Targeted Therapies in Oncology

Combination Therapy Is Required for Cure

The goals of combination therapy are numerous. First, because it is difficult to predict which single agent will be most effective against an individual patient's tumor, the use of multiple drugs increases the potential for an initial complete response. Second, combination therapy is likely critical for preventing the development of therapeutic resistance. Third, drug combinations may have additive or synergistic interactions, increasing the total activity of the drugs and potentially enabling lower dosing and reduced toxicity. One example is the successful dose reduction of cytotoxic chemotherapy with the addition of all- trans -retinoic acid (ATRA) for the treatment of patients with acute promyelocytic leukemia (APL). With few exceptions, curative systemic therapy has required the combination of highly active agents rather than combinations of active and inactive agents.

The need for multiagent therapy was first appreciated in the treatment of acute lymphoblastic leukemia (ALL). At best, 60% of patients are estimated to achieve complete remission with single-agent chemotherapy. Despite continued use, almost all patients will relapse within 6 to 9 months with single-agent therapy. However, complete remission rates exceed 95% in patients treated with combination chemotherapy, and cure rates today for children with ALL exceed 80% with multiagent treatment regimens. Similarly, patients with APL had very poor remission and cure rates until the discovery of ATRA differentiation therapy for this disease. ATRA achieved a high complete remission rate in use as a single agent, but virtually all patients ultimately relapsed. It was not until the combination of cytotoxic chemotherapy with ATRA and the recent combination of ATRA with arsenic trioxide that these patients obtained the best long-term survival rates of any acute myeloblastic leukemia (AML) subtype.

Appropriate Dose Intensity Is Required for Improved Outcome

A guiding principle in the delivery of cytotoxic chemotherapeutic agents is the concept of the maximum tolerated dose (MTD): the simple idea that more is better. Drugs are delivered at the maximum dose possible, a dose greater than which unacceptable toxicity is encountered. Here, without a clear cellular biomarker of cytotoxic activity, it is not possible to relate the maximum molecular effect to maximum achievable efficacy. Thus cytotoxic drugs are first evaluated in phase I studies with the primary study goal of establishing the MTD. Subsequent phase II and III trials are then based on this dosing schedule. MTD assumes that some toxicity is acceptable and, in fact, expected in the pursuit of cure. The most extreme example of this concept has been the use of autologous stem cell transplantation to deliver what otherwise would be lethal doses of chemotherapy. In this case the MTD is surpassed by replacing the dose-limiting factor, the hematopoietic system, with stem cells harvested from the patient before delivery of the conditioning chemotherapy. While clearly a brute force approach, the MTD paradigm also established an important corollary: significant side effects can be tolerated when therapy is administered with curative intent (discussed later).

Many retrospective studies of chemotherapy dose intensity and outcome support this premise, particularly in the pediatric setting. In high-stage neuroblastoma the addition of dose intensification with stem cell transplantation has improved the outcome for patients with metastatic disease. In pediatric ALL several studies suggest that dose intensification is important for long-term outcome, including one randomized study com­paring standard versus dose-reduced methotrexate with mercaptopurine. Moreover, in Ewing sarcoma chemotherapy intensification through interval compression in a randomized trial was also associated with an improved outcome in patients with localized disease. However, fewer prospective studies than retrospective studies have been performed to address the issue of dose intensity and outcome for pediatric tumors. Significantly, while these studies begin to emerge, some previous assumptions about MTD have come into question. For example, retrospective studies support the notion that doxorubicin or methotrexate dose intensity is important for outcome in patients with localized high-grade osteosarcoma. However, prospective studies involving large patient numbers have called into question the role of further dose intensification for this disease. Furthermore, randomized trials for malignancies such as breast cancer have demonstrated that massively increasing dose with stem cell transplantion does not always improve long-term survival. Well-controlled prospective trials are critical to address the issue of dose intensification, even for cytotoxic agents.

