Molecularly Targeted Therapies and Biotherapeutics

Malcolm A. Smith

INTRODUCTION

With the approval of the BCR-ABL inhibitor imatinib for the treatment of chronic myeloid leukemia (CML) by the Food and Drug Administration (FDA) on May 10, 2001, the era of molecularly targeted therapy for cancer had undeniably arrived. The remarkable remissions induced by this well-tolerated, orally administered agent in most patients with chronic phase CML raised hopes for a dramatic transformation in the way that cancer is treated. The principle that an agent targeted at the underlying molecular defect(s) of a particular cancer can have potent anticancer activity while producing minimal toxicity has powerful appeal. The activity of all-trans retinoic acid previously demonstrated against acute promyelocytic leukemia (APL) supports the validity of this principle.

Subsequent to the dramatic successes of imatinib for chronic phase CML, a number of other molecularly targeted agents have shown activity against specific cancers or molecularly defined subsets of patients within a given diagnosis, including anaplastic lymphoma kinase (ALK) inhibitors for ALK-rearranged lung cancer and anaplastic large cell lymphoma, BRAF inhibitors for BRAF-mutated melanoma, mTOR inhibitors for subependymal giant cell astrocytoma, and VEGFR2 targeted agents for renal cell carcinoma, (Table 11.1). These positive examples, however, are counterbalanced by the cautionary observations that many agents advertised as molecularly targeted agents have shown limited activity at tolerated doses and that there remain many cancers (including most pediatric cancers) for which no effective molecularly targeted agents have been identified.

A basic question is how to define “molecularly targeted agent.” Identifying the cellular moiety that an agent targets qualifies an agent as “molecularly targeted” only in a tautological sense. By this criterion, most conventional cytotoxic agents can be considered molecularly targeted agents, as illustrated by the high potency with which Vinca alkaloids block tubulin assembly and with which epipodophyllotoxins inhibit topoisomerase II. Rather, the primary distinguishing characteristic of a successful molecularly targeted agent is the dramatic differential dependence of the cancer cell compared with normal cells upon the agent’s target for growth and survival. This differential dependence can generally be related to a specific genomic alteration in the cancer cell, with the alteration commonly being in the gene for the agent’s target or a gene in a pathway closely related to the agent’s target. As the degree of differential dependence between cancer cells and normal cells on a specific target or pathway may vary widely, there will be gradations in the extent to which targeted therapies are able to effectively eliminate cancer cells while causing little harm to normal cells.

The prioritization of molecularly targeted agents will differ between adult and pediatric cancers in part because of the distinctive biology of cancers arising in children compared with those of adults as illustrated by the distinctive patterns of mutations that each possesses. Another important difference is that the primary therapeutic goal for childhood cancers is cure, not palliation, whereas for many adult cancers sustained palliation is an important objective. For adults with cancer, cytostatic agents that slow tumor growth for a year or two may be considered quite valuable if they can prolong survival while allowing acceptable quality of life. For children, achieving stable disease or delaying disease progression for a year or two is at best a modest success. This fundamental distinction between the relative benefit of cure versus palliation for adults and children with cancer has implications both in terms of the cellular pathways targeted for intervention and in terms of the design of clinical trials to evaluate molecularly targeted agents in children. Given the goal of increasing cure rates through the use of molecularly targeted therapies, the identification of critical pathways that promote survival and block apoptosis for specific childhood cancers is a key step in selecting candidate agents for evaluation.

A sea change in cancer drug development occurred in the past decade as gene sequencing transitioned from a research method that was limited to a few genes at a time to a clinical test that can be applied to hundreds or thousands of genes at a time as well as to whole genomes. As a result of this technological advance, thousands of childhood cancer specimens have been sequenced, leading to the identification of the vast majority of recurring mutations present in the cancers that arise in children. The enormous body of data developed has provided remarkable insights into the biology of childhood cancers. At the same time, the new knowledge has provided the sobering realization that most childhood cancers do not have recurring mutations in genes that are at the present time considered targetable.

This chapter examines the potential applications of molecularly targeted agents in the treatment of childhood cancers. The chapter addresses general questions relevant to the development of these agents in children with cancer, noting important lessons learned from the development of molecularly targeted agents for adult indications. A key issue that pervades the chapter is the need for effective prioritization methods to determine which new agents should be introduced into the pediatric setting. The large numbers of molecularly targeted agents in development for adult cancers highlight the critical need to make wise decisions about which of these agents should be brought into phase 1 evaluation and then moved forward for phase 2 and 3 evaluations against specific childhood cancers.

BASIC PRINCIPLES FOR THE CLINICAL DEVELOPMENT OF MOLECULARLY TARGETED AGENTS

Identifying Childhood Cancer Therapeutic Targets

Oncogene Addiction through Genomic Alterations

The clinical success from pursuing a therapeutic target is dependent on the extent to which cancer cells are differentially dependent upon the activity of the target and its associated signaling pathways for survival and/or proliferation. The term “oncogene addiction” describes the dependence of cancer cells on specific activated or overexpressed oncogenes for the maintenance of their malignant phenotype, while the term “tumor suppressor hypersensitivity” describes the dependence of the malignant phenotype on the continued silencing of specific tumor suppressor genes.¹ Oncogene addiction is central to molecularly targeted therapy, as the greatest successes for molecularly targeted agents have come from

successfully blocking the activity of oncogenes to which cancer cells are addicted.

TABLE 11.1 FDA-Approved Small Molecule Molecularly Targeted Agents (Excluding EGFR and HER2 Targeted Agents)

Primary Target	Agent	Primary Cancers with Indications for Targeted Agents
ABL family kinases	Imatinib	Initial approval by FDA for Ph⁺ CML in 2001, with multiple indications subsequently, including: Ph⁺ CML in chronic phase, accelerated phase, or blast crisis Relapsed or refractory Ph⁺ ALL Myelodysplastic/myeloproliferative diseases (MDS/MPD) associated with PDGFR (platelet-derived growth factor receptor) gene rearrangements Aggressive systemic mastocytosis (ASM) without the D816V c-KIT mutation or with c-KIT mutational status unknown Hypereosinophilic syndrome (HES) and/or chronic eosinophilic leukemia (CEL) with the FIP1L1-PDGFRα fusion kinase and HES and/or CEL that is FIP1L1-PDGFRα fusion kinase negative or unknown Dermatofibrosarcoma protuberans (DFSP) that is unresectable, recurrent, and/or metastatic Gastrointestinal stromal tumors (GIST) that are KIT (CD117) positive and unresectable and/or metastatic malignant
	Dasatinib	2006	Chronic-phase, accelerated-phase, or myelogenous or lymphoid blast-phase CML with resistance or intolerance to prior therapy, including imatinib; Ph⁺ ALL with resistance or intolerance to prior therapy
	Dasatinib	2010	Newly diagnosed Ph⁺ CML in chronic phase
	Nilotinib	2007	Chronic-phase and accelerated-phase Ph⁺ CML in patients resistant or intolerant to prior therapy that included imatinib.
		2010	Newly diagnosed Ph⁺ CML in chronic phase
	Bosutinib	2012	Chronic-, accelerated-, or blast-phase Ph⁺ CML in patients with resistance or intolerance to prior therapy
	Ponatinib	2012	Chronic phase, accelerated phase, or blast phase CML or Ph⁺ ALL that is resistant or intolerant to prior tyrosine kinase inhibitor therapy (NOTE: In 2014, the indication was restricted to T315I-positive CML or T315I-positive Ph⁺ ALL, or to patients with CML or Ph⁺ ALL for whom no other tyrosine kinase inhibitor is indicated.)
ALK	Crizotinib	2011	Metastatic non-small cell lung cancer in patients whose tumors are ALK-positive. Regular approval in 2013
ALK	Ceritinib	2014	ALK-positive metastatic non-small cell lung cancer in patients who experience disease progression on or who are intolerant to crizotinib
BRAF	Vemurafenib	2011	Unresectable or metastatic melanoma with the BRAFV600E mutation
BRAF	Dabrafenib	2013	Unresectable or metastatic melanoma with BRAFV600E mutation
BRAF + MEK	Dabrafenib Trametinib	2014	In combination with dabrafenib to treat patients with unresectable or metastatic melanoma with a BRAF V600E or V600K mutation
BTK	Ibrutinib	2013	Mantle cell lymphoma in patients who have received at least one prior therapy
BTK	Ibrutinib	2014	Chronic lymphocytic leukemia (CLL) in patients who have received at least one prior therapy.
Cereblon	Lenalidomide	2005	Transfusion-dependent anemia due to low or intermediate-1 risk myelodysplastic syndromes (MDS) associated with a deletion 5q cytogenetic abnormality
		2006	Multiple myeloma for use in combination with dexamethasone in patients who have received one prior therapy
		2013	Mantle cell lymphoma in patients whose disease has relapsed or progressed after two prior therapies
	Pomalidomide	2013	Multiple myeloma in patients who have received at least two prior therapies, including lenalidomide and bortezomib
HDACs	Vorinostat	2006	Cutaneous manifestations of cutaneous T-cell lymphoma in patients with progressive, persistent, or recurrent disease on or following two systemic therapies.
	Romidepsin	2009	Cutaneous T-cell lymphoma in patients who have received at least one prior systemic therapy
	Belinostat	2014	Relapsed or refractory peripheral T-cell lymphoma
JAK2	Ruxolitinib	2011	Intermediate and high-risk myelofibrosis, including primary myelofibrosis, postpolycythemia vera myelofibrosis, and postessential thrombocythemia myelofibrosis.
MEK	Trametinib	2013	Unresectable or metastatic melanoma with BRAFV600E or V600K mutation
mTOR	Temsirolimus	2007	Previously untreated patients with advanced renal cell carcinoma
	Everolimus	2009	Advanced renal cell carcinoma after failure of treatment with sunitinib or sorafenib
		2010	Subependymal giant cell astrocytoma (SEGA) associated with tuberous sclerosis (TS) in patients that are not candidates for surgical resection
		2011	Progressive neuroendocrine tumors of pancreatic origin (PNET) in patients with unresectable, locally advanced or metastatic disease
		2012	Postmenopausal women with advanced hormone receptor-positive, HER2-negative breast cancer in combination with exemestane, after failure of treatment with letrozole or anastrozole
PI3K-delta	Idelalisib	2014	Relapsed CLL, in combination with rituximab, in patients for whom rituximab alone would be considered appropriate therapy due to other comorbidities
Proteasome	Bortezomib	2003	Multiple myeloma, with initial approval for a refractory patient population, and with subsequent approvals for patients who had received at least one prior therapy (2005) and for patients with no prior therapy (2008)
	Bortezomib	2006	Mantle cell lymphoma with at least 1 prior therapy
	Carfilzomib	2012	Multiple myeloma with progression while on or after treatment with bortezomib and an immunomodulatory agent
VEGFR2	Sunitinib	2011	Progressive well-differentiated pancreatic neuroendocrine tumors (pNET) in patients with unresectable, locally advanced, or metastatic disease
		2006	Advanced (metastatic) renal cell carcinoma (kidney cancer)
		2006	Gastrointestinal stromal tumor (GIST) after disease progression on, or intolerance to, imatinib mesylate
	Sorafenib	2013	Locally recurrent or metastatic, progressive differentiated thyroid carcinoma (DTC) refractory to radioactive iodine treatment.
		2007	Unresectable hepatocellular carcinoma (HCC)
		2005	Advanced renal cell carcinoma (RCC)
	Pazopanib	2012	Advanced soft tissue sarcoma (STS) who have received prior chemotherapy
	Pazopanib	2009	Advanced renal cell carcinoma
	Axitinib	2012	Advanced renal cell carcinoma after the failure of one prior systemic therapy
	Regorafenib	2012	Metastatic colorectal cancer (mCRC) in patients previously treated with chemotherapy, with an anti-VEGF therapy, and, if KRAS wild type, with an anti-EGFR therapy
	Regorafenib	2013	Advanced gastrointestinal stromal tumors (GIST) that cannot be surgically removed and no longer respond to imatinib and sunitinib
VEGFR2/RET	Vandetanib	2011	Symptomatic or progressive medullary thyroid cancer in patients with unresectable, locally advanced, or metastatic disease.

