From Bench to Bedside with Targeted Therapies

The target for therapy switched from cytoplasmic hormone receptors to growth factor receptors on the cell surface in the early 1980s. Our group first hypothesized that inhibition of a receptor function might inhibit tumor cell growth (Table 45-1 ). We produced a monoclonal antibody, cetuximab, which binds to the EGF receptor, blocks activation of the receptor’s tyrosine kinase by its ligands, EGF and TGFα, and inhibits proliferation of cancer cells in culture and in human tumor xenografts. ^11–13 This was followed by reports on trastuzumab, a monoclonal antibody that binds to the closely related HER-2 receptor. ¹⁴ It was during this period that the term oncogene was first used, and these were the first experimental cancer therapies that targeted the products of an oncogene—in these cases a receptor tyrosine kinase. Since the mid-1980s, many oral chemotherapeutic agents have been developed that target these two tyrosine kinases and many other kinases, located both in receptors and free in the cytoplasm.

Trastuzumab received regulatory approval for the treatment of breast cancer expressing high levels of HER-2 in 1998. ¹⁵ This was followed by imatinib against the product of the BCR-ABL gene translocation, another tyrosine kinase, in patients with chronic myelocytic leukemia in 2001. ¹⁶ Cetuximab entered clinical trials in 1990 and was approved for colorectal cancer in 2004 ^17,18 and for head and neck cancer in 2006. These approaches to experimental therapeutics created a new paradigm for the development of therapies against cancer.

Three breakthroughs in knowledge and technology converged to greatly accelerate this change in therapeutic research on cancer:

As a result of this confluence of discoveries, there is pressure from all stakeholders in cancer care for development of new approaches to the treatment of cancer—using the genomic and molecular analysis of an individual patient’s cancer to select the targeted therapy that is most likely to provide clinical benefit. Although earlier approaches provided many successful chemotherapeutic agents and continue to be important, the new targeted approach to developing new cancer therapies dominates the field today. These new therapies are designed to target a particular gene product that has been shown to contribute to the malignant phenotype of cancer cells in correlative studies of molecular pathological data and clinical data and in laboratory experiments with cultured cells and animal models. ^20,21

There have been some spectacular successes in clinical trials with experimental targeted therapies that enrolled only patients whose cancer was known to harbor an aberrancy in the target of that experimental drug or antibody. These successes have resulted in substantial prolongation of patients’ lives for many months or years.

The first example of this approach to reach the clinic involved trastuzumab, the monoclonal antibody against the HER-2 receptor. It is overexpressed on 20% to 25% of breast cancers and was found to be a biomarker predicting a worse prognosis in these patients. Clinical activity of trastuzumab was demonstrated in early studies only among patients whose cancers expressed high levels of HER-2. Genentech adopted a new paradigm for designing a randomized Phase III clinical trial, investigating standard chemotherapy with or without trastuzumab for treatment of patients with advanced, metastatic breast cancer. Only patients with high expression of HER-2 in their tumors were eligible to enroll. This enriched the trial with patients more likely to respond, and converted what would otherwise have been a negative study (too few responders) into a positive study, leading to FDA approval in 2002. ¹⁵ The optimal clinical situations for the use of trastuzumab have been refined continuously, as a result of dozens of clinical trials over the ensuing decade (see Chapters 36 and 50).

A second and most dramatic example involves the development of the ABL tyrosine kinase inhibitor imatinib for the treatment of patients with chronic myelocytic leukemia (CML), whose leukocytes nearly always carry a BCR/ABL gene rearrangement. Before this treatment, the median survival of patients with this disease was 3½ years. Today most patients on this and follow-up drugs live in remission more than 5 years, and some may reach a normal lifespan. ¹⁶ The incidence of CML is only 4000 cases per year. A drug against this disease was not considered to be a potential “blockbuster” compared with a treatment for lung cancer (over 200,000 cases/year). Movement from the research laboratory in a pharmaceutical company to clinical trials was encouraged by the persistent efforts of an academic clinician scientist, Brian Druker, who performed preclinical studies in his laboratory demonstrating efficacy against CML cells and then led the first clinical trials. The premise, which turned out to be prescient, was that blocking the abnormally active ABL tyrosine kinase—a product of the BCR-ABL rearrangement in patients with early CML—might produce especially effective responses, because this was the only known genetic abnormality in these patients. In contrast, most cancers have many genetic aberrations that contribute to their malignant activities at the time they are first detected.

