High-affinity pan-specific monoclonal antibodies that target cysteinyl leukotrienes and show efficacy in an acute model of colitis[S]

Cysteinyl leukotrienes (CysLTs) are a small family of biological signaling lipids produced by active leukocytes that contribute to diverse inflammatory disease states as a consequence of their engagement with dedicated G protein-coupled receptors. Immunization of mice with a CysLT-modified hapten carrier protein yielded novel monoclonal antibodies that display variable binding affinity to CysLTs. Solution binding assays indicated differing specificities among the antibodies tested, with antibody 10G4 displaying a preference for leukotriene C4 (LTC4). X-ray crystallography of a humanized 10G4 Fab fragment in complex with LTC4 revealed that binding induces a hook-like conformation within the hydrocarbon tail of the lipid arachidonic acid moiety. Specific hydrogen bonding to the LTC4 carboxylate groups further stabilized the complex, while a water molecule mediated a hydrogen bond network that connected the N-terminal arm of l-glutathione to both the arachidonyl carboxylate of LTC4 and the antibody heavy chain. Prophylactic administration of two anti-CysLT antibodies in mice followed by challenge with LTC4 demonstrated their in vivo efficacy against acute inflammation in a vascular permeability model. 10G4 ameliorated the effects of acute dextran sulfate sodium-induced colitis, suggesting that anti-CysLT antibodies could provide a therapeutic benefit in the treatment of inflammatory diseases.

In support of efforts aimed at developing therapeutics to treat inflammatory diseases mediated by CysLT signaling, we have produced pan-specific murine monoclonal antibodies (10G4, 2G9, 9B12, and 14H3) and carried out in vitro binding studies to measure their binding affinity and specificity for CysLTs. In preparation for assessing its clinical potential as a treatment for inflammatory diseases, a humanized version of 10G4 (Hu10G4) was produced and shown to preserve the LTC 4 binding profile of its murine precursor. To understand the source of specificity and to gain insight into its mode of antigen recognition, we have determined the 1.75 Å X-ray cocrystal structure of a Hu10G4 Fab fragment in complex with LTC 4 . Treatment of mice with 10G4 or 2G9 serves to inhibit LTC 4 -dependent movement of fluid across blood vessels into peritoneal tissues. Furthermore, 10G4 administration mitigates disease symptoms of mice in a dextran sulfate sodium (DSS)induced model of acute colitis. This study illustrates that humanized monoclonal antibodies can be produced that target specific CysLTs and convey unique physiological properties that are of potential therapeutic value.

Preparation of LTE 4 -protein conjugates
The immunogen was prepared by incubating 0.22 mg of LTE 4 , 2.5 mg of BCP, and 2.9 mg of BS3 in 90% PBS/10% DMSO for 2 h at room temperature, followed by purification of the proteinlipid conjugate using 7 kDa MWCO Zeba desalting columns equilibrated with Imject purification buffer. The LTE 4 -BCP flow through material was used directly for immunization. LTE 4 was also crosslinked to Imject BSA through BS3 according to the manufacturer's recommendations. This LTE 4 -BSA conjugate was used as the coat material for competition ELISA and KinExA experiments. week-old female Swiss Webster mice were initially immunized by two subcutaneous injections of 0.025 mg (0.05 mg total) of immunogen (LTE 4 -BCP) emulsified in complete Freund's adjuvant. After 21 days, the mice were boosted with a single intraperitoneal injection of 0.05 mg of LTE 4 -BCP emulsified in incomplete Freund's adjuvant. Every week thereafter, the mice received a single intraperitoneal injection of 0.05 mg LTE 4 -BCP emulsified in incomplete Freund's adjuvant for an additional 8 weeks.

Immunization and hybridoma screening
Serum samples were collected 3 days after the second, third, fifth, and ninth boosts and screened for the presence of potential anti-CysLT antibodies by direct ELISA. Briefly, 96-well ELISA plates were coated with 1 g/ml LTE 4 -BSA conjugate diluted in 0.1 M carbonate buffer (pH 9.5) at 4°C overnight. Plates were blocked with 1% BSA in 1× PBS plus 0.1% Tween-20 (blocking buffer) for 1 h at room temperature. Then the plates were washed with 1× PBS and serum samples serially diluted in blocking buffer were added to the wells and incubated for 1 h at room temperature. The plates were washed and a 1:5,000 dilution of HRP conjugated to goat anti-mouse IgG (#1030-05; Southern Biotech) in blocking buffer was added to the wells. After 1 h, the plates were washed and developed using TMB (Invitrogen). The reaction was quenched after 5 min by adding 1.0 M H 2 SO 4 . The absorbance at 450 nm was measured using a Perkin Elmer plate reader (#1420) and the data were analyzed using GraphPad Prism software.
Spleens from mice that displayed high titers were harvested and used to generate hybridomas with the ClonaCell ® -HY hybridoma cloning kit. Once the hybridomas were grown to confluency in medium E, the cell supernatants were collected and the mouse IgG quantified by ELISA.
To confirm that the hybridoma supernatants contained antibodies that bound the unconjugated (native) CysLTs in solution, ELISA plates were coated with LTE 4 -BSA, as described for the direct ELISA above. LTC 4 , LTD 4 , or LTE 4 starting at 1, 10, or 30 M, respectively, was serially diluted (3-fold) using PBS containing 0.1 mg/ml FAF-BSA and 50 ng/ml IgG (9B12, 10G4, 2G9, or 14H3 cell supernatants). The plates were washed and the titrations were transferred into LTE 4 -coated wells and allowed to incubate for 4 h at room temperature or until equilibrium. After washing, the bound IgG was detected using an HRP anti-mouse IgG secondary antibody and the data analyzed as described above.

