Jerome Mettetal

Translational Exposure–Efficacy Modeling to Optimize the Dose and Schedule of Taxanes Combined with the Investigational Aurora A Kinase Inhibitor MLN8237 (Alisertib)

Aurora A kinase orchestrates multiple key activities, allowing cells to transit successfully into and through mitosis. MLN8237 (alisertib) is a selective Aurora A inhibitor that is being evaluated as an anticancer agent in multiple solid tumors and heme-lymphatic malignancies. The antitumor activity of MLN8237 when combined with docetaxel or paclitaxel was evaluated in in vivo models of triple-negative breast cancer grown in immunocompromised mice. Additive and synergistic antitumor activity occurred at multiple doses of MLN8237 and taxanes. Moreover, significant tumor growth delay relative to the single agents was achieved after discontinuing treatment; notably, durable complete responses were observed in some mice. The tumor growth inhibition data generated with multiple dose levels of MLN8237 and paclitaxel were used to generate an exposure–efficacy model. Exposures of MLN8237 and paclitaxel achieved in patients were mapped onto the model after correcting for mouse-to-human variation in plasma protein binding and maximum tolerated exposures. This allowed rank ordering of various combination doses of MLN8237 and paclitaxel to predict which pair would lead to the greatest antitumor activity in clinical studies. The model predicted that 60 and 80 mg/m2 of paclitaxel (every week) in patients lead to similar levels of efficacy, consistent with clinical observations in some cancer indications. The model also supported using the highest dose of MLN8237 that can be achieved, regardless of whether it is combined with 60 or 80 mg/m2 of paciltaxel. The modeling approaches applied in these studies can be used to guide dose-schedule optimization for combination therapies using other therapeutic agents. Mol Cancer Ther; 13(9); 2170–83. ©2014 AACR.

Growth Rate Analysis and Efficient Experimental Design for Tumor Xenograft Studies

Human tumor xenograft studies are the primary means to evaluate the biological activity of anticancer agents in late-stage preclinical drug discovery. The variability in the growth rate of human tumors established in mice and the small sample sizes make rigorous statistical analysis critical. The most commonly used summary of antitumor activity for these studies is the T/C ratio. However, alternative methods based on growth rate modeling can be used. Here, we describe a summary metric called the rate-based T/C, derived by fitting each animals tumor growth to a simple exponential model. The rate-based T/C uses all of the data, in contrast with the traditional T/C, which only uses a single measurement. We compare the rate-based T/C with the traditional T/C and assess their performance through a bootstrap analysis of 219 tumor xenograft studies. We find that the rate-based T/C requires fewer animals to achieve the same power as the traditional T/C. We also compare 14-day studies with 21-day studies and find that 14-day studies are more cost efficient. Finally, we perform a power analysis to determine an appropriate sample size.

Dose Schedule Optimization and the Pharmacokinetic Driver of Neutropenia

Toxicity often limits the utility of oncology drugs, and optimization of dose schedule represents one option for mitigation of this toxicity. Here we explore the schedule-dependency of neutropenia, a common dose-limiting toxicity. To this end, we analyze previously published mathematical models of neutropenia to identify a pharmacokinetic (PK) predictor of the neutrophil nadir, and confirm this PK predictor in an in vivo experimental system. Specifically, we find total AUC and Cmax are poor predictors of the neutrophil nadir, while a PK measure based on the moving average of the drug concentration correlates highly with neutropenia. Further, we confirm this PK parameter for its ability to predict neutropenia in vivo following treatment with different doses and schedules. This work represents an attempt at mechanistically deriving a fundamental understanding of the underlying pharmacokinetic drivers of neutropenia, and provides insights that can be leveraged in a translational setting during schedule selection.