A third example involves oral inhibitors of the EGF receptor tyrosine kinase, gefitinib and erlotinib (TKIs). In a series of very large and expensive randomized trials of chemotherapy plus or minus one or the other of these TKIs in patients with advanced lung cancer, the results showed no significant overall benefit from adding the TKI. However, a few patients had a substantial prolongation of survival. When it was discovered that mutations were present in the gene encoding the EGF receptor in the lung cancers of many of the patients who responded well, ^22–24 clinical trials were repeated with the requirement of an EGF receptor mutation for enrollment. The positive response rates for advanced lung cancer patients with a mutated EGF receptor were confirmed. ²⁵ This time, in the selected patient population, the results were positive in the majority of patients. Thus, a “failed” new targeted therapy was converted into a successful treatment for the 10% of lung cancer patients with mutated EGF receptor genes in their tumors, and the median duration of survival for those patients has doubled. In fact, for these patients the TKI therapy was more effective than chemotherapy, whereas for patients without the mutation, chemotherapy gave better outcomes than treatment with a TKI. ²⁵ This “resurrection of a cancer drug” brought clinical benefits to selected patients and economic benefits to the pharmaceutical companies, which were now able to market their new targeted drugs. This was a wake-up call to the industry, strongly suggesting the value of a new, selective approach to clinical trials with new targeted cancer treatments.

Among the lessons from these examples is the observation that for each case the circumstances were different, and an understanding of both the molecular biology and the clinical disease was critical in the development of a novel therapy that targeted a molecular abnormality in these patients’ cancers.

Von Hoff was the first to report on a clinical trial in which patient enrollment was governed by the result of a molecular analysis of tumors, using immunohistochemistry, fluorescence in situ hybridization, and, primarily, gene expression arrays. Of 66 patients treated with commercially available drugs that were felt to match the detected molecular markers, 27% had a progression-free survival at least 30% longer compared with the response to the previous drug(s) they had received. ²⁶ In this study, the patients served as their own controls, and gene mutations were not analyzed.

Recent examples have demonstrated the utility of identifying the targeted genetic aberration in the tumors of patients who enroll in clinical trials with experimental targeted therapies at the earliest stages.

The Biomarker-integrated Approaches of Targeted Therapy for Lung Cancer Elimination (BATTLE) trial for patients with advanced, heavily treated non–small-cell lung cancer was the first reported study with a panel of experimental drugs that (1) required a fresh tumor biopsy for genetic and molecular assays to detect particular aberrations that could serve as biomarkers for assigning therapy and (2) introduced an adaptive randomization design to assign a patient to receive one of the four experimental therapies that did, or did not, target an abnormality detected in that patient’s cancer. Essentially, this was a panel of four separate Phase II trials, testing the efficacy of “rationally” assigning patients to a particular trial. Importantly, nearly 100% of patients who were offered participation agreed to the biopsy procedure as a condition for receiving access to one of the experimental targeted therapies in a randomized study. The endpoint of the trial was improved duration of progression-free survival, which was subsequently confirmed by improved overall survival. The results showed that there were four situations where the results of the biomarker assays predicted a significant improvement in response, or in one case, a worse response, to an experimental drug targeting that biomarker. ²⁷ This suggested that preselection of patients for a panel of Phase II studies of new therapeutic agents based on biomarkers showing abnormalities in the targets of these agents is useful and may improve clinical outcomes.

Another trial at MD Anderson Cancer Center retrospectively analyzed response rates and survival of patients treated with experimental drugs that matched genetic abnormalities in their cancers. Over a 4-year period, more than 1000 patients with advanced, heavily treated cancers were tested for mutations in hot spots in 11 genes and for loss of expression of PTEN. An experimental drug was often available targeting the detected genetic aberration for most of their genes in ongoing Phase I/II trials. A total of 291 patients with genetic aberrations could be evaluated for response to an experimental targeted therapy. Of these, 175 received a targeted therapy that matched an aberrant gene target in their cancer, and 116 were treated with nonmatched therapy (based on nonavailability of matched therapy at the time). The overall response rate of the first group was 27% (including two complete responses), significantly greater than the second group, which was 5%. Median overall survival was 13.4 months compared to 9.0 months, a 49% increase. ²⁸ In most reports analyzing the results of Phase I trials, the main objective is determination of maximal tolerated dose and toxicities, and the objective response rate to new therapies is less than 10%. This trial, which achieved a much higher response rate, was a nonrandomized study and involved many types of cancer and many separate clinical trials of new drugs at varying doses. Although the results support the benefit of matching experimental therapy to genetic aberrations, this conclusion now must be confirmed in a randomized study. Today a number of cancer centers are carrying out prospective clinical trials using next-generation sequencing technology to assess cancers and using the results for assigning patients to trials of experimental targeted therapies.

