Sara A. Courtneidge
SUMMARY OF KEY POINTS
During embryonic development and in the adult organism, the fate of a cell is decided by the cues it receives from its surroundings. For example, growth factors instruct cells to divide, and extracellular matrix proteins provide survival signals. Other stimuli can cause cells to migrate, to differentiate, to undergo programmed cell death (apoptosis), or to enter a survival state that involves autophagy. Each of these outcomes is initiated by binding of diverse protein and nonprotein ligands to receptors, most of which are localized on the cell surface. Receptor activation results in the recruitment of adaptor molecules and enzymes, particularly protein and lipid kinases. These recruited proteins then relay signals to the nucleus, the cytoskeleton, and other subcellular compartments to affect the response. Each type of receptor initiates a discrete set of signaling pathways, such that different ligands binding to the same cell can have different effects. Furthermore, the same ligand can have different effects on different cell types because of innate differences in the signaling components present in the cells. Thus, the combinatorial action of several intracellular signaling pathways dynamically controls the responses of cells and organs to external cues from the environment.
The first clues that components of intracellular signaling pathways are important in causing cancer came from research on tumor viruses in model systems such as chickens and mice. Many RNA tumor viruses (retroviruses) contain cancer-causing genes called oncogenes, which derive from host sequences. For example, the first retroviral oncogene to be discovered, src, the transforming gene of Rous sarcoma virus, derives from the cellular src gene, which encodes a protein tyrosine kinase. During the transduction event, 3′ sequences of the src gene were lost, resulting in the production of a protein that lacks the regulatory carboxy-terminus. Thus, while the cellular Src protein is normally tightly regulated in cells, the viral form of the protein is constitutively active and transforming. Other examples of oncogenes that are transduced by retroviruses include ras, myc, sis (encoding the growth factor PDGF), akt, raf, fos, and many more. Equally important lessons were learned from the study of DNA tumor viruses. The oncogenes of these viruses do not derive from the host genome. Rather, in this case, the oncogenes associate with and modify the functions of key signaling proteins. For example, most DNA tumor viruses encode proteins that are able to inactivate negative regulators of signaling, such as p53 and Rb. Many can also activate signaling by binding to proteins such as Src, the platelet-derived growth factor (PDGF) receptor, and cyclins. These oncogene studies therefore provided the early tools necessary to dissect the signaling pathways that control cell growth. Other important information, particularly on the control of cell survival, has come from the study of genetically tractable organisms such as fruit flies and the nematode Caenorhabditis elegans. More recently, whole genome sequencing projects have allowed the direct analysis of human clinical specimens for alterations in key signaling pathways.
With the use of the systems and tools described previously, much progress has been made in the last decades in the characterization of intracellular signaling: Several classes of receptor have been defined,all protein kinases (the kinome) have been described, and some signaling pathways are now known in their entirety. Furthermore, there has been intense study of the perturbations that occur in intracellular signaling during cancer progression. This effort has resulted in the definition of new molecular targets for cancer drug discovery. Indeed, some new drugs that target cellular signaling have recently been approved, and many more are in clinical testing. While not yet fully realized, the promise is that defining and targeting signaling pathways responsible for all aspects of the cancer phenotype will result in more potent and less toxic chemotherapies.
This chapter reviews the basic principles of intracellular signaling. Then some examples of receptors and their mechanisms of activation are given (antigen and other immune receptors are not discussed). This is followed by an overview of some common signal transduction intermediates and some selected examples of signal transduction pathways elicited by certain receptors. Finally there is a brief discussion of how these signaling pathways are dysregulated in cancer and how this might be exploited for targeted therapeutic intervention. This is not intended to be a comprehensive list of all receptors and signals; rather, examples have been chosen that have relevance to the cancer phenotype. The references that are provided are intended to point the reader to more detailed and thorough reviews of the topics covered; primary references are provided only for new discoveries that have not yet been the subject of reviews. Furthermore, many of the examples and themes that are briefly described in this chapter are explored in more detail in later chapters. For example, in Chapter 5 , Craig Thompson and Rebecca Elstrom describe in some detail the control of cell death and in Chapter 4 , Jacqueline Lees reviews how signaling pathways feed into the cell cycle. Furthermore, each of the chapters on specific malignancies contains descriptions of the dysregulated signaling pathways that are involved in each disease.
General Principles of Intracellular Signaling
Intracellular signaling is the mechanism by which cues that are present in the extracellular environment are relayed and interpreted by the cell. These external cues can be growth factors that signal a cell to divide, extracellular matrix proteins that promote survival, hormones that change the metabolism of the cell, cytokines that instruct the cell to differentiate, or other signals that promote motility and invasive ability. Complex multicellular organisms have evolved to have a large array of receptors for these external cues, as well as an even larger number of intracellular signaling molecules. Both embryonic development and adult homeostasis require specificity as well as temporal and spatial control of these signaling pathways. Many disease states, including cancer, diabetes, and immune disorders, can arise if these signaling pathways are inadequately controlled.
Intracellular signaling can at first glance seem overwhelmingly complicated. A given receptor can engage a number of different signaling pathways, each eliciting a distinct phenotype. One signaling pathway can affect the output of another, a phenomenon known as cross-talk. The same ligand can have different effects in different cell types. And the activation of a receptor frequently also elicits negative regulatory pathways that are designed to switch the system off after a defined period of signaling.    But this complexity can be reduced somewhat. Experimental observation has shown that the members of any given family of receptors signal in approximately the same way. Furthermore, many receptors use similar components to signal; for example, cytokine, growth factor, and G protein-coupled receptors all activate the MAP kinase pathway and the PI 3-kinase pathway.   Finally, most signaling pathways make use of adaptor or scaffolding proteins. These proteins lack catalytic activity and instead are made up of multiple protein-protein and sometimes also protein-lipid interaction domains.   In this way, adaptor proteins orchestrate the simultaneous activation of a number of pathways following receptor activation. Signaling specificity can be provided by the recruitment of distinct adaptor proteins by different receptors. A schematic of how signal transduction pathways can involve different components is outlined in Figure 2-1 .
