THE JOURNAL OF BIOLOGICAL CHEMISTRY Q 1993 by The American Society for Biochemistry and Molecular Biology, Inc.

Vol. 268, No. 8, Issue of March 15, pp. 5866-5871, 1993 Printed in U.S.A.

Identification of a Major Binding Site for Complement C3 on the IgG1 Heavy Chain*

(Received for publication, August 25, 1992)

Jason M. ShohetSO1, Philip Pembertonll , and Michael C. Carroll$** From the §Department of Pathology, Harvard Medical School, the $.Department of Pathology, Boston University Medical School, and the I( Division of Immunology, Children's Hospital, Boston, Massachusetts 021 15

Activation of the alternative pathway of complement by immune complexes involves the covalent attach- ment of the third component (C3) to the IgG heavy chain. In order to localize the sitelsites of attachment, adducts of human C3-IgG were digested in situ with endoproteinase Lys-C and Staphylococcus aureus V8 protease, and the fragments were analyzed. The di- meric peptide containing the covalent bond, identified by alkylation of the free thiol group (CYS'~'~) with i~do['~C]acetamide, was isolated by high-performance liquid chromatography fractionation. A double se- quence with NHa termini corresponding to position 134 of IgG heavy chain and position 1002 of the C3a' chain was found by analysis with automated Edman degra- dation. The intact dimeric peptide had a mass of 3453 Da and was composed of IgG and C3 fragments with predicted sizes of 23 and 12 residues, respectively. The IgG peptide includes a cluster of six potential acceptor sites for ester bond formation. Thus, it appears that C3 binding is limited to a single region within the CH1 domain of the IgG, heavy chain.

The fixation of complement to immune complexes and a variety of other pathogens by the classical and alternative pathways mediates both the afferent and efferent phases of the immune response (1). Fragments of C3 covalently bound to immunogenic targets provide ligands for complement receptors, i.e. CR1, CR2, CR3, and CR4, which are found on a range of immunologically active cells including macro- phages, B and T lymphocytes, and polymorphonuclear leu- kocytes (2-4). C3 opsonization is critical for induction of antibody response (5-a), B lymphocyte proliferation and memory (9), and antigen processing and presentation as well as immune complex clearance (5, 6). Deficiencies in comple- ment pathway components which lead to inefficient recogni- tion of and fixation to antigenic targets result in decreased antigen clearance and are clearly associated with immune complex-mediated diseases, severe infections, and autoim- mune sequelea (10-12).

Activation of C3 via the classical pathway requires activa- tion of the C1 complex, C4, and C2, whereas, in the alternative pathway, initial C3 activation is thought to be spontaneous. The common structural features of the surfaces of the sub- strates which direct the recognition and activation of the

* This work was supported by Grant HD 17461-08 from the Na- tional Institutes of Health. 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" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

ll A Malcolm Hecht fellow and recipient of an Arthritis Foundation Medical Student Research Award.

** To whom correspondence should be addressed.

alternative pathway of complement remain to be elucidated. However, regulation of C3 activation has been shown to be controlled in part by factor H which inhibits formation and increases the rate of dissociation of the C3 convertase (C3bBb) on autologous nonactivating surfaces (13). Factor H binds much more avidly to C3b associated with nonactivating rather than activating surfaces, and this may be a vital event in regulation (14, 15). Properdin is a positive regulator of C3 binding which apparently acts to stabilize the interaction of the alternative pathway convertase with its target surface (16- 18). These observations imply that specific molecular confor- mations of target surfaces are critical to the binding of C3b and may distinguish an activating surface from a nonactivat- ing surface.

Human complement C3 is a 190-kDa protein consisting of disulfide linked a(115 kDa) and p(75 kDa) chains (19). The mechanism of C3 binding has been well defined and occurs through a transacylation reaction involving a intra-chain thioester group formed between the side chains of C ~ S " ' ~ and G~x" '~ on the C3a' chain (20,21,22). The thioester becomes activated through a conformational rearrangement induced by a proteolytic cleavage of native C3 by the serine protease activity of the C3 convertase (C3HzOBb or C3bBb) which is formed on the surface of immunogenic targets by the alter- native complement activation pathway (23). The transacyla- tion can occur between the y-carbonyl of GIX"'~ and either - OH or -NH2 bearing side chains of the substrate resulting in either ester or amide linkages. Thus, C3 may bind to serine, threonine, and tyrosine side chains and the eNH2 of lysines on the substrate surface (24, 25). However, the resulting covalent linkages are primarily ester bonds as characterized in previous studies (21, 22, 30).

