Rachael L. Brake

Abstract C252: A phase 1, dose-escalation study of MLN0128, an investigational oral mammalian target of rapamycin complex 1/2 (mTORC1/2) catalytic inhibitor, in patients (pts) with advanced non-hematologic malignancies.

Background: MLN0128 is an investigational, potent, and highly selective inhibitor of mTORC1/2, which are integral to cell proliferation, angiogenesis, and cellular metabolism. This first-in-human study ([NCT01058707][1]) aimed to determine the maximum tolerated dose (MTD), dose-limiting toxicity (DLT), safety, pharmacokinetics (PK), pharmacodynamics (PD), and preliminary antitumor activity of oral MLN0128. Methods: Pts aged ≥18 years with advanced solid tumors were enrolled in a 3+3 dose-escalation design initially to receive MLN0128 once daily (QD), and then enrolled in escalating intermittent schedules of once weekly (QW), 3 days/week (QDx3d QW), or 5 days/week (QDx5d QW) dosing, in 28-day cycles. Blood samples were collected at multiple timepoints for PK analysis. PD endpoints were evaluated in skin to determine the effect on mTORC1/2-dependent biomarkers. A preclinical PK-efficacy model was generated (Phoenix NLME v1.1) with tumor xenograft efficacy data, and implemented using clinical PK parameters to simulate tumor volume-time curves for various MLN0128 doses/schedules. Response was assessed by RECIST v1.1. Results: 115 pts received MLN0128 doses in the ranges 2-7 mg QD (n=30), 7-40 mg QW (n=30), 6-20 mg QDx3d QW (n=33), and 7-13 mg QDx5d QW (n=22). Median age was 60 years (range 24-89); 40% were male. The most common tumor types were colorectal (22%), renal (9%), and ovarian (8%) cancer. The MTDs were 6 mg QD, 40 mg QW, 16 mg QDx3d QW, and 10 mg QDx5d QW. Based on the overall safety profiles, the recommended phase 2 doses were 5 mg QD, 40 mg QW, 9 mg QDx3d QW, and 7 mg QDx5d QW. Pts received a median of 2 cycles (range 1-23). The most common drug-related adverse events (AEs) were hyperglycemia (65%), nausea (60%), vomiting (44%), decreased appetite (36%), diarrhea (33%), asthenia (30%), and mucosal inflammation (30%). The most common drug-related grade ≥3 AEs were hyperglycemia (12%), asthenia (9%), and mucosal inflammation (5%). MLN0128 exhibited dose-linear PK with a plasma half-life of ∼8 h, and did not accumulate in plasma with QD dosing. Modeling of preclinical PK-efficacy data and simulation of human tumor volume-time curves using clinical PK parameters suggested potential greater antitumor effect with schedules using more frequent dosing (i.e. QD, QDx5d QW). There was a treatment-related inhibition of mTORC1/2 biomarkers in skin. Two of 10 pts (20%) with renal cancer receiving MLN0128 at 15 and 40 mg QW had a partial response; 5 pts (6%) had stable disease for ≥6 cycles. Conclusions: Based on the safety profile, PK/PD, simulated tumor volume-time curves, and responses, MLN0128 5 mg QD and 40 mg QW were selected for further evaluation in an expansion phase in pts with renal, endometrial, or urothelial cancer. Citation Information: Mol Cancer Ther 2013;12(11 Suppl):C252. Citation Format: Jeffrey R. Infante, Josep Tabernero, Andres Cervantes, Shadia Jalal, Howard A. Burris, Teresa Macarulla, J. Alejandro Perez-Fidalgo, Rachel Neuwirth, Chirag Patel, Esha Gangolli, Rachael Brake, Jeffrey Sturm, Eric H. Westin, Michael Gordon. A phase 1, dose-escalation study of MLN0128, an investigational oral mammalian target of rapamycin complex 1/2 (mTORC1/2) catalytic inhibitor, in patients (pts) with advanced non-hematologic malignancies. [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 C252. [1]: /lookup/external-ref?link_type=CLINTRIALGOV&access_num=NCT01058707&atom=%2Fmolcanther%2F12%2F11_Supplement%2FC252.atom