Abstract 4649: Using pharmacokinetic/efficacy modeling to identify the optimal schedule for MLN0264, an anti- guanylyl cyclase C (GCC) antibody-drug conjugate, in a range of xenograft models

MLN0264 is an investigational antibody-drug conjugate (ADC) that consists of the human anti-guanylyl cyclase C (GCC) antibody linked to a microtubule-disrupting agent (monomethyl auristatin). As ADCs have a very long clearance half-life, the potential exists for a highly infrequent dosing schedule. A quantitative understanding of the relationship between exposure and preclinical antitumor biological activity is thus applied to support dose schedule selection in the clinic. In this study, we develop a pharmacokinetic/efficacy (PK/E) relationship in xenograft models to evaluate the predictive contributions of exposure and xenograft characteristics to MLN0264 biological activity. Single dose pharmacokinetic (PK) data were obtained for a range of time points, and a linear two-compartment PK model was built. Xenograft biological activity studies were conducted in which MLN0264 was administered at various dose levels and dosing schedules to mice bearing one of six different xenograft models. We used multiple linear regression and tumor dynamic modeling to understand the factors contributing to the biological activity. First, the growth rate of each tumor under control and treatment conditions was established by fitting the biological activity data within the dosing period to an exponential growth function. Then, we assessed the relationship of AUC to antitumor biological activity. Within each xenograft model, AUC was strongly correlated with biological activity across a range of schedules (R2>0.9, p Although biological activity within a xenograft model was strongly correlated with exposure, when the data from different xenograft models were pooled together, the correlation was weaker (R2= 0.30, p= 0.0004). This result suggests that, although schedule is not a major determinant of biological activity, other xenograft-specific factors may be contributing to biological activity. To test the contribution of some possible factors, we used multiple linear regression to determine which covariates (such as GCC expression level and baseline growth rate of xenograft models) are predictive of biological activity. The results suggest baseline growth rate does not correlate with biological activity while GCC expression is weakly correlated. This finding formed the basis for the development of a mechanistic model of GCC expression and biological activity. Taken together, this work demonstrates the use of a PK/E framework to identify the scheduling effect for a first-in-man protocol for an ADC. We have also demonstrated that this PK/E framework can be further leveraged to assess the contribution of other potential predictors of ADC biological activity in xenograft models. Citation Format: Shu-Wen Teng, Christopher Zopf, Johnny Yang, Brad Stringer, Julie Zhang, Wen Chyi Shyu, Arijit Chakravarty, Petter Veiby, Jerome Mettetal. Using pharmacokinetic/efficacy modeling to identify the optimal schedule for MLN0264, an anti- guanylyl cyclase C (GCC) antibody-drug conjugate, in a range of xenograft models. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 4649. doi:10.1158/1538-7445.AM2014-4649

Abstract 365: Using radionuclear imaging and mechanistic modeling to assess the therapeutic potential of antibody-drug conjugates (ADCs)

