Abstract
The introduction of specific, molecular-targeted drugs is radically changing cancer treatment. Pharmacodynamics, which measures drug effects on the host, is key during early-phase clinical trials of novel agents to determine the relations between drug dose and target inhibition as well as measure the downstream effects of target inhibition on the cancer. In this article, we describe the application of flow cytometry to the pharmacodynamic monitoring of molecular-targeted agents in leukemia patients. The methods are based on current clinical flow-cytometry applications, with the addition of phosphospecific antibodies to measure the activation states of intracellular signaling elements and the introduction of techniques that maintain drug–target equilibrium during sample preparation. Using this approach, we successfully showed dose-dependent inhibition of c-Kit during a phase I clinical trial treating acute leukemia patients with the novel agent sorafenib. Further refinements identify considerable interpatient variation in signaling activity within leukemic blast populations, suggesting that an individualized approach to treatment based on flow cytometric monitoring might be advantageous. Improvements in sample turnaround offer the potential to introduce real-time pharmacodynamic monitoring during early-phase clinical trials.
Introduction
Medical oncology, which deals with the systemic treatment of cancer, is rapidly evolving from a chemotherapy-based specialty toward a newer, molecular-oriented discipline that is founded in basic science and driven by the rapid development of molecular-targeted drugs. Although molecular oncology is still in its infancy, it is clear that its evolution will require close integration between clinical and laboratory medicine and that each of these medical specialties needs firm grounding in basic science.
Cancer develops as a multistage process from precursor lesions caused by accumulated genetic alterations that result in gains or losses of activity in key signaling cascades. These alterations select for malignant potential by effects on cell growth, differentiation, and survival. Deregulation occurs through signal transduction networks that, although complex, are amenable to drug treatment: these include small molecules that inhibit (or sometimes activate) intracellular kinases, as well as engineered proteins that act on surface receptors or soluble growth factors (Sawyers, 2004).
In the past few years, several molecular-targeted agents have been licensed for the treatment of cancer patients. Examples include the humanized monoclonal antibodies cetuximab (Erbitux) and tristazumab (Herceptin), which inhibit epidermal growth factor receptor (EGFR) and Her2/neu, respectively, and erlotinib (Tarceva) and gefitinib (Iressa), which are small molecule inhibitors of the EGFR kinase domain. However, it should be noted that although clinical benefit using these agents firmly establishes the principles of molecular oncology, for the most part, response rates are less than those seen using standard chemotherapy and are often of short duration (Moore et al., 2007; Shepherd et al., 2005). The most dramatically effective molecular-targeted agent currently used in the clinic is the bcr/abl kinase inhibitor imatinib (Gleevec), which is highly effective at treating chronic myeloid leukemia patients (Druker et al., 2001). But even in the presence of a unique molecular defect driving cancer development, the drug does not produce permanent cures, and imatinib is much less effective against cases of chronic myeloid leukemia (CML) in blast crisis that have accumulated additional molecular defects.
At the present time, molecular oncology is making rapid progress on several fronts. Using increasingly sophisticated analytical tools for genetic analysis and protein interactions, basic scientists continue to discover new molecular mechanisms for cancer development. The establishment of these mechanisms’ clinical relevance is facilitated through the increasing availability of tumor banks and closer interactions with molecular pathologists to study the expression in different forms of cancer and the effects on patient outcome. The elucidation of an important molecular mechanism prompts the development of novel targeting agents by established pharmaceutical companies or the growing biotechnology industry. Molecular pathology techniques developed to establish the clinical relevance of the target can then be further refined to provide a biomarker to predict the likelihood of response in cancer patients.
Pharmacodynamics is a branch of clinical pharmacology that optimizes drug doses by studying beneficial and toxic (toxicodynamic) effects on the host. During early clinical trials, pharmacodynamics establishes dose schedules for efficacy testing, whereas later in patient management, pharmacodynamic monitoring is used to adjust the dose in individual patients (for example, insulin dose is adjusted based on blood glucose monitoring). The pharmacodynamic monitoring of molecular-targeted anticancer agents generally means measuring effects on the immediate drug target (for example, decrease in the phosphorylated substrate of a kinase target) or the broader effects of target inhibition on downsteam signaling pathways, transcriptional regulation, or cell growth and survival mechanisms. It is important to distinguish between these two levels of drug effect, because loss of immediate target inhibition might be the result of inadequate drug dose or selection for a mutant target protein such as occurs in imatinib-resistant CML. On the other hand, the lack of downstream effect would suggest that the drug target was redundant to cancer growth in that particular patient.