Toxicity Is Acceptable When Therapy Is Curative

The oncology community has accepted toxicity as inherent in the mission of cure. With the implementation of dose-intensive regimens, however, comes additional toxicity. Because of the relatively nonspecific nature of traditional chemotherapy, morbidity from these treatment regimens is significant. Acute toxicity takes the form of infection, transfusion dependence, renal injury, and hepatic injury. There are also potential long-term consequences such as cardiac injury, infertility, chronic renal insufficiency, hearing loss, skeletal injury, and secondary malignancies. Ongoing efforts are focused on decreasing these short- and long-term side effects. For example, to minimize acute toxicity, chemotherapy dosing is usually individualized on the basis of body surface area (mg/m ² ). For some cytotoxic agents, such as methotrexate and busulfan, drug levels can be monitored to minimize toxicity. Dose intensity can also be achieved with regional tumor delivery, growth factor support, or rescue with stem cells. Furthermore, better supportive care, such as infection prophylaxis, has enabled tolerance of prolonged immunosuppression. To decrease long-term morbidity, “safe” maximum cumulative doses are being established, such as maximum cumulative doxorubicin to minimize late cardiac effects. Although toxicity can be tolerated, it is essential that the level of toxicity be commensurate with the level of benefit. Marginally effective therapeutic regimens that result in substantial toxicity should not become accepted standards.

Establish Curative Regimens and Attempt To Reduce Toxicity

Treatment strategies for most of the pediatric malignancies, including ALL and the solid tumors, have used disease stratification to decrease dose in treatment regimens. A dose-intensive regimen is initially used to establish efficacy. Then, patients with a good prognosis are treated with less intensive regimens, with the goal of com­parable disease-free outcomes but decreased treatment-related morbidity. This is particularly important for pediatric disease, in which the long-term toxicity of treatment is experienced over a lifetime. Historically, these stratifications were based on factors such as age, stage, and histology, but now they include molecular genetic markers, such as MYCN amplification in neuroblastoma, and MLL rearrangement in pediatric ALL. For example, the National Wilms'Tumor Study evaluated a half-dose chemotherapy regimen for infants with low-risk disease in an attempt to decrease the incidence of toxic deaths without compromising long-term survival. These infants had acceptable toxicity with no toxic deaths, and long-term survival was not compromised. A similar approach has been taken with infants and young children with low-stage neuroblastoma, and therapy for children with ALL has been dose-reduced based on rapid clearance of minimal residual disease.

While our understanding of the molecular basis of malignancy improves, so will our ability to develop fine-grained prognostic categories and patient-tailored treatment regimens, perhaps even for traditional chemotherapy agents. While we implement targeted therapy, it is expected that disease stratification and patient selection will be based on molecular genetic findings related to the credentialed target. In the clinical development of targeted therapies, we must continue to consider these key concepts: combination therapy, appropriate dose intensification, tolerance of toxicity for substantial efficacy and cure, and treatment adjustment within established favorable prognostic groups.

Genetics Track

The genetics track relies on the idea that a cancer develops from a finite number of genetic alterations, that these alterations give the malignant cell a selective advantage, and that these genetic changes are selected for during malignant cell replication. Hence the malignant cell becomes dependent on a set of these changes for continued survival, a process sometimes referred to as oncogene addiction. This cancer cell dependence or addiction should provide a window for drug specificity for the malignant cell compared with normal cells. Targeted therapeutics taking advantage of genetic abnormalities should, in theory, be more injurious to the malignant cell than to the normal cellular counterpart. Advances in cytogenetic studies, high-throughput gene sequencing, detection of copy number alteration, whole genome expression profiling, and proteomic approaches have revolutionized candidate genetic target discovery. Now, tens of thousands of genes in hundreds of tumor samples can be analyzed simultaneously. These candidate targets must then be carefully credentialed for their suitability as a drug target. Is there cancer cell dependence on the genetic abnormality, and can the target be pharmacologically manipulated? Imatinib is one of a number of examples of clinically validated targeted therapeutics exploiting a genetic dependence, in this case the dependence of CML on the activity of the Abelson kinase (ABL). The very existence of the cancer-dependent state is dramatically illustrated by the emergence of resistance mechanisms that largely appear to restore Abelson activity (see further discussion later in this chapter).