Proceedings: AACR Annual Meeting 2014; April 5-9, 2014; San Diego, CAnnAntibody-drug conjugates (ADCs) comprise a drug class that allows for direct delivery of a cytotoxic agent to the target tumor cell. The successful development of an ADC involves the assessment of patient characteristics such as tumor antigen expression and tumor vascularity. There is a need to develop a mechanistic understanding of the effect these tumor parameters have on ADC biological activity. This mechanistic insight can be derived from a mathematical model integrating quantitative preclinical data, and will support ADC development.nnMLN0264 is an investigational ADC consisting of a human anti-guanylyl cyclase C (GCC) antibody linked to the microtubule-disrupting agent monomethyl auristatin (MMAE). Here, we use mathematical modeling, in conjunction with in vivo imaging, to decouple the contribution of different tumor parameters to overall ADC biological activity. We constructed a mathematical model of ADC biological activity by integrating experimental results from (1) in vivo antibody imaging studies, (2) in vitro viability assays, and (3) in vivo xenograft biological activity studies.nnFirst, blood pharmacokinetics and tumor disposition were quantitatively constrained using in vivo radiolabeled antibody single-photon emission computed tomography (SPECT) data for both blood and tumor tissues. SPECT data from three xenografts with various antigen expression levels were used to link antigen expression level to ADC uptake. Second, the relationship between bound GCC receptor concentration and cell viability was established using viability assays run on an engineered cell line (293-GCC) with high antigen expression and high sensitivity to MMAE. Finally, using this relationship, we built a tumor growth dynamics model to describe in vivo xenograft biological activity, and to estimate the growth inhibition coefficient of 293-GCC.nnThis mechanistic model can be used to gain insights into the factors driving response of a tumor that is intrinsically sensitive to MMAE. Our results indicate the process of ADC vascular permeability is one of the limiting factors of ADC disposition. This outcome is reasonable given that a large ADC molecular weight decreases the permeability. Furthermore, the model simulations suggest some tumors that are intrinsically sensitive to MMAE may not be affected by the ADC if GCC antigen expression levels are very low.nnTaken together, the mechanistic model developed here forms the basis of a quantitative understanding for several factors influencing MLN0264 patient selection.nnCitation Format: Shu-Wen Teng, Ozlem Yardibi, Julie Zhang, Donna Cvet, Johnny Yang, Kelly Orcutt, Melissa Gallery, Arijit Chakravarty, Wen Chyi Shyu, Jerome Mettetal, Daniel Bradley, Petter Veiby. Using radionuclear imaging and mechanistic modeling to assess the therapeutic potential of antibody-drug conjugates (ADCs). [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 365. doi:10.1158/1538-7445.AM2014-365

Abstract 372: Choosing the right schedule for progression free survival: A systems pharmacology approach

Drug resistance is a common feature of many cancer therapies and often limits their long-term effectiveness. Most anticancer agents are developed with an initial focus on Objective Response (short-term response), despite clinical evidence that, for most cancers, this does not correlate with long term therapeutic benefit (Progression-Free Survival), due to the emergence of resistance. Viewing cancer as a disease of somatic Darwinian evolution reframes resistance as a consequence of competing populations of tumor cells within a polyclonal tumor, each responding differentially to selection pressure. In this context, the amplification of pre-existing resistance to therapy has been implicated as a common cause of treatment failure. Here we ask how drug scheduling can impact long-term response to therapy, using an evolutionary model of tumor kinetics. First, we apply a dynamic model of tumor kinetics with competing subclones of tumor cells to a published dataset of prostate cancer progression (Treatment thalidomide+docetaxel or ketoconazole+ hydrocortisone+alendronate), and show that it is capable of closely predicting Progression Free Survival. Tumor cells are modeled as being either sensitive or resistant, cell growth is modeled using a simple exponential growth model, and drug effect is incorporated in the model as overall inhibition on the growth-rate. Competition between tumor sub-populations is modeled as frequency-dependent fitness, using a logistic tumor growth model. Next, we use this dynamic model derived from the prostate cancer dataset to understand the effect of schedule on resistance, simulating tumor growth upon drug treatment for different doses and schedules. Beyond a literal simulation of resistance for a given drug schedule, we assess the contributions of resistance made by dose amplitude and dose frequency. Our model predicts that for a graded dose-response curve (Hill slope of 1), more frequent dosing schedules minimize the emergence of resistance. On the other hand, for a steep dose-response curve (Hill slope of 5), for constant dose density, we find that a high infrequent schedule is better than a low frequent dosing schedule for low doses. As the dose density increases, the low frequency schedule again emerges as optimal for minimizing resistance. As a final test, a competition term between the subpopulations is also included in our model, with essentially identical results. We then apply our model to published datasets of clinical response to treatment, and find close agreement with published results. Our generalized approach to this problem provides a framework for dissecting the underlying drivers of the emergence of resistance, and provides a recommendation for the minimization of resistance in a practical setting. Citation Format: Mayankbhai Patel, Jerome Mettetal, Matthew Cohen, Santhosh Palani, Keisuke Kuida, Mark Hixon, Joseph Bolen, Wen Chyi Shyu, Dennis Noe, Arijit Chakravarty. Choosing the right schedule for progression free survival: A systems pharmacology approach. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 372. doi:10.1158/1538-7445.AM2014-372

