Optimization of PDE3A Modulators for SLFN12-Dependent Cancer Cell Killing

Abstract

6-(4-(Diethylamino)-3-nitrophenyl)-5-methyl-4,5-dihydropyridazin-3(2H)-one, or DNMDP, potently and selectively inhibits phosphodiesterases 3A and 3B (PDE3A and PDE3B) and kills cancer cells by inducing PDE3A/B interactions with SFLN12. The structure–activity relationship (SAR) of DNMDP analogs was evaluated using a phenotypic viability assay, resulting in several compounds with suitable pharmacokinetic properties for in vivo analysis. One of these compounds, BRD9500, was active in an SK-MEL-3 xenograft model of cancer.

We previously reported on the results of a differential viability screen, where a p53 mutant lung cancer cell line, NCI-H1734, was killed by 6-(4-(diethylamino)-3-nitrophenyl)-5-methyl-4,5-dihydropyridazin-3(2H)-one, or DNMDP, while a second lung cancer cell line, A549, was not affected.¹ Further screening detected selective cell killing in 22 out of 766 cancer cell lines, correlating with elevated expression of phosphodiesterase 3A (PDE3A). While DNMDP selectively inhibited PDE3, most PDE3 inhibitors had no cell killing effects and in fact rescued cancer cells from DNMDP-induced death. PDE3A immunoprecipitation experiments from HeLa cell lysates showed PDE3A bound to Schlafen family member 12² (SFLN12) in the presence of DNMDP, but not in the presence of trequinsin, a PDE3 inhibitor that does not kill cancer cells. DNMDP-sensitive cell lines were found to express elevated levels of both PDE3A and SLFN12.¹ Since our report, others have discovered that selected PDE3 inhibitors, some of which are known to phenocopy DNMDP, kill a gastrointestinal stromal tumor (GIST) cell line³ and a subset of primary ovarian cancer cells.⁴

Although DNMDP is a very potent and highly selective compound in a cellular cytotoxicity assay, it has structural liabilities making it unsuitable for further development: a dialkylanilino group prone to metabolic instability and a potentially reactive nitro group. We sought to discover analogs that maintained or improved cellular activity while improving pharmacokinetic properties. Furthermore, we needed potent and selective compounds to better examine the relationship between PDE3 and SFLN12, a protein of unknown function, in sensitive cancer cells. Without sufficient quantities of SLFN12 to study in detail, our initial SAR was driven using phenotypic screening, i.e., viability assays.

The more active (R)-enantiomer of DNMDP had previously been synthesized¹ from (R)-6-(4-aminophenyl)-5-methyl-4,5-dihydropyridazin-3(2H)-one (1), a commercially available starting material used for the synthesis of the inotropic drug levosimenden.⁵ Compound 1 was acetylated, then nitrated followed by acetyl hydrolysis. Reductive amination with acetaldehyde produced both the mono- and dialkylated amine, (R)-DNMDP, as a 19:1 ratio of enantiomers as determined by chiral SFC analysis. DNMDP was N-alkylated producing racemic compound 6 (Scheme 1).

Scheme 1. Synthesis of (R)-DNMDP and Analogs.

Reagents and conditions: (a) Ac₂O (91%); (b) 90% HNO₃/H₂SO₄ (19%); (c) NaOH/H₂O/MeOH (quant.); (d) CH₃CHO, NaBH(OAc)₃/DCM (7% combined); (e) NaH, EtI/DMF (61%).

We were concerned initially that the nitro group of DNMDP might lead to promiscuous protein binding. Reductive amination of 1 with acetaldehyde under standard conditions gave des-nitro (R)-DNMDP (7) (Scheme 2). The reaction was performed using racemic 1 as well. Modifying reaction conditions allowed isolation of monoethyl analog 8. Compound 1 was alkylated to make the morpholino analog 9, which was chlorinated adjacent to the morpholine ring (10). Heterocycle 11 was prepared by condensation of 1 with diformyl hydrazine.

Scheme 2. Synthesis of (R)-des-Nitro DNMDP Analogs.

Reagents and conditions: (a) 7 - CH₃CHO, NaBH₃CN/MeOH (82%); 8 - CH₃CHO, then NaBH₃CN/MeOH (8%); (b) (BrCH₂CH₂)₂O, K₂CO₃/DMF (46%); (c) NaOCl/HOAc (40%); (d) CHONHNHCHO (73%).