Lineage Track

Cancer cells often maintain developmental features of the lineage from which they were derived. This is well supported with gene expression studies that report tumor cells more closely resembling the normal lineage counterpart than tumors of a different cell type. The notion that lineage or “legacy” features of a tumor cell may be exploited with targeted therapy is known as lineage addiction . One of the earliest examples of the success of this approach has been the antagonism of ER and its signaling pathway in patients with ER-positive breast cancer. Similarly, the vast majority of patients with prostate cancer derive significant benefit from antagonism of the androgen receptor (AR). In both ER-positive breast cancer and in prostate cancer (virtually all forms of which express AR), the emergence of hormone-refractory disease largely appears to rely on mechanisms that restore steroid receptor signaling. Thus as in the case of imatinib resistance, the dependent state is illustrated by these resistance mechanisms. In many cases lineage dependence may also derive from the normal hierarchy of differentiation beginning with the stem cell.

Host Track

The importance of tumor cell environment to the development and maintenance of the malignant state has become increasingly recognized and represents another potential inroad to targeted therapy. It is possible that tumors, by evolving in a particular environmental context, may become dependent on certain growth factors or neighboring cells for survival in that niche. Preclinical evidence suggests that many tumors require de novo blood vessel development for their growth and maintenance. The identification of these angiogenesis factors has enabled the development of modifiers of the angiogenesis process, leading to the successful clinical application of bevacizumab, an anti–vascular endothelial growth factor (VEGF) antibody, in patients with colorectal cancer and the approval of kinase insert dependent receptor inhibitors (sunitinib and sorafenib) in renal cell carcinoma (see further discussion later in this chapter). The interaction of tumors with the bone microenvironment has also been targeted using inhibitors of osteoclastogenesis. Here, multiple bisphosphonate agents are approved for the prevention of cancer-related fractures.

Synthetic Lethal or Empiric Track

A number of highly efficacious anticancer agents do not conveniently map to the first three categories. Admittedly, we can say that the understanding of the mechanisms of efficacy of such agents remains opaque. Nonetheless, it is quite likely that such agents, including many of the cytotoxic agents described previously, take advantage of aspects of cancer cell biology that remain unknown but may be broadly termed as synthetic lethal interactions. Synthetic lethal genetic interactions are defined as mutations that are lethal in combination but that do not produce a lethal phenotype as a single event. In a similar way, gain-of-fitness alterations in a malignant cell, which give the cell a survival advantage, can sensitize the cell to other stresses that have no consequence in a normal state but may be lethal with the malignancy promoting alteration. Thus cancer-promoting mutations give the malignant cell a survival advantage, but they may also incur a distinct liability that can be exploited therapeutically. Rather than two synthetic lethal genetic events, the clinician can also apply such synthetic lethal concepts to a preceding cancer-related molecular alteration followed by the application of a therapeutic. For example, a striking synthetic lethal relationship was discovered between loss of BRCA1 or BRCA2 function and inhibition of the poly(ADP-ribose) polymerase I enzyme (PARP). Here, loss of PARP1 was found to induce double-strand DNA breaks that required the recruitment of the RAD51 homologous recombination repair complex to the site of these breaks for effective repair. Because it was known that BRCA1 and 2 were required for RAD51 recruitment to such complexes, it was hypothesized that loss of BRCA1 or BRCA2 would create sensitivity to PARP inhibition. This was demonstrated both with siRNA and small-molecule inhibitors of PARP. Moreover, clinical activity of PARP inhibitors has now been seen in patients with BRCA1 and BRCA2 mutations with ovarian or breast cancer.

Target Epidemiology

Preclinical models of cancer are limited in both number and in their qualitative relationship to human cancers. Comprehensive collection of human tumor specimens could, in theory, be used to study a given target in a large array of samples broadly covering tissue distribution and stage distribution (primary versus metastatic sites). In practice, however, there are limited sets of comprehensive tumor collections, and generally the largest tumor-type specific collections are limited to fixed, paraffin-embedded specimens. The availability of tissue samples paired with robust clinical data remains limited. Furthermore, the vast majority of human cancer collections come from surgical resection samples, which do not necessarily reflect the disease state in which a therapeutic will be initially tested. Despite these limitations, molecular epidemiology studies in tissue samples, enabled by tissue microarrays, single nucleotide polymorphism (SNP) arrays, expression microarrays, reverse protein arrays, and next-generation sequencing technologies remain critical to building a case for the cancer relevance of a given target.