One of the most direct predictors of whether a cancer cell is “addicted” to a particular protein or pathway for survival and/or proliferation is whether the cancer has activating mutations, increased copy number, or translocations in genes involving the pathway. The pattern of clinical efficacy of imatinib (an inhibitor of BCR-ABL, KIT, and platelet-derived growth factor receptor [PDGFR]) supports the utility of targeting genes with activating genomic alterations as illustrated by its activity against BCR-ABL expressing CML,² gastrointestinal stromal tumors (GIST) with KIT mutations,³ and hematologic conditions with translocations that create fusion proteins that result in constitutive activation of either PDGFRA or PDGFRB.⁴^,⁵ The clinical activity of the Hedgehog pathway inhibitor vismodegib in patients with advanced basal cell carcinoma,⁶ a cancer characterized by activating mutations in the Hedgehog pathway, is another example in which the presence of activating mutations predicts for oncogene addiction and clinical activity, as is the activity of the ALK inhibitor crizotinib in patients with ALK-rearranged non-small cell lung cancer (NSCLC) and anaplastic large cell lymphoma.⁷ Oncogenes may also be activated through gene amplification, as exemplified by HER2 amplification in breast cancer (identifying susceptibility to trastuzumab and to small molecule HER2 inhibitors such as lapatinib), MET amplification in gastric cancer and lung cancer (identifying susceptibility to small molecule MET inhibitors), and MYCN amplification for neuroblastoma.⁸^,⁹ Copy number alterations can be identified using array-based approaches, as a by-product of DNA sequencing, or by using targeted approaches such as multiplex ligation-dependent probe amplification (MLPA) and fluorescent in situ hybridization (FISH).

There are caveats regarding the relationship between genomic alterations that activate oncogenes and the clinical activity for inhibitors of these activated oncogenes. For example, the relatively modest activity for FLT3 inhibitors against acute myeloid leukemia (AML) with FLT3 activating mutations suggests that not all mutations are equal in their ability to predict the clinical benefit that will derive from modulating a specific molecular target,¹⁰ as does the lack of impact of JAK inhibitors on clonal burden for patients with JAK2-mutated myeloproliferative neoplasms.¹¹ Likewise, for the V600E mutation of BRAF, the high activity of the BRAF inhibitor vemurafenib against V600E-mutated melanoma contrasts with its limited activity against colon cancer with the mutation.¹² The latter example serves as a cautionary note to those who propose that histologic diagnosis be superseded by genomic diagnosis as the basis for treatment decisions, as it indicates that activated oncogenes act within a cellular context that can modulate the cancer cell’s dependence upon oncogenic signaling.

Specific examples of activated oncogenes that serve as therapeutic targets include the NPM-ALK fusion gene for anaplastic large cell lymphoma, ALK point mutations for a subset of neuroblastoma, BRAF genomic alterations for pediatric gliomas, ABL family fusion genes for pediatric ALL, and Hedgehog pathway mutations for medulloblastoma. However, the frequency of genomic alterations in childhood cancers is much lower than that observed in adult cancers, likely a reflection of the many years over which adult cancers develop and accumulate genomic alterations.¹³ An important lesson learned from sequencing thousands of childhood cancer tumor specimens is that oncogenes for which there are clinically available inhibitors appear to be activated through genomic alterations in only a small subset of all childhood cancer cases. This has important consequences for developing molecularly targeted agents for childhood cancers, as the most straightforward path to targeted therapy development through inhibition of oncogenes activated by genomic alterations is currently an option for a minority of patients. Two other approaches gain prominence as a result: identifying inhibitors of pediatric cancer oncogenes that are currently considered “undruggable” (e.g., EWS-FLI1 and
PAX-FKHR fusion proteins) and identifying synthetic lethal relationships between “undruggable” pediatric oncogenes (or tumor suppressors) and small molecule inhibitors.

Gene Expression and Protein Expression

Gene expression profiling, performed using either array-based or RNA sequencing methods, is a powerful research tool that can determine the proportion of tumors expressing the mRNA for potential cellular targets and can identify previously unsuspected targets expressed by a particular cancer type. For example, the widespread expression of FLT3 in childhood acute lymphoblastic leukemia (ALL) cases with MLL rearrangement was identified through gene expression profiling, leading to investigation of FLT3 as a potential therapeutic target for this ALL subtype.¹⁴ Expression profiling can also identify genes that classify subsets of patients with distinct outcomes or clinical characteristics, as exemplified by the identification of a gene expression signature that distinguishes between alveolar and embryonal rhabdomyosarcoma.¹⁵

Particularly important for the identification of new molecular targets is the application of unsupervised clustering methods to identify novel biologically distinctive subsets within a single cancer diagnosis. Illustrative examples include gene expression profiling of childhood B-precursor cell acute lymphoblastic leukemia (ALL), which revealed clinically relevant molecular subtypes of this subset of ALL including a subtype with characteristics of Ph⁺ ALL but lacking the BCR-ABL fusion gene,¹⁶^,¹⁷ and profiling of medulloblastoma that identified four subgroups that have distinctive patterns of gene mutations and chromosomal alterations.¹⁸ Detailed investigation of the Ph-like ALL subset identified by gene expression profiling showed that these cases have a range of genomic alterations, but that almost all cases show an alteration in genes involved in growth factor signaling. These alterations include potentially therapeutically relevant fusion genes involving tyrosine kinases (e.g., EPO, PDGFRB, ABL1, and CSF1R), as well as genomic alterations involving CRLF2, JAK family members, and RAS pathway alterations.

Detailed investigation of medulloblastoma cases has identified the molecular underpinnings of the four subtypes, of which the Wnt subtype is characterized by favorable outcome and the Sonic hedgehog (SHH) group is characterized by mutations in the SHH pathway.¹⁹ The results from characterization of the SHH group illustrate the challenges facing targeted therapy development for childhood cancers. The SHH group shows distinctive mutation profiles by age, with infants having primarily either PTCH1 or suppressor of fused (SUFU) mutations, older children having PTCH1 or GLI2 amplification, and adults having primarily PTCH1 mutations.¹⁹ It is only for adults that a clear majority of patients have genomic lesions that are sensitive to the SHH pathway inhibitor vismodegib, an agent that inhibits Smoothened (SMO). For the pediatric age range, up to 50% of cases have lesions downstream of SMO that are inherently nonresponsive to vismodegib.¹⁹

The methods described above characterize the genome and transcriptome of cancer cells, but additional methods are required to evaluate the proteome, “the protein population of a cell, characterized in terms of localization, post translational modifications, interactions, and turnover, at any given time.”²⁰ Of particular importance for targeted therapeutics development is phosphoproteome, as the phosphorylation status of specific proteins in signaling cascades and networks controls key cellular processes, including metabolism, transcription, cell cycle progression, cytoskeletal rearrangement and cell movement, apoptosis, and differentiation.²¹ The human genome codes for over 500 kinases, which allow cells to rapidly modify the activity and stability of proteins through site-specific phosphorylation.²¹ Because of their central role in cellular signal transduction pathways that are aberrantly activated in cancer cells and because of the relative ease with which small molecule inhibitors of their activity can be developed, kinases play a disproportionately large role in cancer drug development.

Monoclonal antibodies that recognize specific phosphorylation sites in proteins are important tools in evaluating the status of signaling pathways. Phosphospecific antibodies can be used to document the activation status of specific proteins using immunoblotting and immunohistochemical methods, as illustrated by the documentation of phospho-ALK in neuroblastoma cell lines containing activating ALK mutations.²² Application of phosphospecific antibodies to tissue arrays is a particularly efficient way to define phosphorylation profiles of large numbers of clinical specimens, as illustrated by the use of tissue arrays to create a phosphorylation profile of kinases involved in the mTOR and PKC pathways for alveolar and embryonal rhabdomyosarcoma.²³ Phosphospecific muliparameter flow cytometry is an important tool for providing single cell activation state readouts for signaling pathways, both at baseline and following specific stimuli. This method was used to demonstrate a STAT5 signaling signature after exposure to suboptimal concentrations of GM-CSF in samples from patients with juvenile myelomonocytic leukemia (JMML).²⁴

Expression levels of kinases, even with evidence of activation such as phosphorylation at specific amino acids, does not appear to predict for response to kinase inhibitors in the absence of genomic alterations in an underlying kinase gene. For example, a phase 2 trial of imatinib failed to show a relationship between activation of wild-type imatinib-sensitive tyrosine kinases and clinical response to imatinib.²⁵ Similarly, no objective responses to the MEK inhibitor selumetinib in the absence of BRAF mutations were observed for more than 40 childhood cancer xenografts, even though many of the xenografts tested showed evidence of MEK expression and activation as measured by phospho-ERK levels.²⁶

Synthetic Lethality

Given the relative paucity of activated oncogenes in childhood cancers that can be targeted with current molecularly targeted therapies, there is a great need to identify agents that are effective against cancers that show loss of function for specific tumor suppressor genes and for cancers that show activation of oncogenes that are not targetable with existing agents (e.g., MYC and RAS family genes). The concept of synthetic lethality provides a framework to guide the development of agents in each of these scenarios.²⁷ Synthetic lethal interactions were initially utilized in studies of yeast in which the presence of two mutations (neither of which was lethal alone) resulted in cell death. The cancer drug development equivalent of synthetic lethal interactions is the targeted agent that is not toxic to normal cells but is toxic to cells with an activated oncogene or to cells with a tumor suppressor gene deficiency.²⁸

The concept of synthetic lethality is illustrated by the sensitivity of cancers deficient in homologous recombination through BRCA1 and BRCA2 mutations to poly (ADP-ribose) polymerases (PARP) inhibitors. PARP enzymatic activity generates long chains of poly (ADP-ribosyl)ated polymers on selected proteins, and it plays a key role in the repair of single strand DNA breaks through base excision repair (BER). PARP inhibition leads to unrepaired single-stranded DNA breaks that result in stalled replication forks. Following treatment with PARP inhibitors, these lesions are repaired in cells competent for homologous recombination (HR), but they are lethal in cells lacking HR capacity.²⁹^,³⁰ BRCA1 and BRCA2 germline mutations are most commonly observed for women with breast or ovarian cancers, and PARP inhibitors induce objective responses in a substantial portion of such patients.³¹^,³² BRCA1/BRCA2 mutations are not commonly observed in childhood cancers.

