|Home | About | Journals | Submit | Contact Us | Français|
Histone deacetylase (HDAC) inhibitors are powerful tools in understanding epigenetic regulation and have proven especially promising for the treatment of various cancers, but the discovery of potent, isoform-selective HDAC inhibitors has been a major challenge. We recently developed a cyclic α3β-tetrapeptide scaffold for the preparation of HDAC inhibitors with novel selectivity profiles [Montero, A.; Beierle, J. M.; Olsen, C. A.; Ghadiri, M. R. J. Am. Chem. Soc. 2009, 131, 3033]. In this study, we elaborate this scaffold with respect to side chain diversity by synthesizing one-bead–one-compound combinatorial libraries of cyclic tetrapeptide analogs and applying two generations of these focused libraries to the discovery of potent HDAC ligands using a convenient screening platform. Our studies led to the first HDAC6–selective cyclic tetrapeptide analog, which extends the use of cyclic tetrapeptides to the class-II HDAC isoforms. These findings highlight the persistent potential of cyclic tetrapeptides as epigenetic modulators and possible anticancer-drug lead compounds.
Histone deacetylases (HDACsa) are a family of enzymes found in bacteria, fungi, plants, and animals that profoundly affect cellular function. The natural substrates of most HDACs are believed to be Nε-acetyl lysine residues in the tails of histones H2A, H2B, H3 and H4 [supporting information (SI) Figure S1]. The degree of acetylation of these sites affect the packing of nucleosomes in chromatin complexes as well as the recruitment of transcription activator- and repressor-proteins to histone tails, which in turn modulate the extent of gene transcription.1–4 The HDAC enzymes are not restricted to deacetylation of histones, however, and are known to regulate the function of other proteins in the cell.5 For example, HDAC6 has been associated with deacetylation of α-tubulin6 as well as peroxiredoxin I and II,7 and has been shown to rescue neurodegeneration in Drosophila melanogaster.8 To date, eleven Zn2+-dependent human HDAC enzymes have been identified and classified according to their sequence similarity [class-I: HDACs 1, 2, 3, and 8; class-IIa: HDACs 4, 5, 7, and 9; class-IIb: HDACs 6 and 10; class-IV: HDAC11]. Class-III HDACs comprise a series of NAD+-dependent enzymes (Sirtuins 1–7).
Although the majority of HDAC inhibitors in clinical trials are non-selective,9, 10 small-molecule HDAC inhibitors have proven to be promising candidates for therapeutic intervention, especially as tumor suppressors for potential cancer chemotherapies.3, 11–15 Indeed, several candidates have advanced to clinical trials9 and Vorinostat/SAHA was recently approved by the FDA for the treatment of cutaneous T-cell lymphoma.16 Non-selective HDAC inhibitors have also recently been shown to elicit positive effects related to neurological disorders.17–20 Despite these successes with non-selective HDAC inhibitors, the ability to selectively perturb individual HDACs with small molecules would be exceedingly useful in unravelling the complex and highly dynamic network of HDAC signaling and in the design of new and safer drug candidates. However, structure-based efforts to design class- or isoform-selective inhibitors have been hampered by the limited structural information available, which currently comprise cocrystal structures of “HDAC-Like Protein” (HDLP),21 HDAC8,22, 23 and the catalytic domain of HDAC724 with TSA as well as the catalytic domain of HDAC4.25 Although the class-IIa HDACs have considerably lower intrinsic deacetylase activity compared to class-I HDACs against standard substrates,4, 26–28 class-IIa HDACs play pivotal roles in numerous pathways, and they are therefore equally important targets for future selective therapeutic intervention in various diseases.4, 29–32
Nature provides a number of related cyclic scaffolds with HDAC inhibitory activity, including non-ribosomal desipeptides,33 the recently reported marine natural product largazole,34–39 and tetrapeptide natural products such as the trapoxins,40, 41 HC toxins,42, 43 chlamydocin,43 apicidins (1),44–46 and the azumamides (2)47–51 (Figure 1). Numerous analogs of these natural compounds have been investigated as well.52–63 For the medicinal chemist, cyclic tetrapeptides built from all α-amino acids often present significant challenges as drug targets due to poor macrolactamization yields for closing the 12-membered ring and multiple three-dimensional conformations on the NMR time scale.64–66 We were recently able to minimize these shortcomings by developing synthetic HDAC inhibitor scaffolds (3, Figure 1) that have an α→β-amino acid substitution in the backbone to give a 13-membered ring.67 By systematically changing the chirality of the amino acids and the position of the β-amino acid we optimized the scaffold for inhibition of HeLa extract HDAC activity. Moreover, we correlated compound potencies with their high-resolution NMR structures, which allowed us to construct a three-dimensional pharmacophore model. In our previous study, the identities of the side chains were fixed so that we could focus on understanding the effects of peptide backbone modifications on structure and potency. In contrast, here we investigate a variety of side chain functionalities by designing and carrying out total solid-phase synthesis (SPS) of one-bead–one-compound libraries68 of cyclic peptides for the purpose of inhibiting HDAC enzymes. The present study focuses on HDACs with potent deacetylase activity against histones and standard substrates, i.e., class-I HDACs 1, 3, and 8 as well as class-IIb HDAC6. HDAC2 was omitted from this investigation due to the high degree of sequence similarity to HDAC1.