Antibody sequencing, production, and purification
After three rounds of limiting dilution and growth in medium E, the hybridoma subclones were transferred to Iscove's DMEM plus Gibco FBS and Cellgro supplements (no penicillin/streptomycin). Total RNA was isolated from 5 × 10 6 cells using the Nu-cleoSpin RNA kit. mRNA was isolated from total RNA using oligo d(T)25 magnetic beads and used to generate first strand cDNA followed by TdT tailing and PCR amplification, following the manufacturer's protocol for 5′-RACE cloning (Invitrogen). The immunoglobulin heavy chain variable region cDNA was generated using a mouse IgG1k isotype-specific primer (5′-TATG-CAAGGCTTACAACCACA-3′). The TdT-tailed cDNA was PCR amplified using a 5′ anchor primer (5′-GGCCACGCGTCGAC-TAGTACGGGIIGGGIIGGGIIG-3′) with a nested 3′ primer (5′-CA-CAATTTTCTTGTCCACCTTGGTGC-3′). The product of the reaction was purified using a NucleoSpin gel and PCR clean-up kit and sequenced using a reverse primer (5′-CCTTGACCAG-GCATCCCA-3′). The variable domain of the heavy chain was then amplified and inserted as an AgeI/AfeI-digested fragment and ligated into the expression vector, pFUSE-CHIg-mG1. The immunoglobulin kappa chain variable region was amplified using an isotype-specific primer (5′-CTCATTCCTGTTGAAGCTCTT-GACAAT-3′). The TdT-tailed cDNA was PCR amplified using the same 5′ anchor primer as for the heavy chain variable region, plus a kappa chain nested 3′ primer (5′-CTCATTCCTGTT-GAAGCTCTTGACAATGGG-3′). The product of the reaction was purified using a NucleoSpin gel and PCR clean-up kit and sequenced using a reverse primer (5′-AGTTGTTCAAGAAGCA-CACGA-3′). The variable domain of the light chain was then amplified and inserted as an AgeI/BstAPI digested fragment and ligated into the expression vector, pFUSE2-CLIg-mK.
Murine monoclonal antibodies were produced from stable CHO-M cell lines transfected with the heavy and light chain expression vectors described above. E1000 shake flasks were seeded at 0.3 × 10 6 cells/ml in SFM4CHO media and supplemented with ActiCHO feeds. After 10 days in culture, the harvest was clarified by centrifugation (4,000 g) followed by in-line filtration (0.45-0.20 m; Sartorius). The Hu10G4 was produced from transient transfection HEK293 cells, and the supernatant was collected after 5 days in culture.
Anti-CysLT IgG antibodies were purified from the clarified harvests using affinity chromatography (10G4/2G9: protein G resin; Hu10G4: MabSelect protein A resin). Purified IgG was then passed through a Sartobind Q cartridge (Sartorius) before being dialyzed against 1× PBS and concentrated using Amicon stirred cell concentrator with a 50 kDa MWCO membrane (Millipore).

Measurement of antibody:CysLT binding affinity by KinExA
The equilibrium dissociation constant for individual CysLTs was determined by KinExA (Sapidyne Instruments) using a KinExA 3200 instrument equipped with an autosampler. The LT conjugate used to capture the free antibody was prepared by linking 5S-hydroxy-6R-(S-cysteinyl)-7E,9E,11Z,14Z-eicosatetraenoic acid to BSA in 0.1 M sodium phosphate and 0.15 M NaCl (pH 7.2) as described above. The purified LTE 4 -n-sub-BSA conjugate was diluted with running buffer (PBS without calcium and magnesium with 0.002% azide) to 30 g/ml, adsorbed to PMMA beads, and blocked with 150 M fraction V FAF-BSA (Calbiochem).
Aliquots of LTE 4 , LTC 4 , and LTD 4 were transferred to new amber glass vials and the methanol was evaporated using a dry argon stream. The dried CysLT aliquots were resuspended in running buffer containing 15 M FAF-BSA by repeated sonication and vortex mixing to a final CysLT concentration of 0.3 mM.
Samples containing antibody and FAF-BSA (1.5 M final) were prepared with running buffer in silanized glass tubes and the CysLT was added and serially diluted over 14 fractions. Antibody and starting CysLT concentrations varied depending on experiment and ranged from 25 to 500 pM IgG and 5 nM to 3 M, respectively. Sample fractions were equilibrated for >1 h at room temperature before performing equilibrium affinity experiments at 0.5-1 ml/min. Antibody captured on the beads was detected using an AlexaFluor or Dylight goat anti-mouse or human secondary (Jackson ImmunoResearch) depending on the antibody being tested. Each fraction was sampled and analyzed in duplicate using the KinExA Pro software version 3.6.3 (Sapidyne Instruments).

Humanization of 10G4
The variable domains of the murine anti-CysLT monoclonal antibody, 10G4, were humanized by grafting the murine complementarity-determining regions (CDRs) into human framework regions. Suitable acceptor human framework sequences were selected using IgBLAST (16). Human immunoglobulin heavy variable 4-59 (accession number AB019438) was selected for the humanized version of the 10G4 heavy chain variable domain and JH6 (accession number J00256) was used for the heavy chain J region. For the light chain, VKI O12 (accession number X59315) was the human framework selected for the10G4 light chain variable domain. JK2 (accession number J00242) was used for the J region. The CDR sequences were those of the murine antibody 10G4. Humanized 10G4 variants containing various combinations of mutations to the murine amino acid in the framework regions (back mutations) of both the heavy and light chains were constructed, expressed, and evaluated for binding to LTC 4 , LTD 4 , and LTE 4 using ELISA and KinExA. The variant that showed the highest affinity for the CysLTs (Hu10G4) was selected for crossreactivity and crystallization studies.

