Design, synthesis, modeling, biological evaluation and photoaffinity labeling studies of novel series of photoreactive benzamide probes for histone deacetylase 2

Abstract

The design, modeling, synthesis, biological evaluation of a novel series of photoreactive benzamide probes for class I HDAC isoforms is reported. The probes are potent and selective for HDAC1 and 2 and are efficient in crosslinking to HDAC2 as demonstrated by photolabeling experiments. The probes exhibit a time-dependent inhibition of class I HDACs. The inhibitory activities of the probes were influenced by the positioning of the aryl and alkyl azido groups necessary for photocrosslinking and attachment of the biotin tag. The probes inhibited the deacetylation of H4 in MDA-MB-231 cell line, indicating that they are cell permeable and target the nuclear HDACs.

Histone deacetylases (HDACs) are considered viable drug targets for multiple therapeutic applications including cancer and neurological diseases.^{1, 2} Recently, Cravatt et al³ and Gottesfeld et al⁴ described the design and applications of photoaffinity probes for profiling HDACs in native proteomes and live cells. The scaffold of the probes included a portion of a pan HDAC inhibitor suberoyl anilide hydroxamic acid (SAHA), a benzophenone group as a photoreactive group, and an alkyne handle to attach an azide containing reporter tag via (3+2) cycloaddition. Attempts to use the same features based on HDAC1 and 2 selective benzamide scaffolds resulted in probes with HDAC potency above 180 µM in HeLa cell nuclear lysate.⁵

We have already established the Binding (E) nsemble (Pro) filing with (F) photoaffinity (L) abeling approach (BEProFL) where we have experimentally mapped the multiple binding modes of diazide based photoreactive probes for HDACs.⁶ The design of these probes included decoration of HDAC ligands with a 3-azido-5-azidomethylene moiety, a photoaffinity labeling group originally proposed by Suzuki et al⁷ for specific labeling of the catalytic portion of HMG-CoA reductase. The aromatic azido moiety was used as a photoreactive group and the aliphatic azide was well suited for (3+2) cycloaddition with an alkyne moiety of the biotin-containing reporter group. Based on these features, we have successfully designed and synthesized highly potent and selective probes for HDAC3 and HDAC8 and demonstrated that they are cell permeable and exhibit excellent antiproliferative activity against several cancer cell lines.⁸ Our main objective in this study was to design photoreactive benzamide probes for HDAC2 and evaluate their activity/selectivity profile for other class I HDAC isoforms.

We hypothesized that a set of potent and selective benzamide-based probes capable of crosslinking with HDAC2 can be designed by appropriately decorating benzamides 1 and 2 (Fig. 1) with a combination of the aryl and alkyl azides. Both 1 and 2 and their derivatives were reported by Delorme⁹, Miller¹⁰, Gangloff¹¹, and their colleagues to be active and selective inhibitors of HDACs1 and 2.

Figure 1.

Benzamide inhibitors of HDAC 1 and 2

Substituted benzoic acids 3–8, mono-N-Boc protected phenylenediamines 9, 10 and azidoaniline 11 shown in Fig. 2 were chosen as precursors for the synthesis of the photoreactive probes. Acid 4, protected phenylendiamines 9, 10, and intermediates 16, 18 and 19 were synthesized as reported previously,^{7, 8, 10, 12} whereas benzoic acid 3 was available commercially. The synthesis of precursors 5, 6, 7, 8, and 11 and benzoic acids 3–8 followed by deprotection of the resulting N-Boc products to give the final probes in 70–80% overall yield¹³(Scheme 2).

Figure 2.

Amine and acid precursors used for synthesis of photoreactive probes

Scheme 2.

Reagents and conditions: (a) CDI, DBU, TEA, 4-(aminomethyl)benzoic acid, THF, 10 h, 0 °C - rt, 90–92%; (b) EDCI, HOBt, DMF, 12 h, 80 °C, 80–90%; (c) TFA/DCM, 0.5 h, rt, 90–95%.

