|Home | About | Journals | Submit | Contact Us | Français|
We report the identification of novel small molecule agonists of integrin CD11b/CD18, which increased, in a dose-dependent manner, the adhesion of the integrin CD11b/CD18 expressing cells to two physiologically relevant ligands: Fibrinogen and iC3b. Compound 6 showed an ex vivo EC50 of 10.5 μM and in vitro selectivity for binding to the recombinant αA-domain of CD11b/CD18. In silico docking experiments suggest that the compounds recognized a hydrophobic cleft in the ligand-binding αA-domain, implying an allosteric mechanism of modulation of integrin affinity by this novel compound.
The α/βintegrin heterodimers mediate important cellular functions including cell adhesion, migration and signaling.1 β2 integrins, with a common β-subunit (β2, CD18) but distinct α-subunits (CD11a, CD11b, CD11c and CD11d), are critical leukocytic receptors that are important in inflammation and immunity.2 The integrin CD11b/CD18 (also known as Mac-1, CR3 and αMβ2) is the predominant β2 integrin in neutrophils, macrophages and monocytes and mediates pro-inflammatory functions in these cells.3-5 CD11b/CD18 recognizes the complement fragment iC3b, Fibrinogen, and ICAM-1 as ligands, among various others. CD11b/CD18 has been implicated in many inflammatory and autoimmune diseases, such as ischemia-reperfusion injury (including acute renal failure and atherosclerosis), tissue damage, stroke, neointimal thickening in response to vascular injury and the resolution of inflammatory processes.2, 6-9 Thus, there is a considerable potential for agents that modulate the function of CD11b/CD18 as therapeutic agents for the treatment of such inflammatory conditions.
Progress towards identifying small molecules that selectively target integrin CD11b/CD18 has been slow, with only a few reported discoveries,10, 11 primarily due to the lack of good high-throughput screening (HTS) assays. We recently described a simple, no-wash cell-adhesion based HTS assay in the 384-well plate format that is ideal for the discovery of small molecules against the integrin CD11b/CD18.12 Our cell-based assay does not require expensive reagents, is very quick, and provides a quantitative and consistent readout. In this preliminary report, we describe the discovery of several novel small molecule agonists of integrin CD11b/CD18 (Figure 1).
We screened a commercially available library of >13,500 small molecules to identify novel agonists of integrin CD11b/CD18 using our previously described assay,12 which relies on the ability of small molecules to increase adhesion of mammalian K562 cells stably transfected with the wild type integrin CD11b/CD18 (K562 CD11b/CD18) to Fibrinogen, a physiologic ligand of integrin CD11b/CD18. K562 CD11b/CD18 cells showed virtually no binding to immobilized Fibrinogen (Fg) when incubated in the assay buffer (1mM each of physiologic ions Ca2+ and Mg2+ in Tris buffered saline (TBS++)) alone. However, 208 compounds selectively increased adhesion of K562 CD11b/CD18 cells and were identified as hits from this primary screen (hit-rate of ~1.5%).
Next, we “cherry-picked” the top 87 compounds that produced the highest level of cell-adhesion for validation using a secondary assay - small molecule induced binding of K562 CD11b/CD18 cells to two different integrin CD11b/CD18 ligands (Fg and iC3b). 46 compounds showed increased K562 CD11b/CD18 cell adhesion in the secondary assays in a dose dependent fashion and thus were confirmed as agonists hits (hit confirmation rate of ~53%). Majority of the agonists showed the EC50 (Effective Concentration for 50% increase in cell adhesion) values in the 10–30μM range.
Surprisingly, we found that a large subset of the hits contained a central five-membered 2,4-di-oxo-thiazolidine motif.12 A subsequent primary screen with a chemical library of >92,000 compounds also identified a number of similar 2,4-di-oxo-thiazolidine motif containing compounds as hits.14 Furthermore, we found that this family of hit compounds are structurally similar to compounds recently reported as CD11b/CD18 agonists in an independent screen using purified CD11b A-domain (αA-domain),11 helping us quickly narrow down the target binding site for this class of molecules to the CD11b A-domain. We confirmed targeting of the αA-domain by the 2,4-di-oxo-thiazolidine motif containing compounds using binding assays with purified recombinant αA-domain, where these compounds increased binding of αA-domain to immobilized Fg.15 Additionally, binding was selective as cells not expressing CD11b/CD18 did not show any appreciable binding and the binding of CD11b/CD18 expressing cells could be blocked with known blocking monoclonal antibodies (mAbs) 44a3 (anti-CD11b) and IB416, 17 (anti-CD18).