Impact of Hydrolysis Mediated Clearance on the Pharmacokinetics of Novel Anaplastic Lymphoma Kinase Inhibitors

Compound 1 [(E)-4-fluoro-N-(6-((4-(2-hydroxypropan-2-yl)piperidin-1-yl)methyl)-1-((1S,4S)-4-(isopropylcarbamoyl)cyclohexyl)-1H-benzo[d]imidazol-2(3H)-ylidene)benzamide], a new, potent, selective anaplastic lymphoma kinase (ALK) inhibitor with potential application for the treatment of cancer, was selected as candidate to advance into efficacy studies in mice. However, the compound underwent mouse-specific enzymatic hydrolysis in plasma to a primary amine product (M1). Subsequent i.v. pharmacokinetics studies in mice showed that compound 1 had high clearance (CL) and a short half-life. Oral dose escalation studies in mice indicated that elimination of compound 1 was saturable, with higher doses achieving sufficient exposures above in vitro IC50. Chemistry efforts to minimize hydrolysis resulted in the discovery of several analogs that were stable in mouse plasma. Three were taken in vivo into mice and showed decreased CL corresponding to increased in vitro stability in plasma. However, the more stable compounds also showed reduced potency against ALK. Kinetic studies in NADPH-fortified and unfortified microsomes and plasma produced submicromolar Km values and could help explain the saturation of elimination observed in vivo. Predictions of CL based on kinetics from hydrolysis and NADPH-dependent pathways produced predicted hepatic CL values of 3.8, 3.0, 1.6, and 1.2 l/h⋅kg for compound 1, compound 2 [(E)-3,5-difluoro-N-(6-((4-(2-hydroxypropan-2-yl)piperidin-1-yl)methyl)-1-((1s,4s)-4-(isopropylcarbamoyl)cyclohexyl)-1H-benzo[d]imidazol-2(3H)-ylidene)benzamide], compound 3 [(E)-3-chloro-5-fluoro-N-(6-((4-(2-hydroxypropan-2-yl)piperidin-1-yl)methyl)-1-((1s,4s)-4-(isopropylcarbamoyl)cyclohexyl)-1H-benzo[d]imidazol-2(3H)-ylidene)benzamide], and compound 4 [(E)-N-(6-((4-(2-hydroxypropan-2-yl)piperidin-1-yl)methyl)-1-((1s,4s)-4-(isopropylcarbamoyl)cyclohexyl)-1H-benzo[d]imidazol-2(3H)-ylidene)-3-(trifluoromethyl)benzamide], respectively. The in vivo observed CLs for compounds 1, 2, 3, and 4 were 5.52, 3.51, 2.14, and 2.66 l/h⋅kg, respectively. These results indicate that in vitro metabolism kinetic data, incorporating contributions from both hydrolysis and NADPH-dependent metabolism, could be used to predict the systemic CL of compounds cleared via hydrolytic pathways provided that the in vitro assays thoroughly investigate the processes, including the contribution of other metabolic pathways and the possibility of saturation kinetics.

Abstract A169: Assessment of genotype-correlated sensitivity to the investigational PI3Kα selective inhibitor MLN1117 in preclinical models.