A pediatric-relevant example of synthetic lethality is the relationship between EZH2 inhibition and SMARCB1 loss of function. SMARCB1 loss of function through deletion or mutation is the sole recurring genomic alteration in rhabdoid tumors.³³ SMARCB1 is a member of a SWI/SNF complex that remodels chromatin by utilizing the energy of ATP to reposition nucleosomes, leading to changes in gene expression.³⁴ Counteracting
the effects of the SWI/SNF complex is the Polycomb Repressor Complex 2(PRC2) that mediates gene silencing through EZH2-catalyzed trimethylation of histone 3 lysine 27 (H3K27) at the promoters of target genes.³⁵ Mice with conditional loss of SMARCB1 in their T-cells rapidly develop T-cell lymphomas, but tumor development is completely suppressed by concomitant loss of EZH2.³⁶ Small molecule inhibitors of EZH2 have been developed and have entered clinical evaluation.³⁷ Treatment of a rhabdoid tumor xenograft with an EZH2 inhibitor led to dose-dependent tumor regression and was associated with the expected reduction in intratumoral trimethylation of H3K27.³⁸

The requirement for DOT1L function for leukemias with MLL gene rearrangement is another pediatric-relevant example of a dependency created by a cancer-causing genomic alteration.³⁹^,⁴⁰ DOT1L is the methyltransferase that modifies histone 3 on lysine 79 (H3K79). H3K79 is a chromatin modification associated with actively transcribed genes, and the DOT1L containing complex that forms with various MLL-fusion genes leads to H3K79 methylation at the promoter regions of many genes that are MLL-fusion gene targets. Genetic inactivation of DOT1L led to down-regulation of MLL-fusion gene targets while leaving global gene expression largely unaffected, and it blocked MLL-fusion-driven leukemia in vivo.⁴⁰ A small molecule inhibitor of DOT1L induced complete regressions in a xenograft model of MLL leukemia, and this agent has entered clinical evaluation.⁴¹

Target Validation and In Vitro and In Vivo Testing

While the presence of gene mutations, increased gene copy number, or translocations involving a gene are a priori evidence for the importance of a specific gene for a cancer, functional validation is required before accepting the gene product as a therapeutic target. This requires observing the effect of specifically inhibiting (or activating) the gene product of interest in relevant cancer cell lines and in vivo models. Various methods have been used to evaluate the therapeutic potential of modulating the activity of particular proteins or signaling pathways, including blocking antibodies (primarily applicable to protein receptors and their ligands), dominant negative mutants that interfere with function when overexpressed, genetic models in which the pathway’s activity is either repressed or overexpressed, nucleic-based methods (e.g., RNA interference [RNAi]) to down-regulate protein expression, and chemical inhibitors. The example of the insulin-like growth factor I (IGF-1) receptor as a potential therapeutic target for Ewing sarcoma and rhabdomyosarcoma illustrates how these methods have been applied in the pediatric setting. Data from experiments using blocking antibodies, overexpression of a dominant negative IGF-1 receptor (IGF-1R),⁴² and cell lines derived from genetically engineered mice null for IGF-1 receptor expression support the potential utility of targeting IGF-1 pathway signaling for these two pediatric cancers.

In vitro testing of candidate targeted agents against panels of molecularly characterized cancer cell lines can provide additional evidence for target specificity and relevance.⁴³ A hallmark of the in vitro sensitivity profile of molecularly targeted agents is that they show highly selective activity against cancer cell lines. As an example, an ALK kinase inhibitor was tested against a panel of 602 cancer cell lines, with most being largely refractory to treatment but with a small subset of lines (approximately 2%) displaying marked sensitivity.⁴³ The sensitive cell lines were highly enriched for neuroblastoma, anaplastic large cell lymphoma, and NSCLC, cancer types now known to contain genomic alterations resulting in ALK activation.

Sorafenib is an example of a multitargeted kinase that shows nanomolar range growth inhibition of cell lines that have activation of specific receptor tyrosine kinases, including FLT3, KIT, RET, and PDGFR.⁴⁴^,⁴⁵ Sorafenib also potently blocks VEGFR2 signaling at similar concentrations. By contrast, at much higher concentrations (1 to 10 µM), sorafenib shows broad activity against a range of adult cancer cell lines and pediatric cancer cell lines.⁴⁵^,⁴⁶ This example illustrates the importance of cautiously interpreting claims of activity based on in vitro results using concentrations far in excess of those required to inhibit cell lines with known dependence upon the agent’s target(s), as these observations likely have little or no clinical relevance due to nonspecific toxic effects that preclude achieving such high levels in the clinical setting. In the cases of sorafenib, which is highly protein bound, micromolar concentrations are achieved in patients, but the unbound drug levels are in the nanomolar range, and it is the latter that are most related to IC⁵⁰ values obtained in vitro in low serum testing conditions.⁴⁵^,⁴⁷

Interpreting the clinical relevance of in vitro activity data must take into consideration the drug concentrations that are achievable in the clinic. When in vitro activity requires 72- to 96-hour exposure to concentrations that can be achieved only for brief periods in the clinical setting (or not achieved at all), then skepticism about the presence of a therapeutic window for the agent is warranted. In vitro testing results for the HDAC inhibitor vorinostat illustrate this concept. Concentrations of vorinostat from 2.5 µM to 10.0 µM have been commonly used to demonstrate the in vitro effects of vorinostat, even though the maximum achievable drug concentrations in humans are in the 1 to 2 µM range.⁴⁸^,⁴⁹^,⁵⁰^,⁵¹ Concentrations above 1 µM are only briefly maintained because of the short half-life of vorinostat (1.5 to 2.0 hours).⁵⁰^,⁵² In vitro exposure to vorinostat for 3 to 6 hours at 1 to 10 µM concentrations fails to elicit significant growth inhibition, and exposure for 24 or more hours is required for significant effects.⁵³ Hence, it is not surprising that despite the broad in vitro cytotoxic activity of vorinostat at micromolar concentrations, its activity in the clinic is very restricted (primarily cutaneous T-cell lymphoma).⁵¹

A final step in target validation is confirmation of antitumor activity in relevant in vivo models.⁵⁴ While genetically engineered models have provided important insights into cancer biology as described previously, in vivo drug testing has primarily, though not exclusively, been performed using xenograft models. An exception to this generalization is the central role that genetic models of medulloblastoma heterozygous for Ptch1 have played in the development of Hedgehog pathway inhibitors.⁵⁵ The reliability of xenograft models has been questioned,⁵⁶ leading to attempts to increase their predictive value, including: molecularly characterizing xenografts to confirm that they are biologically similar to the cancers that they are claimed to represent;⁵⁷ utilizing panels of xenografts for a given diagnosis to better reflect the clinical heterogeneity within the diagnosis;⁵⁷^,⁵⁸ utilizing xenografts established by direct transplantation into immunocompromised mice to reduce the selective pressures forced by the requirement of initial in vitro culture;⁵⁹^,⁶⁰ and ensuring that the systemic drug exposures at which activity is observed in preclinical models are comparable to those that are achievable in humans.⁶¹ Xenograft models with specific genomic alterations (e.g., ALL xenografts with BCR-ABL or JAK2 mutations, neuroblastoma xenografts with ALK mutations, etc.) are especially useful in confirming the in vivo relevance of both therapeutic targets and of agents that block action of the targets.

A common reason for in vivo preclinical testing results to overpredict for clinical activity is that mice can tolerate higher drug exposures for tested agents in comparison with humans, leading to overestimates of activity as a result of human xenografts being exposed to higher drug concentrations than comparable tumors are exposed to in humans. As an example, mice tolerate much higher levels of topoisomerase I inhibitors than humans, leading to high estimates of activity for this class of agents against a range of xenograft models using the murine maximum tolerated dose, but without translation of activity in the clinic for the corresponding cancer diagnoses. It is possible to minimize false positive results based on greater tolerance of mice for anticancer agents by reducing the dose of the agents so that the drug levels in mice more closely approximate those achievable in humans.⁶¹ When this was done for topotecan, the preclinical activity closely mirrored that observed in the clinic.⁶¹ In recognition of the importance of addressing differences in drug exposure between mice and humans,
it is common practice for pharmaceutical companies to perform detailed pharmacokinetic-pharmacodynamic (PK-PD) modeling to characterize the relationship between drug concentration and antitumor activity in dose-ranging xenograft experiments.⁶²^,⁶³ Predictions of clinical activity based on simulations of tumor growth inhibition (TGI) in xenografts at clinically relevant doses and schedules can be made by replacing the murine pharmacokinetics, which are used to build the PK-PD model, with human pharmacokinetics. These predictions based on human pharmacokinetic data are more accurate predictors of clinical activity than predictions based on activity at the murine maximum tolerated dose.⁶²

A key factor in enhancing the predictive value of in vivo preclinical models for molecularly targeted agents is the use of clinically relevant criteria for claiming antitumor activity. It is common for reports of in vivo testing of new agents to claim activity for tumor growth delay, which in the clinical setting would be defined as progressive disease. For childhood cancer in vivo preclinical testing, tumor regressions are felt to be the best indicator of likely clinically relevant activity for childhood cancers. The goal of treatment for childhood cancers is cure, and the pathway to cure must go through complete response, whether achieved by chemotherapy, radiotherapy, or surgery. Examples of agents shown to induce objective responses both in preclinical models and in the clinic include crizotinib for tumors with ALK fusion proteins,⁷^,⁶⁴ dasatinib for BCR-ABL ALL,⁶⁵ sorafenib for FLT3-ITD leukemias, ⁶⁶^,⁶⁷^,⁶⁸ and MEK inhibitors for BRAF-mutated cancers.²⁶^,⁶⁹

Systematic approaches to testing novel agents against pediatric preclinical models, such as that being employed by the Pediatric Preclinical Testing Program (PPTP), are important for the development of molecularly targeted agents. The PPTP has established a set of molecularly characterized cell lines and xenografts that represent common types of childhood cancers.⁵⁷ Standard PPTP testing procedures involve testing novel agents against the 23 cell lines of its in vitro panel and against more than forty solid tumor and ALL xenografts, with most childhood cancers represented by 4 to 8 xenografts. Activity signals have been observed for some targeted agents, including activity of the BCL-2 inhibitor navitoclax against ALL xenografts, the activity of the Aurora A kinase inhibitor MLN-8237 against neuroblastoma xenografts,⁷⁰^,⁷¹ the activity of the MEK inhibitor selumetinib against a BRAF-mutated astrocytoma xenograft,²⁶ and the activity of the MDM2 inhibitor RG7112 against multiple ALL xenografts and selected solid tumors.⁷² Conversely, some targeted agents that are effective against adult cancers have limited activity against the PPTP’s preclinical models.

