Structure-activity relationship studies in a new series of 2-amino-N-phenylacetamide inhibitors of Slack potassium channels

Abstract

In this Letter we describe structure-activity relationship (SAR) studies conducted in five distinct regions of a new 2-amino-N-phenylacetamides series of Slack potassium channel inhibitors exemplified by recently disclosed high-throughput screening (HTS) hit VU0606170 (4). New analogs were screened in a thallium (Tl⁺) flux assay in HEK-293 cells stably expressing wild-type human (WT) Slack. Selected analogs were screened in Tl⁺ flux versus A934T Slack and other Slo family members Slick and Maxi-K and evaluated in whole-cell electrophysiology (EP) assays using an automated patch clamp system. Results revealed the series to have flat SAR with significant structural modifications resulting in a loss of Slack activity. More minor changes led to compounds with Slack activity and Slo family selectivity similar to the HTS hit.

Malignant migrating partial seizure of infancy (MMPSI), also known as epilepsy of infancy with migrating focal seizures (EIMFS), is a rare, serious form of infantile epilepsy.¹ Patients with MMPSI are typically pharmacoresistant with seizure onset developing within the first six months of life and a 25% mortality rate within the first year. Approximately half of MMPSI cases have been linked to multiple de novo gain-of-function (GOF) heterozygous missense mutations in the KCNT1 gene.^{2, 3} KCNT1 encodes for the sodium-activated potassium channel known as K_Na1.1 or Slack (Sequence Like A Calcium-Activated K⁺ Channel). Slack is a member of the Slo family (Slo2.2) of K⁺ channels along with Slo1 (Maxi-K), Slo2.1 (Slick), and Slo3. Slack channels are essential regulators of electrical activity and distributed throughout the central nervous system, where they play important roles in affecting neuronal excitability.^3–7 Slack channels are characterized by 4 subunits with six hydrophobic transmembrane domains (S1-S6) and a pore-forming domain between S5 and S6. Also present are an extended C-terminal cytoplasmic domain, containing a regulator of potassium conductance (RCK) domain and a nicotinamide adenine dinucleotide-binding (NAD⁺) domain.^{2, 6, 8–11}

Several recent reports have documented more than 30 mutations in Slack. The GOF mutations correlated with MMPSI are associated with an increase in Slack current.^{3, 12–15} In addition to MMPSI, KCNT1 GOF mutations are associated with other epilepsy disorders, including autosomal dominant sleep-related hypermotor epilepsy (ADSHE),¹⁶ early-onset epileptic encephalopathy (EOEE),¹⁷ myoclonic-atonic epilepsy,¹⁸ temporal lobe epilepsy with intellectual disability,¹⁹ and autosomal dominant nocturnal frontal lobe epilepsy (ADNFLE).^{20, 21} Multiple potential mechanisms by which these GOF mutations in Slack channels lead to increased neuronal excitability and rate of firing have been put forth. For example, activation of Slack channels leads to increased amplitude of AHP, resulting in shorter action potentials and increased neuronal firing frequency.^{12, 22} Likewise, GOF Slack mutations may reduce the excitability of inhibitory interneurons, ultimately leading to decreased inhibitory tone.^{8, 13} Finally, Slack also has non-conducting functions, and interacts with downstream signaling pathways that may be disrupted by mutations.^{22, 23}

