Discovery of CDZ173 (Leniolisib), Representing a Structurally Novel Class of PI3K Delta-Selective Inhibitors

Abstract

The predominant expression of phosphoinositide 3-kinase δ (PI3Kδ) in leukocytes and its critical role in B and T cell functions led to the hypothesis that selective inhibitors of this isoform would have potential as therapeutics for the treatment of allergic and inflammatory disease. Targeting specifically PI3Kδ should avoid potential side effects associated with the ubiquitously expressed PI3Kα and β isoforms. We disclose how morphing the heterocyclic core of previously discovered 4,6-diaryl quinazolines to a significantly less lipophilic 5,6,7,8-tetrahydropyrido[4,3-d]pyrimidine, followed by replacement of one of the phenyl groups with a pyrrolidine-3-amine, led to a compound series with an optimal on-target profile and good ADME properties. A final lipophilicity adjustment led to the discovery of CDZ173 (leniolisib), a potent PI3Kδ selective inhibitor with suitable properties and efficacy for clinical development as an anti-inflammatory therapeutic. In vitro, CDZ173 inhibits a large spectrum of immune cell functions, as demonstrated in B and T cells, neutrophils, monocytes, basophils, plasmocytoid dendritic cells, and mast cells. In vivo, CDZ173 inhibits B cell activation in rats and monkeys in a concentration- and time-dependent manner. After prophylactic or therapeutic dosing, CDZ173 potently inhibited antigen-specific antibody production and reduced disease symptoms in a rat collagen-induced arthritis model. Structurally, CDZ173 differs significantly from the first generation of PI3Kδ and PI3Kγδ-selective clinical compounds. Therefore, CDZ173 could differentiate by a more favorable safety profile. CDZ173 is currently in clinical studies in patients suffering from primary Sjögren’s syndrome and in APDS/PASLI, a disease caused by gain-of-function mutations of PI3Kδ.

Phosphoinositide-3-kinase δ (PI3Kδ) is a lipid kinase composed of an enzymatic p110δ subunit and a regulatory p85 subunit expressed predominantly in hematopoietic cells. PI3Kδ catalyzes the intracellular conversion of phosphatidylinositol 4,5-bisphosphate (PIP2) to phosphatidylinositol 3,4,5-trisphosphate (PIP3) downstream of a number of receptor tyrosine kinases (e.g., BCR, TCR, FcεR1)² as well as some G protein-coupled receptors (GPCR; e.g., CXCR5).³

Inhibition of PI3Kδ has been shown to be beneficial for the treatment of hematological malignancies where the PI3K/AKT pathway is hyperactive.⁵ The importance of PI3Kδ for the function of cells involved in inflammatory respiratory diseases (T cells, mast cells, neutrophils)^6,7 has guided the development of inhaled and oral PI3Kδ inhibitors for the treatment of asthma or chronic obstructive pulmonary disease.⁹⁻¹¹ Moreover, PI3Kδ is overactive in patients with autoimmune diseases such as rheumatoid arthritis¹² and systemic lupus erythematosus,¹⁴ where PI3Kδ inhibitors have been effective in respective murine models.^15,16 In patients with activated PI3Kδ syndrome (APDS), mutations in the PIK3CD gene encoding p110δ lead to an immunodeficiency characterized by recurrent infections and prominent lymphoproliferation.³⁷ In summary, there is strong evidence that PI3Kδ inhibition is a promising therapeutic principle for the treatment of APDS as well as inflammatory and autoimmune diseases,^19,20 prompting numerous companies to engage in programs geared toward the identification of potent and selective PI3Kδ inhibitors.^11,22−29

