CHEMMEDCHEM COMMUNICATIONS DOI: 10.1002/cmdc.201300424
Tailoring Small Molecules for an Allosteric Site on Procaspase-6 Jeremy Murray,*[a] Anthony M. Giannetti,[a] Micah Steffek,[a] Paul Gibbons,[a] Brian R. Hearn,[b] Frederick Cohen,[a] Christine Tam,[a] Christine Pozniak,[a] Brandon Bravo,[a] Joe Lewcock,[a] Priyadarshini Jaishankar,[b] Cuong Q. Ly,[a] Xianrui Zhao,[a] Yinyan Tang,[b] Preeti Chugha,[b] Michelle R. Arkin,[b] John Flygare,[a] and Adam R. Renslo*[b] Although they represent attractive therapeutic targets, caspases have so far proven recalcitrant to the development of drugs targeting the active site. Allosteric modulation of caspase activity is an alternate strategy that potentially avoids the need for anionic and electrophilic functionality present in most activesite inhibitors. Caspase-6 has been implicated in neurodegenerative disease, including Huntington’s and Alzheimer’s diseases. Herein we describe a fragment-based lead discovery effort focused on caspase-6 in its active and zymogen forms. Fragments were identified for procaspase-6 using surface plasmon resonance methods and subsequently shown by X-ray crystallography to bind a putative allosteric site at the dimer interface. A fragment-merging strategy was employed to produce nanomolar-affinity ligands that contact residues in the L2 loop at the dimer interface, significantly stabilizing procaspase6. Because rearrangement of the L2 loop is required for caspase-6 activation, our results suggest a strategy for the allosteric control of caspase activation with drug-like small molecules.
Caspases are aspartate-specific cysteine proteases that function in a variety of cellular processes, including inflammation, apoptosis, and likely also in development and neurodegeneration.[1] Caspases can be subdivided into so-called activator and effector caspases. Activator caspases are expressed as monomers and become active upon formation of C2-symmetric dimers. The effector caspases include caspases-3, 6, and 7 and are distinguished in that they are expressed as dimeric zymogens (procaspases) that require proteolytic processing to become activated. The careful control of caspase activation via dimerization and/or proteolytic processing prevents spurious activa-
[a] Dr. J. Murray, Dr. A. M. Giannetti, M. Steffek, P. Gibbons, Dr. F. Cohen, C. Tam, C. Pozniak, B. Bravo, Dr. J. Lewcock, C. Q. Ly, X. Zhao, Dr. J. Flygare Departments of Structural Biology, Biochemical Pharmacology, Neuroscience, and Discovery Chemistry Genentech, Inc., 1 DNA Way, South San Francisco, CA 94080 (USA) E-mail: [email protected] [b] Dr. B. R. Hearn, P. Jaishankar, Dr. Y. Tang, Dr. P. Chugha, Prof. M. R. Arkin, Prof. A. R. Renslo Department of Pharmaceutical Chemistry and Small Molecule Discovery Center, University of California, San Francisco 1700 4th Street, San Francisco, CA 94158 (USA) E-mail: [email protected] Supporting information for this article is available on the WWW under http://dx.doi.org/10.1002/cmdc.201300424.
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
tion of the apoptosis cascade and deleterious consequences for the cell.[2, 3] Active-site-directed caspase inhibitors are generally anionic, electrophilic, and poorly selective, making them non-ideal as drug leads and at best crude tools for understanding the role of specific caspases in disease.[4] These same properties make active-site inhibitors particularly problematic as lead candidates for neurodegenerative diseases, for which therapeutics must cross the blood–brain barrier. The search for more drug-like leads has led to a focus on molecules that bind to zymogen or zymogen-like forms of these enzymes. Hence, disulfide-tethered small molecules have been found that bind an allosteric site at the dimer interface of caspases-1 and 7, stabilizing a zymogen-like conformation and inhibiting the active enzyme in vitro.[5, 6] Small-molecule activators of procaspases have also been identified, although these do not appear to act by binding at the dimer interface.[7, 8] Most recently, a phage-derived peptide (pep419) has revealed a novel pH-dependent dimer– tetramer equilibrium that appears to regulate procaspase-6 activation specifically.[9] While more promising with regard to isoform selectivity, pep419 has clear limitations for application in cellular studies and as a therapeutic lead. We sought to apply fragment-based discovery using biophysical methods and X-ray crystallography[10–12] to probe the active and zymogen forms of caspase-6 for allosteric smallmolecule binding sites that might modulate caspase-6 activity and/or activation of procaspase-6. If such compounds could be identified, we expected them to be more tractable leads for neurodegenerative disorders than traditional active-site-directed inhibitors. Herein we report the presence of a zymogenspecific small-molecule binding site at the dimer interface of procaspase-6 and describe multiple nanomolar-affinity ligands specifically tailored for this site. Caspase-6 is unique among effector caspases in being capable of self-activation[13] by cleavage of the inter-subunit linker at Asp193 and rearrangement of the L2 loop. Notably, the small molecules identified in our study significantly engage residues Tyr198 and Thr199 of the L2 loop. This fact, combined with our finding that the compounds significantly stabilize the protein, suggest that control of procaspase activation with small molecules merits further study as a therapeutic strategy. A fragment screen was carried out using surface plasmon resonance (SPR) methods as described previously.[14, 15] Approximately 2300 compounds were screened at 50 mm against both wild-type active and zymogen (C163A) forms of caspase-6 in ChemMedChem 2014, 9, 73 – 77
73
CHEMMEDCHEM COMMUNICATIONS parallel. Hits from either primary screen were tested in dose– response up to 200 mm to yield KD and ligand efficiency (LE) values (LE: binding affinity per non-hydrogen atom (NHA); [kcal mol 1 NHA 1]). This yielded 84 binders to procaspase-6 with KD values ranging from 3 to 2300 mm. Interestingly, the majority of screening hits bound to either zymogen or to active caspase-6, with very few showing similar affinities for both proteins (Supporting Information figure 1). This result indicates that the two forms of the protein possess unique small-molecule binding sites and supports the notion that there are zymogen-specific binding features in caspase-6, as in other caspases. The majority of fragment hits were mono- or bicyclic aromatic compounds such as 1 and 2 (Figure 1). By combining features of these fragments, we identified additional analogues (3–5) that possess superior ligand efficiencies.