Cross-reactivity
Aliquots of LTE 4 and LTC 4 were dried down under argon. Each lipid was biotinylated at a ratio of 20:1 biotin:lipid using Pierce EZ-link NHS-LC-LC-Biotin kit according to manufacturer's instructions (Thermo). ELISA plates (96-well; Greiner) were coated with 0.5 g/ml coating material (Hu10G4: goat anti-human IgG, Fc-specific Jackson #109-005-098; 10G4 and 2G9: goat anti-mouse IgG, Fc-specific Jackson #115-005-071) diluted in 0.1 M carbonate buffer (pH 9.5). Plates were sealed with thermal adhesive and allowed to incubate at 4°C overnight. Plates were washed with 1× PBS plus 0.05% Tween-20 and then blocked with 1% BSA in 1× PBS plus 0.1% Tween-20 for 1 h at room temperature. The plates were washed and primary antibody diluted in 1× PBS (Hu10G4, 100 ng/ml; 10G4, 50 ng/ml; 2G9, 50 ng/ml) was added to the plate and allowed to incubate for 1 h at room temperature. The plates were washed. All reference and test competitors were purchased from Cayman Chemicals with the exception of l-cysteine (Pierce), and l-cysteine-l-glycine (Sigma). A 12 point 3-fold dilution series of reference competitor starting at 1 M LTC 4 and 10 M LTE 4 was used for 10G4/Hu10G4 and 2G9, respectively. A 12 point 3-fold dilution series of test competitor starting at 30 M was used for Hu10G4, 10G4, and 2G9. All reference and test competitors were diluted in 0.5 nM tracer in 0.5× PBS plus1 mg/ml BSA (Hu10G4 and 10G4, LTC 4 -LC-LC-biotin; 2G9, LTE 4 -LC-LC-biotin). Diluted competitor solution was applied to the plate and allowed to incubate for 21 h at room temperature. The plates were washed and a 1:60,000 dilution of secondary antibody in blocking solution (peroxidase-conjugated streptavidin Jackson #016-030-084) was allowed to incubate on the plates for 15 min. The plates were washed and developed by allowing cold TMB (Invitrogen) to incubate on the plates for approximately 5 min before the reaction was quenched by addition of 1.0 M H 2 SO 4 . Plates were read at 450 nm on a Perkin Elmer plate reader (#1420) and the data were analyzed using GraphPad Prism software.

Fab production and complex formation
The full-length humanized 10G4 monoclonal antibody was produced from transient 293-F cells (Invitrogen) using a 3:1 light:heavy chain ratio of plasmid. Cultures were collected after 5 days. Antibodies were purified by protein-A affinity chromatography using MabSelect resin, dialyzed against 1× PBS, and concentrated to 11.25 mg/ml using Centricon-30 centrifugal concentrators (Millipore). Concentration was determined by absorption at 280 nm. Purified full-length antibody was incubated for 24 h at 37°C in a 160:1 ratio with activated papain (Thermo) in digestion buffer [20 mM cysteine-HCl, 50 mM sodium phosphate (pH 7.2), 20 mM EDTA]. The Fab was purified by protein-A affinity chromatography using Pro-Sep-vA resin (Millipore) and purity was confirmed by SDS-PAGE. The purified Fab was dialyzed against 50 mM Tris-HCl and 150 mM sodium chloride (pH 7.5). The purified Fab was concentrated to 14.4 mg/ml using Centricon-30 centrifugal concentrators (Millipore), sterile filtered, and stored at 4°C until ready for use. Eighty nanomoles of LTC 4 (Cayman) were dried down in a glass vial under argon gas. The sample was then resuspended in 30 l 1× sample buffer [50 mM Tris-HCl, 150 mM sodium chloride (pH 7.5)]. The sample was vortexed vigorously for 2 min, then sonicated 5-10 min and vortexed for an additional 2 min to completely resuspend lipid. Thirty-two nanomoles of total Fab were added to lipid solution and allowed to incubate on a nutator for 24 h at 4°C. The Fab:LTC 4 complex was collected through a 0.22 m cellulose acetate filter (Corning) and stored at 4°C until use. The final complex concentration was 11.3 mg/ml.

Complex cocrystallization
Fab:LTC 4 complex cocrystals were grown by the vapor diffusion hanging drop method using the microcrystal additive screening approach (17). Briefly, 1 l of antibody complex was combined with 1 l reservoir containing 0.2 M ammonium sulfate, 0.1 M sodium cacodylate trihydrate (pH 6.5), and 30% PEG 8000 and incubated in the sealed well of a tissue culture plate at room temperature over 0.5 ml of reservoir solution. Needle-like crystals grew after 7 days. A microseed stock was prepared by resuspending the entire crystallization drop in 100 l fresh reservoir solution followed by vigorous vortexing and storage at 20°C. The microseed stock (0.6 l) was then added to a drop containing 2.4 l of Fab:LTC 4 complex and 0.8 l of a new reservoir solution composed of 0.15 M zinc acetate dihydrate and 20% PEG 3350 and the resulting drop was sealed and incubated at room temperature over 0.5 ml of this second reservoir solution. Single crystals of dimensions 0.2 × 0.1 × 0.04 mm 3 formed in roughly 5 days.

X-ray crystallography
Crystals were harvested directly from the mother liquor using nylon loops and transferred to a cryopreservative solution containing 20% ethylene glycol and flash-cooled by plunging into liquid nitrogen. X-ray diffraction data were collected on an ADSC Quantum 200 CCD detector at ALS synchrotron beamline 5.0.2 at LBNL. A complete set of diffraction data was collected and processed in HKL-2000 (18). The data indexed to a primitive monoclinic lattice and the clear absence of reflection intensities at the 2n+1 values in 0k0 indicated a P2 1 space group. Matthew's coefficient analysis suggested that with one complex in the asymmetric unit, the solvent content was 42%. Statistics for the scaled intensity data are presented in Table 3.
The Hu10G4 Fab:LTC4 X-ray cocrystal structure was solved by molecular replacement. Atomic coordinates for LT1009, the humanized anti-sphingosine-1-phosphate antibody (PDB ID: 3I9G), were modified by removal of all nonbonded atoms and deletion of the six CDR loops. The resulting model was employed as a probe using PHASER from the CCP4i package (19,20). Breaking the probe into two rigid ensembles, one containing both the heavy and light chain variable domains and the other composed of the two constant domains, produced a single clear solution after rotation and translation functions were run against working data within the 10.0-4.0 Å range. Statistics for molecular replacement are included in Table 3.
Rigid body maximum likelihood refinement with REFMAC5 was carried out against 50.0-3.40 Å data resulting in a starting R cryst of 43.6% and R free of 43.2% (21). Initial restrained refinement with overall refinement of B factors was next carried out to 2.70 Å and CDR loops and side chains were built using 2F O -F C and F O -F C difference electron density maps in COOT (22). The resulting model was refined to 2.10 Å resolution and density for the ligand was clearly seen in the resulting F O -F C electron density map. Idealized coordinates and a library file for LTC 4 were created using the Sketcher module in CCP4i and placed in the map density by hand in COOT. Refinement against working data to 2.0 Å with individual B factor refinement clearly revealed that the two ends of the glutathione moiety were incorrectly placed. Moreover, three strong spherical peaks could be observed with  > 10 in the resulting F O -F C map, each corresponding to a zinc ion. The ligand was manually rebuilt and three zinc ions, as well as individual water molecules, were placed by hand and refined against all working data to 1.90 Å. Model building was completed and refined by maximum likelihood with individual isotropic B factor refinement against all working data to 1.75 Å. Model stereochemistry was validated by MolProbity in Phenix (23,24). The final model contained light chain amino acids 1-214 and heavy chain residues 1-214, one LTC 4 molecule, three zinc ions, one molecule each of ethylene glycol, polyethylene glycol, and Tris, and 363 waters. The final model refinement statistics and geometry are reported in Table 3. Figures were prepared using PyMOL (25).