Figure 2-1 Schematic of intracellular signaling. A prototype receptor is shown. Dimerization of the receptor by ligand elicits multiple signaling pathways, whose outputs result in changes in gene expression, in the cytoskeleton, and in metabolism. These signaling pathways also elicit the production of negative regulators, which ultimately turn the signal off.
Receptor Activation by Ligand
Most of the receptors that receive cues from the extracellular environment are found in the plasma membrane of the cell, where they initiate signaling for those peptide and protein ligands that cannot cross the lipid bilayer. In contrast, receptors for lipid soluble ligands are found in the cytoplasm and the nucleus. A broad overview of the molecular makeup and mechanism of activation of several different classes of receptor will be described later.
G Protein-Coupled Receptors
The largest class of plasma membrane receptor is the so-called G protein-coupled receptor (GPCR) family, which has more than 1000 members. Ligands include agents that stimulate neurotransmission, light and taste perception, and cell division and differentiation as well as the chemokines, which are ligands that are involved in cell attraction. All GPCRs share a common architecture, with seven transmembrane α-helical domains, connected by both extracellular and intracellular loops. Because of this architecture, GPCRs are sometimes also referred to as serpentine or heptahelical receptors. The extracellular domain is responsible for ligand binding, which causes a conformational change such that the G protein binding to the intracellular domain becomes activated. G proteins are heterotrimeric proteins that consist of α, β, and γ subunits. Activation causes the dissociation of the a subunit from the βγ complex; both separated components then go on to mediate signaling events that are very similar to those initiated by receptor tyrosine kinases.
Receptor Tyrosine Kinases
The next largest class of receptors is the receptor tyrosine kinases (RTKs).    This class consists of approximately 90 members, most of which are involved in the control of cell growth, motility, and differentiation as well as metabolic control. Examples include PDGF receptors, epidermal growth factor (EGF) receptors, the ephrin receptors, hepatocyte growth factor receptors, fibroblast growth factor (FGF) receptors, insulin receptors, and many more. While the ligands for most classes of RTK are now known, some remain “orphans,” with their ligands yet to be discovered. At least one RTK, Her2 (also known as ErbB2), has no known ligand but instead signals by heterodimerization with other EGFr family members.   Also, by sequence analysis of the catalytic domains, several RTKs are predicted to lack catalytic activity. In one of these cases, Her3, the receptor is transphosphorylated when heterodimerized with other EGFr family members (particularly Her2) and acts as an adaptor protein. The majority of the RTKs are single polypeptide chains that contain an extracellular ligand-binding domain, a short hydrophobic transmembrane domain, and a cytoplasmic region containing the kinase domain, as well as other sequences that regulate recycling and/or turnover of the receptor and interaction with signaling molecules. One exception to this rule is the insulin receptor family, which is composed of a disulfide-bonded tetramer, with two identical extracellular ligand binding subunits and two transmembrane subunits with catalytic activity. Another exception is the Met/Hepatocyte growth factor receptor family, in which a single polypeptide chain is cleaved to produce a stable dimer with ligand binding invested in one subunit and catalytic activity in the other. While it is generally thought that the single membrane pass RTKs are monomeric in the absence of ligand,    there is some recent evidence that EGF receptors can exist in a dimeric but inactive state in the absence of ligand.
Some RTK ligands are dimers (e.g., the PDGFs), while others are monomeric (the EGFs). The FGF receptors require both FGF ligands and heparin sulfate proteoglycans for full activation.   Yet otherRTKs, notably Ret and MuSK, do not bind ligands directly, requiring a coreceptor to present the ligand.   Despite these differences, RTKs generally share a common mechanism of activation on ligand binding, which involves dimerization and often further oligomerization to form higher-order structures. Recent crystallographic studies have revealed the molecular mechanisms behind ligand-induced activation and how monomer ligands can induce dimerization. In the case of the EGFr, the dimerization occurs at an interface between two receptor monomers, the ligand presumably serving to initiate the domain rearrangements that are necessary to make these contacts.   In contrast, FGFs, while monomeric, have two receptor-binding sites. Dimerization is thus achieved by interactions between the receptor and the ligand. The heparin sulfate is required to organize and strengthen the interaction of ligand and receptor as well as to provide specificity to the interactions.   Regardless of how the extracellular domains are brought into close apposition, the result is the juxtapositioning of the two catalytic domains and the subsequent transphosphorylation of tyrosine residues in the activation loop of the kinase domain, as well as elsewhere in the intracellular domain. These phosphorylations serve to initiate further conformational changes that stabilize the active form of the enzyme as well as to provide binding sites for the enzymes and adaptor proteins that are described later.
