Tumor Growth and Cell Proliferation In Vivo

Fundamentally, cancer is a disease of accumulation of clonal cells. Abnormal cell proliferation is necessary, although often insufficient, for tumorigenesis. It is the increase in tumor cell number, and thus tumor burden, that ultimately accounts for the adverse effects on the host. Indeed, the goal of most current cancer therapy is to reduce the number of tumor cells and to prevent their further accumulation. To better accomplish this goal, a more complete description of the unique characteristics of tumor cell proliferation is required. This task is made difficult by the fact that the mechanisms that underlie tumor and normal cell proliferation are very similar. In this section, we will review current understanding of the complex molecular mechanisms involved in the regulation of cell proliferation with particular emphasis on the aberrations in this system that occur in malignant cells.

From the perspective of biologic evolution, it is obvious that cells within a multicellular organism like humans have an intrinsic proliferative potential that is in vast excess of that required to meet the requirements of normal growth and development. Cellular life has evolved from single-celled organisms selected for their ability to replicate in a minimum amount of time, within the constraints imposed by the biochemical processes of cell division. Standard laboratory strains of bacteria, for example, can divide every 20 minutes under optimal conditions. The constraints of cell division imposed on multicellular organisms are greater since replication must be carried out with absolute fidelity to maintain the integrity of the organism. Even so, normal human cells can divide as often as once or twice a day in vivo. A cell dividing once a day would generate a cell number equal to the total number of cells in an adult human in less than 2 months. Clearly, then, human cells have inherited a surplus of proliferative capacity from their unicellular ancestors. Multicellular organisms must evolve mechanisms to restrain this proliferative capacity to appropriate times and places. The key in understanding tumor cell proliferation, then, is to characterize these mechanisms and to understand how they fail during tumorigenesis.

The rate of cell proliferation within any population of cells depends on three parameters: (a) the rate of cell division (Tc), (b) the fraction of cells within the population undergoing cell division (growth fraction), and (c) the rate of cell loss from the population due to terminal differentiation or cell death (see next section). Tc represents the time it takes to complete a cell division cycle. The cell division cycle can be divided into two functional phases, S and M phases, and two preparatory phases, G1 and G2 (Fig. 2.1). S phase is defined as the phase in which the DNA is replicated. Under normal circumstances, the time it takes a typical human cell to complete S phase is about 8 hours and is invariant. Fully replicated chromosomes are segregated to each of the two daughter nuclei by the process of mitosis during M phase. The length of M phase is about 1 hour and is also normally invariant. G1 phase precedes S phase, whereas G2 phase precedes M phase. G1 and G2 phases are required for the synthesis of cellular constituents needed to support the following phase and ultimately to complete cell division. In mammalian cells, the length of G2 phase is about 2 hours. The length of G1 phase is highly variable and can range from about 6 hours to several days or longer. The varying length of G1 phase accounts for most of the difference in Tc between different cell types or between cells growing under different conditions.

Figure 2.1

The cell cycle. When a cell is not synthesizing DNA (S phase) or completing mitosis (M phase), it is commonly termed as being a G (gap) phase. Normal cells are capable of resting in a nondividing state, called G0. They can begin one or more cycles of (more...)

Cell Cycle Control

A successful cell division cycle requires the orderly and unidirectional transition from one cell cycle phase to the next. Certain events must be completed before others are begun. For example, beginning mitosis before the completion of DNA replication would obviously be deleterious to the cell. In theory, the ordering of cell cycle events may be accomplished in a manner analogous to the substrate-product relationship of a metabolic, biochemical pathway. ¹ The product of one reaction serves as the substrate and is thus required for the next reaction. Hence, regulation of the system is inherent in the biochemical events of the process itself. The prevailing view, however, is that the timing and ordering of cell cycle transitions is dependent on separate positive and negative regulatory circuits. The regulatory circuits enforce a series of checkpoints, allowing passage only after completion of critical cell cycle events. Two classes of regulatory circuits exist, intrinsic and extrinsic. Intrinsic regulatory pathways are responsible for the precise ordering of cell cycle events. Since the length of S, G2, and M phases in mammalian cells is relatively invariant, the transitions between these phases are controlled predominantly by intrinsic regulatory pathways. Extrinsic regulatory pathways function in response to environmental conditions or in response to detected cell cycle defects. Both types of regulatory circuits can use the same checkpoints. We will focus our attention on the extrinsic regulatory circuits where differences between normal and neoplastic cells are observed.

Passage of the cell cycle checkpoints ultimately requires the activation of intracellular enzymes known as cyclin-dependent kinases (CDKs). CDKs are extremely well conserved through evolution. CDKs exist in all eukaryotic cells from fungi to plants to mammals. In fact, CDKs from human cells can functionally substitute for the enzymes in yeast. The structural and functional conservation of these enzymes through evolution suggests that they are centrally important for the cell cycle in all eukaryotic cells. The requirement for these enzymes for cell cycle transitions has been amply documented, particularly in organisms like yeast that are amenable to genetic manipulation. ¹ Since activation of CDKs is the central event in cell cycle transitions, it is not surprising that their activity is exquisitely regulated at several levels. ² The active CDK holoenzyme is composed of a catalytic subunit and the cyclin regulatory subunit. One level of regulation is that each cyclin protein is synthesized at a particular stage of the cell cycle. For example, cyclin D is synthesized during G1, cyclin E is synthesized in late G1, cyclin A is synthesized during S and G2 phases, and cyclin B is synthesized in G2 and M phases (Fig. 2.2). Therefore, a given catalytic subunit cannot become active until an appropriate cyclin is synthesized. Upon synthesis, a cyclin can assemble with an appropriate catalytic subunit. However, this complex requires phosphorylation on threonine by another regulated kinase, the CDK-activating kinase or CAK. CAK is itself a CDK composed of cyclin H and CDK7 proteins. Hence the levels of CAK influence the activity of assembled CDKs. Another level of regulation is deactivation of the CDK by phosphorylation of its ATP binding site by yet another regulated kinase activity. This kinase activity is unusual in that it has dual specificity for both tyrosine and threonine. A CDK deactivated by phosphorylation of its ATP binding site can be reactivated by a dual specificity phosphatase of the Cdc25 family. In fact, dephosphorylation by these phosphatases may be the rate-limiting step in triggering cell cycle transitions. Another level of regulation is the presence of a diverse family of proteins known as cyclin-dependent kinase inhibitors, or CKIs, that can block activation of CDKs. Two distinct classes of CKIs have been described. One class inhibits multiple CDKs and includes p21CIP1, p27KIP1, and p57KIP2. The other class specifically inhibits cyclin D/CDK4 or 6 CDKs and includes p16INK4, p15INK4B, p18INK4C, and p19INK4D. The synthesis, degradation, and activity of these CKIs are regulated in response to both mitogenic and antimitogenic signals. For example, cell cycle regulation by cell-cell contact or transforming growth factor-𝛃 (TGF-𝛃) is mediated by p27KIP1. ^{3, 4} Once activated, the CDKs that drive the transition into a particular cell cycle phase often need to be deactivated before completion of that phase and transition to the ensuing phase. For example, the CDKs required for initiation of mitosis also prevent exit from mitosis and into G1 phase. The final level of CDK regulation involves their specific degradation in precise order. It is now generally understood that ubiquitin-mediated proteolysis is responsible for this regulation as well as the regulation of a host of other cell cycle regulators. ⁵ Hence, synthesis, post-translational modification, and programed degradation all contribute to the regulation of CDKs.