The PI3K-Akt-mTOR signaling pathway is the most dysregulated pathway in cancer. Specifically, helical and kinase domain mutations in the p110α subunit (E542K, E545K, H1047R) account for greater than 15% mutations in many cancers. In addition, copy number gain of the PIK3CA locus (>2.5 copies) also occur at a high frequency suggesting that these events are also associated with the oncogenic process. MLN1117 is an investigational PI3Kα selective inhibitor that is currently in Phase I dose escalation in subjects with advanced cancer ([NCT01449370][1]). It is widely hypothesized that PI3Kα isoform selective inhibitors will demonstrate considerable therapeutic benefit in PI3K “addicted” tumors; the work presented here is focused on identifying the association of genetic alterations with sensitivity to MLN1117. We have assessed the sensitivity profile both in vitro (n=47) and in vivo (n=29) using human tumor cell lines and primary human tumor xenografts, representing 9 different tumor types. PIK3CA gene point mutations are associated with sensitivity to MLN1117, in both in vitro and in vivo settings. 14 of 17 (82%) PIK3CA mutant cell lines demonstrate an in vitro IC50 of ≤ 5.0 μM and all non mutant models demonstrate an IC50 of > 5.0 μM. In vivo 12 of 15 PIK3CA mutant models including two non-hotspot alterations (R88Q and G106_R108del) demonstrate a TGI value of > 50%. Although copy number gain of PIK3CA is a frequent genetic event and could also be considered as a potential marker of sensitivity to PI3Kα inhibitors, in our assessment this was not apparent. We assessed 25 in vitro models and 16 in vivo models with a gain in the PIK3CA copy number greater than 2.5. Using a cut-off of IC50 of 50% in both settings we did not identify a significant association between PIK3CA amplification and MLN1117 sensitivity. Thus, PIK3CA mutation status is a significant predictor in classifying sensitive vs. non-sensitive models to MLN1117 whereas PIK3CA copy number status is not (compare p=0.00002 vs p = 0.7668). Our assessment of models with PTEN alterations found all models to be insensitive to MLN1117 using our established in vitro criteria whereas several of these models were found to have borderline sensitivity (range 41-50% TGI) in vivo. These results suggest the need for additional exploration of the PTEN role and MLN1117 sensitivity. Additional trends observed in this data set are 1) PIK3CA co-mutated with KRAS is insensitive to MLN1117 in both in vitro and in vivo systems; and 2) complex mutations involving PIK3CA and ARID1A or EIF4B demonstrate a distribution of sensitivities. We present data that support the use of PIK3CA mutation as a patient selection strategy for clinical studies of MLN1117 while the utility of identifying patients with increases in PIK3CA copy number for clinical studies of MLN1117 may be limited. Citation Information: Mol Cancer Ther 2013;12(11 Suppl):A169. Citation Format: Natasha Iartchouk, Erik Koenig, Pooja Shah, Jeff Szwaya, Michael Fitzgerald, Dong Mei Zhang, Yueying Cao, Keli Song, Rachael Brake. Assessment of genotype-correlated sensitivity to the investigational PI3Kα selective inhibitor MLN1117 in preclinical models. [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 A169. [1]: /lookup/external-ref?link_type=CLINTRIALGOV&access_num=NCT01449370&atom=%2Fmolcanther%2F12%2F11_Supplement%2FA169.atom

Abstract A41: Automated immunohistochemistry of phosphobiomarkers: Case study of MTOR (MLN0128) and PI3Kα (MLN1117) investigational inhibitors, single agent and in combination, on xenografts and mouse skin