Serine and Threonine Kinase Receptors
The proteins in this family act as receptors for transforming growth factor (TGF)-β, activins and inhibins, and bone morphogenetic protein. Members of this family of receptors, which together are often known as the TGFβ superfamily, play key roles during development and morphogenesis, as well as in cell cycle progression, motility and wound healing, and immune surveillance. The ligands can be homodimers or heterodimers, and they interact with two single-pass membrane proteins called type I and type II receptors. Both receptor subunits have intrinsic serine/threonine kinase activity in the cytoplasmic domain. Signaling is thought to be initiated by ligand-induced oligomerization of two type I receptors with two type II receptors. The type II receptor, which has constitutive kinase activity, phosphorylates and activates the type I receptor, which then goes on to phosphorylate substrate proteins.
Integrins are heterodimers of an α and a β subunit, each of which is membrane-spanning. Mammals have 18 α subunits and 8 β subunits that can combine to form 24 distinct heterodimers. Integrins derive their name from their ability to bind to ligands outside the cell and cytoskeletal components inside the cell and so integrate the two environments. Their extracellular ligands can be either extracellular matrix proteins or cell surface proteins on neighboring cells. In their low-affinity state, the extracellular domains of the β subunits adopt a “bent” conformation. Ligand binding generates a higher-affinity, straightened structure of the β subunit, corresponding allosteric changes in the extracellular domain of the α subunit, followed by conformational changes in the transmembrane and cytoplasmic domains. Clustering of integrins also occurs because of the multivalent nature of the ligands. The net result is a separation of the cytoplasmic domains of the α and β subunits, which mutational studies have shown is required for this “outside-in” signaling. Integrins do not contain intrinsic kinase activity but instead signal through associated cytoplasmic kinases such as Src and FAK.
Unlike any other class of membrane receptor, integrins can also signal from the cytoplasm to the extracellular space, so-called inside-out signaling. In this form of signaling, cytoplasmic signal transduction pathways, for example, caused by activation of the EGFr or Frizzled receptors, generate a conformational change in the integrin extracellular domain, resulting in increased adhesion to extracellular ligands and subsequent outside-in signaling. Although the exact mechanism by which this conformational change is transmitted was unclear until very recently, it was known that the cytoskeletal protein talin was required. Talin binds to the cytoplasmic tails of most β integrins through its PTB-like domain. Reducing talin expression or preventing its association with the β subunit is sufficient to inhibit inside-out signaling. Talin adopts an autoinhibited structure in quiescent cells. Intracellular signaling pathways release this inhibition, allowing it to bind to and change the conformation of the integrin, thus promoting high-affinity interactions of the integrins with their ligands.
The cytokine receptor superfamily consists of the receptors for growth hormone, prolactin, erythropoietin, thrombopoetin, G-CSF, and the interleukins as well as the interferon receptors.    They are grouped together because of their use of functionally related receptors. Most receptors use a common γ chain, together with variant β chains and sometimes α chains. Structural homologies among these chains include four conserved cysteine residues in the extracellular domain and a WSXWS motif located near the transmembrane domain. The cytoplasmic domains of these subunits lack any catalytic activity but instead have two regions of low homology, called “box 1” and “box 2.” Although the exact functions of these boxes are not known, it is clear from mutagenesis studies that they are necessary for receptor function. Activation of cytokine receptors occurs when ligand binding causes dimerization, perhaps by promoting a rotational switch near the WSXWS motif. Kinases of the JAK family, which are constitutively associated with the cytoplasmic domains, then transphosphorylate the receptors to initiate signaling. The most important substrates of the JAKs are the Stat proteins, which have an SH2 domain and a transcriptional activation domain. Signaling by this class of receptors thus involves tyrosine phosphorylation-induced dimerization of the Stats, which renders them transcriptionally competent.
Another class of cytokine receptors is represented by the tumor necrosis factor (TNF) receptor superfamily, also known as death receptors, which control apoptosis in response to exogenous signals.Members of this family include the TNF receptors 1 and 2, Fas, CD40, and TRAIL and are simple single-pass transmembrane proteins, with cysteine-rich ligand-binding domains on the outside and either a death domain (R1 receptor) or a TRAF-binding domain (R2 domain) on the inside. They have no intrinsic catalytic activity. It was originally thought that TNF activated its receptor by inducing trimerization, but it is now known that this class of receptors breaks the ligand-induced oligomerization rule. Instead, the TNF receptors are found as preformed trimers on the cell surface. This is thought to be required for their function. For example, TNF-α signals through TNF-R1 to initiate cell death and through TNF-R2 to activate NFkB. If the ligand were to initiate the formation of trimers, mixed oligomers would form in any cell that expresses both receptors. Since only a trimer of death domain-containing receptors is functional, this would result in the suppression of death signals. Thus, preassembly of receptor complexes might promote the formation of homotypic receptors and confer more specificity in the response. As with other, more conventional receptors, ligand binding is thought to elicit conformational changes that render the cytoplasmic domains competent for signal transduction.
Frizzled receptors are the binding partners for the Wnt family of ligands. This is a large class of peptides that control a wide array of developmental processes and have also been implicated in cancer. Of particular current interest is the role of Wnts in the control of stem cell fate.   Frizzled receptors have a cysteine-rich extracellular Wnt-binding domain, followed by seven transmembrane domains and a short cytoplasmic tail. Frizzled alone is unable to transduce Wnt signals. Rather, members of the LRP family, which are single-pass transmembrane proteins, are required as coreceptors. The current model for the way in which Wnts activate their receptors is an interesting and novel one in which it is postulated that Wnts bring LRP and Frizzled into close proximity. In support of this model, coexpression of chimeric fusion proteins that cause the close apposition of Frizzled and LRP is sufficient to induce Wnt-independent signal transduction.
