Identification of Inhibitors of Integrin Cytoplasmic Domain Interactions With Syk

Leukocyte inflammatory responses require integrin cell-adhesion molecule signaling through spleen tyrosine kinase (Syk), a non-receptor kinase that binds directly to integrin β-chain cytoplasmic domains. Here, we developed a high-throughput screen to identify small molecule inhibitors of the Syk-integrin cytoplasmic domain interactions. Screening small molecule compound libraries identified the β-lactam antibiotics cefsulodin and ceftazidime, which inhibited integrin β-subunit cytoplasmic domain binding to the tandem SH2 domains of Syk (IC50 range, 1.02–4.9 µM). Modeling suggested antagonist binding to Syk outside the pITAM binding site. Ceftazidime inhibited integrin signaling via Syk, including inhibition of adhesion-dependent upregulation of interleukin-1β and monocyte chemoattractant protein-1, but did not inhibit ITAM-dependent phosphorylation of Syk mediated by FcγRI signaling. Our results demonstrate a novel means to target Syk independent of its kinase and pITAM binding sites such that integrin signaling via this kinase is abrogated but ITAM-dependent signaling remains intact. As integrin signaling through Syk is essential for leukocyte activation, this may represent a novel approach to target inflammation.


INTRODUCTION
Integrin cell adhesion molecules are expressed on all cell types and are essential for proper organism development and tissue homeostasis (1,2). In addition, integrins play a significant role in the pathophysiology of numerous diseases including those of the immune system (3)(4)(5)(6)(7)(8). Because they regulate key steps in leukocyte recruitment into tissue, including leukocyte tethering, firm adhesion, diapedesis, and migration (9), integrins have been directly implicated in inflammation and autoimmunity. Furthermore, integrins not only provide adhesive contacts, but also serve as signaling molecules that are essential for leukocyte activation within sites of inflammation (10,11).
Originally identified as a key signaling component of the B-cell antigen receptor and Fc-receptors (12), the non-receptor spleen tyrosine kinase (Syk) is now recognized as an essential integrin signaling component in inflammation (13,14). One of the earliest events in integrin signaling is the activation of Syk (15)(16)(17). Integrin signaling via Syk is involved in neutrophil spreading, respiratory burst and degranulation (11), costimulation of the expression of interleukin (IL)-1b in monocytes (18), and extracellular signal-regulated protein kinase activation in macrophages (10,19).
Integrin activation of Syk is a result of the direct interaction between integrin cytoplasmic domains and the N-terminal SH2 domains of Syk (20,21) that results in Syk clustering and either transactivation (22) or activation by associated src family kinases (22). Immune response receptor activation of Syk requires interaction between Syk tandem SH2 domains and immunoreceptor tyrosinebased activation motif (ITAM)-containing adaptor proteins such as DAP12 or FcRg (13,14,23). Current models suggest that the direct association between integrin cytoplasmic domains and Syk allows for Syk recruitment into integrin signaling complexes that contain ITAM-bearing adaptor proteins, which results in maximal activation of Syk and downstream effectors (13,14,22,23).
The integrin:Syk signaling axis is a promising area for therapeutic intervention. Identifying antagonists of integrin cytoplasmic domain interactions with Syk would provide new molecular tools to elucidate the nature of integrin and ITAMcontaining adaptor molecule co-signaling through Syk in leukocyte activation. Here, we describe the development of high-throughput screening (HTS) systems that were used to identify inhibitors of integrin cytoplasmic domain interactions with Syk. These inhibitors, when incorporated into cells, impede integrin signaling through Syk but do not prevent FcgRI signaling, demonstrating that specific integrin proximal signaling pathways can be targeted while leaving other signaling pathways intact.

Cell Lines
Sources of Cell Lines THP-1 cells were sourced from ATCC. The cell line was derived from a male source.

Protein Purification
Syk expression construct-transduced BL21(DE3) cells were grown overnight at 37°C in Luria broth medium (Invitrogen) in the presence of 100 ug/ml ampicillin. The overnight culture was then diluted with LB medium to an OD600 of 0.05 and further grown at 37°C. Once the culture reached an OD600 of 0.4-0.5, protein expression was induced by adding isopropylthio-b-galactoside to a final concentration of 0.1 mM. After 3 h of induction at 37°C or overnight at 25°C, the cells were collected by centrifugation at 12,000 × g for 20 min and stored at −20°C. Frozen pellets were suspended (30 ml per L of culture) in phosphate-buffered saline (PBS) lysis buffer (1.8 mM KH 2 PO4, 10.1 mM Na2HPO4, 137 mM NaCl, 2.7 mM KCl, pH 7.4) with complete, EDTA-free protease inhibitors (Roche, cat#11873580001), 2 mM phenylmethylsulfonyl fluoride (PMSF), 2 mM dithiothreitol (DTT), and 1% Triton-X 100. The collected cells were disrupted through sonication (Qsonica, 50% amplitude, 15-s pulse with 30 s intervals for 3 min, repeated 3-5×). The lysate was clarified at 16,000 × g for  Figure S2).

