Introduction

The underlying basis of the cancer phenotype is deregulated cell growth, which stems from two main hallmarks of cancer: uncontrolled proliferation and loss of programmed cell death (enhanced survival). In normal cells, these processes are tightly controlled by the discrete integration of signaling cascades that translate extra- and intracellular cues into specific output responses. These signaling pathways are often initiated on binding of ligand to the extracellular domain of a receptor, followed by recruitment of adaptor proteins or kinases that activate an intracellular cascading network of protein and lipid intermediaries that produce a cellular response. The specificity, amplitude, and duration of signaling are critical for normal cellular function, and these constraints are often abrogated in human cancer.

Investigation of the signal transduction pathways that regulate normal cellular functions has revealed that key components of these networks are commonly altered in cancer cells by mutation, amplification/deletion, chromosomal translocation, overexpression, or epigenetic silencing. These alterations lead to constitutive activation or suppression of signaling that underlie the various hallmarks of the cancer phenotype. In this chapter, we will review the major signal transduction cascades, with a focus on the ones that are commonly altered in human cancers. Individual sections will highlight signaling intermediaries that have been validated as drug targets in patients with cancer.

Receptor Tyrosine Kinase Signaling

The receptor tyrosine kinases (RTKs) comprise a family of transmembrane cell surface receptors that transduce extracellular signals internally to promote growth or survival and/or regulate other cellular phenotypes.1,2 Members of this protein family share a similar modular domain structure. Growth factors bind to the extracellular ligand-binding domain of RTKs and induce dimerization of two receptor monomers, juxtaposing the intracellular tyrosine kinase domains of each monomer.³ This process results in transphosphorylation of tyrosine residues within the cytoplasmic domains of the RTK dimer. After transphosphorylation, a variety of intracellular proteins are recruited to the activated RTK through Src homology 2 (SH2) domains that recognize the phosphotyrosine plus a specific amino acid sequence motif C-terminal to the tyrosine residues.^4,5

More than 117 SH2 domains have been characterized, each with unique phosphotyrosine sequence specificities.⁶ Each domain is part of a larger adaptor protein involved in transducing extracellular signals to activate, or in some cases suppress, specific intracellular signaling cascades. Thus the complement of signaling pathways that a given RTK regulates is dictated by the profile of phosphorylated tyrosine residues plus flanking amino acids within their intracellular domains.^7,8 However, more than one adaptor protein can often recognize individual context-dependent phosphotyrosine motifs within an RTK, underscoring how this system is designed to provide both specificity and diversity of intracellular signaling.

Recruitment of signaling intermediaries to the plasma membrane facilitates their interaction with membrane-bound proteins responsible for stimulating a diverse array of downstream pathways (Figs. 2-1 and 2-2). As an example, the lipid kinase phosphatidylinositol 3 kinase (PI3K), described in more detail in a later section, recognizes and binds to a pattern of phosphorylated tyrosine residues found within multiple activated RTKs through the SH2 domain located in its p85 regulatory subunit. Binding of the p85 regulatory subunit in turn results in activation of its kinase activity. Approximately 20 classes of RTKs have been defined based on growth factor specificity. This section will focus on the RTK classes for which specific cancer therapies exist or are in development.

G-Protein Coupled Receptors Signaling

G-protein coupled receptors (GPCRs) are seven transmembrane domain–containing proteins that transduce ligand-specific signals across the plasma membrane to mediate numerous physiological processes including sensory perception, immunologic responses, neurotransmission, weight regulation, and cardiovascular activity.¹²⁹ GPCRs also regulate basic cellular functions including growth, motility, differentiation, and gene transcription. The GPCR family comprises more than 800 receptors, which are the targets of more than 30% of all FDA-approved drugs, though few to date have found a role as anticancer therapies.^130,131 Given that GPCRs activate many of the signaling cascades that are deregulated in human cancer, it is not surprising that recent studies have implicated GPCRs in cancer initiation and progression.^134–134

GPCRs can be categorized into five or six families, depending on the nomenclature used.¹³⁵ Using a more recent phylogenetic classification, the five major families are represented by the acronym GRAFS: Glutamate, Rhodopsin, Adhesion, Frizzled/Taste2, and Secretin.¹³⁶ The rhodopsin family of receptors (also referred to as class A GPCRs) is the largest class, with more than 670 members. Crystallization of the bovine rhodopsin receptor in 2000 provided the first high-resolution insight into the structure of GPCRs.¹³⁷ In general, rhodopsin family of receptors have short N-termini. Included in this family are the α group (histamine, dopamine, serotonin, adrenoceptors, muscarinic, prostanoid, and cannabinoid receptors), the β group (endothelin, gonadotropin-releasing hormone, and neuropeptide-Y), the γ group (opioid, somatostatin, and angiotensin), and the δ group (P2RYs, glycoprotein-binding FSHR/TSHR/LHCGR, PARs, and olfactory receptors).¹³⁸

The 15 secretin receptors (class B) all have conserved cysteines in the first and second extracellular loop, and most have three cysteine bridges in the N-termini. This class includes the calcitonin-like, corticotrophin-releasing hormone, the glucagon-like, gastric inhibitory polypeptide, growth hormone–releasing hormone, adenylate cyclase activating polypeptide, parathyroid hormone, secretin, and the vasoactive intestinal peptide receptors.¹³⁹ The adhesion family (also included in class B, according to another classification system) has distinctly long N-termini, and only 3 of 33 receptors in the family have known ligands (EGF-like module containing mucin-like receptors [EMR]-2/3/4)¹⁴⁰. The glutamate receptor family (class C) binds ligands in their N-termini, which has a complex two-domain folded structure bridged by disulphide bonds. Common receptors in this family of 22 include the glutamate, GABABRs, calcium-sensing, the sweet and umami taste (TAS1R1-3), and the GPCR6 receptors.^141,142 The Frizzled (Fz) receptors (FZD1-10 and SMO; see the last two paragraphs of this section for an in-depth discussion of signaling) bind Wnt ligands in the extracellular domain region containing nine conserved cysteine residues.¹⁴³ The taste2 receptors (25 members of T2R, bitter taste) have varying sequence homologies that likely allow the sensing of thousands of distinct bitter tastes.^144,145 Overall, GPCRs share the seven transmembrane domain structure but have different regions of conservation. All of the families include orphan receptors, which are related by sequence and structure, but have no identified ligand to date.

All GPCRs have seven α-helical domains that weave through the plasma membrane and are interconnected by flexible extra- and intracellular segments, thus engendering the synonyms serpentine or heptahelical receptors. The specificity of biological response initiated by each GPCR depends on (1) ligand recognition; (2) the distinct structure of each receptor class and subclass; and (3) ligand-directed binding of specific cytosolic enzymes and adapters that initiate a plethora of intracellular signaling cascades (general and cancer-specific examples will be discussed further later in this chapter).

Operationally, ligand binding on the GPCR extracellular surface induces a conformational change in the receptor, mainly via transmembrane helices 5 and 6, which creates a deep pocket in the intracellular face of the receptor.148–148 This cleft enables binding and activation of heterotrimeric G proteins, which consist of an inactive, guanosine diphosphate (GDP)-bound Gα subunit and a Gβγ-subunit dimer, which act as a molecular switch.^149,150 Activated GPCRs promote GDP for guanosine triphosphate (GTP) exchange on the Gα nucleotide binding site.¹⁵¹ This GTP-bound Gα dissociates from Gβγ and the receptor, and then both activated subunits proceed to initiate signaling cascades. There are four members of the Gα family, Gα_s, Gα_i/o, Gα_q/11, and Gα_12/13, which can be further subtyped, and each can stimulate several downstream effectors. Moreover, each GPCR can couple to multiple Gα family members, thus generating a complex pattern of intracellular signaling.