The hypothesis that C3 binds to a specific region of the IgG heavy chain has been investigated using several different experimental approaches. Using rabbit ovalbumin-antioval- bumin aggregates to activate the alternative pathway, Gadd and Reid (26) showed that C3 binds covalently both to the antigen and to the intact IgG heavy chain. It was suggested that C3b bound to the Fd region of IgG, since F(ab'), frag- ments partially blocked binding to the intact heavy chain as had been suggested for C4 (27). Further support for binding in the Fd region came from Takata et al. (28). They demon- strated a high molecular weight band on SDS-PAGE' which, upon treatment with hydroxylamine, released fragments of C3a' chain (65 kDa) and IgG heavy chain (or the presumed Fd fragment when F(ab')z was used for immune precipitates). These results support the contention that C3 binds to a

The abbreviations used are: PAGE, polyacrylamide gel electro- phoresis; CH1, first constant domain on the Ig heavy chain; HPLC, high-pressure liquid chromatography; IgGl, immunoglobulin G, iso- t,ype 1; Lys-C, endoproteinase Lys-C; PVDF, polyvinylidene difluo- ride; SV8, S. aureus protease V8.

5866

A Major Binding Site for Complement C3 on IgGl Heavy Chain 5867

defined region of the IgG heavy chain within immune com- plexes and is not randomly deposited along the IgG constant domains. In contrast, Anton et al. (29) have proposed that C3 binds to both the CH1 and CH3 domains of IgG based on analysis of pepsin fragments on SDS-PAGE. Recently, Shohet et al. (30) used an autologous human system, i.e. human aggregated IgG and human complement, to localize the binding of C3 through an ester linkage to an 18-kDa region of the IgGl heavy chain encompassing the CH1 domain. This recent result was achieved with an IgGl myeloma protein from a different patient than that used in the present study.

The results of this study further localize the binding to a maximum stretch of 23 residues (134-156) within the CH1 domain by in situ digestion with endoproteinase Lys-C (Lys- C) and Staphylococcus aureus V8 protease (SV8). This region was found to include a cluster of six hydroxyl containing amino acids which would provide acceptor sites for ester linkage to the nascent C3b molecule.

MATERIALS AND METHODS

Reagents and Chemicals-All electrophoresis chemicals were from Bio-Rad. N-Iodoacetyl-N'-(5-sulfo-l-naphthyl)ethylenediamine, i~do[l-'~C]acetamide, and carrier free lZ5I were obtained from Amer- sham Corp. HPLC grade acetonitrile was obtained from Aldrich Chemical Co. All other buffers and chemicals were obtained from Sigma. Endoproteinase Lys-C (EC 3.4.99.30) was obtained from Pro- mega. s. aureus V8 protease (EC 3.4.21.19) and all other enzymes were from Worthington Biochemicals.

IgG Purification and Alkylation-IgG1 was precipitated from plas- mapheresis fractions of a myeloma patient (Goddrum) (gift of Dr. Peter Schur, Brigham and Womens Hospital, Boston) as previously described and concentrated to 20 mg/ml and alkylated with iodoac- temide before heat aggregation (56 "C for 20 min) (30).

Formation, Labeling, and Purification of C3. ZgG Adducts-C3 . IgG adducts were formed by adding heat-aggregated IgGl to whole human serum to activate the alternative pathway as described previously (30). The free thiol ( C ~ S ' ~ ' ~ ) in the C3 ct chain was labeled by adding either iod~[l-'~CC]acetamide or N-iodoacetyl-N"(5-sulfo-l-naph- thy1)ethylenediamine to the serum at the same time as the IgG,.

Peptide Transfer to PVDF Membranes-Preparative gels were "electrotransferred onto PVDF membranes (either Immobilon" from Millipore or Problott'" from Applied Biosystems) at 12 V for 1 h. Adduct bands were identified by ultraviolet fluorescence of marker lanes containing adducts alkylated with the fluorescent reagent N-iodoacetyl-N"(5-sulfo-l-naphthyl)ethylenediamine. Individual bands were cut out for proteolytic digestion.

Sequencing-Automated Edman degradation was performed with an Applied Biosystems Model 477A automated sequenator with on- line microbore HPLC phenylthiohydantoin amino acid detection. Phenylthiohydantoin amino acids were determined by reversed-phase high-performance liquid chromatography as described (31).

Proteolytic Digestion of Adducts Tramferred to PVDF Mem- branes-Adducts were reduced by incubation at 37 "C for 1 h in 10 ml of 6 M guanidine, 200 mM Tris-HC1, 25 mM dithiothreitol, pH 7.2. Iodoacetamide was added to 80 mM, and the membrane was incubated at room temperature for an additional 30 min. Following washing with Hz0 they were incubated with PVP-40 for 30 min at room temperature (PVP-40 solution: 0.25 g of PVP-40, 0.6 ml of glacial acetic acid in 50 ml of H20). The PVP-40 was then removed by extensive washing with deionized HzO. The membrane was suspended in 250 pl of digestion buffer per 15 adduct bands (i.e. from 15 lanes isolated from preparative SDS-PAGE), and the appropriate enzymes were added for digestion. Assuming an estimate of between 10 and 20 pg of adduct protein per preparative band, enzyme-to-substrate ratios of between 1 : lO and 1:20 by weight were used. After digestion the supernatant was removed and saved in a tube treated with silane to limit binding; and the peptides were eluted from the bands with 2 X 300 pl of 40% acetonitrile and then 2 X 250 p1 of 40% acetonitrile with 0.1% trifluoroacetic acid. All the elution supernatants are com- bined with the original supernatant, and the volume was reduced to 150 pl by Speed-Vac. Following concentration, trifluoroacetic acid was added to 0.1%.