Diazotization of 1 followed by halogenation produced the fluoro (12), chloro (13), and iodo (14) derivatives, respectively (Scheme 3). Compound 13 (racemic) is a known PDE3 inhibitor.⁶ The iodide 14 reacted at high temperature with N-methyl piperazine to produce 15. Suzuki coupling with 14 followed by hydrogenation gave analogs with cyclic, acyclic, sp², and sp³ carbon substituents on the para position of the phenyl ring (16–19). Heterocycle 20 was prepared similarly.

Scheme 3. Synthesis of (R)-DNMDP Analogs.

Reagents and conditions: (Top) (a) NaNO₂, HCl/water, then NaBF₄ (12, 15%) or CuCl₂ (13, 77%) or KI (14, 49%); (b) N-methyl piperazine, NMP, 160 °C microwave (20%). (Bottom) (a) Pd(PPh₃)₄, Na₂CO₃/H₂O/THF, propene (16, 90%) or cyclohexene (17, 90%); (b) H₂, 10% Pd/C, MeOH (60% 18, 63% 19); (c) 5-pyrimidine boronic acid, Pd(PPh₃)₄, Na₂CO₃/H₂O/dioxane (68%)

Substituent effects on the phenyl ring were examined. The nitro group of DNMDP was reduced, and the amine product (21) was acetylated (22) (Scheme 4). Reducing the nitro group of 3 gave 23, subsequent heating produced benzimidazole 24.⁷

Scheme 4. Synthesis of 3-Amino Substituted Analogs.

Reagents and conditions: (a) H₂, 10% Pd/C/MeOH (65%); (b) Ac₂O (90%); (c) H₂, 10% Pd/C/MeOH (92%); (d) toluene, reflux (54%).

Halogenation of acetamide 2 provided the chloride (25) and bromide (26), respectively (Scheme 5). The bromide underwent Sonogashira coupling with trimethylsilylacetylene generating the protected alkyne, which was deprotected with fluoride ion to give 27 or hydrolyzed directly to ketone 28.

Scheme 5. Synthesis of Acetylene and Acetyl Analogs.

Reagents and conditions: (a) Br₂, CH₂Cl₂ (15%); (b) NaOCl/HOAc (41%); (c) 28, TMSCCH, Pd₂(dba)₃, Et₃N/DMF (80%); (d) TBAF/THF (52%); (e) HCOOH/H₂O (83%).

Fluorinated analogs were prepared starting with 3,4-difluoropropiophenone, which was first alkylated with ethyl bromoacetate then condensed with hydrazine to give 29 (Scheme 6). The 4-fluoro group was displaced by refluxing morpholine to give 30. Racemic 29 and 30 were separated into enantiomers with chiral SCF chromatography. Only one enantiomer of both 29 and 30 was active against HeLa cells, and the active enantiomer of 29 was converted to the active enantiomer of 30. The absolute stereochemistry of the separated enantiomers was determined by low yielding reductive removal of the fluorine atom from 30 to give 9. Compound 9 obtained was compared to racemic and (R)-9 by chiral SCF chromatography to assign stereochemistry at the chiral center. The dichloro analog of 29 was enantioselectively synthesized via amide 25. The amine resulting from hydrolysis was diazotized and converted to dichloro analog 31 (Scheme 6).

Scheme 6. Synthesis of Dihalogenated Analogs.

Reagents and conditions: (Top) (a) LiHMDS, THF, −78 °C, then BrCH₂COOEt (30%); (b) hydrazine/EtOH (29%); (c) morpholine, 130 °C (53%). (Bottom) (a) NaOH/MeOH (98%); (b) NaNO₂/HCl then CuCl₂ dihydrate/CH₃CN, 80 °C (58%).

Isosteric replacement of chiral 6-methyl dihydropyridazinones by achiral 5,5-dimethylpyrazolones has maintained PDE3/4 activities in other systems.⁸ Synthesis of this heterocycle began with ethyl 4-nitrobenzoyl acetate, which was dimethylated, condensed with hydrazine, and reduced to produce aniline 32 (Scheme 7). Conversion of the amino group of 32 to the diethylamino (33), morpholino (34), and chloro (35) groups was done similarly to the corresponding dihydropyridazinones.

Scheme 7. Synthesis of Dimethylpyrazolones.

Reagents and conditions: (a) NaH, MeI/THF (40%); (b) hydrazine/EtOH (64%); (c) H₂, Pd/C/EtOH (94%); (d) CH₃CHO, NaBH₃CN/MeOH (56%); (e) (BrCH₂CH₂)₂O, K₂CO₃/DMF (33%); (f) NaNO₂, HCl/water, then CuCl dihydrate/CH₃CN, 80 °C (72%).

