Charcot-Leyden Crystal Protein (Galectin-10) Is Not a Dual Function Galectin with Lysophospholipase Activity but Binds a Lysophospholipase Inhibitor in a Novel Structural Fashion*

Charcot-Leyden crystal (CLC) protein, initially reported to possess weak lysophospholipase activity, is still considered to be the eosinophil's lysophospholipase, but it shows no sequence similarities to any known lysophospholipases. In contrast, CLC protein has moderate sequence similarity, conserved genomic organization, and near structural identity to members of the galectin superfamily, and it has been designated galectin-10. To definitively determine whether or not CLC protein is a lysophospholipase, we reassessed its enzymatic activity in peripheral blood eosinophils and an eosinophil myelocyte cell line (AML14.3D10). Antibody affinity chromatography was used to fully deplete CLC protein from eosinophil lysates. The CLC-depleted lysates retained their full lysophospholipase activity, and this activity could be blocked by sulfhydryl group-reactive inhibitors, N-ethylmaleimide andp-chloromercuribenzenesulfonate, previously reported to inhibit the eosinophil enzyme. In contrast, the affinity-purified CLC protein lacked significant lysophospholipase activity. X-ray crystallographic structures of CLC protein in complex with the inhibitors showed that p-chloromercuribenzenesulfonate bound CLC protein via disulfide bonds with Cys29 and with Cys57 near the carbohydrate recognition domain (CRD), whereas N-ethylmaleimide bound to the galectin-10 CRD via ring stacking interactions with Trp72, in a manner highly analogous to mannose binding to this CRD. Antibodies to rat pancreatic lysophospholipase identified a protein in eosinophil and AML14.3D10 cell lysates, comparable in size with human pancreatic lysophospholipase, which co-purifies in small quantities with CLC protein. Ligand blotting of human and murine eosinophil lysates with CLC protein as probe showed that it binds proteins also recognized by antibodies to pancreatic lysophospholipase. Our results definitively show that CLC protein is not one of the eosinophil's lysophospholipases but that it does interact with eosinophil lysophospholipases and known inhibitors of this lipolytic activity.

Charcot-Leyden crystals (CLCs), 1 first reported more than 150 years ago by Charcot and Robin (1), are found in a variety of tissues, body fluids, and secretions as hallmarks of allergic inflammation involving eosinophils or basophils. CLC protein, the unique autocrystallizing constituent that forms these distinctive hexagonal bipyramidal crystals, became synonymous with eosinophil lysophospholipase (LPLase) when chromatographic purification of this enzymatic activity yielded protein preparations that formed CLC and contained the CLC protein (2)(3)(4). However, our cloning of the CLC cDNA revealed a polypeptide of 16.5 kDa with 142 amino acids that bore no similarities to any known sequences of LPLases, phospholipases, or other lipolytic enzymes but had sequence similarities to members of the superfamily of S-type lactose-binding and IgE-binding animal lectins (5), now collectively referred to as galectins (6,7). Our subsequent determination of the x-ray crystallographic structure of CLC protein showed that its overall tertiary fold was highly similar to that of the prototype galectins (8), members of the superfamily of ␤-galactosidebinding animal lectins. The structure of CLC protein provided details of a carbohydrate recognition domain (CRD) with both similarities to and differences from other members of the galectin family. These structural findings, the observation of weak affinity for lactosamine-containing sugars (8,9) and a highly conserved gene structure comparable with other galectins (10), resulted in CLC protein being designated galectin-10 (11).
The x-ray crystal structure of CLC protein at 1.8-Å resolution, determined from crystals identical in morphology to the crystals found in vivo, showed that the protein has a "jelly-roll" motif that results from a tight association between a fivestranded and a six-stranded ␤-sheet joined by two 3 10 helices at the two ends (8). The overall structural fold of CLC protein is highly similar to that of galectin-1, -2, and -3 and most similar to galectin-7 (12), but the protein shares only moderate amino acid sequence identity with members of this superfamily. The galectin-10 amino acid sequence is most similar to galectin-3 and -4 (25-30% identity) and less similar to galectin-1 and -2 (15-20% identity) (8), with the BESTFIT algorithm providing similarities of 43-48% for galectins-1, -2, -3, and -7 (9). CLC protein has a CRD containing 9 of the 13 conserved residues in the CRD of galectin-1, -2, and -7, sharing 6 of 8 residues directly involved in ␤-galactoside binding by these galectins (8). Initial carbohydrate binding studies using solid-phased simple sugars (8,9) suggested that CLC protein had specific, albeit weak, binding activity for N-acetyl-D-glucosamine and lactose (8,9), but the mode of carbohydrate binding was not established. However, our recent studies clearly show that these weak carbohydrate binding activities can be attributed to galectin-10 binding to the cross-linked agarose (or Sepharose) matrix and not to the solid-phased lactosamine-containing simple sugars themselves (13). Most recently, using x-ray crystallographic approaches, we have demonstrated that CLC protein does not show any affinity for ␤-galactosides but can bind mannose (in the crystal) in a manner that is unique and different from the binding of lactosamine-containing sugars by other related galectins (14). A detailed analysis of the topology and chemical nature of CLC protein's CRD in complex with mannose provided the first evidence of possible carbohydrate recognition by this unique lectin. The natural carbohydratecontaining ligand(s) of CLC protein has not as yet been identified, and the predicted bifunctional role of CLC protein as both a galectin and LPLase in eosinophil-mediated inflammatory responses remains speculative.
The CLC protein is still considered to be the eosinophil LPLase despite the lack of amino acid sequence or structural homology with any other LPLases. The protein was originally reported to exhibit divalent cation-independent weak LPLase activity with a Michaelis constant (K m ) of 22 M and V max of 50 nmol/h/mg at 37°C for lysopalmitoylphosphatidylcholine substrate (15). The crystal structure of CLC protein in the absence of bound lysophosphatide substrate did not provide direct information about the location of an LPLase active site, although we did suggest a potential catalytic triad based purely on structural considerations (8). However, the nature and mechanism of LPLase activity for CLC protein has remained highly speculative in view of its structural and other similarities to the animal galectins. In the present study, we have reassessed the LPLase activity of CLC protein/galantin-10 in human peripheral blood eosinophils and in the eosinophil-differentiated myelocyte cell line AML14.3D10. Our results show that CLC protein is not the eosinophil's LPLase and that it does not possess LPLase activity of its own, but that it does interact with known inhibitors of eosinophil LPLase activity via amino acid residues in its CRD. However, CLC protein does have the capacity to bind to human and murine LPLases that are recognized by antibodies to pancreatic LPLase. The enzyme responsible for the majority of LPLase activity expressed by the human eosinophil is probably identical to the 75-kDa pancreatic enzyme (16) that we and others have detected in the eosinophil at both the protein and mRNA levels (17).

