Truncated Forms of Mannose-binding Protein Multimerize and Bind to ...

THEJOURNAL OF BIOLOGICAL CHEMISTRY ( 0

Vol. 268,No. 1, Issue of January 5, pp. 364-370.1993 Printed in U. S.A.

1993 by The American Society for Biochemistry and Molecular Biology, Inc.

Truncated Formsof Mannose-binding Protein Multimerize andBind to Mannose-rich Salmonella montevideo but Fail to Activate Complement in Vitro* (Received for publication, May 29, 1992)

Jo E. SchweinleSj, Miyako NishiyasuS, Tie QiangDing*, Kedarnath Sastryll, StephenD. Gilliesll, and R. Alan B. Ezekowitzll From the $InfectiousDisease Section, Department of Internal Medicine, Yale University School of Medicine, New Haven, Connecticut 06510 and the TDivision of Hematology and Oncology, Children’s Hospital, Harvard Medical School, Boston, Massachusetts 021 15

Human serum (MBP) and human recombinant (rMBP) mannose-binding proteinbindto mannoseserum-resistant rich, Salmonella montevideo (SH5770), enhance C3 deposition, and render the organisms serum-sensitive. We investigated structural features of MBP necessary for this effect. MBP has a cysteine-rich amino-terminal region, a collagen-like region, and a carboxyl-terminal carbohydrate-recognition domain. We prepared carbohydrate-recognition domains lacking the other two domains either by deletion mutagenesis (AMBP, 16 kDa) or by collagenase digestion of whole rMBP (cdMBP, 16-18 kDa). Whole and truncated MBP were detected on immunoblot by specific monoclonal antibodies that recognizeboth bound and freeMBP. rMBP enhanced C3 deposition on SH5770 %fold, while cdMBP and AMBP did notincrease C 3 deposition over controllevels. All forms of MBP boundto SH5770 by enzyme-linked immunosorbent assayand by measuring binding of radiolabeledwhole and truncated MBP. Binding by’“I-AMBP was inhibitedby mannan and by MBP. Thus failure of truncated MBPs to enhance C3 deposition was not due to failure tospecifically bind carbohydrateresidues. To determine themolecular form of truncated MBP in nondenaturing conditions, ‘261-AMBP was centrifuged through a 5-20% sucrose-density gradient. The peak of lZ6I-AMBP sedimented to estimated Sz0+, 2.01, but larger multimers also were present. Multimers bound SH5770 with higher affinity than monomers. We conclude that carbohydrate-recognitionregions of MBP produced by collagenase digestionor by deletion mutagenesis are sufficientforligandbinding. However,thecollagen-likeregion is necessary for MBP enhancementof C3 deposition on SH5770.

Human mannose-binding protein (MBP)’ is an acute-phase reactantthat belongs totheC-typeanimal lectin family. Producedinthe liver andpresentinserum,it circulates primarily as large (200-600 kDa) polymers composed of 32kDamonomers (1-4).Each monomer can be divided into three regions: 1) the amino-terminal region, rich in cysteine residues that form disulfide bonds to stabilize the multimers ( 5 ) ,2) a nonfibrillar collagen-likeregion composed of repeated Gly-X-Y sequences similar in overall structure to Clq, pulmonary surfactant apolipoprotein, the macrophage scavenger receptor, and acetylcholinesterase (6); and 3) the carboxyl terminus that recognizes and binds to mannose and N-acetylglucosamine residues and shares sequence homology with other C-type lectins (6). Several studies (7-13) suggest that MBP is important in primary immunesurveillance for pathogens bearing mannose and/or N-acetylglucosamine residues on theirsurfaces. Kuhlman et al. (8) reported that MBP directly opsonizes Salmonella montevideo for ingestion and killing by polymorphonuclear leukocytes and monocytes independently of antibodies andcomplement.Using a mannose-rich organism, we (7) showed that MBP enhances alternative pathway C3 deposition on serum-resistantS. montevideo and renders itsensitive to serum bactericidal activity. Ikeda et al. (11)and Lu et al. (12) reported complement activation by MBP via the classical pathway. Each of these investigators used a different experimental approach to show complementactivation,perhaps explaining their differentconclusions. Although the role of MBP in complement activation is well established (7, 11-14), the mechanism of this effect and the structural requirements within MBPnecessary for activation are unknown. Furthermore, domains necessary for classical or alternative pathway activation are not defined. In this report, we characterize truncated forms of MBP missing the collagen-like tail and show that truncated MBP fails to enhance complement activation despite binding to mannoserich S. montevideo. MATERIALSANDMETHODS

* This work was supported by National Institutesof Health Grant A130286 (to J. E. S.) and by a National Institutes of Health grant and a grant-in-aid from the Bristol Myers Squibb Medical Institute (to R. A. B. E.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “aduertisement” inaccordance with 18 U.S.C. Section 1734 solely to indicate this fact. T o whom correspondence should be addressed: Infectious Disease Section, Dept. of Internal Medicine, Yale University School of Medicine, P. 0. Box 3333, LC1 810, 60 Davenport St., New Haven, CT 06510. Tel.: 203-785-7576; Fax: 203-785-3864.

Bacteria-S.montevideo SH5770(SH5770) was usedinall the experiments for several reasons. 1) The lipopolysaccharide is manThe abbreviations used are: MBP, human mannose-binding protein; rMBP, recombinant human MBP; AMBP, cloned truncated MBP;cdMBP,truncated collagenase-digested MBP; SH5770, S. montevideo serotype SH5770; HBSS, Hanks’ balanced salt solution; ADS, normal human serumadsorbed with SH5770; PCR, polymerase chain reaction; PAGE, polyacrylamide gel electrophoresis; PBS, phosphate-buffered saline; mAb, monoclonal antibody; ELISA, enzymelinked immunosorbent assay.