Pharmacodynamic monitoring has become mandatory during early-phase clinical trials of molecular-targeted anticancer agents. It requires the application of laboratory methods that are different from those used to predict sensitivity. A further consideration in pharmacodynamics is the need for multiple tumor samples during treatment. This is much more readily achieved in blood, so the detailed study of pharmacodynamic effects in leukemia provides a window to learn basic principles that might then be applied more generally to solid-tumor patients. Flow cytometry is a well-established technique for studying leukemia that has recently been further developed by our group and others for the pharmacodynamic monitoring of molecular targeted agents during early-phase clinical trials (Green et al., 2006; Tong et al., 2006). The purpose of this article is to review this exciting new area and to discuss the potential for further development including eventual introduction as a routine clinical test.
Methodology
The general principles of signal transduction analysis by flow cytometry, based on the use of phosphospecific antibodies, are the subject of several recent articles (Chow et al., 2001; Chow et al., 2006; Jacobberger et al., 2003; Perez and Nolan, 2002) and are summarized here in outline form. Usually the pattern of changes that occur in response to activators and inhibitors of signaling pathways is more informative than a single snapshot of a steady state (Chow et al., 2001; Perez and Nolan, 2002). Because signaling is a dynamic process that involves the rapid addition and removal of phosphate groups from the target protein, normally, the samples are fixed at a specific time following these additions. Fixation destroys kinase and phosphatase activities as well as preserves intracellular proteins and their phosphorylated state. Surface membrane permeabilization is then done by treating with alcohols or detergents before staining with phosphospecific antibodies. There have been significant improvements in the affinity and specificity of phosphospecific antibodies suitable for flow cytometry, and these are now available for a wide range of relevant markers.
Special Considerations for Pharmacodynamic Monitoring
Because the large majority of targeted agents currently under development inhibit rather than activate signaling pathways, their pharmacodynamic effects are shown by a decrease in the levels of phosphorylated epitopes during drug treatment. The underlying concepts are illustrated in Figure 1, which shows the acute activation of the extracellular-regulated kinase (ERK) pathway in CD3-positive lymphocytes, using phorbol ester (PMA), and the effects of pretreatment with the upstream inhibitor U0126.
Although we and others have identified constitutive activation of signaling pathways in blood samples obtained from acute and chronic leukemia patients (Chow et al., 2006; Pallis et al., 2003), the levels are often insufficient to provide a baseline for pharmacodynamic monitoring of a pathway inhibitor. Therefore, the clinical protocols usually involve processing blood samples with acute ex vivo activation using a relevant growth factor. In contrast, blood samples from chronic myeloid leukemia (CML) patients show readily detectable phosphorylation of signaling elements downstream of the abnormal bcr/abl tyrosine kinase such as STAT5 (signal transducer and activator of transcription). Preliminary results from studies of CML patients show that during drug treatment with the bcr/abl inhibitor imatinib, the levels of phosphorylation decrease, suggesting that this approach might be developed as a routine test to allow dose optimization or the early detection of drug resistance (Goolsby et al., 2005; Jacobberger et al., 2003). As shown in Figure 2, a significant reduction of the in vivo levels of phosphorylated STAT5 (p-STAT5) in peripheral blood CD34+ cells in a chronic-phase CML patient was seen postinitiation of imatinib therapy (top two panels). In vitro treatment of the post-therapy sample with imatinib (bottom panel) demonstrated a further decrease in p-STAT5 level, suggesting that a higher dose of imatinib might have been beneficial for this patient.
A detailed sample preparation protocol is described in a recent publication by our group (Chow et al., 2005). This addresses several major technical problems encountered in pharmacodynamic monitoring. Because most kinase inhibitors show reversible binding to their target, whole blood processing is needed to maintain the drug–target equilibrium. Excessive fixation renders red blood cells difficult to lyse, but brief treatment with 4% formaldehyde is compatible with subsequent red-cell lysis using Triton X-100, which also has the effect of permeabilizing the white blood cell membrane to allow intracellular antibody staining. This approach is suitable for many relevant phosphorylated epitopes and also gives excellent preservation of light scatter and surface staining with most of the standard surface phenotypic markers.
There is now a large number of high-quality monoclonal phosphospecific antibodies labeled with a wide range of fluorescence stains, allowing the development of complex staining protocols that examine signaling networks rather than individual signaling elements. As well as providing powerful basic research techniques, these sophisticated flow-cytometry protocols have exciting potential for clinical diagnosis and monitoring. However, it should be noted that there is a wide range in the optimum fixation conditions for intracellular antigen staining, which sets constraints on the selection of phosphoepitopes. This is particularly relevant for pharmacodynamic monitoring, which usually requires a large dynamic range in signal strength.