EXPERIMENTAL PROCEDURES
Cell Lines and Cell Culture-The AML14.3D10 eosinophil-differentiated cell line (18) was kindly provided by Drs. Cassandra Paul and Michael Baumman (Wright State University, Dayton, OH). AML14.3D10 is a fully differentiated eosinophilic myelocyte subline of the AML14 and AML14.eos (19 -21) cell lines that continues to proliferate and maintain a differentiated eosinophil phenotype, including secondary (specific) granules, without cytokine supplementation (18). It displays many of the characteristics of mature peripheral blood eosinophils, including constitutive expression of the major protein mediators of the eosinophil such as the granule cationic proteins (major basic protein, eosinophil peroxidase, eosinophil-derived neurotoxin, and eosinophil cationic protein) and CLC protein (18). In addition, it can be induced to express chemokine receptors (e.g. for eotaxins) (22) and expresses granulocyte-macrophage colony-stimulating factor that drives its proliferation, differentiation, and survival in culture (23). The mature nature of this cell line and its utility for studies of eosinophil biology were described in a recent review (21). AML14.3D10 cells, which express CLC protein in amounts comparable with that of peripheral blood eosinophils (21), were maintained at between 3 ϫ 10 5 and 1 ϫ 10 6 cells/ml in RPMI 1640 medium supplemented with L-glutamine (2 mM), ␤-mercaptoethanol (0.05 mM), sodium pyruvate (1 mM), and 8% fetal bovine serum (Invitrogen).
Purification of Human and Murine Eosinophils-Blood (120 -480 ml; up to 1 unit), from normal, nonallergic, healthy donors was used to isolate eosinophils following methods originally described by Hansel et al. (24). Informed consent was obtained from blood donors according to the guidelines established by the Institutional Review Board of the University of Illinois at Chicago. Sedimentation of erythrocytes was performed at room temperature for 1 h by mixing 50 ml of peripheral blood with 10 ml of Macrodex (6% dextran 70 in 0.9% sodium chloride) and 200 l of 0.5 M EDTA. The leukocyte-containing plasma fraction was harvested, diluted 1:1 with phosphate-buffered saline, and overlaid on Ficoll-Paque gradients (25 ml of cell mixture onto 15 ml of Ficoll-Paque 400 (Amersham Biosciences), followed by centrifugation for 30 min at room temperature. The granulocyte pellets were harvested and washed with phosphate-buffered saline. The remaining erythrocytes were lysed using brief exposure (10 s) to distilled water, and the eosinophils were isolated using a Miltenyi MACS CD16 kit to remove neutrophils following the methods described by the manufacturer (Miltenyi Biotec, Auburn, CA) (25). Total cell counts and eosinophil counts were performed using Randolph's stain. Human eosinophil purity was routinely Ͼ95% as assessed by differential counts of Wright's/Giemsastained cytocentrifuge slides, with eosinophil viability Ͼ95% as determined by trypan blue dye exclusion. Murine blood eosinophils (kindly supplied by the laboratory of Dr. James J. Lee, Mayo Clinic, Scottsdale, AZ) for whole cell lysates for CLC protein ligand blotting were purified from IL-5 transgenic mice to Ͼ98% purity using density gradient centrifugation on Percoll (Amersham Biosciences) followed by negative selection using Miltenyi MACS magnetic beads (Miltenyi Biotec) (26).
Preparation of Whole Cell Lysates-Cell lysates were prepared from 1 ϫ 10 8 to 1 ϫ 10 9 AML14.3D10 cells or 1 ϫ 10 7 to 5 ϫ 10 7 blood eosinophils by sonication (four 15-s sonication bursts with 1-min intervals for cooling) on ice using a model 450 sonifier (Branson Ultrasonic Corp., Danbury, CT) with a microtip at a constant duty cycle and power setting of 3.5. The crude sonicates were centrifuged at 10,000 ϫ g at 4°C for 30 min, and the soluble, cell-free supernatants were used for CLC protein depletion experiments including LPLase activity measurements, depletion, and affinity purification of the protein.
CLC Depletion by Antibody Affinity Purification-Rabbit IgG antibodies to crystal-derived eosinophil CLC protein were affinity-purified on a solid-phase CLC-Sepharose 4B column as previously described (27,28). A solid-phased CLC column was first prepared using CLC protein isolated by crystallization (in vitro formation of Charcot-Leyden crystals from human eosinophil whole cell lysates (29)); the crystal-derived CLC protein preparation contained a single homogeneous 16.5-kDa band on overloaded, silver-stained or Coomassie Blue-stained SDS-PAGE gels. The resolubilized, crystal-derived CLC protein was solidphased to CNBr-activated Sepharose 4B resin using standard methodology (Amersham Biosciences). Antibodies to CLC protein in the serum from a rabbit immunized with pure crystal-derived CLC protein was affinity-purified using the solid-phased CLC-Sepharose column using standard methodology. The affinity-purified anti-CLC IgG antibodies were then chemically cross-linked to Protein A-Sepharose (Sigma) using dimethylpimylimidate as the cross-linker. Lysates of peripheral blood eosinophils or AML14.3D10 eosinophil-differentiated cells that express CLC protein were prepared by sonication as above and fractionated on the anti-CLC column at 50% or less of the column's CLC protein-binding capacity, first determined by saturating the column with CLC protein to quantitate the column's maximum CLC-binding capacity. The CLC-depleted flow-through fractions from the column were collected, concentrated to the original starting volume of whole cell lysate applied to the affinity column, and saved for determination of CLC content and for LPLase activity measurements. The CLC protein bound to the anti-CLC affinity column was eluted with 100 mM triethylamine (pH 11.5). The eluted CLC protein-containing fractions were dialyzed into 100 mM Tris-HCl buffer (pH 7.5) containing 2 mM EDTA and concentrated to the starting volume of the original eosinophil or AML14.3D10 cell lysate using Centricon 10 (Amicon) centrifugal concentrators. The starting cell lysate, the CLC-depleted flow-through fractions from the anti-CLC affinity column, and the eluted, highly purified CLC protein from the same fractionation were all analyzed for CLC content by a highly sensitive CLC protein radioimmunoassay (29), by SDS-PAGE, and by Western blotting using SuperSignal enhanced chemiluminescence reagents (Pierce) for detection and for LPLase activity. Unless otherwise indicated in the figures, equivalent volumes of the starting whole cell lysate, the pooled and concentrated CLC-depleted flow-through, and the eluted affinity-purified CLC protein were analyzed by SDS-PAGE and Western blotting.