To guide the design of the first-generation library, we first surveyed the effect of the Zn2+-coordinating group and its distance from the peptide core on our selection of HDAC enzymes. Although we had previously established that changing the Zn2+-coordinating functionality in peptides 3a–c accommodated leaps in potency using HeLa cell nuclear extract,67 we wished to determine more specifically how the inhibitors acted against the individual HDAC isoforms. We therefore tested 3a–c against our panel of recombinant human HDACs as well as against a HeLa cell cytosolic extract (Table 1). The inhibition trend described previously for 3a–c against HeLa nuclear extract (hydroxamic acid > ethylketone > acid Zn2+-coordinating functionality) was also observed for the cytosolic extract as well as HDACs 1 and 3. The IC50 values indicated that the predominant source of HDAC activity in HeLa nuclear extract is HDAC1, while HDAC3 seems to be more dominant in HeLa cytosolic extract, which corresponds well with the fact that HDAC3 is known to shuttle into the cytoplasm.69 Furthermore, the observed IC50 values indicated that the requirements for inhibition of HDAC8 are different from the other HDACs in class-I, in agreement with previous findings.10, 22, 23 Interestingly, the ketone 3b, like apicidin, did not inhibit HDAC6 in the concentration range tested, while acid 3a proved to be a micromolar inhibitor of HDAC6. This might be explained by an interaction between the carboxylate of 3a and a free binding site His in HDAC6 (Figure S2). For comparison, HDAC8 contains an active site His residue (His143) that forms a contact with an adjacent Asp (Asp183). Although HDAC6 does contain a corresponding active site His residue, the corresponding Asp is absent, so it is possible that the active site His in HDAC6 is free to interact with the carboxylate in 3a (Figure S2). This finding was encouraging, since the carboxylic acid Zn2+-coordinating group would provide a useful handle for resin anchoring to synthesize focused libraries of inhibitors via efficient and robust chemical SPS techniques.
To reduce the synthetic effort required for assessing the effect of different lengths of the Zn2+-coordinating side chain, we designed a series of alkylated cysteine-derived analogs (4a–d), which enabled evaluation of the distance requirements for the different enzyme isoforms (Figure 1 and Figure S3). The use of alkylated cysteine containing compounds has precedence in the literature for probing HDACs.70,71 It was previously shown that changing the linker length by plus/minus one methylene unit significantly decreased the potency of inhibition of HDAC activity in HeLa cell extract;52 however, the effect on individual HDAC isoforms was not investigated in that study. Furthermore, largazole analogs with different linker lengths have recently been investigated.61 With the sulphur atom in compound 4a extending the linker length by only ~0.28 Å compared to the methylene in 3b it was not surprising to find that this compound behaved similarly to 3b. The corresponding compound with a linker of one fewer methylene unit, 4b, showed total loss of activity. Likewise, the shortened linkers in the carboxylic acids 4c and 4d also rendered these peptides less potent than their parent compound (3a) across our selection of enzymes. It may be noted, however, that a comparison of compounds 3a vs. 4c reveals a ~40-fold decrease in potency against HDAC1 and only a ~3-fold decrease against HDAC8. The length of the linker may thus be a factor to consider when designing isoform selective HDAC ligands. The design of so-called “linker-less” hydroxamic acid HDAC inhibitors with selectivity for HDAC8 was recently reported,72 supporting the suggestion that shortened linkers might prove effective against this isoform. Partial decomposition of these thioether compounds was observed upon prolonged storage (presumably due to oxidation of the thioether); nonetheless, they are useful as a discovery tool giving easy access to a variety different linker lengths and Zn2+-coordinating moieties.