There are four mammalian Notch receptors, which play important roles in development and tissue homeostasis, by specifying cell fates and creating boundaries between different cells.     Each receptor is a single-pass transmembrane protein, with an EGF repeat domain that binds ligand in the extracellular domain, and a cytoplasmic domain that lacks catalytic activity but instead contains several ankyrin repeats and a C-terminal PEST domain. During transport through the Golgi, the receptor is cleaved into ligand-binding and transmembrane domains that remain associated with each other through noncovalent interactions. The ligands for Notch are called DSLs (for delta-serrate-lag2 family), and they are also single-pass transmembrane proteins with EGF-like repeats in their extracellular domains. Signaling between Notch and DSLs therefore takes place when one cell carrying the ligand and one cell expressing the receptor are in close apposition. Unlike the other receptors discussed here, in which ligand binding usually controls oligomerization of the receptor and phosphorylation of the cytoplasmic domains, activation of the Notch receptors is accomplished by regulated and specific proteolysis. The binding of ligand to Notch makes the receptor susceptible to cleavage first by an ADAMs family metalloprotease in the extracellular domain and then by a γ-secretase in the transmembrane domain. This frees the intracellular domain of Notch, which translocates to the nucleus, where, together with a DNA-binding protein called CSL and coactivators, it activates transcription. Uniquely among the plasma membrane receptors, then, the Notch receptor family does not use a series of signal transduction pathways to exert their effects but rather affects gene expression directly.
Nuclear Hormone Receptors
Some receptors, particularly those for estradiol, progesterone, androgen, glucocorticoids, polyunsaturated fatty acids, and the retinoids, predominantly exert their effects by acting as transcriptional activators and/or repressors. These nuclear hormone receptors are regulators of many aspects of homeostasis, as well as sexual development. Many have been implicated in disease, including cancer, lipid disorders, and diabetes.   In addition, several are the targets of important drugs; for example, PPARg is the target of the antidiabetes drug rosiglitazone. Most of the nuclear hormone receptors consist of a single polypeptide chain with a DNA-binding domain, a transactivation domain, a ligand-binding domain, and sequences that mediate interaction with coregulators. While some receptors can bind DNA with high affinity as monomers, most require dimerization. This requirement is particularly true of the retinoid receptor family, in which the exact nature of the heterodimers that are formed dictates the transcriptional output. The ligands for the nuclear hormone receptors either diffuse passively into the cell or are produced within the cell during normal metabolism. In some cases, association between receptor and ligand takes place in the nucleus. In other cases, particularly the sex hormone receptors, the receptor is cytoplasmic and bound to chaperone proteins in the absence of ligand. Ligand binding dissociates the chaperones and allows translocation to the nucleus. Once they are nuclear, ligand-dimer complexes interact with DNA and recruit transcriptional regulators to effect their responses. Thus, in this canonical mechanism of action, nuclear hormone receptors as a class do not require intracellular signal transduction pathways. However, there are a growing number of cases, particularly well studied in the case of the estrogen receptor, in which hormone can also bind a plasma membrane receptor.   Although the identity of the membrane receptor is currently not established (some investigators think that it is the same as the nuclear form, whereas others have suggested it is a GPCR), it is clear that this form of the receptor can elicit canonical signaling involving Src, MAP kinases, and lipid kinases in a similar way to RTKs. The roles of these two forms of receptor signaling, which are often called genomic and nongenomic, particularly in cancer causation, awaits further clarification.
Components of Intracellular Signaling Pathways
With the exception of the nuclear hormone receptors and the Notch receptors, ligand-activated receptors do not have direct mechanisms of action. Rather, their activation results in the initiation of one or more signal transduction cascades that ultimately change the phenotype of the cell. Many general principles and some individual elements of these signal transduction cascades are shared by different classes of receptor. Some of the more commonly used strategies are described in overview here.
Perhaps one of most common features shared by most receptor signal transduction cascades is the use of adaptor proteins, which are defined as proteins with protein-protein interaction and sometimes also protein-lipid interaction domains. Common protein interaction motifs include SH2 and some PTB domains, which bind to phosphotyrosine-containing peptides; SH3 domains, which bind to proline-rich ligands; WW and WD40 domains, which bind phosphorylated serine and threonine residues, respectively  ; and PDZ domains, which bind carboxy-terminal sequences of proteins.   In each case, neighboring amino acids provide specificity to the interaction. For example, the SH2 domain of Grb2 preferentially binds a phosphotyrosine followed by the sequence XN, where X is any amino acids and N is asparagine. The SH2 domain of Src, in contrast, prefers the sequence Y(p)EEI. PH, PX, and FYVE domains bind phosphoinositides, particularly those phosphorylated in the 3, 4, and/or 5 positions by phosphatidylinositol (PI) kinases.   
Since most adaptor proteins have multiple interaction domains, they act as scaffolds that cluster together distinct signaling molecules. Adaptor proteins can serve to give specificity to receptor signaling by clustering distinct sets of signaling proteins in a particular subcellular location. In some instances, the recruited proteins might modify each other's activity, for example, when a kinase phosphorylates a coassociated protein. Some adaptor proteins are selective for certain receptor-signaling systems. For example, death domain-containing adaptors are used exclusively by the superfamily of TNF receptors,  the insulin receptor substrate family of proteins is dedicated to the insulin receptor and cytokine receptor families,   and FRS2 is involved in FGF and nerve growth factor signaling.Others, such as Grb2 and Shc and the Gab, Dok, and Vav families, are used by multiple receptor types.