www.chemmedchem.org
Figure 2. X-ray crystal structures of fragment hits bound to procaspase-6: a) Co-crystal structure of fragment 6 bound to procaspase-6 at 1.9 resolution; 1.9 2 Fo Fc electron density is shown as a grey mesh contoured at 1 s. b) Relative binding orientations of fragments 1 (purple) and 6 (blue) in the small-molecule binding site of procaspase-6, from the respective co-crystal structures. A representative SPR binding curve for 6 is shown in Supporting Information figure 1.
the phenol OH and the backbone NH groups of Thr199. (Figure 2 a). The phenol OH in 6 also participates in an intramolecular hydrogen bond with the pyridine ring nitrogen atom, an interaction that likely stabilizes the binding conformation. Indeed, SAR of the benzylic side chain confirmed the favorable effects of a phenolic OH function positioned as in 6. The introduction of a fluorine atom at the distal end of the phenolic ring, as in analogue 7, also proved favorable (Figure 3). Com-
Figure 1. Structures of fragments 1–5: Binding affinities (KD) for procaspse-6 are provided, along with ligand efficiencies (LE, kcal mol 1 NHA 1).
X-ray crystallography was used to identify fragment binding site(s) by soaking individual fragment hits into procaspase6C163A crystals. The co-crystal structures of fragments 1–6 revealed a novel binding site at the dimer interface that is only present in the pro-form of caspase-6. The pyridopyrimidine ring of 1 and pyridine ring of 6 bind a hydrophobic crevice, sandwiched on both faces by p–p stacking interactions with Tyr198 side chains from both the A and B monomers (Figure 2). The co-crystal structures of fragments 3–5 demonstrated that this dimer interface site can accommodate a range of aromatic heterocycles. Analogues of 3–5 that differ in substitution of the exocyclic amine exhibited similar KD values and adopted distinct binding orientations. This accommodation of alternate binding orientations produced some uncertainty as to the best approach to optimize interactions with this site. Therefore, we turned our attention to fragment 6, which was unique among the hits in that it possesses an aromatic side chain that projects deeper into the binding site at the dimer interface (Figure 2 a). The X-ray crystal structure of 6 revealed additional sites for interaction that extend along the hydrophobic crevice that include favorable interactions with the side chain methylene groups of Glu214, and a hydrogen bond interaction between 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Figure 3. Structures fragment 6 and related synthesized analogues 7–10: Binding affinities (KD) for procaspse-6 are provided, along with ligand efficiencies (LE, kcal mol 1 NHA 1).
putational docking of 7 to procaspase-6 suggested that the additional binding affinity may result from a multipolar C F C=O interaction[16] with the backbone amide carbonyl group of Ala195. Superimposing the X-ray co-crystal structures of fragments 1 and 6 (Figure 2 b) immediately suggested a ‘fragment merging’ strategy in which the methyl group of 6 would be replaced with hetero/aromatic ring systems in an effort to better engage the Tyr198A/B residues at the dimer interface. Appending or fusing a second hetero/aromatic ring to the pyridine ring of 6 indeed produced analogues with improved binding ChemMedChem 2014, 9, 73 – 77
74
CHEMMEDCHEM COMMUNICATIONS
www.chemmedchem.org
affinity (Figure 3). Interestingly, analogues in which the new aryl ring is fused into a heterobiaryl system (e.g., as in 9 and 10) had superior LE than analogue 8, in which a pyrimidine ring is appended by a single bond. Modeling of 8 in the binding site revealed a significant dihedral angle between the pyrimidine and pyridine rings, an arrangement that does not fully exploit stacking interactions with Tyr198A/B. Compound 8 is thus an example of a ‘strained merged fragment’[17] wherein the merging of overlapping fragments produces internal steric strain that, in the case of 8, limits optimal contact with binding site residues. We considered that analogue 11, a regioisomer of 8, might decrease the dihedral angle of the pyrimidine–pyridine ring system by formation of intramolecular hydrogen bonding interactions (Table 1). Gratifyingly, analogue 11 was found to be
Table 1. Binding affinities and ligand efficiency values for 8 and related analogues 11–13.