The inhibitory profile of the probes against class I HDAC isoforms was determined using a fluorogenic assay and the results are given in Table 1. The inhibition of HDAC8 was measured using the fluorogenic acetylated substrate Fluor de Lys and purified recombinant human HDAC8 from E.coli,¹⁴ whereas the inhibition of HDAC1-3 was measured using fluorogenic acetylated substrate Boc- L-Lys(Ac)-AMC and commercially available recombinant human HDAC1-3.¹⁵ We also explored the effect of preincubation with HDAC1, 2, 3, and 8 as it was previously observed that the potency of the benzamide-based HDAC inhibitors increased with preincubation with HDAC1-3.^{11, 16} The maximum incubation time was chosen on the basis of stability of HDAC proteins in the conditions used to determine IC₅₀ values. We found that for HDAC1, 3, and 8 the maximum incubation time was 3 hours, whereas HDAC2 protein was stable for 24 hours. The IC₅₀ values of ligands 1 and 2 determined in this study vary from those reported previously.^{10, 11} We attribute this discrepancy to the differences in the assay conditions, the protein sources, substrates, and preincubation times. The analysis of SAR was facilitated by docking all the probes to HDAC2 (PDB:3MAX),¹¹ HDAC3 (PDB:4A69),¹⁷ and HDAC8 (PDB: 1T69)¹⁸ using GOLD v.5.1.^{19, 20}

Table 1.

Potency of the probes against class I HDAC isoforms

#	HDAC1 IC50(nM)		HDAC2 IC50(nM)			HDAC3 %Inhibition(10µM)		HDAC8 %Inhibition(10µM)
	Preincubation time		Preincubation time			Preincubation time		Preincubation time
	5 min	3 h	5 min	3 h	24 h	5 min	3 h	5 min	3 h
1a	2100 ± 44	140 ± 34	5130 ± 470	1050 ± 81	210 ± 17	6.6	35	NA	NA
1b	1200 ± 85	70 ± 5.3	3200 ± 260	690 ± 57	110 ± 36	8.7	56	NA	4.7
1c	5600 ± 520	1400 ± 160	18000 ± 910	5200 ± 400	2400 ± 74	2.2	27	NA	11
1d	21000 ± 100	13000 ± 320	32000 ± 1700	21000 ± 720	10000 ± 490	NA	24	NA	14
1e	23000 ± 1400	2700 ± 69	34000 ± 2300	6500 ± 120	830 ± 28	4.3	16	NA	22
1f	18000 ± 130	3800 ± 54	17000 ± 4400	6600 ± 140	750 ± 81	7.6	12	NA	6.6
1g	96000 ± 1600	55000 ± 1300	120000 ± 4900	94000 ± 530	77000 ± 4100	2.0	6.5	NA	26
2a	3800 ± 120	780 ± 22	3800 ± 540	1000 ± 70	320 ± 32	7.8	48	NA	NA
2b	2500 ± 280	990 ± 53	7000 ± 250	1100 ± 25	350 ± 16	11	46	NA	25
2c	2800 ± 240	1210 ± 68	7100 ± 220	1000 ± 50	300 ± 77	21	47	NA	22
1	410 ± 16	52 ± 4.3	1200 ± 93	350 ± 15	140 ± 8	26	96	NA	24
2	14500 ± 1300	1880 ± 5.2	38000 ± 2000	14000 ± 1030	740 ± 49	8.8	40	NA	3.0
SAHA	29 ± 1.6	34 ± 3.2	200 ± 14	ND	260 ± 4.3	100	100	100	100

NA – no inhibition up to 10µM concentration, ND - not determined. Data are mean ± SD of three independent experiments.