Cytoplasmic Tyrosine Kinases
Some receptors have intrinsic tyrosine kinase activity, while others, such as GPCRs and cytokine receptors, do not. Yet each of these receptor classes uses tyrosine phosphorylation as a signal. They do this by recruiting, or being stably associated with, cytoplasmic tyrosine kinases (CTKs). There are several subfamilies of CTKs, the most important for signal transduction being the Src family, the Abl family, the Tec family, the FAK family, and the JAK family. These kinases become activated by association with the receptor, usually through dimerization and transphosphorylation. In these cases, the CTKs can be considered to be noncovalently associated receptor subunits that provide the required catalytic activity. But it is interesting that even the RTKs recruit CTKs, which is required for their full function.   An example of this, the recruitment of Src family kinases by the PDGFr, is given later. While these observations led some researchers to speculate that RTKs were able to phosphorylate only associated proteins and therefore required CTKs to phosphorylate all downstream signaling components, it now seems clear that different tyrosine kinase families have broadly different substrate specificities and that some signal transduction pathways are dependent on the substrates phosphorylated by the RTK and some on the CTK substrates. In keeping with this, cytokine receptors and GPCRs also recruit and/or activate more than one class of CTK.
Cytoplasmic Serine and Threonine Kinases
There are 518 protein kinases in the human genome. Of these, 90 are tyrosine kinases, either RTKs or cytoplasmic kinases as described previously. All other members of the kinome phosphorylate serine and threonine residues. While many of these enzymes have housekeeping functions in the control of metabolism, DNA replication, and so on, many more are obligate proximal members of signal transduction cascades from receptors. The best studied of these kinases are the ones that make up what is called the MAP kinase (MAPK) pathway. MAPKs are small single-subunit serine/threonine kinases. They have a number of substrates, but chief among them are transcription factors, whose subcellular localization and activity are regulated by MAPK phosphorylation. MAPKs fall into three classes: the ERKs, the JNKs, and the p38 family. All of these enzymes are normally inactive in quiescent cells, but become activated upon growth and/or stress stimulation of the cells. This activation is accomplished by enzymes that are generally known as MAPK kinases (MKKs). These are dual-specificity enzymes; they phosphorylate MAPKs on both a tyrosine and a threonine in the activation loop phosphorylation, causing a large increase in catalytic activity of the MAPKs. While there is some cross-talk, different MKKs appear to be somewhat selective in activating the different classes of MAPKs (Fig. 2-2 ). The MKKs in turn need to be phosphorylated to be activated, which is accomplished by a broad range of serine/threonine kinases, generally called MKK kinases (MKKKs), from a number of different families. While the combinatorial complexity of these cascades of kinases could be staggering, recent evidence suggests that scaffold proteins exist that provide specificity and regulation to each cascade.
Figure 2-2 Mitogen-activated protein kinase (MAPK) cascades. Intracellular signaling from various receptors involves activation of members of the mitogen-activated kinases. The MAPKs consist of three families of related kinases (Erks, p38s, and Jnks). Activation of MAPKs is mediated by phosphorylation by a series of related kinases, MKKs (MAPK kinases), which are activated by a broad spectrum of kinases through phosphorylation. Activated MAPKs translocate to the nucleus and phosphorylate a variety of structurally unrelated transcription factors. The transcription factors and gene expression are thus activated.
The best studied of all the MAPK pathways, and one that is central to all mitogenic signaling from receptors, is the one known as the Ras-Raf-MAPK pathway. The Ras family consists of three small GTPases: H-, N-, and K-Ras. Of these, K-Ras has been implicated in many human cancers. For example, more that 90% of pancreatic cancers and approximately 50% of colon cancers have an activating mutation in K-Ras.      In unstimulated cells, Ras is found in an inactive, GDP-bound form. Many ligands activate Ras, by stimulating the exchange of GDP for GTP, which is accomplished by activating proteins called guanine nucleotide exchange factors and inhibiting proteins called GTPase-activating proteins. In the case of growth factor signaling, the guanine nucleotide exchange factor that is involved is called Sos, which is usually associated in the cytoplasm with an adaptor protein called Grb2. The GTPase-activating protein that is involved is p120 Gap. On receptor activation, two different types of adaptor protein, called Shc and Gab, become recruited to receptors in the membrane and phosphorylated on several tyrosine residues. Several of these sites are in YxN motifs, which are canonical binding sites for the Grb2 SH2 domain. Thus, receptor activation results in the following cascade of signaling: Shc (or Gab) > Grb2:Sos > Ras. Once Ras is in the GTP-bound state, it initiates a number of different signaling pathways, including PI 3-kinase activation, and cytoskeletal changes.     But of particular importance, it recruits the serine/threonine kinase Raf to the plasma membrane, where other signals, including tyrosine phosphorylation, activate its intrinsic kinase activity. Raf then goes on to phosphorylate and activate MEK 1 and 2, which in turn activate ERK1 and 2.
Space does not allow a full description of all serine/threonine kinases that are involved in signal transduction here. But there is one enzyme complex that has attracted much recent attention: mTor (mammalian target of rapamycin).       mTor is a signal integrator, linking information about nutrients, energy status, and growth factor stimulation to outputs such as protein synthesis, ribosome biogenesis, metabolism, and cell survival and proliferation ( Fig. 2-3 ). There are two structurally distinct mTor complexes, with proteins known as Raptor and Rictor. The mTor/Raptor complex is regulated by a small GTPase known as Rheb, which is normally kept in an inactive state by a GTPase-activating protein consisting of a complex of two tumor suppressors TSC1 and TSC2. A variety of signals, including signals from Akt, ERKs, and other serine/threonine kinases, inactivate the TSC1/2 complex, thus activating Rheb and in turn mTor/Raptor. The signaling outputs from this complex include mRNA translation, ribosome biogenesis, and autophagy through phosphorylation of substrates such as 4E-BP1 and ribosomal S6K. The mTor/Rictor complex is less well understood, but it is activated by RTKs and plays an important role in activating Akt.