Our first-generation library comprised a focused 252-member one-bead–one-compound library aimed at exploring the importance of the side chain functionalities for potent HDAC inhibition. Based on our preliminary studies and to facilitate solid phase synthesis, we opted to employ the carboxylic acid Zn2+-coordinating residue for this library. Albeit not the most potent Zn2+-coordinating moiety, the carboxylic acid was considered more appropriate than the hydroxamic acid, since the latter gave rise to poor isoform selectivity for the cyclic tetrapeptides tested.52 The ethylketone containing ligands seemed ideal for inhibition of class-I HDACs 1–3, but showed no inhibitory activity against HDAC6, and the total solid-phase synthetic route to give the ethylketone containing ligands required post cleavage purification (Figure S4), which was not ideal for a high-throughput library screening scheme.
Previous SAR studies involving cyclic tetrapeptides as HDAC inhibitors have indicated that the aromatic residue vicinal to the Zn2+-coordinating amino acid (position aa2 in Figure 2) is important for potent inhibition.58, 59 We therefore decided to probe this position extensively in our initial library. Aromatic residues of varied bulkiness and polarity as well as non-aromatic hydrophobic residues were included. For position aa3, four different β3-amino acids were chosen, including a β3-Phe residue to mimic the Phe side chain found in trapoxin and chlamydocin. Residues commonly found in natural product HDAC inhibitors (e.g., Aib, D-Pip, Pro, and Val) were included along with Phe at position aa4. To synthesize the library, the first residue (N-Fmoc-(S)-Asu-O-Allyl) was anchored to a macrobead 2-chlorotrityl solid support via its carboxylic acid side chain using protocols previously described developed for regular SPS resins.67, 73 The choice of macrobeads as solid support afforded facile handling during the split–pool synthesis procedure and enabled an economical small-scale preparation of each compound while still affording enough material to test each product in several assays in 96-well microtiter plates (Figure 2). After the total synthesis of the 252-member library of ligands including on-resin cyclization, individual beads were arrayed in 96-well plates, cleaved by treatment with a TFA containing cocktail, and dissolved in DMSO to give stock solutions. After screening the library for HDAC inhibitory activity using HeLa nuclear extract, the obtained hits were analyzed and sequenced using an LC-MS/MS procedure previously developed in our lab.74
From the initial screen and retesting of the first-generation library for inhibition of HDAC activity in HeLa cell extract, we obtained eight confirmed hits (six different structures) as shown in Figure 3A. Semi-logarithmic plots were obtained from the inhibition data at two different dilutions of the initial stock solutions, from which approximate IC50 values for these peptides were calculated using GraphPad Prism. Interestingly, the two most potent compounds (8a and 13) closely resembled our parent compound (3a), having only a single substitution in the aromatic residue at the aa2 site (Trp→naphthylalanine in 8a or Trp→Phe in 13). Furthermore, one of the hits with intermediate potency (compound 12) contained both the aa2–Trp and the aa4–D-Pip residues as found in the natural product apicidin A (1b). The abundance of the various side chains by position on the scaffold (Figure 3B) revealed that the aa2 position required an aromatic side chain as anticipated; the naphthylalanine residue, which is similar to Trp in size, was the most abundant in the hit series. At position aa3 the β3Leu residue was more abundant than the aromatic residues included in the library, and at position aa4 no consensus could be derived, indicating that the identity of the side chain at this position is less significant for inhibition of HeLa cell HDAC activity.
To confirm the results from the library screening, the four identified naphthylalanine containing compounds (8a, 9–11) were resynthesized on ~50 μmol scale, purified, and retested to obtain IC50 values against HeLa nuclear extract (Table 2). Their IC50 values confirmed the approximated values from the initial screening data, supporting that the library screening format is appropriate for the discovery of novel HDAC inhibitors. To derive more detailed SAR information from this initial hit series, compound 14 was also prepared along with the ethylketone analog (8b) of compound 8a (Figure S5 and Table 2). The unnatural naphthylalanine residue was confirmed to be a good alternative to Trp at position aa2. Furthermore, the relative IC50 values for carboxylic acid (8a) vs ethylketone (8b) showed the expected trend, as ethylketone 8b was 10–30-fold more potent against HeLa extract and HDACs 1 and 3. A decrease in potency was observed for HDAC6, providing further evidence that ketones are not favorable for inhibition of this enzyme isoform. The SAR also suggested that aromatic side chains at position aa3 are not favorable for inhibition of class-I HDACs (9 vs 10 and 11 vs 14). Interestingly, compound 10 was two-fold more potent against HDAC1 compared to our parent carboxylic acid (3a) and was somewhat less potent against HDAC3, thus providing a ~10-fold selectivity index for HDAC1 vs HDAC3 as well as potent inhibition of HDAC1 (68 nM). The substitution of the Ala→Phe residue in position aa4 may therefore give rise to an increased selectivity for HDAC1.