Classic GPCR activation of Gα_s stimulates adenylyl cyclase, which generates the second messenger 3′-5′-cyclic adenosine monophosphate (cAMP).152,153 cAMP activates multiple downstream effectors including cAMP-gated ion channels, EPAC, a guanine nucleotide exchange factor for RAP1/2 (which functions in cell adhesion and junction formation), and protein kinase A (PKA).^{129,149,154–157} cAMP binds to PKA regulatory subunits, releasing catalytic subunits and triggering the activation of cytosolic and nuclear substrates, including the transcription factor cAMP response element binding protein (CREB), which can induce proliferation and differentiation, among other phenotypes, depending on cell origin.^160–160 Gα_s also activates the SRC tyrosine kinase and the guanosine triphosphatase (GTPase) activity of tubulin.^149,154 GPCRs coupled with Gα_i/o typically work in opposition to Gα_s by inhibiting adenylyl cyclase and decreasing cAMP levels.^129,161 In addition, some Gα_i/o isoforms can signal to K⁺ and Ca⁺ channels, increase cyclic guanosine monophosphate (cGMP) phosphodiesterases, interact with RAP1GAP1, and cross talk to the MAPK pathway (as described later in this chapter).¹²⁹

Members of the Gα_q/11 family activate phospholipase C–β, which catalyzes the hydrolysis of PIP₂ to yield IP₃ and DAG.¹⁶² IP₃ mobilizes calcium from intracellular stores, whereas DAG activates some isoforms of protein kinase C (PKC).^163,164 Both Gα_q/11 and Gα_12/13 activate a variety of RhoGEFs (p115-RHOGEF, PDZ-RHOGEF, LARG, LBC, and AKAP-LBC) and thus regulate Rho activity (mainly RHO-A) and its contribution to actin stress fiber formation, cell shape and polarity, cell adhesion and migration, gene transcription, and cell cycle progression.¹⁶⁵ Additionally, Gα_12/13 activates the Na⁺/H⁻ exchanger, inducible nitric oxide synthase, phospholipase D, E-cadherin, radixin, and protein phosphatase 5. Gβγ signaling is equally complex, as the active dimer stimulates G-protein–regulated inwardly rectifying K⁺ channels, adenylyl cyclase (types II, IV, and VII), PLC-β, PI3K-γ, SRC, and GPCR kinases (GRKs; see two paragraphs below), while it inhibits adenylyl cyclase (type I), some Ca⁺ channels, and calmodulin, stabilizes Gα in the GDP-bound, inactive state, and helps specify coupling to the proper Gα member.^{129,149,150,154}

Gα signaling is switched off by a family of GTPase-activating proteins called regulators of G–protein signaling, or RGS proteins, which enhance the intrinsic rate of GTP hydrolysis by greater than 1000-fold.166,167 Some RGS proteins enhance signaling through RhoGEFs and act as scaffolds for other signaling cascades (e.g., tether to Raf and MEK2). The GPCR effectors PKA and PKC also contribute to receptor inhibition by phosphorylating their cognate activated GPCR and thereby uncoupling and inactivating Gα_s/Gα_q in a classic negative feedback loop.^168,169

GPCRs can also function via G-protein–independent mechanisms by binding to the arrestin family of cytosolic adapter proteins.170,171 GPCR-coupled arrestins have pleiotropic cellular roles including (1) dampening G-protein signaling by scaffolding enzymes that degrade G-protein second messengers; (2) desensitizing receptors by binding GRK-phosphorylated GPCRs and sterically blocking access to further Gα subunits; (3) mediating GPCR trafficking and endocytosis to clathrin-coated pits; and (4) acting as a scaffold for multiple MAPK cascades. For example, MEK1 is engaged by a tethered complex of RAF-1, ERK2, and β-arrestin to facilitate mitogenic signaling.^172,173

GPCRs are well-established drug targets for antihistamine, antacid, cardiovascular, and antipsychotic drugs, pain suppressants, and antihypertension therapies.130,131 Although less well appreciated as drug targets in cancer, increasing evidence indicates that dysregulated GPCR signaling contributes to cancer initiation and progression. For example, in 1986, the wild-type MAS1 gene, which encodes the MAS GPCR, was reported to induce transformation by coupling to the small G-protein RAC.^174,175 Thus although mutational activation of GPCRs and G proteins may be infrequent events in human tumors, overexpression of these receptors/adapters, loss of negative regulators, or excessive ligand in the tumor microenvironment may contribute to the deregulated proliferation, angiogenesis, migration, and inflammation that is characteristic of many cancers.

Aberrant signaling by thrombin, chemokines, lysophosphatidic acid (LPA), gastrin-releasing peptide (GRP), endothelin 1, and sphingosine-1-phosphate (SIP), along with prostaglandin E2, bradykinin, angiotensin II, interleukin-8 (IL-8), estrogen, orexin, and the Hedgehog (Hh) and Wnt ligands through their associated GPCRs, have been implicated in the development of a number of cancers.132,166,167 Numerous inhibitors of GPCR signaling are also being studied in the clinic, including ZD4054, ABT-627 and Bosentan (target: endothelin A/B receptors), BKT-140 (target: CXCR4 receptor), and CXCR2 ligands (target: CXCR2 receptor).¹³³ Detailed in the following four paragraphs are a few specific examples of GPCRs that have been shown to play a role in cancer initiation and/or progression.

A connection between GPCRs and EGFR signaling has been established in both normal cell physiology and in colon, lung, breast, ovarian, prostate, and head and neck cancer development.176–179 Ligand-bound GPCRs activate SRC, PKC, Ca⁺ channels, and PKA intermediaries, which stimulate proteolytic cleavage and release of membrane-tethered growth factors that bind and thereby transactivate EGFR. Specifically, estrogen binding to the GPCR GRP30 facilitates matrix metalloproteinase–2 and –9 (MMP2 and MMP9) cleavage of the growth factor precursor pro-HB-EGF, thus initiating HB-EGF–mediated EGFR transactivation.¹⁷⁷ Through a related mechanism, LPA-, SIP-, and thrombin-activated GPCRs transactivate EGFR in breast cancer cells via growth factor shedding of tumor necrosis factor (TNF)-α through the action of TACE/ADAM17 zinc-dependent proteases.¹⁷⁹ Thrombin-mediated N-terminal cleavage and activation of proteinase-activated receptor 1 (PAR1), which can act through EGFR, has been found to promote metastasis and invasiveness in melanoma, breast, colon, and prostate cancers.^133,134 Intriguingly, MMP1 was found to function similarly to thrombin in activating PAR1 and promoting breast cancer tumorigenesis and invasion.^180,181 This cross talk between GPCRs and EGFR provides a rationale for the combinatorial use of GPCR agonists/antagonists and EGFR inhibitors in patients with EGFR-driven lung and colorectal cancers (see the RTK section).

Aberrant GPCR signaling through Gα_12/13 also contributes to tumorigenesis by enhancing cancer cell migration, invasion, angiogenesis, and metastasis.¹⁸² Ligand-activated LPA, PAR1, SIP, thromboxane A2 (TP), CXC chemokine (CXCR4), and prostaglandin E2 receptors couple to Gα_12/13 and RhoGEFs to hyperactivate RhoA (described previously), which elicits these progression-associated phenotypes in glioma, melanoma, lung, breast, and ovarian cancers.^133,134 Overexpression of Gα_12/13 and RhoA in breast, prostate, and colon cancers also promotes metastasis by decreasing cell adhesion.¹⁶⁵ RhoGEF inhibitors, RhoGTPase inhibitors, inhibitors of prenylation that could indirectly impair proper Rho localization (e.g., statins and farnesyl/geranylgeranyl transferase inhibitors), and inhibitors of kinases downstream of Rho (e.g., ROCK, LIMK, MRCK, and PAK) have all been developed and tested in biochemical, cell line, and mouse experiments; however, a clear benefit to the use of such compounds in patients with cancer has yet to be established.¹⁸³

Two other prominent examples of dysregulated GPCR signaling in human cancer are the Hh/Smoothened (SMO) and Wnt/Fz signaling pathways.132,133,184 Secreted Hh ligands (first identified based on their roles in normal development and stem cell homeostasis, with sonic hedgehog [SHH] being the most ubiquitous) bind to the 12-pass transmembrane receptor Patched (PTCH), which relieves its repression of the GPCR SMO.¹⁸⁵ Activated SMO couples to Gα_i and Gα₁₂, which regulate the glioma-associated oncogene homologue (GLI) transcription factors, which in turn regulate proliferative, survival, and differentiation signals involving cyclin D1, MYC, BCL2, and the forkhead transcription factors, to name a few.^186,187 Mutations in SHH, PTCH, and SMO are found in patients with inherited and sporadic basal-cell carcinomas,^188,189 whereas overexpression of Hh ligands have been shown to result in hyperactivation of the pathway in breast, colon, prostate, and pancreatic ductal adenocarcinomas.^184,190 Vismodegib (GDC-0449), a selective SMO receptor inhibitor, was approved in January 2012 for the treatment of advanced basal-cell carcinomas (Table 2-1).^191,192 Several other Hh/SMO inhibitors are currently being tested in patients with cancer, mostly for advanced and/or metastatic solid tumors, namely saridegib (IPI-926), BMS-833923 (XL139), LEQ506, LDE225, PF-04449913, and TAK-441.^133,193

