|Home | About | Journals | Submit | Contact Us | Français|
CXCL12 is an attractive target for clinical therapy because of its involvement in autoimmune diseases, cancer growth, metastasis, and neovascularization. Tyrosine sulfation at three positions in the CXCR4 N-terminus is crucial for specific, high-affinity CXCL12 binding. An NMR structure of the complex between the CXCL12 dimer and a sulfotyrosine-containing CXCR4 fragment enabled high-throughput in silico screening for inhibitors of the chemokine-receptor interface. A total of 1.4 million compounds from the ZINC database were docked into a cleft on the CXCL12 surface normally occupied by sulfotyrosine 21 (sY21), and five were selected for experimental screening. NMR titrations with CXCL12 revealed that four compounds occupy the sY21 site, one of which binds with a Kd of 64 µM. This compound selectively inhibits SDF1-induced CXCR4 signaling in THP1 cells. Our results suggest that sulfotyrosine recognition sites can be targeted for the development of novel chemokine inhibitors.
Chemokines are a family of small, secreted proteins that orchestrate cell migration by activating a set of G-protein coupled receptors (GPCRs). The immune system relies on chemokine signaling to direct lymphocyte homing, orchestrate inflammatory responses, and stimulate wound healing. Outside of these normal functions, chemokines and their receptors also participate in numerous disease states, including HIV/AIDS, asthma, autoimmune diseases, and cancer. Most drug discovery research is directed at GPCRs1, and therapeutic modulation of chemokine signaling is correspondingly directed at the receptors rather than the ligands. Small molecule antagonists targeting chemokine receptors are at various stages of development; the HIV entry inhibitor Maraviroc, which blocks the CCR5 coreceptor, was recently approved by the FDA for clinical use2. Chemokine variants and peptidomimetics are also viewed as potential inhibitors3.
Because it directs stem cell homing4 and participates in nearly every aspect of cancer progression – growth, metastasis, and neovascularization5 – the CXCL12/CXCR4 signaling axis is of increasing interest for drug discovery. In principle, inhibitors targeting the chemokine ligand would also be useful, but small (<10 kDa) proteins traditionally have been considered too small to be “druggable”. However, Fesik and coworkers successfully used NMR-based fragment screening to identify micromolar ligands for FKBP12, a 12 kDa protein, and defined structure-activity relationships (SAR) that enabled the subsequent design of potent nanomolar inhibitor6. The serendipitous discovery by Wells and colleagues of an IL-2 inhibitor that binds the cytokine ligand rather than the receptor demonstrated that shallow, solvent-exposed clefts on small, secreted proteins can serve as legitimate sites for drug discovery7. The recent report of a chalcone that binds CXCL12 and prevents CXCR4 activation suggests that chemokines are legitimate targets for inhibition8.
Tyrosine O-sulfation is an important posttranslational modification in the N-terminal extracellular domain of chemokine receptors that contributes to specific chemokine recognition. CXCR4 sulfation at residues 7, 12 and 21 enhances its interaction with CXCL129–11, and the NMR structure of a soluble dimeric CXCL12:CXCR4 complex revealed a specific binding pocket for each sulfotyrosine9. In a recent NMR study using full-length CXCR4, methyl-containing side chains in all three sulfotyrosine recognition sites exhibited saturation transfer effects12, reinforcing the functional relevance of CXCL12:CXCR4 contacts we observed in the soluble complex and validated by mutagenesis9. Of the three CXCR4 sulfotyrosines, sY21 was reported to make the largest contribution to CXCL12 binding9,11. Consequently, we hypothesized that small molecules targeting the sY21 site could act as chemokine inhibitors and designed a structure-based screen for compounds that bind CXCL12 and prevent CXCR4 signaling.
We performed an in silico screen of compounds from the ZINC virtual compound library using DOCK 3.5.54 at the site on CXCL12 occupied by CXCR4 residues D20 and sY21 (Figure 1A) in our NMR structure of the complex (PDB ID 2K05). After examining 1000 compounds with the best docking scores, five that appeared most complementary to the sY21 site were selected for NMR titrations with [U-15N]-CXCL12 to assess binding affinity and specificity.