Vascular permeability
C57BL/6 mice were obtained from Charles River Breeding Laboratories. Test antibody (1.5 nmol) (2G9 or 10G4) was combined with 1.5 nmol LTC 4 and left to incubate overnight. Mice were injected intravenously with 0.2 ml 0.2% Evan's Blue in PBS (Sigma) and then immediately injected intraperitoneally with saline containing no antibody (control), nonspecific antibody plus LTC 4 , or either 2G9 or 10G4 antibody plus LTC 4 . After 30 min, mice were euthanized by isoflurane overdose and the peritoneum was washed with 1.5 ml PBS. The wash was then centrifuged at 500 g for 10 min. The optical density of the wash was read at 610 nm (Spectronic Genesys 2).
For prophylactic subcutaneous injection experiments, mice were injected with vehicle (0.5% DMSO in PBS), nonspecific antibody, or 10G4 (30 mg/kg) 24 h prior to lipid treatment. Mice were injected intravenously with 0.2 ml 0.2% Evan's Blue in PBS. Immediately after this step, mice were injected with 1.5 nmol LTC 4 . After 30 min, mice were euthanized by isoflurane overdose and the peritoneum was washed and prepared for spectrophotometry analysis, as described previously.

Acute DSS-induced colitis
Female C57BL/6 mice (17-20 g) (Shanghai SLAC Laboratory Animal Co. Ltd.) were housed five to a cage in a standard vivarium (PharmaLegacy Laboratories) at 20-26°C at a relative humidity of 40-70%. The animals had ad libitum access to rodent food (Shanghai SLAC Laboratory Animal Co. Ltd.) and water (Phar-maLegacy Laboratories).
Animals were randomly divided into seven treatment groups (n = 10 per group) on day 1: group 1, control; group 2, vehicle (PBS); group 3, cyclosporine A (CsA); group 4, LT1014; group 5, LT1002; group 6, 10G4; and group 7, 2G9. Groups 2-7 received 2% DSS in their drinking water for 6 days followed by regular drinking water for 4 days. Group 1 received regular drinking water for 10 days. Animals in group 1 remained the untreated controls while mice in groups 2-7 received respective vehicle, reference article, or test article.
Body weight and stool consistency (0, normal; 2, loose stools; 4, diarrhea) were monitored daily throughout the study. All animals were euthanized by CO 2 asphyxiation followed by cervical dislocation on day 10. The colon was harvested free from surrounding tissue. Colon length (from the ileocecal valve to the anus) and weight were measured. The colon was longitudinally bi-sectioned, one piece processed as Swiss-roll and immersed in neutral buffered 10% formalin, and the other piece snap-frozen in liquid nitrogen and stored at 80°C. The frozen colon tissue was minced on ice and homogenized in 0.5% hexadecyltrimethylammonium bromide dissolved in 50 mM phosphate buffer to measure myeloperoxidase activity. The fixed colon tissue was embedded in paraffin, sectioned at 4 m thickness, and stained with hematoxylin and eosin. In a blinded fashion that treatment information was unknown to the slide reader, the stained slides were examined by a histopathologist (Phar-maLegacy Laboratories) and assessed with histopathological scoring for inflammatory cell infiltration (0, occasional or no infiltrate; 1, infiltration into lamina propria; 2, infiltration into submucosa; 3, transmural infiltration) and tissue damage (0, no mucosal damage; 1, focal crypt lesions; 2, mucosal erosions or ulcerations; 3, extensive damage affecting the submucosa). Data were expressed as group means ± SEM for body weight and colon length. Statistical analyses were performed on each parameter using GraphPad Prism, SPSS, or Sigmaplot. A value of P < 0.05 was considered statistically significant. All animal studies were conducted in conformity with the Public Health Service Policy on Humane Care and Use of Laboratory Animals, incorporated in the Institute for Laboratory Animal Research Guide for Care and Use of Laboratory Animals and approved by the Institutional Review Board of BTS Research, San Diego under IACUC 1111-03.