364

MBP Activation of Complement

365

nose-rich, each repeating 0-polysaccharide antigenic subunit bearing four mannose residues (15). 2) MBP binds to SH5770 (S), enhances C3 deposition on its surface, and renders the organism sensitive to serum bactericidal activity (7). 3) In the absence of MBP SH5770 is ollaprn 8 t a l k CRD resistant tokilling in nonimmune human serum, although it activates the alternative complement pathway (7, 16). Before eachexperimentSH5770 was grown to mid-logarithmic phase in tryptic soy broth, washed twice with Hanks' balanced salt solution containing0.15 mM CaCI, and 1mM MgCL (HBSS++), then resuspended a t a concentration of 2 X lo9 bacteria/ml (Am 2.00) in HBSS containing 20 mM CaC12 (HBSS/20 mM CaC1,) unless otherIgh SP S t a l k CRD wise stated. Serum-Fresh serum from normalvolunteers was immediately chilled to 0 "C and adsorbed twice at 0 "C with ice-cold, glutaraldehyde-fixed SH5770 (1 X 10" bacteria/ml serum) to remove naturally occurring antibodies (ADS).ADS stored at -70 "C maintained at least 80% of the starting hemolytic titer of C3. Some ADS was heated to 56 "C for 30 min to inactivate complement. Antibodies-Polyclonal antisera to MBP was raised in a rabbit and used without further fractionation. was It monospecificas determined by immunoelectrophoresis against normal human serum. It reacted with both monomeric and multimeric forms of whole and truncated M B P by immunoblot. Murine monoclonal antibodies were produced with specificity for free MBP only (mAb6), or for bound and free MBP (mAb3). Both monoclonal antibodies recognized monomeric and polymeric forms of whole and truncated MBP on immunoblot. FIG. 1. Mannose-binding protein constructs. [ l J ,MBP cDNA Control murine monoclonal antibodies were mAb M732 (gift of Miwas stably transfected into SP, cells, and the recombinant protein chael Wessels, Channing Laboratory, Beth Israel Hospital, Boston, harvested from the supernatant had very similar physical chemical Massachusetts) directed against GroupB type 111 streptococci. properties to serum MBP detailed in Ref. 14. [2J, a truncated MBP Complement C3"Functionally active human C3 was prepared from that encodesfor thestalkandcarbohydrate recognition domain fresh plasma by the procedure of Hammer et al.(17) with minor (CRD) was prepared as described under "Materials and Methods" modifications.C3 was radiolabeled withNalz5I usingIodo-Beads (Pierce Chemical Co.) according to the manufacturer's instructions. and stably transfected into SP2 cells. The resultant protein is characterized as shown in Figs. 2 and 6. 131, truncated MBP was derived Specific activity ranged from 1.7 X 10' to 2.2 X lo6 cpm/pg. Mannose-binding Proteins-Human MBP was purified from out- by collagenase digestion (19) as shown in Fig. 2. Igh SP, immuno(18). Recombinant MBP globulin heavy chain signal peptide. datedplasmaas describedpreviously (rMBP, Fig. 1, panel I ) was produced as reported earlier (a), except in this generation rMPB was made in mousemyelomacells (14) collagen-like tail was produced by digestion using methods described by Drickamer et al. (19). Briefly, 250 p1 of rMBP (300 pg/ml) were instead of Chinese hamster ovary cells. Truncated MBP (AMBP, Fig. 1, panel 2) lacking the collagen tail mixed with 250 p l of collagenase (43 pg/ml, 2.6 units/ml) from was made from the rMBP clone (Fig. 1, panel 1 ) by deletion muta- Clostridium histolyticunz (Calbiochem, Behring Diagnostics, La Jolla, genesis. AMBP includes the so-called stalk aswell as the18 invariant CA). Both enzyme and substrate were in 200 mM triethylammonium regions that are conserved in C-type lectins so that the whole con- buffer (200 mM triethylamine, 50 mM CaClZ, acetic acid to pH 7.8, struct effectively constitutes the carbohydrate recognition domain. containing 0.05% Triton X-100). The mixture was incubated with This construct was made by fusing two polymerase chain reaction agitation for 24 h at 37 "C.Nondigested and polymeric forms of (PCR) products, one encoding the immunoglobulin heavychain leader cdMBPalongwith collagenase(110 kDa, molecular mass) were peptide and the other encoding the carbohydrate recognition domain separated from monomericcdMBP (16- and 18-kDa products) by of humanMBP.ThefirstPCRproduct of 147 base pairs was centrifugation in a Centricon 30 (Amicon Division, W. R. Grace & generated using a CDMS oligonucleotide (5'-GTC GAC TGC TTA Co., Danvers,MA)concentrationsystem, repeatedly dilutingthe retentate with buffer to obtain as much cdMBP as possible in the ACT GGC TTA TCG AAA) and KS20 (5'-GAT ACA GCT TGG GAA T G T TG). The other product of 474 base pairs was generated filtrate. The filtrate was subsequently concentrated ina Centricon 10 using oligonucleotidesKS22(5'-AGC T G T ATC CTT TCC AGC (Amicon) concentration tube to 100 pg/ml. CCG GAT GGT GAT AGT AGC CT) and 48-llB (5'-TAT CTC The proteins were solubilized in sample buffer with or without 2GAG TCA GAT AGG GAA CTC ACA GAC GGC). The underlined mercaptoethanol by boiling for 8 min, then analyzed by electrophosequence in KS22 oligonucleotide corresponds to a region comple- resis on 12.5% SDS-polyacrylamide gels (SDS-PAGE). Insome cases mentary to that found in KS20. The template was CDMSvector the proteinswere stained withCoomassie Blue.In others, the proteins having a cDNA fragment with the entire coding region of human were immediately transblotted to nitrocellulose then probedwith MBP towhich XhoI linkers had been added andwhich had its signal various antibodies specific for MBP. peptide sequencereplaced withthatfrom immunoglobulinheavy Complement Actiuation-Washed bacteria (300 p l , 2 X lo9 orgachain leader. PCR was performed in a Perkin Elmer thermal cycler, nisms/ml in HBSS/2O mM CaCl2) were first mixed with 300 p1 of which consisted of 30 cycles of 94 "C for 15 s for denaturation; 55 "C HBSS/20 mM CaCl2 or 300 pl of HBSS/20 mM CaClz containing 20 for 15 s for annealing, and 72 "C for 15 s for extension followed by 10 pg/ml huMBP, rMBP, cdMBP, or AMBP in 1.5-ml microcentrifuge minutes final extension a t 72 "C. The PCR productswere purified by tubes, and incubateda t 25 "C for 1 h. The bacteria were then washed electrophoresis through low melting point agarose gel and isolation and resuspended in HBSS++ at 1 X lo9 bacteria/ml. Bacteria (250 of DNA by centrifuging the gel slice using SPIN-X tube (Costar, p l ) and 250 p1 of 5% serumwere mixed (final concentrationof bacteria Cambridge, MA) a t 4 "C. One microliter of each product was used to = 5 X 108/ml, final concentration serum = 2.5%) and incubated at set upa new PCR using theCDM8and48-llBprimersunder 37 "C tumbling end over end for 30 min. Triplicate 100-pI aliquots of conditions described above, except for 5 initial cycles with annealing each sample were added to 400 p1 of HBSS at 0 "C (ice-water bath) temperature of 45 "C. The final PCR productencoded 17 amino acids then centrifuged a t 12,500 X g for 5 min in a microcentrifuge (Micorresponding toheavy chain signal peptide plus 3 extra aminoacids: crospin 24S, Sorvall Instruments, Du Pont). The supernatant was Leu, Ser, Ser. The extra amino acids were to ensureefficient cleavage aspirated andsaved, the tipof the tube containing the bacterial pellet after the signal peptide because in a previous construct where amino was clipped, and eachwas counted separately in a CobraAuto-Gamma acid 81, proline, marked the NH, terminus of the truncated protein, radiation counter (Packard, Downers Grove, IL). Number of moleinefficient processing was observed (results not shown). AMBP as cules of C3 specifically bound per organism was calculated as dewell as whole MBP was radiolabeledby the same method as C3 scribed (20). iodination (see above) to 7.52 X IO3 cpm/pg and 1 X IO5 cpm/pg for Other complement activation experiments were performed to deuse in some experiments. termine if any molecular form of AMBP was able to activatecompleA second truncated MBP (cdMBP, Fig. 1, panel 3 ) lacking the ment. These studieswere executed exactly as described above except