Background: Immunohistochemistry provides visual context to protein modulation and is relied upon to elucidate mechanism of action, marker translocation and pharmacodynamics (PD) in drug discovery. Markers of the PI3K-MTOR pathway have been monitored in this way to aid in the clinical development of various inhibitors targeting this pathway. Here, assays were developed using preclinical xenograft PD studies to explore the feasibility of automated staining of pAKT 473, pS6, p4EBP1 and pRb. In preparation for clinical assessment of these markers in human skin samples; these markers were also examined in normal mouse skin. p4EBP1 was selected to examine the reproducibility of these automated assays and to explore the correlation between pathway inhibition in PD xenograft tissues and murine skin tissues. The automated staining technique was coupled with whole slide scanning and digital image analysis for quantitative comparison. Methods: Assays were chosen for evaluation based on known biology. FFPE PTEN-null breast cancer primary human tumor xenografts with high, medium and low biomarker expression, with companion skin samples, were selected for feasibility and reproducibility testing. Whole slide scanning and reproducibility analysis was executed using Aperio Scanscope XT and Spectrum software. Statistical analysis methods of mixed effect linear models and two sample t-tests were applied. Xenograft and skin samples from NCI-H1048 tumor bearing Nude mice treated with MTOR (MLN0128)and PI3Kα (MLN1117) investigational inhibitors were immunohistochemically stained for p4EBP1. Definiens object-based imaging software was used to isolate and analyze viable tumor area and exclude necrotic and stromal components in xenografts. A customized Definiens® algorithm was used on skin to identify the epithelial layer for analysis. Results: After evaluation, baseline biomarker expression in the xenograft and skin samples determined marker selection for future steps. Baseline p4EBP1 levels were high in both xenograft and skin. The reproducibility testing yielded consistent results in both xenograft and skin, validating autostaining use for PD studies. Phosphomarker p4EBP1 expression was inhibited after MLN0128 administration, with increased reduction after combination with MLN1117. Development of a novel analysis solution using Definiens software for skin samples permitted the area of interest to be restricted to the epithelial layer and excluded other epidermal features with similar morphometric characteristics. The p4EBP1 modulation was concordant between skin and tumor though more pronounced in skin, indicating that skin is a suitable surrogate tissue for determining modulation of this PI3K-MTOR pathway marker. Conclusions: Autostaining platforms provide reproducible results and stain consistently for PI3K-MTOR phosphomarkers on both xenografts and mouse skin samples. Utilizing p4EBP1, target effect was observed after MTOR and PI3Kα administration. This development and evaluation of the phosphobiomarker using automated staining technology and advanced image analysis supports preclinical PD assessment and combination clinical trials as well as influenced timepoint selection for clinical assessment. Citation Format: Anna Kreshock, Natalia Iartchouk, Yueying Cao, Jeffrey Szwaya, Feng Gao, Christopher Simpson, Evan Luongo, Kristine Burke, Esha Gangolli, Keisuke Kuida, Rachael Brake, Stephen Tirrell, Vaishali Shinde. Automated immunohistochemistry of phosphobiomarkers: Case study of MTOR (MLN0128) and PI3Kα (MLN1117) investigational inhibitors, single agent and in combination, on xenografts and mouse skin. [abstract]. In: Proceedings of the AACR Special Conference: Targeting the PI3K-mTOR Network in Cancer; Sep 14-17, 2014; Philadelphia, PA. Philadelphia (PA): AACR; Mol Cancer Ther 2015;14(7 Suppl):Abstract nr A41.

Abstract 791: Rational dose optimization for multi-drug cocktails

Proceedings: AACR Annual Meeting 2014; April 5-9, 2014; San Diego, CA The design of an optimal combination therapy can be thought of as a set of tradeoffs designed to maximize the efficacy of the combination for a given toxicity budget. As anticancer agents (even targeted ones) often have overlapping toxicities, an efficient approach to making efficacy-toxicity tradeoffs is critical. In this context, a major impediment is that the search space grows exponentially with each additional drug while in vivo studies have a fixed upper size limit. Here, we model the efficacy and toxicity landscapes of multidrug cocktails, and describe an approach to visualize and design efficient studies for three-drug combinations. We further demonstrate the application of our approach in a practical drug development context by showing its impact on actual study design, and validation with in vivo datasets. As our approach relies on equations from first principles, it is in theory extensible to an arbitrary number of drugs, subject to the practical constraints of drug dosing. First, we demonstrate an approach to visually determine the optimal dose combination for a three-drug combination with fully overlapping toxicity. Efficacy and toxicity isoboles visualized as surfaces in three-drug space demonstrate the optimal dose combination, which corresponds to the point of tangency between the MTD and efficacy isobole surfaces. Next, we derive an analytical method to efficiently estimate this point with a minimum of data, and without using graphical methods (important for extending the work beyond three dimensions). The combined efficacy and toxicity of a three-drug combination were modeled from first principles using isobolograms as a sum of the single-agent PK/Efficacy (PK/E) relationships and four combination terms (three binary and one ternary) to account for potential interactions between the drugs. 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. Through both analytical and numerical approaches, we determined the the single- and double-agent efficacy parameters uniquely determine the dose escalation path which provides the best estimate of combination efficacy. Simulated experiments demonstrate the optimal observation path up to MTD performs acceptably compared to a uniformly-gridded exposure space. Finally, we demonstrate the application of this approach in a proposed experimental design for three-drug combinations, and validate it with experimental data. While developing multi-drug cocktails poses many scientific and operational challenges, the approach presented here provides a straightforward route to preclinical testing and validation through the design of parsimonious studies that can explicitly define the contribution of each individual drug (and each two drug combination) to the overall efficacy and toxicity landscape of the cocktail. Citation Format: Christopher J. Zopf, Andrew Chen, Santhosh Palani, Rachael Brake, Mark Manfredi, Jeffrey Ecsedy, Wen Chyi Shyu, Arijit Chakravarty. Rational dose optimization for multi-drug cocktails. [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 791. doi:10.1158/1538-7445.AM2014-791