Compd[a] 8 11 12 13
R1
R2 – H F F
– H H NH2
KD [mm][b]
LE[c]
24 2.9 0.47 0.38
0.30 0.36 0.40 0.39
[a] The 2-pyrimidyl analogues 11–13 were designed to stabilize a coplanar binding conformation not available to regioisomeric 5-pyrimidyl analogue 8; SPR data for compound 12 binding to procaspase-6 are shown in Supporting Information figure 1. [b] Binding affinity for procaspase-6. [c] Ligand efficiency in kcal mol 1 NHA 1.
~ 10-fold more potent than 8, and possessed an LE value superior even to fused analogues such as 9 and 10. Further modifying analogue 11 with the introduction of a fluorine atom (as done previously in 7), produced analogue 12, with a KD value of 480 nm and LE of 0.40. Finally, the introduction of an amino group at a solvent-exposed position of the pyrimidine ring afforded analogue 13 with a slightly improved KD value of 380 nm. A co-crystal structure of 12 bound to procaspase-6 was determined to 1.8 resolution and confirmed the design approach (Figure 4). Hence, compound 12 binds in an analogous fashion to its progenitor 6, with the phenolic side chain forming the same contacts with Glu214A and Thr199A. The predicted C F C=O multipolar interaction with Ala195 proposed for 7 on the basis of modeling was confirmed in the X-ray structure of 12. The distance between the fluorine and the carbonyl carbon is 3.1 , and the angle between the plane of the peptide bond of Ala195 and the fluorine is 72.28, representing geometries consistent with favorable multipolar interactions.[16] Finally, the pyrimidine and pyridine rings are observed in a coplanar conformation with the pyrimidine ring engaged in stacking interactions with Tyr198A/B. The integra 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Figure 4. X-ray co-crystal structure of 12 bound to procaspase-6 (1.8 resolution): The multipolar C F C=O interaction between 12 and the Ala195 backbone C=O group is shown as magenta dashes. Two intramolecular hydrogen bonding interactions that stabilize the binding conformation are shown as black dashes. The 2 Fo Fc electron density contoured at 1 s is shown as a grey wire mesh.
tion of design, modeling, and X-ray crystallography proved a powerful combination for the atom-efficient conversion of fragments to leads, yielding a ~ 300-fold improvement in binding affinity with the addition of just six new heavy atoms. The discovery of nanomolar-affinity leads like 12 and 13 serves to validate a previously unknown small-molecule binding site on procaspase-6 that is able to accommodate druglike small molecules. The druggability of this site was further confirmed by the success of a second, chemically distinct strategy for the optimization of fragment 6. In this second approach, we sought to exploit the symmetry of procaspase-6 and the location of the small-molecule binding site straddling the C2 axis. It was clear from the co-crystal structure of 6 that the binding pocket occupied by the phenolic side chain is duplicated on the opposite side of the binding site. Accordingly, symmetrical analogue 14 was synthesized and, as predicted, found to bind with its two phenolic side chains projecting into analogous binding pockets on either side of the C2-symmetric interface (Figure 5). While this was a highly encouraging finding, the inferior LE value of 14 relative to its progenitor 6 suggested that the central pyrimidine ring in 14 is not ideally suited to span both halves of the binding site (an alternative explanation is that the protein is not exactly symmetric). Modeling was used to evaluate new designs in which the central pyrimidine ring of 14 was replaced by slightly larger ring systems, including pyrrolopyridine, as found in early analogue 10. After some experimentation, we identified pseudosymmetric analogues 15–17 that exhibited low or sub-micromolar affinity for procaspase-6 and LE values superior to that of symmetric analogue 14 (Table 2). Apparently, the central heteroaromatic scaffolding in 15–17 is better suited to properly position the flanking benzylic side chains in their respective binding sites on opposite sides of the dimer interface. The favorable effects of phenolic and fluorine substitution in 16 and 17 is consistent with the effect observed in 7 and 12, suggesting that the benzylic side chains in these analogues occupy the same binding pocket. ChemMedChem 2014, 9, 73 – 77
75
CHEMMEDCHEM COMMUNICATIONS
www.chemmedchem.org
Figure 6. Melting temperatures (Tagg) of procaspase-6C163A determined at various pH values either alone or in the presence of compounds 6, 12, 13, or 16 (100 mm to 1 mm, depending on KD). Figure 5. Structure of symmetrical analogue 14 (top) and 1.7 resolution Xray co-crystal structure of 14, illustrating the C2 symmetry of the dimer interface site. The 2 Fo Fc electron density contoured at 1 s is shown as a grey wire mesh.
Table 2. Binding affinities and ligand efficiency values for pseudosymmetric analogues 15–17 compared with 10.
Compd 10 15 16 17
R1/R2
R3/R4
– H/H H/OH F/H
– OH/H H/F OH/F
KD [mm][a]
LE[b]
37 2.6 0.55 0.61
0.34 0.29 0.31 0.30
[a] Binding affinity for procaspase-6. [b] Ligand efficiency in kcal mol 1 NHA 1.