Monoclonal antibodies bind LTC 4 , LTD 4 , and LTE 4 with high affinity
Four monoclonal hybridomas (10G4, 2G9, 9B12, and 14H3) were generated by screening mice immunized with LTE 4 conjugated to protein carriers via disuccinimidyl homobifunctional cross-linking (26). LTE 4 was chosen on the basis that it is the stable end product of CysLT extracellular metabolism (27). Interestingly, IgG secreted from these hybridomas displayed markedly different binding properties to unconjugated LTC 4 , LTD 4 , and LTE 4 , based on a competition ELISA with LTE 4 -BSA-coated plates (Fig. 1). Comparison of the four antibodies for target binding revealed that 10G4 bound LTC 4 with the highest affinity, while 2G9 and 9B12 appeared to exhibit a preference for binding LTE 4 . Monoclonal antibodies from these three clones showed similar binding to native LTD 4 using this competition format. The 14H3 clone displayed relatively weak binding to all three lipids, so it was excluded from further characterization.
The apparent high-affinity CysLT binding observed with the competition ELISA above was confirmed in solution by KinExA. In this assay, the amount of free antibody present in an antibody:lipid complex solution at equilibrium was measured after capture on lipid-coated beads. Equilibrium dissociation binding constants (K d ) for the purified mouse monoclonal antibodies 10G4, 2G9, and 9B12, as well as Hu10G4, prepared by grafting the six CDR loops from 10G4 onto a human IgG scaffold and then mutating seven additional residues back to the murine sequence are reported in Table 1. The 10G4 and Hu10G4 antibodies demonstrated picomolar range binding affinity for each of the three CysLTs tested. 10G4 bound with highest affinity to LTC 4 , with a measured K d value of less than 10 picomolar. As observed previously by ELISA, the 2G9 and 9B12 antibodies exhibited different target binding profiles. The 2G9 bound preferentially to LTE 4 with a measured K d value of roughly 50 picomolar. The K d values for 9B12 binding to LTC 4 , LTD 4 , and LTE 4 were all higher than those of 2G9, so 9B12 was excluded from further target specificity studies.

The 2G9 and 10G4 are pan-specific antibodies against CysLTs
Using a competition ELISA with a biotinylated tracer molecule, we next evaluated the ability of a wide range of related lipids and CysLT 1 R receptor antagonist compounds to bind the immobilized antibodies (supplemental Table  S1). The IC 50 values for each unlabeled competitor normalized as a percentage relative to unlabeled LTE 4 and LTC 4 for 2G9 and 10G4, respectively, are listed in Table 2.
Expansion of the study to include all of these compounds paints a clear picture of target binding specificity and reveals that, although 2G9 and 10G4 were generated using the same immunization and screening processes, these monoclonal antibodies exhibited different relative binding preferences for LTC 4 and its metabolites. Antibody 2G9 bound with highest affinity to LTF 4 and N-acetyl LTE 4 , followed by LTE 4 > LTD 4 = LTC 4 , while 10G4 bound these compounds in the following order: LTC 4 > LTF 4 > N-acetyl LTE 4 > LTD 4 > LTE 4 . The specificity profile suggests that both 2G9 and 10G4 recognize the fatty acid hydrocarbon chain and the sulfidopeptide group on the respective CysLTs, and that both substituents of the CysLTs must be present in a proper conformation for high affinity binding. However, it is also clear that the mode of CysLT binding differs between 2G9 and 10G4.
In addition to the rank order of binding affinity for the main CysLTs, 2G9 and 10G4 also showed differences for other sulfidopepide-containing LTs. Antibody 2G9 exhibited weak, but measurable, binding to eoxin E 4 and C 4 , whereas 10G4 failed to display any detectable binding to these compounds. Likewise, 2G9 suffered an 10-fold loss in binding when the C11 cis double bond in LTC 4 isomerized to the trans conformation, where 10G4 showed a >100-fold loss of binding due to this alteration. Methyl esterification of the C1 carboxylate group resulted in a significant decrease in the ability of LTE 4 or LTC 4 to cross-react with preformed CysLT complexes with either 2G9 or 10G4, suggesting that this carboxylate group contributes significantly to antigen binding by both antibodies. This limited structure-activityrelationship study suggests that, although both antibodies Ninety-five percent confidence intervals are given within parentheses. UD, unable to be determined.  specifically target CysLTs, 10G4 shows a preference for LTC 4 and displays greater selectivity for its preferred target antigen than does 2G9.

Hu10G4 Fab:LTC 4 complex crystal structure
Humanization of 10G4 resulted in a monoclonal antibody, Hu10G4, with similar affinity and specificity properties as compared with the parental murine 10G4 antibody ( Table 2, Fig. 1). In order to elucidate the mechanism of CysLT selective binding by 10G4, we crystallized and determined the three-dimensional structure of the Hu10G4 Fab fragment in complex with LTC 4 (Fig. 3A). The X-ray cocrystal structure was solved by molecular replacement and refined to 1.75 Å resolution resulting in a model with excellent geometry and stereochemistry ( Table 3).
The Hu10G4:LTC 4 complex crystallographic model reveals that LTC 4 binds to the antigen binding site primarily by burying its hydrocarbon tail. The combination of trans double bonds at carbons 7 and 9 followed immediately by cis double bonds at carbons 11 and 14 serves to support adoption by the lipid hydrocarbon tail of a hook-like structure that inserts deep within the antigen binding site between the heavy and light chain variable domains. The glutamate residue of the LTC 4 glutathione moiety extends away from the antigen binding site. Analysis of relative temperature (B) factors and poor electron density map quality at this residue suggests that the glutamate -carbon exhibits conformational flexibility relative to the remainder of the bound lipid antigen molecule (Fig. 3B). This is not too surprising, as this region corresponds to the end of the LTE 4 immunogen that was anchored to its hapten carrier protein during immunization. Both the arachidonate and glutathionyl glycine carboxylic acid groups participate in hydrogen bond networks involving the side chains of amino acid residues from CDR loops of both the antibody heavy and light chains. In its bound conformation, the LTC 4 lipid presents 542.1 Å 2 molecular surface area (probe radius 1.00 Å). Binding of LTC 4 to Hu10G4 excluds 275.5 Å 2 (50.83%) of its molecular surface (Fig. 3C) (28). The degree to which antibody binding buries LTC 4 is typical, based on a published analysis of the occluded surfaces of diverse antibody-bound nonpeptide antigens relative to their respective unbound molecular surface areas (29).

