Localization of the Site on the Complement Component C1q Required for the Stimulation of Neutrophil Superoxide Production*

C1q, the recognition subunit of the classical comple- ment pathway, interacts with specific cell surface molecules via its collagen-like region (C1q-CLR). This bind- ing of C1q to neutrophils triggers the generation of toxic oxygen species. To identify the site on C1q that interacts with the neutrophil C1q receptor, C1q was isolated, digested with pepsin to produce C1q-CLR, and further cleaved with either trypsin or endoproteinase Lys-C. The resulting fragments were separated by gel filtration chromatography and analyzed functionally (activation of the respiratory burst in neutrophils) and structurally. Cleavage of C1q-CLR with endoproteinase Lys-C did not alter its ability to trigger neutrophil superoxide produc- tion. However, when C1q-CLR was incubated with trypsin under conditions permitting optimal cleavage, the ability of C1q-CLR to stimulate superoxide production in neutrophils was completely abrogated. Fractionation of the digests obtained with the two enzymes and identification by amino acid sequencing permitted localiza- tion of the receptor interaction site to a specific region of the C1q-CLR. Circular dichroism analyses demon- strated that cleavage by trypsin does not denature the remaining uncleaved collagen-like structure, suggest- ing that after trypsin treatment, the loss of activity was not due to a loss of secondary structure of the molecule. However, irreversible heat denaturation of C1q-CLR also abrogated all activity. Thus, a specific conforma- tion conferred by the collagen triple helix constitutes the functional receptor interaction site. These data should direct the design of future specific therapeutic reagents ) at 280 nm of 6.8 for C1q (28) and 2.1 for C1q-CLR (27). Alternatively, after enzymatic digestion, the protein concentration of the purified fragments was determined using the bicinchoninic acid protein assay (Pierce) with a C1q-CLR prepara-tion of known concentration (determined using the above extinction coefficient) used as a standard. All proteins were stored at (cid:50) 70 °C. Preliminary experiments demonstrated that optimal conditions for trypsin digestion of C1q-CLR were achieved when TPCK-trypsin was added to C1q-CLR (2.5–4.5 mg/ml in TBS, pH 7.5) to produce an enzyme:substrate ratio (E:S) of 1:10 (w:w) and incubated at 50 °C for 30 min. Endo Lys-C was incubated with C1q-CLR at similar concentra- tions in TBS, pH 7.5 (E:S, 1:200, w:w) at 37 °C for 2 h. Digestions were terminated by placing the samples on ice. Aliquots of the samples were then electrophoresed under nonreducing conditions using the Tricine SDS-PAGE system (10% acrylamide) described by Scha¨gger and von Jagow (29). Fragments produced by enzymatic digestion were charac-terized and separated by gel filtration chromatography using a Super- ose 12 HR 10/30 (Pharmacia Biotech Inc.) column equilibrated in TBS, pH 7.4 (in the presence or the absence of 5 m M CaCl 2 ). The void volume of the column was 8.1 ml, and the total column volume was 25 ml. Circular dichroism was recorded at ambient temperature using a Jasco J720 spectropolarimeter. Data were collected at 0.5-nm intervals, and 4 scans were averaged but not smoothed. A cell of 0.5-mm path length was used. Some circular dichroism spectra were obtained using a Jasco 500C circular dichrometer with a temperature-controlled cuvette. Pro- tein concentrations used were 0.4–1 mg/ml. All data

C1q, the recognition subunit of the classical complement pathway, interacts with specific cell surface molecules via its collagen-like region (C1q-CLR). This binding of C1q to neutrophils triggers the generation of toxic oxygen species. To identify the site on C1q that interacts with the neutrophil C1q receptor, C1q was isolated, digested with pepsin to produce C1q-CLR, and further cleaved with either trypsin or endoproteinase Lys-C. The resulting fragments were separated by gel filtration chromatography and analyzed functionally (activation of the respiratory burst in neutrophils) and structurally. Cleavage of C1q-CLR with endoproteinase Lys-C did not alter its ability to trigger neutrophil superoxide production. However, when C1q-CLR was incubated with trypsin under conditions permitting optimal cleavage, the ability of C1q-CLR to stimulate superoxide production in neutrophils was completely abrogated. Fractionation of the digests obtained with the two enzymes and identification by amino acid sequencing permitted localization of the receptor interaction site to a specific region of the C1q-CLR. Circular dichroism analyses demonstrated that cleavage by trypsin does not denature the remaining uncleaved collagen-like structure, suggesting that after trypsin treatment, the loss of activity was not due to a loss of secondary structure of the molecule. However, irreversible heat denaturation of C1q-CLR also abrogated all activity. Thus, a specific conformation conferred by the collagen triple helix constitutes the functional receptor interaction site. These data should direct the design of future specific therapeutic reagents to selectively modulate this response.