Discovering High-Priority Combinations Involving Molecularly Targeted Agents

A reasonable assumption supported by mathematical modeling is that optimal benefit of molecularly targeted agents will require their use in combination with standard chemotherapy regimens or with other targeted agents.⁷³ That said, there are very few examples in which an agent that was ineffective as monotherapy succeeded when used in combination. The concept that a novel agent without activity as a single agent should be tested in combination with known effective agents or with other novel agents should be discouraged except in situations with compelling biological rationale, as this strategy will be unsuccessful unless the agent produces a change in cancer cells (but not normal cells) that allows other agents to exert their effect in a more potent way.

There are multiple molecular mechanisms by which drug combinations may result in synergistic, additive, or antagonistic effects.⁷⁴ An enormous challenge in oncology drug development is prioritizing among the millions of potential two- or three-drug combinations that could potentially be studied using the hundreds of oncology agents in clinical development. RNAi screens can provide useful information to guide selection of combinations for further evaluation.⁷⁵ For example, RNAi chemosensitization screens can identify genes that when suppressed at the translational level sensitize cancer cells to chemotherapy agents, as illustrated by the use of such a screen to demonstrate that suppression of the PDK1 pathway sensitizes breast cancer cells to tamoxifen.⁷⁶ Highthroughput genome-scale screening may also be accomplished using the bacterial clustered regularly interspaced short palindromic repeats (CRISPR)-Cas9 system.⁷⁷^,⁷⁸

In vitro testing can be used to identify drug combinations that include molecularly targeted agents for possible prioritization for clinical evaluation. As examples, agents that block BCL-2 family signaling potentiate the activity of a range of chemotherapy agents in vitro,⁷⁹ and mTOR inhibitors potentiate the activity of glucocorticoids against multiple ALL cell lines.⁸⁰^,⁸¹ A critical cautionary note in assessing the clinical relevance of in vitro observations of potentiation or synergy is that these experiments provide little or no information about whether the potentiation is cancer cell specific, and hence combination in vitro testing results provide little or no evidence regarding whether a therapeutic window for tested combinations exists. Clinical development of inhibitors of MGMT (e.g., O6-benzylguanine and lomeguatrib), agents that robustly potentiate the in vitro activity of nitrosoureas against cancer cell lines, illustrates this issue. The concern with agents like MGMT inhibitors that potentiate the toxicity of DNA damaging/interactive agents is that the potentiation may be equally observed for cancer cells and for normal tissues. Unfortunately for both O6-benzylguanine and lomeguatrib, this concern was realized, as the addition of these agents to alkylating agents (e.g., nitrosoureas and temozolomide) in patients led to increased toxicity requiring dose reductions. As a result, the combinations were no more effective than the chemotherapy agent given alone at its standard single agent dose.⁸²^,⁸³^,⁸⁴^,⁸⁵

In vivo combination testing is particularly challenging, as the multiple doses and schedules of each agent that could potentially be evaluated in combination create a daunting challenge. “Therapeutic synergy” is a useful concept to apply for in vivo combination testing, with a positive claim for therapeutic synergy made when the antitumor effect of the combination is greater than the effect of either of the single agents used at their maximum tolerated doses.⁸⁶ Many published combination preclinical studies have failed to compare the combination effect with the effect of the individual agents administered at their maximum tolerated dose (MTD), which may be one reason for previous incorrect predictions of clinical activity for combinations in which a novel agent is added to a standard agent. In the clinical setting, combinations in which a novel agent is added to a standard agent often require substantial reductions in the dose of the standard agent, as described above for MGMT inhibitors. The therapeutic synergy concept directly addresses this concern in the preclinical setting by comparing the activity of the combination with the activity of the single agents at their MTD. An evaluation of the mTOR inhibitor rapamycin with standard agents illustrates the concept of therapeutic synergy.⁸¹ The standard chemotherapy agents studied were cyclophosphamide and vincristine, and for each of these agents, their combination with rapamycin was more effective than any of the single agents used alone at their MTD.⁸¹ These findings stimulated the development of a clinical trial for children with recurrent rhabdomyosarcoma evaluating the addition of either the mTOR inhibitor temsirolimus or the antiangiogenic agent bevacizumab to vinblastine and cyclophosphamide. The results from this trial demonstrated that the temsirolimus-containing regimen produced significantly longer event-free survival than the bevacizumab-containing arm.⁸⁷

There is great interest in novel/novel combinations in which two molecularly targeted agents are used together to broaden and increase the depth of the clinical activity of the combination. An example of an effective novel/novel combination is the BRAF inhibitor dabrafenib plus the MEK inhibitor trametinib for patients with melanoma with BRAF V600 mutations. Both agents show single agent activity for BRAF V600-mutated melanoma through inhibition of MAPK pathway signaling, and preclinical and clinical results documented the greater efficacy of the combination
compared with either single agent.⁸⁸^,⁸⁹ The dual pathway inhibition achieved with concomitant administration of BRAF and MEK inhibitors limits the paradoxical BRAF inhibitor-induced mitogen-activated protein kinase (MAPK) signaling in cells.⁹⁰ This reduction of the paradoxically increased MAPK signaling may contribute to the enhanced anticancer effect of the combination as well as to the reduced occurrence of skin lesions (e.g., cutaneous squamous cell carcinomas) associated with BRAF inhibitors used as single agents.⁸⁸^,⁹⁰ Two limitations are important to note: the combination does not extend the range of activity of dabrafenib or trametinib used as single agents, and resistance to the combination generally develops within 1 to 2 years.

Efforts to therapeutically exploit concurrent inhibition of the RAS/MAPK pathway (e.g., through the use of MEK inhibitors) and the PI3K pathway illustrate the challenges that face clinical development of novel/novel combinations.⁹¹ Both pathways promote cell proliferation and survival, and there is extensive crosstalk between the pathways such that inhibition of one pathway can lead to compensatory activation of the other. Multiple laboratories have presented preclinical evidence supporting the synergistic activity of combinations of MEK inhibitors with PI3K pathway inhibitors.⁹²^,⁹³ However, while concurrent inhibition of both pathways has the potential to exhibit favorable efficacy compared with inhibition of either pathway alone, in the clinic combinations that block both pathways appear to induce greater toxicity that limits the administration of adequate doses of the pathway inhibitors.⁹⁴^,⁹⁵ Another example of a novel/novel combination that has progressed to clinical evaluation is the combination of a MEK inhibitor with a CDK4/6 inhibitor for patients with NRAS mutated melanoma. Preclinical testing showed that MEK inhibition induced apoptosis but not cell cycle arrest for NRAS-mutant melanoma, and that combined pharmacological inhibition of MEK and CDK4/6 in vivo led to substantial synergy in therapeutic efficacy.⁹⁶ An evaluation of the clinical tolerance of this combination is being conducted using the MEK inhibitor trametinib and the CDK4/6 inhibitor palbociclib (NCT02065063).

Clinical experience has shown that the most effective combinations are those that are developed from agents with robust single agent activity and that can be combined together at their recommended single agent doses. The positive experience of adding imatinib to standard leukemia chemotherapy children with Ph⁺ ALL illustrates this point, as both imatinib and standard chemotherapy are active against this disease and they can be combined together without reductions in dosing.⁹⁷ Other examples that illustrate this paradigm include the addition of tretinoin to standard chemotherapy for APL,⁹⁸ the addition of rituximab to standard chemotherapy for adults with diffuse large B-cell leukemia,⁹⁹ and the addition of trastuzumab to chemotherapy for women with breast cancer.¹⁰⁰

Resistance to Molecularly Targeted Agents

Most molecularly targeted agents are effective against only a select population of patients whose tumors have the precise genomic alterations that confer sensitivity to the specific agent. While intrinsic resistance to targeted agents is expected in the absence of the specific genomic lesions associated with sensitivity, intrinsic resistance can also occur in the presence of the requisite genomic lesion. Melanoma with BRAF V600E mutation illustrates this point, as only one-half of patients with this lesion treated with a BRAF inhibitor show an objective response.¹⁰¹^,¹⁰² Limited response of melanoma patients to initial treatment with a BRAF inhibitor has been associated with coexisting genomic alterations such as PTEN deletion and with rapidly developing adaptive resistance mechanisms (e.g., through upregulation of receptor tyrosine kinase signaling pathways).¹⁰³^,¹⁰⁴ The low response rates to vemurafenib for BRAF-mutated colorectal cancer result from intrinsic resistance due to feedback activation of EGFR when BRAF is inhibited.¹²

Acquired resistance to molecularly targeted agents following a period of clinical response is the norm for targeted agents used to treat advanced cancers. For example, patients with NSCLC and EGFR mutations or ALK translocations and patients with BRAF-mutated melanoma typically show progressive disease within one to two years of treatment initiation.¹⁰⁵ The resistance that develops to single-agent targeted therapies following treatment with an effective targeted agent is virtually assured given the large number of cancer cells present at diagnosis and the biological heterogeneity that is present within the multitude of clones extant prior to treatment initiation.⁷³^,¹⁰⁶ Effective targeted agents provide strong selection pressure that provides a competitive advantage for the resistant clones, which over time grow out and eventually cause clinical resistance.

Acquired resistance can be divided into pharmacological mechanisms of resistance and biological mechanisms of resistance. Pharmacological resistance is the result of failure of drug delivery to the target. Examples include lack of compliance, changes in gastric absorption (e.g., by the initiation of antacid use), and the induction of metabolizing enzymes. The development of CNS metastases due to failure of adequate CNS penetration from blood-brain barrier effects can also be considered as pharmacological resistance, in that the resulting lower CNS drug levels allow micrometastatic disease present in the CNS to progress despite clinical benefit outside of the CNS. These pharmacological resistance mechanisms are not distinct for molecularly targeted agents and can also confer resistance to standard cytotoxic treatments.