All the newly synthesized benzamide-based probes had activity ranging between 70 nM and 55 µM and 110 nM and 77 µM for HDAC1 and HDAC2, respectively. All of the probes demonstrated a robust 2–40-fold increase in inhibition of HDAC1 and 2 upon preincubation with the enzymes for 3 h and 24 h, respectively (Table 2). Consistent with the previously reported observation,¹¹ SAHA, a hydroxamate-based inhibitor, did not exhibit time-dependent inhibition. Similar trends were observed with HDAC3 and HDAC8. In the discussion below we will use only IC₅₀’s obtained at the maximum preincubation time, unless specified otherwise.

Table 2.

Ratios of IC₅₀ for HDACs 1, 2 and 3 with respect to preincubation time and selectivity for HDAC1 vs. 2. ND - not determined

#	HDAC1 IC50 ratio 3 h/5 min	HDAC2 IC50 ratio 3 h/5 min	HDAC2 IC50 ratio 24 h/5 min	HDAC3 %inhibition ratio 3 h/ 5 min	HDAC2/HDAC1 IC50 ratio 3 h/ 3 h	HDAC2/HDAC1 IC50 ratio 24 h/ 3 h
1a	15	4.9	24	5.3	7.5	1.5
1b	17	4.6	29	6.4	9.9	1.6
1c	4.0	3.5	7.5	12	3.7	1.7
1d	1.6	1.5	3.2	-	1.6	0.77
1e	8.5	5.2	41	3.7	2.4	0.31
1f	4.7	2.6	23	1.6	1.7	0.20
1g	1.7	1.3	1.6	3.3	1.7	1.4
2a	4.9	3.8	12	6.2	1.3	0.41
2b	2.5	6.4	20	4.2	1.1	0.35
2c	2.3	7.1	24	2.2	0.82	0.25
1	7.9	3.4	8.6	3.7	6.7	2.7
2	7.8	2.7	52	4.5	7.5	0.39
SAHA	0.85	ND	0.77	1.0	ND	7.7

In general, the probes exhibited better activity and selectivity for HDAC1 and 2 as compared to HDAC3 and HDAC8 (Table 1). The most HDAC1 and 2 potent probe 1b had an estimated 100- and 1000-fold selectivity for HDAC1 and 2 as compared to HDAC3 and HDAC8, respectively. In the case of HDAC8, no inhibition was observed after 5 min, whereas inhibition of HDAC3 varied from 2% for 1g to 21% for 2c at 10µM is shown in Schemes 1 and 2. The synthesis of the probes 1a–g and 2a–b proceeded through an efficient carbodiimide based coupling reaction between mono-N-Boc protected phenylendiamines 9–11concentration of the inhibitors. After preincubation for 3 h, inhibition of HDAC3 and HDAC8 by the probes varied from 6.5 % for 1g to 56 % for 1b and from 4.7% for 1b to 26 % for 1g, respectively. Similarly to the probes, ligand 1 showed pronounced inhibition of HDAC3 and HDAC8, 96% and 24%, respectively, and ligand 2 inhibited 40% of activity of HDAC3 and only 3% of activity of HDAC8. Despite the similarity of probes 2a–c, 2a did not inhibit HDAC8 at 10 µM, whereas both 2b and 2c inhibited 25% and 22% of activity of HDAC8, respectively.

Scheme 1.

Reagents and conditions: (a) NaNO₂, HCl, NaN₃, 5 h, 0 °C- rt, 85%; (b) SOCl₂, MeOH, 8 h, 0 °C - rt, 87%; (c) K₂CO₃, 2-azidoethyl-4-methylbenzene sulphonate, acetone, 5 h, reflux, 77%; (d) THF/H₂O (1:1), KOH, 10 h, 70 °C, 92%; (e) NaN₃, Sod. ascorbate, CuI, N,N-dimethylethane-1,2-diamine, EtOH/H₂O, reflux, 92%; (f) i: NaN₃, CH₃CN, reflux, 80% ; ii: oxalyl chloride, DCM, 6 h; (g) methyl 4-aminobenzoate, Pyridine, DCM, 0 °C- rt, 86% ; (h) 2N NaOH, THF/H₂O (8:2), 2 h, rt, 93%.