Biotinylated GST
We expressed and purified GST following the Protein Purification protocol above (pGEX-4T-1 in DH5a as per manufacturer's instructions, GE Healthcare Lifesciences), and then we modified it with biotin linked through a (PEG) 3   (Solulink Sulfo ChromaLink Biotin, Solulink, cat#B-1007). Covalent protein modification is expected to occur between lysine side chains and the water soluble sulfo-NHS ester moiety. The ChromaLink reagent contains an internal bis aryl hydrazone (l max 354 nm) that allows for simple quantitation by UV-vis analysis (Thermo Scientific Nanodrop spectrometer). The molar substitution ratio was calculated from the ratio of the bis aryl chromophore concentration to protein concentration, both using Beer's Law, and was determined to be approximately one, or, on average, one biotin per GST. Integrity of the biotinylated protein was verified by SDS-PAGE and AlphaScreen assays (Supplemental Figures S2 and S3).

Primary and Null Amplified Luminescent Proximity
Homogeneous Assay (AlphaScreen) The following HTS assay was developed to identify antagonists of the interaction between the integrin b3 cytoplasmic domain and the non-receptor tyrosine kinase Syk. The assay uses the AlphaScreen assay format (26)(27)(28)(29)(30) as a measure of the binding between the integrin b3 cytoplasmic domain residues 716-762 and the tandem SH2 domains of Syk (residues 6-370). b3 peptides were synthesized with a penultimate N-terminal biotinylated lysine residue ( Figure 1D), and Syk residues were expressed and purified as a GST-fusion protein [e.g. GST-Syk(6-370)] ( Figure 1E). Background signals in this assay were determined by omitting the integrin b3 peptide. Specific assay conditions are described in Table 1. Briefly, anti-GST conjugated AlphaScreen acceptor beads (Perkin Elmer cat# 6760603) were coated with GST-Syk(6-370) and plated in Optiplate-384-well plates (Perkin Elmer). Then, either dimethyl sulfoxide control or compound was added, followed by biotinylated b3 peptide, and avidin-coated AlphaScreen donor beads. After a 4-h incubation period at room temperature (while protected from light), assays plates were read on an Infinite M1000 (Tecan) multiplate reader (Ex. 680/30; Em. 570/100).

ELISA-Based Orthogonal Assays
For ELISA-based assays, 96-well plates (Costar cat# 9018) were coated overnight at 4°C with neutravidin (Thermo Scientific, cat#31050) in 10 mM Tris pH 9 and 100 mM NaCl. After overnight incubation, non-specific binding sites were blocked with bovine serum albumin (BSA) (2% w/v) for at least 1 h at room temperature. Plates were washed in 25 mM HEPES pH 7.5, 100 mM NaCl, 0.1% BSA, and 0.01% TritonX-100 (assay buffer), and indicated biotinylated b3 peptide was incubated for at least 2 h at room temperature (0.5-3 ug/ml in di-H 2 O). After washing in the assay buffer, GST-fusion proteins were added (50 ul, indicated concentrations) and incubated at room temperature for 1 h in assay buffer. For compound inhibition studies, compounds were added to GST-fusion proteins (at indicated concentrations) before they were added to plates.  All reagents were dispensed using a Thermo Multidrop combi, sequence 3-5. 6 TopSeal-A (cat# 6050195 from Perkin Elmer), centrifuge at 1,000 rpms for 1 min before the 4-h incubation at room temperature in the dark 7 Plate reader with integrated, dedicated AlphaScreen module and laser light source.
Adhesive Cover (cat#4360954) and centrifuged for 2 min at 800 × g at room temperature. Thermal denaturation was measured on a Quant Studio 6 Flex Real-time PCR system (Applied Biosystems). Reaction conditions were as follows: initial 2:00 min hold at 25°C; ramp up in 1°C increments to a final temperature of 95°C; 2:00 min hold at 95°C; run method used was " Step and Hold" 1:00 min intervals.