Next, we determined the selectivity of compounds for integrin CD11b/CD18 over highly homologous integrin CD11a/CD18 (also known as LFA-1). Since both compounds 1 and 2 (Figure 2) produced high levels of K562 CD11b/CD18 cell-adhesion in secondary assays, we obtained pure 1 and 2 in the powder form from commercial vendors for these assays. We also generated K562 cells stably transfected with wild type integrin CD11a/CD18 (K562 CD11a/CD18). Next, we measured the ability of 1 and 2 to increase cell adhesion to immobilized ICAM-1, a physiologic ligand of integrin CD11a/CD18. Although 1 was not highly soluble in aqueous buffers, 2 showed two-fold higher selectivity for integrin CD11b/CD18 over CD11a/CD18, with EC50 values of 13.6±5μM with K562 CD11b/CD18 cells. This is in contrast with the previously described compounds,11 which showed equal binding to both integrins in our assays.
In order to determine the exact binding mode of 2 in the αA-domain, we explored the structure-activity relationship (SAR) of various substitutions on the core 5-(furan-2ylmethylene)-2,4-di-oxo-thiazolidine motif that was common between compounds 1, 2 and a number of other primary hits. We obtained a limited series of compounds with various R1 and R2 substitutions (Table I) from commercial sources and determined their effect on K562 CD11b/CD18 cell adhesion to its natural ligand Fg. The EC50 values of these derivatives are listed in Table I.
In general, substitutions at the C-5 position of the furan ring (R2 substituents) had the largest effect on agonist potency. Non-aromatic (3–4) or non-conjugated (5) substituents that disrupted the pi-conjugation with the planar furanyl ring were strongly dis-favored. Planar aromatic rings were preferred and non-substituted phenyl ring (6) was most preferred over aliphatic or polar heteroatoms at ortho or the para positions of the phenyl ring (7–11). For the R1 substituents at the N-3 position of the thaizolidine ring, shortening the length of the substituted ester (from ethyl to methyl) (12–13), and shortening the aliphatic chain length (14–15) was highly dis-favored. Similarly, substitution of the aliphatic chain with a phenyl ring was dis-favored (16). Long-chain, bulky residues were also dis-favored at R1 (17–20). However, compound 21 containing methylene substituted small aromatic ring bound to a level similar to 6. Conversely, a co-substitution of benzyl at R1 with a highly electron-withdrawing and bulky para-substituted aromatic at R2 was highly dis-favored (22).
Compound 6 also showed selective binding to the purified recombinant αA-domain by increasing its binding to immobilized Fg15 as well as a high selectivity for integrin CD11b/CD18 (EC50 = 10.5±5μM) over CD11a/CD18 (no appreciable binding). Thus, we believe that we have discovered a novel and unique integrin CD11b/CD18 selective agonist.
To further evaluate the compounds we also calculated various physicochemical descriptors using Schrodinger QikProp program. Our best compounds (1, 2, 6, 7, 8) have good predicted Caco-2 cell permeability and human oral absorption. Among them 6 has a slightly better clogP and better predicted solubility. Because it is more active than 2 and smaller, it also has the highest ligand efficiency (BEI = 14).18
Next, in order to gain insights into potential binding pockets for these small molecules in the αA-domain, we conducted in-silico docking experiments. The high-resolution three-dimensional structure of CD11b A-domain in both its closed (inactive) and open (active, ligand-competent) conformations is available from PDB.19-21 However, the α7 helix in αA (that creates part of a hydrophobic pocket known as Socket for Isoleucine (SILEN) in CD11b21 or IDAS in CD11a22 and that shows the highest conformational change upon αA-activation19, 23, 24) is shorter by three residues in the three-dimensional structures of the open form20, 21 as compared to that of the closed form of αA.19, 20 As the newly discovered agonists are predicted to bind in this region and stabilize this conformation of αA,11 we constructed a model of the open (active, ligand-competent) conformation of the CD11b A-domain by manually extending the α7 helix in the high resolution structure of CD11b A-domain20, 21 by three additional residues from the structure of the closed form followed by hydrogen bond optimization and constrained (Impref) minimization as implemented in the Maestro protein preparation facility (Schrodinger Inc, Portland).