Genetic Alterations.

As illustrated by the success of ATRA, trastuzumab, and imatinib, direct genetic alteration leading to gene activation of a given drug target remains perhaps the best evidence for the causal role of a given gene in carcinogenesis. Genetic alterations that are highly frequent in a given tumor type and are present in the majority of cells are a strong indicator of causality. In the era of large-scale resequencing projects, it is critical that one distinguishes “driver” mutations (causal mutations) from so-called passenger mutations. In addition, driver mutations that arise late in the course of disease, are present in only a minor fraction of the tumors, or are present in only a fraction of tumor cells in a given single tumor must be given less weight. These findings would suggest a reduced likelihood of this lesion possessing a strong causal link to the initiation of the disease state, although they may pose an eventual mechanism of drug resistance. The types of genetic alterations (e.g., point mutations, focal amplifications, broad amplifications, and translocations) are also distinguished with respect to the extent to which they can be used to establish causality or to infer dependence.

Gain-of-Function Genetic Alterations.

Gain-of-function genetic alterations include translocations, activating point mutations, and increases in gene copy number. Translocations were among the first discovered genetic alterations initially described using chromosome spreads. Approaches to the discovery of chromosomal rearrangements include exon microarrays, which can potentially detect imbalanced or amplified translocations, cancer outlier profile analysis (COPA), paired-end resequencing of BAC clones from a genomic library (giving a structural map of a cancer genome), and newer methods of single-molecule sequencing (discussed later). Translocations are highly informative with respect to causality because they typically alter only two genes, and because background or random translocation events are relatively rare, with chromothripsis (massive DNA rearrangements affecting one or a few chromosomes that occur as a result of one catastrophic event) as a notable exception.

Amplifications are now routinely detected by methods of DNA copy number analyses: comparative genomic hybridization (CGH) array, high-density SNP arrays, and whole exome or whole genome sequencing. There are two general types of increases in gene copy: focal amplifications and broad copy number increases covering an entire chromosome or chromosome arm. With sufficient sample numbers it is possible to discern one to several genes contained within a focal amplicon. Typically, the minimal region of amplification is a guide. However, methods for detecting copy number alterations (e.g., genomic identification of significant targets in cancer [GISTIC]) also use the amplitude of amplification as an important parameter. Although it is not definitive proof, amplification lends evidence to a causal relationship between a given gene within an amplicon and a cancer state. In some amplicons, multiple genes remain strong candidates for oncogenes targeted by the amplification. One notable example is the 11q13 amplification with genes including Cyclin D1 , FGF3 , and FGF4 .

Larger-scale chromosomal alterations leading to trisomy or gain of an entire chromosome arm are frequent yet are far more difficult to study. As such, identifying single genes (or drug targets) within such broad regions remains a challenge. In this case the number of genes implicated in such large-scale alterations is so large that almost no importance can be ascribed to the cancer relevance of a gene contained within such broad chromosomal-level alterations. Outlier analysis may begin to allow the elucidation of multigene targeting by such broad regions and is illustrated by the analysis of trisomy 7 in glioma revealing a dependence on cMET and its ligand HGF, both located on chromosome 7.