Compounds 15 and 16, in which the least structurally restrictive residue (aa4) was substituted with Nε-Ac-lysine or Aoda, respectively, were also prepared and tested (Table 2) so that we could ascertain how such substitutions would influence binding to the HDACs before designing a second-generation library. The Lys(Ac)–aa4 residue in compound 15 was inspired by the two consecutive lysine residues K381 and K382 known to be involved in acetylation/deacetylation events in the transcription factor p53. Moreover, the Lys(Ac)-Lys(Ac) motif has been used in a commercially available fluorescent HDAC8 substrate [His-Arg-Lys(Ac)-Lys(Ac)-AMC, www.biomol.com], which led us to hypothesize that such a motif might provide an improved degree of recognition by HDAC8. The Aoda–Aoda sequence in 16 provides a double isosteric version of this feature. Indeed, improved activity against HDAC8 was observed for compound 15 (150 nM) and to a much lesser extent for 16 (1,400 nM); however, the IC50 values for inhibition of HDACs 1 and 3 were still in the same range as observed for parent compounds 3b and apicidin (Table 2).
Based on the SAR information gathered from the first generation library, we decided to focus on the diversification of the aa3 and aa4 positions in the second-generation library in the hopes of reducing bias for class-I HDAC inhibitory activity. In addition to amino acids found in the various non-ribosomal cyclic tetrapeptide inhibitors and their analogs (mostly hydrophobic residues), a number of hydrophilic residues (i.e., β3Ser, Thr, Arg, and Lys) were included due to the presence of these side chains in the histone tails (Figure S1). Aspartic acid was also included due to its presence vicinal to an Nε-acetylated lysine in α-tubulin, which is known to be a substrate for HDAC6.
A 525-member second-generation library (Figure 4) was next synthesized on macrobeads as described for the initial library. The compounds were initially screened for HDAC inhibition activity in HeLa cell extract, and then both libraries were counter-screened against HDAC6. As before, the hits were confirmed by retesting the stock solutions in duplicate as well as performing LC-MS/MS sequencing and MALDI-TOF MS, which furnished another nine hit structures (for full hit series, see Figure S6 and Figure S7). All the hits contained either Trp or naphthylalanine at the aa2 position. In addition to the results already discussed for the first generation library above, Tyr at the aa3 position was observed in a number of hits, while Phe(F5) or Arg appeared at the aa4 position. Unfortunately, no hits against HDAC6 were observed that were not also inhibitors of HeLa HDAC activity. Three peptides that were hits against both HeLa extract and HDAC6, all of which had an Arg residue at the aa4 position, were resynthesized (compounds 17–19) along with two analogs having either a Phe(F5) or a Lys(Ac) residue at the aa4 position, respectively (compounds 20, 21) [Figure 5 and Table 2]. The obtained IC50 values revealed that the unnatural Phe(F5) residue, which was present in a number of hits from the HeLa HDAC activity screening, resulted in low potency of compound 21 over the whole selection of enzyme isoforms tested, so other hit structures containing this residue were not pursued further.
The compounds differing only in the identity of the residue at position aa3 (17–19) revealed interesting SAR information. Notably, the compound having the β3Tyr–Arg motif at positions aa3–aa4 (19) was the most potent carboxylic acid-containing inhibitor of HDAC6 identified so far, and had also the best selectivity index (HDAC3/HDAC6 = 0.65) in the series, albeit still in favor of the class-I HDACs. Since HDAC6 is known to deacetylate α-tubulin, we decided to combine the β3Tyr–Arg motif with an Asp residue at the aa2 position to give compounds 22 and 23. Interestingly, a successful reversal of the isoform selectivity pattern was achieved with compound 23, as the selectivity indices for HDAC1/HDAC6, HDAC3/HDAC6, and HDAC8/HDAC6 were approximately 3, 4, and 6, respectively, with potent inhibition of HDAC6 (IC50 = 39 nM). Compound 23 is, to the best of our knowledge, the first cyclic tetrapeptide analog exhibiting isoform selectivity toward a class-II HDAC. Moreover, its selectivity index is comparable to that observed for tubacin (HDAC1/HDAC6 selectivity index of 7),75–77 and its potency is in the nanomolar range as recently reported for a series of thiolate containing inhibitors78, 79 inspired by an HDAC6 selective substrate.80,81 More potent (picomolar) HDAC6 selective inhibitors containing a phenylisoxazole surface recognition group have also been reported recently.82 The ethylketone analog 24 containing the aa4–Phe residue as observed in compound 10 was prepared in the hopes that higher potency against HDAC1 could be obtained while still retaining the selectivity profile discussed for 10, but peptide 24 was unfortunately too hydrophobic and hence insoluble under the assay conditions. As expected based on previously observed trends for hydroxamic acid compound 3c, the hydroxamic acid containing homolog 25 was a potent inhibitor of both classes of HDACs.