& J

366

MBP Activation of Complement

that each pool (Fig. 6, pools A - D ) of AMBPwas recalcified and immediately used individually in place of MBP in the preincubation step of SH5770 prior to incubation in serum and 'z51-C3. M B P Binding toBacteria Detected by ELISA"SH5770 were boiled for 10 min to kill and heat-fix the organisms. After centrifugation, organisms were resuspended (2 X lO'/ml) in calcium carbonate buffer, pH 9.55 (15 mM NaC03, 30 mM NaHC03, 3 mM NaN,). Fifty microliters of the bacterial suspension was added to each well of a 96-well microtiter plate (1 X lo' organisms/well), and were allowed to adhere at 37 "C for a 1.5-h shaking. After washing the organisms with HBSS/ZO mM CaC12, 50 pl of MBP (10 pglml), cdMBP (5 pg/ ml), or AMBP (5 pglml) in HBSSI20 mM CaCI,were added and incubated 1 h a t 25 "C. The wells were washed with PBS-Tween (PBS containing 0.5 ml/liter polyoxyethylene-sorbitanmonolaurate (Tween 20, Sigma)), blocked with 5% bovine serum albumin at 4 "C overnight, then washed again with PBS-Tween. Increasing dilutions from 1:1,000 to 1:15,000 of mAb3 (shown by immunoblot to recognize cdMBP and by immunoblot and double immunodiffusion to recognize AMBP, data not shown) were added and incubated at 25 "C for 1 h. Wells were again washed with PBS, and horseradish peroxidasederivatized goat anti-murine IgG(1:250 dilution, Miles Scientific, Naperville, IL) was added and incubated at 25 "C for 1h. Wells were washed with PBS, and substrate o-phenyldiamine (Bio-Rad) was added and allowed to develop. The reaction was stopped with addition of 4%H2SO4. Plates were read on an Emax precision microplate reader (Molecular Devices Corp., Menlo Park, CA) at 450 nm. The controls for nonspecific binding ofmAb3 to SH5770 werewells containing buffer with no MBP. The controls for nonspecific antibody binding to MBP were wells containing SH5770 and MBP thatwere then incubated with murine monoclonal antibody M732 (M732) directed against Group B, type I11 streptococci. To express specific binding, optical density of the wells containing buffer without MBP were subtracted from sample wells. Optical densities of the wells probed with mAb M732 are graphically depicted in the results. Binding of Iz5I-MBP to SH5770"In other experiments, fractions of '"1-AMBP separated by sucrose-density sedimentation (see below) were combined to form poolsA, B, C, and D. One hundred microliters of each pool (0.093 pg of pool A, 0.669 pg of pool B, 1.526 pg of pool C, or 1.88 pg of pool D) were mixed with -1 X 10' bacteria in PBS/ 20 mM CaClz (final volume 300 p l ) , then incubated for 20 min at 25 "C rotating end over end. Samples were centrifuged at 12,500 X g for 10 min, and the supernatantsaved. Sides of the tubes were washed with 1.25 ml of buffer, second supernatant added to the first, and supernatants andpellets counted. Results were expressed as percentage of total '"1-MBP that bound to the bacterial pellet. In some experiments, 3.3 mg of mannan was added to bacteria simultaneously with each pool to compete with mannose residues on bacteria for specific binding. Mannan,rather than a chelating agent such as EDTA, was used to define background binding because EDTA disrupts Gram-negative bacterial cell walls (reviewed in Ref. 21). Lipopolysaccharides are released, potentially exposing cell wall constituents not normally available to interact with MBP. Binding to these constituents would result in artificially elevated levels of nonspecific binding. Alternatively, loss of the lipopolysaccharide might result in artificially low background binding. Separation of Molecular Forms of A M B P and MBP-Fifty micrograms of '"I-AMBP, 100 pg of nonlabeled AMBP, 14.8 pgof lZ5IrMBP, or 158 pgof rMBP were diluted in 5% sucrose in PBS containing 0.1% gelatin to a final volume of 500 pl, then layered over step gradients of 5-20% sucrose in the same buffer in Beckman UltraClear centrifuge tubes (25 X 89 mm,Beckman Instruments Inc., Palo Alto, CA). Tubes were loaded into a chilled Beckman SW 28 rotor, and centrifuged at 4 "C in aBeckman L8-70 Ultracentrifuge a t 27,500 rpm for 24 h. Seven-drop fractions were collected from the bottom of the tube. In some cases lZ6I-AMBPfractions were combined to make four pools (Fig. 6, pool A, fractions 20-40; B , fractions 40-60; C, fractions 60-80; and D,fractions 80-loo), and each pool individually subjected to a second sucrose density gradient centrifugation. Each centrifuged pool wasfractionated andanalyzed for radioactivity and/ or protein and for binding to SH5770 (see above). Individual fractions of "'I-rMBP were analyzed for polymeric state by nonreducing SDSPAGE on 5-20% polyacrylamide gradient gels that were dried and used for autoradiography. Protein molecular weight standards were sedimented at thesame time on identical sucrose gradients. A mixture 7.0), ofbovine thyroglobulin (Szo., 19.2), bovine y globulin (S~O.,,, 3.55), and myoglobin (Szo,w2.04), all from chicken ovalbumin (S20.w Bio-Rad, was diluted to a final volume of 500 pl in 5% sucrose/PBS/ gelatin (total protein 17 mg). Catalase (Sz0,, ll.O), /3-galactosidase

