[3a,4]-Dihydropyrazolo[1,5a]pyrimidines: Novel, Potent, and Selective Phosphatidylinositol-3-kinase β Inhibitors

Abstract

A series of novel [3a,4]dihydropyrazolo[1,5a]pyrimidines were identified, which were highly potent and selective inhibitors of PI3Kβ. The template afforded the opportunity to develop novel SAR for both the hinge-binding (R₃) and back-pocket (R₄) substitutents. While cellular potency was relatively modest due to high protein binding, the series displayed low clearance in rat, mouse, and monkey.

The phosphatidylinositol 3-kinases (PI3Ks) are members of a larger family of lipid kinases that phosphorylate the inositol moiety of membrane-associated phosphatidyl inositols.¹ PI3Ks catalyze the conversion of phosphatidylinositol-4,5-diphosphate (PIP₂) to phosphatidylinositol-3,4,5-triphosphate (PIP₃), which allows recruitment of proteins containing the PIP₃-binding pleckstrin homology (PH) domain to the plasma membrane. The PI3K class I family comprises four isoforms (α, β, δ, and γ), with the α and β isoforms most widely expressed. Each is a heterodimer consisting of an isoform-specific p110 catalytic subunit and a regulatory subunit.

PI3Ks are an important mediator of receptor tyrosine kinase (RTK) and G-protein-coupled receptor (GPCR) signal transduction and are involved in the regulation of cellular metabolism, survival, and proliferation.² Aberrant activation of the PI3K pathway is strongly associated with malignant transformation, through either overexpression of PI3Kα, mutations that lead to constitutive activity of PI3Kα, or loss of function of the tumor suppressor PTEN (phosphatase and tensin homologue), which opposes PI3K function by converting PIP₃ to PIP₂.³⁻⁹ While the PI3K isoforms share the same substrate specificity, they have distinct biochemical roles and cellular functions.¹⁰ In particular, there is considerable evidence that in PTEN-null tumors, malignant transformation is primarily driven by the PI3Kβ isoform.¹¹⁻¹⁵ The development of PI3Kβ-selective inhibitors is therefore an attractive target for the treatment of PTEN-null tumors.

We have previously disclosed imidazopyrimidones (1) and triazolopyrimidones (2), as shown in Figure 1, which are potent and selective inhibitors of PI3Kβ.^16,17 However, the compounds were not suitable for in vivo studies due to high clearance in rat and mouse. Both 1 and 2 contain an embedded benzylamine moiety, a potential site for metabolism, which might contribute to the high clearance. We therefore designed a pyrazolo-pyrimidine scaffold, in which the benzyl group is attached to a carbon atom rather than a nitrogen, in an attempt to improve clearance.

Figure 1.

Imidazopyrimidone (1), triazolopyrimidone (2), and pyrazolopyrimidine (3) PI3Kβ inhibitors.

Pyrazolopyrimidines were generally prepared according to previously reported methods¹⁸⁻²⁰ as shown in Scheme 1. An α-cyanocarbonyl compound I is alkylated²¹ with a benzyl bromide II and then condensed with hydrazine to afford the key aminopyrazole intermediate IV. The aminopyrazole is then condensed with a β-carbonyl-containing ester V to afford the 7-hydroxypyrazolopyrimidines VI. Where V is diethyl malonate, the resulting dihydroxypyrimidine (VI, R₃ = OH) is chlorinated with POCl₃, and the chlorines are selectively displaced with hydroxide and then morpholine to afford the final products VII. To introduce the R₄ substituent, VII is again chlorinated with POCl₃, and chlorine is either displaced with nucleophiles (methoxide or amines) or carbonylated, and the resulting functional groups are elaborated according to standard methodology.

Scheme 1. Synthesis of Pyrazolopyrimidines.

Reagents and conditions: (a) K₂CO₃, THF, reflux. (b) NH₂NH₂·H₂O, EtOH, reflux. (c) NaOEt, EtOH, reflux or HOAc, 140°, microwave. (d) POCl₃, N,N-diethylaniline, 110°. (e) 1 N NaOH, THF. (f) Morpholine, EtOH, 140°, microwave. (g) RNH₂, MeOH, 135° or NaOMe, MeOH, 0–40°, or Pd(OAc)₂, dppf, CO, Et₃N, DMF/THF, 70°.