To derive more detailed SAR related to compounds 17, 19, and 20, we also prepared peptides 26 and 27 containing the Lys–aa4 and the Lys(Ac)–aa4 residues, respectively. The potencies of these compounds confirm that an aromatic residue at position aa3 results in decreased potency against the class-I HDACs, as previously observed for 9–14. Furthermore, the results show that substitution of the Lys(Ac)–aa4 residue (27) for an Arg–aa4 (17) or a Lys–aa4 residue (26) had little effect on the inhibitory activities measured against the entire selection of HDACs. It was recently reported that an Arg residue vicinal to the Lys(Ac) promoted binding of synthetic histone substrates to HDAC8,83 which also corresponds well with the observation that most of the lysine residues known to participate in acetylation/deacetylation events in the tails of histones H3, H4, and H2A have Arg residues in their vicinity (Figure S1). Taking into account that compounds 17, 26, and 27 are also 4–5 fold more potent against HDAC3 and 3–4 fold more potent against HDAC6 when compared to their Ala–aa4 analog (8b), our results suggest that the presence of an Arg residue vicinal to the Zn2+-coordinating amino acid may also promote binding to other HDAC isoforms (i.e., HDACs 3 and 6 in this study). To check if the Lys(Ac) residue in 27 was deacetylated under the conditions of our assay, we carried out the standard HeLa extract assay with this compound (at concentrations of 10 μM or 1 μM), but instead of the trypsin development step, we quenched the reaction with MeOH and analyzed the reaction mixtures by LC-MS. These experiments showed the peak expected for compound 27, but no trace of the corresponding deacetylated analog (26).
Finally, the composition of our most potent inhibitor of HDAC8 (compound 15) was combined with the cysteine derived Zn2+-coordinating amino acid containing one methylene unit fewer than Asu (as in 4c) to give peptide 28. Although the HDAC3/HDAC8 selectivity was reversed, the potency against HDAC8 was not considered satisfactory. Efforts to design cyclic tetrapeptide inhibitors with improved potency as well as selectivity for HDAC8 are now in progress in our laboratory. The SAR information obtained is summarized in Figure 6.
To assess the effectiveness of our novel inhibitors in cells, we treated Jurkat cells (T-cell leukemia) with a selection of compounds (apicidin 1a, 10, 15, and 23) for 24 hours and western blotted for histone H3, acetylated histone H3 (K9+K14), and acetylated α-tubulin as previously described.84 The presence of acetylated tubulin is an indicator of HDAC6 inhibition, whereas the presence of acetylated histone H3 indicates inhibition of the class-I HDACs. Similarly, as expected based on the IC50 values obtained with the recombinant enzyme isoforms, the two ketone containing compounds (apicidin and 15) exhibited potent histone deacetylase activity and only minor tubulin deacetylase activity. We were also pleased to find that library hit 10, containing the carboxylic acid Zn2+-coordinating group, affected the acetylated histone levels in the cells compared to the DMSO control, whereas the effect on acetylated α-tubulin levels was considerably less pronounced in relation to the corresponding DMSO control. As expected considering the in vitro enzyme inhibition data, 10 was less potent than both apicidin and 15 (Figure 7). Finally, the HDAC6 selectivity of compound 23 was confirmed by the presence of relatively high levels of acetylated α-tubulin, with comparably low levels of histone H3 acetylation.
Growth inhibition of various cancer cell lines by selected compounds was next determined using the XTT assay (Table 3).67, 85 Except for the highly potent hydroxamic-acid-containing broad spectrum HDAC inhibitors (3c and 25), which were also broadly cytotoxic, the general trend seemed to be that cytotoxicity correlated largely with potent HDAC1 inhibition. Our most potent carboxylic acid HDAC inhibitor (10) did not potently inhibit the growth in any of our series of cancer cells, which is in accord with the relatively weak inhibition of HDAC activity in Jurkat cells compared to compound 15 and apicidin indicated by the western blot. Interestingly, compound 15, which is an analog of 3b with slightly lower potency against HDAC1 (~2-fold) and significantly higher potency against HDAC8 (~15-fold), was less cytotoxic than 3b towards HeLa cells, indicating that HDAC8 may not be important for cell survival in this cell line. Moreover, compound 23, which is a >250-fold more potent inhibitor of HDAC6 than 3b and only ~5-fold less potent against HDAC1, was less cytotoxic towards HeLa cells than 3b, indicating that HDAC6 is less important than HDAC1 for cell survival. Clearly, these findings require further support, and differences in cell permeability may also play a major role in the different cytoselectivity profiles observed. In Jurkat cells, for example, compound 8b proved slightly more potent than the hydroxamic acid inhibitor 3c, even though 3c is a more potent inhibitor of all the HDACs tested in vitro. This observation may be explained by an expected more favorable logP value for compound 8b. The trend observed for these compounds against HeLa, K-562, and KYO-1 cells seem to better reflect the HDAC inhibitory activities, confirming that several factors play important roles in the growth inhibition process. Taken together with the western blots, these results show that our compounds may find use as specific small molecule probes to regulate histone and/or tubulin acetylation at sub-cytotoxic concentrations in cells. Finally, we were pleased to find that by reintroducing the strong Zn2+-coordinating hydroxamic acid moiety in one of the hit compounds to give 25, we were able to obtain potent growth inhibition in a number of cell lines. This provides further evidence that the effects of the cyclic peptide core and the Zn2+-coordinating moiety on potency may be considered somewhat additive in this class of compounds.
