Identification of human S100A9 as a novel target for treatment of autoimmune disease via binding to quinoline-3-carboxamides

Abstract

Despite more than 25 years of research, the molecular targets of quinoline-3-carboxamides have been elusive although these compounds are currently in Phase II and III development for treatment of autoimmune/inflammatory diseases in humans. Using photoaffinity cross-linking of a radioactively labelled quinoline-3-carboxamide compound, we could determine a direct association between human S100A9 and quinoline-3-carboxamides. This interaction was strictly dependent on both Zn++ and Ca++. We also show that S100A9 in the presence of Zn++ and Ca++ is an efficient ligand of receptor for advanced glycation end products (RAGE) and also an endogenous Toll ligand in that it shows a highly specific interaction with TLR4/MD2. Both these interactions are inhibited by quinoline-3-carboxamides. A clear structure-activity relationship (SAR) emerged with regard to the binding of quinoline-3-carboxamides to S100A9, as well as these compounds potency to inhibit interactions with RAGE or TLR4/MD2. The same SAR was observed when the compound's ability to inhibit acute experimental autoimmune encephalomyelitis in mice in vivo was analysed. Quinoline-3-carboxamides would also inhibit TNFalpha release in a S100A9-dependent model in vivo, as would antibodies raised against the quinoline-3-carboxamide-binding domain of S100A9. Thus, S100A9 appears to be a focal molecule in the control of autoimmune disease via its interactions with proinflammatory mediators. The specific binding of quinoline-3-carboxamides to S100A9 explains the immunomodulatory activity of this class of compounds and defines S100A9 as a novel target for treatment of human autoimmune diseases.

Figure 1. Human S100A9 Is a Target Protein for Quinolines

(A) The basic structure of the quinoline compounds is shown. In the lower part of the panel, the specific modifications made in order to use these compounds as probes to isolate the target protein are shown. (B) A two-dimensional gel is shown in which the indicated spots (boxed) were subsequently identified as S100A9. The protein in all three spots was isolated separately and homogenously identified as S100A9. (C) Sensorgrams obtained after injection of 25–200 nM human S100A9 over immobilised ABR-224649 (left panel). Sensorgrams from top to bottom represent: 200, 150, 100, 75, 50, 37.5, and 25 nM S100A9 and nonspecific binding (NSB), i.e., sample buffer without S100A9. In this particular experiment, injection time was 6 min at a flow rate of 30 μl/min, and regeneration was performed with a 30-μl pulse of HBS-P buffer containing 3 mM EDTA (HBS-EP). Start injection of sample: association phase (1), running buffer: dissociation phase (2), regeneration solution (3), and running buffer again (4) constitute an analysis cycle. HBS-P with 1 mM Ca²⁺ and 10 μM Zn²⁺ was used as running and sample buffer. In the right panel, responses at steady state (after subtraction of signal in reference flow cell) were plotted versus concentration of S100A9 yielding half-maximal binding at 85 nM (r ² = 1.00). (D) Binding of homo- and heterodimeric human S100A8 and S100A9 to immobilized ABR-224649 at a concentration of 100 nM (based on their homo- or heterodimeric molecular weight). The response at late association phase was calculated and plotted in ascending order of response magnitude. (E) Displacement of S100A9 binding to immobilised ABR-224649 by ABR-215757 is shown. S100A9 was injected for 3 min at 100 nM (i.e., at ≈B _max/2 concentration) ± 7.81–1,000 μM 215757, and responses at late association were plotted against the concentration of competitor. An IC₅₀ value of 37 μM was calculated in this experiment using a one-site competition model (r ² = 1.00). The amino-linker compound, ABR-224649, showed a very similar ability to displace binding as ABR-215757 when coinjected with S100A9 over the surface (unpublished data). (F) Effect of Ca²⁺ and Zn²⁺ on binding of S100A9 to ABR-224649 is shown. S100A9, 100 nM, was injected in HBS-P buffer containing either a fixed concentration of Ca²⁺ (1 mM) or Zn²⁺ (10 μM), with Zn²⁺ and Ca²⁺ concentrations titrated from 0–50 μM and 0–2,000 μM, respectively. Responses at late association phase were plotted versus metal ion concentration, and EC₅₀ values of 5.5 μM for Zn²⁺ and 193 μM for Ca²⁺ were calculated using a sigmoidal dose-response model.

Figure 2. Human S100A9 Binds Human RAGE

(A) Left: sensorgrams showing association and dissociation phase after injection of 100 nM hS100A9 (upper) and hS100A8/9 (lower) over RAGE immobilized at a density of approximately 3,000 RU (solid lines). Curves were fit to a 1:1 binding model with mass transfer (dotted line) and yielded K_D values of 38 and 9.4 nM with maximum responses at 1,200 and 220 RU for homodimeric and heterodimeric hS100A9, respectively. The somewhat lower affinity for homodimeric S100A9 is due to a higher off-rate. Right: responses at late association phase plotted against the concentration (Conc.) of S100A9 showing half-maximal binding at approximately 100 nM. (B) Zn²⁺ and Ca²⁺ concentrations were titrated as described in Figure 1F. EC₅₀ values of 7.1 μM for Zn²⁺ and 146 μM for Ca²⁺ were calculated using a sigmoidal dose-response model. (C) Displacement curve showing inhibition of 100 nM S100A9 binding to the RAGE surface with 7.81–1,000 μM ABR-215757 as competitor. Competitor in the absence of S100A9 showed background or minimal direct binding to RAGE (unpublished data). In this experiment, the resulting displacement curve yielded 50% inhibition at 26 μM. Dashed line indicates signal for S100A9 in the absence of competitor. (D) Binding of 100 nM S100 proteins (injected for 3 min at 30 μl/min) to immobilised RAGE in the presence of 10 μM Zn²⁺ and 1 mM Ca²⁺ with responses calculated from sensorgrams at late association phase. Comp, competitor.