Figure 2-3 mTor and PI 3-K signaling. The figure illustrates some of the key components of the mTor and PI 3-K signaling pathways and highlights the central role that the serine/threonine kinases Akt and mTor play in the control of several intracellular signaling events. The three classes of PI 3-K are activated by a variety of cell surface receptors and make different lipid products, which bind to lipid-binding domains in a variety of signaling molecules. Lipid binding activates signaling by recruiting the signaling molecules to intracellular or plasma membranes or by causing a conformational change. Once Akt is activated by PDK1 and other serine/threonine kinases, it phosphorylates a number of substrates to cause profound changes in cell homeostasis. mTor responds to changes in nutrients and growth factors and exerts its effects via forming stable complexes with either Rictor or Raptor.
As we saw earlier, many adaptor proteins have domains that interact with phosphorylated phosphoinositides, which frequently serves to recruit them to the plasma membrane or subcellular organelles such as endosomes. In other cases, phosphoinositide binding serves to activate the catalytic activity of the enzyme to which it is bound. The most important lipid modifiers in signal transduction pathways are a family of enzymes known as phosphatidylinositol 3-kinases (PI 3-Ks).   These come in three classes, based on sequence alignment and substrate preference. Class I enzymes have two subunits, one with catalytic activity and one that specifies the association of the enzyme with other signaling molecules. They generally generate PI 3,4,5-P3 (PIP3) from PI 4,5-P2. Class II enzymes have a single subunit, which contains both catalytic activity and regulatory sequences. They predominantly generate PI 3-P and PI 3,4-P2. Class III has just one member, the Vps34 protein, which was originally identified as a gene required for vacuolar sorting in budding yeast and generates PI 3-P. Relatively little is known about the functions of the class II and class III enzymes in mammals, although Vps34 has recently been implicated in mTor regulation and autophagy, suggesting a role in the response of cells to nutrients. Class II enzymes bind clathrin, suggesting a possible role in receptor trafficking and/or endocytosis. But it is the class I family members that have been most intensively studied for their roles in intracellular signaling pathways and cancer.     
Class I enzymes can be further subdivided into Class IA and Class IB. All members have structurally related catalytic subunits. But the regulatory subunits of class IA members all contain 2 SH2 domains, and some also have an SH3 domain, whereas in class IB, the regulatory subunits are structurally distinct. Class IA members are activated by RTKs, whereas class IB members are activated by GPCRs. In all cases, activation occurs when the PI 3-K is recruited to the membrane by regulatory domain sequences, for example, the SH2 domains in the case of class IA. This brings the enzyme into close proximity with its substrate and allows catalysis to take place. PIP3 then acts as a lipid second messenger by binding to PH domains in a variety of proteins, including the serine/threonine kinases PDK1 and Akt (seeFig. 2-3 ). One important target of Akt is the FOXO family of transcription factors, which are sequestered in the cytoplasm and inactivated by phosphorylation, thus inhibiting gluconeogenesis. Other metabolic targets include the glucose transporter Glut4, glycogen synthase kinase 3, and ATP citrate lyase. Akt also has effects on the cell cycle via its inhibitory effects on FOXO and GSK3 as well as by directly phosphorylating proapoptotic proteins such as BAD. In concert with the small GTPases Rac and Cdc42, class I enzymes also control actin dynamics and thus regulate cell polarity and motility.
Negative Regulators of Signaling
Given the power of signal transduction pathways to regulate all aspects of a cell's phenotype, it is vital that these pathways be tightly controlled. Indeed, in the absence of such control, cancer often arises. We have seen that some control is provided in the recruitment and activation of defined signaling modulators. But there are also other layers of control that ensure that the signal is switched off in a timely manner. A number of different mechanisms are used. The first is at the level of the receptor itself. Following ligand binding, most receptors are internalized via mechanisms using clathrin-coated pits and endocytosis.     In some cases, this results in degradation of both receptor and ligand in lysosomes; in other cases, the unoccupied receptor is recycled back to the plasma membrane. Both RTKs and some cytoplasmic tyrosine kinases are also regulated by ubiquitin-mediated degradation at the proteasome.   In the case of the EGFr and Src, it is the activation of these proteins that initiates the degradation by recruiting an E3 ubiquitin ligase called Cbl via SH2 domain interactions.     This ensures that the proteasome selectivity degrades the active forms of the enzymes. Two other inhibitory mechanisms are also elicited by the very signaling pathways that they inactivate. For example, the activation of MAPK-signaling pathways results in the transcriptional activation of a class of enzymes called MAPK phosphatases (MPKs), which, as the name suggests, dephosphorylate and inactivate MAPKs.   Also, the JAK-Stat signaling pathway results in the production of the SOCS (suppressor of cytokine signaling) proteins, which have an SH2 domain and a region known as a SOCS box. The SOCS proteins are thought to inhibit signaling by binding to specific phosphotyrosine-containing motifs on receptors and the JAK kinases and perhaps initiating ubiquitin-mediated degradation.   Finally, the action of both phosphoprotein and phospholipid phosphatases can also serve to regulate signaling. Of particular importance is the phosphoinositide phosphatase and tumor suppressor known as PTEN, whose loss has been associated with malignancy in many cancers, particularly glioblastomas and prostate cancer.  