Lysophospholipase Activity Assay-Lysophospholipase activity was measured using a radioenzyme assay as described previously without any modifications (15,30). LPLase activity was measured in 50 -200-l samples of whole cell lysates, the pooled and concentrated CLC-depleted flow-through, and the eluted affinity-purified CLC protein or bacterially expressed recombinant CLC protein in a final volume of 1 ml of buffer containing 100 mM Tris-HCl (pH 7.5), 0.01% Tween 20, and 2 mM EDTA. The reaction was initiated by the addition of 192 nmol of palmitoyl lysophosphatidylcholine containing 1 nmol of L-1-[1-14 C]palmitoyl lysophosphatidylcholine substrate with a specific activity of ϳ50 mCi/mmol (Amersham Biosciences). After incubation for 1 h at 37°C, reactions were terminated by immediate extraction with acidified isopropyl alcohol/heptane/sulfuric acid (40:10:1; v/v/v) as previously described (31). Free palmitic acid partitioned in the upper extraction layer was quantitated by scintillation counting of 1 ml of the organic layer, and the mol of released [ 14 C]palmitic acid were determined. One unit of enzyme activity represents 1 mol of fatty acid released/h at 37°C and pH 7.5. The identity of the released reaction product was confirmed by thin layer chromatography (15,30). All LPLase assays included a positive control enzyme, Vibrio sp. phospholipase B (Sigma), run as an internal standard to control for substrate quality and interassay variation in measuring LPLase activity. Results are expressed as the percentage of activity in whole cell lysates, as mol of substrate cleaved/h, or nmol of substrate cleaved/h/10 6 eosinophils or AML14.3D10 cells.
Inhibition of LPLase Activity by Sulfhydryl Group-reactive Agents-Two sulfhydryl group-reactive agents, N-ethylmaleimide (NEM) and p-chloromercuribenzenesulfonate (pCMBS) (Sigma), both previously shown to be potent inhibitors of eosinophil LPLase activity (15), were utilized at concentrations ranging from 10 Ϫ2 to 10 Ϫ5 M. Concentrations of NEM and pCMBS in this range were previously shown to inhibit LPLase activity in eosinophil sonicates by more than 95% (15). For inhibition studies, NEM and pCMBS were added to whole cell lysates (sonicates), CLC-depleted lysates, or purified CLC protein for 30 min at 37°C, prior to assaying LPLase activity.
Antibodies to Native CLC Protein, Rat Pancreatic Lysophospholipase, and Western Blotting-Rabbit polyclonal IgG antibody to crystal-derived CLC protein was prepared and utilized as previously described (27,28,(32)(33)(34). This affinity-purified antibody recognizes a single, 16.5-kDa CLC protein band on Western blots of peripheral blood eosinophil and basophil whole cell lysates and AML14.3D10 cell lysates as well as preparations of antibody affinity-purified and crystal-derived CLC protein. Polyclonal antibody to recombinant rat pancreatic LPLase, kindly provided by Drs. Ralf Kleene and Michael Schrader (Philipps University, Marburg, Germany), was prepared in chickens and used at a dilution of 1:1000. This antibody specifically recognizes a 75-kDa protein in rat zymogen granules (35). Affinity-purified proteins, whole cell lysates, and rat zymogen granule extracts were resolved using 10% SDS-PAGE and transferred to Immobilon polyvinylidene difluoride membranes (Millipore Corp., Bedford, MA) using a Milliblot Graphite Electroblotter II semidry blotting apparatus (Millipore). The blots were blocked overnight at 4°C in 5% Carnation nonfat dry milk in TBS and then incubated with the primary anti-CLC or anti-rat pancreatic LPLase antibody diluted as indicated above in 5% milk in TBS for 1 h at 4°C. The blots were washed 3 times for 15 min each in TBS and incubated with affinity purified, horseradish peroxidase-conjugated goat anti-rabbit IgG (Bio-Rad) at a 1:3,000 dilution or donkey antichicken IgG antibodies (Jackson ImmunoResearch Laboratories, Inc.) at a 1:10,000 dilution in 5% milk in TBS for 45 min at room temperature. Immunoblots were washed three times for 15 min each in TBS and then visualized using the SuperSignal enhanced chemiluminescence kit reagents (Pierce) and Kodak BioMax MR film (Eastman Kodak Co.).
Ligand Blotting with CLC Protein-AML14.3D10 cells and purified mouse and human peripheral blood eosinophils were resuspended in phosphate-buffered saline containing 10 mM ␤-mercaptoethanol, 2 mM EDTA, 10% glycerol, and protease inhibitor mixture (Roche Molecular Biochemicals). After sonication and centrifugation as described above, duplicate sets of whole cell lysates (6 g of total protein/sample) were resolved on each half of a 10% SDS-PAGE gel and electroblotted onto polyvinylidene difluoride membrane (Millipore). After blocking with 5% nonfat dry milk (Carnation), half of the membrane was incubated for 1 h with 10 g of crystal-derived CLC protein solubilized in 100 mM Tris-HCl buffer (pH 7.5), followed by extensive washing with TBS containing 0.2% Tween 20. The membrane half probed with CLC protein, along with the duplicate control membrane, was then probed with affinity-purified polyclonal rabbit IgG anti-CLC protein antibody, followed by horseradish peroxidase-conjugated goat anti-rabbit IgG (Santa Cruz Biotechnology), and the blots were developed using SuperSignal enhanced chemiluminescent substrate (Pierce). Both halves of the membrane were realigned and exposed to Kodak Biomax MR film. The control half of the membrane that had not been probed first with CLC protein was then reprobed with chicken anti-rat recombinant pancreatic LPLase antibody and horseradish peroxidase-conjugated donkey anti-chicken IgY (Jackson ImmunoResearch Laboratories).