The dimer interface site revealed in these studies is comprised of residues from the L2 loop of procaspase-6 and is directly linked to the site of intramolecular cleavage (Asp193) that results in procaspase-6 self-activation.[13] Because activation also requires reorganization of the L2 loop, we were interested in determining whether compounds such as 12 and 16 stabilized the protein via their interaction with the L2 loop. We used thermal shift analysis by differential static light scattering[18] to evaluate temperature-dependent aggregation (Tagg) of procaspase-6 in the presence of fragments and lead compounds, across a pH range that has been shown to affect the dimer–tetramer equilibrium of procaspase-6.[9] Optimized leads such as 12, 13, and 16 significantly stabilized the protein, returning Tagg values 5–7 8C higher across a pH range of pH 5–8 (Figure 6). In contrast, the much less potent fragment hit 6 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
only marginally increased Tagg. Thus, sufficiently potent smallmolecule binders of the dimer interface site have the ability to stabilize procaspase-6 conformation and might therefore inhibit the conformational rearrangement of the L2 loop that renders a fully activated caspase-6. We have described the design of small-molecule ligands for a previously uncharacterized small-molecule binding site on procaspase-6. The combination of fragment-based screening by SPR, X-ray structure determination, and computationally aided chemical design allowed the rapid elaboration of initial fragments into high-affinity leads that significantly stabilize procaspase-6. These results make clear the power of fragmentbased approaches to lead discovery, which allow the targeting of binding sites with unknown function, for which the development of traditional screening assays is not feasible. Compounds such as 12, 13, 16, and 17 represent the first noncovalent, nanomolar-affinity small molecules for the dimer interface site in any caspase family member. Unlike typical active-site caspase inhibitors, the small-molecule ligands identified in this study are neither electrophilic nor anionic, and thus represent much more attractive lead compounds from a drug discovery perspective. Overall, the studies reported herein represent an important first step toward the development of therapeutics that modulate caspase activity and/or activation via allosteric mechanisms.
Experimental Section X-ray diffraction data and all biochemical, biophysical, and synthetic experimental methods are provided in the Supporting Information.
Keywords: caspase modulation · caspase-6 · drug design · fragment merging · fragment screening [1] M. Lamkanfi, N. Festjens, W. Declercq, T. Vanden Berghe, P. Vandenabeele, Cell Death Differ. 2007, 14, 44.
ChemMedChem 2014, 9, 73 – 77
76
CHEMMEDCHEM COMMUNICATIONS [2] S. J. Riedl, G. S. Salvesen, Nat. Rev. Mol. Cell Biol. 2007, 8, 405. [3] E. C. Crawford, J. A. Wells, Annu. Rev. Biochem. 2011, 80, 1055. [4] S. MacKenzie, J. L. Schipper, C. A. Clark, Curr. Opin. Drug Discovery Dev. 2010, 13, 568. [5] J. A. Hardy, J. Lam, T. J. Nguyen, T. O’Brien, J. A. Wells, Proc. Natl. Acad. Sci. USA 2004, 101, 12461. [6] J. M. Scheer, M. J. Romanowski, J. A. Wells, Proc. Natl. Acad. Sci. USA 2006, 103, 7595. [7] K. S. Putt, G. W. Chen, J. M. Pearson, J. S. Sandhortst, M. S. Hoagland, J. T. Kwon, S. K. Hwang, H. Jin, M. I. Churchwell, M. H. Cho, D. R. Doerge, W. G. Helferich, P. J. Hergenrother, Nat. Chem. Biol. 2006, 2, 543. [8] D. W. Wolan, J. A. Zorn, D. C. Gray, J. A. Wells, Science 2009, 326, 853. [9] K. Stanger, M. Steffek, L. Zhou, C. D. Pozniak, C. Quan, Y. Franke, J. Tom, C. Tam, M. J. Elliot, J. W. Lewcock, Y. Zhang, J. Murray, R. Hannoush, Nat. Chem. Biol. 2012, 8, 655. [10] C. W. Murray, T. L. Blundell, Curr. Opin. Struct. Biol. 2010, 20, 497. [11] D. E. Scott, A. G. Coyne, S. A. Hudson, C. Abell, Biochemistry 2012, 51, 4990. [12] M. Hennig, A. Ruf, W. Huber, Top. Curr. Chem. 2012, 317, 115.
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemmedchem.org [13] X. J. Wang, Q. Cao, X. Liu, K. T. Wang, W. Mi, Y. Zhang, L. F. Li, A. C. LeBlanc, X. D. Su, EMBO Rep. 2010, 11, 841. [14] A. M. Giannetti, Methods Enzymol. 2011, 493, 169. [15] C. E. Heise, J. Murray, K. E. Augustyn, B. Bravo, P. Chugha, F. Cohen, A. M. Giannetti, P. Gibbons, R. N. Hannoush, B. R. Hearn, P. Jaishankar, C. Q. Ly, K. Shah, K. Stanger, M. Steffek, Y. Tang, X. Zhao, J. W. Lewcock, A. R. Renslo, J. Flygare, M. R. Arkin, PLoS One 2012, 7, e50864. [16] R. Paulini, K. Muller, F. Diederich, Angew. Chem. 2005, 117, 1820; Angew. Chem. Int. Ed. 2005, 44, 1788. [17] S. A. Hudson, S. Surade, A. G. Coyne, K. J. McLean, D. Leys, A. W. Munro, C. Abell, ChemMedChem 2013, 8, 1451. [18] M. Vedadi, F. H. Niesen, A. Allali-Hassani, O. Y. Fedorov, P. J. Finerty, G. A. Wasney, R. Yeung, C. Arrowsmith, L. J. Ball, H. Berglund, Proc. Natl. Acad. Sci. USA 2006, 103, 15835.