Subtle nucleic acid level alterations, including point mutations and small insertions and deletions, are an important mechanism for the activation of oncogenes. Point mutations leading to the constitutive guanosine triphosphate (GTP)–bound and hence activated form of N-RAS and K-RAS, as well as genetic events leading to activation of the BRAF kinase in melanoma, thyroid cancer, colorectal cancer, pediatric pilocytic astrocytoma, and Langerhans cell histiocytosis are exemplars of this mechanism. Large-scale sequencing projects that unveiled new mutations activating oncogenes such as EGFR and PIK3CA have led to national and international cancer genome sequencing projects. From these studies it has become apparent that the detection of passenger mutations, constituting the background rate of mutation, can pose a significant challenge. These passenger mutations must be separated from the more prevalent and recurrent driver mutations. Such passenger mutations may be exceedingly high in certain types of tumors, particularly those bearing mutations in mismatch repair genes. Indeed, in one case inactivating mutations in the mismatch repair gene MSH6 were discovered in glioma tumors recurring after treatment with temozolomide. Here the sequencing of MSH6 was triggered by the very high rate of observed passenger mutations in treated versus untreated tumors. These data suggest that cautious interpretation of low-frequency events is warranted when considering mutation analysis from larger-scale modalities. A framework for context-specific mutation analysis has been provided by a systematic study of mutation data across many cancer types in The Cancer Genome Atlas (TCGA) project and others, and these studies are revealing remarkably simple genomes in many pediatric cancers ( Fig. 44-1 ).

Loss-of-Function Genetic Alterations.

Loss-of-function alterations do not typically directly indicate or point to a specific gene as a candidate drug target. However, such alterations may predict sensitivity to agents acting against protein targets contained within downstream or parallel pathways. For example, clinical trials are under way testing inhibitors of mammalian target of rapamycin (mTOR) function against sporadic tumors or hereditable cancer predisposition syndromes resulting from loss of the tumor suppressors PTEN and TSC1 or TSC2. In both cases loss-of-function mutations in these tumor-suppressor genes are thought to lead to constitutive activation and hence constitutive dependence on mTOR (discussed later in this chapter).

Similarly, in the hedgehog pathway, a key development pathway regulating cell patterning, deletions or mutations in the gene encoding the Patched receptor (PTCH) are associated with hereditary and sporadic basal cell carcinomas and sporadic medulloblastoma. Loss of PTCH, a receptor for the ligands in the hedgehog family, leads to constitutive activation of the atypical G protein–coupled receptor (GPCR)–like protein Smoothened (Smo). Cyclopamine, a naturally derived inhibitor of Smo, and other, newer small-molecule inhibitors, have had significant activity in preclinical models lacking PTCH gene function and in clinical trials. These examples illustrate that identification of loss-of-function mutations can reveal downstream target molecules in critical cancer pathways.

Translocations are typically thought of as gain-of-function genetic events in that they act dominantly. However, they can functionally inactivate endogenous genes through the creation of dominant-negative protein fusions. For example, the PML-RARα fusion protein is believed to act, at least in part, as a dominant-negative inhibitor of endogenous RARα function. This type of molecular action can be difficult to elucidate in functional experiments. Nonetheless, it is worth remembering this possibility when evaluating proteins or genes as drug targets when they are involved in new or existing translocations.

Chromosomal-level alterations leading to loss-of-function events include homozygous deletions, heterozygous deletions, and loss of heterozygosity. Homozygous deletions are typically highly focal in their nature. When the boundaries are defined with large sample sets, in many cases, only a few genes are found in the overlapping region. For example, in the case of focal deletions on 9p, the typical minimal deleted regions in tumors such as melanoma contain only CDKN2A and CDKN2B . Heterozygous deletions are typically larger and usually involve entire chromosomes or chromosome arms. As in the case of amplifications, such large-scale broad genetic alterations make it difficult to ascertain the cancer relevance of any single gene contained within a larger region. Loss of heterozygosity (LOH) can occur as an event associated with heterozygous deletion leaving only one parental chromosome behind, which is thus homozygous. LOH can also occur when duplication of the remaining parental chromosome creates a situation known as uniparental disomy. This is significant because it may indicate preservation of a mutated allele, leading to two mutant alleles of a given gene. In the case of the tumor suppressor gene p53 , copy-neutral LOH is common.

Inactivating point mutations are a common means of gene inactivation. In this case stop codons are the most efficient mechanism for disrupting gene function. As a result, tumor suppressor gene mutations often are enriched for stop or truncating mutations. This creates certain technical and cost difficulties in sequencing genes for inactivating mutations because it is necessary to sequence all coding exons, consider sequencing of the regulatory elements governing splicing, and consider the detection of promoter mutations. As a result of these difficulties, precise clinical analysis of tumor suppressor gene mutations remains challenging even with the application of next-generation sequencing technologies.