Two recent highly successful Phase II trials of experimental targeted drugs took an approach to testing a new experimental therapy similar to that used in the registration trial of trastuzumab, the trials of TKIs in patients with mutated EGF receptors, and in most of the patients with CML treated with imatinib. Each trial studied a single drug. Large numbers of patients were screened for enrollment, based on the presence of the genetic aberration that the experimental drug was designed to target.

In the first clinical trial, crizotinib, which already was in trials targeting the EML4-ALK rearrangement in lymphoma, was tested for activity against patients with this rearrangement in their lung cancer. This was found to occur in less than 5% of lung cancers. More than 1700 patients were screened at multiple institutions to populate a Phase II trial with 82 preselected patients. The drug was successful, with a response rate of 57%, an additional 35% achieving stable disease, and a median time to progression of greater than 6 months at the time of publication. ²⁹ These results led to successful follow-up trials and FDA approval, within only 5 years of discovering this genetic aberration in small numbers of lung cancer patients (the drug was already available).

Another clinical trial explored the efficacy of vemurafenib which was designed to block the activity of BRAF, a gene that was reported to be mutated in many patients with melanoma. The complete plus partial response rate with advanced melanoma patients whose cancer had the targeted BRAF mutation was over 70%, with a median duration of 7 months. ³⁰ FDA approval followed soon afterward.

These were remarkable results in patients with advanced cancers that were well known to be minimally responsive or resistant to chemotherapy. It is clear that genomic screening can improve response rates with targeted therapies, benefiting patients and shortening the time required for testing an experimental therapy. However, we will not know how generalizable this conclusion is until the results of all such targeted trials—positive and negative—are reported and shared. Negative trials of experimental drugs are not typically reported, although they inform decision making by the pharmaceutical company that is developing them. Meanwhile, the results just described support pursuing this approach to testing targeted experimental therapies in a screened subset of cancer patients.

For these reasons, clinical testing for genomic aberrations in most medical centers currently is limited to assays that measure mutations (insertions, deletions, or substitutions of nucleotides) in selected genes, alterations in gene copy number, and structural rearrangements in selected genes ³² (Figure 45-3 ). Available experimental or approved drugs that target the products of aberrant genes are currently limited to a few dozen genes. At the author’s institution, we currently are sequencing 740 hot spots in 46 genes. This detects only the first of the three categories of genomic analysis listed above. To capture all three categories and expand the gene coverage, we are planning to move to the targeted deep sequencing of the exons (open reading frames) of a few hundred genes. Both of these approaches to sequencing enable multiplexed assays on many genes to be run on a single extracted sample of tumor DNA. This is a huge advance over the technology just a few years ago, when selected regions of genes were sequenced in a separate assay for each individual mutation, and the amount of tumor tissue available rapidly became a limiting constraint. A number of cancer centers have started programs assaying DNA in patient’s cancers to guide therapy (e.g., Refs. 33, 34).

With 800 experimental targeted drugs in the clinical pipeline and a few dozen already approved, current second-generation sequencing of advanced solid tumors yields a match between a genetic aberration and a targeted drug about 40% of the time. The yield is higher in reported cases of deep sequencing with more complete genomic analysis. The ingenuity of companies developing new technologies for detecting genetic aberrations guarantees that advancements will continue to enable more rapid and complete analysis for less cost. The limiting factors have become the need for bioinformatics and computational tools and for trained personnel to use these tools to interpret the sequencing data.

Two key rate-limiting gaps in scientific knowledge were identified recently in a report from the President’s Council of Advisors on Science and Technology. ³⁵ These were (1) identifying and validating the protein targets that are critical to the survival of cancer cells and are “druggable,” and (2) predicting more effectively the efficacy and toxicity of candidate drugs before high investments of funds and time are made in clinical trials.