Cellular SAR analysis was performed by treating DNMDP-sensitive HeLa cells with compounds for 3 days at a dose range of 1 nM through 10 μM, with a counterscreen using DNMDP-insensitive A549 cells (Tables 1 and 2). No compounds tested against A549 cells displayed activity at concentrations up to 10 μM (data not shown)

Table 1. HeLa Cell Viability of DNMDP Analogs^a.

compound	EC50 (nM)	compound	EC50 (nM)
DNMDP	6.9	18	16
(R)-DNMDP	3.8	19	7.7
1	>1000	20	13
5	1.1	21	240
6	>1000	22	>1000
(±)-7	8.8	23	>1000
(R)-7	3.3	24	>1000
8	71	(±)-29	64
(±)-9	36	(R)-29	22
(R)-9	13	(S)-29	>1000
10	4.5	(±)-30	2.8
11	>1000	(R)-30	1.6
12	410	(S)-30	>1000
13	33	31	4.5
14	7.1	32	>1000
15	310	33	>1000
16	8.8	34	>1000
17	2.2	35	>1000

^an ≥ 3; std. error ≤ 20%.

Table 2. HeLa Cell Viability of Acylated Amines^a.

compound	X	EC50 (nM)
2	H	>1000
3	NO2	4.9
23	NH2	>1000
25	Cl	24
26	Br	21
27	CCH	65
28	Ac	130

^an ≥ 3; std. error ≤ 20%.

Several active compounds, (R)-7, (R)-30, and 31, were tested in HeLa cells at effective doses and treated with varying levels of trequinsin (PDE IC₅₀ 0.25 nM¹⁰). As was seen previously with DNMDP,¹ dose-dependent cell rescue was observed indicating a similar mechanism of action.

We previously reported that the (R)-enantiomer of DNMDP is 200–500 times more active, depending upon the cell line tested, than the (S)-enantiomer.¹ This stereochemical dependence was seen in four new examples, where the (R)-enantiomer was more active than the (S)-enantiomer or the racemate (7, 9, 29, 30).

Removing one ethyl group from DNMDP (5) had little effect on potency, while N-ethylating the dihydropyridazinone ring (6) diminished activity. Removing the nitro group from DNMDP did not have a deleterious effect (7). Reduction of the nitro group of DNMDP (21) reduced activity and acylation of the resulting amine (22) removed activity.

The diethylamino group of (R)-des-nitro DNMDP (7) could be replaced by a morpholine ring (9). The N-methyl-piperazine analog (15) displayed much less activity. The presence of a fluorine atom adjacent to the morpholine nitrogen (30), as found in the antibiotic linezolid,¹¹ improved activity in the racemic and (R)-examples, while the (S)-enantiomer was still inactive. Substitution of fluorine with chlorine maintained activity (10). Replacement of the diethylamino group with heterocycles on (R)-des-nitro DNMDP (7) led to both active (20) and inactive (24, 11) compounds.

Replacing the diethylamino group of des-nitro DNMDP (7) with halogens chloro (13) and iodo (14) was permitted, while the fluoro analog was significantly less active (12). As was the case with a 4-morpholino group, adding an additional halogen at the 3-position improved activity (29/12, 31/13). Substitution with carbon substituents (16–19) retained cellular activity.

The dimethylpyrrolidinones (32–35) were inactive.

Intermediate acetamides were tested (Table 2). Contrary to the case with the diethylamino analogs (7 and DNMDP), the des-nitro analog 2 had no activity, whereas the nitrated analog 3 was very potent. Electron withdrawing substituents improved activity, but an amino group was not tolerated (23).

We extended our cellular testing beyond HeLa cells to the DNMDP-sensitive NCI-H2122 (lung cancer) and COLO741 (melanoma) cell lines, as well as the DNMDP-insensitive HCT116 (colorectal cancer) and IMR90 (normal lung fibroblast) cell lines. The results mirror those using HeLa and A549 cells; active compounds are only slightly less active against H2122 and COLO741 cell lines, and all the compounds are inactive against the HCT116 and IMR-90 cells (Supporting Information Table 1).