Peptide Affinity Pull-Down Assays
THP-1 cells were washed in PIPES buffer (10 mM PIPES, 50 mM NaCl pH 6.8) and resuspended at 10 × 10 6 cells/ml in PIPES lysis buffer [PIPES buffer containing 1% Triton X-100, 1 mM EDTA, 1 mM Na 3 VO 4 , 50 mM NaF, 150 mM sucrose, protease inhibitors (cOmplete, Mini, EDTA-free Protease Inhibitor Cocktail, Roche, cat#11836170001)] and 0.1 mM Nethylmaleimide. The cells were incubated overnight in lysis buffer at 4°C with end-over-end rotation. Lysates were then passed through a 21-gauge needle and centrifuged at 13,000 × g for 30 min at 4°C. Lysate protein concentrations were determined by using Quick Start Bradford Dye reagent (BioRad cat#5000202). Cell lysates (100 µg/ml) were treated with or without drug (at indicated concentrations) for 2 h at room temperature before the peptide affinity assays. In parallel, 300 µl high-capacity Neutravidin agarose resin (Thermo Scientific, cat#29204) was washed once with 5 ml of PIPES lysis buffer and incubated (2 h at room temperature) in 1 ml of PIPES lysis buffer containing 250 µg of integrin b3 cytoplasmic domain peptide and 20% FBS. b3 peptide-bound beads were then washed once with PIPES lysis buffer and resuspended in 600 µl PIPES lysis buffer. Cell lysate treated with or without drug was then incubated with 100 µl b3 bound beads for 1 h at room temperature followed by washing (3×) with PIPES lysis buffer. Endogenous Syk bound to b3 peptide beads was eluted with 40 µl of reduced SDS sample buffer. Eluted samples were run on SDS-PAGE (4-20%, BioRad, cat# 4561093) and immunoblotted for Syk using anti-Syk antibody at 1:1,000 (4D10; SantaCruz) dilution followed by GAM-HRP at 1:5,000 and detected with SuperSignal West Pico (Thermo Scientific, cat# 34580). Western blots were scanned, and band intensity was measured using ImageJ (31) to compare and determine the effect of the drug on Syk/b3 binding. Beads alone were used as a negative control.
In Figure 1C, GAM was immobilized in all wells (2 h at 37°C in 200 mM Na 2 HCO 3 pH 9.2, blocked for 1 h in 2% BSA), followed by the addition of Poly-L-Lysine (PLL) or indicated monoclonal antibodies (5 ug/ml) in Tris-buffered saline (TBS) overnight at 4°C. After washing wells in serum-free RPMI, untreated or drug-loaded THP-1 cells (5 × 10 6 cells per well) were plated in standard serum-free RPMI (1.5 ml total volume), allowed to adhere for 10 min at 37°C 5% CO 2, and then immediately lysed in RIPA buffer (1% Triton X-100, 50 mM Tris-HCl pH 7.2, 150 mM NaCl, 0.25% NaDOC, 0.05% SDS, 2 mM PMSF, 1 mM NaF, 1 mM Na 3 VO 4 , and protease inhibitor cocktail) (AEBSF, Aprotinin, E-64, Bestatin, Leupeptin and Pepstatin; VWR Cat#97063-010). Cells kept in suspension served as non-adherent controls and were lysed under the same conditions. For Syk immunoprecipitations, 5 × 10 6 THP-1 cell equivalents were incubated overnight at 4°C with monoclonal antibody 4D10 at 4 ug/mg protein and were rotated. After overnight incubation, protein G-sepharose (20 ul packed beads washed in RIPA) (GE Healthcare cat# 17061801) was added to lysates, which were incubated for 2 h at 4°C, while rotating. Sepharose was washed 3× in RIPA by centrifugation, pelleted, and then resuspended in SDS-PAGE sample buffer (reducing). For whole cell lysates not subjected to immunoprecipitation, approximately 1 × 10 6 cell equivalents were loaded for SDS-PAGE under reducing conditions (4-20%, Thermo Scientific, cat#25204). After electrophoresis, proteins were transferred to nitrocellulose for 1 h 100 V room temperature, blocked in 5% non-fat milk in TBS-T, and probed with the indicated primary antibody at 1:1,000 overnight at 4°C, followed by species-appropriate HRP secondary antibody (Southern Biotech) at 1:5,000 (30 min at room temperature), and detected using Pierce West Pico (Thermo Scientific cat# 34580). For blot stripping, nitrocellulose was subjected to 0.4 N NaOH for 20 min at room temperature, blocked as described above, and re-probed with the indicated antibodies (primary antibody at 1:1,000 for 1 h at room temperature, followed by HRP-conjugated secondary antibody at 1:5,000 for 30 min at room temperature).

Quantitative PCR
THP-1 cells (2 × 10 6 per treatment group) that were glycerol loaded with ceftazidime (100 µM) or vehicle control (as described above) were plated on CS-1 peptide (10 µg/ml; 6well plates) in serum-free RPMI-1640. Control cells were also kept in suspension. Plated and suspension cells were incubated for 4 h at 37°C. After incubation, the medium was removed, and RNA was isolated from cells with a Qiagen RNA extraction kit. After cDNA synthesis, quantitative real-time PCR was used to determine the relative gene expression levels of IL-1b and monocyte chemoattractant protein (MCP)-1 in THP-1 cells adhered to CS-1 in the presence of vehicle or ceftazidime relative to that of cells in suspension. The cycle threshold (Ct) results generated were analyzed according to the comparative Ct method (2 -DDCt ), where DCt is the difference between the CT value of target versus GAPDH control, and DDCt is the difference between the DCt values of a given treated adherent cell sample and that of cells in suspension. Statistical differences between the two treatment groups for each target were determined by a twotailed, unpaired Student t test using GraphPad Prism 7.04 statistical software.