The new fields of systems biology and computational biology have emerged to address the first of these questions. Through extensive computer-based analysis of data from biological studies on the biochemical steps that activate or suppress molecular processes in cells, investigators have been able to draw complex maps of interconnected pathways that regulate the activities of proteins in those cells. Candidate “drivers” that form pivotal controlling nodes in these interconnected pathways are identified in a number of ways: (1) They may connect multiple signaling pathways; (2) they may be recurrently mutated in human cancer specimens; (3) introduction of specific RNAi’s can block a signaling pathway that changes major phenotypic functions in the cell (such as maintaining viability); and (4) knockin or knockout of the gene’s activity in genetically engineered mouse models (GEMMs) results in molecular alterations, changing the phenotype in significant ways that affect malignant behavior. ^36,37 It is important to note that up to now, most targeted therapies discussed in this chapter were developed with earlier approaches that involved identification of abnormal functioning of a target in biological experiments with cancer cells, not because of detection of aberrant genes in broad screens of human cancers.

The Human Cancer Genome Atlas and additional data collected from genomic testing of cancers in multiple centers will identify more genetic aberrations that are cancer drivers and potential targets for new drugs. These new approaches will also enable identification of driver genes that are present at low frequencies—for example, less than 5%—in few types or only one type of cancer. It must be cautioned that a mutation in a known oncogene does not necessarily mean that it is a driver in a particular cancer, and a gene could have “driver” status in a cancer without a genetic aberration, working instead through amplification or an epigenetic mechanism, or through changes in expression of message or posttranslational modification of the protein it encodes. It also must be emphasized that a driver molecule may be activated by a variety of mechanisms related to its situation in one or more activated regulatory pathways ³⁸ (Figure 45-4 ).

The ultimate driver gene is one which is so critical to a cancer cell’s survival that when its function is blocked, the cell dies, typically by apoptosis. The observation of this phenomenon led to the “addiction hypothesis” of Weinstein, who postulated that in cancer cells that are driven by a mutation in a critical regulatory pathway, alternative pathways that could promote or bypass the driving gene’s activity may be downregulated to the point where, unlike the situation in normal cells, they may be irretrievably damaged when the driving, addicting gene is suddenly blocked by a targeted therapy. ³⁹

The identification of a driver gene does not invariably lead to creation of a drug effective against the gene’s product. The tyrosine kinase oncogenes have turned out to be “druggable” targets, and a few dozen new agents against these targets are in clinical use or in clinical trials—both antibodies against receptors on the cell surface and their ligands, and low molecular-weight drugs that act intracellularly. Suppressor genes such as p53, genes controlling transcription factors such as MYC, and genes whose products have more challenging molecular characteristics such as RAS have turned out to be less “druggable.” However, by knowing the upstream molecules that interact with these gene products, or the downstream molecules that they activate, targets may be identified that counteract the activity of these genes and are more “druggable.” ⁴⁰

It was initially hoped that a drug which successfully targeted the product of an aberrant gene in one type of cancer would be effective in other types of cancer with the same genetic abnormality. However, this has not always turned out to be the case, because different types of cells can express pathways that may bypass the drug’s target. An example is found with vemurafenib, which, as noted, is effective against the majority of melanomas expressing a mutated BRAF gene at V600E, but is not effective against colon cancers with the same genetic aberration. ⁴¹ This has been investigated by screening a panel of colon carcinoma cell lines bearing V600E RAF mutations with a short-hairpin RNA (shRNA) library representing the full complement of 518 human kinases, in the presence or absence of the BRAF inhibitor PLX4032. It was found that shRNA vectors targeting and eliminating EGF receptors converted resistant cells into becoming sensitive to inhibition by the BRAF inhibitor. ⁴² This effect was duplicated by treating cells with a drug or antibody against the EGF receptor. Thus inhibition of the EGF receptor tyrosine kinase appears to be required to permit sensitivity to this BRAF inhibitor in colorectal cancer cells. Most melanoma cells do not have the active EGF receptors that are seen in typical epithelial cells, and therefore are sensitive to the drug inhibiting BRAF. Clinical trials are testing this important experimental observation in patients with colorectal cancer bearing the V600E BRAF mutation, by adding an EGF receptor inhibitor to a BRAF inhibitor.