A synthetic tetrapeptide scaffold having an α→β-amino acid substitution in the backbone has been employed for the identification of moderately selective inhibitors for the HDAC isoforms 1 and 6. Using an all (S)-configured α3β ring type as the lead compound, a semi-high-throughput screening scheme was developed for assaying ligands from focused one-bead–one-compound libraries. The present study shows that this screening platform in combination with structural information from natural as well as synthetic HDAC substrates and inhibitors is useful for the discovery of new HDAC inhibitors. The obtained SAR information revealed a range of potent compounds displaying novel side chain patterns and containing unnatural amino acid residues. Depending on the HDAC isoform in question, different amino acid positions on the scaffold proved more or less important for potency, which will aid the future design of selective ligands, help identify optimal sites for labeling, and enable the design of efficient probes for activity-based proteomics or affinity matrix pull down studies.41, 86, 87
Our data also indicate that the cyclic tetrapeptide motif not only provides a highly modular scaffold, but also that this class of compounds can be further modulated by changing the Zn2+-coordinating moiety in order to finetune potency, HDAC isoform selectivity, or cytoselectivity. Importantly, compounds exhibiting potent HDAC inhibitory activity and isoform selectivity were identified and shown to work in cell culture. It is remarkable that from this relatively small selection of compounds it was possible to design ligands with completely reversed selectivity profiles as compared to the known natural product tetrapeptide inhibitors. We therefore envision that this discovery platform may find additional use for the design of class- and isoform-selective HDAC inhibitors targeting other enzyme isoforms. It will be interesting to explore trends in cytoselectivity between different cell lines in greater detail with these ligands, because such studies will help delineate the significance of individual HDACs in various cancers. This knowledge may aid the design of targeted drug candidates with lower toxicity and fewer side effects.
HDAC1, HDAC3–NCoR2, HDAC6, and HDAC8 were purchased from BPS Biosciences (San Diego, CA). Trypsin (TPCK treated, from bovine pancreas; 12,500 units/mg) was from Sigma-Aldrich (Milwaukee, WI). 2-Cl-trityl chloride resin macrobeads were from Peptides International (Louisville, KY), Weinreb amide resin was from Novabiochem (San Diego, CA), and 2-chlorotrityl N-Fmoc-hydroxylamine resin was from Sigma-Aldrich (Milwaukee, WI). β-Amino acids were purchased from Peptech Corporation (Burlington, MA). All α-amino acids, chemicals, and solvents were used as received from suppliers. HDAC assay buffer was prepared as described in the Biomol International protocol AK-503 [Tris/Cl (50 mM), NaCl (137 mM), KCl (2.7 mM), MgCl2 (1 mM), pH 8.0]. Histone H3 antibody was from Abcam (Cambridge, MA), acetylated K9+K14 histone antibody was from Upstate (Temecula, CA), and acetylated α-tubulin antibody was from Sigma (St. Louis, MO).