No broadly effective and safe drug therapy option is presently available for the treatment of MMPSI. Drugs such as quinidine, bepridil, and cofilium modulate Slack currents in vitro, and attempts have been made to repurpose them for KCNT1-associated epilepsies; however, these compounds are hampered by a lack of selectivity, toxicity, and suboptimal drug metabolism and pharmacokinetic (DMPK) properties.^{4, 8, 14, 24–30} Notably, recent reports have described new compounds from diverse chemotypes that inhibit Slack function (Figure 1).³¹ For example, virtual screening of a commercial library employing a cryo-electron microscopy-derived structure of chicken K_Na1.1 identified six Slack inhibitors with micromolar potency, each one predicted to function via blockade of the channel pore.³² Compounds 1-2 are representative analogs with reported IC₅₀ values of 3.2 and 3.0 µM, respectively, in a whole-cell patch clamp assay using HEK-293 cells expressing human wild-type (WT) Slack channels. In addition, a high-throughput screen (HTS) and ensuing optimization effort identified orally available 1,2,4-oxadiazole 3, which normalizes the EEG phenotype in a mouse model of Kcnt1 GOF.³³ Compound 3 was a potent Slack inhibitor with a reported IC₅₀ of 40 nM in an automated patch clamp assay in HEK-TREX cells stably expressing human WT Slack channels. Finally, we recently reported the results of our own HTS utilizing a thallium (Tl⁺) flux assay in HEK-293 cells stably expressing human WT and selected mutant Slack channels. The 2-amino-N-phenylacetamide VU0606170 (4) was identified through this campaign and possessed low micromolar potency for both WT and A934T Slack in Tl⁺ flux and whole-cell automated patch-clamp assays. VU0606170 (4) has an attractive selectivity profile and decreased the firing rate in an overexcited, spontaneously firing cortical neuronal culture consistent with an antiepileptic effect.³⁴ In this Letter, we report the results of structure-activity relationship (SAR) studies within the 2-amino-N-phenylacetamide chemotype exemplified by 4.

Figure 1.

Recently reported small molecule inhibitors of Slack channels: predicted pore blockers 1-2, in vivo tool 3, and HTS hit 4 (VU0606170).

Desiring to rapidly evaluate the potential tractability of VU0606170 (4), we decided to pursue SAR development in five regions of the scaffold (Figure 2). Where amenable, we employed late-stage penultimate intermediates that facilitated the synthesis of small compound libraries. Scheme 1 was used to evaluate SAR in the eastern ring of the scaffold. Commercial sulfamoyl chloride 5 was reacted with 1-Boc-piperazine 6 to afford sulfamide 7. Cleavage of the protecting group under acidic conditions gave secondary amine 8 as the trifluoroacetic acid salt. Alkylation of 8 with ethyl bromoacetate was accomplished with mild heating to provide ester 9. Hydrolysis of the ester with aqueous lithium hydroxide afforded acid and penultimate intermediate 10. Coupling with a variety of aryl amines provided final amide analogs 11-29 and 31-36. Synthesis of analog 30 via amide coupling proved ineffective. Thus, 2,5-dichloroaniline 37 was alkylated with chloroacetyl chloride to afford 38, which was subsequently reacted with intermediate 8 to provide 30.

Figure 2.

Regions for SAR development in the VU0606170 scaffold.

Scheme 1.

Reagents and conditions: (a) 1-Boc-piperazine (6), DIEA, CH₂Cl₂, 81%; (b) TFA, CH₂Cl₂, 97%; (c) Ethyl bromoacetate, K₂CO₃, DMF, 40 ºC, 92%; (d) LiOH, H₂O, THF, 99%; (e) ArNH₂, HATU, DIEA, DMF, 4–64%; (f) chloroacetyl chloride, NEt₃, DMAP, CH₂Cl₂; (g) 8, K₂CO₃, KI, DMF, 50 ºC, 7% (2 steps).

Screening of all new compounds versus WT Slack in our Tl⁺ flux assay utilized optimized buffer conditions that were slightly modified from those previously published to afford a larger signal amplitude (see Supporting Information).³⁴ An initial scan at all positions employing common functional groups quickly alerted us to the potential for mode switching in the scaffold as most of the analogs were weak activators of the channel (Table 1). Such an observation is consistent with compounds in this chemotype modulating Slack function by a mechanism other than simple pore blockade. A few analogs (15, 18, 19, and 22) displayed a mixed profile, appearing to behave as either weak inhibitors or inactive at higher concentrations and weak activators at low concentrations. Although the reason for this profile is not known, a possible explanation is that there are two binding sites in the channel and occupancy of the first site is weakly activating while occupancy of both is inhibiting. Alternatively, there may be two distinct binding sites, one activating and the other inhibiting. Given that 2-methoxy analog 21 was the only compound that maintained the inhibitory profile observed with VU0606170 (4), we redirected our efforts toward analogs that maintained the 2,5-disubstituted pattern of 4 with either the 2-methoxy group (24-28) or the 5-chloro group (29-35) held constant (Table 2).