Selected Examples of Signal Transduction Pathways
Throughout this chapter, we have touched on the concept that many extracellular ligands engage multiple signaling pathways and that there are some common themes to the signaling outputs that emerge. For example, stimulation with most ligands results in profound changes in gene transcription by activating transcriptional regulators such as Fos, Myc, NFkB, Stats, Smads, hormone receptors, and so on. Many factors elicit changes in the cytoskeleton, polarity, and movement of cells. Alterations in metabolism, hormone responses, differentiation, and survival are other common outputs. In this section, a few examples of receptor-initiated signal transduction pathways will be given, illustrating the concepts that have been discussed throughout.
Receptor Tyrosine Kinase PDGFr
There are five PDGF ligands encoded by four different genes (A, B, C and D), and two different receptors encoded by unique genes (α and β). Each ligand is a dimer, with known isoforms including AA, AB, BB, CC, and DD. PDGF AA can bind only to a homodimeric αα receptor, whereas AB, BB, and CC can bind both αα and αβ receptors. PDGF DD binds predominantly to ββ receptors. PDGF acts predominantly on cells on mesenchymal origin and promotes cell growth, migration, and survival. Alterations in PDGFr signaling play a role in sarcomas and gliomas.   Furthermore, chromosomal translocations that cause the constitutive activation of the kinase domain of PDGFr are found in some leukemias and lymphomas.
Once activated by dimerization and transphosphorylated on several tyrosine residues, the PDGFr recruits at least 10 different SH2 domain-containing signaling molecules to these sites ( Fig. 2-4 ).    One class of binding proteins are members of the Src family of tyrosine kinases (SFKs). These enzymes become activated by this association and then phosphorylate a number of other signaling molecules. While the mechanistic details have yet to be worked out, one important effect of SFK activation is the stabilization of myc mRNA and subsequent transcription. Another important signaling cassette to be recruited is the class IA PI 3-K, which associates directly with the receptor via the SH2 domains on its regulator subunit; the signaling cascades that were discussed earlier are then initiated. Recruitment and phosphorylation of the adaptor protein Shc, together with the subsequent binding of Grb2/Sos, allow the Ras/MAPK pathway to become activated. Other signaling effectors that are more selective for RTKs such as the PDGFr include phospholipase Cg, the adaptor protein Nck, and the tyrosine phosphatase Shp2, as well as the negative regulators Ras-GAP and Cbl.
Figure 2-4 Recruitment of SH2 domain-containing proteins by activated PDGF receptors. The figure illustrates how the ligand PDGF, which is a dimer, causes the dimerization and transphosphorylation of its receptor on a number of tyrosine residues, in the juxtamembrane region, in the kinase insert region, on the activation loop, and on the carboxy-terminal tail sequences. Each one of these phosphorylated tyrosines then binds selectively to SH2 domain-containing proteins, which then initiate a number of different intracellular signaling pathways. The selectivity of SH2 domain binding to each tyrosine is determined by the amino acid sequences immediately proximal to each tyrosine.
It is important to note that the example of signaling that is illustrated here—the recruitment of multiple signaling cascades both to the receptor itself and to adaptor proteins—is a general theme that is seen throughout RTK signaling. Specificity is provided by the nature of the phosphorylation sites on the receptor and the domain makeup of the adaptor proteins that are recruited, as well as the complement of signaling proteins that are expressed in any given cell type. Furthermore, other receptors, such as integrins and GPCRs, use very similar strategies. In these cases, cytoplasmic tyrosine kinases (FAK and Src in the case of integrins and Src and Tec family members in the case of GPCRs) are responsible for generating the SH2 domain recruitment sites.
As we saw earlier, the receptors in this family are transmembrane serine/threonine kinases, with ligand stimulating the phosphorylation of the type I receptor by the type II subunit, which activates the intrinsic kinase activity of the type I subunit. The best-characterized intracellular effectors of TGFβ signaling are the Smad proteins.      There are three types of Smads: R-Smads, a common Smad (in vertebrates this is known as Smad4), and inhibitory Smads. The R-Smads and Smad4 have two conserved domains called MH1 and MH2, joined by linker sequences. In addition, the R-Smads have carboxy-terminal serine phosphorylation sites. On ligand binding, the receptors recruit and phosphorylate one or more R-Smads. The conformational changes that are induced by this phosphorylation release the R-Smads from the receptor and allow the formation of a trimeric complex of two R-Smads with Smad4. This complex translocates to the nucleus, where, along with coactivators such as CBP/p300 and factors such as FOXO and Forkhead, transcription is activated. The specificity of the transcriptional response is provided by the R-Smads that are present in the complex and thus the coactivators and transcription factors that are recruited. In turn, recruitment of individual R-Smads is determined by the exact nature of the heterotetramer receptors that are formed in response to the diverse TGFβ family ligands. In the cases in which the ligand is TGFβ itself or activin, efficient recruitment of R-Smads to the receptor also requires an adaptor protein called SARA, which has a FYVE domain that interacts with PI 3-P, and a Smad binding site.   The inhibitory Smads interfere with signaling by binding to type I receptors and interfering with R-Smad recruitment.
Although the Smads are the only known mechanism by which TGFβ receptors modulate transcription, other signal transduction pathways also control Smad activation. For example, R-Smads are phosphorylated by MAPKs, which may prevent nuclear localization of the Smads and transcriptional responses. In contrast, phosphorylation by JNK enhances nuclear translocation and activity of Smad3. Other kinases that have been reported to phosphorylate Smads include cell cycle kinases such as Cdk2 and Cdk4, PKC, and CaMKII. In this way, signaling from other receptors can greatly influence the transcriptional responses to TGFβ receptor stimulation.