Conformational repositioning of the α7 helix upon activation, which appears to be stabilized upon agonist binding suggests that the agonists bind in the region between helix α7 and α1 and the central β sheet.11, 19 This has also been suggested by a previous report.11 Therefore we utilized the above optimized structure of the αA-domain in the open conformation to initiate compound docking. In the apo structure this activation sensitive α7 helix region is spatially crowded by many hydrophobic residues lining the pocket. We applied an induced fit docking procedure implemented in the Schrodinger software suite in which initial docking with a softened potential to generate an ensemble of possible poses is followed by receptor optimization and ligand re-docking.25 This protocol resulted in a high scoring pose of 2 (Z configuration) in which the carbonyl oxygens of the 2,4-di-oxo-thiazolidine core are fixed by Ser133 and Thr169 and the hydrophobic 2,4-dichlorophenyl moiety is interacting in the hydrophobic pocket. In a stable 6 ns all-atom explicit solvent molecular dynamics simulation at increased temperature (using Desmond by DEShaw Research)26 the α7 helix adjusts only very slightly. The induced-fit docking receptor was used to dock additional structures using Schrodinger Glide Program.27 The obtained poses were then rescored using MM-GB/SA methodology28 allowing receptor flexibility to obtain more accurate estimates of relative binding free energies. The resulting binding hypothesis of the best compound 6 is shown in Figure 3. As we expected, the hydrophobic phenyl furanyl moiety (the C5-substituent on the thiazolidine ring) is burried in a hydrophobic pocket lined by residues L312, I308, L305 (α7 helix), L164, V160, F156 (α1 helix), and Y267, I269, I236, V238, I236, I135 (central beta sheet). The relative free binding energy of 6 is slightly lower compared to 2 in agreement with the experimental data. This structural model also explains why compound 13 (highly similar to 2 and 6) is inactive, as in this binding mode the αC carbon of the ethylcarboxylate moiety at N-3 position of the central thiazolidine ring of 2 is in close proximity to Ser133, Thr169, and Asp132 (less than 2.5 Å), which creates a tight fit and does not tolerate the larger methyl group at αC that is present in 13 but is absent in 2 and 6.
For the most comparable compounds 2, 6, 7 and 8, the lower activity of the sterically more demanding compound can (at least partially) be attributed to increased receptor and/or ligand strain. The SAR and the binding hypothesis suggest that one hydrophobic interaction is critical. Compounds with two polar ends are inactive (9, 10, 11, 12). The interaction in the hydrophobic pocket appears quite sensitive to sterical demand and the overall size of the molecule. For example, in case of the smaller ethyl acetate N-3 substituent, larger (2, 7, 8) as well as smaller (6) phenyl furanyl substituents are tolerated (although the smallest is the most active) while for structures with larger N-3 substituents (such as 17 and 20) only the unsubstituted phenyl furanyl is active. Thus, the in silico docking studies suggest a reasonable hypothesis for the binding of these novel allosteric agonists of integrin CD11b/CD18. Additionally, our proposed model is consistent with a previous modeling study with other CD11b A-domain agonists.11 However, as can be expected from an in silico model, the model presented here can not in all cases quantitatively explain the subtle differences in activity and it should be considered a hypothesis that will be further evaluated by extensive molecular dynamics and SAR studies in the future. During the various induced fit docking studies we obtained other (although lower scoring) poses. For example, one model showed the compound 6 “flipped” along its long, vertical axis, as shown in Figure 3B. However in all cases the hydrophobic moiety interacts in the same region described and illustrated in Figure 3. Future structural studies will determine the validity of this model as well as the reasons for selectivity of agonists 2 and 6 for integrin CD11b/CD18 over highly homologous integrin CD11a/CD18.
To summarize, we have identified a series of novel substituted 5-(furan-2ylmethylene)-2,4-di-oxo-thiazolidine motif containing agonists of integrin CD11b/CD18. Several molecules within this class showed good binding to the ligand binding αA-domain of CD11b/CD18 and are currently subject to further evaluation. These compounds may serve as mechanistic chemical biology probes for modulating β2 integrin function in vivo. The proposed structural model suggests that the novel agonists increase integrin CD11b/CD18 dependent cell adhesion by binding αA in the activation-sensitive F-α7 region and allosterically stabilizing it in the open conformation. β2 integrin agonists can modulate recruitment of leukocytes and lymphocytes to the sites of inflammation via a novel mechanism - by promoting cell adhesion and delaying cell de-adhesion in an integrin dependent fashion.11, 12, 29 Thus, these newly identified integrin CD11b/CD18 agonists are also potential leads for the development of unique anti-inflammatory agents.
This work was supported in part by NIH Grants DK068253 and NS053659 and by The Edward W. and Stella C. Van Houten Memorial Fund. We thank Dr. Jun Y. Park and Dr. Jack Rosa for generous help and discussions with the cell-based assays; Dr. Caroline Shamu and the rest of the ICCB staff for their support in the implementation of the HTS assay and Prof. M. Amin Arnaout for helpful discussions. This work was supported with resources of the Center for Computational Science at the University of Miami (# 162). We also thank Istvan Kolossvary from DEShaw Research for help setting up Desmond.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.