-

(S20,w 6.8), albumin (Sz0,,,, 4.73), and carbonic anhydrase (S%,,,,2.8) obtained from Sigma were mixed in 5% sucrose/PBS/gelatin (final volume 500 pl) prior to sedimentation (sedimentation constants are taken from Ref. 22). The standards were treated identically to the MBP samples, except at theend of the experiment, 50 p1 of each BioRad protein fraction was transferred to a microtiter plate well and read on an ELISA reader at absorbance 405 h to detect the myoglobin peak. Because the high sucrose content interfered with our quantitative protein assay, 50 pl of each protein standard fraction was analyzed by SDS-PAGE and stained with CoomassieBlue to ascertain which fraction contained the peak of each protein. Sedimentation constants of the protein standards were used to estimate the sedimentation constants of the MBPs. Inhibition of A M B P Binding toSH5770 by Mannan-To determine if AMBP was binding specifically to mannose residues on SH5770, 100 p1 SH5770 (1.5 X 10' bacteria) were mixed with each of two 100pl aliquots of each pool (A, B, C, or D) of radiolabeled AMBP. '"1AMBP in PBS/2O mMCaC1, (100 pl)was added simultaneously either with 100 p1 PBS/2O mM CaCl, or 100 pl PBS/2O mM CaC1, containing mannan(1mg/ml, final concentration 3.33 mg/ml, Sigma, final total volume 300 pl). The mixtures were then treated exactly as described under "Binding of '251-MBP." Competitive Inhibition of M B P and A M B P Binding to SH5770T o compare relative affinities of MBP and AMBP, we performed experiments in which rMBP was used to inhibit binding of lZ5IAMBP to SH5770. One hundred microliters of '"1-AMBP (4.9 pg in PBS/2O mM CaClZ)was added simultaneously with 100 p1 of increasing concentrations of nonradiolabeled rMBP (0-9.4pg in PBS/2O mM CaC1,) or AMBP (0-250 pg/ml in PBS/20 mM CaC12)to 1.5 X 10' SH5770 in 100 GI PBS/BO mM CaC12 (total mixture volume 300 pl). The mixture was then incubated at 37 "C for 60 min, washed and assayed inexperiments conducted as described in the radioassay above. Results are presented as percentage inhibition of '"1-AMBP binding. Reciprocal experiments of lZ5I-AMBPinhibition of rMBP binding also were performed. RESULTS

Truncated Forms of MBP-Non-treated rMBP(maintained at 4 "C), mock-digested rMBP (incubated at 37 "C in the absence of collagenase), AMBP, and exhaustively digested cdMBP (all in 50 mM CaCl,) wereanalyzed by SDS-polyacrylamide gel electrophoresis (Fig. 2). Under nonreducing conditions (Fig. 2 A ) , rMBP and mock-digested rMBP migrated primarily as doublets of m 82 and 88 kDa as well as two bands of larger multimeric forms and much larger polymers that did not enter the gel (not visible on the gel in Fig. 2, since the stacking gel has been removed). AMBP migrated as 16 kDa, and cdMBP as a doublet of 16 and 18 kDa. Upon reduction with 2% 2-mercaptoethanol (Fig. 2B), rMBP and mock-digested MBP migrated as 45 kDa consistent with incompletely reduced trimers. Protein bands at 45 kDa and in some cases at 66 kDa were frequently seen on analysis by SDS-PAGE and by immunoblot of MBP or rMBP reduced with 2-mercaptoethanol (not shown). Complete reduction of full-length MBP may require treatment with 6 M guanidine hydrochloride in 0.1 M Tris, pH 8.5, containing 70 mM 2-mercaptoethanol at 37 "C for 4 h (19) or boiling in 50 mM dithiothreitol for 5 min (14). In contrast, the majority of AMBP migrated as 25 kDa and cdMBP as a broad band from 25 to 27 kDa when electrophoresed under reducing conditions. Both truncated MBPs and whole rMBP bound monoclonal and polyclonal antibodies as determined by ELISA (see Fig. 4) and immunoblot (data not shown). These results show that truncated forms of MBP produced by collagenase digestion of rMBP or by deletion mutagenesis of the recombinant MBP clone were of similar molecular mass and did not form disulfide-linked multimers. C3 Binding to Bacteria Preincubated in MBP-We studied binding of C3 to SH5770 in the presence of MBP, AMBP, or cdMBP (Fig. 3). Organisms were incubated with MBP, rMBP, AMBP, cdMBP, or buffer, then incubated in 2.5% ADS. After