Biological resistance mechanisms are generally distinctive to the specific molecularly targeted agent under consideration and include genomic alterations in the drug target as well as genomic alterations unrelated to the drug target that result in activation of bypass signaling pathways or reactivation of the targeted pathway. For tyrosine kinase inhibitors, acquisition of mutations in the targeted kinase that confer drug resistance is a common mechanism leading to disease progression. For EGFR mutant NSCLC treated with either gefitinib or erlotinib, more than 50% of progressing tumors have acquired the EGFR T790M mutation.¹⁰⁵^,¹⁰⁷ This mutation is analogous to “gatekeeper” mutations in other kinases, including the T315I mutation in BCR-ABL that confers resistance to imatinib and other ABL inhibitors,¹⁰⁸ the T670I mutation in KIT that confers imatinib resistance for GIST,¹⁰⁹ and the T674I in PDGFRA that confers imatinib resistance in hypereosinophilic syndrome.¹¹⁰ For crizotinib, acquired resistance among NSCLC patients with EML4-ALK fusion gene results from a variety of ALK mutations in approximately one-third of progressing cases, with no individual mutation predominating.¹¹¹^,¹¹² The acquisition of new BRAF point mutations is uncommon in BRAF inhibitor resistant melanoma. A distinctive acquired resistance mechanism for BRAF inhibitors is alternative splicing of BRAF V600E resulting in deletion of the RAS-binding domain of BRAF and leading to enhanced dimerization in cells with low levels of RAS activation.¹¹³^,¹¹⁴ Target kinase gene amplification is also observed as a resistance mechanism for BRAF inhibitors for BRAF V600 mutated melanoma and for crizotinib for ALK-rearranged NSCLC.¹¹²^,¹¹³

Genomic alterations not involving the target kinase but leading to reactivation of the target pathway are a resistance mechanism observed for BRAF V600 mutated melanoma. Only a minority of melanoma cases with acquired resistance to BRAF inhibitors have genomic alterations involving BRAF (e.g, alternative splicing or BRAF amplification), and yet the majority of cases show reactivation of the MAPK pathway. Some of these cases can be explained by mutations in RAS genes (primarily NRAS) or in a smaller percentage of cases by MEK1/MEK2 mutations.¹¹³

Genomic alterations leading to activation of bypass signaling pathways that allow cancer cells to circumvent a targeted agent’s pathway inhibition are an important resistance mechanism for some kinase inhibitors. MET amplification and HER amplification are observed in small percentages of patients with acquired resistance to EGFR TKI therapy.¹⁰⁷ For BRAF inhibitors for V600 mutated melanoma, PI3K-PTEN-AKT pathway mutations are observed as an acquired resistance pathway in approximately 20% of patients.¹¹³ Increased expression and activation of receptor tyrosine
kinases has been observed as another bypass resistance mechanism for patients with melanoma progressing on BRAF inhibitors.¹¹⁵^,¹¹⁶

The rapid development of resistance to effective targeted therapies when used as single agents for patients with advanced cancers is being addressed in several ways. It should be noted that CML is an exception to the rapid development of resistance, as many patients can be maintained on TKI therapy for years without disease progression.¹¹⁷ For those targeted agents for which the primary resistance mechanism relates to secondary genomic alterations in the drug target, developing alternative drugs that circumvent resistance has been an effective approach. For example, the T315I mutation of BCR-ABL is a primary mechanism of resistance to kinase inhibitors such as dasatinib and nilotinib, and a TKI that is effective against this mutation (ponatinib) is highly active in the clinic for patients with the T315I mutation.¹¹⁸ Similarly, third generation EGFR inhibitors (e.g., AZD9291 and CO-1686) selectively target mutant EGFR and are effective against the T790M mutation, and they are active in NSCLC patients who have developed the T790M mutation following initial treatment with erlotinib or gefitinib.¹¹⁹^,¹²⁰ Ceritinib is a second generation ALK inhibitor that is active against many, but not all, of the ALK mutations that confer resistance to crizotinib.¹²¹ It induced objective responses in over 50% of patients with advanced, ALK-rearranged NSCLC, including those who had disease progression during crizotinib treatment and those who had resistance mutations in ALK.¹²²

Utilizing combinations involving targeted agents is another approach being pursued to address the resistance that almost inevitably develops to treatment with a single targeted agent. While the strategy of dual therapy utilizing non-cross-resistant, effective targeted agents has theoretical validity,⁷³ the clinical application of combination therapies of targeted agents is challenging, as discussed previously. Such combinations may not be tolerated because of enhanced toxicities, limiting the ability to deliver effective doses of one or both agents. Additionally, the apparent multitude of potential bypass mechanisms that some cancers can use to circumvent blockade of oncogenic pathways creates challenges. From a pediatric perspective, a more promising approach is combining effective targeted agents with effective chemotherapy. Most childhood cancers show responsiveness to chemotherapy, and the high response rates of effective targeted agents following chemotherapy (both in adults and children) supports the lack of non-cross resistance between these targeted agents and chemotherapy. Proof of principle for this strategy is the marked improvement in longterm disease-free survival achieved with the addition of imatinib to standard chemotherapy for children with Ph⁺ ALL.⁹⁷ Another example of application of this strategy is an ongoing clinical trial (NCT01979536) in which crizotinib, which is highly active as a single agent for relapsed/recurrent anaplastic large cell lymphoma, is being added to a standard chemotherapy regimen for children with this disease.

Identifying Appropriate Doses and Schedules of Molecularly Targeted Agents

In the early years of targeted therapy development, there was enthusiasm for the concept of determining the “optimal biological dose” (OBD) for targeted agents rather than their maximum tolerated dose (MTD).¹²³ The OBD concept was based on the premise that the narrow therapeutic window for conventional cytotoxic agents would not apply to targeted agents, which because of their selectivity were hypothesized to show much greater differences between doses with biological and clinical activity and doses that produced unacceptable toxicity. However, experience with small molecule inhibitors developed as targeted agents has shown that most have narrow therapeutic windows, when a therapeutic window exists at all.¹²³ Examples of adverse events that limit dosing of targeted agents include rashes for EGFR inhibitors and multitargeted kinases,¹²⁴ stomatitis for mTOR inhibitors,¹²⁵ hypertension for VEGF pathway inhibitors,¹²⁶ diarrhea for Notch pathway inhibitors,¹²⁷^,¹²⁸ and vascular occlusion for the kinase inhibitor ponatinib.¹²⁹ Furthermore, since many small molecule targeted agents require continuous exposure for maximum effect, low-grade toxicities (e.g., rash or stomatitis) that may be tolerable when treatment is intermittent become intolerable when both treatment and toxicity are chronic. In more than 60% of cases, Phase 1 trials of small molecule targeted agents have had dosing limited by unacceptable adverse events.¹³⁰^,¹³¹ Monoclonal antibodies differ from small molecule inhibitors in that their dosing is most commonly not limited by toxicity, but rather by reaching a dose that saturates their targeted binding sites.¹²³

Successful use of a molecularly targeted agent in the clinical setting implies that the activity of the target has been successfully inhibited or blocked by the agent, with the depth of inhibition and the duration of inhibition being sufficient to achieve the desired effect. The depth and duration of inhibition required for antitumor effect can be empirically determined in preclinical in vivo models in which pharmacodynamic measures of target inhibition are measured at selected time points following agent administration and correlated to antitumor activity. Pharmacokinetic correlates of antitumor activity can also be determined, so that clinical development of the agent can proceed with both pharmacodynamic and pharmacokinetic benchmarks that need to be met in early phase clinical trials for the agent.¹³² As multiple specimens for pharmacokinetic data are easier to obtain than multiple specimens for pharmacodynamic studies, it is beneficial to have robust preclinical data defining the pharmacokinetic exposures associated with anticancer activity so that these exposures can be targeted in the initial clinical trials of targeted agents.

Practical and logistical issues complicate the approach of basing dosing decisions on target effects within tumor tissue. Sequential tumor biopsies are needed to assess target modulation, with pretherapy and posttherapy sampling required. Intratumoral heterogeneity may produce differences between pre- and posttreatment tissues unrelated to agent effects, a factor that is exacerbated by the small size of tissue specimens often obtained with sequential biopsies.¹³³ Variability may also be introduced by the postexcision processing and storage of biopsy specimens, as changes in the status of signaling pathways in tumor specimens can occur quickly following loss of blood supply.¹³³ A final source of variability is that associated with the assay procedures themselves. As with any clinical laboratory procedure, obtaining interpretable data requires that issues such as assay sensitivity, specificity, precision, accuracy, and linearity be satisfactorily addressed. It is essential that laboratory methods be adequately studied and validated in preclinical models prior to their application in clinical trials to inform dosing decisions.¹³³

Agent effects on surrogate tissues or on blood components may be used to inform dosing decisions. Some on-target effects on surrogate tissues may be clinically evaluable, such as the skin pigmentation changes caused by KIT inhibitors, the characteristic rash associated with EGFR inhibitors, and the rapid decrease in platelets that follows treatment with the BCL-2 inhibitor navitoclax.¹³⁴^,¹³⁵^,¹³⁶ Examples of changes in plasma proteins include the increase in VEGF that occur following treatment with small molecule inhibitors of VEGFR2 and the compensatory increase in IGF-1 and human growth hormone levels that follows administration of antibodies directed against the IGF-1R.¹³⁷^,¹³⁸ Differences between tumor and surrogate tissue in both extrinsic factors (e.g., degree of vascularization) and intrinsic cellular factors (e.g., the activity of drug efflux pumps and other resistance mechanisms) may limit the correlation between surrogate and tumor tissue molecular response to the targeted agent.

The complexities described above illustrate the challenges in performing phase 1 studies with dose escalation based on biologic endpoints determined by testing tumor tissue. As an example, while pharmacodynamic effects of the PARP inhibitor of veliparib could be documented at doses as low as 25 mg,¹³⁹ results of a phase 1 study showed that activity appeared greatest at the MTD (400 mg
twice daily) for patients with BRCA mutations.¹⁴⁰ An argument can be made for developing a targeted agent utilizing the highest dose with acceptable toxicity.¹⁴¹ Under the assumption that the effect of a targeted agent may plateau (but will not decrease) with increasing dose, this approach ensures that further testing of the agent will be performed at a dose with maximum activity and acceptable toxicity.¹⁴¹ However, even when dosing decisions are based primarily on toxicity or agent effect on surrogate tissue, it is important to define within one or more adult patient populations the effect of the agent on its target in tumor tissue at the dose and schedule being brought forward for further clinical evaluation. Documenting that the agent consistently achieves its desired target effect in tumor tissue for at least some cancers enhances confidence in the decision to go forward with further clinical development of the agent.

Precision Medicine and Clinical Trial Designs for Molecularly Targeted Agents

“Precision medicine” has been defined as “the tailoring of medical treatment to the individual characteristics of each patient.”¹⁴² As noted in a National Academy of Sciences report, precision medicine “does not literally mean the creation of drugs or medical devices that are unique to a patient, but rather the ability to classify individuals into subpopulations that differ in their susceptibility to a particular disease, in the biology and/or prognosis of those diseases they may develop, or in their response to a specific treatment.” Inherent in this concept are two components: a treatment that provides specific benefit for a subpopulation of patients and a well-validated diagnostic test to reliably identify this group of patients. The latter component highlights the importance of the companion diagnostics that must accompany clinical development of molecularly targeted agents.