The death receptors are the means by which extracellular signals are linked to the apoptotic machinery of the cell, a process known as instructional apoptosis. One of the best characterized of these receptors is called Fas (also known as CD95 or Apo1); it will be used as illustration.     Trimerization of Fas by its ligand clusters the receptor's death domains, leading to the recruitment of the protein FADD by death domain interaction. The death effector domain in FADD then binds to the inactive, zymogen form of caspase-8. The oligomerization of caspase-8 that ensues causes the self-activation of caspase-8 by proteolysis, resulting in the subsequent cleavage and activation of effector caspases such as caspase-9 and commitment to apoptosis.
The last example to be considered is that of Wnt signaling through Frizzled and its coreceptors, the LRPs. One of the most important signaling components in Wnt signaling is the β-catenin protein, which accumulates in the cells in response to the ligand and activates transcription by associating with the DNA-binding protein TCF.    In unstimulated cells, TCF is in a complex with the negative regulator Groucho and is transcriptionally inactive. In these same cells, β-catenin turns over rapidly, a consequence of its recruitment by a so-called destruction complex. This complex contains Axin, APC, and GSK3, a serine/threonine kinase. Phosphorylation of β-catenin by GSK3 causes β-catenin to become ubiquitinated and targeted for destruction by the proteasome. The binding of Wnt to its ligand causes the phosphorylation of LRP by GSK3 and CK1 and the recruitment of the Axin/APC/GSK3 complex to the receptor ( Fig. 2-5 ). This is presumably sufficient to prevent the phosphorylation of β-catenin, thus allowing its concentration to rise, enter the nucleus to associate with TCF, and activate transcription. Another key intermediate in Wnt signaling is Dsh, which is required upstream of Axin/APC/GSK3. Dsh is also a cytoplasmic protein that becomes recruited to the Frizzle/LRP complex in response to Wnts, but the exact mechanism by which it participates in signaling is unknown at this time.
Figure 2-5 Canonical intracellular signaling from Wnt receptors. In unstimulated cells, a complex of Axin, APC, and GSK (glycogen synthase kinase) phosphorylates β-catenin, which causes its degradation via the proteasome pathway. When Wnt ligand is present, Frizzled and Lrp associate and create binding site for axin, APC, and GSK3 in cooperation with disheveled (Dsh). Phosphorylation of β-catenin is thus prevented, which allows its accumulation and transport into the nucleus. Once in the nucleus, β-catenin displaces the association of Groucho with the transcriptional activator TCF, allowing transcription of several target genes. In some cancer cells, particularly from the colon, mutations in APC prevent the downregulation of β-catenin and thus allow constitutive TCF signaling.
In an intriguing new twist to the story of Wnt signaling, it was recently found that Wnts can act as ligands for the atypical RTKs, Ryk, Ror1, and Ror2.    Little is yet known about the signal transduction pathways that are elicited by Wnt2 binding to these RTKs or about the cellular outputs they specify.
CLINICAL RELEVANCE AND APPLICATIONS
From research conducted over the last two decades, it is clear that many human cancers have their origins in dysregulated signaling pathways.    Some of the earliest oncogenic events to be discovered in humans were mutated and activated K-Ras, chromosomal translocations that result in the overexpression of Myc and the production of the Bcr-Abl tyrosine kinase, and chromosomal amplification of the RTK Her2. Since then, many other activating mutations in signaling pathway components have been discovered in cancer cells. Examples include activating mutations in the EGFr in some non-small-cell lung cancers, in β-Raf in most melanomas as well as other tumor types, in Jak2 in myelodysplastic syndromes,   and in PI 3-K, particularly in breast cancers.  Furthermore, chromosomal translocations involving the RTKs Kit, Flt3, and PDGFr are detected in gastrointestinal stromal tumors and some lymphomas. Other mechanisms that cancer cells use to promote their growth and survival include the overexpression of the antiapoptotic protein Bcl-2 in lymphoma,   activating mutations in G proteins in pituitary tumors,   the acquisition of insensitivity to the inhibitory growth effects of TGFβ while maintaining the positive signals in many carcinomas,    and the overexpression of estrogen or androgen receptors in breast and prostate cancers, respectively.      
The discovery of each of these activated signaling pathways has been rapidly followed by attempts by both the academic community and the pharmaceutical industry to develop new therapeutics targeting these events.     Some of these attempts have yet to be successful despite enormous effort; for example, the Ras GTPase has so far proved intractable to small molecule inhibition.   But for other targets, there have been successes, the most notable being the development of imatinib (target: Abl, PDGFr) and later dasatinib (target: Abl, SFKs) for the treatment of chronic myelogenous leukemia (see Chapter 108 ). Other therapeutics that target signal transduction pathways include antiestrogens and antiandrogens for breast and prostate cancer, respectively  ; trastuzumab for Her2-positive breast cancer; gefitinib and erlotinib (target: EGFr) for lung cancer; sorafenib (targets: VEGFr and Raf) for renal cell carcinoma; and sunitinib (targets: VEGFr, PDGFr, Kit) for renal cell carcinoma and gastrointestinal stromal tumor.  
It can be anticipated that in the future, most cancers will be classified not just according to anatomic site of origin and histopathology, but also on the basis of the genetic alterations that are present in the tumor, including those in signaling pathways. New therapeutic modalities are likely to include strategies (small molecules, antibodies, microRNA, etc.) to target these activated signaling pathways.