Figure 2.2

Cyclin-dependent kinase regulation of cell cycle transitions. A. The phases of the cell division cycle are shown. Transition from one phase to the next requires transit of a checkpoint, like the restriction point (R), during the G0/G1 to S phase transition. (more...)

The G0 to S Checkpoints

As discussed above, the time it takes to progress through the S, G2, and M phases of the cell cycle is relatively invariant. The length of G1 phase on the other hand is variable. In addition, cells can exit the cell cycle for extended periods of time and mammalian cells do so during the G1 phase of the cell cycle. Cells that have exited the cell cycle are said to be in a G0 state, or “quiescent.” Most cells in adults are in G0. This absence from the cell division cycle can be temporary or permanent, as is the case with terminally differentiated cells like neurons. Cells in G0 can be very active functionally and metabolically, and proliferation of G0 cells can be initiated by changes in cell density, the presence of mitogens or growth factors, or the supply of nutrients. These cells then enter the cell cycle, beginning a sequence of events that culminates in cell division. Hence, the G0/G1 to S phase transition is highly regulated, and the result of this regulation, by and large, determines the Tc and growth fraction of a population of cells.

Like other cell cycle transitions, the transition from G0 to S phase as cells re-enter the cell cycle is regulated by two major checkpoints: competence and the restriction point (R). These checkpoints are located approximately 12 and 2 hours before the start of the S phase, respectively. At least three growth factors, provided in serum, are required sequentially to transit these checkpoints following resumption of proliferation of fibroblasts: platelet-derived growth factor (PDGF), epidermal growth factor (EGF), and insulin-like growth factor (IGF-1). As mentioned above, extracellular TGF-𝛃 has the opposite effect. It inhibits the growth of various epithelial cells by modulating expression of CKIs. Paracrine production of TGF-𝛃 could limit growth of both normal and cancer cells, and experimental models suggest that it may play a role in the regression of breast cancers in response to hormonal or drug therapies. ⁶ Once the R point has been passed, the cell is committed to a round of cell division. The cell now completes S, G2, and M phases without the need for growth factors or even additional protein synthesis. Once initiated, the cell cycle is not free running; the competence and restriction checkpoints must be passed in each subsequent G1 phase, thus requiring the continued presence of growth factors. The switching of cells back and forth between quiescence and cycling depends on extracellular conditions and is regulated differently in normal and tumor cells.

Growth factor receptors are complex, large proteins that span the plasma membrane. They have a specific domain that recognizes the growth factor on the outside of the cell, and their cytoplasmic portion may have an enzymatic function, such as a protein tyrosine kinase. Binding of a growth factor or ligand to its receptor can induce transmission of a signal to the cytoplasm through activation of the kinase. ⁷ The next step is a transduction of the cytoplasmic signal to the cell nucleus. This is accomplished by a heterogeneous group of molecules known as second messengers and includes various proteins that are phosphorylated by kinases, small molecules such as inositol phosphates and cyclic AMP, and ions, including Ca++, H+, and Zn++. Within the nucleus, genes are then activated in response to these second messengers. As an illustration of this general scheme, upon binding of EGF to the extracellular domains of its receptor, autophosphorylation occurs in the intracellular domain of the protein. This phosphorylated domain facilitates the formation of a protein complex containing Grb2 and Sos. This complex activates the Ras protein by catalyzing exchange of Ras-bound GDP for GTP. The GTP-bound form of Ras activates the c-Raf kinase. This kinase then triggers a phosphorylation cascade ultimately activating mitogen-activated protein kinase. This kinase may phosphorylate and regulate transcription factors such as Jun, Fos, and Myc.

A variety of proteins are produced during G1 after cells leave quiescence. Some are enzymes that expand metabolic functions lost by G0 cells, such as those providing energy, and more ribosomes are made for rapid protein synthesis. Others have so-called housekeeping functions that keep both quiescent and growing cells in metabolic balance. Only a few proteins appear to be key regulatory molecules. For example, enzymes are required for the synthesis of isoprenoids, which are necessary for activity of the Ras oncogene, and for the synthesis of polyamines, which have many functions including ionic binding to nucleic acids. The Ras oncogene product is synthesized as a precursor protein that requires post-translational processing to become biologically active and capable of transforming mammalian cells. Farnesylation appears to be a critical modification of the Ras protein, and drugs that inhibit farnesyl-protein transferase can block Ras-dependent transformation. These agents have been proposed as a potential new class of therapy for cancer. ⁸ Enzymes involved in the synthesis of DNA, such as thymidine kinase and DNA polymerase, as well as histones, are synthesized just prior to the S phase. These enzyme molecules relocate at the beginning of DNA synthesis, moving from the cytoplasm into the nucleus. A variety of experiments show that DNA is made by a high-molecular-weight, multienzyme complex. ^{9, 10} This complex contains many enzymes known to be involved in the process of DNA replication, but its size and other features are still a matter of debate. The onset of DNA replication has been investigated recently with in vitro systems, and these studies reveal that the synthesis of helicase enzymes, which possess DNA-unwinding ability, may provide the final factor for initiating the S phase. ¹¹ After DNA synthesis has commenced, cell growth becomes relatively independent of external controls. The daughter cells, now in the G1 phase, will then either pass through another cycle or arrest in a quiescent G0 state, depending, once more, on external conditions. If these conditions are not adequate, the cell will become arrested before it reinitiates DNA synthesis.

How is re-entry into the cell cycle from G0 ultimately controlled? Like other cell cycle transitions, activation of CDKs are required. G0 cells are devoid of significant CDK activity. In the presence of mitogenic growth factors, expression of D-type cyclins (cyclins D1, D2, and D3) is stimulated and continues throughout G1 phase as long as the growth factors are present. ¹² D-type cyclins complex with either CDK4 or CDK6 catalytic subunits to form a holoenzyme modified by CAK. All of the relevant substrates for cyclin D/CDK4 or 6 have probably not been enumerated. However, one important substrate is likely the retinoblastoma tumor-suppressor protein (Rb). Rb is constitutively expressed and constrains cells from progressing through the G1 phase of the cell cycle. ¹³ Rb complexes with many cellular proteins including the E2F transcription factors. When in complex with E2F, Rb represses transcription from E2F-dependent promoters. Upon phosphorylation by cyclin D/CDK4 or 6, ¹⁴ Rb loses its ability to restrain the cell cycle. ¹⁵ This response is presumably because it can no longer complex with E2F and represses E2F-dependent transcription. ¹⁶ The E2F family contains at least five members (E2F-1 through E2F-5). The E2F proteins function as transcriptional activators when in heterodimeric complex with one of the E2F-related proteins DP-1, 2, or 3. The heterodimeric complex binds a specific DNA sequence and activates transcription from the promoters of many genes important for S phase including dihydrofolate reductase, DNA polymerase α, and thymidine kinase. ¹⁷ Perhaps most importantly, E2F influences the expression of cyclin E. ¹⁸ In fact, of many E2F-dependent genes, cyclin E is the only one deregulated upon loss of Rb in normal cells. ¹⁹ Cyclin E expression begins in late G1 phase and complexes with CDK2. Cyclin E/CDK2 activity is necessary and sufficient for the start of S phase. ²⁰ Forced expression of cyclin E/CDK2 activity can trigger the start of S phase in the absence of Rb phosphorylation and derepression of other E2F-dependent genes, suggesting that cyclin E is the primary target for regulation by cyclin D/CDK4 or 6, Rb, and E2F. ^{21, 22}