The U.S. FDA has articulated a policy that requires the coapproval of a diagnostic with a therapeutic product when the diagnostic is essential to the safe and effective use of the therapeutic product.¹⁴³ Examples of coapproval include trastuzumab and its companion diagnostic HercepTest^TM, vemurafenib and its companion diagnostic for BRAF V600 mutation, and crizotinib and its companion diagnostic using ALK break apart FISH probes.¹⁴³ These examples illustrate a “one drug, one test” pathway for coapproval of a therapeutic agent and its companion diagnostic, but this regulatory pathway will certainly yield to diagnostics that test for multiple biomarkers simultaneously. The application of “next generation sequencing” (NGS) to cancer specimens best illustrates this new pathway. While research applications of NGS have become widespread, the rigorous validation steps required for clinical application, including determining suitable analytical sensitivity, specificity, accuracy, and precision across the reportable range of genomic alterations, are more recent.¹⁴⁴ Capabilities now exist for using NGS methods to reliably characterize clinically relevant base substitutions, short insertions and deletions (indels), copy number alterations, and selected fusions in a single test across hundreds of cancer-related genes using routine formalin-fixed and paraffin-embedded (FFPE) clinical specimens.¹⁴⁴

The ability to simultaneously determine for clinical use the status of hundreds of genomic alterations and the realization that only small subsets of patients will have any given genomic alteration has given rise to the development of new clinical trial designs for molecularly targeted anticancer agents. One example is the “basket” clinical trial design, which is a signal finding design in which eligibility criteria are agnostic to histology and only require the presence of a specific biological characteristic or genomic alteration.¹⁴⁵ The Imatinib Target Exploration Consortium Study B2225 was an early example of this design, as it allowed enrollment of patients with any malignancy with evidence for a potential role of one or more imatinib-sensitive tyrosine kinases.²⁵ Forty different malignancies were enrolled, with objective responses observed in five diagnoses with known genomic mechanisms of activation of imatinib target kinases. The initial pediatric clinical trial evaluating crizotinib (NCT00939770) used a basket-like approach for its phase 2 expansion, as patients with known genomic activation of ALK were eligible regardless of histology.¹⁴⁶ The pediatric phase 1-2 evaluation of vemurafenib used a similar design, with enrollment limited to patients with BRAF V600 mutation without regard for histology.¹⁴⁷

The “umbrella” clinical trial design enrolls patients and then assigns them to one of multiple treatments based on the results of biomarker evaluations.¹⁴⁸ An example is the Lung Cancer Master Protocol (Lung-MAP, S1400, NCT02154490) for patients with advanced squamous cell carcinoma whose tumors have progressed after frontline therapy.¹⁴⁸ Patients have genomic analysis performed at study entry and are then randomly assigned to either an agent targeting the abnormal pathway or to standard second-line therapy as follows: PI3KCA mutations (PI3K inhibitor GDC-0032), alterations in CDK4/6, CCND1, CCND2, and CCND3 (CDK4/6 inhibitor palbociclib), alterations in FGFR1, FGFR2, and FGFR3 (FGFR inhibitor AZD4547), and alterations in HGF/c-MET (anti-HGF monoclonal antibody rilotumumab). Patients without an abnormality in one of these pathways are randomly assigned to either an immunotherapeutic approach using antiprogrammed cell death 1 ligand 1 (PD-L1) or standard chemotherapy (docetaxel or gemcitabine). A phase 2-3 design is used for each of the randomizations, such that if encouraging activity is observed in the initial phase 2 component for any biomarker/agent group, accrual can be extended to 300 to 400 patients for a definitive assessment of the agent for this genomically defined patient population.

The NCI MATCH clinical trial uses an “umbrella” design with multiple molecularly based phase II studies embedded within the overall trial.¹⁴⁸ Adult cancer patients with any solid tumor or lymphoma that has progressed after at least one standard therapy for metastatic or advanced disease are eligible and must undergo biopsy at disease recurrence to have their tumors characterized for actionable genomic alterations. Patients with a genomic alteration that matches one of the 20 to 25 study agents are assigned to the treatment arm for this agent. Agent/gene pairs are selected on the basis of data showing the targeted agent has demonstrated activity in a human tumor carrying the genomic abnormality. Agents known to be inactive against a certain histology (e.g., BRAF inhibitors targeting V600E in colon cancer) are not evaluated against that histology. The protocol prescribes rules for assigning patients to agents based on the molecular findings from their biopsy, such that a tumor board is not required for decision making. Target accrual for each treatment arm is approximately 30 evaluable patients, with all meeting the molecular eligibility criteria but having a mixture of malignancies.

The ALCHEMIST trial is focused on the adjuvant treatment of early-stage adenocarcinoma of the lung.¹⁴⁸ Among this patient population, approximately 15% have EGFR mutation and 5% have ALK gene fusion. Patients are screened for these genomic alterations, and those with EGFR mutation and randomized to receive erlotinib or not following standard adjuvant chemotherapy, while those with ALK gene fusions are similarly randomized to receive crizotinib or not following standard adjuvant chemotherapy. The primary outcome measure for each randomization is overall survival. To obtain the required number of patients for each randomization (approximately 400 patients per randomization), 8000 patients will need to be screened.

The ALCHEMIST trial highlights the challenges that face application of the precision medicine concept in the childhood cancer setting. The proportion of patients with adenocarcinoma of the lung that have EGFR mutations is similar to the proportion of neuroblastoma patients that have activating ALK mutations. While it is feasible to screen 8000 patients with lung cancer to identify the approximately 400 with EGFR mutation required to conduct a randomized phase 3 trial, to attempt the same for children with high-risk neuroblastoma and ALK mutations would require decades of accrual. Identifying activity signals of single agent activity is not the problem, since 20 or fewer patients are required
to convincingly document the ability of an agent to induce objective responses in a biomarker-defined patient population. Standard pediatric phase 2 trial designs include 20 to 25 patients,¹⁴⁹ and the activity of crizotinib was clearly documented by fewer than 10 patients enrolled into an expansion cohort of a phase 1 study.¹⁴⁶ The challenge is identifying the contribution of a targeted agent when it is added to standard therapy for a small, genomically defined patient population. The standard therapy has some level of effectiveness, and the traditional approach of conducting a randomized phase III trial to reliably identify the contribution that the agent makes when added to standard therapy is not feasible because of patient numbers.

The development of imatinib for Ph⁺ ALL in children illustrates one approach to addressing this challenge.¹⁵⁰ In this case, results from a single-arm study convincingly demonstrated that imatinib added benefit to standard therapy. Three factors related to this example are noteworthy. First, imatinib had substantial single agent activity against the target patient population, increasing confidence that any effects observed with its addition to standard therapy were likely to be true effects. Second, there was a reasonably large, recent historical control population that allowed a comparison to be made between outcome for standard therapy with and without the addition of imatinib. Finally, the treatment effect observed with the addition of imatinib was very large, with 3-year EFS of 80% ± 11% with the addition of imatinib, more than twice that of historical controls (35% ± 4%).¹⁵⁰ The extent to which the imatinib example can be replicated for other agent/biomarker combinations will depend in part on the extent to which these factors are similarly represented. An example of applying the historical control approach to identify the contribution of novel agents added to standard therapy is the ANHL12P1 (NCT01979536), a randomized phase 2 clinical trial for children with anaplastic large cell lymphoma. This population is characterized by their genomic lesion (an ALK fusion gene) and by their uniform expression of the surface protein CD30, making these patients responsive to both crizotinib and brentuximab vedotin.¹⁴⁶^,¹⁵¹ Patients enrolled on ANHL12P1 receive standard therapy plus either crizotinib or brentuximab vedotin, and a total of approximately 140 patients are to be enrolled. The primary outcome measure is a comparison of the EFS for each arm to the estimated EFS for chemotherapy alone, such that with 70 or fewer patients per arm, one or both arms may be identified as superior to chemotherapy alone.

An alternative clinical trial design for small patient populations is to randomize and target large treatment effects, with or without inflated Type I error. When large treatment effects are targeted, small patient numbers are required, as illustrated by the evaluation of a scorpion antivenom in children using a randomized design with a minimum sample size of 14 patients.¹⁵² Likewise, using type I error rates greater than the standard two-sided 0.05 reduces the numbers of patients required for any given targeted effect size.¹⁵³ Because historical controls are often not available for genomically defined patient populations, randomization may be essential for identifying the contribution of a targeted agent when it is added to standard therapy.

Examples of applications of the precision medicine concept to pediatric oncology, in addition to those described above, include a clinical trial of the MEK inhibitor selumetinib for patients with low-grade astrocytoma (NCT01089101) and clinical trials of the Smoothened (SMO) inhibitors vismodegib and sonidegib for children with sonic hedgehog (SHH) pathway activated medulloblastoma (NCT01239316 and NCT01708174, respectively). In the former, a phase 2 expansion cohort is evaluating the activity of selumetinib for different genomically defined groups, including patients with the BRAF V600E mutation and/or KIAA1549-BRAF fusion and patients with neurofibromatosis 1 (NF-1). The evaluations of Hedgehog pathway inhibitors for medulloblastoma are complicated by the relatively small patient population (<30% of all medulloblastoma) and their overall relatively favorable prognosis.¹⁵⁴ Additionally, genomic alterations in several different genes (e.g., SUFU, PTCH1, GLI2) result in SHH pathway activation, and some of these genomic alterations (e.g., SUFU mutations and GLI2 amplification) are not responsive to Smoothened inhibitors.¹⁹

Special Considerations for Clinical Trials for Molecularly Targeted Agents in Children

Allowable Risk to Children in Clinical Research

Unlike adults considering research participation, younger children lack the capacity to decide whether they want to accept risks of harm in return for uncertain benefits or for the common good.¹⁵⁵ This lack of capacity has important implications for how molecularly targeted agents can be studied in children with cancer. In recognition of this lack of capacity, children are afforded special protections from research risk under federal regulation 45 CFR §46, Subpart D and the related FDA regulation 21 CFR §50, Subpart D. Subpart D defines allowable research based on the risk to subjects associated with the research components of the clinical trial, and it has a direct impact on how molecularly targeted agents can be developed for children with cancer. Children may participate in research that poses greater than minimal risk, but only if “the risk is justified by the anticipated benefit to the subject and the relationship of the risk to benefit is at least as favorable as any available alternative approach” (45 CFR §46.405 and 21 CFR §50.52). Thus, children with cancer who have exhausted known effective therapy options may participate in phase 1 trials since they offer the potential for direct benefit.¹⁵⁶ However, tissue collection for correlative biological studies is generally performed for research purposes only and is therefore without the prospect of direct benefit. Subpart D stipulates that children with disorders or conditions may participate in research with “no prospect of direct benefit” to individual subjects provided: 1) the risk represents no more than a minor increase over minimal risk, 2) the intervention or procedure presents experiences to subjects that are reasonably commensurate with those inherent in their actual or expected medical setting, and 3) the intervention is likely to yield generalizable knowledge about the subject’s disorder that is of vital importance for the understanding or amelioration of the subject’s disorder (45 CFR §46.406 and 21 CFR §50.53). Thus, the risks associated with procedures for collecting tissue for biological studies to evaluate the effects of a molecularly targeted agent in a child receiving the agent must be no greater than a minor increase over minimal risk in order to be approved under 45 CFR §46.406, effectively ruling out biopsy of solid tumors strictly for research purposes other than for those with superficial locations.

