic target of rapamycin

mTORC1

mTOR complex 1

mTORC2

mTOR complex 2

NF1

Neurofibromatosis type 1

NLGN3

neuroligin-3

NSCs

neural stem cells

OPG

optic pathway glioma

PDPK1/PDK1

phosphoinositide dependent protein kinase-1

PI3K

phosphoinositide 3-kinase

PIP₂

phosphatidylinositol (4,5)-biphosphate

PIP₃

phosphatidylinositol (3,4,5)-triphosphate

PKCα,β, γ

protein kinase Cα,β, γ

PKCζ

protein kinase C-zeta

PNS

peripheral nervous system

PRAS40

proline-rich in AKT substrate of 40 kDa

Protor

protein observed with rictor

PTEN

phosphate and tensin homolog

Rac1

ras-related C3 botulinum toxin substrate 1

RAF

rapidly accelerated fibrosarcoma

Raptor

regulatory-associated protein of mTOR

RAS

rat sarcoma

Rictor

rapamycin-insensitive companion of mTOR

p90-RSK

ribosomal S6 kinase

SCF

stem cell factor

S6K

p70 S6 kinase

SGK1

serum- and glucocorticoid-induced protein kinase 1

TSC

tuberous sclerosis complex