While the mechanism of action of our compounds involves inducing PDE3/SLFN12 interaction, we nonetheless tested some of our compounds for biochemical inhibition of PDE3A and PDE3B (Table 3).¹² No major selectivity for one isozyme over the other was observed, and the HeLa activity generally mirrored the biochemical results. Halogenation of the phenyl ring (9 to (R)-30) increased PDE3A/B biochemical inhibition 10-fold. The cellular SAR generally agreed with reported SAR of PDE3 inhibition¹³ with the (R)-enantiomers being more potent.¹⁴ However, as was seen previously with commercially available PDE3 inhibitors, some DNMDP analogs, e.g., 2, were potent PDE3 inhibitors with no cellular activity. The increase in cellular potency relative to biochemical activity for active compounds (HeLa EC₅₀s 2.4–15-fold lower than PDE3A/B average IC₅₀s) is consistent with neomorphic modulation, due to complex formation with SLFN12 in this case (see below).

Table 3. PDE3 Inhibition and HeLa Viability^a.

	IC50 (nM)
compound	PDE3A	PDE3B	EC50, HeLa (nM)
DNMDP	25	100	6.9
9	120	260	13
(R)-30	10	27	1.6
31	8	14	4.5
15	2000	3500	310
2	24	16	>1000
25	8	10	24

^aIC₅₀s are an average of two values.

Attempts to find other molecular targets of our compounds were unsuccessful. A Millipore Kinase Profiler screen of DNMDP found no inhibition (conc. 10 μM) against 234 kinases.¹ A Eurofins Lead Profiling screen of compound 9 (conc. 10 μM, 68 assays) found no interactions with a nonkinase target (Supporting Information Table 2).

Previously we used mass spectrometry proteomic analysis to demonstrate that DNMDP binds to PDE3A, inducing PDE3A binding to SLFN12, which results in cell killing, whereas nontoxic PDE3 inhibitors do not induce PDE3A/SLFN12 complex formation.¹ To observe compound-induced PDE3/SLFN12 interaction with newer compounds we transfected HeLa cells with a plasmid that expresses SLFN12 fused with a V5 epitope tag. We then treated the cells with DNMDP or (R)-30 (10 μM). After 8 h of compound treatment, the cells were lysed, and endogenous PDE3A protein was immunoprecipitated using an anti-PDE3A antibody. SLFN12 coimmunoprecipitation was analyzed by immunoblotting with an anti-V5 antibody to detect the SLFN12-V5 fusion protein. The SLFN12-V5 was clearly detected with the anti-V5 antibody for both DNMDP and (R)-30, indicating that both compounds stabilize the PDE3A-SLFN12 interaction (Figure 1).

Figure 1.

Detection of PDE3A/SLFN12 binding.

Cell death with DNMDP occurs via apoptosis by an as of yet determined mechanism.¹ A recent article reports PDE3A/SLFN12 interaction being induced by high concentration of 17-β-estradiol, leading to HeLa cell death, which can be rescued by the PDE3 inhibitors cilostazole and trequinsin, as seen with DNMDP-induced HeLa cell death.¹⁵ The stabilization of SFLN12 leads to a decrease in translation of antiapoptotic proteins leading to cell death.

The in vitro pharmacokinetic properties of selected active compounds were determined (Table 4). Overall the properties were good, and all compounds were highly soluble (>70 μM). Removal of the nitro group from DNMDP (7) and replacing the diethylamino group with a morpholine (9) improved microsomal stability. Addition of a fluorine atom (30) was well tolerated. Halogenated analogs (13) were stable to microsomes as was one heterocyclic analog (20). Other carbon substituted analogs (16, 18) were unstable to microsomes. Plasma protein binding of the analogs generally correlated with cLogP, with the more polar compounds displaying less protein binding. Plasma stability, in mouse and human, was >80% at 5 h for all compounds tested.

Table 4. Pharmacokinetic in Vitro Properties of Select Compounds^a.

	microsomal stability (% remaining at 1 h)		plasma protein binding (% bound)		plasma stability (% remaining at 5 h)
Compound	h	m	h	m	h	m
DNMDP	95	23	87	92	85	87
7	89	26	90	90	104	96
9	94	99	47	42	87	105
(±)-30	109	102	57	55	80	101
(±)-29	108	82	56	52	93	96
13	81	92	87	81	93	86
20	83	105	56	54	104	108

^ah = human, m = mouse; values are average of three values.

Compound (R)-30 was profiled to determine its suitability for animal studies. Compound (R)-30 does not inhibit other phosphodiesterases tested (Supporting Information Table 3). No inhibition in a cytochrome P450 panel (CYPs 1A2, 2C8, 2C9, 2D6, and 3A4) was detected when tested at high concentrations. Compound (R)-30 is soluble (ca. 1 mM), permeable (Caco2 A-B 279 nm/s; B-A 198 nm/s, efflux ratio 0.71), and stable in solution at various pH levels (1, 7, and 10) for extended time periods. An Ames test was negative, and no hERG interaction was detected at high concentration. (R)-30 showed high plasma levels in mice after iv (1 mg/kg) as well as po (2 mg/kg) dosing over eight hours making it a valuable candidate for in vivo xenograft testing (Supporting Information Table 4).

HeLa proved to be a convenient and robust cell line for viability assays, but for xenograft experiments we chose to work with a cancer cell line more relevant to therapeutic opportunities. Viability assays were performed under identical procedures using the SK-MEL-3 melanoma cell line with eight compounds (Table 5). Gratifyingly, the SK-MEL-3 EC₅₀ values were similar to those of HeLa.

Table 5. Viability with SK-MEL-3 Cells.

compound	EC50 (nM)	compound	EC50 (nM)
DNMDP	12	15	684
13	29	9	11
(R)-30	1.3	31	5.6
3	4.2	25	31

^an = 3; std. error < 10%.

The antitumor activity of compound (R)-30 was evaluated in tumor xenografts derived from SK-MEL-3 melanoma cells that were subcutaneously inoculated into female NMRI nude mice (Figure 2). (R)-30 was applied orally at 10 and 20 mg/kg twice daily (2QD) and at 50 mg/kg once per day (QD). We observed inhibition of tumor growth upon all (R)-30 treatments, achieving the strongest antitumor activity at 50 mg/kg QD with a T/C_rel.area of 0.09 and T/C_weight of 0.16 (p < 0.001 and p < 0.05 vs vehicle, respectively). All treatments were well tolerated without critical body weight loss (>10%) or toxicities (Supporting Information Figure 1).

Figure 2.

Antitumor efficacy of (R)-30 in the SK-MEL-3 tumor model in NMRI nude mice.

Plasma levels of (R)-30 were monitored throughout the xenograft experiment (Figure 3). With a dose of 50 mg/kg QD, the unbound plasma level of (R)-30 at 24 h was >100-fold the EC₅₀ against SK-MEL-3 cells, and at doses of 10 and 20 mg/kg 2QD, the EC₅₀ was covered for >30-fold at 24 h (Figure 3).

Figure 3.

Unbound plasma levels of (R)-30 on day 45 after final dose to SK-MEL-3 bearing NMRI nude mice. Curves for lower dose 2QD were simulated.

In summary, starting with our initial screening hit DNMDP, compounds were optimized using a cellular viability assay, making them not only more potent but also improving upon their pharmacological properties. The increase in cellular potency correlated with an increase in PDE3 inhibition, though most known PDE3 inhibitors do not kill HeLa cells. Our advanced compound, (R)-30/BRD9500, induced PDE3A/SLFN12 binding in HeLa cells and was active with oral dosing in a mouse xenograft model of melanoma. Current studies are focused on determining the mechanism of action of small molecule modulators of PDE3 and SLFN12 leading to cancer cell death.

Glossary

ABBREVIATIONS

PDE3

phosphodiesterase 3

SLFN12

Schlafen family member 12

DNMDP

6-(4-(diethylamino)-3-nitrophenyl)-5-methyl-4,5-dihydropyridazin-3(2H)-one

NMP

N-methyl-2-pyrrolidinone

TMSCCH

trimethylsilylacetylene

dba

dibenzylideneacetone

TBAF

tetrabutylammonium fluoride

LiHMDS

lithium hexamethyldisilazane

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsmedchemlett.9b00360.

• Synthesis and characterization of all compounds, biological assay procedures, and in vivo experiment details (PDF)

T.A.L., W.Y., and S.G. designed and synthesized compounds, L.d.W., X.W., and M.L. tested compounds, P.L. supervised pharmacokinetic studies, A.W. and C.K. performed the in vivo study, and T.A.L., S.L.S., H.G., and M.M. directed the project and wrote the manuscript, which has been approved by all the authors.

The authors would like to thank the US National Institutes of Health’s Molecular Libraries Program Centers Network (MLPCN) (grant number 3U45HG005032–05S1 to S.L.S.)¹⁶ and Bayer for financial support.

The authors declare no competing financial interest.