The secreted Wnt glycolipoprotein ligands activate the single transmembrane low-density lipoprotein related coreceptors LRP5/6 and the GPCR-like transmembrane protein Fz, which are phosphorylated and likely couple to Gα_q and Gα_o, respectively, to activate the cytoplasmic scaffold Dishevelled.196–196 Dishevelled in turn inhibits the β-catenin degradation complex (which consists of adenomatous polyposis coli (APC), axin, CKI-α, and GSK-3β, and the E3 ubiquitin ligase β-TrCP).¹⁹⁷ This inhibition results in accumulation of β-catenin, which translocates to the nucleus, where it induces T-cell factor/lymphoid enhancer factor (TCF/LEF)-mediated transcription of genes important for cell differentiation and proliferation, including myc, cyclin D1, VEGF, FGF4/18, E-cadherin, COX2, and members of the Wnt cascade itself.^196,198,199 Noncanonical Wnt/Fz pathways include signaling to the transcription factors nuclear factor of activated T cells (NFAT) via Gα_q/i/Gβγ/PLC/PKC/Ca⁺ (Wnt-calcium pathway) and activator protein–1 (AP1) via RHO/RAC/JNK (Planar cell polarity pathway).^132,200 These pathways regulate cell polarity and migration and are implicated in cancer metastasis.²⁰⁰ Aberrations in canonical Wnt signaling promote tumorigenesis in melanoma, colon, liver, ovarian, and prostate cancers.²⁰¹ Specifically, loss-of-function mutations or truncations in APC and AXIN1/2 and gain-of-functions mutations in β-catenin are found in almost all colorectal cancers, with APC alteration found in more than 85%.^196,202 Germline mutations in the APC gene are also the basis for the inherited cancer predisposition syndrome familial adenomatous polyposis (FAP).²⁰³ Efforts to develop selective inhibitors of Wnt signaling are ongoing and will be aided by current endeavors to crystallize members of the Wnt cascade.²⁰⁴ Of note, nonsteroidal anti-inflammatory drugs (NSAIDs) have shown some promise in modulating Wnt signaling, likely by inhibiting the Wnt-output gene COX2 or by enhancing E-cadherin signaling.^196,205 NSAIDs are approved by the FDA for the treatment of patients with FAP.^204,206

Cytokine Receptor Signaling

Cytokines are protein and glycoprotein ligands secreted by immune cells that initiate diverse and often opposing effects based on target cell lineage. Processes regulated by cytokine signaling include cell proliferation, differentiation, survival, inflammation, angiogenesis, antiviral activity, and modulation of immune function. Cytokines signal in an autocrine and/or paracrine fashion and can be subclassified by protein structure into four families, which total more than 100 members: hematopoietins, interferons (IFNs), chemokines, and the tumor necrosis factor (TNF) superfamily. The hematopoietin family consists of interleukins (IL-1 to IL-31), growth hormone, prolactin, erythropoietin, thrombopoietin, leptin, granulocyte and granulocyte-macrophage colony stimulating factor, and a few others.²⁰⁷ The majority bind to either class I or II cytokine receptors, which are single transmembrane glycoproteins that lack kinase activity and diverge in their extracellular domains to specify ligand binding.^208,209

Class I receptors function as a cluster of two or three subunits that each have two sets of conserved cysteine pairs and a WSXWS motif in their external domains. Class II receptors lack the WSXWS motif and one of the class I conserved cysteine pairs, but they contain conserved proline, tryptophan, and an additional two conserved cysteine residues. Ligand binding causes aggregation of the γ chain (also called γc, CD132), which is common to many cytokines, and the β chain (also known as IL-2R-β, IL-15R-β, or CD122).210,211 In the case of specific cytokines, such as IL-2, association with a third subunit, the α chain (also called IL-2R-α, CD25, or Tac antigen), allows for high-affinity ligand binding.^212,213

The IFN family members are divided into types I, II, and III classes.214,215 Most IFNs are type I and can be further subtyped.²¹⁶ All type I IFNs bind the type I IFN receptor, which consists of two subunits (IFNAR1 and IFNAR2). The sole type II IFN, IFN-γ, binds the type II IFN receptor, which is composed of the IFNGR1 and IFNGR2 subunits. Both types of IFN receptors belong to the class II cytokine receptor family. The IFN family also includes a third branch, the IFN-like molecules, IFN-λ1 (IL-29), IFN-λ2 (IL-28A), and IFN-λ3 (IL-28B), which display some structural overlap with interleukins and the antiviral properties of IFNs and bind a distinct receptor made up of IFNLR1/IL-28R-α and IL-10R-β.²¹⁷ Given that cytokine receptors are promiscuous and bind multiple cytokine ligands, there is a high degree of redundancy in the output profiles of individual cytokines. This redundancy serves to amplify and sustain signaling downstream of these transitory stimuli.

Hematopoietin or IFN binding induces oligomerization of cytokine receptor subunits and autophosphorylation and activation of the Janus activated kinase (JAK) family of intracellular tyrosine kinases (TYK2, JAK1-3), which are constitutively bound to box I and box II α-helix motifs in the receptor cytoplasmic tail.²¹⁸ Activated JAK proteins phosphorylate tyrosine residues on the cytokine receptor chains, which classically bind the STAT family of transcription factors (STAT1-6).^207,219 Docking of STATs facilitates their phosphorylation by JAKs, which in turn causes STAT dimerization, nuclear translocation, and alterations in gene transcription. There is also cross talk between STAT signaling and the NF-κB and SMAD signaling pathways.²²⁰ STATs are also activated by receptor tyrosine kinases, SRC, and ABL, and STAT activation also plays a role in transformation initiated by these oncogenes.²²¹ JAK/STAT activity is regulated by several posttranslational modifications, as well as by the suppressor of cytokine signaling (SOCS) and protein inhibitor of activated STAT (PIAS) proteins.²²⁰ In addition to JAK-STAT signaling, cytokine receptors can transduce messages through LCK and SYK (Src-family kinases), BCL2, and via PI3K/AKT– and RAS/RAF–mediated upregulation of FOS- and JUN-dependent transcription.^221,222

The 29-member TNF superfamily of receptors and their corresponding 19 ligands induce inflammation, in addition to prosurvival and proapoptotic phenotypes.223,224 TNF ligands are transmembrane proteins that function as membrane-integrated or cleaved, soluble trimers that bind and activate preformed, single-transmembrane TNF receptor (TNFR) trimers on the cell surface.²²⁵ Conserved cysteine residues in the TNFR external domains dictate ligand binding, and the presence of death domains, TRAF-interacting motifs, or neither (decoy) dictate downstream signaling. For example, on apoptotic stimuli, ligand activation of the TNFR1 and DR3 receptors recruits the adaptor TRADD (TNFR-associated death domain) via death domains, which in turn binds to FADD (Fas-associated protein with death domain).^223,224,226 FADD binds procaspase-8 and procaspase-10 via death effector domains and induces their cleavage to form active enzymes that cleave caspase-3 and induce apoptosis. TRADD binding is also capable of inducing the intrinsic apoptotic cascade (mitochondrial release of ROS, cytochrome C, and BAX, which leads to caspase-9 and caspase-3 activation and apoptosis).^223,227

Under proliferative stimuli, TNF-α–dependent activation of TNFR1/TRADD and TNFR2 converge on the recruitment of TNFR-associated factor 2 (TRAF2). TRAF2 binding sequentially recruits the receptor interacting protein, TGF-β–activated kinase 1 (TAK1), IκB kinase trimer, which functions to degrade IκBα (inhibitor of NF-κB-α) and in turn facilitate the activation and translocation of NF-κB.223,224 NF-κB in turn induces mediators of inflammation and cytoprotective phenotypes.²²⁸ Alternatively, TAK1 can signal to MAPK kinase (MKK) 3/6 and to two MAP kinases, ERK (proproliferation) and p38-α (to activate the transcription factor AP1).²²⁹ Upstream, TRAF2 can additionally trigger MAP/ERK kinase kinase 1 (MEKK1), MKK7, and c-Jun activating kinase (JNK), the recruitment of which also converge to activate AP1 and thus regulate proliferation and survival.^232–232 There is an important avenue of cross talk between the TNF-directed NF-κB and JNK pathways, the balance of which ultimately decides cell fate.²³³

Deregulation of cytokines leads to numerous inflammatory and autoimmune diseases, such as rheumatoid arthritis and systemic lupus erythematosus. Biotherapies including monoclonal antibodies that bind TNF-α (e.g., adalimumab and infliximab) are used to treat these autoimmune diseases.²²⁵ In cancer, dysregulation of cytokine signaling promotes chronic inflammatory signals and prevents the immune system from attacking cancer cells. Unfortunately, because of their pleiotropic effects, which are often cell-type and microenvironment specific, therapeutic strategies that modulate cytokine signaling have demonstrated only modest clinical activity in patients with cancer. For example, targeting the IL-2R-α subunit with radionuclide-bound humanized antibodies to IL-2R-α (daclizumab) in patients with leukemias that overexpress IL-2R-α was associated with increased time to relapse (7 of 16 patients had partial remissions lasting an average of 9.2 months, and 2 of 16 had complete remissions lasting 36+ months).²³⁴ Conversely, administration of recombinant IL-2 (aldesleukin) induced tumor cell death through IL-2–mediated activation of T cells and B cells.²³⁵ The recombinant human INF-α-2a mimetic, Roferon-A, has been shown to inhibit tumor growth in patients with melanoma and hairy cell leukemia, but such treatments have now been supplanted in these diseases by more active therapies.²³⁶ Moreover, TNF-α has demonstrated antitumor activity but cannot be administered systemically to patients because of its toxic effects. Greater success has been observed with focal delivery of TNF-α to limbs with soft tissue sarcomas and melanomas through isolated limb perfusion techniques. ^237,238 Anti-TRAIL therapies have also recently entered clinical testing (i.e., the anti-TRAIL monoclonal antibody HGS-ETR1) because TRAIL has been shown to have selective proapoptotic effects in tumor versus normal cells, with a more favorable side effect profile than TNF-α-directed therapies.²²⁶

Downstream components of the cytokine signaling cascades are also being explored as targets for drug development. Gain-of-function mutations in the JAK2 and MPL genes, the latter of which encodes the thrombopoietin receptor, are common in myeloproliferative neoplasms, and the selective JAK1/2 kinase inhibitor ruxolitinib (INCB018424) has recently been approved for this indication (Table 2-1).^241–241 STAT3 is frequently hyperactivated in persons with cancer, because it is a downstream effector of both cytokine receptors and mutated/amplified tyrosine kinases.²²¹ Constitutive activation of STAT5 and STAT6 play a critical role in BCR-ABL–driven chronic myelogenous leukemia and IL-13–driven lymphomas and leukemias.²²¹ On the basis of these findings, direct inhibitors of STAT signaling are currently in development.