Gangloff et al¹¹ suggested that the time-dependent inhibition in the case of HDAC2 may be explained by the gradual disruption of the internal hydrogen bond between the aniline hydrogen and carbonyl oxygen in the unbound form of the ligand so as to form a bidentate complex with Zn²⁺ ion in the bound form. After a preincubation 3 h, increase in inhibition of HDAC1 and HDAC2 by the probes varied from 1.6-fold for 1d to 17-fold for 1b and from 1.3-fold for 1g to 7.1-fold for 2c respectively (Table 2). After preincubation for 24 h, inhibition of HDAC2 further improved to 1.6-fold for 1g to 41-fold for 1e. A comparison of the IC₅₀ ratios for 3 h vs. 5 min and 24 h vs 5 min for HDAC2 (Table 2) shows that the weakest inhibitors 1c, 1d, and 1g exhibit the least pronounced change in their IC₅₀ with time. A somewhat similar but less pronounced trend is observed in the case of HDAC1. In general, the trends observed in our case seem to be consistent with the explanation for the time-dependent inhibition given by Gangloff et al.¹¹ The difference in the time-dependent inhibition by the probes that have the same substituent binding to the “foot pocket”, e.g. 1a, 1d, 1e, 2a, and 2b, at 3 h vs 5 min and 24 h vs 5 min suggests that additional factors should be taken into account. Overall ability of the ligands to adopt the necessary conformation for induced fit may play a role in addition to the conformational flexibility of the benzamide portion of the ligands. HDAC1 is highly homologous to HDAC2, and, therefore, its time-dependent inhibition may be explained in a similar fashion. However, neither HDAC3 nor HDAC8 were reported to have crystal structures that would contain a binding pocket similar to the “foot pocket” of HDAC2. The docking of the probes to HDAC2 showed that their binding poses are essentially the same as that of ligand 1, i.e. the aniline nitrogen and the amide oxygen form a bi-dentate chelate with Zn²⁺, whereas the bi-aryl portion occupies the “foot pocket” (Fig. 3 and 4).

Figure 3.

Probe 2c docked into the active site of (A) HDAC2, (B) HDAC3, and (C) HDAC8.

Figure 4.

Overlay of compounds 1b (green), 1c (magenta), 1g (cyan) and 2c (gold)in the active site of HDAC2.

A comparison of the docking pose of probe 2c in HDAC2, HDAC3, and HDAC8 shows that, unlike HDAC2 (Fig. 3A), HDAC3 (Fig. 3B) and HDAC8 (Fig. 3C) cannot accommodate 2c such that it can form a bi-dentate complex with Zn²⁺ in the catalytic site. The binding site of HDAC3 in 4A69 is too small for 2c and the probe is mostly resides outside the binding site. In HDAC8 in 1T69, the binding site is too short and has a somewhat different shape compared to HDAC2. None of the docking poses of 2c coordinates with Zn²⁺ despite the proximity of the groups necessary for coordination. After a co-minimization of 2c with the HDAC8, only coordination between the carbonyl oxygen of 2c and Zn²⁺ was observed. Interestingly, although the residues in the foot pocket of HDAC2 and the corresponding residues in HDAC3 (according to sequence alignment) are the same, the recent X-ray apo-structure of HDAC3¹⁷ did not contain a “foot pocket”. Schwabe et al¹⁷ noted that the HDAC3 structure was crystallized in the absence of the ligand and, therefore, may not be representative of the actual protein-ligand complex interactions. In our opinion, the similarity of the time-dependent inhibition of HDAC2 and HDAC3 and HDAC8 suggests that the latter two isoforms may also adopt the conformation with a “foot pocket” that can accommodate the benzamide-based ligands. The relatively low inhibition of HDAC8 compared to HDAC1-3 may be rationalized by the difference in the residues at the entrance to the “foot-pocket” that imposes different steric and electrostatic requirements on the R⁴ substituent. In HDAC8, the opening to the putative “foot pocket” is hindered by the presence of bulky sidechain of Trp127 as shown in Fig. 3, whereas in HDAC1, 2 and 3 the corresponding residue Leu144 is less bulky and more flexible and makes the “foot pocket” more accessible to the ligands. This is also indirectly supported by the SAR - probe 1g is consistently the least active against HDAC1-3 but its inhibition of HDAC8 is comparable to that of 1, 1e, 2b, and 2c.