HPLC Purification of Proteolytic Peptides-Samples were loaded onto a 4 X 250-mm C2/C1p reversed-phase column (Pharmacia

SuperPac'" Pep-S) equilibrated in 2% aqueous acetonitrile with 0.1% trifluoroacetic acid. Peptides were eluted with a linear gradient from 0.5 to 100% buffer B at a flow rate of 0.5 ml/min for 2 h. Buffer B was 75% acetonitrile with 0.08-0.12% trifluoroacetic acid titrated SO that the ODzl4 was at base line. Peptides were detected at 214 nm or by using an in-line fluorescence detector (McPherson Model FL- 750B). Samples were rechromatographed on a linear gradient from 2 to 50% B in 2 h.

Moss Spectrometry-Matrix-assisted laser desorption time-of- flight mass spectra (32) of HPLC fraction 51 were acquired with a VESTEC VT2000 laser desorption time-of-flight mass spectrometer (VESTEC Corp., Houston, TX), operating at 30-kV accelerating voltage and equipped with a NdYAG laser (Lumonics Inc., Ottawa, Canada). Sample ionization was achieved by 353-nm irradiation, with 8-11s laser pulses at a repetition rate of 5 Hz; each spectrum is the sum of 40 laser shots (33). Sinapinic acid (3,5-dimethoxy-4-hydrox- ycinnamic acid) was used as the matrix (34).

Radioisotope Detection-Iodo['4C]acetamide was detected either by scintillation counting or by Beta-scope'" detection. For counting in the Beta-scope, 5-10% of each HPLC fraction was spotted onto a sheet of PVDF membrane and dried at 37 "C. The blotted sheet was then analyzed. Standard markers were used to position the original spots with the Beta-scope printout.

RESULTS AND DISCUSSION

Previous results demonstrated that heat-aggregated IgGl would activate the alternative pathway when added to human serum. It was also found that the free thiol (CYS'~'~; pro-C3 numbering) which becomes available on activation of C3 could be labeled with either I4C or fluorescent alkylating reagents and this would provide a convenient marker for the adduct peptide. Following fractionation of the adducts by SDS- PAGE, they were digested in situ with CNBr, and the labeled peptides were analyzed by amino acid sequencing. C3 was found to bind to the heavy chain in the region between residues 81-216 (30).

In order to further localize the region of binding, a method was developed to improve enzyme access to the adducts under conditions favoring complete proteolysis. Adducts fraction- ated on SDS-PAGE were transferred to PVDF membranes, and the adduct bands were excised for digestion. Peptides were then generated by digestion with various enzyme com- binations and analyzed by HPLC, hydroxylamine treatment, amino acid sequencing and mass spectrometry.

Efficient enzymatic digestion depended upon buffer, pH, and extent of denaturation and reduction of the adducts. Membrane-associated C3a-IgG heavy chain adducts could be effectively reduced and alkylated with dithiothreitol and iodo- acetamide in the presence of six molar guanidine. In addition, nonspecific binding sites were blocked with PVP-40 prevent- ing adsorption of the proteolytic enzymes on the PVDF mem- brane (see "Materials and Methods") (35). The ester linkage is more susceptible to hydrolysis at alkaline pH, therefore, these steps were all done below pH 8.0. These techniques significantly increased the yield of adduct peptides as meas- ured by optical density and counts/min eluted from the PVDF membrane.

Several different enzymes were used, alone and in combi- nation; including trypsin, chymotrypsin, Lys-C, elastase, and SV8. After digestion, eluates were concentrated and fraction- ated by chromatography on a narrow bore C2/C18 HPLC column as described under "Materials and Methods." The 14C-labeled fragments were followed by scintillation counting or digital "autoradiography" with a Betagen Beta-scope". The general strategy used for these digests was to identify any 14C- containing peaks, rechromatograph them, and analyze them by automated amino acid sequencing. SV8 in combination with Lys-C or trypsin gave the best result. However, in an effort to limit the number of total peptides, SV8 and Lys-C were used predominantly. The chromatography results of the

5868 A Major Binding Site for Complement C3 on IgG, Heavy Chain

SV8/Lys-C digest are illustrated in Fig. 1. When SV8 and Lys-C were used together, a single radio-

labeled peak at fraction 51 was consistently observed in multiple separate experiments. While more than 85% of the counts loaded on HPLC were eluted from the HPLC column, the yield of 14C-labeled material from the membrane was about 25%. This peak was chromatographed again using a shallower gradient and 88% of the radioactivity loaded onto the HPLC column was recovered. A single major 14C-labeled peak was also observed with SV8/chymotrypsin, SVB/trypsin, or SV8 alone (see Table I for recovery yields).