Inhibition of NF-κB, a downstream mediator required for the transforming properties of many oncogenes and cytokines, has been shown to induce apoptosis in leukemias and lymphomas and enhance chemotherapy and radiation response.²²⁸ JNK1 is upregulated in hepatocellular carcinoma and prostate cancer, whereas p38α activity is either lost (in hepatocellular carcinoma) or activated in numerous cancers.²³⁰ Several inhibitors of JNK/p38 are currently in clinical testing as anticancer agents, including ARRY-614, ARRY-797 Semapimod, and Talmapimod (SCIO-469).²³⁰

Serine/Threonine Receptor Signaling

Receptor serine/threonine kinases are exemplified by the TGF-β type I and II receptors.242,243 Ligands for these receptors include the TGF-β superfamily, which comprises the TGF-β1-3 isoforms, activins, inhibins, Nodal and Lefty, as well as the more distantly related bone morphogenic proteins, growth and differentiation factors, and Mullerian inhibitory substance. TGF-β ligands regulate a diverse array of physiological processes including growth, proliferation, survival, hormone release, and differentiation.^242,244 They thus have a central role in embryonic patterning, tissue development, and morphogenesis. Signaling mediated by TGF-β, the namesake and most studied ligand of the superfamily, will be used to exemplify the general structure of the serine/threonine kinase cascades.

TGF-β is ubiquitously expressed and requires a multistep maturation and secretion process to be functional and bioavailable. Initially translated as an immature proprotein, the prodomain (called latency-associated protein) is cleaved and noncovalently bound to the remaining mature form of TGF-β.²⁴⁵ Covalently bound mature, active dimers are further bound to latent TGF-β binding protein (LTBP) such that this complex is sequestered by LTBP binding to the extracellular matrix until appropriate signals initiate matrix metalloproteinase, plasmin, or thrombin-dependent cleavage of LTBP and subsequent release of active TGF-β dimers.^245,246 TGF-β dimers bind constitutively active, single transmembrane, homodimeric TGF-β type II receptors (TβRII). Once bound by ligand, TβRII forms a complex with TGF-β type I (TβRI) receptor homodimers, thus creating a receptor heterotetramer.^242,244 TβRII receptors phosphorylate and activate the intrinsic kinase activity of TβRI, which in turn phosphorylates serine residues in the C-terminal–SSXS motif of receptor-activated Smad proteins (R-SMAD2 and R-SMAD3 for TGF-β).²⁴⁷ Activated R-SMAD2 and R-SMAD3 then form a heterotrimer with SMAD4, which then localizes to the nucleus, where it both binds DNA and partners with other transcription factors (i.e., forkhead, homeobox, zinc-finger, AP-Ets, and bHLH family transcription factors) and cofactors (e.g., p300 and CBP).^248,249 These Smad-containing complexes then target selective promoter elements with high affinity, leading to the induction or repression of hundreds of genes, depending on cell context. For example, TGF-β induces the expression of 4EBP1 and the cyclin-dependent kinase inhibitors INK4B and CDKN1A (p21) and represses Myc to elicit growth inhibitory effects.²⁵⁰ It also activates death-associated protein kinase, growth arrest and DNA damage-inducible 45β, and BIM to promote apoptosis and PDGF in smooth muscle cells to enhance proliferation, among other effects.²⁵¹ TGF-β also signals through many non-Smad effectors, including Shc, which can result in enhancement of RAS/ERK signaling and TRAF6, which activates the TAK1/MKK3&6/JNK/p38 cascades.²⁵² Through indirect mediators, TGF-β can also activate Src, Rho, and PI3K, and through pathway cross talk, it can activate the Wnt, Hh, and Notch cascades.²⁵³

Alterations in TGF-β signaling are common in cancer. 251,254 For example, mutational inactivation of TβRII is common in colorectal cancers with microsatellite instability.²⁵⁵ Mutations and loss of SMAD4 expression are common in colorectal, pancreas, and head and neck cancers.^256,257 Conversely, increased TGF-β expression occurs in breast, prostate, and colorectal cancers and has been associated with cancer progression and the development of metastases.²⁵¹ It has been proposed that TGF-β signaling promotes invasion and metastasis by promoting epithelial-to-mesenchymal transition.²⁵⁰ TGF-β has also been shown to play a role in maintaining the tumor-initiating cell or cancer stem cell population in gliomas, leukemias, and breast cancer.²⁵¹

Intense efforts are underway to develop therapeutic strategies that inhibit the tumorigenic properties of TGF-β signaling. These include the development of selective TβRI inhibitors, TGF-β blocking antibodies, and soluble TGF-β antisense therapies (AP12009).258,259 However, the development of inhibitors of TGF-β signaling has been confounded by the ability of TGF-β to both promote and suppress tumor progression in a context-specific manner.^260,261

Notch Receptor Signaling

The mammalian Notch receptors, NOTCH-1, -2, -3, and -4, are single-pass transmembrane receptors that are functionally unique with regard to receptor processing, signal activation, and transduction.264–264 Notch receptors are translated as immature receptors that undergo S1 cleavage by furin-like convertase during Golgi trafficking.²⁶⁵ This process generates two subunits that are retethered at the cell membrane by noncovalent bonds in the heterodimerization domain.^263,266 The extracellular domain subunit consists of ligand binding EGF-like repeats, a heterodimer domain, three negative regulatory LIN12 and Notch repeats containing numerous cysteines, and a hydrophobic transmembrane-interacting region. The other subunit contains the transmembrane domain and the intracellular domain, which has ankyrin repeats and a RAM domain central to the Notch receptor’s unusual activity as a direct transcription factor, and a PEST motif important for the quick termination of Notch activity and ubiquitin-mediated degradation.

Deltalike (DLL) 1, 3, and 4 and Jagged (JAG) 1 and 2 are the five transmembrane-protein ligands for the Notch receptors.²⁶⁷ A Notch receptor on one cell binds to DLL or JAG ligand on an adjacent cell. This process results in a change in Notch receptor conformation, which facilitates a series of proteolytic cleavages required for receptor activation. First, the ADAM17/TACE (a disintegrin and metalloprotease 17/TNF-α converting enzyme) metalloprotease performs S2 cleavage of the extracellular domain.^268,269 This cleavage allows subsequent S3 cleavage of the transmembrane domain by the γ-secretase complex and release of the Notch intracellular domain (NICD).²⁷⁰ The active NICD translocates to the nucleus, where it binds to and converts the repressor complex CSL (CBF1, Suppressor of Hairless, Lag-1) into a transcriptional activator that recruits coactivators, including the Mastermind-like family proteins and p300. NICD target genes include the HES (hairy enhancer of split) and HRT/HEY (hair-related transcription factor) transcriptional repressors, as well as cyclin D1, MYC, CDKN1A (p21), NF-κB, and the Notch receptors and ligands themselves, among others.^262,266,271 Notch is also heavily involved in cross talk with other pathways, including receptor tyrosine kinases, PI3K, RAS/ERK, JAK/STAT, WNT, Hh, TGF-β/SMAD, and p53.