Cravatt et.al⁵ attributed the low potency of benzophenone based benzamide probes to the positioning of the photoreactive group. Based on their observations, we decided to carry out a small SAR study to explore how the positioning of the aryl azide and aliphatic azide affects the potency and selectivity of the probes. Despite the presence of additional azido groups, probes 1a and 1b were comparable in potency to ligand 1 and probes 2a–c were more than 2–2.5-fold more potent than ligand 2 for HDAC1 and 2. Probes 1c–g were 27–1000 and 5–550-fold less potent than ligand 1 for HDAC1 and 2 respectively. In general, probes 1a–g were found to be less potent than ligand 1 for both HDAC3 and 8, whereas compounds 2a–c and ligand 2 demonstrated comparable potency against HDAC3. In HDAC8, the diazide probes 2b and 2c appear to be more potent than ligand 2 and the monoazide probe 2a was inactive. To gain insights into the plausible explanations for the difference in potency we compared the docking poses of the probes with that of ligand 1. We observed that the meta-substituents R¹ and R³ are too close to the residues Phe210, Gly154, Phe155, Leu276, and Asp104 (Fig. 4).

As a result of this steric interference, the probes are forced to adopt a conformation where the face-to-face π-π stacking between ring B of the probes and Phe155 is disrupted. The loss of these π-π stacking interactions may explain relatively poor potency of mono meta-substituted probes 1e and 1f, 830 and 750 nM respectively, and especially 3,5-disubstituted probes 1c and 1d, 2.4 µM and 10 µM respectively, compared to probes with no meta substituents 1a and 1b, 210 and 110 nM, respectively. The width and shape of the gorge region appears to be important to gain potency and isoform selectivity as demonstrated by Kozikowski et al²¹ in design of tubastatin A, a selective inhibitor of HDAC6. Placement of the aromatic azido group in the “foot-pocket” in 1g led to poor potency for HDAC1 and 2, 55 µM and 77 µM respectively, slightly less pronounced decrease in inhibition of HDAC3 but not HDAC8. The docking showed that the R⁴ azido substituent fits well in the “foot pocket” and occupies the same space as the R⁴ phenyl and 2-thiophenyl substituents in 1a–f. This observation appears to be consistent with the SAR found by Methot et al,²² where non-polar aromatic substituents were found to be preferable compared to polar and/or relatively small substituents R⁴. The additional interactions between the carbamate appendage in probes 2a, 2b and 2c and Tyr209 of HDAC2 identified by docking did not contribute to potency of these ligands, suggesting that this appendage is likely to remain solvent exposed.

Next we investigated whether the newly designed probes are capable of crosslinking HDAC2. Photoaffinity labeling studies were conducted with the probes using commercially available recombinant His-tagged HDAC2. The probes (25 µM) were preincubated with HDAC2 (1.25 µM) for 24 h in photolabeling buffer, exposed to 254 nm UV light for 3×1 min with 1 min resting. A commercially available strained cyclooctyne based biotin tag (BT) was attached to the HDAC2-probe adduct using (3+2) cycloaddition reaction and the biotinylated HDAC2 was visualized by streptavidin-HRP and western blot analysis (Fig. 5). The loading was confirmed by using nickel-HRP, which recognized the His-tag of the recombinant HDAC2 protein.²³ To ensure that the biotinylation was primarily driven through interactions of the probes with the binding site of HDAC2, we performed competition experiments of the probes with a known potent HDAC inhibitor Trichostatin A (125 µM), which has an IC₅₀ of 68 nM for HDAC2.⁸ All of the diazide probes showed a pronounced decrease in biotinylation in the presence of 5-fold molar excess of the competing ligand. The decrease was slightly less pronounced in the case of weakly potent probe 1g.

