Edna F. Choo

Hematopoietic progenitor kinase 1 (HPK1) serves a key immunosuppressive role as a negative regulator of T-cell receptor (TCR) signaling. HPK1 loss-of-function is associated with augmentation of immune function and has demonstrated synergy with immune checkpoint inhibitors in syngeneic mouse cancer models. These data offer compelling evidence for the use of selective small molecule inhibitors of HPK1 in cancer immunotherapy. We identified a novel series of isoquinoline HPK1 inhibitors through fragment-based screening that displayed promising levels of biochemical potency and activity in functional cell-based assays. We used structure-based drug design to introduce key selectivity elements while simultaneously addressing pharmacokinetic liabilities. These efforts culminated in a molecule demonstrating subnanomolar biochemical inhibition of HPK1 and strong in vitro augmentation of TCR signaling in primary human T-cells. Further profiling of this molecule revealed excellent kinase selectivity (347/356 kinases in vitro safety profile, and good projected human pharmacokinetics.

Hematopoietic progenitor kinase 1 (HPK1) is implicated as a negative regulator of T-cell receptor-induced T-cell activation. Studies using HPK1 kinase-dead knock-in animals have demonstrated the loss of HPK1 kinase activity resulted in an increase in T-cell function and tumor growth inhibition in glioma models. Herein, we describe the discovery of a series of small molecule inhibitors of HPK1. Using a structure-based drug design approach, the kinase selectivity of the molecules was significantly improved by inducing and stabilizing an unusual P-loop folded binding mode. The metabolic liabilities of the initial 7-azaindole high-throughput screening hit were mitigated by addressing a key metabolic soft spot along with physicochemical property-based optimization. The resulting spiro-azaindoline HPK1 inhibitors demonstrated improved in vitro ADME properties and the ability to induce cytokine production in primary human T-cells.

Herein we describe the discovery of A-1331852, a first-in-class orally active BCL-X_L inhibitor that selectively and potently induces apoptosis in BCL-X_L dependent tumor cells. This molecule was generated by re-engineering our previously reported BCL-X_L inhibitor A-1155463 using structure-based drug design. Key design elements included rigidification of the A-1155463 pharmacophore and introduction of sp³-rich moieties capable of generating highly productive interactions within the key P4 pocket of BCL-X_L. A-1331852 has since been used as a critical tool molecule for further exploring BCL-2 family protein biology, while also representing an attractive entry into a drug discovery program.

IRAK4 kinase activity transduces signaling from multiple IL-1Rs and TLRs to regulate cytokines and chemokines implicated in inflammatory diseases. As such, there is high interest in identifying selective IRAK4 inhibitors for the treatment of these disorders. We previously reported the discovery of potent and selective dihydrobenzofuran inhibitors of IRAK4. Subsequent studies, however, showed inconsistent inhibition in disease-relevant pharmacodynamic models. Herein, we describe application of a human whole blood assay to the discovery of a series of benzolactam IRAK4 inhibitors. We identified potent molecule 19 that achieves robust in vivo inhibition of cytokines relevant to human disease.

A series of pyrazolopyrimidine inhibitors of IRAK4 were developed from a high-throughput screen (HTS). Modification of an HTS hit led to a series of bicyclic heterocycles with improved potency and kinase selectivity but lacking sufficient solubility to progress in vivo. Structure-based drug design, informed by cocrystal structures with the protein and small-molecule crystal structures, yielded a series of dihydrobenzofurans. This semisaturated bicycle provided superior druglike properties while maintaining excellent potency and selectivity. Improved physicochemical properties allowed for progression into in vivo experiments, where lead molecules exhibited low clearance and showed target-based inhibition of IRAK4 signaling in an inflammation-mediated PK/PD mouse model.

The epigenetic regulator CBP/P300 presents a novel therapeutic target for oncology. Previously, we disclosed the development of potent and selective CBP bromodomain inhibitors by first identifying pharmacophores that bind the KAc region and then building into the LPF shelf. Herein, we report the “hybridization” of a variety of KAc-binding fragments with a tetrahydroquinoline scaffold that makes optimal interactions with the LPF shelf, imparting enhanced potency and selectivity to the hybridized ligand. To demonstrate the utility of our hybridization approach, two analogues containing unique Asn binders and the optimized tetrahydroquinoline moiety were rapidly optimized to yield single-digit nanomolar inhibitors of CBP with exquisite selectivity over BRD4(1) and the broader bromodomain family.

The utility of animal models in understanding drug disposition and extrapolating to human in a direct manner is limited, confounded by differences in enzyme and transporter substrate specificity and expression between species. Conversely, certain clinical pharmacology or mechanistic studies cannot be conducted for ethical and practical reasons. Thus, humanized animal models could be useful to bridge the gap between preclinical and clinical pharmacology and gain relevant insight into the determinants of drug disposition and drug–drug interaction (DDI).