Notch’s causal role in tumorigenesis was established on its discovery in 1991 when a translocation event in acute T-cell lymphoblastic leukemia/lymphoma (T-ALL) was identified that fused the T-cell receptor-β promoter/enhancer elements to Notch, generating a truncated form of NOTCH1 that was constitutively nuclear and active.²⁷² Activating NOTCH1 mutations have since been found to occur in approximately 60% of patients with T-ALL.²⁷³ Subsequent studies have determined that Notch and its ligands have both oncogenic and tumor suppressor properties in several hematologic and solid tumor malignancies including melanoma, glioblastoma, breast, lung, colorectal, and pancreatic cancers, depending on cellular context.^263,266 Thus far, efforts to develop inhibitors of Notch signaling have focused primarily on inhibiting cleavage and thus activation of Notch. Specifically, selective inhibitors of gamma-secretase are currently being tested in early-stage clinical trials.²⁶²

Nuclear Hormone Receptor Signaling

The nuclear hormone superfamily is composed of hormone receptors and orphan receptors (receptors for which no ligand has yet been identified). The nuclear hormone receptors are characterized structurally by a ligand-binding domain, a DNA-binding domain, and a hinge region that connects the ligand and DNA-binding domains.²⁷⁴ Nuclear hormone receptors are classified into four subtypes based on ligand specificity: steroid, retinoid X receptor, monomeric or tethered orphan receptors, and dimeric orphan receptors.²⁷⁵ The steroid receptor ligands include estrogen, progesterone, androgen, and growth hormone. Ligand binding occurs in the cytoplasm and results in receptor homodimerization followed by nuclear translocation. Once translocated into the nucleus, the ligand-bound receptor acts as a transcription factor that modulates the expression of several downstream proteins through binding to steroid response elements, which are conserved nucleotide sequences within the regulatory regions of genes.²⁷⁶ In contrast to the steroid receptors, the retinoid X receptors form heterodimeric complexes with other partners, including the retinoic acid receptor, the thyroid hormone receptor, and vitamin D receptors.

Hormone receptor blockade is a cornerstone in the treatment of estrogen receptor (ER)–expressing and progesterone receptor–expressing breast cancers. The selective estrogen response modulator tamoxifen competes with estradiol for binding to ER. Notably, tamoxifen binding results in ER dimerization, nuclear translocation, and receptor binding to estrogen response elements in the promoter regions of estradiol-target genes. It is thought that the ER : tamoxifen complex recruits transcriptional corepressors in contrast to ER : estradiol binding, which recruits transcriptional coactivators.276,277 Although tamoxifen has antiproliferative effects in ER-expressing breast cancer cells, it causes hypertrophy and neoplastic transformation of endometrial tissue, likely as a result of cell type and context-specific recruitment of transcriptional coactivators that are differentially expressed in these tissues.²⁷⁶

Dihydrotestosterone is the primary ligand of the androgen receptor (AR). AR blockade by antiandrogens such as bicalutamide is a commonly utilized therapeutic modality for patients with locally advanced or metastatic prostate cancer. Bicalutamide competes with dihydrotestosterone for binding to cytoplasmic AR.²⁷⁸ The mechanism by which bicalutamide-bound AR inhibits androgen-dependent gene transcription is unclear but may involve the recruitment of transcriptional corepressors, as well as histone modifications, which lead to tighter chromatin binding and therefore reduced access to promoter regions by transcription factors. Medical or surgical castration, which reduces circulating testosterone and thus dihydrotestosterone levels, is also a mainstay of prostate cancer treatment. Enzalutamide, a nonsteroidal, small molecule antagonist of the androgen receptor that inhibits AR nuclear translocation and DNA binding, was recently approved for the treatment of patients with metastatic prostate cancer who have progressed after treatment with bicalutamide and medical castration (Table 2-1).²⁷⁹

Integrin Receptor Signaling

The integrin receptor family regulates cell adhesion, migration, invasion, and cell survival.280,281 Integrin receptors are heterodimeric molecules consisting of combinations of α and β subunits. Each combination dictates the spectrum of extracellular matrix components to which these receptors bind. Once bound to the extracellular matrix, the receptors recruit multiple proteins to the cell membrane, including cytoskeletal molecules such as paxillin and vinculin that form focal adhesions to extracellular matrix components.²⁸² Unlike the RTK family, integrin receptors do not possess intrinsic kinase activity but rather promote signaling by facilitating the activation of kinases such as SRC or focal adhesion kinase.²⁸³ Integrins are also unique in that they participate in bidirectional signaling. This so-called “inside-out” signaling occurs when extracellular ligands such as cytokines trigger a signaling cascade that leads to a conformational change in the β subunit cytoplasmic tail of the integrin receptor that is transduced to the extracellular component of the receptor, resulting in increased affinity for portions of the extracellular matrix. Conversely, “outside-in” signaling involves binding of ligand to integrins that stimulate the activation of multiple intracellular signaling pathways.²⁸⁴

Integrins are expressed on cancer cells and have been shown to promote disease progression.280,283,285 Integrins are also present on stromal cells, including pericytes (which promote endothelial cell growth and proliferation and thus angiogenesis) and fibroblasts, where they influence the surrounding microenvironment and thus indirectly stimulate tumor growth and proliferation. For example, vascular cell adhesion molecule 1 is expressed on pericytes and binds to the integrin receptor α4β1, which is found on the surface of endothelial cells, resulting in pericyte recruitment to sites of vascular maturation.²⁸⁶ Integrin signaling also plays a role in the activation of MMP–2,²⁸⁷ which promotes cell invasion and has been shown to regulate cyclin D and cyclin-dependent kinase inhibitor expression, thereby controlling cell cycle progression.²⁸⁸ Finally, integrin receptor activation can lead to increased secretion of growth factors, which then stimulate tumor invasion through autocrine and paracrine mechanisms.²⁸⁷

Tumors that express integrin receptors include melanomas, glioblastomas, and breast cancers. In melanoma, the αvβ3 and α5β1 integrin receptors promote vertical growth and metastatic spread to lymph nodes.289,290 In glioblastoma, αvβ3 and αvβ5 are expressed mainly at the edge of tumors, suggesting a role in tumor invasion.²⁹¹ The expression of the α6β4 and αvβ3 integrin receptors in breast cancer is associated with higher grade and tumor size,²⁹² the development of bone metastases,²⁹³ and decreased survival.²⁹⁴ Clinical trials of monoclonal antibodies that target integrin receptors are underway in several cancer types. For example, etaracizumab, a humanized monoclonal antibody targeting αvβ3 integrin, is being tested in solid tumors and has shown activity in patients with metastatic melanoma.²⁹⁵ Cilengitide, an inhibitor of the αvβ3 and αvβ5 integrins, inhibits angiogenesis and tumor cell proliferation in preclinical studies and is being tested in a phase 3 trial in combination with temozolomide and radiation in patients with glioblastoma.²⁸⁷

The SRC family of intracellular, non-receptor tyrosine kinase proteins is composed of 11 members (Src, Fyn, Yes, Blk, Yrk, Frk/Rak, Fgr, Hck, Lck, Srm, and Lyn).²⁹⁶ They share common structural features, including the so-called SH domains 1 to 4. The SH1 domain includes the kinase domain. Only Src, Fyn, and Yes are expressed ubiquitously, whereas the tissue distribution of the latter six is more restricted.²⁹⁷ Together, Src family kinases have pleiotropic roles in cellular proliferation, apoptosis, differentiation, motility, adhesion, angiogenesis, and immunity.^298,299

SRC is by far the most intensively studied family member and was the first gene observed to have oncogenic potential. Peyton Rous was awarded the Nobel Prize for a series of experiments showing that a transmissible factor was present in avian sarcomas that was capable of initiating tumors in recipient birds. Five decades later the viral oncogene v-Src was identified as the oncogenic factor in the Rous sarcoma virus.302–302 Bishop and Varmus later showed that v-Src was a mutant form of the cellular protooncogene c-SRC, and Hunter and colleagues showed that its transformative capacity was dependent on its tyrosine kinase activity.^13,14,303

c-SRC, hereafter called SRC, is regulated in a number of ways. First, Src has a myristoylation site in its N-terminus that is necessary for membrane localization and that promotes its interaction with nearby membrane-bound effectors.³⁰⁴ The SH2 and SH3 domains facilitate protein-protein interactions and conformational changes in the protein. Inactive SRC is maintained in a closed conformation, with phosphorylated Y530 (mediated by CSK, C-terminal SRC, and CSK homology kinases) interacting with the folded over SH2 domain.³⁰⁵ The closed, inactive confirmation of SRC is further stabilized by proline-rich segments of the kinase domain associating with the SH3 domain.³⁰⁶ SRC activation requires dephosphorylation of Y530, likely by PTPα/γ/1β (protein tyrosine phosphatase α/γ/1β) or SHP1/2, which allows the kinase to assume an open conformation.^296,307 Autophosphorylation of Y419 in the activation loop of the kinase domain also promotes full activity,³⁰⁸ whereas binding of focal adhesion kinase and CRK-associated substrate to the SH2 domain induce SRC activation and link SRC signaling to the regulation of focal adhesion, actin reorganization, and migratory phenotypes.^309,310 SRC is a downstream mediator of numerous receptor families including receptor tyrosine kinases, integrin receptors, hormone receptors, cytokines receptors, and GCPRs and promotes signaling through the PI3K/AKT, RAS/MAP kinase, and JAK/STAT cascades, among others.^{296,297,299,307,311–315}

Although SRC mutations are rare in human cancers, SRC is frequently activated as a consequence of other mutational events in colorectal, breast, esophageal, gastric, pancreatic, hepatocellular, ovarian, and lung cancers.²⁹⁶ In colorectal and hepatocellular carcinomas, SRC activation occurs in the setting of concomitant loss of CSK.^318–318 The tyrosine kinase inhibitor dasatinib, which is used in the treatment of chronic myelogenous leukemia and Philadelphia-chromosome–positive acute lymphoid leukemia, inhibits Src family kinases, as well as BCR-ABL, KIT, Ephrin A2 receptor, and PDGFR (Table 2-1).^321–321 Additional dual SRC/ABL and SRC selective inhibitors that are in early clinical testing include bosutinib (SKI-606), saracatinib (AZD0530), ponatinib (AP24534), XL-228, KX2-391, AZM475271, XL99, TG100435/100855, and DCC2036. Most have shown limited single-agent activity and are being developed as combination therapies.^{299,307,313–315}

RAS/MAP Kinase Pathway Signaling

First identified more than 30 years ago as the oncogenes responsible for the transforming potential of the Harvey and Kirsten murine sarcoma retroviruses (Ha-MSV and Ki-MSV), Ras proteins are guanine nucleotide binding proteins.330–334 Ras proteins have intrinsic GTPase activity and cycle between inactive GDP-bound and active GTP-bound states. GDP/GTP exchange thus allows Ras proteins to function as binary molecular switches.