A double sequence of C3 and IgG heavy chain was found on amino acid sequence analysis of SV8/Lys-C fraction 51 (see Table I1 and Fig. 2). The C3 sequence started at position 1002 and extended through the residue 1013 which is the site of covalent attachment. Automated Edman degradation does not routinely detect cysteine due to cyclization and derivati- zation, however, the radiolabeled C ~ S ' ~ ' ~ was confirmed by autoradiography of the phenylthiohydantoin-derived amino acids recovered from the automated amino acid sequenator. The IgGl sequence begins at position 134 of the heavy chain. The same sequence was produced in three separate experi- ments, and the relative yields of the C3 and putative heavy chain residues suggest they are present in a 1:1 molar ratio. In every preparation of fraction 51 analyzed, the yield ab- ruptly dropped off at cycle 12 to values close to background. This would be expected if that cycle is the COOH terminus of the C3 portion of a dimeric peptide or if the covalent

, . . . . , . . :... -.,.. .. .. - .. . . . .. . . . . . _. ~ . . . .

. . . . - . . . . . . . . . .. . I .. . ,. . .I . .. .. , , - . . , . .. . .

~ - . . .... . .. .. ~. .... . . _ ,.. ~. -. . . . . ~ . . .. ... - ......._,. _ . - _ ., . . . .. ... . . . ... . . : - , - . , .- . - , . . .. . . - . , . . - . , . . . .

.- . . . ..' . . .. .. . ... 1 q ) : ; : " . . ... . - . . . . - . . - -1 -.' . . . .

47 40 49 50 51 52 53 54

. . . ._ . , . . . . . . , . .,. . . .. . .. , . .. - . .

FIG. 1. Purification of proteolytic fragments of C3 . IgG ad- ducts by HPLC. A illustrates the C2/C18 reversed-phase chromat- ograph of a complete digest with SV8/Lys-C (Pharmacia SuperPactm Pep-S, see "Materials and Methods" for conditions). The arrow indicates fraction 51. B demonstrates the single "C-labeled peak at fraction 51 detected by autoradiography. No other radiolabeled frac- tions were detected. C illustrates the purification of fraction 51 by repeat chromatography. All peptides were detected at 214 nm.

TABLE I Yields of peptides eluted from PVDF membranes

Enzyme digest cpm x 10" % total* Amount recovered' %

SV~/LYS-C 50 25 88 SV8/Chymotrypsin 46 21 95 SV8/trypsin 62 28 70 SV8 alone 6 2-3 NDd

a Counts per min are uncorrected counts from Beta-scope" readout. Between 5 and 10% of each sample was counted. Background varied between 0.4 and 0.6 cpm.

* Counts eluted from PVDF/counts left on PVDF + counts eluted from PVDF X 100.

Counts recovered in fraction(s) of HPLC/counts loaded X 100. ND, not determined.

TABLE I1 Sequence comparison of SW/Lys-C-deriued peptide with IgG, heavy

chain and C3a'

SV8'endo-Lys (amino acid 1002) (amino acld 134) C3a' I&!

. . . .

S(11) S S G(14.5), S(7.0) G S E(6.0), C" C" G(12.5) G E(6.1), I(4.0) E

The cystine (C) at residue 9 was identified by detection of '*C label recovered from the sequencer fraction collector. The calculated yield for each amino acid is reported in parentheses in picomoles.

A. IgG CHI domain 8o

S-R-N-D-S-K-N-T-L-F-L-Q-H-D-S-L-R-P-E-D-T-G-V-Y-F-C-A-R-

D-G-G-H-G-F-C-S-S-A-S-C-F-G-P-D-Y-W-G-Q-G-T-P-V-T-V-S-S-

90

IO0 110 120

130 140 A-S-T-K-G-P-SIV-F-P-L-A-P-S-S-K-S-T-S-G-G-T-A-A-L-G-C-L-V

I50

16" I , " 4 no ~ 1 ~ - ~ - ~ - ~ - ~ - ~ - ~ - ~ - ~ - ~ - ~ - ~ - ~ - ~ - ~ - ~ ~ ~ - ~ - ~ - ~ - ~ - ~ - ~ - ~ - ~ - ~ -

, "" , I " . "