Table 1.

Eastern amide analogs from initial scan

No.	R	Modea	IC50 / EC50 (µM)b	Efficacy (%)b,c
4	2-OMe, 5-Cl	Inh	2.4	100
11	H	Act	>10d	21
12	2-F	Act	>10d	30
13	3-F	Act	4.6	13
14	4-F	Act	>10d	30
15	2-Cl		not determinede
16	3-Cl	Act	7.6	38
17	4-Cl	Act	>10d	58
18	2- Me		not determinede
19	3- Me		not determinede
20	4- Me	Act	4.1	13
21	2-OMe	Inh	>10d	85
22	3-OMe		not determinede
23	4-OMe	Act	>10d	15

^aInh = inhibitor; Act = activator; NA = not active

^bConcentration-response curve (CRC) from Tl⁺ flux assay in HEK-293 cells expressing WT Slack

^cAmplitude of response in the presence of 30 µM test compound as a percentage of the maximum response for VU0606170 (inhibitors) or loxapine (activators)

^dCRC does not plateau within the concentration range tested

^eCompounds appear to be weak inhibitors/inactive at high concentrations and weak activators at low concentrations

Table 2.

Eastern amide analogs from focused scan

No.	R	Modea	IC50 / EC50 (µM)b	Efficacy (%)b,c
24	F	Inh	4.9	102
25	Br	Inh	1.4	52
26	Me	Inh	>10d	82
27	CF3	Inh	2.4	89
28	t-Bu		not determinede
29	F	Act	3.8	<10
30	Cl		inactive
31	Me	Inh	>10d	44
32	OEt	Inh	7.1	98
33	O(n-Pr)	Inh	4.3	95
34	O(i-Pr)	Inh	5.1	92
35	OCH2(c-Pr)	Inh	7.5	59
36	—		inactive

^aInh = inhibitor; Act = activator; NA = not active

^bCRC from Tl⁺ flux assay in HEK-293 cells expressing WT Slack

^cAmplitude of response in the presence of 30 µM test compound as a percentage of the maximum response for VU0606170 (inhibitors) or loxapine (activators)

^dCRC does not plateau within the concentration range tested

^eCompound appears to be a weak inhibitor/inactive at high concentrations and weak activator at low concentrations

Replacement of the 5-chloro group with other electron-withdrawing substituents maintained the inhibitory profile with bromo (25) and trifluoromethyl (27) analogs showing similar potency to VU0606170 (4). Notably, while the potency was similar, 25 demonstrated approximately half of the efficacy observed with 4 and 27 as evidenced by clear plateauing of the CRC well above the maximum observed with the control compound. Replacement of the 2-methoxy group with halogens (29 and 30) or methyl (31) was not effective. Analogs 32-35 explored lengthening and branching of the alkyl ether substituent. In each case, the compounds were inhibitors; however, potency was 2–3 fold lower than 4. Finally, analog 36 demonstrated the lack of tolerance for an aromatic nitrogen atom.

To explore SAR in the western ring and associated linker 1, we employed chemistry outlined in Scheme 2. Reaction of 1-Boc-piperazine 6 with ethyl bromoacetate under basic conditions afforded ester 39. Hydrolysis of the ester with lithium hydroxide provided lithium salt intermediate 40. Coupling of 40 with 5-chloro-2-methoxyaniline proceeded with HATU. Removal of the Boc protecting group gave secondary amine 42 as its trifluoroacetic acid salt. Reaction of 42 with sulfamoyl chlorides provided sulfamides 43 and 44, while reaction with sulfonyl chlorides yielded sulfonamides 45-48. To access urea analogs 50-55, we first converted 42 to 4-nitrophenylcarbamate 49. Treatment of 49 with various cyclic secondary amines under microwave heating afforded the desired ureas 50-55.³⁵

Scheme 2.