The human genome includes three RAS genes that encode four homologous proteins (HRAS, NRAS, and the alternative splice variants KRAS4A and KRAS4B) with highly conserved N-terminal and variable C-terminal regions. After stimulation of cells by serum growth factors, cytokines, hormones, and neurotransmitters, Ras undergoes a series of C-terminal (C¹⁸⁶AAX) posttranslational modifications that result in its localization to specific membrane microdomains.³³⁵ Membrane localization is required for the transforming properties of Ras, because mutation of Ras at C¹⁸⁶ results in cytosolic localization and protein inactivation, whereas Ras activity can be rescued by myristoylation, which promotes membrane localization.^338–338 Guanine nucleotide exchange factors (GEFs) promote Ras activation by binding to GDP-bound Ras and facilitating the release of GDP and the binding of GTP. SOS1, RAS-GRF (dual-specificity GEFs for Ras and Rac), and RAS-GRP (stimulated by DAG/phorbol esters and Ca⁺) are the mostly highly characterized Ras-GEFs.^341–341 Using RTK stimulation as an example, ligand binding induces RTK dimerization and autophosphorylation of tyrosine residues in the receptor cytoplasmic tail. Phosphotyrosine docking sites recruit scaffold proteins such as SHC and permit interaction with the SH2 domains of adaptor proteins such as GRB2.³⁴² GRB2 in turn recruits SOS1 via its SH3 domain, thereby positioning SOS near membrane-anchored Ras (Fig. 2-1).^342–346 SOS1 in turn activates Ras via its CDC25 homology (RasGEF) domain and N-terminal Ras exchanger motif (REM or RasGEFN domain).

Ras inactivation is catalyzed by GTPase-activating proteins (GAPs), which enhance the intrinsic GTPase activity of Ras proteins by 100,000-fold.³⁴⁷ Ras GAPs, which include p120-RASGAP, neurofibromin, GAP1^IP4BP, and CAPRI, negatively regulate Ras activity and thus function as tumor suppressors.^331,335

Binding of GTP to Ras induces conformational changes in the switch I (loop 2 residues 30 to 38) and switch II (helix 2 and loop 4 residues 60 to 76) domains, which facilitate the association of Ras with regulators and downstream effectors.348,349 Ras directly interacts with more than 20 effector proteins, of which the Raf kinases, PI3Ks, and RALGDS are the best characterized. The canonical RAS/RAF/MEK/ERK (classical MAP kinase) cascade is by far the most extensively characterized Ras effector pathway. This prototype of a three-tiered kinase signaling cascade exemplifies numerous Ras-dependent MAPK cascades that respond to diverse signals, including cell stress and cytokine signaling (see the section on cytokines for details of the JNK and p38 pathways).³⁵⁰ The Raf protein family (which represents the top-tier MAPK kinase kinases, or MEKK) is comprised of three differentially expressed isoforms, ARAF, BRAF, and CRAF (RAF-1).^351,352 Raf, via its Ras binding domain and cysteine rich domain, interacts with GTP-bound Ras.^353–356 Binding of Raf to GTP-bound Ras results in Raf localization to the plasma membrane and its subsequent phosphorylation and activation.³⁵⁷ The mechanisms of Raf1 and B-Raf phosphorylation and activation are distinct and are derived from the summation of signaling inputs from small G proteins (Ras, Rac, CDC42, and RAP1), kinases (activating inputs: SRC/PAK/PKC; inhibitory inputs: PKA/AKT/SGK), isoform homodimerization and heterodimerization, phosphatases (PP2 and PP2A), scaffolding proteins (e.g., KSR, Raf kinase-inhibitory protein [RKIP], and HSP90), and cofactors (14-3-3), with phosphorylation events regulating critical aspects of Raf activation.^352,358 In brief, Ras binding to RAF-1 releases 14-3-3, a negative regulator that binds to basally phosphorylated residues S259 and S261 and sequesters RAF-1 in the cytosol in a closed, inactive conformation. Liberation from 14-3-3 exposes the PKA/AKT/SGK–dependent inhibitory phosphorylation on S259, facilitating its dephosphorylation by protein phosphatase 2A.^359–363 Loss of this inhibitory phosphorylation primes RAF-1 for Ras, PAK, SRC, growth factor, and integrin-stimulated activating phosphorylations on S338, Y341, T491, and S494.^364–368 ARAF is activated similarly to RAF-1. Notably, BRAF activation requires fewer steps because of its constitutive phosphorylation at S445, a site analogous to S338 in RAF-1.^358,369 Binding to Ras-GTP stimulates BRAF phosphorylation at critical residues in its activation loop, T599 and S602 (analogous to T491 and S494 in RAF-1).^370,371 The BRAF isoform is the most potent activator of ERK pathway output.^352,369,371

Activated Raf proteins bind and phosphorylate MEK1 and MEK2 (mitogen-activated protein kinase/extracellular signal-regulated kinases 1 and 2, or MAPKK, or MAP2K) on serines 217 and 221.³⁷² Activated MEK, a dual-specificity threonine/tyrosine kinase, in turn phosphorylates MAPK/ERK1 and -2 on threonine 183 and tyrosine 185, inducing a conformational change, activation, and dissociation from MEK.³⁷³ Activated ERK then phosphorylates substrates in the cytosol (e.g., p90RSK) and in the nucleus (such as the transcription factors Elk1, Ets2, Fos, Jun, ATF2, AP1, Myc, and CREB1), which promote proliferation and survival.^374,375 RAF-1 has also been shown to mediate suppression of apoptosis through non–MEK-dependent interactions with ASK1, MST2, and MEKK1/NF-κB.³⁷⁶

Ras/Raf signaling triggers numerous regulatory mechanisms, including classic negative feedback loops, which serve to attenuate pathway output.³⁷⁷ The Sprouty and Sprouty-related EVH1-domain–containing protein (Spred) proteins (encoded by the SPRY1, -2, -3, and -4 and SPRED1, -2, and -3 genes, respectively) inhibit the cascade at the level of Ras and Raf activation.^378–381 The dual-specificity phosphatases (MKPs/DUSPs) dephosphorylate MAP kinases, including ERK.^382,383 In addition, increased pathway activity induces ERK-dependent negative phosphorylations on BRAF (at S151, S750, T401, and T753) and on RAF-1 (at S29, S43, S289, S296, S301, and S642) that abrogate interactions with Ras and homodimer and heterodimer formation.^384,385 Ras signaling is further regulated by RKIP, which disrupts the RAF-1:MEK interaction and impedes mitogenic-signal propagation, which represses KSR-dependent scaffolding of Raf : MEK, among other functions.^{351,386–388}

In its active, GTP-bound state, Ras alternatively binds to the p110-α catalytic subunit of class I PI3Ks,389,390 causing activation of its lipid kinase activity, thereby generating PIP₃, which in turn stimulates the proproliferative and prosurvival kinases PDK1 and AKT (see the section on PI3K signaling for details). Ras-dependent activation of PI3K can further stimulate Rac, a Rho family GTPase involved in regulation of actin and NF-κB.^330,391,392 A third major class of Ras effectors is the group of GEFs for RAL-A and RAL-B, namely RALGDS, RGL, and RGL2.^395–395 These Ral exchange factors stimulate phospholipase D and the CDC42/Rac-GAP Ral binding protein 1, RalBP1, and inhibit forkhead transcription factors to regulate transcription, vesicular trafficking, and cell-cycle progression.