Figure 5.

Characterization of biotinylated HDAC2 and His-tagged HDAC2 using streptavidin-HRP and nickel-HRP. Western blot analysis of diazide probes 1c, 1d, 1e, 1f, 2b, 2c and 1g (25 µM) photocrosslinked to HDAC 2 (1.25 µM) in the presence or absence of 125 µM of Trichostatin A using streptavidin-HRP and nickel-HRP. Shown is the representative western blot of three independent experiments.

We also confirmed that our probes 1a–f and 2a–c are cell permeable and capable of inhibiting nuclear HDACs by monitoring the acetylation status of histone H4 in MDA-MB-231 breast cancer cell line using previously published procedure.²⁴ H4 is a known nuclear target for HDAC1 and HDAC2 in this cell line.²⁵ All of the probes inhibited deacetylation of histone H4 at 50 µM concentration after a 24 h treatment (Fig. 6A and B)

Figure 6.

Western blot detection of acetyl H4 in MDA-MB-231 cell lines following a 24h treatment with probes at 50 µM (A) Treatment of cells with probes 1a, 1b, 1c, 1d, 1e, suberoyl anilide hydroxamic acid(SAHA) and parent ligand 1.(B) Treatment of cells with probes 1f, 2a, 2b, 2c, suberoyl anilide hydroxamic acid (SAHA) and parent ligand 1. Shown is a representative blot of three independent experiments.

In conclusion, two benzamide scaffolds were successfully explored for design of novel HDAC2 nanomolar potent and selective photoreactive probes suitable for further BEProFL experiments. A total of 10 monoazide and diazide containing benzamide probes were synthesized and tested for their inhibitory activity against class I HDAC isoforms. All the probes are readily accessible in few synthetic steps carried out in a convergent manner. The inhibition was measured at two time points for HDAC1, 3, and 8 and three points for HDAC2. The probes exhibited a 2–40 fold increase in inhibition with respect to time for HDAC1 and 2 and modest increase was observed for HDAC3 and 8. Time-dependent inhibition of HDAC1, 3, and 8 suggests that these isoforms may also adopt the “foot pocket” conformation similar to that of HDAC2 to accommodate the benzamide ligands. The most potent probes exhibit nanomolar activity against HDAC1 and 2. Probe 1b has an IC₅₀ of 70 nM and 110 nM for HDAC1 and 2, respectively, and shows an estimated 100–1000-fold selectivity for HDAC1 and 2 as compared to HDAC3 and 8. The most active diazide probes 2b and 2c have an IC₅₀ of 0.9 and 1.2 µM and 300 and 350 nM for HDAC1 and 2 respectively and show approximately 30-fold selectivity for HDAC1 and 2 as compared to HDAC3 and 8. Docking studies with HDAC2 indicated that the placement of the azido groups meta but not para to the benzamide group in ring B leads to unfavorable for π-π stacking between ring B and Phe155 orientation of the ligand. Consistent with earlier reports, the presence of the bi-aryl moiety in the “foot-pocket” was found to be essential for maintaining potency for HDAC1-3. On the other hand, 1g, a probe that lacks the bi-aryl portion found in 1a–f and 2a–c, was found to be superior to the biaryl-containing probes in inhibiting of HDAC8. As demonstrated by our photolabeling experiments, all the diazide probes efficiently photocrosslinked with recombinant HDAC2. Cell based studies show that the benzamide probes are able to enter the cell nucleus and trigger accumulation of acetylated H4. Presently, the probes are being extensively used in mapping the binding site of HDAC2 via proteomics experiments. Cell based photolabeling experiments are currently underway to understand how these probes bind to HDAC complexes in cells.