Structural determinants of LTC 4 binding affinity and specificity by Hu10G4
A detailed analysis of the noncovalent interactions between atoms of the Hu10G4 antibody Fab fragment and its LTC 4 lipid antigen reveals the chemical details of its binding affinity and selectivity. In all, side chains of sixteen amino acids from each of the six heavy and light chain CDR loops come within close contact of the bound LTC 4 antigen [the Hu10G4 antibody heavy and light chains are numbered according to the system of Kabat et al. (30); the letters "L" and "H" immediately prior to amino acid numbers indicate that they derive from the light or heavy chains, respectively]. Nine of these (TyrH33, SerH34, IleH37, TrpH47, TyrH52, AlaH96, ProH98, ArgH99, and TrpH101) emanate from the heavy chain and seven (LeuL36, ArgL46, TyrL49, LeuL89, TyrL91, ArgL96, and PheL98) are from the light chain. Two additional heavy chain amino acids (AsnH35A and AsnH50) contact the antigen indirectly through watermediated hydrogen bonds (Fig. 4A).
A complex hydrogen bond network stabilizes the carboxylate group of the LTC 4 arachidonate moiety (Fig. 4B). This supports our observations that methylation at this carboxylate group decreases affinity of LTC 4 binding to Hu10G4. ArgL96 and TyrH52 are both positioned within hydrogen bonding distance of the same carboxylate oxygen atom. An ordered water molecule, bonded to the side chains of AsnH35A and AsnH50, is also positioned to share a hydrogen atom with this oxygen atom. The second arachidonyl carboxylate oxygen atom is located within hydrogen bonding distance of the SerH34 side chain. Interestingly, the amide nitrogen of SerH34 positions a second water molecule to form hydrogen bonds with both the arachidonyl carboxylate group and the amide group oxygen from the glutamyl isopeptide linkage of the LTC 4 l-glutathione moiety. Thus, this second water molecule links CDR-H3 to both the arachidonate carboxylate group and glutamate of LTC 4 . As removal of this glutamate converts LTC 4 to LTD 4 , it is likely that this water-mediated hydrogen bond network contributes significantly to the binding preference for LTC 4 exhibited by the Hu10G4 antibody. This notion is supported by our observation that N-acetyl LTE 4 , which contains an acetyl group that can accept the water-mediated hydrogen bond, recovers a significant amount of the binding affinity lost in the conversion of LTC 4 to LTD 4 .
Other than the indirect water-mediated contact to the amide carbonyl oxygen, the remainder of the LTC 4 glutathionyl glutamate residue does not contact Hu10G4. The glutathione cysteine and glycine amino acids reside within close contact to TyrL49 and TyrH33, respectively, and the ProH98 side chain at the tip of the CDR-H3 loop provides a "spine" over which the LTC 4 antigen drapes like a coat on a rack. The side chain of ArgL46 is positioned within hydrogen bonding distance to the amide carbonyl oxygen of the peptide bond linking cysteine and glycine and the glycine terminal carboxylate group contacts the side chain of ArgH99 (Fig. 4C). This latter interaction likely explains the further decrease in binding affinity observed upon removal of glycine in the conversion of LTD 4 to LTE 4 , as well as the slight decrease in observed binding affinity toward LTF 4 (a CysLT in which the glycine residue only is removed from LTC 4 ).
The majority of the LTC 4 molecular surface that becomes buried upon binding to Hu10G4 is the hydrocarbon tail from the lipid arachidonate moiety. The tail adopts a hook-like conformation that fits snugly into a hydrophobic cavity formed at the antibody binding site by side chains from light chain residues, LeuL36, TyrL49, LeuL89, TyrL89, and PheL98 and heavy chain residues, IleH37, TrpH47, AlaH96, and TrpH101 (Fig. 4D). A buried water molecule, held in place by hydrogen bonds from the indole nitrogen of TrpH101 and backbone carbonyl oxygen atoms from ProH98 and ArgH99, is also found within the hydrophobic binding pocket. The predisposition of the LTC 4 hydrocarbon tail to adopt its observed conformation appears to play a vital role in imbuing Hu10G4 with its antigen binding affinity and specificity. The observed inability of Hu10G4 to bind to polyunsaturated lipids with altered double bond patterns, such as PGE 2 and LTB 4 (the product of the 5-LO catalyzed reaction of arachidonic acid), serves to support this hypothesis. It also explains why isomerization from cis to trans of the C11 double bond in LTC 4 severely decreases binding affinity. Finally, in addition to its role in supporting the hydrophobic antigen binding pocket, TyrL91 also participates in a water-mediated hydrogen bond with the hydroxyl group from the C5 position on the LTC 4 arachidonate.

Anti-CysLT antibodies inhibit vasopermeability during acute inflammation
LTC 4 has been shown to induce Evan's Blue dye extravasation into a variety of tissues. This includes the respiratory and utero-genital tracts and into the peritoneal cavity after either intravenous or intraperitoneal administration, respectively (31,32). Shortly following intraperitoneal injection in mice, LTC 4 induces vascular permeability in a dose-dependent manner and likely helps to mediate the fluid accumulation phases of the acute inflammatory response (31). Using the mouse model, we evaluated the potency of 2G9 and 10G4 to inhibit the activity of LTC 4 to  (19). d Calculated against a cross-validation set of 5.0% of data selected at random prior to refinement. e Calculated by MolProbity (24). f Combines clashscore, rotamer, and Ramachandran evaluations to a single score, normalized to the same scale as X-ray resolution (24). induce plasma exudation into the peritoneal cavity by monitoring dye extravasation (Fig. 5).
Initially, LTC 4 was premixed with equimolar or molar excess of antibody 2G9 and then injected into the peritoneum of mice that had received prior intravenous administration of 0.2% Evan's Blue dye. Dye extravasation was measured 30 min following injection. A statistically significant decrease of dye extravasation was observed for the groups of mice that received 2G9 (P = 0.0011 for equimolar and P < 0.0001 for excess 2G9) compared with the isotype-matched control antibody group (Fig. 5A). 10G4 also appeared to be effective at blocking LTC 4 -induced plasma exudation and restricting dye extravasation to levels similar to the naïve group (Fig. 5B).
After demonstrating that preincubation of 2G9 and 10G4 with LTC 4 neutralized the biological action of LTC 4 , we next investigated to determine whether 10G4 showed activity when administered prophylactically. To this end, two groups of mice were dosed subcutaneously with 30 mg/ml 10G4 or control antibody 24 h prior to LTC 4 treatment and dye extravasation was monitored as above. A significant (P < 0.0001) decrease in measured peritoneal dye was observed for the mice receiving 10G4 pretreatment compared with the control antibody treatment group (Fig. 5B). These data suggest that our anti-CysLT antibodies hold potential as therapeutic agents that effectively block the action of LTC 4 in modulating vascular permeability during the acute phase of inflammation.