Several additional Ras effectors have been identified. These Ras effectors include (1) PLC-ε, which generates IP₃ and DAG and regulates calcium release and PKC activation; (2) T-cell lymphoma invasion and metastasis-1, which facilitates actin reorganization via Rac; (3) AF6, which contributes to cytoskeletal changes; (4) RIN1, which regulates endocytosis; and (5) RASSF and NORE1, which have been shown to regulate apoptosis and cell cycle progression.330,335 Many other nondirect connections allow Ras to impact the cellular microenvironment, metabolic signaling, autophagy, inflammation, and immune responses. Overall, the complex effects of Ras activation result in a diversity of context-dependent phenotypes ranging from cell proliferation to cell death that are influenced by a wide variety of extracellular stimuli and intricately woven layers of regulation.

The potent oncogenic effects of Ras signaling are highlighted by the high prevalence of mutations in the RAS genes and their proximal downstream effectors.335,396–398 RAS mutations are found in approximately 30% of all human tumors, the majority of which (85%) occur in the KRAS isoform. KRAS is frequently mutated in pancreatic cancers (58%), colorectal and biliary tree cancers (33% and 31%, respectively), and non–small cell lung adenocarcinomas (17%). HRAS mutations are most common in low-grade bladder cancers (11%), whereas NRas mutation is a frequent event in melanoma (18%) and biliary tree cancers (11%). Mutations that lock Ras in its GTP-bound state confer oncogenic potential. Point (missense) mutations in residues 12, 13, and 61 (exons 2 and 3) generate a constitutively active Ras oncoprotein by abrogating intrinsic GTPase activity and inhibiting GAP binding.³⁹⁹ Other mutations, including those at residues 117 and 146, contribute to Ras activation by increasing GDP-to-GTP exchange.^400,401 Germline mutations of several pathway components have been shown to be the underlying cause of the neurofibromatosis type 1 (NF1), Noonan, Costello, and cardio-facio-cutaneous syndromes and other developmental disorders.³⁹⁶ More recently, somatic mutations of the Ras-GAP NF1 have been shown to be highly prevalent in several tumor types, including glioblastomas.⁴⁰² Mutations in BRaf are also common in human cancer and typically occur in a mutually exclusive pattern with Ras mutations. BRAF mutations have been identified in approximately 8% of all cancers, most notably in melanoma (43%), papillary thyroid (39%), biliary tree (14%), colorectal (12%), and ovarian (12%) cancers.^{351,352,376,396} A single valine to glutamic acid substitution at residue 600 (V600E) accounts for more than 80% of all BRAF mutations.

Given the high prevalence of Ras pathway alterations in human cancers, significant attention has been directed toward the development of selective inhibitors of this pathway.377,397,403,404 To date, clinically effective direct inhibitors of oncogenic Ras have yet to be identified. The inability to directly target Ras has been attributed to the high affinity of Ras-GTP interaction. Extensive efforts were thus directed toward inhibiting the posttranslation modifications required for Ras activation. Specifically, inhibitors of the enzyme farnesyl transferase, which regulates Ras localization, were tested in randomized phase 3 trials but were found to be inactive.⁴⁰⁵ The failure of farnesyl transferase inhibitors in KRAS-mutant tumors was predicted by the preclinical observation that geranyl geranyl modification can substitute for farnesylation in targeting KRAS and NRAS to the plasma membrane.^406,407

As an alternative, more recent efforts have focused on the development of selective inhibitors of key kinase effectors of Ras transformation. Selective Raf inhibitors were recently shown to have unprecedented clinical activity in melanoma (Table 2-1).^408,409 Both vemurafenib (PLX4032) and dabrafenib (GSK2118436) induce tumor regressions in most patients with BRAF (V600E) mutant melanoma.^410–414 Notably, these agents selectively inhibit Raf signaling in BRaf mutant tumors and are thus inactive in Ras mutant tumors.⁴¹⁵ The selective MEK inhibitor trametinib (GSK1120212) was also recently shown to improve survival compared with chemotherapy in patients with advanced BRAF mutant melanomas (Table 2-1).⁴¹⁶ Furthermore, the combination of an Raf and MEK inhibitor had greater activity than did Raf inhibitor monotherapy in the same population.⁴¹⁷ Alternative strategies for inhibiting MAPK signaling include selective inhibitors of ERK and HSP90 inhibitors that induce the degradation of CRAF and mutant BRAF.⁴¹⁸ Combinatorial approaches are also being actively pursued, given the modest activity of MEK inhibitors in patients with Ras mutant tumors and the frequent co-occurrence of Ras and PI3K pathway alterations in multiple cancer types.^421–421

The PI3K/AKT/mTOR Pathway Signaling

The PI3K/AKT/mTOR pathway is a key regulator of growth factor–mediated proliferation and survival. Several extracellular growth factors stimulate PI3K by binding to their cognate RTKs or GPCRs.⁴²² Activated PI3K phosphorylates the 3-OH group of the inositol ring of phosphatidylinositol, catalyzing the conversion of PIP₂ to PIP₃. PIP₃ then binds to the pleckstrin-homology (PH) domain of multiple proteins, facilitating their recruitment to the plasma membrane and thus regulating their function (Fig. 2-2).

Class I PI3Ks are heterodimers consisting of a catalytic subunit (p110) and a regulatory subunit (p85) and are involved primarily in the generation of PIP₃.⁴²³ There are four isoforms of p110, designated p110-α, -β, -γ, and -δ, and five p85 isoforms.⁴²⁴ Analogous to the activation of the Ras pathway detailed previously, the p85 regulatory subunit of PI3K associates with phosphorylated tyrosine residues located on the intracellular domains of RTKs through an SH2 domain, resulting in allosteric activation of the p110 catalytic subunit. PI3Ks can also be activated indirectly through the adaptor protein GRB2 after its association with the scaffolding protein GAB1. PI3K pathway activity is negatively regulated by the tumor suppressor PTEN which is a dual lipid and protein phosphatase that dephosphorylates PIP₃, converting it back to PIP₂.^425,426

The best characterized effectors of PI3K are the three members of the serine/threonine kinase Akt family (AKT1, AKT2, and AKT3). Both Akt and phosphoinositide-dependent kinase 1 (PDK1) are recruited by PIP₃ to the plasma membrane via their PH domain, where Akt is phosphorylated at Thr308 by PDK1, resulting in its activation.429–429 Phosphorylation at a second residue, Ser473, by mTOR complex 2 (mTORC2) further enhances Akt activity.⁴³⁰ Activated Akt promotes cellular proliferation, survival, and other phenotypes through activation of multiple downstream effectors. Proliferative effects are regulated through phosphorylation and inhibition of GSK3-β, which phosphorylates cyclin D1 and marks it for degradation; FOXO4, a transcription factor that regulates the expression of the cyclin-dependent kinase (CDK) inhibitor p27; and p21, a second CDK inhibitor that, on phosphorylation, translocates from the nucleus to the cytoplasm, where it regulates cell survival.^431–434 In sum, these actions promote the expression of cyclin D1, which drives progression of cells through the cell cycle and the downregulation of cyclin-dependent kinase inhibitors that suppress cell cycle progression.

Akt-mediated antiapoptotic effects occur through phosphorylation and inhibition of BAD, which negatively regulates the antiapoptotic protein BCL-X_L; caspase 9, a proapoptotic protease; and the FOXO1 transcription factor, which regulates the expression of proapoptotic genes.437–437 Akt also controls cell survival by upregulating NF-κB activity through phosphorylation and activation of Iκb kinase, which marks Iκb, an inhibitor of NF-κB, for degradation.^438,439 Once released from IκB, NF-κB translocates into the nucleus, where it regulates a multitude of genes involved in cell survival. Akt also phosphorylates and activates MDM2, an E3 ubiquitin ligase that binds to the proapoptotic tumor suppressor p53 and directs it for proteosomal degradation.⁴⁴⁰

Although Akt activation mediates many of the transforming effects of PI3K activation, Akt-independent effectors of PI3K activation have also been identified and likely play important roles in the development and progression of some cancers. These Akt-independent effectors of PI3K activation include CDC42 and RAC1, which are involved in motility and reorganization of the cytoskeleton and the serum glucocorticoid kinase family of serine-threonine kinases, which promote cell survival.441,442