M B P Activation of Complement

367

of equimolar amounts of whole MBP by ELISA. All three forms of MBP, whole rMBP, AMBP, and cdMBP,specifically bound (Fig. 4) toSH5770 as detectedby mAb3 that recognized 200 bound and free MBP. Low nonspecific binding of negative 97.4control mAb M732 is also shown inFig. 4. Subsequent binding 68assays with "'I-AMBP (see below) confirmed that truncated MBP binds toSH5770. Thus failure to enhance complement 43 activation was not due to failure of truncated MBP to bind 29to the targetorganism. 184Molecular Form of AMBP-The inability of bound AMBP to activate complementmay reflect either the requirementof A NOT REDUCED B REDUCED FIG.2. SDS-polyacrylamidegel electrophoresis of different the collagen-like tail for complement activation or for multiforms of MBP. For 24 h prior to electrophoresis, rMBPwas either merization of the molecule. Therefore, we examined the momaintained at 4 "C, incubated a t 37 "C, or incubated a t 37 "C in the lecular form of AMBP, and the relationship of form to bacexpresence of 10 pg/ml collagenase. Gelatin (150 p g ) was incubated in terial binding and complement activation. These initial the presence or absence of 10pg/mlcollagenase as a control for periments were designed to determine whether AMBP existed collagenase activity. AMBP was made as described under "Materials as monomers, polymers or both under nondenaturing condiand Methods." Each was solubilized in sample buffer and electropho- tions. We also wanted to compare the sedimentation of whole resed on a 12.5% polyacrylamide gel under nonreducing (panel A ) and reducing (panel R ) conditions. Panel A, in the absence of colla- MBP and truncated MBPin nondenaturing conditions. In order to compare the size andsedimentationrate of genase, rMBP migrated as a doublet a t 82 and 88 kDa plus large nondenatured molecules, '2sI-AMBP or ""I-rMBP were cenmultimers, with (-/+) or without (-/-) incubation. Incubation in the presence of collagenase (+/+) produced a doublet a t 16 and 18 trifuged through 5-20% sucrose-density step-gradients in the kDa). AMBP migrated a t 16 kDa, exactly opposite the smaller diges- absence of calcium. The peak of "'I-rMBP sedimented at a tion product. Gelatin (-/+) migrated as two easily seen high molec- rate of approximately S20,ar 2.75 (Fig. 5), consistent with the ular weight bands and a long smear of protein not clearly apparent known molecular mass of monomeric MBP of 28-32 kDa (1, in this photograph. Digested (+/+) gelatin was not visible, confirming 12). The slope of the ascending limb of rMBP implied that activity of the collagenase. Panel R, under reducing conditions, (-/ -)rMBP and (-/+)rMBP migrated a t 45 kDa, probably representing there were multimeric forms, and their presence was verified incompletely reduced trimers.AMBP migrated a t 25 kDa, a t the by SDS-PAGE and autoradiography (not shown). In fractions lower end of the 25-27-kDa broad bandof (+/+)rMBP in the adjacent 43-47, containing low concentrations of sucrose, there were lane. Gelatin lanes again confirm collagenase activity. 29-32-kDa monomers. In addition, there were bands consistent with trimers andlarger complexes that barely entered the 10000 8000

5c m

?

0

6000

0

3 0

4000 I 0

I n e

0

0

n

2000

0

cdMBP

0 .\MBP

hlBP

Bullcr

FIG.3. Specific C 3 binding to SH5770. SH5770 were preincubated in buffer, cdMBP, AMBP, or rMBP, then in 2.5% ADS, exactly asdescribed under "Materials andMethods." MBP-sensitized SH5770 bound more than 3.8% (9,876 molecules C3/organism) total C3 in the assay, while cdMBP (668 molecules C3/organism)and AMRP (705 molecules C3/organism) bound no more C3 than the control organism preincubated in buffer (1,237 molecules C3/organism). This experiment is representative of four separate experiments performed in triplicate.

exposuretoMBPorrMBP(notshown), 8-foldmoreC3 (9,876 molecules C3/organism) boundtoSH5770thanin control samples (1,237 molecules C3/organism). By contrast, cdMBP (668 molecules CS/organism) and AMBP (705 molecules C3/organism) did not increase C3 deposition over control levels. Thus loss of the collagen-like tail from the carbohydrate recognition region removed the ability of MBP to enhance complement activation. M B P Binding to SH5770-To examine thepossibility that failure of AMBP to activate complement on SH5770 was due to theinability of MBP tobind in the absenceof the collagenlike tail, we compared binding of truncated MBP to binding

20

1 0.0

\\ 5.0

10.0

15.0

Dilution of Anti-YEP ( x 10E3)

FIG.4. Binding of various forms of MBP to SH5770.MBP b i n d i e t o SH5770 wasevaluated by ELISAexactly as described under"Materialsand Methods." rMBP, AMBP, orcdMBP were incubated in wells of microtiterplates precoatedwithheat-killed SH5770. Subsequentincubation in decreasing concentrations of monoclonal anti-MBP detected bound MBP regardless of the form present (solid lines). MRP-containing wells probed with mAb M732 (dashed lines) are shown as negative controls. Specific binding was derived by subtracting optical densities of wells with no MBP probed with mAh3 from optical densities of wells containing MBP. Optical density of these second control wells fell within the outer limits of the results seen graphed with nonrelevant mAb "732. Positive controls were wells containing SH5770 and no MBP that were probed with specific anti-SH5770 antisera (not shown). Thisresult is representative of three separate experiments. These results were duplicated with mAb3 derivatized with horseradish peroxidase and with rabbit polyclonal anti-MBP antibody.

MBP Activation of Complement

368 s 11.0

7.0

20.w

4.73

2.8 2.04

0 0

40

20

60

80

100

Fractions

2oool 1000

4.. . . . . . . . . . . . . . 0

I..