N-Fmoc-(S)-2-Asu(t-Bu)-OAllyl (3 equiv with respect to the resin) was stirred in TFA–CH2Cl2 for 1 h,73 the solvents were evaporated in vacuo, and the residue was co-evaporated several times with CH2Cl2–hexane and dried for 2 h under high vacuum. The TFA-free residue was taken up in CH2Cl2 (2 mL/100 mg) and i-Pr2EtN (20 equiv), and was shaken with 2-Cl-trityl chloride macrobeads in a fritted syringe for 16 h. The resin was washed with DMF, MeOH, and CH2Cl2 (3× each), and the loading was determined by UV measurement (290 nm) of the chromophore in an Fmoc deprotected sample (Novabiochem procedure). The bulk of the resin was then Fmoc deprotected with DMF–piperidine (3:1, 2 × 20 min), washed with DMF, MeOH, and CH2Cl2 (3× each), and dried in vacuo. Three cycles of combinatorial split–pool Fmoc peptide synthesis were then performed in 3 mL fritted syringes, applying the Fmoc protected amino acids (5 equiv), DIC (5 equiv), HOBt (5 equiv), and bromophenol blue (2 equiv) as internal indicator in each coupling step.88 Standard Fmoc deprotection and washing protocols were applied. After coupling of the fourth amino acid, the resins were washed with DMF, MeOH, CH2Cl2, and CHCl3–N-methylmorpholine degassed with argon (3× each). The allyl group was then cleaved with Pd(Ph3P)4 (0.2 equiv) in degassed CHCl3–N-methylmorpholine (4 mL) by shaking for 16 h in the dark, after which time the resin was washed with DMF, 1% sodium dimethylthiocarbamate in DMF, and then neat DMF again (2× each), followed by Fmoc deprotection and standard washing. Finally, the resin-bound peptides were cyclized (2 × 12 h) with HATU (5 equiv) and i-PrEt2N (20 equiv) in DMF (3 mL), washed extensively, and dried under high vacuum. The beads were then divided into individual wells in round-bottomed 96-well polypropylene microtiter plates (Fisherbrand), leaving a column free for the standards and controls during screening. To each well was added ~200 μL of a cleavage cocktail [TFA–CH2Cl2–i-Pr3SiH–H2O (48:48:2:2)] and the plates were left for 16 h. Residual solvents were removed by placing the plates under vacuum in a desiccater for 24 h, before DMSO was added to give stock solutions (1 mM theoretical concentration based on the initial loading and beads/mg) of each peptide for screening and sequencing.
Compounds were prepared on solid phase with cyclization on the resin (for hydroxamic acids) or in solution (for ethylketones) as previously described.67 Linear precursors for carboxylic-acid-containing pseudopeptides were prepared by standard Fmoc SPPS and cleaved using 2,2,2-trifluoroethanol–HOAc–CH2Cl2 (2:2:6) in order to keep the side chain protecting groups in place until after cyclization in solution. All compounds were purified to homogeneity by preparative reversed-phase HPLC (C-18 column) and the obtained fractions were lyophilized to give the compounds as fluffy solids that were characterized by LC-MS and high resolution ESI-TOF MS (Table 4) as well as 1H NMR for selected compounds. The purities of the compounds were determined by analytical HPLC with UV detection at 230 nm, using a Phenomenex Luna 150 × 4.6 mm C18(2) column (5μ) eluted at a rate of 1.5 mL/min. Injection volumes were 20 μL of approximately 1 mg/mL solution. The gradient elution system consisted of eluent A (H2O–MeCN–TFA 99:1:0.1) and eluent B (MeCN–H2O–TFA 90:10:0.07) rising linearly from 0% to 100% of B during 30 min. All cyclic peptides were ≥95% pure, with the exceptions of compounds 4d, 9, 23, and 27, which were ≥90% pure. The HPLC purity of each compound is given in Table 4. DMSO stock solutions were prepared using UV [ε(Trp) = 5690 M−1 × cm−1 (280 nm) and ε(Tyr) = 1280 M−1 × cm−1 (280 nm)] to adjust the concentrations. Stock solutions of compounds lacking these aromatic chromophores were prepared by weight.
The IC50 values against HeLa extracts were determined using the standard Biomol HDAC fluorimetric assay protocol (AK-503).85 For inhibition of recombinant human HDACs the dose–response experiments with internal controls were performed in black or white low binding NUNC 96-well microtiter plates. The dilution series were prepared in Milli-Q water from 1 mM DMSO stock solutions. The appropriate dilution of the inhibitor (10 μL of a 5× concentration solution) was added to each well followed by HDAC assay buffer (37.5 μL) containing Ac-Lys(Ac)-AMC substrate (67 μM) and bovine serum albumin (0.75 mg/mL). For HDAC8 the Arg-His-Lys(Ac)-Lys(Ac)-AMC substrate (27 μM) was used. Finally, a solution of the appropriate HDAC (2.5 μL) was added and the plate was incubated at 37 °C for 30 min in the dark [HDAC1, 0.09 μg/μL; HDAC3, 0.03 μg/μL; HDAC6, 0.011 μg/μL; HDAC8, 0.4 μg/μL]. Then trypsin (50 μL, 0.4 mg/mL) was added and the assay development was allowed to proceed for 15 min at RT, before the plate was read using a Tecan GENios plate reader with excitation at 360 nm and detecting emision at 460 nm.