Defining Recommended Phase 2 Doses (RP2D) for Molecularly Targeted Agents in Children

How then should phase 1 studies of molecularly targeted agents be conducted in children? If a molecularly targeted agent studied in adults has a clear dose-limiting toxicity and further development of the agent is using a dose based on tolerability, then it is appropriate for the initial study in children to use the standard pediatric phase 1 design that calls for beginning at approximately 80% of the adult MTD or RP2D and then escalating the dose based on tolerability. As the pediatric MTD is likely to be close to the adult MTD, dose escalation can generally be limited to one or two dose levels above the adult MTD.¹⁵⁷^,¹⁵⁸ Pharmacokinetic data should be obtained to compare with data from adults and to compare with systemic exposures associated with activity in preclinical models (when available). For targeted agents for which an MTD was not defined in adults because of tolerability at doses that met target-effect endpoints or pharmacokinetic parameter endpoints prior to reaching MTD, dose escalation in the initial pediatric study can be minimized. The dose tolerated in adults can be studied as the initial dose, with perhaps exploration of one or two higher doses. Comparison of drug levels between children and adults at the
tested dose levels can then allow selection of an appropriate dose for further evaluation in children.

Some molecularly targeted agents have clinical development plans that are primarily for use in combination with standard agents (e.g., inhibitors of DNA repair that are developed to potentiate the effects of chemotherapy agents). Rather than conduct a phase 1 study of the targeted agent administered alone and then a second phase 1 study in which the targeted agent is used in combination with additional agents, a single phase 1 study of the targeted agent studied in combination with other agents can be performed. For such trials, it is helpful, when possible, to initially administer the agent alone in each patient to evaluate the pharmacokinetics of the targeted agent in isolation, followed by administration of the agent in combination with other agents. Correct attribution of adverse events to a new agent that is added to a standard drug regimen can be difficult. This task is made easier when the standard regimen has a well-defined pattern of toxicity and when the new agent and the standard regimen have nonoverlapping toxicity profiles.

How can the effect of molecularly targeted agents on their target at their RP2D be evaluated, either directly or indirectly, in children? As noted above, sequential biopsies of tumor tissue will have limited applicability in children because of the research risks associated with biopsy procedures, and, hence, alternative methods are required. An indirect estimate of effect can be achieved by targeting drug exposure levels associated with the desired level of tumor target modulation in preclinical models or in adults. This approach is analogous to that taken in the development of the topoisomerase I inhibitor irinotecan, in which systemic exposures of irinotecan associated with anticancer activity against pediatric xenografts were targeted in a phase 1 trial of irinotecan. Achievement of the intended irinotecan serum concentrations in patients was associated with objective responses. The strategy of targeting drug levels known to be efficacious in preclinical models was also successfully applied on a phase 2 evaluation of topotecan for children with neuroblastoma.⁶¹ Another indirect approach for assessing target modulation is to utilize surrogate tissues (e.g., skin, buccal mucosa, peripheral blood mononuclear cells) for evaluating the biological effects of agents. The caveats described previously for the use of surrogate tissues in adult clinical trials apply equally in the pediatric setting, and the contribution of studies using surrogate tissues is questionable.

The biological effects of targeted agents on cancer cells in children can be directly studied in certain situations. Patients with leukemia provide the greatest opportunities for sequential evaluations of agent effects on cancer cells, given the relative ease with which either peripheral blood or bone marrow can be sampled. For patients with solid tumors, circulating cancer cells have been proposed as a potential resource for sequentially evaluating targeted agent effect.¹⁵⁹ Questions about the general applicability of this approach include whether sufficient numbers of circulating cancer cells can be collected to support the necessary laboratory testing and whether cancer cells circulating in the blood respond to treatment in the same way as do cancer cells within tumor masses. Another approach for studying the effect of targeted agents on tumor tissue for children with solid tumors is to plan administration of the agent around the timing of a clinically indicated surgical procedure.¹⁶⁰ Tumor tissue collected after administration of the agent for a defined period can be analyzed to determine the biological effects of the agent on the tumor. Given the risk to patients that would result from pretreatment tumor biopsies, comparison in most cases will need to be made with tumor tissues collected from patients not receiving the agent, rather than to tumor tissue from the same patient collected prior to treatment. A prerequisite for application of this design is convincing data that administration of the agent is unlikely to complicate either the surgical procedure or the recovery following surgery.

The application of molecular imaging methods, though to date limited, may become increasingly central to the development of molecularly targeted agents in children and may allow childhood cancer researchers to circumvent the inherent limitations resulting from restrictions on performing serial tumor biopsies in children. These imaging methods may allow researchers to sequentially monitor whether an agent had its intended biological effect on its target, and if so, how long this effect was maintained. To date, the use of imaging to monitor pharmacodynamic effects in pediatric clinical trials has been primarily focused on evaluating the effects of antiangiogenic agents.¹⁶¹

Toxicity Concerns for the Use of Molecularly Targeted Agents in Children

There are general issues about toxicities associated with molecularly targeted agents that apply to both adults and children. While presentations touting the promise of molecularly targeted agents often include reference to their lack of toxicity, this perception does not match reality for many “molecularly targeted” agents. Although these agents often lack the myelosuppression and gastrointestinal toxicities associated with cytotoxic agents that target rapidly cycling cells, many of the agents induce significant mechanism-based adverse events that affect their clinical development. For example, Notch pathway inhibitors cause diarrhea as a result of the goblet cell hyperplasia that occurs when Notch signaling is blocked.¹²⁷ Inhibitors of BCL-X_L cause a rapid-onset thrombocytopenia as a result of the dependence of platelets on BCL-X_L for survival in the circulation.¹⁶²^,¹⁶³ Cardiac toxicity has been associated with trastuzumab and with multitargeted receptor tyrosine kinases.¹⁶⁴ Some kinase inhibitors, including mTOR inhibitors, cause rashes and stomatitis that limit dosing. Thus, assumptions of lack of toxicity are misplaced for molecularly targeted agents, and as with conventional cytotoxic agents, the acute and long-term risks from their use will need to be carefully weighed against the evidence for their benefit.

The maxim that children are not simply young adults applies to the clinical development of molecularly targeted agents, just as it does during development of conventional cytotoxic agents. As with any therapeutic agent, children may tolerate molecularly targeted agents differently than adults because of age-related changes in physiology that result in pharmacokinetic differences between children and adults. In addition, molecularly targeted agents may specifically interfere with critical developmental pathways and block the progression from immature to mature adult tissues. For example, the central roles of the Wnt and Hedgehog signaling pathways in bone development will have to be considered as agents blocking these pathways enter clinical evaluation,¹⁶⁵^,¹⁶⁶ a point that was emphasized by the permanent defects in bone structure in young mice caused by transient inhibition of the Hedgehog pathway.¹⁶⁷

The effect of VEGF inhibitors on the growth of immature animals illustrates the distinctive challenges for the pediatric development of molecularly targeted agents.¹⁶⁸ Vascularization of cartilage in long bones is observed during three periods in the life of vertebrates: during late embryonic development, during periods of rapid growth in immature animals as capillary structures invade at the growth plate regions, and during adulthood when angiogenesis is activated for bone remodeling in response to bone injury or other pathologic conditions.¹⁶⁹ Both antibodies that sequester vascular VEGF and small molecule inhibitors of VEGF receptors produce a characteristic effect on growth plates in immature mice and nonhuman primates.¹⁶⁸^,¹⁷⁰^,¹⁷¹ In these animals, blood vessel invasion of growth plate cartilage is markedly suppressed, concomitant with impaired trabecular bone formation and marked expansion of the hypertrophic chondrocyte zone. Importantly, changes associated with relatively brief intervals of angiogenesis inhibition are reversible.¹⁷¹

Bone effects have been observed in a small number of children treated with VEGF-pathway targeted agents. An infant with cutaneovisceral angiomatosis with thrombocytopenia (CAT) syndrome
receiving bevacizumab developed asymptomatic metaphyseal bone lesions that reversed following cessation of bevacizumab.¹⁷² Physeal widening was observed in the pediatric phase 1 trial of pazopanib.¹⁷³ An 11-year-old Tanner stage 2 patient showed a threefold expansion by cycle 10 of pazopanib as well as a decrease in height velocity. Three other patients on this trial (ages 11, 8, and 4 years) also showed radiographic growth plate widening after four cycles of pazopanib, but were removed as a result of progressive disease. That physeal widening has not been reported for other VEGF pathway inhibitors studied in children with cancer may reflect the small number of children studied and the limited duration of treatment and follow-up that is common for phase 1 clinical trials. ¹⁷⁴^,¹⁷⁵^,¹⁷⁶ Three children with recurrent brain tumors treated with bevacizumab in combination with irinotecan developed osteonecrosis in the wrist or the knee, adding another bone toxicity to monitor for in children treated with VEGF pathway inhibitors.¹⁷⁷ The limited clinical experience with VEGF pathway inhibitors in children with cancer prevents the drawing of conclusions about the long-term impact of this class of agents on bone growth in children.

Growth retardation has been documented in children receiving long-term treatment with imatinib for CML.¹⁷⁸^,¹⁷⁹^,¹⁸⁰ The growth inhibitory effects of imatinib mesylate appear to be most pronounced in prepubertal children, compared with pubertal children,¹⁷⁸^,¹⁷⁹ and one study reported that growth velocity tended to recuperate in prepubertal children with growth impairment, as they reached pubertal age.¹⁷⁸ There is evidence for imatinib-induced growth failure in children that results from perturbations of the growth hormone (GH):IGF-1 axis, with reduced IGF-1 and IGFBP-3 levels and reduced response in growth hormone stimulation tests observed in children receiving imatinib.¹⁸⁰^,¹⁸¹^,¹⁸² Of note, second generation tyrosine kinase inhibitors such as dasatinib and bosutinib reduced IGFBP-3 levels in rats in a similar manner as imatinib, suggesting that the effects on the GH:IGF-1 axis may represent a class effect.¹⁸² It is unknown whether GH replacement therapy for children with CML treated with imatinib is safe and effective at reversing growth retardation.

The acute and long-term effects of molecularly targeted agents described above highlight the need for careful monitoring of children receiving these agents. To date many children who have received molecularly targeted agents have had relapsed/refractory disease, and long-term survival was not expected. As more and more children with curative conditions receive these agents as part of frontline therapy (e.g., imatinib for CML, and Ph⁺ ALL and crizotinib for anaplastic large cell lymphoma), it will be particularly important to implement long-term follow-up programs so that late effects can be identified as promptly as possible and prevention and remedial actions taken.