10G4 protects against DSS-induced acute colitis in mice
In order to further assess the in vivo efficacy of anti-CysLT antibodies to treat acute inflammation, we next tested to determine whether treatment of mice with 10G4 or 2G9 could provide protection against acute inflammation brought on by administration of DSS, which induces colitis in mice (33,34). C57BL/6 mice that were fed with a 2% DSS water solution for six days followed by regular drinking water during days 7-10 showed a significant decrease in body weight, increased colon weight, increased colon weight versus colon length versus body weight ratio, and increased colon weight versus body weight relative to mice from which DSS was withheld. Pretreatment of the mice with antibody 10G4 served to significantly lessen the symptoms of acute colitis relative to DSS-treated mice that received either no antibody or an isotype control antibody (LT1014) that did not target CysLTs. 10G4 protected the DSS-treated mice from weight loss to the same extent as administration the immunosuppressive drug, CsA (Fig. 6A). The same correlations were observed when mice were analyzed and assessed by their disease activity index score, which takes into account body weight, stool consistency, and severity of rectal bleeding (Fig. 6B).
Histopathological analysis of extracted colon tissue revealed decreased infiltration of inflammatory cells in DSS-treated mice that were administered 10G4 relative to vehicle-or isotype control antibody-treated mice (Fig. 6C). Finally, decreased tissue damage was measured in colon tissue extracted from the cohorts of DSS-treated mice that were administered 10G4 or CsA (Fig. 6D).