190 200 L-(l-S-S-G-L-Y-S-L-S-S-V-V-l-Vp-S-S-S-S-L-G-T-~-l-Y-l-C-N-

210

B. C3 alpha chain 980 990 1000

I-L-L-Q-G-1-P-V-A-a-n-T-E-D-A-V-D-A-E-R-L-K- 1010 1020

H-L-I-V-T-P-S-G-C-G-E-€ N-M-I-G-M-T-P-1-I-V-I-A-

V-H-Y-L-D-E-T-E-0-W-E-K

FIG. 2. Primary structures of IgGl heavy chain and C3 a chain demonstrating the regions of binding (37, 38). The IgG peptide, identified by sequencing, begins at position 134. Based on predicted molecular weight the peptide extends through position 156. The C3 peptide begins at position 1002 and extends at least through position 1013. This peptide contains the thioester moiety including the glutamic acid residue at 1013 which covalently attaches to IgG.

linkage blocks further sequencing due to an unanticipated cyclization reaction. Several experiments were initiated to further characterize this putative dimeric peptide.

Hydroxylamine treatment of fraction 51 followed by analy- sis on HPLC yielded the expected result of the appearance of two new peaks. One of the new peaks was 14C-labeled and presumably represents the free C3 a' chain fragment; while the second peak was putative free IgG heavy chain fragment (data not shown). The presence of protein in these peaks was confirmed by amino acid analysis. However, hydroxylamine treatment of some preparations of fraction 51 did not alter the elution time of the 14C-labeled material. Longer incubation with hydroxylamine or the presence of 6 M guanidine hydro- chloride did not affect the results. Several possible explana- tions exist and they are: alterations of intermolecular forces resulting in strong noncovalent association of the peptide products, enrichment of hydroxylamine-resistant bonds dur- ing purification, disulfide bond formation involving cysteine residue 153, or possibly 0-N intramolecular shifts resulting in amide bond formation (insensitive to hydroxylamine).

Alternatively, fraction 51 represents two unlinked peptides that copurify under the HPLC conditions used. The I4C- labeled material would therefore represent a single fragment of free C3a chain. Complete release of the C3-heavy chain ester bond in the intact adduct was not observed in previous experiments (26,28,30). Based on previous results, it appears

A Major Binding Site for Complement C3 on IgG, Heavy Chain 5869

m l z 1300 1800 2 3 0 0 2800 3300 3800 46.1 5 0 . 0 5 3 . 6

FIG. 3. Time of flight spectrum from mass spectrometry of fraction 61. The major ions present are at m/z 3382, 3453, and 3497 and represent the intact dimeric peptide (see text). The dimeric peptide double ions which are detected at 'A the m/z ratio are also labeled. The ion at m/z 1296 corresponds with the calculated mass of the C3 peptide determined by amino acid sequencing. The minor peaks at 2120 and 1164 are unassigned. The predicted IgG mass of 2175 is not seen (see "Results").

P S 3 1 . 4 3 7 . o 4 1 . 8

that the covalent bond becomes less sensitive to hydroxyl- amine as the adduct is reduced in size (50% release of dimeric peptide versus 75% release of intact adduct).' The sensitivity of detection of the Beta-scope'" under the conditions used was sufficient to detect less than 15% of the counts measured in faction 51. Thus, if fraction 51 represented up to 85% release, the other 15% of the total 14C material should have been detected in another peak. In addition, the molar ratios of the C3 fragment and the IgG residues are approximately 1:l after accounting for the relative sequencing efficiencies for each amino acid. I t is unlikely that two copurifying peptides would be present in such similar amounts.

The combinations of SVB/trypsin and SVB/chymotrypsin were also used to digest the C3-1gG1 adducts. With SV8/ trypsin the same single-radiolabeled fraction as with SV8/ Lys-C was observed as determined by HPLC elution time and sequence analysis. This would be expected since there are no arginines within the peptides from residues 134-156 of IgG or 1002-1013 of C3. The sequence of this peptide was the same as with SV8/Lys-C except for a higher background of unas- signed residues (data not shown). The SVB/chymotrypsin digest gave one major fraction corresponding to fraction 51 and two minor l4C-labeled fractions (41 and 48) on the Beta- scope'". Sequence analysis of these fractions (41 and 48) yielded only a short piece of C3 sequence and no heavy chain sequence (data not shown).

These two digests support the results of the SVB/Lys-C digest since in each the same major single peak was resolved; C3 bound to multiple sites on the heavy chain would not be expected to yield a single I4C-labeled peak on reversed-phase HPLC. Digests with multiple enzyme combinations (each containing SV8 protease) yielding identical sequence results strongly suggest that the V8 protease is responsible for the cleavage sites close to the C3 binding site. When SV8 alone was used to digest the adducts, only 2-3% of the counts could be eluted from the PVDF membrane. HPLC analysis of this digest resulted in two radiolabeled fractions (52 and 55) but did not yield sufficient material for sequencing or mass-spec analysis. The proteolytic action of trypsin and chymotrypsin probably facilitated sufficient degradation of the globular IgG. C3 complex to allow access for SV8 to the region of

* J. M. Shohet, P. Pemberton, and M. C. Carroll, unpublished observation.

covalent binding. Cleavage a t Lys residues observed with chymotrypsin are attributed to contaminating trypsin activ- ity.