THERAPIES TARGETED TO APOPTOSIS PATHWAYS

A hallmark of cancer is the ability to evade apoptosis (programmed cell death).¹⁸³ The acquisition by cancer cells of defects in programmed cell death that provide a survival advantage during tumorigenesis can also provide cancer cells with intrinsic resistance to treatment approaches such as chemotherapy and radiation therapy. An obvious approach to enhancing the effectiveness of cancer therapy is by manipulating the cancer cell’s dysregulated apoptosis pathways to favor cell death. While conceptually simple, achieving clinical benefit through this approach is far from trivial. Apoptotic pathways are complex with multiple steps at which progression toward cell death may be blocked. In addition, in order for a therapeutic window to exist, the specific apoptosis pathway target of an agent must be one upon which cancer cells are critically dependent and one for which normal cells have much less need. The two primary apoptotic pathways, the extrinsic (death receptor) pathway and the intrinsic (mitochondrial) pathway, are described in Figure 11.1.

BCL-2 Family Proteins as Therapeutic Targets

Oligomerization of the proapoptotic BCL-2 family members BAX and BAK in the mitochondrial membrane is a critical step in the mitochondrial apoptosis pathway.¹⁸⁴^,¹⁸⁵ BAX and BAK oligomerization is dependent upon the overall balance between the expression and activation of the remaining more than twenty BCL-2 family proteins. Each of these proteins contains one or more BCL-2 homology (BH) domains, and they can be divided into those that promote survival and those that promote apoptosis. The prosurvival family members (e.g., BCL-2, BCL-XL, MCL-1, and BFL-1) contain BH domains 1-4 and act directly or indirectly to block the apoptosis-inducing activities of BAX and BAK. The proapoptotic BCL-2 family members can be divided into “effectors” (BAX and BAK), “direct activators” (BID, BIM, and potentially PUMA), and “sensitizers” (BAD, BIK, BMF, HRK, PUMA, and NOXA). Like the prosurvival BCL-2 family members, the proapoptotic family members BAX and BAK contain four BH domains (BH1-BH4), while the direct activators and sensitizers contain a single BH domain, BH3.

Apoptosis occurs when BAX and BAK change their structure from inert monomers into homo-oligomers, leading to mitochondrial outer membrane permeabilization (MOMP) and cytochrome C release into the cytoplasm, resulting in caspase activation.¹⁸⁵ This transition can be initiated by binding of the BH3-only activators BID and BIM to the effectors, BAX and BAK. The induction of apoptosis is limited by binding of prosurvival BCL-2 members to BAX and BAK as well as to activators such as BID and BIM. The remaining BH3-only BCL-2 family members that serve a sensitizing role promote apoptosis by competing for binding to prosurvival BCL-2 family members, thereby freeing the effectors and activators to induce apoptosis.

In considering therapeutic applications related to BCL-2 family proteins, an important concept is that some cancers are primed to undergo apoptosis because of their expression of proapoptotic BH3-only proteins.¹⁸⁶ Because of their high level of expression of BH3-only proteins, these cancers are addicted to the concurrent high-level expression of prosurvival BCL-2 family members. For these cancers, inhibitors of antiapoptotic BCL-2 members can result in apoptosis and in potential clinical benefit. As an example, chronic lymphocytic leukemia (CLL) is a lymphoid malignancy that appears to be primed for apoptosis through specific inhibition of BCL-2.¹⁸⁷^,¹⁸⁸^,¹⁸⁹ Likewise, many ALL leukemia cells are dependent for survival on the apoptotic block provided by BCL-2 family expression;¹⁹⁰ however, the relative dependence upon specific BCL-2 family members may vary by subtype. ALL cases with MLL gene rearrangement appear to be particularly responsive to BCL-2 inhibition.¹⁸⁷^,¹⁹¹ For T-cell ALL, inhibition of BCL-XL may be required for most cases for apoptosis induction, with only cases with the early T-cell phenotype (ETP) showing sensitivity to BCL-2 inhibition alone.¹⁹²^,¹⁹³ These examples highlight the complexities of modulating BCL-2 family members as a therapeutic strategy.

Agents targeting antiapoptotic BCL-2 family members have entered clinical evaluation. While the BCL-2 antisense agent, oblimersen (G3139, Genasense), was the first BCL-2 inhibitor to proceed to extensive clinical testing, results for this agent were disappointing,¹⁹⁴^,¹⁹⁵^,¹⁹⁶^,¹⁹⁷ likely a result of inadequate levels of target inhibition. Small molecule BCL-2 family inhibitors are in clinical evaluation, with these agents differing in the range of prosurvival BCL-2 family members that they effectively inhibit. Navitoclax (ABT-263) is an orally bioavailable BH3 mimetic that potently inhibits the prosurvival activity of BCL-2, BCL-X_L, and BCL-W, but not that of MCL-1 or BFL-1.¹⁹⁸ Navitoclax-induced regressions as a single agent against small-cell lung cancer and ALL xenograft models.¹⁹⁸^,¹⁹⁹ The PPTP studied navitoclax against its childhood cancer preclinical models. Limited single agent activity was observed against solid tumor xenografts, but remissioninducing activity was noted for approximately 50% of the ALL xenografts tested.⁷⁰ Extended testing of ALL models showed
remission-inducing activity across multiple ALL subtypes, with resistance to navitoclax appearing related to MCL-1, a BCL-2 prosurvival family member that navitoclax does not inhibit.²⁰⁰

Figure 11.1 Simplified schema of extrinsic and intrinsic apoptosis pathways. The extrinsic (death receptor) pathway is initiated by ligation and clustering of members of the death receptor superfamily (e.g., tumor necrosis factor [TNF] receptor I, Fas, and the TNF-related apoptosis-inducing ligand [TRAIL/Apo-2L] receptors TRAIL-R1 [DR4] and TRAIL-R2 [DR5]). Receptor clustering leads to formation of death-inducing signal complexes (DISC) by recruitment of adapter proteins (e.g., Fas-associated death domain [FADD]), which can then attract procaspase-8. The procaspase-8 that is recruited to the complex is converted by proteolytic cleavage to caspase-8, which can then activate downstream effector caspases leading to apoptosis. The pathway to apoptosis in the presence of Smac mimetics requires TNF receptor signaling, and is initiated by cIAP1/2 auto-ubiquitination and degradation, which eventually leads to apoptosis through caspase-8 activation in a complex with RIP1 and FADD or to necroptosis via a complex with RIP1 and RIP3. The intrinsic (mitochondrial) pathway is responsive to internal toxic stimuli (e.g., DNA damage, disruption of microtubules, etc.). These stimuli result in increased activity of proapoptotic BH3-only “sensitizers” members of the BCL-2 family, which inhibit BCL-2 and BCL-XL function, leading to oligomerization of BAX and BAK in mitochondrial membranes. Alternatively, the toxic stimuli can increase activity of the BH3-only “direct activators,” which also results in BAX/BAK oligomerization in the mitochondrial membrane. The interaction of BAX and BAK with the mitochondrial membrane results in the release into the cytoplasm of cytochrome c and other proapoptotic factors, leading to the formation of a cytoplasmic complex (the “apoptosome”) that includes cytochrome c and apoptotic protease-activating factor-1 (Apaf-1). Formation of the apoptosome results in the production of active caspase-9, which then activates downstream effector caspases. Cross-talk between the two apoptosis pathways occurs by caspases-8 cleavage and activation of the BH3-only protein BID, which can then bind to BAX and BAK, promoting their mitochondrial membrane insertion and oligomerization and leading to activation of the intrinsic pathway. XIAP can suppress apoptosis by binding to and inhibiting caspases.

Navitoclax demonstrated impressive activity in patients with CLL and selected types of non-Hodgkin lymphoma in the phase 1 setting.²⁰¹^,²⁰² Despite promising preclinical data for small cell lung cancer, navitoclax showed little single-agent activity for patients with this cancer,²⁰³ and its primary utility appears to be for lymphoid malignancies such as CLL. Navitoclax dose escalation was limited by rapid-onset, reversible thrombocytopenia.²⁰¹^,²⁰² Thrombocytopenia is an on-target effect of navitoclax, as BCL-X_L plays an essential role in promoting platelet survival.¹⁶³ While the single agent activity of navitoclax against ALL xenografts, its activity in adults with CLL, and its ability to potentiate the cytotoxic effects of standard antileukemia agents all support testing of navitoclax for childhood ALL,⁷⁰^,¹⁹⁸^,²⁰⁰^,²⁰⁴ the navitoclax-induced thrombocytopenia has limited its clinical evaluation in this setting.

ABT-199 was developed to circumvent the platelet liabilities associated with navitoclax. ABT-199 specifically inhibits BCL-2 at nanomolar concentrations with little effect on BCL-XL. As would be predicted by this inhibitory profile, ABT-199 showed minimal in vitro and in vivo effects on platelets.²⁰⁵ Remission-inducing in vivo activity for ABT-199 was demonstrated for selected leukemia and lymphoma preclinical models, and it also enhanced the activity of standard of care agents such as rituximab and bendamustine when used in combination with them.¹⁸⁹^,¹⁹³^,²⁰⁵ In the clinic, thrombocytopenia has not been dose limiting for ABT-199, in contrast to clinical experience with navitoclax. ABT-199 induced a high rate of response in patients with relapsed/refractory CLL, with approximately 80% showing objective responses, of which approximately 20% were complete responses.²⁰⁶ A challenge in developing ABT-199 for pediatric leukemias will be enriching for patient populations most likely to respond, as illustrated by the finding of preclinical activity of ABT-199 for T-cell ALL that was restricted to the ETP subset.¹⁹²^,¹⁹³

Inhibitor of Apoptosis Proteins (IAP) as Therapeutic Targets

IAPs are evolutionarily conserved cytoplasmic proteins that were initially studied for their ability to suppress apoptosis through direct inhibition of activated caspases.²⁰⁷ Eight human IAPs have been identified and all share a protein domain, the baculovirus IAP repeat (BIR), that is important for antiapoptotic activity. XIAP is the only member of the human IAP family that actually inhibits apoptosis through interactions with caspases, and it inhibits
caspases at both the initiation phase (caspase-9) and the execution phase (caspase-3 and -7).²⁰⁸ A more recently recognized oncogenic role for IAP family members is in regulating NF-κB signaling, as described below.

The second mitochondria-derived activator of caspases (Smac) is central to understanding the cellular role of IAPs.²⁰⁷^,²⁰⁸ Smac under normal conditions is sequestered in mitochondria, but in the presence of mitochondrial dysfunction or apoptosis is released into the cytosol. The N-terminal tetrapeptide (AVPI) of cytosolic Smac selectively binds to IAPs through their BIR domains, and in doing so competes for the interaction between XIAP and caspases-3, -7, and -9.

Only gold members can continue reading. Log In or Register to continue