10

. I . .

40 20

. . . . .. . , . . . . . . . . . , . ... . . . . ., . . . . .. . I 30 Fractions

50

FIG. 5. Molecular form of rMBP by sucrose-density gradient centrifugation. lz5I-rMBPwas layered over a step gradient from 5 to 20% sucrose, then centrifuged a t 28,000 rpm in a Beckman L8-70 ultracentrifuge a t 4 "C for 24 h. Seven-drop fractions were collected and evaluatedfor radioactivity. The peak of '251-MBP radioactivity sedimented at a rate of S20.w 2.75. Inset contains a plot of the log ofSzO,uI uersus percentage of gradient traversed by molecular weight standards.

0 20

0

40

60

80

io0

Frec11ons

gel (pentamers or larger). Thisobservation suggeststhat upon standing at4 "C after separationfrom larger multimers in the gradient, monomeric rMBP associated to form polymers. In contrast to rMBP, the AMBP peak sedimented at a rate approximately Sz0,,,,2.0 (Fig. 6 A ) , consistent with the molecular mass of monomeric AMBP of 16 kDa (estimatedby SDSPAGE). The sloping ascending limb of the AMBP peak also suggested that multimeric AMBP complexes also were present. This was tested by preparing pools A through D as indicated in Fig. 6A, and resedimenting pools individually on 5-20% sucrose gradients (Fig. 6B). Pool A sedimented at 9.9 Sz0,,,,,pool B at 5.4 Szo,ru, pool C at 3.02 Sz0,,,,,and pool D at 2.01 Szo,w. In each case, the peak of the pool sedimented at approximately the mid-pointof the fractions itwas composed of, confirming that oligomers as well as monomers of AMBP were present in the sample. We therefore asked whether all forms of AMBP would bind to SH5770. Relationship of Molecular Form of M B P to Bacterial Binding-Binding of monomers and multimersof AMBP to mannose-rich bacteria was tested (Fig. 7). After incubation with SH5770, bacteria-associated 'T-AMBP from pooled fractions A, B, C, or D was measured and expressed as percentage of total lz5I-AMBPin that pool. In some experiments, mannan was added simultaneously with AMBP to SH5770. SpecificallyboundlZ5I-AMBP was calculated by subtracting percentage '251-AMBPbound in the presence of mannan from the percent bound in the absence of mannan. Pool A specifically bound 18%of lZ5I-AMBP,pool B 11.3%, pool C 3.5%, and pool D 3.7%. Theseresults show that AMBP binds specifically to mannose residues on SH5770 and also suggest that polymers of AMBP bearing multiple carbohydrate-recognition domains bindto the mannose-rich bacterial surfaces with higher affinity than the monomeric forms. Pentameric/hexameric forms of rMBP, but not smaller of the multimers, may substitute for Clq to initiate activation classical complement pathway (11-13). Removal of the collagen-like tail of Clq by collagenase digestion eliminates its ability to initiate complement activation (23) but does not remove the ability of truncated Clq to multimerize and to bind to antigen-antibodycomplexes (24). Likewise, a mixture

FIG. 6. Sucrose density gradient centrifugation of "'1AMBP. Panel A , lZ5I-AMBPwas separated according to density as described in Fig. 5. The peak of lZ5I-AMBP radioactivity sedimented a t a rate of Szo,w2.0. Panel B, pools A, B, C, and D prepared from fractions from initial sedimentationof '"I-AMBP (panel A ) , were resedimented individually on 5-20% sucrose density gradients. Pool A (inset) sedimented at a rate of 9.9 S ~ Opool , ~ , B a t 5.4 Szo,w, pool C a t 3.02 S20,m,and pool D at 2.01 S20,w. Pool A, squares with open centers; pool B, diamonds with open centers; pool C, solid squares; pool D, solid triangles. Protein standards inpanel A are plotted exactly as in Fig. 5. To maintain clarity, only representative protein standards are designated on the graph. 20

A

B

C

D

Pools of 1251-AMBP from Sucrose Density Gradient

FIG. 7. Binding of pools A, B, C, or D to SH5770.SH5770 (1.5 X 108bacteria) were incubated tumbling at 37 "C for 20 min with pool A, B, C, or D. Bacteria with bound '"I-AMBP were separated from unbound "'1-AMBP by centrifugation. Pellet and supernatant radioactivity were counted separately and percent lZ5I-AMBP associated with the bacterial pellet calculated. In some experiments, 3.3 mg of mannan was added simultaneously with each pool. Specific binding was defined as the difference between total binding and binding in the presence of mannan. Results depicted are representative of 5 separate experiments.

of truncated MBP of all multimeric states was unable to support complement activation despite binding to mannoserich substrates. In order to determine whether multimeric AMBP supported