Modeling
All docking was done in triplicate for both large and focused search boxes and was initialized with random seeds. Autodock Tools was used to prepare ceftazidime for docking into the input molecular model of SYK, prepared using standard protocols (34,35) from crystallographic structure 4FL2 (36). Docking was completed with Autodock Vina 1.1.2 (37, 38) using an initial unbiased search box (Supplementary Figure S7) with center (−11.6, 3.6, 34.0) Å and size (90,80,80) Å and then refined by using a focused box encompassing the putative binding sites, identified from the first round. Resultant PDBQT files were split into each PDB using split2pdb and Babel. TCL scripts were used to automate running VMD post processing and generating MATLAB input for KMeans analysis in either four or six bins. Clusters for each bin were then analyzed to generate heat maps with normalized contact numbers stored into the Beta-Factor PDB field for easy structure visualization using either VMD 1.9.2 or PyMOL 2.2.2. The best scored poses were further evaluated using molecular dynamics simulations in GROMACS (2016.5intel-2018-GPU-enabled). GROMACS was run using standard established protocols, but, in short, the GROMOS9643a1 force field was used with a system prepared from the 4FL2 complex with ceftazidime from the best scored Autodock results. The total system charge was neutralized in an explicit water box with 10 Å distance between the protein/ceftazidime complex and the box edge for periodic boundary conditions. Energy minimization was performed to remove interatomic clashes and then NVT ensemble sampling for equilibration was Vcompleted to define the initial conditions of 10 ns MD simulation production runs.

Quantification and Statistical Analysis
All statistical analysis and details for experiments can be found in the figure legends, the Methods, and the Results. Microsoft PowerPoint was used to generate the figures. GraphPad Prism was used to generate graphs in 1G-L, 2A-C, 3A-D, 4A, 4C, 5D, 6B-D, S1, S3B, S3D, S3E, S4, S5, and S6. FlowJo was used to generate Figure 4B. MATLAB was used to complete KMeans statistical analysis of docking poses. VMD was used to generate Figures 5A-C, E, and S7. PyMOL was used to generate Figures 5F, G, and GRACE was used to generate the graph in Figure 5H.