Reagents and conditions: (a) Ethyl bromoacetate, K₂CO₃, DMF, 68%; (b) LiOH, H₂O, THF, 98%; (c) 5-chloro-2-methoxyaniline, HATU, DIEA, DMF, 61%; (d) TFA, CH₂Cl₂, 100%; (e) R²SO₂Cl, DIEA, CH₂Cl₂; 41–58%; (f) 4-nitrophenyl chloroformate, CH₂Cl₂, 0 ºC; (g) cyclic secondary amine, NMP, µwave, 200 ºC, 25 min, 32–51% (2 steps).

The most modest changes in the western ring (43-45) were not tolerated (Table 3). Benzene and 4-chlorobenzene sulfonamides 46 and 48 exhibited potency on par with 4; however, efficacy was diminished considerably. Surprisingly, 4-methylbenzene sulfonamide 47 was inactive. For the most part, the urea analogs were weak activators of the channel (50, 51, 54, and 55) or displayed a mixed profile (53). Only analog 52 functioned as a weak inhibitor.

Table 3.

Western ring and linker 1 analogs

No.	R	Modea	IC50 / EC50 (µM)b	Efficacy (%)b,c
43	piperidin-1-yl		inactive
44	4-morpholino		inactive
45	cyclohexyl		inactive
46	phenyl	Inh	1.1	32
47	4-methylphenyl		inactive
48	4-chlorophenyl	Inh	2.3	16
50	3,3-difluoropyrrolidin-1-yl	Act	>10d	<10
51	piperidin-1-yl	Act	>10d	10
52	4-propylpiperidin-1-yl	Inh	>10d	58
53	4-isopropylpiperidin-1-yl		not determinede
54	4,4-dimethylpiperidin-1-yl	Act	9.0	23
55	azepan-1-yl	Act	>10d	17

^aInh = inhibitor; Act = activator; NA = not active

^bCRC from Tl⁺ flux assay in HEK-293 cells expressing WT Slack

^cAmplitude of response in the presence of 30 µM test compound as a percentage of the maximum response for VU0606170 (inhibitors) or loxapine (activators)

^dCRC does not plateau within the concentration range tested

^eCompound appears to be a weak inhibitor/inactive at high concentrations and weak activator at low concentrations

We previously reported that relocation of the carbonyl group on linker 2 from the carbon adjacent to the aniline nitrogen to the carbon adjacent to the piperazine nitrogen was not tolerated.³⁴ In fact, we employed this regioisomer of VU0606170 known as VU0849686 as a negative control. Still, we were interested in exploring substitution of the linker and investigated stereospecific methylated analogs via Scheme 3. Following a method described in the patent literature,³⁶ we converted commercially available chiral alcohol 56 to its corresponding triflate 57, which was reacted with secondary amine 8 to provide 58 with inversion of stereochemistry. Hydrolysis of the methyl ester was accomplished with 6N hydrochloric acid and heating to provide acid 59 as its hydrochloride salt. Coupling with 5-chloro-2-methoxyaniline and 2-methoxy-5-(trifluoromethyl)aniline afforded analogs 60 and 61, respectively. The same sequence was repeated utilizing methyl (S)-2-hydroxypropanoate instead of 56 to afford the R-enantiomer analogs 62 and 63 (Table 4). No evidence of significant preference for one enantiomer versus the other was noted. In both cases the trifluoromethyl analogs (61 and 63) were slightly more potent than the chloro analogs (60 and 62); however, all four compounds were 1.5 to 2-fold less potent than the hit VU0606170 (4).

Scheme 3.

Reagents and conditions: (a) Tf₂O, 2,6-lutidine, CH₂Cl₂, −70 ºC to r.t., 53%; (b) 8, K₂CO₃, CH₂Cl₂, H₂O, 0 ºC to r.t., 71%; (c) 6N HCl, 100 ºC, 56%; (d) 5-chloro-2-methoxyaniline or 2-methoxy-5-(trifluoromethyl)aniline, HATU, DIEA, DMF, 35% (60), 63% (61).

Table 4.