Radioimmunoassay for CLC Protein-CLC protein levels in whole cell AML14.3D10 lysates, affinity column fractions, and preparations of affinity-purified CLC protein were determined by a highly sensitive double antibody radioimmunoassay that detects CLC protein in the low nanogram range (0.1-10 ng/ml) (29,36).
Reverse Transcription (RT)-PCR and DNA Sequencing-Total RNA, extracted from AML14.3D10 cells using TRIZOL reagent (Invitrogen) was used as the template. Reverse transcription was performed according to the instructions of the First-Strand cDNA Synthesis Kit from Amersham Biosciences using the oligonucleotide 5Ј-GGCAAAGTTG-GTCCAGTAGGC-3Ј as the primer, which is complementary to nucleotides 1475-1496 of the human carboxyl ester lipase cDNA (16), which is identical in sequence to human pancreatic LPLase (16). An aliquot of the reverse transcription reaction product was then used as the template for a first round PCR using Taq polymerase (Invitrogen), a forward primer (5Ј-GTCACCTTCAACTACCGTGTC-3Ј) complementary to human carboxyl ester lipase cDNA nucleotides 494 -515, and the above reverse primer to amplify a 1001-bp cDNA fragment. To ensure that the amplified 1001-bp fragment was derived from cDNA (mRNA) and not genomic DNA contaminating the AML14.3D10 RNA, the primers used for PCR were chosen to be complementary to regions crossing multiple intron-exon boundaries of the carboxyl ester lipase genomic DNA sequence located 5 kb apart (including six introns and seven exons) (37). The first round RT-PCR product was purified using a Qiaquick Gel Extraction Kit (Qiagen) and used as template for second round PCR using the same primers. Rat pancreatic LPLase cDNA (38), kindly provided by Drs. William Rutter and Jang Han of Chiron Corp., was used as a positive template control. The second round PCR product was sequenced in its entirety on both strands at the University of Illinois at Chicago DNA Core Facility using an ABI 373 DNA sequencer. The confirmed sequence was identical to the human carboxyl ester lipase/ pancreatic LPLase cDNA, nucleotides 565-1496 (16).
Crystallization of Complexes of CLC Protein with Sulfhydryl Groupreactive Inhibitors-Crystals of CLC protein were obtained as described previously (8). CLC protein in complex with NEM or pCMBS was obtained by soaking native CLC protein crystals in 50 mM NEM or 50 mM pCMBS for 3 days. Crystals of CLC protein complexed with the inhibitors were also obtained by co-crystallization using conditions for native crystal growth but with the addition of a 2 mM concentration of the inhibitor in the crystallization buffer.
X-ray Data Collection and Processing-Diffraction data were collected at room temperature using the Synchrotron Radiation Source at Daresbury on station PX9.6 ( ϭ 0.87 Å) for the CLC protein-NEM complex to 1.8-Å resolution and on station PX7.2 ( ϭ 1.488 Å) for the CLC protein-pCMBS complex to 2.3 Å. An additional CLC protein-pCMBS data set was collected to 1.8-Å resolution on station X31 ( ϭ 0.92 Å) at EMBL-Hamburg, Germany. Data for the CLC protein-NEM complex were also collected to 2.7-Å resolution in house on a 30-cm diameter MAR-Research image plate mounted on an Enraf-Nonius x-ray generator. Raw data images were indexed, integrated, and corrected for Lorentz and polarization effects using the program DENZO (39). All data were scaled and merged using the program SCALEPACK (39). Intensities were truncated to amplitudes by the TRUNCATE program (40).
Structural Refinement-The 1.8-Å resolution native CLC protein structure (Protein Data Bank code 1LCL (8)) was used as a starting model for the refinement of each complex. Alternating cycles of manual building, conventional positional refinement, and the simulated annealing method as implemented in X-PLOR (41) improved each model, while bulk-solvent correction as implemented in X-PLOR 3.851 allowed all measured data (from 20 to 1.8 Å) to be used in the refinement. The program O (42) was used to adjust the model to fit SigmaA-weighted 2 ͉F o ͉ Ϫ ͉F c ͉ calc electron density maps (43). During the final stages of refinement, water molecules were inserted into the models only if there were peaks in the ͉F o ͉ Ϫ ͉F c ͉ electron density maps with heights greater than 3 and they were at hydrogen bond-forming distances from appropriate atoms. 2 ͉F o ͉ Ϫ ͉F c ͉ calc maps were also used to verify the presence of these water molecules. In each case, the inhibitor molecule was included during the final stages of refinement. For pCMBS, the coordinates from the Protein Data Bank (code 1XZC (44)) were used, and for NEM, a manually built model using standard stereochemical rules was used. In the CLC protein-pCMBS complex, there was no density in the electron density map for the last two C-terminal residues, and they were excluded from the model. The program PROCHECK (45) was used to assess the quality of final structures. Analysis of the Ramachandran ( Ϫ ) plot showed that all residues were in the allowed regions.