DNA Damage-Induced Checkpoints

When nuclear DNA is damaged, normal cells initiate a response that includes cell cycle arrest, apoptotic cell death, and transcriptional induction of genes involved in DNA repair. Induction of apoptosis is an important response to DNA damage and is discussed in detail below. Normal cells in G1 phase prior to the R point will arrest in G1 phase upon sensing DNA damage. This arrest is presumably induced to prevent the replication of damaged DNA. Replication of damaged DNA can result in the incorporation of heritable genetic mutations. If cells are past the R point or within S phase, DNA replication is slowed, again to allow time for DNA repair. If cells sense DNA damage while in G2 phase, a G2 cell cycle arrest will occur. Different types of DNA damage can interfere with normal mitosis, resulting in heritable genetic mutations or cell death.

The tumor-suppressor genes ATM and p53 play an important part in responses to damaged DNA. ^{23, 24} For example, cells containing mutations in p53 fail to arrest in G1 or undergo apoptosis efficiently upon irradiation. Cells containing mutations in ATM are also deficient for cell cycle arrest as well as some forms of DNA repair. The p53 protein functions as a transcription factor by binding specific DNA sequences and regulating transcription from promoters containing those sequences. In normal cells, DNA damage induces an increase in p53 levels by inhibiting the normal rapid turnover of the protein. The p53 protein is normally targeted for ubiquitin-dependent proteolysis by association with the Mdm2 protein. This association is inhibited by phosphorylation of p53 on specific amino-terminal residues that is triggered by DNA damage. ²⁵ Phosphorylation on these amino terminal residues also facilitates dephosphorylation or acetylation of p53 carboxy-terminal residues. These modifications increase the affinity of p53 for its DNA binding site by distinct mechanisms ^{26, 27} and hence increase its ability to activate transcription. The transcription of a number of genes can be affected by activation of p53. However, the ability of p53 to directly increase expression of p21CIP1 is probably important for p53-dependent G1 cell cycle arrest observed upon DNA damage. As discussed above, p21CIP1 is a CKI that can inhibit the activity of multiple CDKs, including cyclin D/CDK4 or 6 as well as cyclin E/CDK2. In summary, DNA damage generates a signal that can activate p53 by post-translational modification. Increased p53 activity upregulates p21CIP1, which prevents activation of CDKs, required for the G1 to S transition.

ATM is the gene whose mutation is responsible for ataxiatelangiectasia. Immunodeficiency, progressive cerebellar ataxia, radiosensitivity, cell cycle checkpoint defects, and cancer predisposition characterize this disease. ATM encodes a protein containing a phosphatidyl-inositol 3-kinase-like domain, implicating it in signal transduction. Like p53 mutant cells, mutant ATM cells are defective in the G1/S checkpoint activated after radiation-induced DNA damage. This defect is attributable to the lack of p53 activation that normally occurs, suggesting that ATM may participate in the same pathway as p53. ATM protein, and the related ATR protein, can, in fact, associate with and phosphorylate p53 at its amino-terminal sites. ^{28, 29} The p53-directed kinase activity of the ATM protein is itself activated by DNA damage. ATM protein, therefore, contributes to the activation and stabilization of p53 by phosphorylating amino-terminal sites during the radiation-induced DNA damage response. ATM protein may play a role in sensing DNA damage and generating the DNA damage signal.

Environmental agents like radiation or DNA-damaging chemicals most commonly induce DNA damage. Rarely, DNA damage can be generated by mistakes in the normal execution of the cell cycle. However, a form of DNA damage eventually occurs in all normal cells as they suffer replicative senescence. Normal cells have a limited replicative lifespan both in vivo and in vitro; a cell can undergo only a finite number of cell divisions. This limit is thought to be imposed, at least in part, by levels of telomerase activity. ³⁰ Telomerase is an enzymatic activity within cells that is required to maintain the integrity of DNA ends. ³¹ DNA polymerases involved in DNA replication synthesize DNA in the 5′ to 3′ direction and require a primer and template. The requirement for a primer ensures that some genetic information will be lost from the 5′ end of DNA during each round of DNA replication. Telomerase adds DNA of a particular sequence to the ends of DNA without the need for a separate primer or template, thus protecting cells from loss of genetic information. Telomeric DNA also protects chromosomes from degradation or recombination. Without telomeric DNA, chromosomes become unstable. As normal cells become senescent, they lose telomerase activity and their cell division cycle is arrested. This cell cycle arrest may be mediated by the DNA damage checkpoint since shortened telomeric DNA is associated with DNA strand breaks that may be sensed as damaged DNA. Consistent with this hypothesis, shortening of telomeric DNA triggers a p53-dependent cell cycle arrest by accumulation of single stranded DNA. ³²

Checkpoint Defects in Tumor Cells

To maintain tissue homeostasis and to support normal development, each organ maintains tight controls over Tc, growth fraction, and cell loss. Physiologic stimuli can alter these parameters in normal tissues, leading to increased tissue growth, but this growth will cease when the stimulus is withdrawn or a new steady state is achieved. In contrast to normal cells, however, tumor cells continue to proliferate even in the absence of proliferative signals. Although tumor cells proliferate under inappropriate conditions, they do not necessarily proliferate faster than normal cells. In fact, some normal tissues grow faster than cancers under physiologic conditions (Table 2.1). Biopsy samples from normal, inflammatory, and neoplastic lesions of the lung, cervix, vocal cord, or pharynx have been analyzed for the rate of cell proliferation; these studies showed that benign inflammatory lesions can grow over 20 times faster than cancer in a discrete time and place. ^{33– 35} Similarly, rapid proliferation of human lymphoid cells is induced by immunostimulants, and growth kinetics of these cells are similar to those observed in high-grade lymphomas. ³⁶ So, it is not simply rapid growth at a single time and place that distinguishes neoplasia but rather growth that is not restrained to appropriate times and places.

Table 2.1

Growth Parameters of Human Neoplasms and Normal Tissues.

It generally is believed that neoplastic cells multiply exponentially during the early phases of tumor cell growth. As the tumor mass increases, however, the rate of growth declines. Measuring tumor growth over time describes a curve with an exponential increase in the early period, then a flattening out of the growth rate over time (i.e., Gompertzian curve). ³⁷ Several mechanisms have been invoked to explain this change in growth rate with larger tumors: (a) decrease in the growth fraction, (b) increase in cell loss (i.e., exfoliation, necrosis), (c) nutritional depletion of tumor cells resulting from outgrowth of available blood supply, or (d) lengthening of Tc. Experimental tumor models suggest that cell cycle time changes only slightly when tumor growth decreases. ³⁸ Under adverse conditions, tumor cells often leave the growth fraction and enter a nongrowing state (G0 or prolonged G1) (see Fig. 2.1), although these same cells can re-enter the division cycle when conditions improve or when stimulated by growth factors. Therefore, the mass doubling time of tumors is correlated with the growth fraction (Table 2.2).