Received: October 25, 2013 Published online on November 20, 2013
ChemMedChem 2014, 9, 73 – 77
77
Tailoring Small Molecules for an Allosteric Site on Procaspase-6 Jeremy Murray,*[a] Anthony M. Giannetti,[a] Micah Steffek,[a] Paul Gibbons,[a] Brian R. Hearn,[b] Frederick Cohen,[a] Christine Tam,[a] Christine Pozniak,[a] Brandon Bravo,[a] Joe Lewcock,[a] Priyadarshini Jaishankar,[b] Cuong Q. Ly,[a] Xianrui Zhao,[a] Yinyan Tang,[b] Preeti Chugha,[b] Michelle R. Arkin,[b] John Flygare,[a] and Adam R. Renslo*[b] Although they represent attractive therapeutic targets, caspases have so far proven recalcitrant to the development of drugs targeting the active site. Allosteric modulation of caspase activity is an alternate strategy that potentially avoids the need for anionic and electrophilic functionality present in most activesite inhibitors. Caspase-6 has been implicated in neurodegenerative disease, including Huntington’s and Alzheimer’s diseases. Herein we describe a fragment-based lead discovery effort focused on caspase-6 in its active and zymogen forms. Fragments were identified for procaspase-6 using surface plasmon resonance methods and subsequently shown by X-ray crystallography to bind a putative allosteric site at the dimer interface. A fragment-merging strategy was employed to produce nanomolar-affinity ligands that contact residues in the L2 loop at the dimer interface, significantly stabilizing procaspase6. Because rearrangement of the L2 loop is required for caspase-6 activation, our results suggest a strategy for the allosteric control of caspase activation with drug-like small molecules.
Caspases are aspartate-specific cysteine proteases that function in a variety of cellular processes, including inflammation, apoptosis, and likely also in development and neurodegeneration.[1] Caspases can be subdivided into so-called activator and effector caspases. Activator caspases are expressed as monomers and become active upon formation of C2-symmetric dimers. The effector caspases include caspases-3, 6, and 7 and are distinguished in that they are expressed as dimeric zymogens (procaspases) that require proteolytic processing to become activated. The careful control of caspase activation via dimerization and/or proteolytic processing prevents spurious activa-
[a] Dr. J. Murray, Dr. A. M. Giannetti, M. Steffek, P. Gibbons, Dr. F. Cohen, C. Tam, C. Pozniak, B. Bravo, Dr. J. Lewcock, C. Q. Ly, X. Zhao, Dr. J. Flygare Departments of Structural Biology, Biochemical Pharmacology, Neuroscience, and Discovery Chemistry Genentech, Inc., 1 DNA Way, South San Francisco, CA 94080 (USA) E-mail: [email protected] [b] Dr. B. R. Hearn, P. Jaishankar, Dr. Y. Tang, Dr. P. Chugha, Prof. M. R. Arkin, Prof. A. R. Renslo Department of Pharmaceutical Chemistry and Small Molecule Discovery Center, University of California, San Francisco 1700 4th Street, San Francisco, CA 94158 (USA) E-mail: [email protected] Supporting information for this article is available on the WWW under http://dx.doi.org/10.1002/cmdc.201300424.
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
tion of the apoptosis cascade and deleterious consequences for the cell.[2, 3] Active-site-directed caspase inhibitors are generally anionic, electrophilic, and poorly selective, making them non-ideal as drug leads and at best crude tools for understanding the role of specific caspases in disease.[4] These same properties make active-site inhibitors particularly problematic as lead candidates for neurodegenerative diseases, for which therapeutics must cross the blood–brain barrier. The search for more drug-like leads has led to a focus on molecules that bind to zymogen or zymogen-like forms of these enzymes. Hence, disulfide-tethered small molecules have been found that bind an allosteric site at the dimer interface of caspases-1 and 7, stabilizing a zymogen-like conformation and inhibiting the active enzyme in vitro.[5, 6] Small-molecule activators of procaspases have also been identified, although these do not appear to act by binding at the dimer interface.[7, 8] Most recently, a phage-derived peptide (pep419) has revealed a novel pH-dependent dimer– tetramer equilibrium that appears to regulate procaspase-6 activation specifically.[9] While more promising with regard to isoform selectivity, pep419 has clear limitations for application in cellular studies and as a therapeutic lead. We sought to apply fragment-based discovery using biophysical methods and X-ray crystallography[10–12] to probe the active and zymogen forms of caspase-6 for allosteric smallmolecule binding sites that might modulate caspase-6 activity and/or activation of procaspase-6. If such compounds could be identified, we expected them to be more tractable leads for neurodegenerative disorders than traditional active-site-directed inhibitors. Herein we report the presence of a zymogenspecific small-molecule binding site at the dimer interface of procaspase-6 and describe multiple nanomolar-affinity ligands specifically tailored for this site. Caspase-6 is unique among effector caspases in being capable of self-activation[13] by cleavage of the inter-subunit linker at Asp193 and rearrangement of the L2 loop. Notably, the small molecules identified in our study significantly engage residues Tyr198 and Thr199 of the L2 loop. This fact, combined with our finding that the compounds significantly stabilize the protein, suggest that control of procaspase activation with small molecules merits further study as a therapeutic strategy. A fragment screen was carried out using surface plasmon resonance (SPR) methods as described previously.[14, 15] Approximately 2300 compounds were screened at 50 mm against both wild-type active and zymogen (C163A) forms of caspase-6 in ChemMedChem 2014, 9, 73 – 77
73
CHEMMEDCHEM COMMUNICATIONS parallel. Hits from either primary screen were tested in dose– response up to 200 mm to yield KD and ligand efficiency (LE) values (LE: binding affinity per non-hydrogen atom (NHA); [kcal mol 1 NHA 1]). This yielded 84 binders to procaspase-6 with KD values ranging from 3 to 2300 mm. Interestingly, the majority of screening hits bound to either zymogen or to active caspase-6, with very few showing similar affinities for both proteins (Supporting Information figure 1). This result indicates that the two forms of the protein possess unique small-molecule binding sites and supports the notion that there are zymogen-specific binding features in caspase-6, as in other caspases. The majority of fragment hits were mono- or bicyclic aromatic compounds such as 1 and 2 (Figure 1). By combining features of these fragments, we identified additional analogues (3–5) that possess superior ligand efficiencies.