Abstract C176: The combination of mTORC1/2 and PI3Kα inhibition alleviates PI3K pathway reactivation and leads to significant antitumor activity in multiple preclinical xenograft models.

In many cancers, monotherapy targeted against signal transduction can lead to pathway reactivation by feedback mechanisms. The phosphoinositide 3-kinase (PI3K) pathway has been studied in this regard. Allosteric inhibitors of mammalian target of rapamycin complex 1 (mTORC1), such as rapamycin analogs, interrupt the normal negative regulation of insulin receptor substrate-1 (IRS-1) mediated by ribosomal protein S6 kinase (S6K), leading to pathway reactivation, as demonstrated by AKT phosphorylation and activation of its downstream substrates. Recent data suggest that administration of catalytic mTOR inhibitors can lead to similar feedback. Using in vitro models we have shown that the investigational catalytic mTORC1/2 inhibitor MLN0128 can lead to pathway reactivation, as measured by increases in AKT phosphorylation at threonine 308 (T308) over time. Furthermore we have shown pre-clinically that administration of the investigational isoform-selective PI3Kα inhibitor MLN1117 mitigates this pathway reactivation. In in vitro studies the combination of MLN0128 plus MLN1117 resulted in greater inhibition of key pathway regulators (pAKT, pS6, and p4EBP1) than either single agent alone, and a sustained decrease in AKT (T308) phosphorylation. The combination also induced greater apoptosis pre-clinically, as indicated by increases in cleaved poly (ADP-ribose) polymerase (PARP). The antiproliferative effects of this combination were assessed in 14 human tumor cell lines known to demonstrate sensitivity to single-agent MLN0128. In all settings (across tumor types and genotypes), administration of MLN0128 plus MLN1117 resulted in additive antiproliferative activity, except in two breast cancer models in which the activity of the combination was considered to show synergistic activity. In in vivo studies, concurrent oral administration of MLN0128 plus MLN1117 resulted in additive or synergistic antitumor activity in human tumor xenograft models. The combined antitumor activity correlated with modulation of pharmacodynamic markers and was observed irrespective of the dosing schedules tested or genotypes of the models, which included PI3K-mutant breast, PIK3CA/KRAS-mutant colon, PTEN-null breast, and EGFR-mutant non-small-cell lung cancer.The results from these preclinical experiments suggest that, in certain patient populations, co-administration of MLN0128 plus MLN1117 may lead to greater therapeutic effect than either single agent alone. The flexibility of being able to schedule these compounds independently and the isoform-selective profile of MLN1117 could provide distinct clinical advantages over other mechanisms of PI3K pathway inhibition. Together these data support the rationale that has led to the investigation of the safety of this combination clinically in an ongoing phase I study ([NCT01899053][1]). Citation Information: Mol Cancer Ther 2013;12(11 Suppl):C176. Citation Format: Rachael L. Brake, Robyn Fabrey, Jeff Szwaya, Michael Fitzgerald, Natasha Iartchouk, Xin Guo, Keisuke Kuida, Fabian Zohren, Mark Manfredi. The combination of mTORC1/2 and PI3Kα inhibition alleviates PI3K pathway reactivation and leads to significant antitumor activity in multiple preclinical xenograft models. [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 C176. [1]: /lookup/external-ref?link_type=CLINTRIALGOV&access_num=NCT01899053&atom=%2Fmolcanther%2F12%2F11_Supplement%2FC176.atom