Discovery of a Series of 5,11-Dihydro-6H-benzo[e]pyrimido[5,4-b][1,4]diazepin-6-ones as Selective PI3K-δ/γ Inhibitors

Abstract

Dual inhibition of PI3K-δ and PI3K-γ is an established therapeutic strategy for treatment of hematological malignancies. Reported molecules targeting PI3K-δ/γ selectively are chemically similar and based upon isoquinolin-1(2H)-one or quinazolin-4(3H)-one scaffolds. Here we report a chemically distinct series of potent, selective PI3K-δ/γ inhibitors based on a 5,11-dihydro-6H-benzo[e]pyrimido[5,4-b][1,4]diazepin-6-one scaffold with comparable biochemical potency and cellular effects on PI3K signaling. We envisage these molecules will provide useful leads for development of next-generation PI3K-δ/γ targeting therapeutics.

PI3K-δ and PI3K-γ are members of the Class I Type IA and Class I Type IB family of phosphatidylinositol-4,5-bisphosphate 3-kinases (PI3Ks). Unlike the related PI3K-α/-β, which are ubiquitously expressed, PI3K-δ and PI3K-γ are expressed primarily in leukocytes and perform a number of roles in regulation of the immune system. PI3K-δ has been shown to be involved in B-cell activation, proliferation, homing, and retention in lymphoid tissues; PI3K-γ regulates T-cell proliferation and cytokine production.¹

PI3K-δ and PI3K-γ are the dominantly expressed PI3K isoforms in B- and T-cells, respectively, where they are key nodes in the PI3K/Akt/mTOR pathway. This pathway is misregulated in a number of blood-borne cancers including chronic lymphocytic leukemia (CLL), follicular lymphoma (FL), and indolent non-Hodgkin’s lymphoma (iNHL).¹

PI3K-δ signaling drives malignant B-cell proliferation. Selective inhibition of PI3K-δ using small molecule inhibitor Idelalisib has proven to be an effective treatment for CLL when used in combination with rituximab, a chimeric monoclonal antibody that targets the B-lymphocyte antigen CD20.² PI3K-γ activation is key for inflammatory cell recruitment to tumors, associated with angiogenesis and tumor growth, which can be attenuated by knockdown or pharmacological inhibition of PI3K-γ.^3,4 As these two kinases play distinct and complementary roles in immune function, dual inhibition of PI3K-δ and PI3K-γ is also an attractive strategy for broadly targeting hematological malignancies. Inhibition of PI3K-δ/γ is well tolerated with mild, reversible side effects reported in the clinic.⁵ The dual inhibitor Duvelisib is currently in Phase III clinical trials for CLL, FL and Phase II clinical trials for iNHL, either alone or in combination with monoclonal antibody therapy.⁶ Additionally Duvelisib has potent anti-inflammatory and joint protective effects in murine models of rheumatoid arthritis.⁷ A Phase IIa exploratory clinical trial in mild allergic asthma met several secondary end points demonstrating proof-of-concept that next generation PI3K-δ/γ inhibitors may also prove effective in this disease area.⁸

Currently reported selective dual inhibitors of PI3K-δ/γ are based upon isoquinolin-1(2H)-one or quinazolin-4(3H)-one scaffolds (Figure 1).⁹⁻¹¹ Here we report a chemically distinct series of potent, selective PI3K-δ/γ inhibitors based on a 5,11-dihydro-6H-benzo[e]pyrimido[5,4-b][1,4]diazepin-6-one scaffold with comparable enzymatic potency and cellular effects on PI3K-δ signaling.

Figure 1.

Structures of PI3K-δ/γ selective inhibitors reported in the literature and described in this work.

Throughout the course of a screening campaign designed to identify antileukemic compounds, we observed that compound 1 (FMF-01-085-1) shows antiproliferative activity in T-cell acute lymphocytic leukemia (T-ALL) cell lines (IC₅₀ MOLT4 cells = 33 nM; IC₅₀ Jurkat cells = 166 nM). Subsequent kinome profiling revealed the primary targets of this compound are PI3K-δ/γ (Table 1, Supporting Table 1, Supporting Figure 1), leading us to explore the SAR of this series. Compounds were synthesized according to Scheme 1. Analogues from our initial screen lacking an aryl-sulfonamide showed no inhibitory effects on PI3K-δ/γ (e.g., compound 19, FMF-01-086-2, Supporting Table 2); therefore, we focused our synthetic efforts on compounds containing this moiety.¹²

Table 1. SAR, Isoform Selectivity, and Aurora Kinase Selectivity of PI3K-δ/γ Inhibitors.

^aIC₅₀s measured using ADAPTA assay format (ThermoFisher Scientific).

^bIC₅₀s measured using Z’LYTE assay format (ThermoFisher Scientific). IC₅₀s plotted from the average of duplicate experiments. Errors are reported as ±95% confidence interval.