MBP Activation of Complement

369

teins. Our experiments were designed to distinguish between these possibilities. Multiple lines of evidence suggest that it is the collagenous tail per se that is necessary for complement activation. First, both AMBP and cdMBP bound to SH5770, as assessed by ELISA (Fig. 4) and by direct binding of radiolabeled proteins (Fig. 7). Hence the absence of the collagen-like tail does not prevent binding of the truncated forms to the bacteria. Second, truncated MBP retains the epitopes for monoclonal antibody recognition present in the full-length molecule (Fig.4). Finally, multimeric forms of AMBP bind to SH5770 (Fig. 7) yet fail to activate complement. One inference that can be drawn from these results is that thecollagen-like tail is required to assemble the conformation that enables complement activation. Although it is clear that pentamers or hexamers are essential for classical pathway activation, the oligomerization state allowing alternative pathway activation has not been elucidated. Similar functional loss has been reported for classical pathway activation by Clq after collagenase digestion (23) despite retention of binding by Clq globular heads to antigen-antibody complexes(24). Collagenase-digestedClq, like truncated MBP,tends to form multimers (24). Analogous scenarios might explain loss of classical pathway activity by truncated MBP, but leaves the mechanism of alternative pathway activation unclear. Experiments are underway in our laboratory to define the mechanism of alternative pathway activation. Lu et al. (12) demonstrated that in the absence of Clq, binding of the larger multimers of MBP (pentamer/hexamer forms) to zymosan more efficiently activated complement via the classical pathway. In this case, the tertiarystructural similarity of polymeric MBP to Clqmay have allowed MBP to interact with ClrzClszin such a way that upon binding to DISCUSSION zymosan the conformational alterations necessary for Clr Whole human MBP, whether recombinant or purified from then Cls to become activated were satisfied. These earlier serum, binds to mannose-rich Salmonellae (8) and activates studies did not clarify whether multiple ligand-binding sites the alternative complement pathway (7), a characteristiccor- were sufficient or if the collagen-like region also was essential roborated by the present investigations. MBP also activates for complement activation to proceed. Our experiments demcomplement on mannan-coated sheep erythrocytes (11) and onstrated that the collagenous tail is necessary for compleon zymosan (12) via the classical pathway. The present studies ment activation, regardless of the multimeric state of the show that the collagen-like region, or at least some part of it molecule. Sucrose density sedimentation clearly established that is absolutely required for MBP to activate complement. The collagen-like region of MBP may berequired either for truncated MBP exists as multimers even in the absence of direct interaction with complement proteins or for inducing the collagen-like region, consistent with earlier work (25). a necessary conformation or multimerization within the pro- Predictably, larger multimers with presumably more ligand-

complement activation while monomers or smaller oligomers could not, each recalcifiedpool (Fig. 6) from the sucrose density gradient separation of AMBP wastested individually. None of the pools supported deposition of C3 on SH5770, the mannose-rich test organism (data not shown). Competitive Inhibition of 1251-MBPBinding by MBP or AMBP-We performed studies to determine the relative affinities of lZ5I-AMBPand whole rMBP for the mannose-rich organism. As increasing amounts of rMBP were added to SH5770 simultaneously with aconstantamount of lZ5IAMBP, rMBP inhibited lZ5I-AMBPbinding in a dose-response fashion at all concentrations tested (Fig. 8, panel A ) . Although most of the rMBP used in this assay existed as a monomer (as shown by sucrose density sedimentation, Fig. 5 , and autoradiography, not shown), multimeric forms also were present. Nonetheless, assuming that AMBP and rMBP were each present primarily as monomers, 0.0761 mol of monomeric rMBP was needed to inhibit 50% of the binding of 1 mol of '251-AMBP,demonstrating the greater affinity of whole MBP for the mannose-rich surface. When non-radiolabeled AMBP was used to compete with lZ5I-AMBP,binding to SH5770 was inhibited in a dose-related manner (Fig. 8, panel B ) . Binding of 5 pg of lZ5I-AMBPwas reduced by 50% in the presence of 5 pg of AMBP. Reciprocal inhibition experiments (AMBP inhibition of lZ5I-rMBPbinding, datanot shown) did notdemonstrate AMBP interference with rMBP binding at theconcentrations of AMBP used. A maximum of 7.3 mol of AMBP was added for each mol of lZ5I-rMBPused in the assay. Concentrations ofAMBPwere never sufficient to inhibit the more avidly bound lZ5I-rMBP.

1

YEP Added (pg)

AMBP Added (pg)

FIG. 8. Inhibition of Iz6I-AMBP binding to SH5770. SH5770 (1.5 X 10') were incubated with 4.9 figlZ51-AMBP and increasing amounts of rMBP (0-9.4 pg) or AMBP (0-250 pg) and treated exactly as described in Fig. 7. Results are expressed as percent inhibition of T - A M B P binding. lZ5I-AMBPbinding was inhibited in adose-response fashion by both rMBP (panel A ) and AMBP (panel B ) . The binding of 1 mol of lZ5I-AMBPcould be diminished by 50% with 0.0761 mol of rMBP (panel A , dashed line), while a molar ratio of 1:l (AMBP:'"IAMBP) diminished binding by 50% (panel B, dashed line). In additional studies (not shown) inhibition of "'1-rMBP binding by AMBP was not observed at concentrations tested. However, limited availability of AMBP prevented testing of inhibition with amounts of AMBP sufficient to demonstrate displacement of the high affinity rMBP.