DISCUSSION
The elevated production of CysLTs is associated with diverse inflammatory disease states. Syslová et al. (35) measured a 2-fold increase of LTE 4 and LTC 4 in the exhaled breath condensate of patients with moderate asthma relative to healthy controls. LTE 4 levels were detected at 3-fold higher levels in the urine of Crohn's disease patients relative to healthy subjects and a 2-fold increase was measured in patients with ulcerative colitis (36). Urinary LTE 4 levels increase by eight-fold in patients suffering from chronic lung disease of extreme prematurity (37). A 4-fold increase in serum LTC 4 has been measured in eczema patients, while 16-fold increased LTD 4 levels were reported in the serum of patients with hepatocellular carcinoma (38,39). As CysLTs are known to promote inflammatory signaling upon binding to GPCRs, it is likely that these molecules directly promote inflammatory disease states and that therapeutic strategies aimed at interfering with CysLTs before they engage their receptors might be beneficial to clinicians and the patients they treat.
To date, drug discovery in the field of inflammatory diseases has remained focused primarily on development of small molecule therapeutics. For example, there are currently several small molecule therapeutics on the market for the treatment of asthma, including montelukast (Singulair), zafirlukast (Accolate), and zileuton (Zyflo) (40). Each of these drugs interferes with LT signaling or production as montelukast and zafirlukast target CysLT 1 R and zileuton inhibits the enzyme, 5-LO. However, in recent years, interest in antibody-based therapeutics has increased due to characteristics that set them apart from small molecule and natural product therapies. Antibodies can bind immutable targets with high affinity and specificity, which can be further optimized through antibody engineering. Although antibody therapeutics have many potential advantages, there are still challenges that need to be overcome that include limited penetration of membrane barriers and their potential for immunogenicity.
In this study, we describe a set of antibodies that were isolated after immunization of mice with an LTE 4 -protein conjugate. Of the antibodies tested, 2G9 displayed the strongest binding preference for LTE 4 , with 3.8% cross-reactivity with LTC 4 and 5.4% cross-reactivity with LTD 4 . When tested against a battery of modified CysLT and related compounds, 2G9 was found to bind with even greater preference to LTF 4 (a version of LTC 4 in which the glutathionyl glycine has been removed, but the glutamate remains intact), as well as a version of LTE 4 in which glutathionyl cysteine residue is N-acetylated (supplemental Table S1). Therefore, antibody 2G9 appears to display increased antigen binding affinity by supporting interactions with an amide carbonyl oxygen N-terminal to the glutathionyl cysteine, while the presence of a glutathionyl glycine disrupts this and other 2G9:antigen binding interactions. Surprisingly, 10G4 exhibited strong preferential binding to LTC 4 over LTE 4 (4.0% cross-reactivity) or LTD 4 (0.3% crossreactivity). The cross-reactivity of other molecules was less for 10G4 than for 2G9. Moreover, the binding affinity of LTC 4 for 10G4 was remarkably tight, with observed binding dissociation constants derived from KinExA in the low picomolar range.
As described in the present study, the X-ray crystal structure of the Fab fragment of a humanized version of 10G4 in complex with LTC 4 revealed the source of the specificity of this antibody for its lipid antigen. Antigen binding induced a unique conformation within the unsaturated hydrocarbon tail of LTC 4 , while antibody side chains contributed to a water-mediated network of hydrogen bonds that linked the lipid fatty acyl group to both the glutamate and glycine arms of the covalently attached glutathione moiety. Analysis of the amino acid changes that were likely to have occurred during affinity maturation offered insight into the manner by which 10G4, which was raised by immunization against LTE 4 , bound with such exceptional affinity to LTC 4 . Using the NCBI/IgBLAST server to identify likely mouse germline variable domain sequences that gave rise to 10G4 revealed that all of the mutated amino acids mediated contacts with atoms that were common between LTE 4 and LTC 4 (16). For example, conversion of Phe90 in the murine kappa light chain variable domain germline sequence IgkV9-124*01 to TyrL91 in 10G4 provided a hydrogen bond donor that stabilized a water-mediated contact to the C5-OH of the LTC 4 arachidonic acid moiety, while mutation in the IgHV3-1*02 heavy chain variable domain of His52 to TyrH52 supported hydrogen bonding with the arachidonate C1 carboxylate group. These mutations served to make 10G4 a good antibody (K d = 0.3 nM) against LTE 4 . The additional placement of glutamate and glycine amino acids in the glutathionyl moiety of LTC 4 provided additional beneficial binding interactions that did not result from affinity maturation. For instance, the N atom of 10G4 amino acid SerH34 mediated a water-mediated hydrogen bond network that connected glutamic acid with the arachidonate C1 carboxylate and heavy chain CDR loop 1 and ArgH99 lay within hydrogen bonding distance of the carboxylate of glycine. We speculate that these fortuitous positive interactions resulted in the "good" LTE 4 antibody, 10G4, functioning as an exceptional (K d = 5 pM) antibody against LTC 4 .
Prior to this analysis, only one crystallography study of CysLTs bound to a protein was reported. In 2016, Jablonka et al. (44) published X-ray crystal structures of the lipocalin family protein, LTBP1, from saliva of the "kissing bug," Rhodnius prolixus, a vector for the Chagas' disease pathogen, Trypanosoma cruzi, in complex with either LTD 4 or LTC 4 . The protein, which is thought to sequester CysLTs in Fig. 6. Antibody 10G4 protects C57BL/6 mice in a DSS-induced model of acute inflammation. A: The body weight of mice treated with PBS buffer (Vehicle), a non-CysLT isotype control antibody (LT1014), CsA, or 10G4 was monitored during and after treatment with DSS and relative to DSS-untreated (Control) mice. B: Disease activity index scores for the mice. C: Colon tissue sections stained with hematoxylin and eosin to assess immune cell infiltration from untreated (Control) mice and mice treated with DSS after administration of vehicle, 10G4, or nonspecific antibody LT1014. D: Tissue damage scores assigned to samples of colon tissue removed from untreated (Control) and DSS-fed mice treated with vehicle, CsA, antibody LT1014, 10G4, or 2G9. order to inhibit skin inflammation around the site of a bug bite, binds with nearly equal affinity to LTC 4 , LTD 4 , and LTE 4 . The crystallographic models reveal that LTBP1 binds a CysLT by burying its fatty acyl component at the center of the lipocalin -barrel domain, while the glutathionyl moiety causes conformational change in a loop to "lock" the lipid in place. The extended conformation adopted by the hydrocarbon tail of the bound lipid seems unlikely to convey too much specificity. However, the authors report that TXA 2 , PGD 2 , and arachidonic acid each fail to bind to LTBP1 and they go on to show that the loop conformational change is responsible for the CysLT binding affinity.
On the topic of antibody recognition of small molecule lipid antigens: it is interesting to note that each of the three anti-lipid antibodies we have succeeded in analyzing by complex crystallization and X-ray crystallography in this and previous studies used a significantly different approach for selective binding of its specific antigen. Whereas Hu10G4 bound LTC 4 by combining complementarity to a unique conformation available to the lipid unsaturated hydrocarbon tail with side chain-and water-mediated hydrogen bonds, the LT3015 anti-lysophosphatidic acid antibody relied primarily on amino acid side chains from both heavy and light chain CDR loops to bury the unique phospho headgroup of its lysophosphatidic acid antigen, while leaving the hydrocarbon tail largely exposed to solvent (45). The LT1009 humanized anti-sphingosine-1-phosphate antibody used amino acids from its light chain to partially coordinate a pair of calcium ions that bound with high affinity to an otherwise highly solvent-exposed phosphosphingosine head group, while the heavy chain nearly completely engulfed the hydrocarbon tail (46). These differences are partially explained by the various strategies used to anchor the different lipid antigens to their carrier proteins for immunization. In all three cases the atom through which the lipid was anchored to hapten is exposed to solvent. However, it seems likely that the systems studied thus far illustrate only a few of the possible chemical strategies by which antibodies can selectively bind small molecule lipids. While the relatively few existing structural models show convincingly that selective binding of lipids by antibodies is possible, future studies to identify and structurally characterize additional novel high-specificity antibody:lipid complexes are needed to expand this field and serve as structural templates for theoretical and engineering projects on specific lipid binding.
We show by two in vivo models of acute inflammation that anti-CysLT antibodies can convey beneficial outcomes. The decrease in LTC 4 -dependent dye extravasation observed in mice treated with antibody 2G9 was, on the surface, surprising because we had previously found that antibody to bind preferentially to LTE 4 . However, we did observe that 2G9 bound with somewhat significant affinity to LTC 4 (Table 1). Moreover, LTC 4 was metabolized to LTE 4 , which is a potent activator of inflammatory signaling through CysLT 3 R. We suspect that the observed effect of 2G9 relative to control antibody on dye extravasation is a result of a combination of its nanomolar binding affinity for LTC 4 and its ability to compete with the CysLT 3 R for binding to trace amounts of LTE 4 . Although not tested directly against 2G9, the LTC 4 -specific antibody, 10G4, provided a stronger and more statistically powerful protection against the effects of LTC 4 injection on dye extravasation when measured either after subcutaneous or intraperitoneal injection of the antibody relative to saline vehicle or nonspecific antibody. Consequently, 10G4 was further tested in the acute colitis model, where it was found to convey improvement to overall health and colon tissue histopathology in DSS-treated mice.
Our biochemical, structural, and in vivo observations support the development of potential clinical applications for anti-CysLT antibodies. We propose that by binding to free CysLTs, the antibodies prevent their LT antigens from engaging their cognate GPCRs, thus preventing activation of signaling pathways that lead to increased vascular permeability, acute inflammation, and eventually tissue remodeling and fibrosis. It bears mentioning that studies aimed at determining the efficacy of anti-CysLT antibodies against chronic inflammatory conditions did not yield conclusive results.