Table 2.2

Correlation between Mass Doubling Time and Growth Fraction.

The biochemistry of growth appears to be very similar qualitatively in tumor and normal cells. ³⁹ Despite numerous efforts, universal differences in biochemical machinery have not yet been discovered between normal and tumor cells. The fundamental difference probably lies in a relaxation of the regulation of cell growth. ^{9, 38} For example, normal cells generally are quiescent at physiologic levels of growth factors, whereas related tumor cells are able to proliferate under these conditions. In some experimental models, tumor cells proliferate in the absence of or at very low levels of growth factors. Further, fibroblast-derived tumor cells are less sensitive than normal cells to the presence of other cells in their immediate vicinity. Normal cells typically cease proliferation when the in vitro culture becomes confluent, but tumor cells can reach several-fold higher densities in culture. Also, cells of normal solid tissue lie on a secreted extracellular matrix (ECM) that is composed of various proteins that stimulate cell growth. ⁴⁰ Tumor cells often are partly or completely independent of ECM for optimal growth, and they may secrete little matrix material. ⁴¹

What molecular defects bring about the relaxed growth requirements in neoplastic cells? Defects can occur at several levels. For example, limiting growth factors may not be needed because tumor cells inappropriately produce their own (i.e., an autocrine mechanism). Alternatively, receptors may be produced in excess, as is the case for EGF receptors in numerous clinical tumors, leading to adequate stimulation at the low growth factor concentrations found in vivo. Moreover, mutations that alter intracellular signaling mechanisms may bypass growth factor dependence. Mutated forms of proto-oncogenes and inactivated tumor-suppressor genes can activate growth in these ways. We will focus here on defects that occur in cell cycle regulatory proteins that enforce the checkpoints discussed above.

Like normal cells, the transit of cell cycle checkpoints in cancer cells ultimately requires the activation of CDKs. Due to the complexity of CDK regulation, defects leading to inappropriate activation of CDKs can occur at several levels. The overexpression of cyclin D1 has been detected in many human cancers due to gene amplification or translocation of the cyclin D1 gene. ⁴² The cyclin D1 gene is located on chromosome 11q13. This chromosomal region is amplified in a wide variety of human cancers including small-cell lung tumors (10%), primary breast cancers (13%), bladder cancer (15%), esophageal carcinoma (34%), and squamous cell carcinoma of the head and neck (43%) among others. ⁴³ Of course, other potential oncogenes could be contained within the amplified region. However, cyclin D1 is likely important since its expression is consistently elevated in these tumors. Cyclin D1 overexpression can also be observed in tumors, such as sarcomas, colorectal tumors, and melanomas, without amplification of the gene. In some cases, cyclin D1 expression is activated by chromosomal translocation. In parathyroid adenoma, Motokura and colleagues. ⁴⁴ have identified cyclin D1 as being translocated to the parathyroid hormone gene, thereby deregulating cyclin D1 expression. Translocation of the cyclin D1 gene with immunoglobulin heavy chain gene transcriptional control elements has also been observed in B cell lineage mantle cell lymphomas. Cyclin D1 is a growth factor responsive cyclin that plays an important role in regulating the G0/S checkpoint. Deregulated expression of cyclin D1 could inappropriately increase cyclin D1/CDK4 activity and drive transit of the checkpoint even in the absence of growth factors. Direct evidence that forced expression of cyclin D1 can facilitate tumorigenesis has been obtained from transgenic mice in which overexpression of cyclin D1 has been targeted to the mammary epithelium. These mice develop ductal hyperproliferation and eventual mammary tumor formation. ⁴⁵

CDK activation can also be accomplished by inactivation of CKIs. Genetic mutation of CKI genes has also been observed frequently in human cancer. In this scenario, loss of a CKI relieves one constraint on the activation of CDKs and provides a proliferation stimulus. In particular, the INK4 locus within chromosomal region 9p21 is one of the most frequently mutated areas in human cancers. ⁴⁶ This locus is also frequently methylated in some tumor types including bladder cancer and leukemia. Extensive methylation of DNA prevents efficient transcription of genes within the methylated region, thus silencing gene expression. Three proteins are encoded by the INK4 locus including the CKIs p16INK4a and p15INK4b as well as p19ARF (see below). It is likely that p16INK4a is a bona fide tumor-suppressor gene since many of the mutations detected in tumors specifically target expression of this protein, and because germline mutations that specifically map to p16INK4a have been detected in kindreds with familial melanoma and pancreatic adenocarcinoma. In addition, mutations in CDK4 that prevent binding with p16INK4a, thus relieving it of p16-mediated inhibition, have also been found in melanoma-prone families. Loss of p16INK4a may facilitate activation of cyclin D1/CDK4 or 6, which is likely to affect regulation of the G0/S checkpoint. Mutation of other CKIs in human cancer is rare, suggesting that they may be required for execution of the cell cycle. However, expression of p27KIP1 is inversely correlated with clinical outcome in a limited number of cancers, including melanoma and carcinoma of the oral cavity.

CDK activation also requires dephosphorylation of inhibitory threonine/tyrosine phosphorylation sites by the Cdc25 family of dual specificity phosphatases. In vitro evidence exists that the Cdc25 family memers are potential oncogenes. ⁴⁷ Forced expression of Cdc25 can cooperate with Ha-Ras or loss of Rb to induce oncogenic transformation of primary cells. Overexpression of Cdc25 has also been detected in some primary human tumors. Cdc25A may be a direct transcriptional target for the myc oncogene. ⁴⁸ Inappropriately high Cdc25 levels may provide an oncogenic stimulus by inappropriately activating CDK activity.

One of the most important genes involved in human cancer is the Rb tumor-suppressor gene. An interesting feature of retinoblastoma is that close to 40% of cases are hereditary, and susceptibility to retinoblastoma is inherited as a simple autosomal dominant trait with high (90%) penetrance. The simple genetics of retinoblastoma has provided the means to molecularly clone the gene responsible; mutational inactivation of both alleles of Rb is necessary and sufficient for retinoblastoma. ⁴⁹ Mutation of Rb is observed at high frequency in osteosarcoma and soft-tissue sarcoma as well. Rb mutations can also be detected in a wide variety of clinically important cancers including carcinoma of the breast, prostate, bladder, kidney, liver, pancreas, cervix, and lung, as well as leukemia. Further, expression of wild-type Rb cDNA in cancer cells can inhibit their tumorigenicity. ⁵⁰ As mentioned above, cyclin D/CDK4 or 6 phosphorylation, which, in turn, is regulated by p16INK4a, inhibits Rb function. This finding suggests that these three proteins function in the same biochemical pathway (Fig. 2.3). Support for this functional interrelation comes from the observation that deregulation of any one of these proteins greatly decreases the likelihood of detecting defects in the other proteins. For example, tumor cells that lose p16INK4a or overexpress cyclin D1 generally retain wild-type Rb. Cells lacking wild-type Rb typically express normal levels of cyclin D1 and p16INK4a. In addition, induction of cell cycle arrest by forced expression of p16INK4a only occurs in cells that contain functional Rb. If mutations in any of the members of this pathway are considered, disruption of this p16INK4a/cyclin D1/CDK4 or 6/Rb pathway may occur in most human cancers. Since this pathway is important for regulation of the G0 to S phase transition, it has a major influence on the growth fraction of normal tissues.