RESULTS
CLC Protein-depleted Lysates of AML14.3D10 Eosinophils and Blood Eosinophils Retain Their Lysophospholipase Activity-To determine whether CLC protein contributes to any of the LPLase enzyme activity expressed by the eosinophil, either through expression of LPLase activity itself or as a cofactor for an alternative eosinophil LPLase, we sought to fully deplete CLC protein from blood eosinophil and AML14.3D10 whole cell sonicates and then assess the remaining enzyme activity relative to both the starting whole cell lysate and the isolated, purified CLC protein. To accomplish an absolute depletion of CLC protein from AML14.3D10 eosinophil lysates, we passed an amount of the cell lysate, containing CLC protein at ϳ50% or less of the maximum binding capacity, over an anti-CLC affinity column. This approach resulted in the complete depletion of CLC protein from AML14.3D10 eosinophil lysates as shown both by Western blotting (Fig. 1A) and by a sensitive double antibody radioimmunoassay for CLC protein (data not shown). Analysis of the column flow-through from the anti-CLC affinity column by SDS-PAGE and Western blotting showed that it no longer contained any detectable CLC protein (Fig. 1A). Likewise, analysis of the column flow-through fractions showed that CLC protein levels were below the detectable limits of the CLC protein radioimmunoassay, which is capable of detecting the protein in amounts as low as 1 ng/ml. To assess the effect of CLC depletion on the LPLase activity present in AML14.3D10 eosinophil lysates, the starting cell lysate, CLCdepleted cell lysate (column flow-through), and the CLC protein eluted from the antibody affinity column were assayed for LPLase activity (Table I). The CLC-depleted column flowthrough fraction retained 99.5% of the LPLase activity present in the starting cell lysate, with less than 0.5% of the starting LPLase activity associated with the CLC protein eluted from the column. Furthermore, both the starting AML14.3D10 eosinophil lysate and CLC-depleted lysate expressed LPLase activity that cleaved L-1-[1-14 C]palmitoyl lysophosphatidylcholine substrate at an average rate of 1.55 and 1.74 mol/h, respectively, compared with an activity of only 0.008 mol/h for the affinity-purified CLC protein (Table I).
Results with the AML14.3D10 eosinophil cell line were confirmed using peripheral blood eosinophils purified from normal donors (Fig. 1B). For blood eosinophils, the CLC-depleted column flow-through retained 91% of the LPLase activity present in the starting whole cell lysate, whereas the purified CLC protein from the same separation was essentially enzymatically inactive, expressing only 0.1% of the LPLase activity present in the whole eosinophil lysate (Table I). Of interest, blood eosinophils expressed ϳ10-fold greater LPLase activity on a per cell basis compared with the AML14.3D10 eosinophil cell line (Table I).
CLC-depleted AML14.3D10 Eosinophil Cell Lysates Contain Lysophospholipase Activity That Is Inhibitable by NEM and pCMBS-The LPLase activity present in eosinophils was previously shown to be inhibitable by sulfhydryl group-reactive agents such as NEM and pCMBS (15). In contrast, the LPLase activity of chromatographically purified CLC protein was FIG. 1. Western blot and lysophospholipase activity measurements of antibody affinity column depletion of CLC protein/ galectin-10 from AML14.3D10 and blood eosinophil whole cell lysates. The soluble fraction of whole cell sonicates of AML14.3D10 cells (A) or peripheral blood eosinophils (B), containing both LPLase activity and CLC protein (in an amount less than or equal to 50% of the column's maximum capacity for binding CLC protein), was fractionated on an anti-CLC protein antibody affinity column. Proteins were separated by SDS-PAGE and immunoblotted using affinity-purified IgG antibody to CLC protein (panels below each bar graph). Fractions were analyzed for LPLase activity and are plotted as a percentage of the activity in the whole cell lysates (100%). The fractions analyzed include the starting whole cell lysate, the pooled nonbinding flow-through fractions, and the CLC protein eluted from the column using 100 mM triethylamine, pH 11. Sample volumes for lysophospholipase assays and Western blotting were equilibrated to that of the starting whole cell lysate, and equal volumes were analyzed. Crystal-derived CLC protein was used as the protein standard. As shown for AML14.3D10 cells (A), the column flow-through showed no signal whatsoever (Western blot; top) and was entirely depleted of CLC protein, also as demonstrated by radioimmunoassay (Table I) shown to be resistant to inhibition by these reagents (15). We therefore sought to determine whether CLC protein-depleted AML14.3D10 eosinophil lysates contained LPLase activity that could be inhibited by NEM and pCMBS and whether affinitypurified CLC protein from these cell lysates retained any LPLase activity that could likewise be inhibited by these sulfhydryl group-reactive agents. The starting AML14.3D10 eosinophil lysate, the CLC-depleted lysate (flow-through fraction), and the eluted, antibody affinity-purified CLC protein fractions, prepared as described in Fig. 1 and Table I, were assayed for LPLase activity in the absence or presence of NEM ( Fig. 2A) or pCMBS (Fig. 2B). Lysophospholipase activity was evaluated and plotted as a percentage of the enzyme activity of the starting whole cell lysate in the absence of inhibitor for each concentration of the inhibitor tested (Fig. 2). The LPLase activity in both the starting whole cell lysate and the CLC-depleted cell lysate was inhibited in a dose-dependent fashion by both NEM and pCMBS, whereas the minimal residual LPLase activity associated with the affinity-purified CLC protein fraction (Table I) was not significantly inhibited by any concentration of NEM or pCMBS tested (Fig. 2).
Crystal Structure of CLC Protein-pCMBS and CLC Protein-NEM Complexes-The failure of pCMBS and NEM to inhibit the residual LPLase activity that remained associated with affinity-purified CLC protein prompted us to evaluate the binding of these agents to CLC protein at the structural level. We sought to determine 1) whether or not pCMBS and NEM bound to CLC protein via its two cysteine sulfhydryls and 2) whether the inhibitors altered the tertiary structure of CLC protein upon binding. Complexes of CLC protein with pCMBS or NEM were prepared by soaking native crystals of CLC protein in 50 mM pCMBS or NEM for 3 days. Crystals of CLC protein complexed with the inhibitors were also obtained by co-crystallization in the presence of a 2 mM concentration of the inhibitors. The CLC protein-pCMBS complex crystals were hexagonal and grew within 1 week at 16°C (space group P6 5 Tables II and III, respectively. As anticipated, pCMBS bound to the Cys 29 and Cys 57 sulfhydryls of CLC protein via its mercury atom (Fig. 3), with essentially no significant alterations to the tertiary structure of the protein (Fig.  3, Table IV). However the binding of pCMBS to Cys 57 caused a flip in conformation of the Arg 60 side chain, enabling a van der Waals interaction with the ligand. The position occupied by the Arg 60 side chain in the native unliganded structure is substituted by a water molecule (Fig. 3C). In contrast, the binding of pCMBS to Cys 29 causes displacement of two water molecules (Fig. 3B). Both pCMBS and NEM, when bound to Cys 29 , cause a 90°rotation to the C␤-S␥ bond of Cys 29 with respect to its position in the free structure. NEM was bound via its atom C2 to the Cys 29 sulfhydryl of CLC protein, an interaction that also involves the formation of a hydrogen bond between the O1 atom of NEM and Asp 85 (O␦1) of CLC protein (Fig. 4B, Table  IV). This interaction compensates for the loss of a water-mediated hydrogen bond in the unliganded native structure. In contrast to pCMBS, NEM was not found to be bound to Cys 57 of CLC protein but instead stacked against the tryptophan (Trp 72 ) side chain in the CRD via van der Waals interactions. This was further stabilized through hydrogen bond interaction of the C2 atom of NEM with His 53 N⑀2 and the O2 atom of NEM with Glu 33 (O⑀2) from a symmetry-related molecule. No significant conformational change was observed in the CRD of CLC protein upon NEM binding apart from the displacement of a network of four water molecules (Fig. 4C).