Abstract 4948: PET/CT clinical protocol design for the novel, first in class 68Ga labeled guanylyl cyclase C targeted peptide MLN6907 ([68Ga]MLN6907)

Acquisition and interpretation guidelines for clinical PET/CT imaging in oncology have been designed for whole-body 18F-FDG imaging and may not be optimized for assessment of other PET imaging tracers. Here we describe a methodology of PET/CT study design for the novel first in class 68Ga-labeled Guanylyl cyclase C (GCC) targeted peptide, [68Ga]MLN6907, based on a combination of in vitro, ex vivo, and in vivo preclinical imaging studies and model-based estimation of tumor parameters from simulated clinical PET data. GCC, a protein expressed in GI malignancies, is being targeted by the antibody drug conjugate MLN0264. GCC is also expressed on the healthy apical surface of the intestinal epithelium, which should be inaccessible to intravascular treatment. [68Ga]MLN6907 binds GCC with high affinity and is being developed as an imaging biomarker in an effort to help identify patients likely to respond to GCC-targeted therapy. In a series of experiments, the peptide affinity, internalization rate, and clearance were determined in patient-derived CRC xenografts with varied tumor microenvironmental phenotype. In addition to supporting the clinical development of the imaging agent, this data was used in combination with simulated clinical list-mode PET data to evaluate tumor parameter estimability under several clinically viable acquisition and reconstruction conditions. Specifically, liver CRC metastases of varying tumor diameter, antigen density, and vascularity were simulated in combination with PET imaging acquisition duration and reconstruction with and without partial-volume correction. Tumor, liver, and background time-activity curves (TACs) were generated from the reconstructed data and analyzed using a distributed tumor model to estimate the known tumor antigen density and vascularity. Analysis of the simulation studies revealed: 1) Partial volume correction is required for accurate antigen density and vascularity estimation; 2) Parameter estimation was most accurate within a tumor size range of 1-5 cm; 3) Parameter estimation was robust for all tested TAC reconstruction durations (e.g., 2, 3, 5, and 10 min); 4) Parameter estimation was optimal for common clinical acquisition times of 30-90 minutes; 5) Antigen density estimation was less accurate in poorly vascularized tumors. For the translation of a novel clinical biomarker, well controlled preclinical studies are critical; and, in this case, the findings combined with the distributed tumor model simulations directly guided the clinical image protocol. This rational and data-driven approach has the ability to not only improve the estimation of tumor properties in human subjects but also to guide the design of first-in-human oncology clinical imaging protocols with novel biomarkers. Citation Format: Jacob Y. Hesterman, Kelly D. Orcutt, Ozlem Yardibi, Jerome T. Mettetal, Shu-Wen Teng, Donna Cvet, Jack Hoppin, Thea Kalebic, Daniel P. Bradley. PET/CT clinical protocol design for the novel, first in class 68Ga labeled guanylyl cyclase C targeted peptide MLN6907 ([68Ga]MLN6907). [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 4948. doi:10.1158/1538-7445.AM2014-4948

Abstract A137: Combinatorial therapy design in cancer as a directed attack on the hubs of a scale-free network.