We have recently disclosed the optimization of novel 4,6-diaryl quinazolines as PI3Kδ-selective inhibitors.³⁰ These efforts led to discovery of quinazoline 1a as a potent and isoform selective inhibitor of PI3Kδ (Table 1). We could demonstrate that compound 1a is effective in down-regulating PI3Kδ-mediated signaling in vitro and in vivo, ultimately resulting in dose-dependent efficacy in a mechanistic rodent model after oral dosing. However, when the amorphous material of compound 1a was exposed to an aqueous environment, a crystalline trihydrate polymorph formed over time. This trihydrate had low aqueous solubility, and we hypothesized that solubility-limited absorption was the reason for the underproportional exposure increase, which we observed with compound 1a in higher dose rat PK studies.³⁰ When we directly compared trihydrate vs amorphous material in rat PK studies, we indeed observed a significant drop in exposure (ca. 5-fold, data not shown). A polymorph with poor aqueous solubility that is formed under an aqueous environment poses a serious development hurdle as inconsistent or erratic exposure levels become more likely. This not only impacts preclinical efficacy or toxicological studies, but also poses a risk in later clinical development (safety risk if exposure is too high, lacking efficacy risk if too low). It became the project goal to identify a replacement for compound 1a with an improved solubility of crystalline material.

Table 1. Influence of 5,6,7,8-Tetrahydropyrido[4,3-d]pyrimidine Core Introduction on Biochemical/Cellular Potency, PI3K Isoform Selectivity, and in Vitro ADME Parameters.

^aMean of a minimum of two independent experiments; standard deviation for pIC₅₀ values < 0.3.

^bKGlo format.

^cADAPTA format (values comparable to KGlo).

^dInhibition of pAkt formation in Rat-1 cells.

^eInhibition of anti-IgM induced mCD86 expression on mouse splenocytes.

^fInhibition of anti-IgM/IL-4 induced rCD86 expression in 50% rat blood.

^gEffective permeability [10^–6 cm·s^–1] (pH = 6.8), n = 1.

^hHigh-throughput log P measurement with immobilized artificial membranes, n = 1.

ⁱIntrinsic clearance in incubations of rat liver microsomes [μL·min^–1·mg^–1 protein], n = 1.

^jHigh-throughput equilibrium solubility determination [mM] (pH = 6.8); nd = not determined.

^kCDZ173 (leniolisib).

We recently disclosed that the 4-aryl spacer could be substituted by a pyrrolidineoxy substituent, which led to an improved membrane permeability profile and increased ligand efficiency.³¹ When the core quinazoline was changed to other heterocycles that still contained the hinge binding nitrogen, namely, quinolines, 1,5-naphthyridines, and cinnolines, we observed that a substantial modulation of the lipophilicity was the consequence. Not every derivative, including the compounds that are presented in this manuscript, showed a tight correlation of log D with other parameters that are known to be modulated by lipophilicity, such as solubility, membrane permeability, and microsomal stability. Nevertheless, we noticed some consistent impact by lipophilicity changes, which we outline in the following section.

As part of our core scaffold modifications we extended our investigations to partially saturated bicyclic systems, and in this context, we discovered 5,6,7,8-tetrahydropyrido[4,3-d]pyrimidine (THPP) as a suitable core that decreased lipophilicity significantly. This is illustrated by comparing quinazoline 1b (HT-logP = 2.7) to the similarly substituted THPP analogue 2a (HT-logP = 0.7); the lipophilicity drop of two orders of magnitude resulted in a marked decrease of membrane permeability (PAMPA log Pe = −4.9 × 10^–6 → 6.7 × 10^–6 cm·s^–1 at pH = 6.8) and a loss in cellular activity (PI3Kδ IC₅₀ = 2.63 μM). Replacing one of the amide bonds by an ether bridge (2a → 2b) increased lipophilicity (HT-logP = 2.2), reduced topological polar surface area (tPSA= 91 → 80 Ų) and improved membrane permeability, and the biochemical activity translates better into cellular assays. When we introduced the pyrrolidineoxy-substituents on the THPP core we found that the overall biochemical and cellular PI3Kδ activities for our first analogues such as 3a were moderate. At the same time, the lipophilicity space appeared highly promising, and ADME properties such as membrane permeability, solubility, and stability in liver microsomes were favorable. By SAR expansion, we learned that introducing a CF₃ group in the 3-position of the methoxypyridine (3a → 3b) increased PI3Kδ potency by a factor of 5. This modification was accompanied by a lipophilicity increase by almost one order of magnitude, and solubility and microsomal stability were negatively impacted. For oxygen-linked derivatives (3a–d), the optimal balance was found combining a methyl group in the 3-position of the pyridine with a tetrahydropyran group as R² substituent (compound 3c). Similar to the pair 3a/3b, an introduction of a CF₃ group (3c → 3d) led to an increased in vitro clearance and reduced solubility.