Methylated linker 2 analogs

No.	*	R	Modea	IC50 / EC50 (µM)b	Efficacy (%)b,c
60	S	Cl	Inh	4.4	93
61	S	CF3	Inh	3.4	95
62	R	Cl	Inh	4.9	98
63	R	CF3	Inh	3.5	100

^aInh = inhibitor; Act = activator; NA = not active

^bCRC from Tl⁺ flux assay in HEK-293 cells expressing WT Slack

^cAmplitude of response in the presence of 30 µM test compound as a percentage of the maximum response for VU0606170

In the core of the scaffold, we sought to evaluate the impact of substitution as well as the tolerability for various piperazine isosteres.^{35, 37, 38} In some cases the core monomers were not symmetric (e.g., 64 in Scheme 4), and we thus employed complementary routes to arrive at two distinct analogs. Specifically, monomer 64 was converted to sulfamide 65 as described previously. Removal of the protecting group to afford 66 and subsequent alkylation with 67 gave analog 68. The same three-step sequence was also carried out in reverse order to arrive at analog 71. The same complementary approach was likewise followed for monomers 72 and 73, while symmetric monomers 74-76 were simply reacted in the manner used to prepare analog 71. Two additional core analogs (80 and 85) followed slightly modified procedures (Scheme 5). In the case of acyclic analog 80, the use of a protecting group was avoided by employing an excess of diamine 77. The remaining steps were as previously described. Preparation of piperidine analog 85 utilized previously outlined chemistry but was initiated with starting material 82, which already contained an ester group.

Scheme 4.

Reagents and conditions: (a) 5, DIEA, CH₂Cl₂, 61% (65), 98% (71); (b) TFA, CH₂Cl₂, 96% (66), 94% (70); (c) 67, DIEA, CH₃CN, 75 ºC, 66% (68), 81% (69).

Scheme 5.

Reagents and conditions: (a) 5, CH₂Cl₂, 99%; (b) Ethyl bromoacetate, K₂CO₃, DMF, 33%; (c) LiOH, H₂O, THF, 95% (80), 99% (84); (d) 5-chloro-2-methoxyaniline, HATU, DIEA, DMF, 40% (81), 24% (85); (e) 5, DIEA, CH₂Cl₂, 93%.

Methyl substitution of the piperazine core on the carbon adjacent to the sulfamide nitrogen (68 and 86) significantly reduced potency (Table 5). On the other hand, methyl substitution of the other carbon was well tolerated (71 and 87), with (R)-enantiomer 87 being equipotent to hit compound 4. Acyclic analog 81 retained inhibitory activity but exhibited reduced potency, likely due to increased degrees of rotational freedom. Piperidine 85 and spirocycles 88 and 89 were weak activators while spirocycle 90 was a weak inhibitor. Octahydropyrrolo[3,4-c]pyrrole 91 was the only analog in this set devoid of activity at Slack. Finally, expansion of the ring in the form of homopiperazine 92 gave a moderately potent analog.

Table 5.

Core analogs

No.	CORE	Modea	IC50 / EC50 (µM)b	Efficacy (%)b,c
68		Inh	>10d	20
86		Inh	13	58
71		Inh	3.5	93
87		Inh	2.4	101
81		Inh	>10d	103
85		Act	9.3	25
88		Act	>10d	17
89		Act	10	32
90		Inh	>10d	69
91			inactive
92		Inh	6.2	106

^aInh = inhibitor; Act = activator; NA = not active

^bCRC from Tl⁺ flux assay in HEK-293 cells expressing WT Slack

^cAmplitude of response in the presence of 30 µM test compound as a percentage of the maximum response for VU0606170 (inhibitors) or loxapine (activators)

^dCRC does not plateau within the concentration range tested

Our SAR studies in the VU0606170 (4) chemotype to date have not identified analogs that improve upon the potency of the hit compound. Still, we have identified several analogs that were equipotent or near equipotent with 4.⁸ Thus, we decided to take a closer look at the pharmacological profiles of select analogs. First, we screened nine select inhibitors against A934T mutant Slack, a variant commonly associated with MMPSI (Table 6).^{2, 3} All of the molecules functioned as inhibitors and were equipotent or more potent versus A934T as compared to WT with the majority of the molecules being slightly more potent versus A934T. Such a result may indicate that our inhibitors block an open confirmation of the channel, a phenomena observed with quinidine in other variants with Slack GOF mutations.⁸

Table 6.