The IgG sequence data indicate a cleavage between Ser'33 and Val'34 which would not be normally predicted with the enzymes used. There are several possible explanations for this. There may be contaminating protease activity in the SV8 preparation used in our study. Alternatively, the associ- ation of the adduct complex to the PVDF membrane could alter the peptide conformation making this Ser-Val bond unusually susceptible to hydrolysis. Conformational changes induced by the intimate association of IgG and C3 may also influence the stability of this bond.

Further characterization of the SVB/Lys-C fraction 51 was performed by mass spectrometry which yielded three large masses of 3382, 3453, and 3497 Da (see Fig. 3). Some hetero- geneity would be expected due to alkylation of the C3 CYS '~ '~ as well as possible amino or carboxy peptidase activity of the proteolytic enzymes. However, the primary peak was observed a t 3453 Da. In addition, three other distinct peptide masses of 2120, 1164, and 1296 Da were detected.

The sum of the masses of C3 and IgG should equal the masses of the dimeric peptide and a water molecule (water of hydration lost during original ester bond formation). The ion a t m/z 1296 (Fig. 3) corresponds to the 12-residue proteolytic fragment of C3 (cleavage a t Lys'O0' and G ~ U ' ~ ' ~ ) with C ~ S ' ~ ' ~ alkylated (calculated m/z 1298). The other two ions (mlz 1164 and 2120) do not correspond exactly to either C3 or IgG peptide fragments but could represent the expected hetero- geneity. Alternatively, they may be contaminants or partially degraded fragments of the adduct. Assuming that the mass of the intact dimeric peptide is at most 3453 Da and that the C3 portion of the dimeric peptide is 1296, the predicted mass of the IgG portion is approximately 2175 Da ((3453 + 18) - 1296 = 2175). Thus, the mass spectrometry data in conjunc- tion with the sequence data limit the size of the IgG portion of the dimeric peptide to about 2175 Da. The IgG heavy chain fragment extending from Val'34 t o L Y S ' ~ ~ (mass approximately 2266 Da from consensus sequence) is close to this value. The exact mass of the 134-156 fragment is unknown since the sequence of the IgGl used is not known. However, a single amino acid substitution from the consensus sequence such as Phe to Gly would change the calculated mass of this fragment

A Major Binding Site for Complement C3 on IgG, Heavy Chain 5870

A.

B.

FIG. 4. IgGl crystal structure of the Kol crystal illustrating the putative region of C3 binding. A shows the position of residues 134-156 within the entire IgGl molecule. The right-hand globular domain represents the variable region containing the antigen binding site. On the left is the first constant domain leading to the hinge region. The CH1 region from residue 134 to 156 is colored green except for the hydroxyl groups of the serine and threonine residues which are colored yellow. The small arrow indicates serine 145. B illustrates an expanded space-filling view of the region of interest with the same color scheme as in A. This region contains a cluster of residues with hydroxyl side chains (see Fig. 2) which are clearly exposed to solvent on the surface of the molecule and appear to be unhindered by major steric interactions. Therefore, these residues are available acceptor sites for C3 trankacylation.

TABLE I11 Comparison of human IgG (yl. y2, 7% and y4 subclasses), IgA, and

IgM sequences within the CHI domain (residues 134-156)

Ig Residue

134....... .. 140 . . . . . . . . . . . . . . . . . 150 . . . . . . . . . 156

to 2174. The 134-156 fragment is also predicted from the sequence data (Table 11), and cleavage at Lys 156. It should be noted that, although there may be isotypic variation within this region, the Val'34 and Lys15'j residues are highly conserved within and between isotype classes (see Table I11 and Ref. 36).

These results prompted an analysis of the crystal structure of IgGl to determine the presence of solvent-exposed hy-

droxyl-containing residues in the region of CH1 extending from residue 134 to 156. Graphic analysis was performed on a Silicon Graphics Personal Iris" system using the PEER program (R. Bruccoleri, Bristol Meyers Squibb). Fig. 4 illus- trates the position of this region of the Kol Fab crystal (Brookhaven crystal data bank number 2FB4) (36). The six amino acids containing hydroxyl side chains (4 serines and 2 threonines) in this region are clustered in the region 140 through 148. The hydroxyl side chains of the serine residues 141, 143, 145, and threonine 144 are all clearly exposed to solvent and appear to protrude from the rest of the molecule. There is also a lysine at position 142 whose amino side chain faces inward, unexposed to solvent. Therefore, this residue is not expected to be a potential site for amide bond formation with C3. However, this possibility could not be ruled out. Lack of solvent exposure may also explain why Lys-C did not cleave the lysine-serine bond at position 142-143. Alterna- tively, the presence of the C3 heavy chain covalently bound to this region could have blocked enzyme access to this lysine.