Mapping of ligand-induced 1H-15N HSQC shift perturbations (Figure 1B) indicated that three of the five compounds (ZINC IDs 1709621, 4202287, 16954065) bound weakly but specifically to CXCL12 in the CXCR4 sY21 site (Figure 1C). Perturbations induced by ZINC compound 4900356, while significant, were consistent with nonspecific interactions at multiple sites.
Compared to the other four compounds, ZINC 310454 produced larger shifts for more CXCL12 residues (Figure 2A). While the pattern of shift perturbations is distinct from the other molecules, their distribution on the CXCL12 surface is consistent with the docking pose for this compound (Figure 2B), which is considerably larger than the others.
Nonlinear fitting of chemical shift perturbations showed that 310454 bound CXCL12 significantly more tightly (Kd = 64 ± 15 µM) than the other compounds, similar to the affinities exhibited by successful ‘hits’ from other NMR-based drug discovery efforts. To further probe the structure-activity relationships of 310454-mediated CXCL12 inhibition, we measured the binding of five related compounds by NMR (Figure 2C and Supporting Figure 2). Removal or replacement of the carboxylic acid with a methyl ketone or bromine established the critical importance of this functional group, as evidenced by complete loss of binding. Indeed, the presence of one or more carboxylic acid groups was a common feature of the original five compounds identified through in silico screening. Substitution of the naphthyl group with a phenyl ring lowered the affinity by ~10-fold and altered the pattern of shift perturbations in a manner consistent with the predicted binding mode (Figure 2B). Thus, both charged and hydrophobic interactions contribute the affinity and specificity of 310454 for the CXCL12 sulfotyrosine binding pocket.
To test 310454 as an inhibitor of CXCL12-mediated signaling, we measured CXCR4 activation by monitoring intracellular Ca2+ levels in THP-1 cells, which express high levels of CXCR4 and CCR2. Addition of 100 µM 310454 alone induced no Ca2+-flux in THP-1 cells. Preincubation of chemokine with 100 µM 310454 had no effect on MCP-1/CCR2 signaling but abolished the CXCL12-mediated Ca2+-flux response (Figure 2D). We conclude that 310454 is a selective inhibitor of CXCL12 that acts by blocking a key interaction with sY21 in the CXCR4 N-terminus.
Our results reinforce the concept that small cytokines are viable drug targets8. The uncertainty in side chain positions associated with NMR structure ensembles is often viewed as a barrier to successful in silico ligand screening. However, using a single NMR-derived conformer, we exploited details of CXCR4 sY21 recognition by CXCL12 to search for compounds that would occupy the same site and satisfy a similar set of polar and hydrophobic contacts. Four unrelated compounds bind the selected site, and one, 3-(naphthalene-2-carbonylthio carbamoylamino)benzoic acid, inhibits CXCL12 activation of its receptor CXCR4 at micromolar concentrations. SAR analysis confirmed the orientation of the bound ligand and demonstrated a requirement for the benzoic acid functional group, suggesting that it may mimic the negatively-charged sulfotyrosine. We speculate that sulfotyrosine binding sites on other chemokines and elsewhere on the CXCL12 surface could be similarly targeted with small molecules. By linking micromolar ligands that bind adjacent sites, it may be possible to design novel, high-affinity inhibitors of CXCL12 and other chemokines.
This work was supported by NIH grants AI058072 and AI063325 and a State of Wisconsin cancer research grant (BFV), an American Cancer Society Spin Odyssey Postdoctoral Fellowship from the New England Division (CTV), and startup funds from UWW (CTV) and USF (YC). We thank Brian Shoichet for computing resources and Emmanuel Smith for assistance with docking.
Supporting information available: Detailed methods, CXCL12-associated docking poses of all compounds, and HSQC spectra. This material is available free via the Internet at www.acs.pubs.org.