When we switched the ether linker to an amine (3e–h), two ADME parameters were markedly affected: reduced in vitro clearance and reduced PAMPA permeability (compare compounds 3b vs 3h, 3c vs 3e, and 3d vs 3f). In general, amine-linked analogues were also chemically more stable toward acid catalyzed hydrolysis compared to their oxygen-linked counterparts. In contrast to the oxygen-linked derivatives, the optimal overall property space for amine-linked analogues was achieved by a more lipophilic R¹/R² substitution pattern, such as that shown for propionamide derivatives 3g and 3h. Compound 3h was found to have the optimal profile, hydrophilic enough to ensure good solubility and metabolic stability, yet not too polar to allow for favorable membrane permeability. This compound was selected for further profiling and later became our clinical candidate now known as CDZ173 (leniolisib).

Compared to compound 1a, CDZ173 is similar with respect to the overall on-target/selectivity profile, moderate rat liver microsomal turnover, and solubility of amorphous material. In contrast to compound 1a, CDZ173 shows good PAMPA permeability, enabling favorable absorption as we will discuss later. In addition, the solubility of crystalline CDZ173 (free base) was sufficient to support all pharmacological and early toxicological experiments. At a later stage, a crystalline phosphate salt with an >500-fold solubility improvement was identified in a salt screen (see Supplementary Table 1) and chosen as the clinical service form in order to reduce the risk for exposure variability and/or food effects.

At the time when the optimization work was conducted, we did not have any X-ray cocrystal structures for PI3Kδ selective ligands. As a surrogate, we used a homology model based on PI3Kγ. As this model was not able to explain the reasons for the observed selectivity, we optimized PI3K activity empirically using biochemical and cellular assay results.³¹ The first structures of selective ligands complexed with PI3Kδ became available only after CDZ173 had been identified. It was only at this point that a solid selectivity rational on the molecular level could be developed (Figure 1). A comparison of the structures of quinazoline 1 bound to PI3Kα and PI3Kδ revealed differences in the binding modes of the inhibitor to the ATP sites of both isoforms.³⁰ Whereas the acetyl-piperazine group of the ligand stacks to the side chain of W760 in PI3Kδ, the corresponding interaction to W780 in PI3Kα is prevented by R770 in PI3Kα, and the ligand is bound in an extended conformation to this isoform (Figure 1, left panel). This loss of van der Waals interactions between the inhibitor and W780 in PI3Kα correlates with considerable loss in potency of the inhibitor for PI3Kα compared to PI3Kδ. We hypothesized that maintaining the stacking interaction with W760 (PI3Kδ) while preventing the ligand from adopting an extended conformation would offer a design principle for PI3Kδ inhibitors with exquisite selectivity over PI3Kα. To test this hypothesis, we obtained the cocrystal structure of CDZ173 bound to the ATP site of PI3Kδ (Figure 1, right panel). This structure reveals that the propionamide group of CDZ173 stacks with W760 in a similar way to the acetyl-piperazine group of compound 1a. This binding mode of CDZ173 confirms our docking results³¹ and strengthens our hypothesis that the (S)-enantiomer is preferred regarding PI3Kδ potency and isoform selectivity. Supplementary Figure 10 shows that the binding modes observed for compounds 1a and CDZ173 clearly differ from previously discovered PI3Kδ inhibitors. “Propeller-shaped” inhibitors such as GS-643624 bind to a pocket that is induced by insertion of the quinazoline group of GS-643624 between W760 (PI3Kδ) and M752 (PI3Kδ) (“WM-pocket”).³² In the cocrystal structures of compounds 1 and CDZ173, M752 (PI3Kδ) occupies the area where the quinazoline of GS-643624 is bound. The WM-pocket is not induced for our inhibitor series.

Figure 1.