Integrin Activation of Syk and Direct
Binding of Syk to the Integrin b3 Cytoplasmic Domain Syk is phosphorylated upon integrin engagement of the ligand (17,18), which can involve its direct interaction with integrin cytoplasmic domains (20, 21) ( Figure 1A). When the myelomonocytic cell line THP-1 is plated on the integrin a4b1 substrate VCAM-1, phosphorylation of Syk on multiple tyrosine residues (Tyr323, Tyr352, Tyr 525/526) is rapid and prolonged ( Figure 1B). Integrin-induced phosphorylation of Syk can be recapitulated with plastic immobilized monoclonal antibodies specific for the integrin b1 subunit (mAb TS2/16), the b2 subunit (mAb 76C3), or the integrin aVb3 (mAb LM609) ( Figure 1C). This phosphorylation is independent of the potential interaction of cellular Fc receptors with immobilized goat-anti-mouse capture antibody, as forced non-specific adhesion to GAMcoated surfaces with the use of poly-L-Lysine does not result in Syk phosphorylation ( Figure 1C). Model peptides of the integrin cytoplasmic domains have demonstrated their direct interactions with intracellular effector molecules, including Syk (39). To develop a binding assay between integrin cytoplasmic domains and Syk, we synthesized integrin b-chain cytoplasmic domains with an N-terminal Gly-Lys-Gly-Gly sequence in which the Lys residue was modified with a biotin-conjugated long-chain (LC) carbon linker ( Figure 1D). Syk kinase is modular and comprises N-terminal tandem SH2 domains followed by a large kinase domain ( Figure 1E) (40). The integrin cytoplasmic domain binding site localizes to a region in Syk within the tandem SH2 domains (residues 6-370) (20). Although integrin b1, b2, and b3 cytoplasmic domains can directly interact with Syk, b3 was chosen for assay development because its relative binding affinity is higher than that of the other two integrin cytoplasmic domains (21). We developed an HTS based on the AlphaScreen format because it is a homogeneous system amenable to miniaturization and has been used to screen protein:protein interactions (26)(27)(28)(29). A schematic of the assay format (including the null screen) is provided in Figure 1F. To first demonstrate binding between the integrin b3 cytoplasmic domain constructs and GST-Syk(6-370), we performed ELISAbased assays in which plastic adsorbed neutravidin was used to capture biotinylated b3 peptide, and GST-Syk(6-370) binding was determined by using anti-GST-HRP conjugated antibodies. GST-Syk(6-370) binding to the integrin b3 cytoplasmic domain was dose-dependent, saturable, and specific ( Figure 1G), demonstrating that the synthetic peptide was appropriate for further assay development. For AlphaScreen assay development, avidin-coated donor beads were used to capture biotinylated b3 cytoplasmic domain peptide. Anti-GST-conjugated acceptor beads were used to capture GST-Syk(6-370). The binding between the integrin b3 cytoplasmic domain peptide and Syk juxtaposed the donor and acceptor beads, resulting in a fluorescent signal after donor beads are activated at 680 nm. b3 cytoplasmic domain interaction with GST-Syk(6-370) was dose dependent and selective, as no signal was generated when purified GST was conjugated to the acceptor beads ( Figure 1H). Furthermore, the GST-Syk(6-370) binding curve demonstrated the classic biphasic "hook" effect, whereby excess GST-Syk(6-370) competes in binding to acceptor beads and b3 peptide resulting in a diminished signal. A shortened b3 cytoplasmic domain peptide [b3(sh)] (Figures 1D, I) also binds GST-Syk(6-370) (20) (Supplemental Figure S1). Deletion of the four Cterminal residues of the integrin b3 cytoplasmic domain decreases Syk binding (20). This peptide was synthesized (b3sh759X, Figure 1D) and tested in an ELISA-based format, which demonstrated decreased binding to GST-Syk(6-370) ( Figure 1I). When tested in the AlphaScreen format, decreased binding was also observed ( Figure 1J), again demonstrating specificity of the assay. The N-terminal SH2 domain of Syk is sufficient for interaction with integrin cytoplasmic domains (21). As such, GST-Syk(1-99) binds integrin b3 cytoplasmic domain ( Figure 1K). As a final test of AlphaScreen assay specificity, we removed the GST tag from GST-Syk(6-116) by thrombin cleavage, and we used this purified isolated N-terminal SH2 domain (Supplemental Figure S2) as a competitive inhibitor in the AlphaScreen assay. As demonstrated in Figure 1L, Syk(1-116) attenuated the signal in a dose-dependent fashion. The AlphaScreen assay is susceptible to false-positive "hits" from compounds that can, for example, quench fluorescent signal or react with singlet oxygen. Thus, a counter-screen was developed to identify false-positive hits ( Figure 1F, right schematic and Supplemental Figure S3A). For this assay, we expressed and purified GST and then modified it with biotin linked through a (PEG) 3 spacer (Solulink Sulfo ChromaLink Biotin) to a molar substitution ratio of approximately 1. This was calculated based on absorption of the bis-aryl hydrazone in the biotinylation linker domain (Supplemental Figure S3B) and GST protein concentration. Integrity of the biotinylated protein was verified by SDS-PAGE (Supplemental Figure S3C). Biotinylated GST demonstrated a dose-dependent hook effect, and unmodified GST did not generate a signal (Supplemental Figure S3D). Soluble unmodified GST was used as an inhibitor to demonstrate assay specificity (Supplemental Figure S3E). The primary b3:GST-Syk(6-370) AlphaScreen and false-positive counter-screen appeared sufficiently tractable to translate into an HTS format.

HTS Assay Validation and Screen of Prestwick Chemical Library Collection
In a checkerboard dose-response analysis of biotinylated b3 peptide and GST-Syk(6-370), concentrations of 100 nM peptide and 0.3 nM GST-Syk(6-370) gave the highest signal: background ratios (between 10 and 20 fold) for the screen. When these analyte concentrations were tested in an automated HTS system, the signal:background ratios ranged from 6 to 10 fold. In the automated systems, we tested adding reagents in a different order. A maximal signal was generated when reagents were added in a sequential manner starting with the addition of GST-Syk(6-370) to anti-GST acceptor beads (with a 1-h incubation at room temperature), followed by the addition of biotinylated b3 peptide (1-h incubation at room temperature) and then avidin-coated donor beads. Kinetic tests indicated that the maximal signal was completely generated by 4 h after the final addition of donor beads. DMSO sensitivity tests indicated the assay could withstand 1% DMSO with no effect on the maximum signals generated. The pilot HTS was to be performed in the presence of 0.25% DMSO; therefore, this finding was within the limits of the assay, and a pilot screen was initiated. To determine assay reproducibility, we screened the 1,280compound Prestwick Chemical Library Collection on three separate occasions (Figure 2A). The following is a summary of the screening metrics from the primary screen: average maximum signal = 30,908; average SD of maximum signal = 2,052; average S/B ratio = 5.6; and the average Z'-score = 0.7. A false-positive counter-screen was performed once, and the metrics were as follows: average maximum signal = 44,456; average SD of maximum signal = 2,134; average S/B ratio = 29; and the average Z'-score = 0.8. The final concentration of compounds tested was 25 uM. A scatter plot of the assay results comparing all repeat combinations is presented in Figure 2A. From the primary screen, a total of 17 compounds demonstrated an average of ≥50% inhibition of fluorescent signal ( Figure 2B). However, several compounds also demonstrated inhibition in the false-positive counter screen ( Figure 2B, white bars). Using a hit criteria of greater than 50% inhibition of fluorescent signal in the primary screen and a greater than 3-fold difference in inhibition over the false-positive counter screen, we identified 11 preliminary hits ( Figure 2B, indicated by asterisks). The hits from the primary and secondary false-positive screens were next obtained from a separate commercial source and verified for activity. Most of the re-sourced compounds demonstrated some level of inhibitory activity; however, only four of these fulfilled the original hit criteria as outlined above ( Figure 2C, indicated by the number sign). This finding represents a true hit rate of 0.3 percent from the Prestwick library collection. The hits that were identified in the AlphaScreen were the following antibiotics: methacycline, cefsulodin, ceftazidime, and neomycin sulfate. Based on the success of this screen, we also screened the Maybridge Hitfinder library of 14,400 compounds. Although 32 compounds were identified as hits in the primary screen, none of the re-sourced compounds retained the 3-fold difference between primary and null screens (Supplemental Figure S4).