AML14.3D10 Eosinophils Express a Protein That Co-purifies in Small Quantities with CLC Protein on an Anti-CLC Affinity Chromatography Column and Is Recognized by Antibodies to
Recombinant Rat Pancreatic Lysophospholipase-Based on our finding that affinity-purified CLC protein retained less than 1% of the LPLase activity present in AML14.3D10 eosinophil lysates, we sought to determine whether the source of this residual enzyme activity could be due to the co-purification of a LPLase with CLC protein. Eosinophils have previously been reported to express an LPLase that is highly similar to human and rat pancreatic LPLase (17). SDS-PAGE analysis of multiple preparations of antibody affinity-purified CLC protein revealed an ϳ85-kDa protein in AML14.3D10 cells that variably co-purified with CLC protein in small amounts (Fig. 5A). To determine whether this minor ϳ85-kDa contaminant of the affinity-purified CLC protein was similar or identical to the previously described pancreatic-like LPLase of the eosinophil, and affinity-purified CLC protein The soluble fraction of whole cell sonicates of AML14.3D10 or peripheral blood eosinophils, containing both LPLase activity and CLC protein (in an amount equivalent to 50% or less of the column's maximum capacity for binding CLC protein), was fractionated on an anti-CLC protein antibody affinity column. The fractions analyzed included the starting whole cell lysate, the nonbinding, CLC-depleted flow-through fraction, and the CLC protein eluted from the column using 100 mM triethylamine, pH 11.5. we performed Western blot analysis on preparations of affinitypurified CLC protein using an antibody specific for recombinant rat pancreatic LPLase (Fig. 5B), which is Ͼ98% similar to human pancreatic LPLase at the amino acid level. The antibody to recombinant rat LPLase recognized an ϳ85-kDa protein in both the affinity-purified preparations of CLC protein and in whole cell sonicates of AML14.3D10 eosinophils; this protein had identical mobility to that of the pancreatic LPLase present in a rat zymogen granule extract used as the positive control and size marker for Western blot analysis (Fig. 5B).
Only a small percentage of the pancreatic LPLase in AML14.3D10 eosinophil lysates co-purifies with CLC protein on the immunoaffinity column; the majority elutes in the nonbinding flow-through fraction. Comparable results were obtained for antibody affinity column purified blood eosinophil CLC protein, in which an ϳ75-kDa minor contaminant copurified with the protein and was likewise recognized by antibody to recombinant rat pancreatic LPLase (results not shown). Preimmunization serum used as a negative control for Western analysis did not show any reactivity whatsoever with the same preparations of affinity-purified CLC protein, AML14.3D10 or eosinophil whole cell lysates, or rat zymogen granule extract (results not shown). AML14.3D10 Eosinophils Express an mRNA Encoding Human Pancreatic Lysophospholipase-Based on our detection of an ϳ85-kDa protein in AML14.3D10 eosinophil lysates that was recognized by antibody to recombinant rat pancreatic LPLase, and in lesser amounts in preparations of affinitypurified CLC protein, we sought to determine whether human eosinophils express mRNA encoding a LPLase similar or identical in sequence to that of pancreatic LPLase. Oligonucleotide primers, complementary to 21-bp sequences of both rat and human pancreatic LPLase, were utilized to assess the presence of mRNA encoding the pancreatic enzyme in total RNA preparations from AML14.3D10 eosinophilic myelocytes using RT-PCR. The primers were selected based on their identity to conserved coding sequences of both the rat and human pancreatic LPLase and crossed multiple intron-exon boundaries of the LPLase gene sequence located 5 kb apart (including six introns   FIG. 3. Structural details of CLC protein-pCMBS complex. A, ribbon representation of the CLC structure in complex with pCMBS bound to both Cys 29 and Cys 57 . B, details of pCMBS bound to Cys 29 . C, details of pCMBS bound to Cys 57 . The native CLC protein structure in the absence of pCMBS is shown in gray, while the complex structure is shown in standard element colors. Differences in water structure are observed in the liganded (green) and native unliganded structure (cyan). The mercury atom is purple. The dashed lines represent the covalent linkage between the mercury atom and sulfhydryl group atom of Cys 29 and Cys 57 . Amino acids are represented with one-letter codes.

FIG. 2. CLC-depleted AML14.3D10 eosinophil cell lysates contain lysophospholipase activity that is inhibitable by NEM and pCMBS.
The soluble fraction of whole cell sonicates of AML14.3D10 eosinophils, containing both LPLase activity and CLC protein, was fractionated on an anti-CLC column as described in Fig. 1. The starting whole cell lysate, CLC-depleted cell lysate, and purified CLC protein, all from the same affinity column, were assayed for LPLase activity in the absence or presence of the indicated concentrations of NEM (A) or pCMBS (B). LPLase activity is plotted as a percentage of the enzyme activity of the starting whole cell lysate in the absence of inhibitor for each concentration of the inhibitor tested. Results from two independent experiments are shown. and seven exons), precluding the amplification of the correct size 1-kb fragment from genomic DNA that might contaminate the RNA samples in small amounts. An appropriately sized 1-kb cDNA fragment, identical in size to that of the PCRamplified fragment obtained using a rat pancreatic LPLase cDNA template as positive control, was obtained following both a first round RT-PCR (not shown) and a second round PCR amplification (Fig. 6). To ensure that the 1-kb RT-PCR-amplified fragment corresponded to that of human pancreatic LPLase and did not represent contamination by the rat pancreatic LPLase control template utilized in these experiments, we sequenced the second round PCR-amplified fragment obtained from preparations of AML14.3D10 eosinophil mRNA. The nucleotide sequence of the 1-kb amplified cDNA fragment (Fig. 6, lane 4) was identical to that of human pancreatic LPLase, confirming the eosinophil expression of this lipolytic enzyme. The need for a second round PCR amplification to obtain a strong signal visible by ethidium bromide staining suggests that the pancreatic LPLase mRNA is of low abundance in the eosinophil, as found in a prior report (17).