(Left) Structural alignment of PI3Kα and PI3Kδ with compound 1a bound to the ATP sites. The protein part of PI3Kδ/compound 1a (PDB 5IS5)³⁰ is depicted as a gray surface. Amino acid side chains T750, M752, W760, and the inhibitor are represented as blue sticks for PI3Kδ. The corresponding amino acid and inhibitor structures in PI3Kα (PDB 5ITD)³⁰ are shown in yellow. (Right) Comparison of the binding modes of CDZ173 (green sticks) and compound 1a (blue sticks) to the active site of PI3Kδ (gray surface with amino acids T750, M752, and W760 colored in red, yellow, and magenta, respectively.

As outlined in Table 1, the inhibitory effect on PI3Kα, -β, and -δ isoforms on the cellular level was routinely assessed in Rat-1 cells transfected with human myr-p110 PI3K isoforms using the PI3K-dependent phosphorylation of AKT (pAKT) as a readout. The PI3Kδ activity of CDZ173 in this artificial assay system translated well into rodent B-cell activation assays (murine splenocytes and rat whole blood), indicating sufficient fraction unbound and minimal unspecific binding (Table 1). CDZ173 potently inhibits formation of proximal pathway markers (pAKT), and also, anti-IgM ± IL-4-induced expression of distal markers of B cell activation (CD69) and costimulation (CD86) across species using whole blood or isolated cells (Table 2). In mouse, anti-IgM-induced B cell proliferation is also potently blocked. These results confirm and extend published findings with PI3Kδ deficient mice and PI3Kδ selective tool compounds.³³ The effects of CDZ173 on T cell responses are shown in Table 2. The mixed lymphocyte reaction (MLR) is a model for allogeneic T cell activation in vitro. The MLR with both human peripheral blood mononuclear cells (PBMCs) and mouse splenocytes is potently inhibited by CDZ173. The aCD3-stimulated proliferation of T cells from human PBMC is also inhibited, and the reduced dependency on PI3Kδ in the presence of aCD28 costimulation could be confirmed.³³ The differentiation of murine and human aCD3/aCD28 mAb-stimulated T cells into Th2 or Th17 cells is significantly blocked. In summary, CDZ173 potently blocks PI3Kδ-dependent signaling in B and T cells across species and matrices.

Table 2. Cellular Inhibition of PI3Kδ-Dependent B and T Cell Functions in Different Species by CDZ173.

	B cellsa								T cells
species	humanb			cynob	ratc		moused		humane					moused
stimulus	aIgM	aIgM/IL-4		aIgM	aIgM/IL-4		aIgM		Allo MLRf	aCD3g	aCD3/aCD28f			Allo MLRf	aCD3/aCD28f
readout	pAKT	CD86	CD69	pAKT	pAKT	CD86	Prol.g	CD86	Prol.g	Prol.g	Th2	Th17	Prol.g	Prol.g	Th2	Th17
IC50 [μM]	0.144	0.202	0.193	0.084	0.150	0.099	0.007	0.048	0.079	0.159	0.095	0.073	1.15	0.033	0.010	0.101

^aData from at least two independent experiments are shown; standard deviation for pIC₅₀ values <0.3.

^bMeasured in 90% whole blood.

^cMeasured in 50% whole blood.

^dMeasured in splenocytes.

^eMeasured in PBMCs.

^fMean of at least four independent experiments.

^gProliferation (Prol.) measured as ³H-Thymidine uptake.

CDZ173 shows a 30-fold cellular selectivity over PI3Kα (Table 1, cellular assay section in Supporting Information), which translates to a preservation of insulin homeostasis in mice up to a dose as high as 100 mg/kg bid (Supplementary Figure 9). While the cellular isoform selectivity against PI3Kα and -β was determined in Rat-1 cells as already described, PI3Kγ activity has to be tested differently due to its distinct signaling downstream of GPCR. We measured the PI3Kγ-dependent activation of U937 monocytes by the GPCR-ligand MIP-1α, and CDZ173 showed no activity up to the highest test concentration (IC₅₀ > 7.4 μM). In addition, the cellular activity on PI3Kγ was assessed via the activation of mouse bone marrow-derived mast cells (BMMCs) stimulated with adenosine where CDZ173 was a very weak inhibitor (IC₅₀ = 7.8 μM). When tested in biochemical assays against the Class III PI3K Vps34 and enzymes of the wider PI3K family (mTOR and PI4K), CDZ173 was inactive up to the highest test concentration (IC₅₀ > 9.1 μM). While DNA-PK was moderately inhibited by CDZ173 (biochemical IC₅₀ = 0.88 μM), this activity did not translate to an inhibition of p53 in a stressed p53RE-bla HCT116 reporter cell line further down in the pathway of DNA repair (IC₅₀ > 3 μM).