Figure 2.3

The Rb and p53 growth control pathways. Underphosphorylated and active Rb in complex with transcription factors like E2F represses the transcription of genes required for entry into S phase. Upon mitogenic stimulation, synthesis of cyclin D increases (more...)

The p53 gene is the most frequently mutated gene in human cancer. ⁵¹ Germ line p53 mutation is involved in the cancer-prone Li-Fraumeni syndrome. ⁵² Mice lacking p53 due to genetically engineered disruption are also cancer prone. Wild-type p53 is critically important for operation of the DNA damage-induced checkpoint (see above). Upon sensing DNA damage, p53 is activated, resulting in either G1 cell cycle arrest or apoptosis. These responses either allow time for the cell to repair the damage or to rid the body of cells with damaged DNA. Loss of p53 function, therefore, decreases genomic stability. Loss of genomic stability can increase the accumulation of additional genetic mutations required for neoplastic transformation. The Mdm2 gene encodes a protein that binds p53 and targets it destruction by the ubiquitin-proteosome pathway. Too much Mdm2 protein may be analogous to p53 inactivation since any p53 synthesized would be rapidly degraded. Mdm2 was originally identified as an oncogene amplified in a spontaneously transformed mouse cell line. Overexpression of Mdm2 mediated by gene amplification can also be detected in human cancer, particularly sarcoma. ⁵³ Interestingly, the p19ARF protein encoded by the INK4a locus also regulates p53 function. ⁵⁴ The p19ARF protein can bind Mdm2 and prevent Mdm2 from targeting p53 for degradation. Consistent with the ability of p19ARF to activate p53, forced expression of p19ARF can cause a p53-dependent cell cycle arrest. As discussed previously, mutations of the INK4 locus that inactivate p19ARF, as well as p16INK4a, are commonly observed in human cancer. Inactivation of p19ARF may contribute to tumorigenesis since Mdm2-mediated degradation of p53 would be unimpeded. The functional interrelation between p19ARF, Mdm2, and p53 defines another cell cycle checkpoint control pathway (see Fig. 2.3.) Deficiencies in this pathway also play a vital role in neoplastic transformation.

Although cancer cells use the same cell cycle machinery as normal cells, the cell cycle checkpoints in tumor cells are relaxed. Of the scores of proto-oncogenes and tumor-suppressor genes that have been identified to date, most function in signal transduction pathways that mediate mitogenic stimulation. These signal transduction pathways eventually converge on the cell cycle checkpoint that controlsthe G0/G1 to S phase transition and activate appropriate CDKs.Influencing the transit of this checkpoint has a major influence on the proliferation of normal and tumor cells by affecting both Tc and growth fraction. Despite the number and variety of these genes involved in signal transduction, relaxation of the G1/G0 to S checkpoint controls in tumor cells is mediated, for the most part, by disruption of two pathways, the Rb and p53 growth control pathways. These two genes, individually, are the most frequently mutated in human cancer cells. Disruption of the Rb or p53 pathways probably occurs in virtually every human cancer.

Differentiation and Cell Proliferation

Differentiation begins shortly after the first few cell divisions that follow fertilization. Throughout development, and in adult organisms, the ability of a cell to proliferate is intimately connected to its state of differentiation. Adult tissues generally express a variety of factors that act to maintain both the proliferation and the differentiation status of the cells. These include secreted molecules, transmembrane receptors, intracellular signaling molecules, and transcription factors. For example, myoD ⁵⁷ and c/EBP-a ⁵⁸ are nuclear factors that activate the transcription of muscle- and adipocyte-specific genes, respectively; in addition, both proteins are potent inhibitors of cell proliferation.

In early embryos, cell proliferation is the primary means by which the cell mass increases. As the organism develops, however, proliferation becomes restricted. Some differentiated cells continue to proliferate, but others irreversibly lose this ability. Embryonic cells often display traits that confer on them a selective growth advantage over that of an adult cell. They proliferate vigorously, are capable of extensive migration, secrete factors that increase the local supply of blood, and produce enzymes capable of degrading basement membranes. These traits also are characteristic of tumor cells, including the ability to increase local blood supply (i.e., angiogenesis). Recent data suggest that tumor angiogenesis is an important, negative prognostic indicator for carcinomas and leukemias. ^{59, 60} Angiogenesis is now being investigated as a potential target for cancer therapy. ⁶¹ Thus, in adult organisms, mutations or conditions that activate portions of embryonic programs for gene expression or inactivate portions of the adult program can produce cells with many properties of malignant tumor cells. ⁶²

Stem Cells

Stem cells have the capacity for both self-renewal (i.e., proliferation without a change in phenotype) and differentiation (i.e., changing into a new phenotype). Some stem cells have already undergone considerable differentiation, so further differentiation is restricted to a single cell type or lineage. Other stem cells are multipotent and differentiate into a variety of cell types (i.e., hematopoietic stem cells). It has been difficult to demonstrate cells in adults that are totipotent (i.e., capable of differentiating into most or all of the cell types that comprise the organism), but the recent cloning of animals from mature cells demonstrates the persistence of stem cell characteristics even in fully differential cells. ⁶³ Also, recently, neuronal stem cells were shown to produce a variety of blood cell types ⁶⁴ and adult human mesenchymal stem cells that are present in adult marrow were shown to have the potential to differentiate to lineages of mesenchymal tissues, including bone, cartilage, fat, tendon, muscle, and marrow stroma. ⁶⁵ Conversely, primitive hematopoietic stem cells can give rise to muscle cells. ⁶⁸

In general, stem cell differentiation results in two types of changes: the expression of specialized, differentiation-specific gene products and a partial or complete restriction of the cell’s capacity for further proliferation. It then follows that another mechanism by which tumor cells might arise is through mutations that render a stem cell partly or wholly unable to differentiate.

Some cells, particularly in adults, are terminally differentiated. These cells are irreversibly blocked in their ability to proliferate, although they may perform specialized functions for a long period of time. Tumors of terminally differentiated cells are not found. Thus, tumors of mature muscle or nerve cells do not occur, although tumors of less differentiated myoblastic or neuronal stem cells do. Cell proliferation appears to be incompatible with the expression of a terminally differentiated program of gene expression. Thus, irreversible arrest of cell division and expression of the terminally differentiated phenotype are interdependent.

In many tissues, continuous proliferation is restricted to a subpopulation of cells, the stem cells, which undergo self-renewal, as well as differentiation, into cell types with a more restrictive proliferative potential. It then follows that mutations or conditions that interfere with the differentiation of stem cells will result in unbalanced proliferation and, thus, uncontrolled growth of the tissue. Mutations that drive proliferation are associated with an accumulation and overgrowth of less-differentiated cells in the tissue. A common feature of tumor cells is their failure to differentiate terminally under appropriate conditions either in vivo or in culture. ^{67– 69}

Extracellular Factors That Control Differentiation

During embryogenesis and in a number of adult tissues, differentiation depends on external factors. These include insoluble factors such as ECM and both the proximity and type of neighboring cells as well as a growing list of soluble factors. In model systems, differentiation can be induced by a variety of biologic agents and drugs (Table 2.3.). Both the ECM and differentiation-promoting soluble factors may be produced in an autocrine or paracrine fashion.

Table 2.3

Induction of Differentiation in Culture.