Analysis of the rest of the Kol CH1 crystal structure reveals two more clusters of hydroxyl-containing amino acids at positions 189-193 and 199-201 which are exposed to solvent and therefore are potential binding sites. The CH2 and CH3 crystal structures are not available for graphic analysis; how- ever, the primary structure of these regions does not show any clusters of serine or threonine residues similar to those in the CH1 region.

Three observations from the SVS/Lys-C digest indicate that C3 binds within the region of IgG heavy chain extending from amino acid 134 to 156. First, a single radiolabeled peak was observed from several different digests suggesting that C3 binds to a defined region of heavy chain. Second, a double NH2-terminal sequence corresponding to positions 1002 on the C3a' chain and position 134 on the heavy chain was observed. This sequence abruptly stops at the glutamyl resi- due at position 1013 suggesting a C3 fragment of 12 residues (1002-1013). Third, the finding of unique peptide masses that correspond well to peptides of C3 and heavy chain starting at the observed NH2 termini supports the two previous obser- vations. The observed mass of 3453 Da for the intact adduct minus the mass of the sequenced C3 portion limits the size of the IgG portion to a maximum of 23 residues. The results of these experiments demonstrate a major binding site for C3 within the CH1 of IgGl beginning at residue 134 and possibly extending through residue 156.

Since the complete sequence of the IgG, used in this study has not been determined, a potential difficulty is that it might represent an abnormal variant. The IgG used was a mono- clonal IgGl myeloma protein (Goddrum). When this IgGl is aggregated by heat, it has been shown to activate both the classical (39) and alternative complement pathways as ex- pected (this work). Using a myeloma protein from a different patient (Nichol), we have also found that C3 binds to a region encompassing the CH1 domain (30). Thus, it is unlikely that the IgGl (Goddrum) used in this work is sufficiently mutated from normal human polyclonal IgG to affect the location and specificity of C3 binding.

The region of IgG, CH1 domain extending from residue 134 to 156 shares homology with the other subclasses of IgG as well as with the same region in the heavy chains of the IgM and IgA molecules as illustrated in Table 111. Recently, Valim and Lachmann (40) have shown that antibody isotype and subclass significantly affect the efficiency of classical and alternative pathway activation independent of antibody spec- ificity. It may well be that immunoglobulin isotype structural variation affects the access of C3 to the region of binding

A Major Binding Site for Complement C3 on IgG, Heavy Chain 5871

within the CH1 domain, altering the binding efficiency of C3. Alternatively, the acceptor site for C3 binding may vary among the isotypes. Molecular modeling of the overall inter- action between C3 and the IgG subclasses may provide insight into the mechanism which produces the observed variation of complement activating potential of different types of immune complexes.

The finding that C3 binds to IgGl within a narrow region of 23 residues in the CH1 domain rather than to multiple sites throughout the heavy chain contrasts to the binding observed for another thioester containing serum protein a-2 macroglobulin (41). a-2 macroglobulin is a serine protease inhibitor which functions by cross-linking to proteases on cleavage of its bait region. Multiple binding sites for a-2 macroglobulin (primarily the €-amino group of lysine) on five different serine proteases have been demonstrated (41). Thus, specificity in binding does not appear to be a general require- ment for its function. In contrast, specificity in the interaction of C3 with other complement proteins and Ig might be ex- pected to facilitate organization of a macromolecular complex. For example, the binding of nascent C3b to C4b in formation of the C5 convertase is specific for a single acceptor site on the C4 ,8 chain (42). It will be important to determine if C3 binding to the other isotypes of immunoglobulin is also spe- cific for a single region and if only a single acceptor site is utilized.

Acknowledgments-We thank Dr. Fred Rosen for critical review of this manuscript and Linda Fucci for performing the amino acid sequencing. We also thank Joel Schildbach for assistance with the computer graphical analysis. Mass spectral data were provided by Dr. I. A. Papayannopolis of the M. I. T. Mass Spectrometry Facility which is supported by National Institutes of Health Grant RR00317 ( to K. Biemann).

REFERENCES 1. Muller-Eberhard, H. J. (1988) Annu. Rev. Biochem. 67,321-347 2. Fearon, D. T. (1980) J. Exp. Med. 162,20-36 3. Fearon, D. T. & Ahearn, J. M. (1989) Curr. Top. Microbiol. Immunol. 163 ,

4. Ross, G. D. & Rabellino E. M. (1979) Fed. Proc. 38 , 1467-1475 5. Arvieux, J., Yssel, H. & ’Colomb, M. G. (1988) Immunology 66,229-235

83-98

6. Schifferli, J., Ng, Y. C. & Peters, D. K. (1986) N. Engl. J. Med. 316 , 488-

8. Ochs, H. D., Wedgwood, R. J., Frank, M. M., Heller, S. R. & Hosea, S. W. 7. Mastumoto, A. K., et al. (1991) J. Exp. Med. 173 , 55-64

9. Klaus, G. G . & Humphrey, J. H. (1977) J. Immunol. 136,4100-4106

490

(1983) Clin. Exp. Immunol. 63,208-216

10. Ross, S. C. & Densen, P. (1984) Medicine 63,243-273 11. Bitter-Suermann, D. & Burger, R. (1989) Curr. Top. Microbiol. Immunol.