Scheme 1. Synthetic Route for Synthesis of 5,11-Dihydro-6H-benzo[e]pyrimido[5,4-b][1,4]diazepin-6-ones.

Reaction conditions: (i) DIEA, 1,4-dioxane, 50 °C; (ii) Fe, AcOH, 50 °C; (iii) NaH, MeI, DMF, 0 °C; (iv) XPhos, Pd₂(dba)₃, Cs₂CO₃,1,4-dioxane, 95 °C.

We have previously reported that the 5,11-dihydro-6H-benzo[e]pyrimido[5,4-b][1,4]diazepin-6-one scaffold is capable of binding to the ATP binding pocket of LRRK2,¹³ ERK5,¹⁴ and AuroraA/B kinases¹⁵ and to the acetyl-lysine binding pocket of the BRD4 bromodomains.¹⁶ However, methylation of the phenyl ring in the tricyclic core is not tolerated by these targets. Kinome profiling at 1 μM compound concentration revealed that 1 has excellent selectivity across the human kinome, with a selectivity score, S₁₀ of 0.013. Importantly, other targets in the PI3K pathway such as Akt, DNA-PK, BTK, and mTOR are not inhibited (Supporting Table 1, Supporting Figure 1) and BRD4 activity is low (BRD4_1 IC₅₀ = 6.0 μM, Supporting Table 3). The compound has some inhibitory effects on PIP5K2C (PIP4K-γ), a lipid kinase with low levels of activity in vitro. In our experience this level of inhibition corresponds to micromolar biochemical IC₅₀.

As some activity is present for PI3K-α (and H1047L/Y mutants) we measured PI3K-α and PI3K-β IC₅₀s to determine the isoform selectivity. Compound 1 is 26-fold selective for PI3K-δ over PI3K-α and 270-fold selective over PI3K-β. The only off-target activity of concern is against Aurora kinases A and B. Enzymatic testing revealed that compound 1 has 30-fold selectivity over Aurora A and 60-fold over Aurora B.

This prompted us to further investigate the factors conferring selectivity to the series (Table 1). Meta substitution of the aniline ring with an N-substituted sulfonamide biases the potency of the compounds toward PI3K-δ/γ (compounds 1, 3, 5, 7, 9, 10, 12). Conversely, the same substituents in the para position improve the Aurora A/B potency and reduce the PI3K-δ/γ potency (compounds 2, 4, 6, 8, 11, 13).

Compounds containing an unsubstituted sulfonamide nitrogen are equipotent against PI3K-δ/γ and AuroraA/B (compounds 14, 16, 17). It has been shown that ortho-substitution adjacent to the hinge-binding motif can remove AuroraA/B activity from this scaffold.¹⁵ Ortho-methylation of the aniline ring of potent compound 16 to give compound 18 shows the expected low AuroraA/B activity but also has dramatically reduced PI3K-δ/γ activity.

Covalent inhibitors have been reported for PI3K-α that target a nonconserved cysteine unique to this isoform.¹⁷ Examination of X-ray crystal structures showed there are no accessible cysteine residues proximal to the ATP binding pocket in PI3K-δ and PI3K-γ (PDB IDs: 4XE0, 4EZJ). As compound 1 contains an acrylamide, we used LC–MS/MS experiments with purified PI3K-δ protein to confirm that 1 does not covalently label the protein. Therefore, we sought to remove this reactive functionality while maintaining on-target potency and kinome selectivity. Gratifyingly compound 9 (FMF-02-109-1) and compound 12 (FMF-02-063-1) both maintained potent inhibition of PI3K-δ/γ and showed comparable selectivity for PI3K-δ against PI3K-α (40-fold, 25-fold) and improved selectivity against PI3K-β (360-fold, 2300-fold), Aurora A (65-fold, 70-fold) and Aurora B (71-fold, 71-fold). Kinome profiling revealed that compounds 9 and 12 also maintain an excellent selectivity profile with S₁₀ of 0.010 and 0.008 respectively (Figure 2A, Supporting Table 1). Additionally low BRD4 activity was observed for all compounds (BRD4_1 IC₅₀ = 18.8 μM, 10.8 μM respectively, Supporting Table 3).

Figure 2.

(A) Kinome-wide selectivity profile of compounds 9 and 12. (B) Comparison of biochemical IC₅₀ values of compounds 9, 12, Duvelisib, and Idelalisib in PI3K-δ and PI3K-γ ADAPTA assays. (C) Effects of inhibitors 9 and 12 on AKT and S6RP phosphorylation in isogenic HMEC lines expressing CA-p110α, CA-p110β, or CA-p110δ treated with the indicated compounds at 0.01, 0.1, or 1 μM for 1 h. Du, Duvelisib; Id, Idelalisib.