Figure 3. Interaction of Human S100A9 with Human TLR4/MD2

(A) Left: binding of 100 nM S100A9 (upper) and S100A8/9 (lower) to immobilized TLR4/MD2 (coupling density ∼3,000 RU) represented as sensorgrams (solid lines). Best fit of curves was obtained with 1:1 model with mass transfer (dotted lines) with K_D values of 2.1 and 3.8 nM and maximum responses at 1,270 and 280 RU, respectively. Right: half-maximal binding at approximately 77 nM was obtained when plotting responses at late association phase against concentration (Conc.) of S100A9. (B) Influence of Zn²⁺ on S100A9 binding to hTLR4/MD-2 at a fixed concentration of Ca²⁺ (1 mM). Half-maximal binding was obtained at 6.4 μM Zn²⁺. (C) Displacement curve showing inhibition of S100A9 binding to immobilised hTLR4/MD2 at 100 nM with the competitor (ABR-215757) added in a concentration range from 7.81 to 1,000 μM. Direct binding of ABR-215757 to hTLR4/MD2 was negligible (unpublished data). In this experiment, an IC₅₀ value of 23 μM (r ² = 0.998) was calculated after fit of data to a one-site competition model. Response for S100A9 in the absence of compound is indicated by the dashed line. (D) Blocking of human S100A9 binding to immobilized TLR4/MD2, RAGE and Q compound by human TLR4/MD2 complex in solution. Comp, competitor.

Figure 4. Analysis of Mouse S100A9 Interaction with Mouse RAGE, Q Compound, and Mouse TLR4/MD2

(A) Sensorgrams generated after injection of 50 and 100 nM mouse S100A9 (3 min at 30 μl/min) over indicated ligands coupled to the sensor chip. (B) Inhibition curves for displacement of 100 nM mouse S100A9 interacting with mouse RAGE and mLPS-Trap, respectively, using ABR-215757 as competitor. Conc, concentration. (C) Binding of 100 nM mouse S100A9 homodimers and S100A8/A9 heterodimers to the indicated surfaces.

Figure 5. PCA Loading Plot of the Relationship between Q Compound S100A9 or S100A8/A9 Binding Activity and aEAE Inhibition

(A) Correlation between the binding strength to S100A9 of the Q compounds (Table S1) and their efficacy in inhibiting aEAE in vivo. (B) Loading plot of the first two principal components (p[1] and p[2]) demonstrating the relationship between S100A9 and S100A8/A9 binding activity and aEAE inhibition. 1: mS100A8/A9 and mRAGE; 2: mS100A8/A9 and mLPS-Trap; 3: mS100A9 and ABR-224649; 4: mS100A9 and mLPS-Trap; 5: mS100A9 and mRAGE/hFc; 6: aEAE; 7: hS100A9 and rhTLR4/MD-2; and 8: hS100A9 and rhRAGE/Fc.

Figure 6. Fab 43/8 Blocks Binding of S100A9 to RAGE and TLR4

(A) FACS analysis of human PBL using 43/8, the S100A8/A9 specific antibody 27E10 and isotype control. FL1, fluorescence channel 1 (green). (B) Inhibition curves for binding of 100 nM human S100A9 to immobilised human RAGE and human TLR4/MD2 in the absence or presence of 25 to 400 nM Fab 43/8. After preincubation of S100A9 ± Fab for 1 h at room temperature in HBS-P buffer containing 1 mM Ca²⁺ and 10 μM Zn²⁺, samples were injected for 3 min at 30 μl/min. Note: molar ratio homodimeric S100A9/monovalent Fab is 8:1, 4:1, 2:1, 1:1, and 0.5:1 at the various Fab concentrations. Conc, concentration. (C) Influence of Ca²⁺ and Zn²⁺ on binding of human S100A9 to amine-coupled intact 43/8 antibody (coupling density ∼5,000 RU). Sensorgrams generated after injection of S100A9 at 100, 200, and 400 nM (3 min at 30 μl/min) over 43/8 in the absence (−Ca/Zn) or presence (+Ca/Zn) of 1 mM Ca²⁺ and 10 μM Zn²⁺.

Figure 7. Inhibition of TNFα Induction by 43/8 Fab and ABR-215757

C57BL/6 mice were treated with PBS, 43/8 Fab, or ABR-215757 for 2 h before being challenged with 3 or 6 μg of LPS intraperitoneally, as indicated. Two hours after LPS challenge, the animals were sacrificed, and the level of TNFα in serum was determined. Error bar indicates standard error of the mean (SEM).