C1q, the recognition subunit of the classical complement pathway, is a 460,000-Da serum protein that has an unusual macromolecular structure and contributes to a variety of functions in the response of the host to infection or injury. As part of the C1 complex, C1q binds to immune complexes or antibodyindependent C1 activators resulting in the initiation of the classical complement cascade (1). However, upon dissociation from the C1r 2 C1s 2 tetramer (2), C1q can interact with cell surface molecules via its collagen-like domain, inducing cellspecific responses. One example of such a C1q receptor-mediated response is the enhancement of phagocytic capacity (3)(4)(5) that occurs when monocytes and culture-derived macrophages interact with C1q. Furthermore, interaction of C1q with neu-trophils (6), eosinophils (7), and vascular smooth muscle cells (8) triggers the generation of bactericidal oxygen species.
C1q shares an unusual macromolecular structure with certain other molecules also known to enhance uptake of specific pathogenic material (9 -13). Like C1q, both pulmonary surfactant protein A (SP-A) 1 and mannose-binding protein (MBP) have collagen-like sequences contiguous with noncollagen-like sequences, short NH 2 -terminal domains containing interchain disulfide bonds, and a single disruption in the Gly-X-Y repeat pattern within the collagen-like sequence (13)(14)(15)) causing a characteristic "kink" or bend in the tertiary structure. Like C1q, SP-A (9) and MBP (16) can enhance Fc receptor-and complement receptor-mediated phagocytosis. However, neither SP-A nor MBP by themselves stimulate superoxide (O 2 . ) production by polymorphonuclear leukocytes (17) in contrast to the readily detectable stimulation by C1q (6,17). This suggests that the ligand interaction sites of the C1q receptor that triggers superoxide production (C1qR O 2 . ) may differ from those sites of the receptor that enhance phagocytic capacity. Also consistent with the hypothesis that these receptors differ in some way is the observation that two monoclonal antibodies have been shown to inhibit the enhancement of phagocytosis mediated by C1q but do not block the production of superoxide by neutrophils (18,19).
Very little is currently known about the ligand requirements for functional interaction with the C1q receptor on any cell types. It is known that the 176,000-Da collagen-like fragment of Clq mediates the functional interaction with most cells (4,20), that C1r 2 C1s 2 blocks that interaction (6), and that heataggregated C1q loses the ability to trigger B cell immunoglobulin secretion (21). In addition, it appears that a multivalent interaction is required for C1q to trigger a response upon binding to the cells (6,21). It is not known whether this is due to a requirement for receptor clustering or a surface-or aggregate-induced conformational alteration. This requirement is similar to that seen with CR1, CR2, Fc ␥ RIII, and many other receptors (22,23). If the receptor for C1q that triggers superoxide production differs from the one that enhances phagocytosis, it is probable that the interaction sites on the ligand, C1q, that mediate these responses are also distinct. Therefore, identification of the receptor interaction sites should permit the selective manipulation of the responses mediated by the different receptors; that is, to enhance phagocytosis without generating extracellular superoxide or to inhibit the production of superoxide without effecting monocyte phagocytic capacity.
The present investigation characterizes a specific region on the collagen-like domain of C1q (C1q-CLR) that interacts with the neutrophil C1q receptor (C1qR O 2 . ) to induce superoxide production. Enzymatic digestion, followed by functional analysis of the purified fragments and amino acid sequencing to unambiguously identify the cleavage sites generated, allowed the localization of a specific region of the C1q-CLR that is critical for the C1q-mediated neutrophil response. Furthermore, heat denaturation, as assessed by circular dichroism analyses, demonstrated that secondary structure of the collagen helix is required for the C1q-induced neutrophil superoxide response, indicating that a single linear amino acid sequence will not suffice for this interaction. Nevertheless, the data suggest that it may be possible to design synthetic antagonists to modulate this interaction and thus limit the production of toxic oxygen radicals.