Hit Potency and Orthogonal Screens
To determine the IC 50 of the hit compounds, we performed doseresponse analyses ( Figure 3A). In the AlphaScreen assay, the most potent drug was neomycin sulfate (IC 50 = 38.0 ± 19 nM), followed by ceftazidime (IC 50 = 1.02 ± 0.2 µM), cefsulodin (IC 50 = 4.9 ± 4.2 µM), and methacycline (IC 50 = 22.4 ± 4.1 µM). The Hill coefficients of dose-response curves were close to 1.0. As an orthogonal screen, we performed ELISA-based assays with bb3(fl) and GST-Syk . Under these assay conditions, methacycline was not soluble at the highest doses tested, and neomycin sulfate did not demonstrate inhibitory activity. Cefsulodin and ceftazidime remained active with a similar rank-order of activity; however, they were less potent in this assay format than in the AlphaScreen format ( Figure 3B). Ceftazidime (the most potent of the two remaining hits) was further tested in assays of whole cell lysates to verify the antagonism of the binding of wild-type full-length Syk to integrin cytoplasmic domains. Neutravidin-coated agarose beads were bound with bb3(fl) and incubated with THP-1 whole cell lysate, with or without ceftazidime. Bead-bound Syk was detected by western blot. Ceftazidime inhibited wild-type Syk interaction with integrin b3 cytoplasmic domains ( Figure 3C). We performed thermal denaturation experiments to demonstrate interactions between hits (ceftazidime and cefsulodin) and GSTfree Syk(6-370). Syk(6-370) melting temperatures (T m ) were measured with and without compound. Cefsulodin and ceftazidime decreased the T m of Syk(6-370) in a dose-dependent fashion ( Figure 3D).

Effects of Ceftazidime on Integrin Signaling via Syk
Integrins of the b1, b2, and b3 families can signal through the tyrosine kinase Syk ( Figure 1C) (17,18,41). To determine whether ceftazidime was a selective inhibitor of integrin cytoplasmic domain interactions with Syk, we performed AlphaScreen assays using b1, b2, and b3. Ceftazidime was equally active against all three integrins (1.62 ± 0.41 µM, 1.53 ± 0.21 µM, and 1.87 ± 0.2 mM, respectively) ( Figure 4A). We used the THP-1 cell line in experiments designed to determine whether antagonists of the integrin b3 cytoplasmic domain interaction with Syk would prevent integrin signaling via Syk. Integrin expression on THP-1 cells was analyzed by flow cytometry to determine expression levels of integrins a4b1, b2, and avb3, and the immune response receptor FcgRI; we wanted to identify integrins that were expressed at similar levels as FcgRI because our goal was to determine if antagonists of integrin:Syk interactions would also affect the canonical activation of Syk that occurs via interaction with dually phosphorylated ITAMs (pITAMs) found within transmembrane adaptor molecules associated with immune response receptors such as FcgRI (42). Of the three integrins tested, integrin a4b1 was expressed at similar levels as FcRgI ( Figure 4B). Because ceftazidime is equally active against all three integrins in vitro ( Figure 4A), we examined the effects of ceftazidime on integrin a4b1 signaling through Syk. The results were negative in initial experiments performed by simply adding ceftazidime to THP-1 cells and examining a4b1 signaling through Syk. We hypothesized this was due to a low accumulation of ceftazidime within cells, which led to the development of a facile approach of loading THP-1 cells with the drug by using glycerol shock. This method was validated with a fluorescent probe, sulfo-Cy5 (Supplemental Figure S5), which demonstrated accumulation of the dye in THP-1 on increasing concentrations of glycerol. In this method, THP-1 cells were loaded with ceftazidime (at indicated doses) or vehicle (DMSO) control, rested for 10 min in serum-free medium (RPMI-1640) containing drug or vehicle control, and then plated on the a4b1 integrin substrate CS-1. Cells kept in suspension demonstrated little phosphorylation of Syk Tyr352 ( Figure 4C). However, when cells adhered to CS-1, prominent tyrosine phosphorylation was observed ( Figure 4C). In the presence of ceftazidime, a4b1induced phosphorylation of Syk was inhibited ( Figure 4C). This was not due to ceftazidime inhibition of THP-1 cell adhesion to CS-1 (Supplemental Figure S6).