Cell-cell and cell-ECM interactions are important for both the induction and maintenance of differentiation in several cell lineages. Although our understanding at a molecular level of insoluble factors is still incomplete, progress has been made in identifying key molecules and pathways through which these factors act. In the case of the ECM, specific cell surface receptors bind to particular components of the ECM. ⁷⁰ It now appears that the binding of an ECM component to its cellular receptor activates an intracellular signal transduction pathway that is analogous to the signaling pathways that have been identified for polypeptide GFs and growth inhibitors. Tumor cells often lose their ability to sense the ECM or neighboring cells. ⁷¹

The soluble factors that regulate differentiation can be broadly classified into those that bind to cell surface receptors and those that freely cross the plasma membrane and bind to cytoplasmic or nuclear receptors. The first class includes molecules such as the fibroblast growth factors (FGFs) (TGF-𝛃 and TGF-𝛃) and hematopoietic factors such as colony-stimulating factor-1 (CSF-1), granulocyte colony-stimulating factor (G-CSF), granulocyte-macrophage colony-stimulating factor (GM-CSF), stem cell factor (SCF), Flt-3 Ligand, and the interleukins. These are all polypeptides, and many were first identified as GFs or growth inhibitors. It now is clear, however, that these factors have a multitude of effects depending on the target cells and the cellular microenvironment. ⁷²

For example, basic FGF was identified as a fibroblast mitogen in brain and pituitary extracts, but recent data suggest that FGF induces mesodermal differentiation in early embryos, is angiogenic, and is a survival factor for endothelial cells. ^{73, 74} FGF also inhibits the differentiation of some cells. Terminal differentiation into mature myotubes cannot occur unless it is withdrawn from proliferating myoblasts. Similarly, TGF-𝛃 was first identified as a stimulator of anchorageindependent growth in mesenchymal cells and later as an inhibitor of epithelial cell proliferation. ^{75, 76} Like FGF, TGF-𝛃 stimulates the differentiation of some cells (i.e., keratinocytes or intestinal epithelial cells) but inhibits differentiation in others (i.e., myoblasts or preadipocytes). In some human tumor cells cultured in vitro or in athymic mice, TGF-𝛃 both inhibits tumor growth and promotes a more differentiated phenotype in the remaining cells. Other studies suggest that TGF-𝛃 induces the expression of one or more inhibitors of cyclin-dependent protein kinases (e.g., p21, p27, p16); inhibition of these kinases, in turn, prevents the phosphorylation and, thus, inactivation of the RB protein, thereby inhibiting cell proliferation. ^{77– 80}

The membrane-permeable regulators of differentiation include retinoic acid (i.e., vitamin A) and its derivatives (RA). ⁸¹ There is strong evidence that concentration gradients of RA are critical for the morphogenesis of some tissues in the early embryo. ⁶³ RA can stimulate or inhibit growth and differentiation depending on the cell type. In general, RA is required for the differentiation of many epithelial cells. It diffuses freely into cells, whereupon it binds to specific nuclear protein receptors (i.e., the retinoic acid receptors [RARs]). In addition, other nuclear proteins, called RAR coregulators, have been found that interact with RARs and modulate their actions in various cell types. ⁸² These differences may explain why specific cells and tissues differ in their responses to RA. Some differentiation-specific genes that are regulated by RA contain sequences specific to the initiation of transcription, ⁸³ to which RA-RAR complexes bind and thereby activate transcription. ^{63, 84} The sex steroids, estrogen and testosterone, may regulate differentiation by similar mechanisms.

Tumor cells often produce factors that affect both growth and differentiation. Basic FGF can confer neoplastic properties when expressed in an inappropriate cell type (e.g., a fibroblast). In addition, inappropriate expression of FGF by one cell may stimulate the growth and affect the differentiation of neighboring cell types. ⁷⁴

Intracellular Regulators

External factors and intrinsic programs of gene expression control cellular differentiation. In either case, the expression of differentiation-specific genes generally is under the control of a small number of master regulatory genes. Genes recently have been identified that are potential “master regulators” of developmental stages and differentiation-specific gene expression. The most globally acting master regulatory genes are known as homeotic genes, which were first identified as genetic loci that determined the developmental and spatial fates of cells in embryos of the fruit fly Drosophila. Similar genes have been identified in the genomes of higher organisms, including humans. Individual homeotic genes are expressed at different times during development and also are expressed in different adult tissues. Some homeotic genes code for extracellular factors, whereas others code for nuclear proteins that are probably transcriptional regulatory factors. Homeotic genes regulate programs of differentiation as opposed to individual differentiation-specific genes. They appear to act by initiating cascades of gene expression that involve regulatory genes having a more restricted range of actions. ^{85, 86} Some homeotic genes may function as tumor suppressors in normal tissues; others may promote tumorigenesis when mutated or deregulated. Mutations in homeotic genes may reactivate portions of an embryonic program of gene expression or suppress portions of an adult program of gene expression. ^{87– 89}

DNA Methylation

In many cases, cells must go through one or more rounds of DNA replication before they can differentiate. This requirement may be because there often is a need to modify the pattern of DNA methylation before differentiation begins. DNA methylation in eukaryotes involves addition of a methyl group to the carbon/5 position of the cytosine ring. ⁹⁰ Changes in DNA methylation commonly are introduced during DNA replication. The methylation of DNA on specific cytosine residues is believed to contribute to the changes in gene expression that occur during development. Presumably, DNA methylation affects gene expression because the transcriptional regulatory proteins that bind to methylated DNA differ from those that bind to unmethylated DNA. Many neoplastic tissues are hypomethylated relative to their normal counterparts, ^{91, 92} and, indeed, pharmacologic agents that alter the pattern of DNA methylation induce differentiation in a number of cultured cell lines. DNA methylation is probably not a universal mechanism for differentiation, however, and some cells can be induced to differentiate with either minimal or no change in cell cycle progression. ⁹³

Differentiation and Cancer Therapy

Analysis of differentiation by tumor cells often provides valuable information for both the diagnosis and therapy of human cancers. As tumor cells grow and die, they can release glycoproteins and other products similar to those of fetal tissues, and these oncofetal products can be detected in serum or other body fluids to assist in diagnosis, follow-up, and selection of therapies. Examples include estrogen receptors and α-lactalbumin in breast cancer, prostate-specific antigen and prostatic acid phosphatase in prostate cancers, and myoglobin and desmin in sarcomas. ⁹⁴

Elevations of these markers in the serum often predict relapse of the neoplasm before any sign by routine examination or radiographic tests. In general, the specificity of such markers for a given neoplasm is poor because minor elevations also occur with inflammatory and other benign conditions or with several types of neoplasms. In carcinoma of unknown primary site, tissue markers for neuroendocrine differentiation select for a subgroup of patients with improved response to chemotherapy. ⁹⁵

Some tumor cells can be induced to differentiate terminally. This has been shown most extensively in cultured cell lines (see Table 2.3) but also in experimental animals. ^{80, 96, 97} After tumor cells have been induced to undergo terminal differentiation, their ability to grow as a tumor often is stably suppressed. In contrast to most anticancer drugs, which have nonspecific toxicity to both normal and cancer cells, drug-induced differentiation can be demonstrated with agents (or drug levels) that exert minimal effects on normal cells. ⁹⁶ These observations have stimulated increased interest in clinical applications of differentiating agents to provide therapeutic gain with minimal toxicity. ¹³⁹

A number of agents known to induce differentiation in various model systems have been used clinically (see Table 2.3). Some are useful only in a particular type of tumor; for example, estrogens and androgens have been useful in treating some breast, prostate, and gynecologic tumors, providing that tumor cells express the appropriate nuclear receptor. Other differentiation-inducing drugs have been more widely studied. For example, high doses of RA, hexamethylene bisacetamide (HMBA), or 5-azacytidine, which is an inhibitor of DNA methylation, can induce differentiation and inhibit the growth of several types of tumors in laboratory models. ⁹⁷ HMBA was found to be active in patients with myelodysplastic syndromes, ⁹⁸ and complete remission was achieved with direct evidence of terminal differentiation of leukemic cells to clonal granulocytes.