Scale-free networks, characterized by a power-law degree distribution, are known to be resilient to random failures, yet vulnerable to directed attacks on their hubs (their most connected nodes). Here, we show that protein-protein interaction networks in cancer are scale-free, validate the effectiveness of hub attacks using published RNAi lethality screens conducted on human cancer cells, and propose a novel approach for combinatorial hub attacks.nnMethods: Proteins involved in 14 different cancer types were extracted from the KEGG database. For each cancer type, a network was constructed by incorporating the experimentally-validated physical interactions of the pertinent KEGG proteins from the BioGRID interaction network. To validate the presence of scale-free cancer networks, we fit the degree distribution of the full network to a power-law. Then, we quantified the degree distribution of the lethality proteins discovered from RNAi screens (conducted on human multiple myeloma and epithelial ovarian cancer cells) and compared it to the rest of the network. A previously-derived mathematical relationship between the destruction of the giant component (a metric of network integrity) and the fraction of hubs removed was applied to the cancer networks to investigate the utility of combination therapy.nnResults: In agreement with previous studies of many biological networks, we found that the 14 cancer types listed in KEGG all demonstrated scale-free behavior. We analyzed previously published, genome-wide RNAi lethality screens in two cancer cells, and found a significant enrichment (≈ 5-fold) of median degree-connectivity among lethal proteins compared to all the proteins in the network. As expected, for a scale-free network, the removal of < 5% of the proteins completely destroyed network integrity. In the case of basal-cell carcinoma, as few as 9 proteins were responsible for the integrity of the entire network. Also, the marginal utility of each successive hub knockout for network destruction remained relatively constant, consistent with a combinatorial approach in tackling cancer.nnConclusion: In this work, we demonstrate that cancer networks are scale-free. The correlation between connectivity and lethality in the investigated cancer networks suggests that the number of neighbor-interactions of a protein may determine its potency as a drug target. We demonstrate a greedy algorithm for combinatorial therapy design based on the principle of simultaneous targeting of high-connectivity nodes. As the network of proteins responsible for toxicity in response to anticancer agents is also likely to be scale-free (and may be uncorrelated with cancer protein networks) the targeted destruction of cancer network hubs may in effect be a random attack on the toxicity nodes. This suggests that a combinatorial approach of targeting multiple proteins with high connectivity in cancer networks may be effective in creating an optimal therapeutic window.nnCitation Information: Mol Cancer Ther 2013;12(11 Suppl):A137.nnCitation Format: Andrew Chen, Arijit Chakravarty, Jerome Mettetal, Wen Chyi Shyu, Joseph Bolen, Santhosh Palani. Combinatorial therapy design in cancer as a directed attack on the hubs of a scale-free network. [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference: Molecular Targets and Cancer Therapeutics; 2013 Oct 19-23; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2013;12(11 Suppl):Abstract nr A137.

Abstract B214: Application of an evolutionary model of cancer cell response to dose-response viability curves to assess the potential for pre-existing resistance.

In recent years, tumor progression in the clinic has been demonstrated to be polyclonal. A single tumor contains diverse subpopulations of cancer cells which under the selective pressure of drug treatment evolve to produce drug resistance. As well, clonal diversity and genomic instability exist in tissue culture cancer cell lines. In this work, we determine the levels of pre-existing resistant subpopulations in cancer cell lines through the application of an evolutionary model of cancer cell response to an in vitro time-course of dose-response viability curves. Methods: We simulated dose response curves versus exposure time for a range of heterogeneous (varying fractions of sensitive and resistance cells) populations of cancer cells with predefined PD parameters. We added Gaussian noise to our estimates and fitted the time-course dose response data with both a single population model and with a dual population model and compared the goodness of the fit, PD parameter and residual variability levels for each of the heterogeneous populations. Model fitting and simulations were performed using NONMEM. Cancer cells were assumed to grow exponentially and the drug effect was modeled as a sigmoidal Emax function. An analytical solution was derived to estimate the population fractions from the plateau levels of the dose response curves. We validated the model predictions against a previously published dataset (Sci. Transl. Med. 3, 90ra59, 2011) where the baseline fraction of sensitive and resistant cells was varied in a controlled manner. Finally, the model was used to predict the percentage of resistant cells in a pre-existing population in response to treatment with signal transduction inhibitors. These predictions were then validated directly using time-lapse microscopy on cells grown in plastic and on soft agar. Results: From the simulations, we determined the study design necessary for modeling dose response relationship for heterogeneous populations to be a time course (0, 24, 48, 72h) viability assay with a range of concentrations spanning low to saturable effect. The dual population model accurately fitted the dose response curves simulated from heterogeneous populations, whereas the single population model showed poor predictability and exhibited large unexplained residual variability. The dual population model also predicted well the PD parameters and growth kinetics of the sensitive and resistance populations when applied to the published dataset with single time course (72h) viability measurement performed on heterogeneous populations with varying levels of resistance sub-fractions. When applied to a prospective in vitro study, the dual population model adequately predicted the fraction of resistance populations preexisting in the cell line tested. Conclusion: Taken together, the modeling approach described here provides a novel evolutionary approach to the assessment of pre-existing resistance to drug treatment that can be rapidly applied for mechanistic investigations of single-agent and combination cancer therapeutics. Citation Information: Mol Cancer Ther 2013;12(11 Suppl):B214. Citation Format: Snehal Samant, Derek Blair, Andrew Chen, Jerome Mettetal, Wen Chyi Shyu, Mark Hixon, Jeffrey Ecsedy, Santhosh Palani, Arijit Chakravarty. Application of an evolutionary model of cancer cell response to dose-response viability curves to assess the potential for pre-existing resistance. [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference: Molecular Targets and Cancer Therapeutics; 2013 Oct 19-23; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2013;12(11 Suppl):Abstract nr B214.