Modeling Ceftazidime Binding to Syk Tandem SH2 Domains
Ceftazidime was docked (Autodock Vina 1.1.2) (34, 35) onto the Syk tandem SH2 domains (PDB 4FL2) (36). Cluster analysis indicated several possible binding sites for ceftazadime within the tandem SH2 domains of SYK. Figures 5A-C show three clusters from both the large (W1) and focused (W2) search windows that span the entire tandem SH2 domain (Supplemental Figure S7). The heat maps identify regions in the intersection between the Nterminal and C-terminal SH2 domains, along with the interdomain A region. The K-means analysis of docking poses shows significant preference for clusters 1 and 2. Cluster 1, located at (−10.9, −5.3, 30.0) Å in the PDB frame, has 43 members that average 3.9 +/− 3.4 Å within that cluster centroid ( Figure 5D, Figure 5A purple sphere 1) and is adjacent but not in direct competition with the residues known to interact between pITAM peptides and the tandem SH2 domain ( Figure 5E; yellow shaded surface) (43). Both cluster 1 and cluster 2 interact with the tandem SH2 domain in areas that are not dominated by interactions with the SYK catalytic domain ( Figure  5E) (36). Molecular dynamics simulations (10 ns each) of the ceftazidime and Syk complex further reinforced that cluster 1 is a

Ceftazidime and Selective Inhibition of Integrin Signaling via Syk
The modeling results suggest ceftazidime binding to Syk tandem SH2 domains would have limited effects on the interaction between pITAM peptides and Syk and on Syk's catalytic activity. In AlphaScreen assays that measured the binding between pITAM and the Syk tandem SH2 domains ( Figure 6A), ceftazidime had limited activity (IC 50 > 300 uM), although free pITAM peptide completely inhibited this interaction (IC 50 = 97.1 ± 29.2 nM; n = 3) ( Figure 6B). As expected, ceftazidime had no effect on the catalytic function of full-length Syk ( Figure 6C) despite testing at 100-fold higher concentration than the AlphaScreen IC 50 . On the basis of these results and modeling, we hypothesized that ceftazidime may inhibit integrin signaling via Syk but leave canonical FcgRI signaling via Syk intact. THP-1 cells were loaded with ceftazidime (100 µM) and plated on a4b1 substrate CS-1 or anti-FcgRI mAb (to crosslink this receptor). Integrin a4b1 induced phosphorylation of Syk, and proline-enriched tyrosine kinase-2 (Pyk-2), a kinase that is downstream of Syk (44), was inhibited by incorporation of ceftazidime ( Figure 6D). Integrin signaling in THP-1 cells resulted in upregulation of pro-inflammatory cytokines such as IL-1b and MCP-1 ( Figure 6E). Similar to phosphorylation of Syk, adhesion-dependent induction of the expression of IL-1b and MCP-1 was significantly inhibited by incorporation of ceftazidime ( Figure 6E). However, when FcgRI was crosslinked on the surface of ceftazidime-loaded cells, tyrosine phosphorylation of Syk was not inhibited ( Figure 6D).