In fresh cultures of human promyelocytic leukemia, retinoid-induced differentiation, similar to the effects seen in passaged leukemia cell lines, has been observed. ⁹⁹ All-trans-retinoic acid has been used in clinical therapy of promyelocytic leukemia with promising results, ^{100, 101} and retinoids have been used in combination with interferon-α to produce responses in patients with squamous cell cancer of the skin or cervix. ¹⁰² Because differentiating agents may inhibit tumor cell growth by multiple mechanisms, it is difficult to prove specific differentiating actions of these agents when used in patients. ^{101, 103} For example, retinoids not only induce differentiation in leukemia; they also downregulate anti-apoptotic genes. ¹⁰⁴ An exception are studies in leukemias where sequential samples are readily available, differentiation markers are well established, and the clonality of differentiated cells can be ascertained by methods such as molecular cytogenetics (FISH). ⁹⁹

Retinoids and other differentiating agents also are used in clinical trials to prevent cancer in patients with premalignant lesions or a high risk for developing cancer of the breast, cervix, colon, skin, lung, or oral cavity. Early results are encouraging, including the reversal of oral leukoplakia and prevention of second neoplasms in patients with treated squamous cell carcinoma of the head and neck. ^{103, 105} Greater knowledge about the molecular basis for the control of differentiation should lead to more accurate predictions, however, as well as rational design of therapies for controlling tumor growth by manipulating the state of differentiation. ^{97, 106, 107}

Central Role of Caspases in Apoptosis

Caspases are zymogens: they exist as inactive polypeptides that can be activated by removal of the regulatory prodomain and assembled into the active heteromeric protease. Currently, the caspase family consists of 13 members (see Table 2.4). They encompass a death domain (DD), a death effector domain (DED), and a caspase-recruitment domain (CARD). ¹¹³ The DD is present in members of the TNF receptor family and is involved in the early events of the signaling pathway. The DED and CARD are critical in the downstream portion of the pathways by recruiting caspases to the plasma membrane before their activation. Recent studies have shown that the apoptotic cascade triggered by cytochrome C and dATP is mediated by binding of caspase-9 to Apaf-1 through CARD/CARD interactions. ¹¹⁴ Caspase-9 becomes activated and, in turn, activates and cleaves caspase-3. NMR spectorscopy data provide evidence that basic/acidic surface polarity in the CARD domain is highly conserved and may represent a general mode for CARD/CARD interaction. ¹¹⁵ Caspase-9 deletion in knockout mice prevents activation of caspase-3 in embryonic brains in vivo, leading to perinatal death with a markedly enlarged and malformed cereburm. ¹¹⁶ Caspase-9-deficient thymocytes show resistance to dexamethasone but not to Fas-mediated apoptosis, implicating a functional diversification of caspase cascades, depending on the external stimulus.

Caspases can be grouped into three subfamilies based on their specificities. Group I, or ICE, subfamily of caspases (caspase-1, -4, and -5) prefer the tetrapeptide sequence WEHD and are believed to play a role mainly in inflammation, whereas members of group II (caspases-2, -3, and -7) and group III (caspases-6, -8, -9, and -10) display specificity for DEXD and (I/L/V)EXD, respectively, and are mainly involved in apoptosis. ^{117– 119} The finding that caspases-8 and -10 each contain two N-terminal located DEDs that enable them to associate with death receptors has placed these two caspases upstream in the apoptotic activation pathway. ^{120– 122} In turn, caspase-3 appears to be a downstream central executioner ^{123– 125} that can directly process pro-caspases-2, -6, -7, and -9. ^{126, 127} Findings by several groups have revealed that the activation of caspase-3 requires the Ced-4 homolog, Apaf-1, and pro-caspase-9, as well as dATP and cytochrome C. ^{128– 130} Hence, caspase-9 is upstream in the pathway and is regulated by Bcl-2 family genes. For murine caspase-11 and -12, no human counterparts have been described so far. Recently, the isolation of human caspase-13 (ERICE) from the ICE subfamily was reported. ¹³¹ The demonstration that activation of ERICE is mediated by caspase-8 has supported a potential downstream role for active ERICE in caspase-8–mediated cell death.

There is also evidence that cell death can proceed in the absence of caspases, perhaps through alterations in the mitochondrial membrane permeability transition (PT). ^{132, 133}

Substrates of caspases

Activated caspases cleave numerous targets resulting in the so-called “death of a thousand cuts.” Caspase targets include cytoskeleton proteins (nuclear lamins, actin, gelsolin), regulators of DNA repair (poly ADP-ribose polymerase [PARP]), degradation of nuclear DNA by activation of DNA-dependent protein kinase, and deactivation of the inhibitor of caspase-activated deoxyribonuclease protein (ICAD) ¹³⁴ ; RNA splicing (U1 protein), nuclear mitotic apparatus protein (NuMA), and cell cycle proteins (including Rb and p21-activated kinase [PAK]). ¹³⁵ Caspases are also involved in extracellular apoptotic events including the cleavage of apoptotic bodies and exposure of phosphatidylserine on the cell surface. Caspase-dependent cleavage of p21 ¹³⁶ and Rb ¹³⁷ during DNA-damage-induced apoptosis provides one of the potential links between apoptosis and cell cycle progression. Recent studies demonstrated that Bcl-2, Bcl-X^L, and XIAP proteins are also substrates for caspase cleavage. ^{138– 140} Cleavage of these proteins releases a C-terminal product that lacks the BH4 domain and acts as a death effector. Hence, once caspase activation has been initiated, proteins that inhibit apoptosis are functionally inactivated and converted into peoapoptotic efectors. Since drug resistance is associated with an inability of tumor cells to undergo apoptosis, direct activation of caspases in cancer cells may be an effective strategy to kill resistant cells. One of the proposed approaches is to induce intracellular cleavage of caspase-1 or caspase-3 by a nontoxic, lipid-permeable, dimeric FK506 analog that binds to the attached FK506-binding proteins, FKBPs. ¹⁴¹ Using this chemically induced dimerization, it was possible to induce rapid apoptosis in a Bcl-X^L-independent manner. Obviously, caspase activators should be selective for cancer cells. Whether these approaches can be realized is presently uncertain.