CLC Protein Binds Lysophospholipases Expressed by Human and Murine Eosinophils and AML14.3D10 Cells: Analysis by
Ligand Blotting-To determine whether CLC protein has the capacity to bind LPLases expressed by blood eosinophils or AML14.3D10 eosinophilic myelocytes, whole cell lysates were analyzed by ligand blotting using crystal-derived native human eosinophil CLC protein as the probe. Although murine eosinophils do not express an orthologue of CLC protein, 2 they have been shown to express considerable LPLase enzymatic activity and were therefore included in these analyses as an additional source of eosinophil LPLases. Whole cell lysates were resolved on SDS-PAGE, electroblotted onto polyvinylidene difluoride membrane, and probed with 1) anti-CLC antibody alone (control blot) or 2) CLC protein followed by anti-CLC protein antibody (CLC ligand blot). The control blot was then reprobed with anti-rat pancreatic LPLase antibody to identify the eosinophil LPLase species (Fig. 7). The CLC protein bound to 75-and 2 S. J. Ackerman, J. Du, and J. J. Lee, unpublished observations.
where I i (h) and I(h) are the ith and the mean measurements of the intensity of reflection h. b R merge is equal to R sym , but it refers to merging data from different crystals and different sources.  55-kDa proteins in human and mouse blood eosinophils, respectively, and both proteins in AML14.3D10 cells (Fig. 7, lanes  4 -6), which are identical in size to the LPLases identified in these samples by the anti-recombinant rat pancreatic LPLase antibody (Fig. 7, lanes 7-9). A longer exposure of the CLC ligand blot (Fig. 7, lanes 4 -6, inset) showed that CLC protein also binds to the 55-kDa protein in human eosinophils. Murine eosinophils and AML14.3D10 cells also express an 85-kDa spe-cies recognized by the anti-LPLase antibody (Fig. 7, lanes 8 and  9, respectively), that weakly binds CLC protein in whole cell lysates (Fig. 7, lanes 5 and 6, inset). These findings indicate that CLC protein has the capacity to selectively interact with pancreatic-like LPLases present in both human and murine eosinophils and eosinophilic cell lines such as AML14.3D10. DISCUSSION CLC protein (galectin-10) accounts for an estimated 7-10% of total cellular protein in the mature blood eosinophil (3,15,29,46), and eosinophils express 3-8-fold more LPLase activity than other granulocytes (neutrophils) or mononuclear leuko-FIG. 5. AML14.3D10 eosinophils express an 85-kDa protein that co-purifies in small quantities with CLC protein on an anti-CLC affinity chromatography column and is recognized by antibodies to recombinant rat pancreatic lysophospholipase. A, CLC protein was purified from whole cell sonicates of AML14.3D10 eosinophils by antibody affinity chromatography. The eluted CLC protein was analyzed by SDS-PAGE, and the gel was stained with Coomassie Blue. The mobility of the CLC protein is indicated, along with the mobility of an ϳ85-kDa co-purifying protein that is comparable with the molecular mass of human pancreatic LPLase. B, Western blot analysis of the purified CLC protein sample shown in A using a chicken anti-recombinant rat pancreatic LPLase that is cross-reactive with human pancreatic LPLase. Rat pancreatic zymogen granule extract was included as an LPLase-positive control (lane 1), and the starting AML14.3D10 eosinophil whole cell sonicate was analyzed (lane 5) for comparison. Lanes 1 and 5 were loaded with 3.6 and 10.5 g of total zymogen granule extract or AML14.3D10 cell lysate protein, and lane 3 was loaded with 5.6 g of the affinity-purified CLC protein pool, respectively.  6. AML14.3D10 eosinophils express an mRNA encoding human pancreatic lysophospholipase. Total RNA was prepared from AML14.3D10 eosinophils, and a 1001-bp cDNA fragment was amplified by RT-PCR using forward and reverse primers identical in sequence to both rat and human pancreatic LPLase (16,17). Reverse transcription utilized a 3Ј to 5Ј reverse primer specific for rat/human pancreatic LPLase. The resulting cDNA was amplified using two 35cycle rounds of PCR. A 1001-bp fragment was weakly visible after the first round of PCR (not shown). This band was eluted from the agarose gel and used as template for a second round of PCR. The ethidium bromide-stained agarose gel shows the second round 1001-bp amplified PCR fragment (lane 4) in comparison with a rat pancreatic LPLase cDNA template (lane 2), the same template without Taq polymerase (lane 3), and 1-kb ladder size markers (lanes 1 and 5). 7. CLC protein binds lysophospholipases expressed by human and murine eosinophils and AML14.3D10 cells: Analysis by ligand blotting. Whole cell lysates of purified human and mouse blood eosinophils and AML14.3D10 eosinophils were analyzed by ligand blotting using crystal-derived native eosinophil CLC protein as the probe. Whole cell lysates were resolved on SDS-PAGE, electroblotted onto polyvinylidene difluoride membrane, and probed with anti-CLC antibody only (control blot, lanes 1-3) or with CLC protein as probe followed by anti-CLC protein antibody (ligand blot; lanes 4 -6). The control blot was reprobed with the anti-recombinant rat pancreatic LPLase antibody (lanes 7-9).
cytes (2,47). As well, CLC protein is the second most abundantly expressed gene at the mRNA level in both developing eosinophil progenitors from umbilical cord blood (48) and the mature eosinophil (49). However, the intracellular and/or extracellular role(s) of CLC protein as a LPLase or galectin in eosinophil biology remains speculative and unresolved (8,50,51). The extremely low specific activity of CLC protein compared with the enzymatic activities of other known mammalian and prokaryotic LPLases (50 -52) and the lack of significant sequence similarities to any other lipolytic enzymes (lipases, phospholipases, or LPLases) (5,9,10) have argued against a role for CLC protein as an eosinophil LPLase. These observations, coupled with the significant similarities of CLC protein to members of the galectin superfamily in terms of amino acid sequence (5), three-dimensional structure (8,14), and organization of the CLC gene (10), prompted us to reevaluate the reported LPLase activity of the protein (2-4, 8, 15, 30).