MATERIALS AND METHODS
Media, Reagents, and Antibodies-Pyrogen-free water (MilliQ-Plus) is used for all laboratory buffers and reagent preparation. TPCK-trypsin was purchased from Worthington (Freehold, NJ), and sequence grade endoproteinase Lys-C (endo Lys-C) was obtained from Boehringer Mannheim. Lymphocyte separation medium was purchased from Organon Teknika Corp. (Durham, N.C.). All other reagents used, except where noted otherwise, were obtained in the highest quality available from Sigma.
Protein Isolation, Enzymatic Digestion, and Biochemical Characterization-C1q was isolated from plasma-derived human serum by the method of Tenner et al. (24) modified as described (25). The preparations used were fully active as determined by hemolytic titration and homogeneous as assessed by SDS-polyacrylamide gel electrophoresis (PAGE). C1q collagen-like fragments (C1q-CLR) were obtained by the digestion of C1q with pepsin using the procedure of Reid (26) as modified by Siegel and Schumaker (27). Protein concentration was determined using an extinction coefficient (E 1% ) at 280 nm of 6.8 for C1q (28) and 2.1 for C1q-CLR (27). Alternatively, after enzymatic digestion, the protein concentration of the purified fragments was determined using the bicinchoninic acid protein assay (Pierce) with a C1q-CLR preparation of known concentration (determined using the above extinction coefficient) used as a standard. All proteins were stored at Ϫ70°C.
Preliminary experiments demonstrated that optimal conditions for trypsin digestion of C1q-CLR were achieved when TPCK-trypsin was added to C1q-CLR (2.5-4.5 mg/ml in TBS, pH 7.5) to produce an enzyme:substrate ratio (E:S) of 1:10 (w:w) and incubated at 50°C for 30 min. Endo Lys-C was incubated with C1q-CLR at similar concentrations in TBS, pH 7.5 (E:S, 1:200, w:w) at 37°C for 2 h. Digestions were terminated by placing the samples on ice. Aliquots of the samples were then electrophoresed under nonreducing conditions using the Tricine SDS-PAGE system (10% acrylamide) described by Schä gger and von Jagow (29). Fragments produced by enzymatic digestion were characterized and separated by gel filtration chromatography using a Superose 12 HR 10/30 (Pharmacia Biotech Inc.) column equilibrated in TBS, pH 7.4 (in the presence or the absence of 5 mM CaCl 2 ). The void volume of the column was 8.1 ml, and the total column volume was 25 ml. Circular dichroism was recorded at ambient temperature using a Jasco J720 spectropolarimeter. Data were collected at 0.5-nm intervals, and 4 scans were averaged but not smoothed. A cell of 0.5-mm path length was used. Some circular dichroism spectra were obtained using a Jasco 500C circular dichrometer with a temperature-controlled cuvette. Protein concentrations used were 0.4 -1 mg/ml. All data were converted to mean residue ellipticity.
Protein Sequencing-The Edman degradation method was carried out in a Hewlett-Packard Protein Sequencing System model G1005A with an on-line analyzer of the amino acid derivatives. The direct loading of samples, e.g. total digests, onto the sequencing cartridge without previous desalting ensured that all peptide components could be recovered and identified. 4-Hydroxyprolines but no hydroxylysines were identified during sequence analysis. Hydroxylysines predicted from the cDNA sequences (30) and previous amino acid sequencing (31) were verified only as the absence of lysine. C1q-CLR was treated with pyroglutamate aminopeptidase (Boehringer Mannheim) to remove the modified glutamate from the NH 2 terminus of the B chain (32) to permit sequencing of that chain.
Cells-Polymorphonuclear leukocytes are isolated from blood drawn from normal volunteers into EDTA-containing syringes or from buffy coats prepared from blood collected with CPDA1 as anticoagulant.