370

MBP Activation of Complement

binding sites, bound mannose-rich Salmonellae with greater cubation, an aliquotof each sample was electrophoresed, and affinity than smaller multimers and monomers, but higher no difference was detected in multimerization stateof rMBP inthepresence or absence of calcium asdetermined by affinity did not influence the ability of truncated MBP to activate complement. In addition, the overlap in sedimenta- migration on SDS-PAGE. However, after 24 h of incubation, tion rates of MBP and AMBP suggests that smaller AMBP the migration of rMBP confirmed that incubation in the exists as multimers with at least as many binding sites as a absence of calcium resulted in separationof the multimersof substantial portionof whole MBP. Thusif the ligand-binding rMBP, causing it to migrate primarily as the 32-kDa monregion were sufficient for complementactivation, at least omer, while in the presence of calcium, it migrated primarily as a doublet at -82 and 88 kDa, plus several much larger some C3 deposition would have been observed. the The collagen-like region also contributes to the stabilityof bands. The loss of multimericstate occurredevenin the MBP-ligandinteraction.TruncatedMBP specifically absence of dilution in thesucrose gradient. In conclusion, our results indicate a requirement for the binds to mannose glycoconjugates and can be successfully competed from its target mannose residues by mannan, just MBP collagen-like region for complement activation. Even without thestabilizing effect of the collagen-like region, globas mannanpreventsinteraction of whole MBP with the mannose-rich targetsurface. However, avidity of AMBP bind- ular carbohydrate-recognition"heads" exist as monomers and ing is less than whole MBP since whole MBP successfully variously sized, noncovalentlylinked multimers that maintain competes withAMBPfor mannose residues, while AMBP their ability to bind specifically to mannose-rich targets. does not inhibit binding of whole MBP at concentrations REFERENCES tested (data not shown). The explanation for this observation 1. Kawasaki, M., Kawasaki, T., and Yamashina,I. (1983) J. Biochem. (Tokyo) 94,937-947 may be that the collagen-like region stabilizes polymerized 2. Wild, J., Robinson, D., and Winchester, B. (1983) Biochem. J. 2 1 0 , 167monomers in the optimal conformationfor ligand binding. 174 3. Summerfield, J. A,, and Taylor, M. E. (1986) Biochim. Biophys. Acta 8 8 3 , The collagen-like tail of human rMBP is cleaved by colla197-206 genase to produce inactivecarbohydrate-recognitionfrag4. Ezekowitz. R. A. B.. Dav. - , L. E.. and Herman. G. A. (1988) J. E m . Med. 1 6 7 , 1034-1046 ments of the carboxyl terminus. Collagenase preferentially 5. Sastry, K., Herman, G. A,, Day, L., Deignan, E., Bruns, G., Morton,C. C., cleaves polypeptides after proline in the amino acid sequence and Ezekowitz, R. A. B. (1989) J. Exp. Med. 170,1175-1189 Drickamer, K. (1988) J. Bid. Chem. 263,9557-9560 Gly-X-Pro-Gly (26). Human MBP has six such repeats in its 6. 7. Schweinle, J. E., Ezekowitz, R. A. B., Tenner, A. J., Kuhlman, M., and Joiner, K. A., (1989) J. Clin. Inuest. 84,1821-1829 collagen tail (4,5 ) and is susceptible to cleavage. Despite 8. Kuhlman, M., Joiner, K., andEzekowitz, R. A. B. (1989) J. Exp. Med. 1 6 9 , exhaustive digestion, cdMBP remained a doublet of 16 and 1733-1745 9. SuDer. M. S.. Thiel. S.. Lu. J., Levinsky, R. J., and Turner, M. W. (1989) 18 kDa underdenaturingconditions of SDS-PAGE. It is Lancet 2 , i236-1239 . unlikely that one of these two bands represents incomplete 10. Sumiya, M., Super, M., Tahona, P., Levinsky, R. J., Arai, T., Turner, M. S., and Summerfield, J. A. (1991) Lancet 337,1569-1570 digestion since under the same conditions collagen-like reIkeda. K.. Sannoh. T.. Kawasaki. N.. Kawasaki., T.., andYamashina. I. gions of ratMBP(19)andClq (23, 24) are completely 11. (1987) J. Biol. Chem: 2 6 2 , 7451-7454 Lu, J., Thiel, S., Wiedemann, H., Timpl, R., and Reid, K. B. M. (1990) J. 12. removed. Produced in mouse myeloma cells, rMBP may be Immunol. 144,2287-2294 glycosylated to varying extents, producing two sizes (82 and 13. Ohta, M., Okada, M., Yamashina, I., andKawasaki, T. (1990) J. Biol. Chem. 2 6 5 , 1980-1984 88 kDa) of rMBP on SDS-PAGE (see doublets of nondigested 14. Super, M., Gillies, S. D., Foley, S., Sastry, K., Schweinle, J. E., Silverman, rMBP in Fig. 2). The cdMBP doublet undernonreducing V. A., and Ezekowitz, R. A. B. (1992) Nature Genet. 2 , 50-55 15. Fuller, N. A,, and Stauh,A. M. (1968) Eur. J. Biochem. 4, 286-300 conditions probably represents digestion of two rMBPs of 16. Liang-Takasaki, C.-J., Saxen, H., Makela, P. H., and Leive, L. (1983) different glycosylation states. Less likely, carbohydrate resiInfect. Immun. 41,563-569 17. Hammer, C. H., Wirtz, G. H., Renfer,L., Gresham, H. D., and Tack, B. F. dues may be lost from a portion of the undigested carboxyl (1981) J. Biol. Chem. 256,3995-4006 terminus during theenzymatic degradation. 18. Kyogashima, M., Krivan, H. C., Schweinle, J. E., Ginsburg, V., and Holt, G. D. (1990) Arch. Blochem. Bzophys. 283,217-222 The apparent inconsistency in form (multimer versus mon19. Drickamer, K., Dordal, M. S., and Reynolds, L. (1986) J. Biol. Chem. 2 6 1 , 6878-6887 omer) of rMBP present on SDS-PAGE and in sucrosedensity N. K., Joiner, K. A., Frank, M. M., and Lieve, L. (1986) J. gradient was caused by differences in experimental conditions. 20. Grossman, Immunol. 1 3 6 , 2208-2215 Hancock, R. E. W. (1984) Annu. Reu. Microbiol. 3 8 , 237-264 21. Our hypothesiswas that presence or absence of calcium was 22. Edsall, J. T. (1953) in The Proteins: Chemistry, Biological Activity and responsible for the apparent differences in multimerization Methods (Neurath, H., and Bailey, K.. eds) Vol. l A , ~. pp. 550-726, Academic Press, New York state of rMBP on SDS-PAGE and in the sucrose-density 23. Knobel, H. R., Heysser, C., Rodrick, C., and Isliker, H.(1974) J. Immunol. gradient. We confirmed the hypothesis by incubating rMBP 1 1 2 , 2094-2101 K. B. M., Lowe, D. M., and Porter, R. R. (1972) Biochem. J. 1 3 0 , in calcium-free PBS at 4 "C, simulatingthe calcium-free 24. Reid, 7A9-763 . .- ." environment of the sucrose gradient experiments,or in trieth- 25. Weis, W. I., Crichlow, G. V.,,Murphy, H. M. K., Hendrickson, W. A,, and Drickamer, K. (1991) J . B ~ o l Chem. . 2 6 6 , 20678-20686 ylammonium buffer containing 50 mM CaCh at 37 " c , simu- 26. Seifter, S., and Harper, E. (1971) in The Enzymes (P. D. Boyer, ed) 3rd Ed., Vol. 3, pp. 649-697, Academic Press, Orlando, FL lating the collagenase-digestion experiment. After a 1-h inI

'