Abstract B213: Toward better preclinical combination studies: Using constrained optimization techniques in combination dosage and study design.

Dose-limiting toxicities limit the potential of many oncology agents, including targeted therapies. Combination therapies provide a path forward for widening the dose-response window, particularly if efficacy combines in an additive or synergistic manner, while toxicity is subadditive or non-overlapping. Here, we model the efficacy and toxicity landscapes of two-drug combinations, and provide analytical and numerical methods to establish optimal combination study design.nnMethods: The combined efficacy and toxicity of a pair of drugs were modeled from first principles using isobolograms as a sum of the single-agent PK/Efficacy (PK/E) relationships and a multiplicative combination term. Borrowing from Microeconomic Utility Theory, we found the point of greatest efficacy along the Maximum Tolerated Dose (MTD) toxicity contour, which corresponds to the efficacy isobole tangentially intersecting the MTD contour. These analytical results were supported by graphical inspection as well as numerical estimation. We used a combination of simulated efficacy studies and analytical methods to identify the most efficient combination study designs.nnResults: A combination of simple PK/E models was sufficient to describe a wide range of observed efficacy and toxicity isobolograms. The model uniquely determined the optimum combination dosage that provides maximal efficacy, and we found the single-agent conditions that can favor drug combination, namely: similar maximal efficacy (Emax) values, saturating relationships, weak sigmoidicity, and strong combination interaction effect. We analytically demonstrated that, for most scenarios, the optimal study design involves a diagonal escalation, resulting from the exploration of a fixed-dose combination with several ascending dose levels. We demonstrated via simulation that this diagonal escalation design depends only on the single-agent PK/E relationships, and not on the degree of interaction between the toxicities of the two agents. We ran an in vivo validation experiment using this diagonal escalation design, demonstrating its practical benefit in an experimental setting.nnConclusion: We demonstrate a modeling approach to combination therapy that solves for optimal dosing and study design. We found that a diagonal, constant-ratio escalation scheme was generally the most optimal for gathering combination information, and that the design depended only on the single-agent dose-response profiles. The combination information is critical for generating the efficacy and toxicity isoboles, which in turn allow us to predict the optimal combination dosage. The methods presented here can allow for the rapid and efficient translational assessment of the added benefit of a given combination in the clinical context.nnCitation Information: Mol Cancer Ther 2013;12(11 Suppl):B213.nnCitation Format: Andrew Chen, Sabrina Collins, Jerome Mettetal, Mark Manfredi, Katherine Galvin, Wen Chyi Shyu, Jeffrey Ecsedy, Santhosh Palani, Arijit Chakravarty. Toward better preclinical combination studies: Using constrained optimization techniques in combination dosage and study design. [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference: Molecular Targets and Cancer Therapeutics; 2013 Oct 19-23; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2013;12(11 Suppl):Abstract nr B213.