To evaluate the pyrazolopyrimidine core, we prepared two 2-methyl-5-morpholine-7-hydroxy analogues with different benzyl substituents at position 3. The pyrazolopyrimidine core introduces the possibility of three tautomeric forms, two of which display an NH moiety, which is not the case in either the imidazopyrimidone (1) or the triazolopyrimidone (2) cores. We were gratified to see that despite these potentially significant structural differences, both pyrazolopyrimidines retained the high PI3Kβ potency and excellent isoform selectivity observed with the other inhibitor series, as shown in Table 1. Because the SAR at this benzyl position had previously been well-established,¹⁶ we were able to focus our further efforts mainly on the 2-methyl-3-trifluoromethyl benzyl analogues while we sought to better define the much less well-established hinge-binding and back-pocket SAR.

Table 1. SAR of 3-Benzyl 7-Hydroxypyrazolopyrimidinines^a.

		PI3K pIC5022
compd	R2	α	β	δ	γ
4a	2-Me-3-CF3-phenyl	6.5	9.4	7.6	6.5
4b	1-naphthyl	6.5	9.0	7.6	6.2

^apIC₅₀ values are the mean of a minimum of two determinations.

A docking pose of 4a in a PI3Kβ homology model¹⁶ is shown in Figure 2. The morpholine oxygen provides the key hydrogen-bonding interaction with the hinge residue Val⁸⁵⁴. The pyrimidine hydroxyl interacts with the back-pocket Tyr⁸³⁹ via a bridging water molecule. A key feature is the interaction of the benzylic side chain with the flexible P-loop, which caps the ATP binding pocket. A conformational change induced by binding of the inhibitor creates a small hydrophobic pocket lined by Met⁷⁷⁹ and Trp⁷⁸⁷, which is ideally suited to accommodate a 2,3-disubstituted phenyl ring. This is consistent with the observed SAR of substituted benzyl moieties and is hypothesized to be a major contributor to the β-isoform selectivity. With this model in mind, we designed analogues to explore the SAR for the other substituents on the pyrazolopyrimidine core.

Figure 2.

Homology model of 4a docked into PI3Kβ. (a) Ribbon diagram showing key residues and interactions. (b) Surface showing hydrophobic pocket induced by the 2,3-disubstituted benzyl group.

We next examined the hinge-binding morpholine moiety. We were particularly interested in analogues, which modulated the basicity of the morpholine either by replacing the nitrogen or by converting it to an amide. The tetrahydropyranyl and pyridyl analogues 5a and 5e were prepared from the corresponding β-keto esters, and the morphlin-3-one analogue 5b was prepared via a Buchwald amidation²³ of the 5-chloro-7-hydroxypyrazolo-pyrimidone intermediate.

The enzymatic activity of these analogues is shown in Table 2. Although the morpholine nitrogen is not involved in any specific interactions with the enzyme, its substitution by carbon (5a) causes a 10-fold drop in activity. The piperidone analogue 5b drops nearly 1000-fold in activity, possibly due to an intramolecular hydrogen bond between the carbonyl and the NH of the pyrimidone tautomer, which could cause the morpholinone ring to adopt an unfavorable conformation. The 2-methylmorpholine substitution had provided equipotent analogues in the imidazopyrimidone series¹⁷ but gave a less potent compound in this series, and resolution of the two enantiomers was not attempted.

Table 2. Enzymatic Activity of 5a–e.

We then examined the effect of substitution at the 2-position of the pyrazolopyridine moiety. According to the homology model, there are no close contacts to the protein in this region, so we expected that a variety of substitutions might be tolerated. Enzymatic activity of the 2-substituted pyrazolopyrimidines is shown in Table 3. The presence of a 2-position substitutent improves activity about 10-fold (5d vs 4a). Small, aliphatic substituents like Me, Et, or CF₃ are all well tolerated and provide similar activity. There was a 10–100-fold reduction in activity with the polar hydroxyl or amino substituents. Interestingly, the loss in activity with the hydroxyl analogue 6f could be overcome by methylation of the oxygen (6g), suggesting that the poor activity of 6f and 6h is less due to changes in the electronics of the pyrazolopyrimidine core and more likely due to introduction of hydrogen bond donors at that position.

Table 3. Enzymatic Activity of 2-Substituted Analogues.