Many of the canonical functions of Akt with regard to cell growth and proliferation are mediated through the mTOR pathway. mTOR is a serine-threonine kinase and a member of the phosphatidylinositol kinase–related kinase family.⁴⁴³ mTOR is a component of two complexes, the rapamycin-sensitive mTORC1 and the rapamycin-insensitive mTORC2 complexes.⁴⁴⁴ Within mTORC1, mTOR associates with RAPTOR (regulatory-associated protein of mTOR) and mLST8, whereas the mTORC2 complex consists of mTOR, RICTOR (rapamycin-insensitive companion of mTOR), SIN1, and mLST8.⁴⁴⁵ In addition to its role in activating Akt as previously described, mTORC2 also controls cytoskeletal changes through regulation of paxillin, RHO, RAC, and PKC-α.⁴⁴⁶

mTORC1 activation is regulated in part by Akt phosphorylation of TSC2. TSC2 forms a heterodimeric complex with TSC1 that acts as a GAP for the small GTPase RHEB, causing accumulation of inactive RHEB-GDP.447,448 TSC2 phosphorylation results in suppression of this GAP activity and subsequent activation and accumulation of RHEB-GTP, which then activates mTORC1. mTORC1 serves as a central nexus for integration of multiple extracellular signals, including oxygen and amino acid levels, growth factors, and stress. Based on these input signals, mTORC1 activity influences cellular growth, metabolism, protein synthesis, and cell cycle progression. One such example is the energy-sensing mechanism of the cell comprised of LKB1 and AMPK. Increasing levels of AMP, a marker of decreased nutrient levels, results in AMPK phosphorylation and activation by LKB1. AMPK phosphorylates TSC2, which, as previously described, inhibits mTORC1 activity, leading to downregulation of protein synthesis and cell growth in response to low nutrient levels.⁴⁴⁹ In part, mTORC1 activation regulates these phenotypes via phosphorylation and activation of p70S6 kinase and inhibition of 4EBP1. The former protein stimulates messenger RNA (mRNA) translation, whereas the latter inhibits translation of mRNA transcripts with a 5′ cap.⁴⁵⁰

Both mTORC1 and S6 kinase also participate in a negative feedback loop in which both proteins activate IRS1, which results in inhibition of insulin-mediated PI3K activation.⁴⁵¹ Akt can also activate mTORC1 in a TSC2-independent manner via phosphorylation of PRAS40, a protein that interacts with mTORC1.

The mechanisms responsible for PI3K pathway activation in cancer are diverse and include activating mutations and amplification of PIK3CA, AKT1, AKT2, and AKT3, loss of PTEN expression/function, loss of TSC1 or TSC2 function, RAS mutation, and dysregulated growth factor receptor and integrin activation, as outlined below.424,452 In human tumors, activation of PI3K is frequently a direct consequence of dysregulated RTK signaling as a result of mutation, amplification, or ligand overexpression. For example, ERBB2 amplification in breast and gastric cancer results in Akt activation.^25,452 Similarly, kinase domain mutations of EGFR induce constitutive Akt activation in lung cancers and glioblastomas, and Akt activity in these tumors is critical for EGFR-mediating transformation.^420,453 Oncogenic RAS mutations also activate PI3K, and PI3K activation is required for Ras-mediated tumorigenesis in genetically engineered mouse models.^{389,454–456}

Mutations in the PIK3CA gene, which encodes the p110-α catalytic subunit of class IA PI3K, are frequently observed in tumors of the colon, breast, brain, stomach, ovary, and other cancers.424,452,457 The most frequent mutations are E542K and E545K, located in the helical domain (exon 9), and H1047R (exon 20), located in the kinase domain.⁴⁵⁸ All three mutants demonstrate increased lipid kinase activity, induce phosphorylation of Akt and its downstream effectors, and can transform chicken embryo fibroblasts.⁴⁵⁹ Exon 9 helical domain mutations block the inhibitory interaction between the p85 regulatory subunit and the p110 subunit and result in constitutive kinase activation.⁴⁶⁰ Exon 20 catalytic domain mutations result in constitutive kinase activation.⁴⁵² PIK3CA mutations commonly co-occur with KRAS mutations and ERBB2 amplification in colon and breast cancers, respectively, and expression of mutant PI3K in breast cell lines, as well as fibroblasts, cause neoplastic transformation.⁴⁶¹ Unlike wild-type p110-α, which lacks oncogenic potential, wild-type overexpression of the other three isoforms can induce transformation of cultured cells.⁴⁶² PIK3R1 encodes the p85 regulatory subunit of PI3K. Alterations within this gene have also been reported in multiple cancers, including glioblastomas and colorectal, endometrial, and ovarian cancers.^402,463

Loss of PTEN function is the most frequently observed PI3K/Akt pathway alteration in human malignancies and is common in tumors of the prostate, breast, ovary, lung, colon, and bladder, as well as in melanomas and glioblastomas.⁴⁵² Loss of PTEN function in tumors is mediated by a diversity of mechanisms including mutation, deletion, posttranslational modification, and promoter hypermethylation.⁴⁵² Dysregulated expression of microRNAs that target the 3′-untranslated region of PTEN has also been shown to induce cell survival and cisplatin resistance.⁴⁶⁴

Because Akt activation enhances proliferation and suppresses apoptosis, constitutive activation would be predicted to have strong oncogenic function. Indeed, Akt was initially identified as a proto-oncogene in the mouse leukemia virus Akt8.⁴⁶⁵ A recurring hot spot mutation in the PH domain of AKT1 (E17K) occurs with low frequency in breast, colorectal, bladder, endometrial, and ovarian cancers.^468–468 This mutation results in constitutive localization of Akt to the plasma membrane without the need for PIP₃ recruitment.⁴⁵⁸ Amplification of the AKT2 gene has been reported in ovarian and pancreatic cancers, and gain of function AKT3 mutations have been reported to occur in melanoma.⁴⁶⁹

Given the significant proportion of malignancies that harbor mutations in the PI3K/AKT/mTOR pathway, a concerted effort is ongoing to identify selective inhibitors of various PI3K pathway components. These agents can be categorized as selective PI3K inhibitors, dual PI3K/mTOR kinase inhibitors, Akt inhibitors, and mTOR inhibitors (Table 2-1). Both isoform-specific and pan-selective inhibitors of PI3K are in development. Although PIK3CA mutations and PTEN loss correlate with a heightened response to some PI3K and Akt inhibitors in preclinical studies, it remains unclear whether mutation status is a clinically useful biomarker of PI3K inhibitor sensitivity.⁴⁷⁰

As previously described, Akt regulates cell growth, proliferation, protein translation, and nutrient metabolism, primarily through the mTOR pathway. Temsirolimus and everolimus, analogues of rapamycin that inhibit the mTORC1 complex, are approved by the FDA for use in patients with renal cell carcinomas, and everolimus is approved for the treatment of patients with pancreatic neuroendocrine tumors (Table 2-1).^471–474 Everolimus also has significant clinical activity in patients with subependymal giant cell astrocytomas that arise in the setting of tuberous sclerosis, an inherited cancer-predisposition syndrome resulting from germline mutations in the TSC1 and TSC2 genes.⁴⁷⁵ Resistance to mTOR inhibition is thought to derive from multiple mechanisms, including activation of parallel signaling networks such as the MAPK pathway.⁴⁷⁶ Furthermore, inhibiting mTORC1 can preferentially lead to an increase in mTORC2 signaling and, as previously described, mTORC2 enhances Akt activation via phosphorylation of the serine 473 residue.⁴⁷⁷

A recent report of a complete response to the mTOR inhibitor everolimus in a patient with metastatic bladder cancer suggests that such agents can induce significant antitumor responses in genetically selected patients.⁴⁷⁸ Whole genome sequencing of this outlier responder’s tumor revealed loss of function alterations in TSC1 and NF2, both of which normally downregulate mTOR activity. Additional responses to everolimus were observed in TSC1 mutant tumors but not to the degree of the complete responder, indicating that co-altered genes likely confer resistance to mTOR inhibitory therapy even in the setting of a predictive biomarker of response. In sum, mTOR inhibitors are likely most effective when used in genetically defined patient populations, with the majority of patients requiring combination therapies. As an example of the latter, everolimus in combination with exemestane was recently approved by the FDA for the treatment of advanced hormone receptor–positive, HER2-negative breast cancer.⁴⁷⁹

Translational Implications

As previously outlined, mutational and epigenetic alterations induce constitutive activation of a broad array of signaling pathways in human tumors. In some instances, the growth and survival of tumor cells have been shown to be exclusively dependent upon a single pathway activated by a mutated oncogene or tumor suppressor, a phenomenon referred to as “oncogene addiction.”⁴⁸⁰ In such instances, targeted inhibitors of such pathways have demonstrated unprecedented clinical activity in molecularly defined subsets of patients (Table 2-1). Examples include imatinib in patients with CML, erlotinib in patients with EGFR mutant non–small cell lung cancer, and vemurafenib in patients with BRAF mutant melanoma.^{28,328,329,411,414,481} Despite these dramatic successes, most patients with cancer have yet to benefit from this approach. Potential explanations for this lack of benefit include the redundant regulation of key downstream mediators of transformation by multiple signaling pathways, the lack of specificity of the drug for the driver alteration, and the continued practice of performing clinical trials of targeted inhibitors in unselected patient populations.

In normal cells, activation of signaling induces negative feedback that serves to rapidly inhibit pathway activation. It has become apparent that abrogation of this physiological feedback is required for transformation mediated by the oncogenes and tumor suppressors previously discussed. “Adaptive resistance,” defined as a rapid reactivation of parallel signaling pathways after relief of negative feedback signals, likely abrogates the effects of selective pathway inhibitors. Because significant drug development efforts are currently focused on the development of targeted inhibitors of oncogene-activated signaling pathways, a detailed understanding of these normal physiological pathways and their dysregulation in cancer will be required of both basic cancer researchers and practicing clinical oncologists.