www.chemmedchem.org
Figure 2. X-ray crystal structures of fragment hits bound to procaspase-6: a) Co-crystal structure of fragment 6 bound to procaspase-6 at 1.9 resolution; 1.9 2 Fo Fc electron density is shown as a grey mesh contoured at 1 s. b) Relative binding orientations of fragments 1 (purple) and 6 (blue) in the small-molecule binding site of procaspase-6, from the respective co-crystal structures. A representative SPR binding curve for 6 is shown in Supporting Information figure 1.
the phenol OH and the backbone NH groups of Thr199. (Figure 2 a). The phenol OH in 6 also participates in an intramolecular hydrogen bond with the pyridine ring nitrogen atom, an interaction that likely stabilizes the binding conformation. Indeed, SAR of the benzylic side chain confirmed the favorable effects of a phenolic OH function positioned as in 6. The introduction of a fluorine atom at the distal end of the phenolic ring, as in analogue 7, also proved favorable (Figure 3). Com-
Figure 1. Structures of fragments 1–5: Binding affinities (KD) for procaspse-6 are provided, along with ligand efficiencies (LE, kcal mol 1 NHA 1).
X-ray crystallography was used to identify fragment binding site(s) by soaking individual fragment hits into procaspase6C163A crystals. The co-crystal structures of fragments 1–6 revealed a novel binding site at the dimer interface that is only present in the pro-form of caspase-6. The pyridopyrimidine ring of 1 and pyridine ring of 6 bind a hydrophobic crevice, sandwiched on both faces by p–p stacking interactions with Tyr198 side chains from both the A and B monomers (Figure 2). The co-crystal structures of fragments 3–5 demonstrated that this dimer interface site can accommodate a range of aromatic heterocycles. Analogues of 3–5 that differ in substitution of the exocyclic amine exhibited similar KD values and adopted distinct binding orientations. This accommodation of alternate binding orientations produced some uncertainty as to the best approach to optimize interactions with this site. Therefore, we turned our attention to fragment 6, which was unique among the hits in that it possesses an aromatic side chain that projects deeper into the binding site at the dimer interface (Figure 2 a). The X-ray crystal structure of 6 revealed additional sites for interaction that extend along the hydrophobic crevice that include favorable interactions with the side chain methylene groups of Glu214, and a hydrogen bond interaction between 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Figure 3. Structures fragment 6 and related synthesized analogues 7–10: Binding affinities (KD) for procaspse-6 are provided, along with ligand efficiencies (LE, kcal mol 1 NHA 1).
putational docking of 7 to procaspase-6 suggested that the additional binding affinity may result from a multipolar C F C=O interaction[16] with the backbone amide carbonyl group of Ala195. Superimposing the X-ray co-crystal structures of fragments 1 and 6 (Figure 2 b) immediately suggested a ‘fragment merging’ strategy in which the methyl group of 6 would be replaced with hetero/aromatic ring systems in an effort to better engage the Tyr198A/B residues at the dimer interface. Appending or fusing a second hetero/aromatic ring to the pyridine ring of 6 indeed produced analogues with improved binding ChemMedChem 2014, 9, 73 – 77
74
CHEMMEDCHEM COMMUNICATIONS
www.chemmedchem.org
affinity (Figure 3). Interestingly, analogues in which the new aryl ring is fused into a heterobiaryl system (e.g., as in 9 and 10) had superior LE than analogue 8, in which a pyrimidine ring is appended by a single bond. Modeling of 8 in the binding site revealed a significant dihedral angle between the pyrimidine and pyridine rings, an arrangement that does not fully exploit stacking interactions with Tyr198A/B. Compound 8 is thus an example of a ‘strained merged fragment’[17] wherein the merging of overlapping fragments produces internal steric strain that, in the case of 8, limits optimal contact with binding site residues. We considered that analogue 11, a regioisomer of 8, might decrease the dihedral angle of the pyrimidine–pyridine ring system by formation of intramolecular hydrogen bonding interactions (Table 1). Gratifyingly, analogue 11 was found to be
Table 1. Binding affinities and ligand efficiency values for 8 and related analogues 11–13.