				PI3K pIC50
compd	R3a	R2b	R1	α	β	δ	γ
5d	A	X	H	6.5	8.3	7.7	6.3
4a	A	X	Me	6.5	9.4	7.6	6.5
6a	B	Y	Et	6.8	9.2	7.7	6.4
6b	C	Y	Et	<5.8	7.9	<6.4	<5.1
6c	A	Y	CF3	6.7	9.4	8.2	6.6
6d	B	Y	CF3	6.2	8.6	7.1	6.4
6e	C	Y	CF3	5.6	8.2	7.0	<5.2
6f	C	X	OH	<5.7	7.2	<6.2	<5.3
6g	B	Y	OMe	6.8	8.9	7.6	6.3
6h	C	Y	NH2	5.8	8.1	6.2	<5.4

^aA = morpholinyl, B = 4-pyridyl, and C = 4-tetrahydropyranyl.

^bX = 2-Me-3-CF₃-phenyl, and Y = 2,3-dichlorophenyl.

The tautomeric nature of the pyrazolopyrimidine template permits substitution at the 7-position, something not possible with the imidazopyrimidone and triazolopyrimidone templates. This allowed us to prepare variously substituted analogues to evaluate the requirement for a hydrogen-bonding interaction with Tyr⁸³⁹, either directly or via a bound water molecule.

The enzymatic activity of the various 7-substituted analogues is shown in Table 4. While the original 7-hydroxy (4a) provided the highest activity, a wide variety of substitutions could be tolerated to varying degrees. In particular, the 7-carboxy analogue 7d retained considerable activity, while the amino, hydroxyl-amine, carboxamide, and urea analogues 7g, 7i, 7e, and 7l also displayed modest potency. Interestingly, both the unsubstituted analogue 7c and the chloro analogue 7b retained measurable activity, indicating that a hydrogen-bonding interaction at that position is not absolutely essential for activity.

Table 4. Enzymatic Activity of 7-Substituted Analogues.

			PI3K pIC50
compd	R4	R1	α	β	δ	γ
4a	OH	Me	6.5	9.4	7.6	6.5
7a	OMe	H	<5.3	6.0	<6.2	<5.3
7b	Cl	H	<5.3	6.5	<6.2	<5.3
7ca	H	Me	<5.8	6.1	<6.3	5.7
7d	CO2H	Me	5.8	8.3	7.4	<5.2
7e	CONH2	Me	<5.2	7.6	<6.2	<5.2
7f	CN	Me	<6.0	<5.9	<6.6	<5.5
7g	NH2	Me	<5.7	7.4	6.8	<5.8
7h	NHMe	Me	<5.8	<5.5	<6.2	<5.2
7ia	NHOH	Me	<5.8	7.4	6.6	<5.2
7ja	NHCH2CH2OH	Me	<5.8	5.9	<6.2	5.3
7k	NHCOCH3	Me	<5.7	<5.9	<6.2	<5.3
7l	NHCONH2	Me	<5.2	7.7	<6.2	<5.2

^a5-THP analogue.

The compounds were also evaluated for cellular activity in two human breast cancer cell lines: a PTEN-null cell line, MDA-MB-468, and a PTEN wild-type cell line, HCC1954. The MDA-MB-468 cells are expected to be sensitive to PI3Kβ inhibition, while the HCC1954 cells should be much less sensitive and serve as a control cell line. Inhibition of proliferation was measured in an anchorage-independent (soft agar) growth assay,¹⁶ and functional activity was determined by measuring inhibition of the PI3K-dependent phosphorylation of AKT at Ser⁴⁷³.^16,17

The cellular activity of selected compounds is shown in Table 5. Both the functional and the antiproliferative activities were unexpectedly poor for this entire class of compounds. Despite a wide variety of structural changes, there did not seem to be any SAR for cellular activity. Both polar (6g,h) and nonpolar substituents (6a,c) at position 2 of the pyrazolopyridine moiety gave similar activities. Substitution of the hinge-binding morpholine with either tetrahydropyran (5a) or pyridine (5e) did not improve cell potency, and replacement of the 7-hydroxy group with polar (7e,l), acidic (7d), or basic substituents (7g) likewise had little effect. The poor cellular activity of the pyrazolopyrimidines is in sharp contrast to the structurally similar imidazopyrimidone (1) and triazolopyrimidone (2) inhibitors, where compounds of similar enzymatic potency demonstrated nanomolar potency in both cellular assays.