CDZ173 showed no activity up to the highest test concentration when tested against CYP isoform assays (3A3, 2D9, 2D6, 2C9),³⁵ a panel of ion channels (including hERG)³⁶ and a protease panel. In a panel of 50 safety related targets (GPCRs, ion channels, transporters), CDZ173 only showed measurable activity for hPDE4D (IC₅₀ = 4.7 μM) and 5HT2B (IC₅₀ = 7.7 μM). Furthermore, CDZ173 was tested against a panel of 442 diverse serine/threonine, tyrosine, and lipid kinases utilizing DiscoverX′ KINOMEscan technology at 10 μM. The only kinase that was inhibited by CDZ173 (in addition to the expected Class I PI3Ks) was RPS6KA5 (76% inhibition, no cellular follow-up as no negative impact on safety/tolerability was anticipated).

The pharmacokinetic profile of CDZ173 is shown in Table 3. In general, the compound is absorbed very quickly across species as can be seen by an early T_max of the oral profiles. Whereas clearance is low to moderate in rats and monkeys, it was found that clearance in dogs is very low resulting in a very high exposure in blood. As dog plasma protein binding is very high (>99%) and restricting distribution of the compound into tissue (Vss = 0.3 L·kg^–1), dogs were excluded as a potential species for toxicological studies. In contrast to compound 1a,³⁰ CDZ173 did showed proportional exposure increase comparing 3 and 30 mg/kg doses in rats.

Table 3. PK Parameters^a for CDZ173 (Crystalline Free Base).

species	ratb	ratb	ratb	dogc	dogc	monkeyd	monkeyd
dose [mg·kg–1]	1 (iv)	3 (po)	30 (po)	0.1 (iv)	0.3 (po)	0.1 (iv)	0.3 (po)
CL [mL·min–1·kg–1]	28 (3)			2 (1)		17 (2)
VSS [L·kg–1]	3.1 (0.8)			0.3 (0.1)		0.6 (0.1)
t1/2 term [h]	1.8 (0.3)			1.5 (0.4)		0.7 (0.4)
AUC d.n.e [nM·h]	1355 (158)	1165 (27)	2107 (655)	20120 (7720)	11690 (6277)	2250 (290)	297 (107)
BAV [%]		86 (2)	>100		58 (31)		13 (5)
Cmax d.n.e [nM]		391 (161)	334 (105)		3410 (1347)		213 (180)
Tmax [h]		0.4 (0.1)	<0.25		0.7 (0.3)		<0.5

^aMean values with standard deviation in brackets.

^bFemale, n = 4.

^cMale, n = 3.

^dOne female, two male.

^ed.n. = dose normalized to 1 mg·kg^–1.