12. Isenman, D. E. (1988) Bailliere’s Clin. Immunol. Allergy 2 , 975-994 13. Whaley, K. & Ruddy, S. (1976) J. Exp. Med. 144,1147-1163 14. Fearon, D. T. & Austen, K. F. (1977) J. Exp. Med. 146,22-33 15. Pangburn, M. K. & Muller-Eberhard, H. J. (1978) Proc. Natl. Acad. Sci.

16. Fearon, D. T. & Austen, K. F. (1975) J. Exp. Med. 142,856-863 17. Muller-Eberhard, H. J. & Gotze, 0. (1972) J. Exp. Med. 136 , 1003-1008 18. Pillemer, L., et al. (1954) Science 120,279-283

20. Tack, B. F. (1985) in Complement (Muller-Eberbard, H. J., and Miecher, 19. Tack, B. F., and Prahl, J. W. (1976) Biochemistry 16,4513-4521

21. Thomas, hf L., inatova, J., Gray, v&. R. & Tack, B. F. (1982) Proc. Natl.

22. Dodds, A. W. & Law, S. K. A. (1988) Complement 6,89-97 23. Pangburn, M. K., Schreiber, R. D. & Muller-Eberhard, H. J. (1980) J. Exp.

24. Levine, R. P., Finn, R. & Gross, R. (1983) Ann. N . Y. Acad. Sci. 4 2 1 , 235-

163,223-233

U. S. A. 76,2416-2420

M., eds) p 49 72, Springer-Verla , New York

Acad. Sci. U. S. A. 79,1054-1058

Med. 164,856-860

25. Law, S. K. A., Minich, T. M. & Levine, R. P. (1984) Biochemistry 2 3 ,

26. Gadd, K. J. & Reid, K. B. M. (1981) Biochem. J. 196,471-480 27. Campbell, R. D., Dodds, A. W. & Porter, R. R. (1980) Biochem. J. 189 ,

245

3267-3272

fi7-m 28. TakaGy Y., Tamura, N. & F’ujita, T. (1984) J. Immunol. 132,2531-2537 29. Anton, L. C., Alcolea, J. M., Sanchez-Corral, P., Marques, G. & Sanchez,

30. Shohet, J. M., Bergamaschini, L., Davis, A. E., and Carroll, M. C. (1991)

31. Tsunasawa, S., Kondo, J., and Sakiyama, F. (1985) J. Bwchem. (Tokyo)

32. Karas, M., & Hillenkamp, F. (1988) A d . Chem. 80,2299-2301 33. Juhasz, P., Papayannopoulis, I. A,, and Biemann, K. (1992) in Proceeding

of the 12th American Peptide Sym osium (Smith, J. A,, and Rivers, J.

34. Beavis, R. C., & Chait, B. T. (1989) Rapid Commun. Mass Spectrom. 3, E., eds) Escom Science Publishers f;. V., The Netherlands, in press

436-439 35. Matsudaira, P. (1987) J. Biol. Chem. 261,10035-10038 36. Marquart, M., Deisenhofer, J. & Huber, R. (1980) J. Mol. Biol. 1 4 1 , 369-

37. De Bruijn, M. H. L. & Fey, G. H. (1985) Proc. Natl. Acad. Sci. U. S. A. 8 2 ,

38. Kabat, E. A,, Wu, T. T. & Bilofsky, H. (1987) Se Uences of Immunoglobulin Chains, National Institutes of Health, Wash. %. C.

39. Carroll, M. C., Fathallah D. M., Bergamaschini, L., Alicot, E. M., Isenman, D. E. (1990) Proc. Nail. Acad. Sci. U. S. A. 87.6868-6872

40. Valim, Y. M. L., and Lachmann, P. J. (1991) Clin. Ex Immunol. 8 4 , l - 8 41. Sottrup-Jensen L. Hansen H. F. Pedersen, H. S. & firistensen, L. (1990)

42. Kim. Y. U., Carroll, M. C., Isenman, D. E., Nonaka M., Pramoon’ago P J. Biol. C h e i . 266, 17723-17757

Takeda, J., Inoue, K. & Kinoshita, T. (1992) J. B’iol. Chem. 26d,4i712 4176

A. (1989) Biochem. J. 267,831-838

J. Biol. Chem. 266,18520-18524

9 7 , 701-704

391

708-712