DISCUSSION
Integrin signaling through the non-receptor tyrosine kinase Syk is due to direct interactions between Syk and integrin b-chain cytoplasmic domains (20). In this report, we have identified a blactam-containing cephalosporin antibiotic ceftazidime that inhibits integrin signaling through Syk without inhibiting  ELISA-based assay formats (45). Ceftazidime inhibited endogenous Syk (from THP-1 cell lysate) binding to b3 cytoplasmic domain peptides immobilized on sepharose beads ( Figure 3C). Thermal denaturation curves of the tandem SH2 domains of Syk in the presence of ceftazidime indicated a destabilizing effect of compound binding ( Figure 3D) (46), which suggests that ceftazidime acts through binding the tandem SH2 domains of Syk and not the integrin cytoplasmic domains. When loaded into THP-1 cells, ceftazidime inhibited integrin-dependent phosphorylation of Syk ( Figure 4C) and adhesion-dependent expression of the inflammatory cytokines IL-1b and MCP-1 ( Figure 6E), independent of the effects on cell adhesion (Supplemental Figure  S6). In summary, ceftazidime inhibited integrin b-chain cytoplasmic domain interactions with Syk in four different assay formats, including cell-based integrin signaling assays. There is debate as to whether integrin signaling via Syk requires Syk SH2 domain interactions with phosphorylated ITAMs, or whether integrins signal via Syk in an ITAM-independent manner. Initial studies in heterologous expression systems demonstrated that integrin signaling through Syk was ITAM independent (20,21,47). Selective inhibition of integrin-dependent phosphorylation of Syk with ceftazidime, while leaving FcgRI-induced phosphorylation of Syk intact, is consistent with this model. Also supporting this model are observations of allbb3 signaling in platelets from Syk R41Afl/fl;PF4-Cre mice that demonstrate normal aIIbb3-dependent phosphorylation of Syk R41A despite abrogated GPVI and CLEC-2 signaling, which requires an intact phosphotyrosine binding N-SH2 domain (48). In vitro studies (22) also support ITAM-independent integrin coupling to Syk; these investigators found binding of Syk to integrin cytoplasmic domains presented in high density (to model integrin clustering) resulted in Syk transphosphorylation independent of ITAM interactions. However, integrin signaling in neutrophils and macrophage was shown to be ITAM-dependent in studies in primary cells in which pITAM signaling of Syk was abrogated by retroviral reconstitution of Syk −/− cells with Syk containing non-functional SH2 domains or when pITAM signaling adaptors such as DAP12 or FcRg were knocked out (10,49). It is unclear whether this represents differences in integrin signaling via Syk in disparate cell types or technical differences in the model systems tested. Nevertheless, the mechanisms by which integrins couple with Syk can clearly be both ITAM-independent and ITAM-dependent, and as described by Mocsai et al. (23), these mechanisms need not be mutually exclusive and most likely occur together as integrin co-signaling with a variety of cell surface receptors appears to converge on Syk and Syk family kinases. By demonstrating that ceftazidime can selectively inhibit integrin signaling through Syk while not altering the ability of FcRs to signal via Syk, we provide further evidence of the independent regulation of this kinase by these two different receptor systems. Syk activation occurs through direct interaction with phosphorylated immune-receptor tyrosine-based activation motifs (50) or through integrin clustering that results in transphosphorylation on tyrosine residues and activation of catalytic function (22). The interaction of Syk with pITAM containing immune response receptors or signaling adaptor molecules is well defined, as is the resultant conformational changes that lead to activation of Syk kinase activity (36,43). However, the interaction between integrin cytoplasmic domains and Syk is less well defined. Although the interaction occurs within the tandem SH2 domains of Syk, integrin cytoplasmic domains and pITAMS do not compete for binding to Syk (20,22) and are in fact additive when used in combination to activate this kinase (22), which suggests independent modes of binding. Modeling of the ceftazidime binding pockets in the Syk tandem SH2 domains identified 2 major interaction clusters that were outside of regions important for interaction between pITAM peptides and Syk ( Figure 5). These appeared to be at an interface between the SH2 domains and interdomain A region. This mode of binding is consistent with the lack of effect of ceftazidime on Syk kinase activity ( Figure 6C), the relatively poor antagonist activity of ceftazidime against pITAM:Syk interactions ( Figure 6B), and the lack of effect of ceftazidime on FcRg1 signaling through Syk ( Figure  6D). Further high-resolution structural and biochemical studies are necessary to definitively identify a mode of binding of antagonists such as cefsulodin and ceftazidime to the Syk tandem SH2 domains (including potential covalent modification), along with integrin bchain cytoplasmic domain interactions with the Syk tandem SH2 domains. These studies may also aid in identifying specific regions in the tandem SH2 domains of Syk that specifically interact with integrin b-chain cytoplasmic domains.
A limitation of this work is that the screens performed were cell free, which precluded identifying small molecule compounds that could both inhibit integrin cytoplasmic domain interactions with Syk and accumulate intracellularly. As such, the compounds identified had to be loaded into THP-1 cells by glycerol shock to demonstrate their effects on integrin signaling in cells ( Figure 6). Studies are currently ongoing to identify small molecule compounds that are predicted to be membrane permeable and that models would predict to have a similar mode of binding to the Syk tandem SH2 domains as the b-lactam antibiotics identified here. Identifying such compounds will be useful as chemical probes to ascertain the importance of integrin signaling via Syk in inflammatory and autoimmune disease models.
Currently, all drugs approved for human use that target integrin cell adhesion molecules are antagonists that prevent cell adhesion. These include the aIIbb3 targeting drugs abciximab (51,52), tirofiban (53,54), and eftifibatide (55); the a4b1 and a4b7 targeting humanized mAb natalizumab (56); the a4b7 humanized mAb vedolizumab (57,58); and the aLb2 drugs efalizumab (59) (withdrawn from the market due to safety concerns) (60) and lifategrast (61). These first-generation integrin antagonists all target integrin ectodomains and thus prevent both cell adhesion and integrin signal transduction. Similarly, inhibitors of Syk that target its kinase domain, such as fostamatinib (approved for use in immune thrombocytopenia), lack selectivity (62). Thus, current inhibitors of Syk have off-target selectivity issues and cannot differentially inhibit signaling via one upstream receptor over the other. As demonstrated here and elsewhere (63,64), identifying antagonists of integrin cytoplasmic domain interactions with intracellular signaling effector molecules or adaptor proteins may lead to selective targeting of integrin signaling pathways that contribute to disease pathogenesis, while limiting the effects on cell adhesion, thereby leaving potentially beneficial integrin signaling pathways intact.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.