Natural Inhibitors of Caspases

The function of caspases, even after activation by cleavage, is subject to inhibition by other physiologic caspase inhibitors, thereby preventing unwanted or accidental proteolysis. Alterations in the expression or function of these proteins may confer resistance of tumor cells to the apoptotic stimuli. Viral proteins including CrmA (inhibitor of caspase-1 and -8) and p35 (which inhibits almost all caspases) were the first described caspase inhibitors. ¹⁴² The decoy protein, FLICE-inhibitory protein (FLIP), which prevents the binding of FLICE to its cofactor FADD, inhibits caspase-8 FADD-like interleukin-1B-converting enzyme (FLICE), which is required for its activation. The recent discovery of the CARD domain-containing protein ARC (apoptosis repressor with caspase recruitment domain) suggests the existence of decoy proteins for other caspases. A new family of proteins known as IAPs (for inhibitors of apoptosis proteins) was identified via homology with the baculovirus IAP genes and includes IAP1, IAP2, NAIP, XIAP, and survivin. ^{142– 147} Survivin seems to preferentially target caspase-9. XIAP, c-IAP1, and c-IAP2 prevent the proteolytic processing of pro-caspases-3, -6, and -7 by preventing the conformational changes of pro-caspase-9 required for downstream activation. ^{148, 149} Additionally, active caspase-3 function is directly inhibited by the binding of cleaved caspase-3 or -7 by XIAP, c-IAP1, c-IAP2, and survivin. These findings suggest that the ratio of caspases to IAPs is likely to be critical. However, since IAPs function by blocking caspase activation, they may not be able to prevent cell death induced by caspase-independent mechanisms.

Loss of IAP-related genes may cause cell death in mammalian cells and certain disorders, that is, NAIP mutations are observed in two-thirds of patients with spinal muscular atrophy ¹⁴⁴ c-IAP2 and a novel gene MLT are rearranged in the t(11,18) ¹⁵⁰ found in MALT lymphomas, potentially conferring a survival advantage to lymphoma cells and in some cases rendering them immune to Fas-induced apoptosis. ¹⁵¹

A new human gene survivin, has been described that encodes a structurally unique IAP apoptosis inhibitor that is undetectable in terminally differentiated adult tissues but prominently expressed in transformed cell lines and in all of the most common human cancers of lung, colon, pancreas, prostate, and breast and in high-grade non-Hodgkin’s lymphomas. ¹⁵² Survivin is the first apoptosis inhibitor that is selectively expressed in the G2-M cell cycle-phase and directly associates with mitotic spindle microtubules. ¹⁵³ Inhibition of the survivin-tubulin interaction by microtubule-disrupting agents such as vincristine or nocodazole or mutagenesis of the caspase-binding BIR domain ¹⁵⁴ results in increased caspase-3 activity and induction of apoptosis in G2-M-synchronized cells. Therefore, survivin appears to be a novel apoptotic guardian of a cell cycle checkpoint. High levels of survivin expression are associated with poor clinical outcome in neuroblastoma and colon and gastric cancers. ^{155– 157} Intriguingly, the coding strand of survivin is extensively complementary to that of effector cell protease receptor-1 (EPR-1), ¹⁵⁸ although they are coded from separate genes located at 17q25. ¹⁵⁹ The finding that downregulation of survivin by overexpression of ERP-1 in vitro increases apoptosis and inhibits growth of transformed cells has supported a potential role for endogenous ERP-1 as a natural antisense ¹⁶⁰ and survivin as a potential new target for apoptosis-based therapy.

Death Receptors

Cells require both internal and external means of regulating the activation of caspases and the death machinery. Cell surface death receptors can, depending on other contextual events, transmit apoptosis signals in response to external stimuli such as death ligands, growth factor withdrawal, or chemotherapeutic agents. Death receptors belong to the tumor necrosis factor (TNF) receptor family and have a characteristic cysteine-rich extracellular domain ¹⁶¹ and a homologous cytoplasmic “death domain” ¹⁶² that initiates apoptotic signaling inside the cell. These receptors can induce apoptotic cell death within hours after ligand binding and may exert their apoptogenic effects differentially in diverse cell types, ¹⁶³ depending on downstream signaling.

Structure and Function of BCL-2-Related Proteins

The Bcl-2 family of proteins consists of both inhibitors and promoters of PCD. There are four important Bcl-2 structural homology motifs: BH1, BH2, and BH3, present in both the anti- and pro-survival subfamilies, and BH4, present only among antiapoptotic proteins. ^{201– 204} Most antiapoptotic proteins contain BH1 and BH2, and those closely resembling Bcl-2 contain all four domains. The proapoptotic proteins form two subfamilies. The Bax group includes Bax, Bak, and Bok (Mtd), resembles Bcl-2, and contains BH1, 2, and 3 domains. The BH3 domain group encompasses seven family members (Bik, Blk, Hrk, BNIP3, Bim^L, Bad, Bid, EGL-1) that possess only the BH3 domain, which is essential for their function.

Many Bcl-2-family proteins can associate with each other through a complex network of homo- and heterodimers ²⁰⁵ that depend on interactions between the BH1, BH2 or BH3 domains. ^{201, 206} Bcl-2 forms homodimers or heterodimers with Bax, Bcl-X^L, Bcl-X^S, Mcl-1, and BAD. ^{205– 208} The ratio of antiapoptotic versus proapoptotic dimers is important in determining resistance of a cell to apoptosis, but, in most cases, the functional significance of these interactions has not been explored. However, mutational analysis studies and concerns about the methods used to determine these interactions have raised questions about the requirement for direct protein-protein interactions between the anti- and proapoptotic BCL2 family members. ^{209– 211} Also, recent analyses of cells expressing various levels of Bcl-2 and Bax have revealed that the degree of protection against apoptosis correlates with the amount of Bcl-2 that is free of Bax, rather than the number of Bcl-2-Bax heterodimers. ²¹² Deletion mutants of Bcl-2 lacking the BH4 domain, which permits interaction with proteins, including Bag-1, Raf-1, calcineurin, p53-binding protein, and Nip1-3, exhibit either loss of function or dominant-inhibitory activity and thereby paradoxically promote apoptosis. ^{213– 215}

Apoptosis Activating Factors (Apafs)

Of the Bcl-2/Bcl-X^L-binding proteins in C elegans, CED-4, which serves to recruit caspases via its N-terminus CARD ²⁵¹ domain and activate them, ²⁵² is probably the most important. ^{252– 258} Zou and colleagues have isolated apoptosis activating factors (Apafs) ²⁵⁹ that result in cleavage and activation of caspase-9. Apaf-1 resembles CED-4 with an amino-terminal CARD domain that can bind directly to caspases ²⁵¹ and a large domain containing 12 WD-40 repeats at the C-terminus thought to interact with Apaf-2/cytochrome c. ATP may activate Apaf-1 by binding to the nucleotide-binding p-loop motif. An essential role for Apaf-1 in the Bcl-2-regulated pathway is supported by observations of reduced apoptosis in the brain and striking craniofacial abnormalities with impaired processing of caspases-2, -3 and -8 in Apaf-1 knockout mice. ^{260, 261} Remarkably, cells from Apaf-1 –/– mice are refractory to apoptotic stimuli controlled by Bcl-2 but respond normally to signals from “death receptors.”

In conclusion, CED-4 or human Apafs are likely to be critical for Bcl-2 function in regulation of the caspases.

Taken together, an intriguing network of genes controlling proliferation, differentiation, and apoptosis has evolved that provides new targets for cancer therapy that were unimaginable only a few years ago.