Dilutions of 10 μM and 1 μM were prepared from each of the 1 mM DMSO stock solutions, and 10 μL of each were added to white NUNC assay plates. In each plate, standards and controls consisted of duplicate wells of: 1) no enzyme (blank); 2) 1 μM of inhibitor 3a; 3) 1 μM of apicidin; or 4) no inhibitor (max read-out) were installed in each plate. For screening of HDAC6 we used TSA (0.1 μM) instead of apicidin as one of the internal standards. Each plate was assayed as described above for the purified peptides, and compounds that were considered hits, i.e., satisfied our criteria for inhibition at both concentrations, were rediluted and retested in duplicate to weed out false positives. Hits that also satisfied this round of screening were then sequenced using automated LC–MS/MS as previously described,74, 89–91 and parent ion masses were confirmed by either negative mode sonic spray ionization (SSI) MS or MALDI-TOF MS.
Jurkat cells were cultivated in RPMI 1640 medium supplemented with 10% fetal bovine serum and antibiotics. At confluency, cells were seeded in six-well plates and incubated with various concentrations of the inhibitor compounds at 37 °C for 24 h. The cells were then washed, harvested, and lysed. The lysates were run on 4–12% polyacrylamide SDS page gels (Invitrogen), transferred to nitrocellulose membranes, and visualized as described previously.84 TSA and the vehicle (DMSO) were used as positive and negative controls, respectively.
Cells were cultivated in flat 96-well tissue culture plates in 90 μL of medium supplemented with 10% FBS and antibiotics. HeLa cells were seeded at a density of 5,000 cells/well, Jurkat and MCF-7 at a density of 10,000 cells/well, and K-562 and KYO-1 at a density of 15,000 cells/well. Twenty-four hours later, 10 μL of medium containing various concentrations of the desired compounds were added in triplicate, and the cells were incubated at 37 °C for 72 h in a 5% CO atmosphere. Cell survival was determined using the XTT (Sigma-aldrich, St. Louis, MO) colorimetric assay. Immediately prior to use, a mixture of XTT (1 mg/mL) and PMS (N-methyl dibenzopyrazine methyl sulfate, 0.383 mg/mL) in phenol red free RPMI was prepared (0.1 mL of PMS per 5 mL of XTT solution). After adding 50 μL of this mixture to each well, plates were incubated at 37 °C for 2 h in the case of HeLa and MCF-7 cells, and 4 h for Jurkat, K-562, and KYO-1 cells. Plates were shaken to evenly distribute the dye in the wells and the absorbance was measured using a Tecan GENios plate reader spectrometer at a wavelength of 485 nm.
We thank the TSRI Mass Spectrometry Facility for high resolution mass spectral analyses, Dr. Ana Montero for cell lines and HeLa cell extracts, Mr. Andrei Loutchnikov and Mr. Devon Cayer for MCF-7 and Jurkat cell lines, respectively, Prof. Joel M. Gottesfeld, Dr. James C. Chou, and Dr. David Herman for helpful discussions and assistance with western blot analyses, as well as Dr. Luke J. Leman for fruitful discussions and invaluable comments on this manuscript. C.A.O. thanks the Benzon Foundation and the Lundbeck Foundation for postdoctoral fellowships as well as the Danish Independent Research Council for a Young Researcher’s Award (274-06-0317). This work was supported in part by a grant from the National Institute of General Medical Sciences (GM52190) and the Skaggs Institute for Chemical Biology.
aAbbreviations: AMC, 7-amino-4-methylcoumarin; Aoda, (S)-2-amino-8-oxo-decanoic acid; Asu, (S)-2-amino-suberic acid; DIC, N,N′-diisopropylcarbodiimide; FBS, fetal bovine serum; HATU, 2- (1H-azabenzotriazol-1-yl)-1,1,3,3,-tetramethyluronium hexafluorophosphate; HDAC, histone deacetylase; HDLP, HDAC-like protein; HPLC, high performance liquid chromatography; MALDI-TOF, matrix-assisted laser desorption/ionization time-of-flight; MS, mass spectrometry; NAD, nicotinamide adenine dinucleotide; Nal, 1-naphthylalanine; NCoR2, nuclear co-receptor 2; NMR, nuclear magnetic resonance; Pip, pipecolic acid; PMS, N-methyldibenzopyrazine methyl sulfate; SAR, structure–activity relationship; SPPS, solid-phase peptide synthesis; SPS, solid-phase synthesis; TSA, trichostatin A; XTT, 3′-[1-(phenylaminocarbonyl)-3,4-tetrazolium]-bis-(4-methoxy-6-nitro)-sulphonic acid.