In order to have a more direct comparison of potency to the currently available clinical compounds we measured the IC₅₀s of Duvelisib and Idelalisib in the ADAPTA assay format (Figure 2B). In the PI3K-δ assay, compounds 1, 9 and 12 are equipotent to Idelalisib, whereas Duvelisib is more effective. In the PI3K-γ assay, compounds 1, 9, and 12 are comparable to Duvelisib. Idelalisib, a PI3K-δ specific inhibitor, is much less potent against PI3K-γ, as expected (Table 1, Figure 2B).

Encouraged by the potency of our inhibitors in comparison to the current best-in-class molecules, we next explored the effect of our compounds on PI3K signaling in isogenic HMEC cell lines where PI3K signaling is driven exclusively by CA-p110-α, CA-p110-β, or CA-p110-δ under serum starved conditions, and compared them to Duvelisib and Idelalisib. Commensurate with their biochemical activities, Idelalisib, 9, and 12 show comparable inhibition of, and selectivity for, PI3K-δ signaling at 10 nM concentration (Figure 2C). Duvelisib is the most potent PI3K-δ inhibitor; however, it is less selective against PI3K-β in a cellular context.

To rationalize the isozyme selectivity we docked 9 and 12 into PI3K-δ (PDB: 4XE0). It is known that exploitation of a selectivity pocket formed by rearrangement of a methionine residue in the ATP binding pockets of PI3K-δ and PI3K-γ can improve isozyme selectivity. This has been achieved to date by the design of molecules that adopt a propellor-like conformation, such as Idelalisib and Duvelisib.¹⁸ In the docking models of 9 and 12 in complex with PI3K-δ the butterfly-like conformation of the benzodiazipinone scaffold causes the tolyl group to occupy the selectivity pocket formed by Met 752 while still allowing the pyrimidine-hinge contact to occur. The H-bond of the sulfonamide to Thr 833 may explain the preference for this functional group and the requirement for 3-substitution vs 4-substitution of the aniline ring (Figure 3, Supporting Figure 3, Table 1).

Figure 3.

Docking model of 12 bound to PI3K-δ.

Finally, we treated leukemia cell lines and patient derived primary CLL cells to examine the effects of PI3K-δ/γ inhibition on cell viability. Compounds 9 and 12 were comparably potent to Duvelisib and more potent than Idelalisib in the tested cell lines. In the CLL cells Idelalisib showed little effect, consistent with previously published reports,¹⁹ as did Duvelisib and compound 12, whereas compound 9 demonstrated a dose-dependent reduction in viability (Table 2, Supporting Figure 2).

Table 2. Activity of PI3K-δ/γ Inhibitors in Cell Viability Assays.

compound ID	IC50 (μM) Jurkat	IC50 (μM) Molt4	IC50 (μM) MV4:11	IC50 (μM) Molm 14	IC50 (μM) Loucy	IC50 (μM) CLLa
9	1.6	1.2	0.96	0.61	0.72	3
12	1.4	1.3	0.62	1.0	0.35	>10
Idelalisib	7.9	10.6	6.3	3.6	8.4	>10
Duvelisib	1.9	2.3	4.4	1.2	0.98	>10

^aPatient derived primary cells. IC₅₀s plotted from average of three replicates.

The series of compounds described in this work represent a novel class of PI3K-δ/γ inhibitors. We were able to develop potent, selective molecules with cellular activity and drug-like properties in the absence of structural information. We envisage these molecules will provide useful leads for development of next-generation PI3K-δ/γ targeting therapeutics. Investigation into the binding mode of these molecules by X-ray crystallography may yield rationale for development of molecules with superior PI3K-δ/γ selectivity using structure-based design.

Glossary

ABBREVIATIONS

PI3K

phosphatidylinositol-4,5-bisphosphate 3-kinase

CLL

chronic lymphocytic leukemia

FL

follicular lymphoma

iNHL

indolent non-Hodgkin’s lymphoma

T-ALL

T-cell acute lymphocytic leukemia

Akt

protein kinase B

BTK

Bruton’s tyrosine kinase

mTOR

mechanistic target of rapamycin

DNA-PK

DNA-dependent protein kinase

BRD4

bromodomain-containing protein 4

PLK1

polo-like kinase 1

LRRK2

leucine-rich repeat kinase 2

ERK5

extracellular-signal-regulated kinase 5

JAK2

Janus Kinase 2

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsmedchemlett.6b00209.

• Biochemical and cellular assay protocols, compound synthesis and characterization, supporting figures, and supporting tables (PDF)

Author Present Address

(X.D.) State Key Laboratory of Cellular Stress Biology, Innovation Center for Cell Signaling Network, School of Life Sciences, Xiamen University, Xiamen, Fujian 361102, China.

The authors declare no competing financial interest.