Because of its favorable pharmacokinetic profile in rodents and monkeys, we tested CDZ173 in various in vivo settings, starting with PK/PD experiments. In rats, CDZ173 inhibits ex vivo anti-IgM/IL-4 -induced proximal pAKT formation in rats in a concentration- and time-dependent manner (see Supplementary Figure 1). Plotting total blood exposure against the PD effect, an EC₅₀ = 83 nM (EC₅₀ free = 5 nM, rat PPB = 94%) was calculated, which is in agreement with the in vitro rat whole blood B cell activation assay result (IC₅₀ = 99 nM; Table 1). In cynomolgus monkeys, the ex vivo activation of CD20+ B cells (anti-IgM/IL-4 stimulated pAKT) was inhibited with an EC₅₀ of ca. 140 nM (see Supplementary Figure 2). As for rat, the exposure/concentration–response profiles are in agreement with the in vitro whole blood assay result (IC₅₀ = 84 nM, Table 2). The concentration-dependent PD effect of CDZ173 translated into a dose-dependent pharmacological inhibition of PI3Kδ as demonstrated by inhibition of the generation of antigen-specific antibody responses to sheep red blood cells (SRBC) in rat (Figure 2). Encouraged by these results, CDZ173 was tested in a rat model of collagen-induced arthritis (rCIA) to assess the correlation of autoantibody levels with reduction of clinical symptoms and histological parameters. CDZ173 significantly inhibited pathogenic antirat collagen antibodies, paw swelling, inflammatory cell infiltration, proteoglycan loss, and joint erosion when dosing started before disease onset. Full efficacy was seen with doses as low as 3 mg/kg bid (Supplementary Figures 3–6). In a therapeutic setting, where CDZ173 was administered when significant paw swelling was already evident, a dose of 10 mg/kg bid significantly ameliorated disease parameters and strongly reduced the degree of established autoantibody responses (Figure 3 and Supplementary Figure 7). The PK profile and the dose normalized AUC observed in the multiple dose experiments was comparable to data obtained from single dose PK studies in naive animals. In an ozone-induced lung inflammation model in mice, CDZ173 dose-dependently inhibited the increase in bronchoalveolar lavage (BAL) neutrophil and macrophage numbers with ED₅₀ values of 16 mg/kg and 40 mg/kg, respectively (Supplementary Figure 8). There was no effect on the ozone-induced plasma leakage, as measured by the increased BAL serum albumin levels (data not shown). In all presented preclinical multidose pharmacology studies, CDZ173 was well tolerated, and there were no apparent signs of toxicity.

Figure 2.

Dose-dependent inhibition of SRBC-specific IgM response by CDZ173 (error bars indicate SD); ***p < 0.001 compared to vehicle (student’s t test, two-tailed distribution).

Figure 3.

Inhibition of paw swelling in a therapeutic rat CIA experiment (dosing after disease onset, error bars indicate SEM). Antigen priming (I°) and boost (2°) indicated with arrows. Treatment period indicated with horizontal line at bottom of graph. *p < 0.01.

Taken together, the in vitro properties and the results in preclinical models of autoimmune and inflammatory diseases demonstrate the potential of CDZ173 to act as a therapeutic for immune-mediated diseases. Because of the big structural differences to the first-generation of clinical PI3Kδ and/or PI3Kγδ-selective inhibitors, CDZ173 could favorably differentiate in terms of its safety profile. After supportive preclinical toxicology studies and a successful phase I clinical campaign, CDZ173 (leniolisib) is currently being explored in clinical trials for APDS (ClinicalTrials.gov identifier: NCT02435173) and primary Sjögren’s Syndrome (NCT02775916).

Glossary

ABBREVIATIONS

APDS

activated PI3Kδ syndrome

BAL

bronchoalveolar lavage

BAV

absolute oral bioavailability

BMMC

bone marrow-derived mast cells

CIA

collagen-induced arthritis

CYP

cytochrome P450

GPCR

G protein-coupled receptor

hERG

human ether-à-go-go-related gene

MLR

mixed lymphocyte reaction

THPP

5,6,7,8-tetrahydropyrido[4,3-d]pyrimidine

mTOR

mechanistic Target Of Rapamycin

PAMPA

parallel artificial membrane permeability assay

PASLI

p110 delta activating mutation causing senescent T cells, lymphadenopathy, and immunodeficiency

PBMC

peripheral blood mononuclear cells

PI3K

phosphoinositide-3-kinase

SAR

structure–activity relationship

SRBC

sheep red blood cell(s)

tPSA

topological polar surface area

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsmedchemlett.7b00293.

• Full descriptions of all biological assays and in vivo studies. Characterization of all compounds. Full experimental procedures for the sequence leading to CDZ173. Crystallographic data collection and refinement statistics for crystal structures (PDF)

Accession Codes

PDB code for the X-ray crystal structure described in this study has been deposited in the Protein Data Bank under the accession code 5O83.

The authors declare no competing financial interest.