Comparison of activity versus WT and A934T Slack for select inhibitors

No.	WT		A934T
	IC50 (µM)a	Efficacy (%)a,b	IC50 (µM)a	Efficacy (%)a,b
4	2.4	100	1.4	100
24	4.9	102	2.3	100
27	2.4	89	1.3	88
33	4.3	95	3.1	95
60	4.4	93	2.7	98
61	3.4	95	2.5	96
62	4.9	98	2.9	106
63	3.5	100	2.8	102
71	3.5	93	2.1	94
87	2.4	101	1.1	100

^aCRC from Tl⁺ flux assay in HEK-293 cells expressing either WT or A934T Slack

^bAmplitude of response in the presence of 30 µM test compound as a percentage of the maximum response for VU0606170

We next chose three of the most potent inhibitors with variations at distinct points of the scaffold for screening versus Slick (Slo2.1) and Maxi-K (Slo1 α1/β3) to assess Slo family selectivity (Table 7). Selectivity versus Maxi-K was generally very good, with 27 and 87 showing weak inhibition and 61 being inactive. Most analogs demonstrated weak inhibition of Slick. While 27 had an IC₅₀ of 4.1 µM, efficacy was low at 29% of the positive control. Thus, these analogs displayed similar selectivity to VU0606170 (4), which is not surprising, given the small structural differences. These same three inhibitors were also examined in a whole-cell electrophysiology (EP) assays using an automated patch clamp system (SyncroPatch 384) (Table 8). Gratifyingly, potency and efficacy in this assay was on par with those observed in Tl⁺ flux in both WT and A934T expressing cells with each of these analogs.

Table 7.

Slo family selective for select Slack inhibitors

No.	Slick (Slo2.1)		Maxi-K (Slo1 α1/β3)
	IC50 (µM)a	Efficacy (%)a,b	IC50 (µM)a	Efficacy (%)a,b
4	>10c	35	inactive
27	4.1	29	>10c	36
61	>10c	34	inactive
87	>10c	50	>10c	32

^aCRC from Tl⁺ flux assay in HEK-293 cells expressing either Slick (Slo2.1) or Maxi-K (Slo1 α1/β3)

^bAmplitude of response in the presence of 30 µM test compound as a percentage of the maximum response for SKF96365 (Slick) or paxilline (Maxi-K)

^cCRC does not plateau within the concentration range tested

Table 8.

Whole-cell electrophysiology (EP) versus WT and A934T Slack for select inhibitors

No.	WT		A934T
	IC50 (µM)a	Efficacy (%)a,b	IC50 (µM)a	Efficacy (%)a,b
4	1.7	100	1.6	100
27	2.3	100	1.1	90
61	1.9	100	1.5	70
87	1.7	100	1.9	100

^aCRC from whole-cell EP in HEK-293 cells expressing either WT or A934T Slack utilizing an automated patch clamp system

^bAmplitude of response in the presence of 30 µM test compound as a percentage of the maximum response for VU0606170

In conclusion, we have executed SAR studies in a new series of 2-amino-N-phenylacetamide inhibitors of Slack potassium channels based on the HTS hit VU0606170. Systematic evaluation of five distinct regions of the chemotype revealed flat SAR with most substantial structural modifications proving deleterious to Slack activity. Some analogs with more modest structural deviations from the hit compound exhibit potency on par with the HTS hit. For example, chiral methyl substituents on the core (87) and linker 2 (61) and exchange of the chloro group of the hit for a trifluoromethyl group (27 and 61) gave potent analogs. Importantly, testing of selected analogs versus the A934T Slack mutant revealed potency on par with the WT, and results in whole-cell EP mirrored those obtained with Tl⁺ flux for the same analogs, validating our primary optimization screening approach utilizing Tl⁺ flux as our frontline assay. Utilization of homology models and docking studies might prove instrumental in the optimization of small molecule Slack inhibitors in the future; however, additional information about the binding site for these compounds would be required to best enable such an approach. The observed mode-switching in this scaffold argues against the compounds functioning as simple pore blockers. Hit evaluation and optimization continues in additional distinct scaffolds discovered through our HTS campaign. Reports from those efforts will be reported in the near future.