The above experiment is one example of many biological studies with cell lines and tumor biopsies that provide explanations for resistance to targeted drugs, involving activation of bypass pathways. For example, MET amplification can lead to resistance against the EGF receptor inhibitor gefitinib by activating ERB-B3. ⁴³ Likewise, the efficacy of MEK inhibitors in basal-subtype breast cancer cells is limited by a feedback loop involving activation of the PI3K pathway by activation of the EGF receptor tyrosine kinase. ⁴⁴ This suggests that a combination of inhibitors of MEK plus inhibitors of EGF receptors or the PI3K pathway may be effective in this situation.

In summary, combination treatments with agents against two different targets, in order to attack a driver gene and a bypass pathway concurrently, may provide a way to overcome resistance. Many clinical trials are testing this hypothesis.

With the technologies available today, an attractive way to investigate the mechanisms of failure (resistance) after initial response to an experimental targeted agent in a clinical trial is to biopsy the recurrent tumor and use genomic and expression array analysis—in comparison with the original tumor—to determine what new event or events have conferred resistance.

The bypass mechanisms that explain resistance to a targeted drug in a primary tumor may be discovered by direct comparison of genomic aberrations in primary tumor specimens from patients who are sensitive, or resistant, to a drug targeting a genetic aberration that they share. For example, clinical responses occur only in a minority of the colon cancer patients treated with the EGF receptor inhibitors cetuximab or panitumumab. It was found that colon cancers with a mutation in K-RAS were not responsive to these EGF receptor inhibitors. ⁴⁵ Furthermore, when resistance to these anti-EGF receptor antibodies developed in colon cancer patients who were initially responsive, biopsy of the recurrence showed that a mutation in RAS (either from outgrowth of a minor subpopulation below the limits of detection in the primary tumor, or occurring de novo) accounted for resistance in the majority of patients. ^46,47 Unfortunately, an effective inhibitor of K-RAS is not available to administer with the EGF receptor inhibitor. Because nearly 50% of colon cancers have mutations in K-RAS, screening for these mutations provides benefits to the patient by avoiding a therapy that will not be beneficial, and to controlling the costs of medical care by avoiding the use of these expensive agents in this clinical situation. Testing for this biomarker is now standard of care for treatment of colon cancer patients.

The complexity of the challenge in identifying genes that can modulate sensitivity and resistance to targeted therapies is exemplified by the many mechanisms of resistance that have been discovered in patients with malignant melanoma who were initially responsive to therapy with vemurafenib against V600E B-RAF, and subsequently relapsed. Six different mechanisms identified in tumor biopsies from these patients are listed in Table 45-2 . ^48–54 In all cases, confirmatory studies were carried out with cell lines containing the V600E B-RAF mutation, to demonstrate that the proposed mechanism was able to confer resistance to the inhibitor. The common theme in this series of observations and in studies of resistance to other targeted therapies is that there are two predominant resistance mechanisms—new aberrations of the targeted gene and activation of bypass pathways.

These data naturally stimulate the formation of hypotheses on how to overcome resistance to drugs that target B-RAF. One approach that was tested recently was the addition of an inhibitor of MEK, administered concurrently with an inhibitor of mutated B-RAF. The rate of complete or partial response with the combination was 76%, as compared with 54% with monotherapy (P = 0.03). ⁵⁵ Of course the hope is that these responses will be more durable, resulting in substantial prolongation of life—but the results will not be available for a few years.

Although driver genes and genes to which cancer cells are addicted are obvious targets for drug development, there are other genes and pathways that do not cause cancer but are important for the survival and growth of cancer cells. This has generated the concept of “non-oncogene addiction.” ⁵⁶ Examples include gene products involved in DNA damage repair, protein chaperones and heat shock proteins, histone modification, and buffering of reactive oxygen species. Attacking these pathways that support the malignant phenotype, in addition to targeting critical oncogenes, may be a fruitful approach to combination therapy of cancer.

The series of experiments just discussed demonstrate the appropriateness of the term precision medicine. This term describes the situation in which a more precise understanding of the various genetic aberrations in a particular patient’s cancer, and their interactions with each other and with the surrounding microenvironment, can provide decision support to the treating oncologist—enabling selection of a targeted drug or drugs that are appropriate for that patient’s illness. In this chapter, the emphasis has been on genomics as the basis of precision medicine. This is because at the present time, genomic biomarkers are the most useful for drug development and for making clinical decisions. In the future, it is anticipated that biomarkers from at least four other sources will contribute more and more to precision medicine: