Further evaluation of novel structural modifications to scaffolds that engender PLD isoform selective inhibition

Abstract

This letter describes the on-going SAR efforts based on two scaffolds, a PLD1-biased piperidinyl benzimidazolone and a PLD2-biased piperidinyl triazaspirone, with the goal of enhancing PLD inhibitory potency and isoform selectivity. Here, we found that addition of an α-methyl moiety within the PLD2-biased piperidinyl triazaspirone scaffold abolished PLD2 preference, while the incorporation of substituents onto the piperdine moiety of the PLD1-biased piperidinyl benzimidazolone, or replacement with a bioisosteric [3.3.0] core, generally retained PLD1 preference, but at diminished significance. The SAR uncovered within these two allosteric PLD inhibitor series further highlights the inherent challenges of developing isoform selective PLD inhibitors.

There are two isoforms of the phosphodiesterase phospholipase D (PLD) in mammals, termed PLD1 and PLD2, that catalyze the production of the lipid second messenger phosphatidic acid.^1-4 From studies in our labs as well as others, we have demonstrated that PLD dysfucntion and/or overexpression can be modulated by small molecule, allosteric inhibitors 2-6 (Fig. 1) with therapetuic benefit in oncology, viral infections and CNS disorders.^1-11 While significant progress has been made since the report of halopemide 1,¹² an atypical antipsychotic, as a dual PLD1/2 inhbitor, the ideal, isoform selective PLD1 and PLD2 inhibitors with profiles suitable as in vivo proof of concept tools remain elusive.^1-12 Here, we report the further chemical optimization and evaluation of novel modifications to the scaffolds represented by 2 and 3, and the unexpected PLD pharmacological findings that resulted.

Figure 1.

Halopemide 1, and our recently reported isoform-selective PLD inhibitors: 2, VU0359595 (1,700-fold PLD1 selective), 3, VU0364739 (75-fold PLD2 selective), 4, ML298 (53-fold PLD2 selective), 5, ML299 (dual PLD1/2 inhibitor) and 6, ML395 (>80-fold PLD2 selective).

Within the triazaspirone-based series, represented by 3-6, we previously reported that incorporation of an (S)- β-methyl moiety (as in ML299, 5), dramiatically enhanced PLD1 inhibitory activity converting a highly PLD2 selective inhibitor series into a dual PLD1/2 inhibitor series.^7,8 This was a general finding within this series, and the magnitude of the impact of PLD1 activity exceeded 230-fold (Fig. 2) within the ML298 (4) molecule, (S)-5. The analogous (R)-enantiomer (R)-5 was similar but of lesser magnitude.⁸ However, we had not yet surveyed the impact of addition of a methyl group in the α-position to the piperidine on PLD1 and PLD2 activity. The synthetic route to access these α-methyl analogs proved to be straightforward (Scheme 1).

Figure 2.

Pharmacological impact of incorporation of an (R)- or (S)-β-methyl group into ML298 (4) afforidng (R)-5 and (S)-5. Both enhance PLD1 activity, but the (R)-enatiomer dramatically decreases PLD2 activity (10-fold) as well.

Scheme 1.

a) tert-butyl-(2-oxopropyl)carbamate, MP-B(OAc)₃, DCE, rt, 16 h (94%); b) i) 4 N HCl/dioxane, MeOH (98%); ii R₂COH, PS-DCC, HOBt, DCM, DIEA (76-92%).

Starting with commercial piperdine triazaspirone 6, a reductive amination with tert-butyl-(2-oxopropyl)carbamate provided the α-methyl material 7. Deprotection and standard amide coupling, with six optimal acids, delivered key racemic α-methyl congeners 8. As shown in Table 1, five of the six analogs, while displaying greater inhibitory activity at PLD2, were essentially non-selective versus PLD1 (only 2- to 5-fold selectivity). Only the 3,4-difluorophenyl amide, 8f, possessed potent PLD2 inhibitory activity (PLD2 IC₅₀ = 17 nM) and 63-fold selectvity versus PLD1 (PLD1 IC₅₀ = 1,080 nM). Despite this selectivity, the absolute potency at PLD1 (~1 μM), rendered 8f not useful as an in vitro/in vivo tool,⁸ as PLD1 would also be inhibited at standard testing concentrations. Therefore, we did not attempt to resolve the α-methyl enantiomers, and efforts focused on other domains of the PLD2-preferring core. The ‘magic methyl’ effect¹³ is very pronounced within this series and has profound impact on PLD1 and PLD2 activity.

Table 1.

Structures and activities of analogs 8.

Cmpd	Ar	PLD1 IC50 (nM)a	PLD2 IC50 (nM)b	Fold PLD2-selective
8a		315	64	5
8b		1,490	290	5
8c		230	120	2
8d		35	18	2
8e		21	6	3
8f		1,080	17	63

^aCellular PLD1 assay with Calu-1 Cells

^bCellular assay with HEK293-gfp PLD2 cells. Each IC₅₀ was determined In triplicate.

In parallel, efforts were being directed at the PLD1-biased piperidinyl benzimidazolone scaffold represented by 2.^1,2,5,6 This series was plagued with ancillary pharmacology, due to the GPCR privileged structure, and poor metaboic stability (MET ID indicated oxidative metabolism on the central piperidine ring).^5,6,9 In an attempt to address these issues, we elected to install a methyl group α to the piperidine nitrogen to block oxidative metabolism (as in 9 and 10), as well as a β-fluorine atom (as in 11) to modulate pK_a and potentially improve ancillary pharmacology at biogenic amine targets. The requisite functionalized piperidine benzimidazolones were prepared as previously described following literature routes^5,6,14 and then elaborated via a reported variation on Scheme 1.^5,6 While we were excited to note that these modifications to the piperdine core retained PLD inhbitiory activity, the compounds were less potent and possessed diminished PLD1 selectivity as compared to the unsubstitiuted congeners (Fig. 4); thus, they did not represent a path forward towards improved in vivo tools. However, a brief metabolic stability assessment showed that 9 was more stable in rat microsomes than the corresponding unsubstituted derivative (Cl_hep = 70 mL/min/kg versus Cl_hep = 36 mL/min/kg), and that 11 was inactive at D₂ (IC₅₀ >10 μM), whereas the des-fluorocongener possessed a D₂ IC₅₀ of 22 nM.

Figure 4.

Pharmacological impact of incorporation of substituents on the central piperidine ring in the piperidine benzimidazoone series of PLD1 selective inhibitors.

Finally, we decided to survey the replacement of the piperidine ring with a bioisoteric substitute, namely a [3.3.0] ring system, or octahydrocyclopenta[c]pyrrole.¹⁵ Commercial 2-fluoronitrobenzene 12 was heated under microwave irradiation with tert-butyl-5-aminohexahydrocyclopenta[c]pyrrole-2(1H)-carboxylate to deliver the S_NAr product 13 in 75% yield.^5,14 A zinc mediated nitro reduction afforded 14 which was then treated with triphosgene to provide the benzimidazolone 15. Acid deprotection and standard elaboration based on Scheme 1 yielded seven novel [3.3.0] analogs 16. ^5,14

As shown in Table 2, this effort, with a well documented piperidine biosiostere, led to a significant loss in PLD1 potency and greatly diminished PLD2 selectivity (only 3- to 24-fold). Moreover, 16g was 9.4-fold PLD2 preferring, whereas in the piperidine-based series, the trans-cyclopropylphenyl amide was ~111-fold PLD1 preferring. This is the first example in a benzimidazolone core that preferential PLD2 inhbition has been observed. Once again, metabolic stability in rat microsomes was slighlty improved relative to the piperdine-based series, but neither PLD potency or isofrom selectivity was acceptable. So, while the rationale was sound for the synthesis and evaluation of novel analogs 9-11 and 16, the unpredictable nature of allosteric SAR once again thwarted our efforts, producing weak, dual PLD1/2 inhibitors.

Table 2.

Structures and activities of analogs 16.

Cmpd	Ar	PLD1 IC50 (nM)a	PLD2 IC50 (nM)b	Fold PLD1-selective
16a		244	1,790	7.3
16b		870	3,565	4
16c		468	2,590	5.5
16d		230	1,235	5.4
16e		72	220	3
16f		820	3,250	3.9
16g		1,080	115	−9.4

^aCellular PLD1 assay with Calu-1 Cells

^bCellular assay with HEK293-gfp PLD2 cells. Each IC₅₀ was determined in triplicate.

In summary, we explored and evaluated novel SAR within two distinct chemotypes that afford isoform selective PLD inhibition. While the medicinal chemistry drivers proved to be correct to address ancillary pharmacology and metabolic stability, the inherent steep SAR of allosteric ligands led to a loss in PLD activity and/or PLD isoform selectivity, precluding the development of improved in vivo tools from this campaign. Interestingly, a pronounced ‘magic methyl’ effect was discovered. Efforts continue, and work is in progress to develop optimal in vivo tool compounds that selectively inhibit either PLD1 ro PLD2.

Figure 3.

PLD1 (Calu-1) and PLD2 (293-PLD2) cell-based assay concentration-response curves (CRCs) for representatvie library memebers 8. A) CRCs for 8e; B) CRCs for 8f; C) CRCs for 8c.

Scheme 2.

Reagents: (a) tert-butyl-5-aminohexahydrocyclopenta[c]pyrrole-2(1H)-carboxylate, Na₂CO₃, KI, cyclohexanol, μw, 180°C, 10 min, 90%; (b) Zn, 1N HCl, MeOH; (c) i) triphosgene, Et₃N, THF, rt, 2 h; ii) 4 N HCl dioxane, rt, 82%.