Results and Perspectives
Phase I Trial of Sorafenib
Sorafenib was originally developed as an inhibitor of raf kinase, which is strongly activated by phorbol ester (PMA). During the early development of the sorafenib pharmacodynamic monitoring protocol, we tested effects using isolated peripheral blood lymphocytes exposed to a range of sorafenib concentrations (Chow et al., 2001). Whereas measurement of pERK by conventional western blotting gave a dose/response to sorafenib similar to that previously reported, using flow cytometry, we found that individual cells showed either no inhibition or complete inhibition of ERK activation and that the proportion of cells showing complete inhibition increased with the dose of sorafenib (Figure 3). A similar biphasic effect was also seen using pMEK as the readout for raf kinase inhibition. Although not explained by standard Michaelis-Menten kinetics, this “all-or-none” response to raf signaling is very reminiscent of previous work in progesterone-activated Xenopus oocytes (Ferrell and Machleder, 1998) and illustrates the potential for flow cytometry as a discovery tool in the basic science of signal transduction.
To provide a more physiological stimulus in the leukemia trial, we also included samples that were activated by Stem Cell Factor (SCF), which activates the normal hematopoietic stem cell receptor c-Kit. The readout for this assay was the extent to which ex vivo ERK phosphorylation in the leukemic blasts was inhibited during treatment with increasing doses of sorafenib (Tong et al., 2006). Figure 4a shows typical results from a patient that were obtained immediately predose and after 3 days of treatment at the highest dose of sorafenib tested (400 mg twice daily). It can be seen that whereas SCF and PMA have similar effects on ERK activation in the blast population predose, after 3 days of sorafenib treatment, there is much greater inhibition of SCF. This differential effect is striking when the entire data set for the dose/response and time course obtained from the clinical trial are summarized (Figure 4b). Here, there is a clear dose-dependent inhibition of SCF but no pharmacodynamic effect using PMA activation. Furthermore, a preliminary analysis of the pharmacokinetic data from this trial shows a strong correlation between the extent to which SCF is inhibited and the concentration of sorafenib in the blood samples. This paradoxical lack of effect on PMA activation is explained by the subsequent finding that sorafenib is a potent inhibitor of multiple receptor tyrosine kinases including c-Kit, in addition to raf kinase, and that there is insufficient free drug available in the peripheral blood to inhibit raf kinase, as a result of extensive binding to plasma proteins (Wilhelm et al., 2004).
Monitoring Phosphorylation of the S6 Ribosomal Protein via mTOR and ERK Pathways
Signaling pathways show complex interactions that can be considered as networks. Hence, flow cytometry protocols that measure the phosphorylation states of multiple signaling elements are more informative than those based on a single marker such as pERK. The schematic shown in Figure 5 shows how the ERK and PI3-kinase/Akt pathways, which are commonly activated in cancer because of mutations involving upstream elements such as ras proteins and growth factor receptors, converge at multiple levels to modulate the mammalian target of rapamycin (mTOR) pathway; a major regulator of protein translation. Phosphorylation of the S6 ribosomal protein is a highly dynamic process that is responsive to inputs from Akt/mTOR and ERK. Whereas normal blood cells show low or negative S6 phosphorylation in the absence of growth factor activation, as illustrated in Figure 6, we found that about half of the cases of acute myeloid leukemia examined showed increased levels that could be inhibited by ex vivo treatment with inhibitors of PI3-kinase, mTOR, or ERK pathways (Chow et al., 2006). Interestingly, we found that the relative sensitivity to these agents varied considerably between individual patients, suggesting that this application might be able to classify leukemia patients based on patterns of abnormal signaling, as previously proposed (Irish et al., 2004). This also suggests the possibility to perform ex vivo sensitivity testing for specific molecular targeted agents as a guide to treatment selection in leukemia patients.
Real-Time Pharmacodynamics
As we move toward standardized laboratory protocols based on the use of a rapid-fixation/red-cell–lysis technique and directly conjugated phosphospecific antibodies, the turnaround time needed to generate the results for a blood sample can be as little as 2 hours. This opens up the exciting possibility to perform routine pharmacodynamic monitoring in real time. Initially, we envisage real-time pharmacodynamics simply as rapid feedback to the clinical trials organization. Combined with ex vivo sensitivity testing of individual patients’ leukemia cells, as described above, and pharmacokinetic data, this would inform if the current dose level was close to that needed for optimum target inhibition. However, the successful implementation of real-time pharmacodynamics might open the possibility for more innovative design of phase I clinical trials, in which the actual dose of drug is escalated in individual patients based on pharmacodynamic effects rather than escalated between cohorts of patients, as is currently done.
It also seems technically feasible to use real-time pharmacodynamics based on flow cytometry in routine clinical practice involving molecular targeted agents in leukemia patients. This would be particularly helpful in situations in which there is significant interpatient variation in the dose of drug needed to achieve optimum effect—for example, because of differences in drug absorption and metabolism. Advances in instrumentation, reagent quality, and automation should allow the development of standardized and cost-efficient protocols suitable for routine clinical laboratory use during the next few years. As well as having immediate benefit for patient care, this would be consistent with the current trend toward closer integration between basic science and laboratory medicine as the field of molecular oncology becomes more clearly defined.