We reassessed the contribution of CLC protein/galectin-10 to eosinophil LPLase activity using antibody affinity chromatography to entirely deplete eosinophil lysates of the protein, a maneuver confirmed by highly sensitive and specific Western blotting (30) and radioimmunoassay (29). We followed this with the demonstration that CLC-depleted eosinophil lysates retained essentially 100% of their LPLase activity. In addition, the LPLase activity in CLC protein-depleted cell lysates retained its ability to be inhibited by previously reported sulfhydryl-reactive inhibitors of LPLases such as NEM and pCMBS (15). In contrast, the affinity-purified CLC protein from the identical separations did not have significant LPLase activity, and any residual low level enzyme activity associated with the protein was insensitive to inhibition by NEM or pCMBS. Weller et al. (15) also reported that the weak LPLase activity of both chromatographically purified eosinophil LPLase activity and crystal-derived CLC protein was minimally or not inhibited by sulfhydryl group-reactive agents such as NEM and pCMBS, agents that were highly effective in inhibiting enzyme activity in whole eosinophil sonicates. Our present analyses confirm and extend these observations, demonstrating that the residual LPLase activity that co-elutes with antibody affinitypurified CLC protein likewise cannot be inhibited by sulfhydryl group reactants such as pCMBS and NEM. To confirm that these inhibitors actually bind to CLC protein via the expected cysteine sulfhydryl interactions and to structurally characterize their binding mechanism, we determined the crystallographic structures for the CLC protein-inhibitor complexes. Our crystal structures showed that both pCMBS and NEM bound to Cys 29 but that only pCMBS bound to Cys 57 near the CLC CRD. In contrast, the interaction of NEM with residues of the CLC CRD was significantly different and of greater interest, since NEM was found to bind in a manner analogous to the binding of mannose in the CLC crystal, involving stacking interactions of the ring portion of NEM against Trp 72 (14), a highly conserved residue in the CRD of the galectin superfamily (53,54).
Our findings unambiguously demonstrate that CLC protein is not the source of eosinophil LPLase activity and, importantly, that galectin-10 is not a dual function member of the galectin superfamily, no members of which have been reported to express any LPLase or other lipolytic activities. Galectin-1 has been evaluated for LPLase activity by two groups using the standard radioenzyme activity assay utilized in the past and present studies, and neither group could detect any LPLase activity. 3 Furthermore, recombinant CLC protein expressed and purified from E. coli does not express significant LPLase activity. 4 These observations, coupled with our findings that 1) crystallization of CLC protein from sonicates of blood eosinophils or COS cells transfected with a CLC protein expression vector results in a loss of the apparent LPLase activity of the protein (30) and 2) homogeneous solutions of CLC protein resolubilized from CLC crystals are enzymatically inactive (8) further support the argument that CLC protein is neither an LPLase nor responsible for the prominent LPLase activity expressed by eosinophils. Our confirmation of pancreatic LPLase mRNA and protein expression by the eosinophil (17), coupled with our finding that this LPLase has the capacity to co-purify in small amounts with CLC protein under standard affinity chromatography conditions, provides a likely explanation for the mistaken identification of CLC protein as the eosinophil LPLase in the early 1980s (2)(3)(4)15).
In the original work by Weller and colleagues (3,4,15) in which eosinophil LPLase was first identified as CLC protein and characterized enzymatically, both the chromatographically purified protein and the protein resolubilized from Charcot-Leyden crystals were reported to express LPLase activities with K m and V max values of 23.7 M and 31.8 nmol/h/mg and 21.9 M and 46.8 nmol/h/mg of CLC protein, respectively (15). In our present work, affinity-purified CLC protein from both AML14.3D10 eosinophils and blood eosinophils expressed some residual LPLase activity (130 and 3.6 nmol/h/mg, respectively), comparable with the published K m values (15). These figures represent a minimally active LPLase compared with the activities of other known eukaryotic (mammalian) (17,52,(55)(56)(57)(58) and prokaryotic (59) LPLases. These enzymes typically have 1,000-fold greater specific activities, falling in the mol/h/mg range, suggesting that CLC protein would at best be a very poor LPLase (60).
One of the enzymes responsible for the expression of the considerable LPLase activity elaborated by eosinophils of humans, rodents, and other species is identical to the 75-kDa enzyme expressed by the exocrine pancreas (17). Holtsberg et al. (17), using an antibody generated against a human pancreatic LPLase peptide sequence that lacks any similarity to the CLC amino acid sequence, demonstrated that they could immunodeplete greater than 80% of the LPLase activity present in blood eosinophil whole cell lysates, again indicating that CLC protein is not a major contributor to the prominent LPLase activity expressed by these cells. Furthermore, it was recently suggested by Wang and Dennis (60) that contamination of 50 g of CLC protein with as little as 1 ng of the highly active 75-kDa pancreatic LPLase expressed by eosinophils would be sufficient to provide the activity levels originally reported by Weller and colleagues for CLC protein (15). Finally, human and murine eosinophils also express a 25-kDa LPLase, 5 first cloned from human brain but more widely distributed in terms of tissue expression (56), which likely accounts for the remainder of the considerable LPLase activity expressed by these granulocytes.
Although CLC protein is clearly not one of the eosinophil's LPLases, our results suggest that it may interact with eosinophil LPLases in vitro, perhaps due to an artifactual interaction that occurs during cell lysis procedures used to purify the enzyme activity (15) or CLC protein itself (8,30). Our ligand blotting results indicate that CLC protein does have the capacity to bind the pancreatic-like LPLase expressed by both human and murine eosinophils. Whether this binding is due to nonspecific hydrophobic interactions, specific protein-protein interactions, or carbohydrate binding activity of this unique galectin remains to be determined. Experiments to determine whether CLC protein and eosinophil LPLase(s) co-localize in resting or activated eosinophils are in progress using immunostaining and may provide more compelling evidence for a physiologically relevant interaction in vivo.