Compd[a] 8 11 12 13
R1
R2 – H F F
– H H NH2
KD [mm][b]
LE[c]
24 2.9 0.47 0.38
0.30 0.36 0.40 0.39
[a] The 2-pyrimidyl analogues 11–13 were designed to stabilize a coplanar binding conformation not available to regioisomeric 5-pyrimidyl analogue 8; SPR data for compound 12 binding to procaspase-6 are shown in Supporting Information figure 1. [b] Binding affinity for procaspase-6. [c] Ligand efficiency in kcal mol 1 NHA 1.
~ 10-fold more potent than 8, and possessed an LE value superior even to fused analogues such as 9 and 10. Further modifying analogue 11 with the introduction of a fluorine atom (as done previously in 7), produced analogue 12, with a KD value of 480 nm and LE of 0.40. Finally, the introduction of an amino group at a solvent-exposed position of the pyrimidine ring afforded analogue 13 with a slightly improved KD value of 380 nm. A co-crystal structure of 12 bound to procaspase-6 was determined to 1.8 resolution and confirmed the design approach (Figure 4). Hence, compound 12 binds in an analogous fashion to its progenitor 6, with the phenolic side chain forming the same contacts with Glu214A and Thr199A. The predicted C F C=O multipolar interaction with Ala195 proposed for 7 on the basis of modeling was confirmed in the X-ray structure of 12. The distance between the fluorine and the carbonyl carbon is 3.1 , and the angle between the plane of the peptide bond of Ala195 and the fluorine is 72.28, representing geometries consistent with favorable multipolar interactions.[16] Finally, the pyrimidine and pyridine rings are observed in a coplanar conformation with the pyrimidine ring engaged in stacking interactions with Tyr198A/B. The integra 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Figure 4. X-ray co-crystal structure of 12 bound to procaspase-6 (1.8 resolution): The multipolar C F C=O interaction between 12 and the Ala195 backbone C=O group is shown as magenta dashes. Two intramolecular hydrogen bonding interactions that stabilize the binding conformation are shown as black dashes. The 2 Fo Fc electron density contoured at 1 s is shown as a grey wire mesh.
tion of design, modeling, and X-ray crystallography proved a powerful combination for the atom-efficient conversion of fragments to leads, yielding a ~ 300-fold improvement in binding affinity with the addition of just six new heavy atoms. The discovery of nanomolar-affinity leads like 12 and 13 serves to validate a previously unknown small-molecule binding site on procaspase-6 that is able to accommodate druglike small molecules. The druggability of this site was further confirmed by the success of a second, chemically distinct strategy for the optimization of fragment 6. In this second approach, we sought to exploit the symmetry of procaspase-6 and the location of the small-molecule binding site straddling the C2 axis. It was clear from the co-crystal structure of 6 that the binding pocket occupied by the phenolic side chain is duplicated on the opposite side of the binding site. Accordingly, symmetrical analogue 14 was synthesized and, as predicted, found to bind with its two phenolic side chains projecting into analogous binding pockets on either side of the C2-symmetric interface (Figure 5). While this was a highly encouraging finding, the inferior LE value of 14 relative to its progenitor 6 suggested that the central pyrimidine ring in 14 is not ideally suited to span both halves of the binding site (an alternative explanation is that the protein is not exactly symmetric). Modeling was used to evaluate new designs in which the central pyrimidine ring of 14 was replaced by slightly larger ring systems, including pyrrolopyridine, as found in early analogue 10. After some experimentation, we identified pseudosymmetric analogues 15–17 that exhibited low or sub-micromolar affinity for procaspase-6 and LE values superior to that of symmetric analogue 14 (Table 2). Apparently, the central heteroaromatic scaffolding in 15–17 is better suited to properly position the flanking benzylic side chains in their respective binding sites on opposite sides of the dimer interface. The favorable effects of phenolic and fluorine substitution in 16 and 17 is consistent with the effect observed in 7 and 12, suggesting that the benzylic side chains in these analogues occupy the same binding pocket. ChemMedChem 2014, 9, 73 – 77
75
CHEMMEDCHEM COMMUNICATIONS
www.chemmedchem.org
Figure 6. Melting temperatures (Tagg) of procaspase-6C163A determined at various pH values either alone or in the presence of compounds 6, 12, 13, or 16 (100 mm to 1 mm, depending on KD). Figure 5. Structure of symmetrical analogue 14 (top) and 1.7 resolution Xray co-crystal structure of 14, illustrating the C2 symmetry of the dimer interface site. The 2 Fo Fc electron density contoured at 1 s is shown as a grey wire mesh.
Table 2. Binding affinities and ligand efficiency values for pseudosymmetric analogues 15–17 compared with 10.
Compd 10 15 16 17
R1/R2
R3/R4
– H/H H/OH F/H
– OH/H H/F OH/F
KD [mm][a]
LE[b]
37 2.6 0.55 0.61
0.34 0.29 0.31 0.30
[a] Binding affinity for procaspase-6. [b] Ligand efficiency in kcal mol 1 NHA 1.