Table 5. Cellular Activity of Selected Pyrazolopyrimidines^a.

	pEC50
	MDA-MB-468		HCC1954
compd	growth	pAKT	growth	pAKT
4a	5.3	5.8	4.8	<4.8
4b	5.4	6.0	4.5	<4.5
5a	4.8	5.0	4.6	<4.5
5d	5.5	6.0	4.8	<4.5
5e	4.8	5.5	<4.5	<4.5
6a	ND	4.7	ND	<4.5
6c	<4.5	5.3	<4.5	<4.5
6g	4.6	4.9	<4.5	<4.5
6h	5.0	5.7	<4.5	<4.5
7d	5.5	6.0	<4.5	<4.5
7e	5.7	<4.5	<4.5	<4.5
7g	4.9	6.0	4.9	<4.5
7l	5.9	<4.5	4.9	<4.5

^aND, not done.

To investigate the possible reasons for this lack of cellular potency, we selected one of the most potent analogues, 4a, for further studies. It has reasonably good aqueous solubility (134 uM) and demonstrated excellent cellular permeability in MDCK cells (753 nM/s). In a high-throughput HPLC-based protein binding assay,²⁴ the estimated protein binding was indistinguishable from that of representative imidazopyrimidones and triazolopyrimidones that displayed excellent cellular activity. However, when we evaluated protein binding by equilibrium dialysis, we found that 4a was extremely highly protein bound, 99.9% or greater in rat, mouse, and human serum. The very small free fraction of drug would be consistent with both the poor cellular activity and the lack of apparent SAR. To test the hypothesis that high protein binding was responsible for the poor cellular activity, compounds were evaluated in the phospho-AKT functional assay in MDA-MB-468 cells in the presence and absence of serum.²⁵ The results are shown in Table 6. We were delighted to see that all of the compounds tested demonstrated significantly improved cellular activity when assayed in the absence of serum. Compound 8 is a potent PI3Kβ inhibitor,²⁶ which has significantly lower protein binding by equilibrium dialysis (rat, 74.6%; mouse, 88.2%; and human, 86.3%) and is included as a control to demonstrate that the presence or absence of serum has very little effect on the cellular activity of a compound that is not highly protein bound.

Table 6. Cellular Activity in MDA-MB-468 Cells in the Presence and Absence of Serum.

		pAKT pEC50
compd	PI3Kβ pIC50	+ serum	– serum
4a	9.4	5.8	7.5
5d	8.3	6.0	7.2
5e	8.6	5.5	7.6
6g	8.9	4.9	6.8
8(23)	9.2	7.8	8.2

Compound 4a was also evaluated for pharmacokinetic properties, and we were pleased to see that it exhibited low clearance in rat, mouse, and monkey and good oral bioavailability and exposure in mouse, as shown in Table 7.²⁷ This improvement in clearance over the imidazopyrimidone and triazolopyrimidone series is undoubtedly due in large part to the high protein binding, and we were therefore unable to determine the contribution of the N-benzyl group to in vivo clearance. While high protein binding does not necessarily correlate with poor in vivo efficacy,²⁸ we desired a compound with an unambiguously higher free fraction for in vivo target validation studies, and 4a was not pursued further.

Table 7. DMPK Profile of 4a.

species	Cl (mL/min/kg)	DNAUC (po)a (ng h/mL/mg/kg)	% F
rat	6.5	ND	ND
mouse	3.7	2321	51
monkey	0.33	ND	ND

^a1% DMSO, 10% encapsin, pH 4.5; ND, not determined.

In summary, we have identified novel pyrazolopyrimidines that are highly potent and selective inhibitors of PI3Kβ. This template allowed facile substitution at both the 5- (hinge-binding) and the 7- (back-pocket) positions, and we were able to demonstrate considerable tolerance to substitution at the 7-position, which is thought to make a key hydrogen bond interaction. Because the SAR has generally been conserved among related series, this may afford the opportunity to fine-tune physicochemical properties in other inhibitor series without sacrificing enzymatic potency. The compounds exhibit only modest cellular potency under standard conditions, which we have shown to be due to high protein binding. Compound 4a displays low clearance and high oral exposure in mouse, but the high protein binding precluded further development.

Supporting Information Available

Synthetic procedures and characterization for all compounds. This material is available free of charge via the Internet at http://pubs.acs.org.

The authors declare the following competing financial interest(s): we have been requested to add the following disclosure:The authors are all employees of GlaxoSmithKline and have a financial interest in GlaxoSmithKline.