Inhibition of the bromodomain of the transcriptional regulator CBP/P300 is an especially interesting new therapeutic approach in oncology. We recently disclosed in vivo chemical tool 1 (GNE-272) for the bromodomain of CBP that was moderately potent and selective over BRD4(1). In pursuit of a more potent and selective CBP inhibitor, we used structure-based design. Constraining the aniline of 1 into a tetrahydroquinoline motif maintained potency and increased selectivity 2-fold. Structure–activity relationship studies coupled with further structure-based design targeting the LPF shelf, BC loop, and KAc regions allowed us to significantly increase potency and selectivity, resulting in the identification of non-CNS penetrant 19 (GNE-781, TR-FRET IC₅₀ = 0.94 nM, BRET IC₅₀ = 6.2 nM; BRD4(1) IC₅₀ = 5100 nΜ) that maintained good in vivo PK properties in multiple species. Compound 19 displays antitumor activity in an AML tumor model and was also shown to decrease Foxp3 transcript levels in a dose dependent manner.

Species differences in the expression, activity, regulation, and substrate specificity of metabolizing enzymes preclude the use of animal models to predict clinical drug–drug interactions (DDIs). The objective of this work is to determine if the transgenic (Tg) Cyp3a^–/–Tg-3A4_Hep/Int and Nr1i2/Nr1i3^–/–-Cyp3a^–/–Tg-PXR-CAR-3A4/3A7_Hep/Int (PXR-CAR-CYP3A4/3A7) mouse models could be used to predict in vivo DDI of 10 drugs; alprazolam, bosutinib, crizotinib, dasatinib, gefitinib, ibrutinib, regorafenib, sorafenib, triazolam, and vandetinib (as victims); with varying magnitudes of reported CYP3A4 clinical DDI. As an assessment of the effect of CYP3A4 inhibition, these drugs were coadministered to Cyp3a^–/–Tg-3A4_Hep/Int mice with the CYP3A inhibitor, itraconazole. For crizotinib, regorafenib, sorafenib, and vandetanib, there was no significant increase of AUC observed; with alprazolam, bosutinib, ibrutinib, dasatinib, and triazolam, pretreatment with itraconazole resulted in a 2-, 4-, 17-, 7-, and 15-fold increase in AUC, respectively. With the exception of gefinitib for which the DDI effect was overpredicted (12-fold in Tg-mice vs 2-fold in the clinic), the magnitude of AUC increase observed in this study was consistent (within 2-fold) with the clinical DDI observed following administration with itraconazole/ketoconazole. As an assessment of CYP3A4 induction, following rifampin pretreatment to PXR-CAR-3A4/3A7_Hep/Int mice, an 8% decrease in vandetanib mean AUC was observed; 39–52% reduction in AUC were observed for dasatinib, ibrutinib, regorafenib, and sorafenib compared to vehicle treated mice. The greatest effect of rifampin induction was observed with alprazolam, bosutinib, crizotinib, gefitinib, and triazolam where 72–91% decrease in AUC were observed. With the exception of vandetanib for which rifampin induction was under-predicted, the magnitude of induction observed in this study was consistent (within 2-fold) with clinical observations. These data sets suggest that, with two exceptions, these transgenic mice models were able to exclude or capture the magnitude of CYP3A4 clinical inhibition and induction. Data generated in transgenic mice may be used to gain confidence and complement in vitro and in silico methods for assessing DDI potential/liability.

Data from the clinical absolute bioavailability (F) study with cobimetinib suggested that F was lower than predicted based on its low hepatic extraction and good absorption. The CYP3A4 transgenic (Tg) mouse model with differential expression of CYP3A4 in the liver (Cyp3a(-/-)Tg-3A4Hep) or intestine (Cyp3a(-/-)Tg-3A4Int) and both liver and intestine (Cyp3a(-/-)Tg-3A4Hep/Int) were used to study the contribution of intestinal metabolism to the F of cobimetinib. In addition, the effect of CYP3A4 inhibition and induction on cobimetinib exposures was tested in the Cyp3a(-/-)Tg-3A4Hep/Int and PXR-CAR-CYP3A4/CYP3A7 mouse models, respectively. After i.v. administration of 1 mg/kg cobimetinib to wild-type [(WT) FVB], Cyp3a(-/-)Tg-3A4Hep, Cyp3a(-/-)Tg-3A4Int, or Cyp3a(-/-)Tg-3A4Hep/Int mice, clearance (CL) (26-35 ml/min/kg) was similar in the CYP3A4 transgenic and WT mice. After oral administration of 5 mg/kg cobimetinib, the area under the curve (AUC) values of cobimetinib in WT, Cyp3a(-/-)Tg-3A4Hep, Cyp3a(-/-)Tg-3A4Int, or Cyp3a(-/-)Tg-3A4Hep/Int mice were 1.35, 3.39, 1.04, and 0.701 μM⋅h, respectively. The approximately 80% lower AUC of cobimetinib in transgenic mice when intestinal CYP3A4 was present suggested that the intestinal first pass contributed to the oral CL of cobimetinib. Oxidative metabolites observed in human circulation were also observed in the transgenic mice. In drug-drug interaction (DDI) studies using Cyp3a(-/-)Tg-3A4Hep/Int mice, 8- and 4-fold increases in oral and i.v. cobimetinib exposure, respectively, were observed with itraconazole co-administration. In PXR-CAR-CYP3A4/CYP3A7 mice, rifampin induction decreased cobimetinib oral exposure by approximately 80%. Collectively, these data support the conclusion that CYP3A4 intestinal metabolism contributes to the oral disposition of cobimetinib and suggest that under certain circumstances the transgenic model may be useful in predicting clinical DDIs.