The dimer interface site revealed in these studies is comprised of residues from the L2 loop of procaspase-6 and is directly linked to the site of intramolecular cleavage (Asp193) that results in procaspase-6 self-activation.[13] Because activation also requires reorganization of the L2 loop, we were interested in determining whether compounds such as 12 and 16 stabilized the protein via their interaction with the L2 loop. We used thermal shift analysis by differential static light scattering[18] to evaluate temperature-dependent aggregation (Tagg) of procaspase-6 in the presence of fragments and lead compounds, across a pH range that has been shown to affect the dimer–tetramer equilibrium of procaspase-6.[9] Optimized leads such as 12, 13, and 16 significantly stabilized the protein, returning Tagg values 5–7 8C higher across a pH range of pH 5–8 (Figure 6). In contrast, the much less potent fragment hit 6 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
only marginally increased Tagg. Thus, sufficiently potent smallmolecule binders of the dimer interface site have the ability to stabilize procaspase-6 conformation and might therefore inhibit the conformational rearrangement of the L2 loop that renders a fully activated caspase-6. We have described the design of small-molecule ligands for a previously uncharacterized small-molecule binding site on procaspase-6. The combination of fragment-based screening by SPR, X-ray structure determination, and computationally aided chemical design allowed the rapid elaboration of initial fragments into high-affinity leads that significantly stabilize procaspase-6. These results make clear the power of fragmentbased approaches to lead discovery, which allow the targeting of binding sites with unknown function, for which the development of traditional screening assays is not feasible. Compounds such as 12, 13, 16, and 17 represent the first noncovalent, nanomolar-affinity small molecules for the dimer interface site in any caspase family member. Unlike typical active-site caspase inhibitors, the small-molecule ligands identified in this study are neither electrophilic nor anionic, and thus represent much more attractive lead compounds from a drug discovery perspective. Overall, the studies reported herein represent an important first step toward the development of therapeutics that modulate caspase activity and/or activation via allosteric mechanisms.
Experimental Section X-ray diffraction data and all biochemical, biophysical, and synthetic experimental methods are provided in the Supporting Information.
Keywords: caspase modulation · caspase-6 · drug design · fragment merging · fragment screening [1] M. Lamkanfi, N. Festjens, W. Declercq, T. Vanden Berghe, P. Vandenabeele, Cell Death Differ. 2007, 14, 44.
ChemMedChem 2014, 9, 73 – 77
76
CHEMMEDCHEM COMMUNICATIONS [2] S. J. Riedl, G. S. Salvesen, Nat. Rev. Mol. Cell Biol. 2007, 8, 405. [3] E. C. Crawford, J. A. Wells, Annu. Rev. Biochem. 2011, 80, 1055. [4] S. MacKenzie, J. L. Schipper, C. A. Clark, Curr. Opin. Drug Discovery Dev. 2010, 13, 568. [5] J. A. Hardy, J. Lam, T. J. Nguyen, T. O’Brien, J. A. Wells, Proc. Natl. Acad. Sci. USA 2004, 101, 12461. [6] J. M. Scheer, M. J. Romanowski, J. A. Wells, Proc. Natl. Acad. Sci. USA 2006, 103, 7595. [7] K. S. Putt, G. W. Chen, J. M. Pearson, J. S. Sandhortst, M. S. Hoagland, J. T. Kwon, S. K. Hwang, H. Jin, M. I. Churchwell, M. H. Cho, D. R. Doerge, W. G. Helferich, P. J. Hergenrother, Nat. Chem. Biol. 2006, 2, 543. [8] D. W. Wolan, J. A. Zorn, D. C. Gray, J. A. Wells, Science 2009, 326, 853. [9] K. Stanger, M. Steffek, L. Zhou, C. D. Pozniak, C. Quan, Y. Franke, J. Tom, C. Tam, M. J. Elliot, J. W. Lewcock, Y. Zhang, J. Murray, R. Hannoush, Nat. Chem. Biol. 2012, 8, 655. [10] C. W. Murray, T. L. Blundell, Curr. Opin. Struct. Biol. 2010, 20, 497. [11] D. E. Scott, A. G. Coyne, S. A. Hudson, C. Abell, Biochemistry 2012, 51, 4990. [12] M. Hennig, A. Ruf, W. Huber, Top. Curr. Chem. 2012, 317, 115.
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemmedchem.org [13] X. J. Wang, Q. Cao, X. Liu, K. T. Wang, W. Mi, Y. Zhang, L. F. Li, A. C. LeBlanc, X. D. Su, EMBO Rep. 2010, 11, 841. [14] A. M. Giannetti, Methods Enzymol. 2011, 493, 169. [15] C. E. Heise, J. Murray, K. E. Augustyn, B. Bravo, P. Chugha, F. Cohen, A. M. Giannetti, P. Gibbons, R. N. Hannoush, B. R. Hearn, P. Jaishankar, C. Q. Ly, K. Shah, K. Stanger, M. Steffek, Y. Tang, X. Zhao, J. W. Lewcock, A. R. Renslo, J. Flygare, M. R. Arkin, PLoS One 2012, 7, e50864. [16] R. Paulini, K. Muller, F. Diederich, Angew. Chem. 2005, 117, 1820; Angew. Chem. Int. Ed. 2005, 44, 1788. [17] S. A. Hudson, S. Surade, A. G. Coyne, K. J. McLean, D. Leys, A. W. Munro, C. Abell, ChemMedChem 2013, 8, 1451. [18] M. Vedadi, F. H. Niesen, A. Allali-Hassani, O. Y. Fedorov, P. J. Finerty, G. A. Wasney, R. Yeung, C. Arrowsmith, L. J. Ball, H. Berglund, Proc. Natl. Acad. Sci. USA 2006, 103, 15835.
Received: October 25, 2013 Published online on November 20, 2013
ChemMedChem 2014, 9, 73 – 77
77