Measurement of Superoxide Production-O 2 . was measured by the superoxide dismutase-inhibitable reduction of cytochrome c adapted to a microplate format (34). 96-well Immulon 2 plates (Dynatech Laboratories, Chantilly, VA) were coated with 10 -100 g/ml C1q-CLR or test protein/fragments diluted in phosphate-buffered saline for 2-3 h at room temperature. After washing the plate with phosphate-buffered saline, the reaction was started by the addition of 100 l of neutrophil suspension (

Irreversible Denaturation of C1q-CLR Abrogates C1q-induced Superoxide Production by Human Neutrophils-It has
been previously well established that the collagen-like portion of the C1q molecule binds to neutrophils and triggers superoxide generation (6,17). Therefore, in all the present studies, C1q was treated with pepsin, and the pepsin-resistant C1q-CLR isolated by gel filtration was used as the functional ligand. To investigate parameters that could determine structural requirements for the functional interaction of C1q with its receptor, conditions that would lead to the loss of secondary structure as assessed by CD were explored. C1q-CLR in either 10 mM acetic acid, TBS, or TBS with 5 mM Ca 2ϩ , pH 7.2, was heated to various temperatures, and the CD spectra were recorded. Fig. 1A shows that when heated in the acidic buffer used traditionally for collagen-like molecules, secondary structure, as assessed by the magnitude of the positive band at 224 nm characteristic of collagen triple helix, was rapidly lost in the 46 -52°C range, similar to that previously reported by Brodsky-Doyle et al. (35). In contrast, when C1q-CLR was heated at pH 7.2 (in either the presence or the absence of Ca 2ϩ ), the loss of structure of the C1q-CLR was more gradual between 46 and 70°C.
To determine to what extent, if any, this denaturation was reversible and if this denaturation altered the ability of C1q-CLR to trigger neutrophil superoxide production, C1q-CLR was incubated at pH 7.2 for 30 min at 56, 79, and 100°C, temperatures known to induce increasingly greater loss of specific secondary structure, and then cooled to ambient temperature. The CD of the cooled material was then compared with that of an unheated control sample to determine if renaturation had occurred. In contrast to that seen with the acidic conditions used by Brodsky-Doyle and colleagues (35) (and repeated by us), the CD spectra of C1q-CLR heated to 56°C at pH 7.2 and subsequently recooled, was essentially identical to that of the untreated control, with the exception of some variability in the negative peak at 197 nm (Fig. 1B). This renaturation after heating was complete after an incubation of 45-60 min on ice. Heating the protein to 79°C, however, resulted in a greater and largely irreversible loss of C1q-CLR secondary structure. Samples heated at 79°C and above did not renature even after days at 4°C (data not shown).
The effect of loss of secondary structure on the ability of C1q-CLR to trigger the generation of superoxide by neutrophils was then examined. Whereas irreversible loss of secondary . production, C1q-CLR heated to 56°C, followed by the recovery of secondary structure upon cooling, retained total functional activity as compared with the untreated C1q-CLR in its capacity to trigger O 2 . (Fig. 2). In this experiment and all superoxide assays performed, neutrophils in uncoated control wells produced no superoxide, with a recording essentially identical to the 79°C/ 100°C samples (data not shown). Trypsin Digestion of Reversibly Denatured C1q-CLR Abolishes the Neutrophil C1q Response-To investigate which region of the C1q-CLR molecule was required for C1q receptormediated triggering of the superoxide response in neutrophils, C1q-CLR was subjected to enzymatic cleavage. The fact that the C1q-CLR molecule could be reversibly denatured by heating at 50°C was used to promote enzymatic cleavage by trypsin, which normally is not an efficient protease for collagen-like structures. C1q-CLR was heated to 50°C, trypsin was added, and the incubation was continued at 50°C for 30 min. SDS-PAGE suggested that all chains of C1q-CLR were cleaved under these conditions (Fig. 3, lanes 3 and 7) compared with no enzyme controls (Fig. 3, lanes 2 and 6). To separate and char-acterize the resultant fragments, the digestion mixture was applied to a Superose 12 column, and the elution of the peak fractions was recorded (Fig. 4). When compared with the elution of undigested C1q-CLR sample (Fig. 4A, arrow; elution volume, 8.6 ml), the main fragment generated by trypsin cleavage was distinctly retained and thus significantly smaller in size (Fig. 4B, arrow; elution volume, 11.6 ml). This elution behavior is consistent with the evidence of cleavage seen by SDS-PAGE analysis but indicates that some of the multichain structure of C1q is maintained (albumin elutes at 12.6 ml).
The major peak of both the undigested C1q-CLR and the trypsin-generated NH 2 -terminal fragment of C1q-CLR (noted with the arrows in Fig. 4) were then assayed for the ability to trigger polymorphonuclear leukocyte superoxide production. The fragment resulting from trypsin digestion had very little activity as compared with the untreated or mock digested control (Fig. 5). This loss of activity did not reflect an inability of the trypsin-generated fragment to bind to the microtiter well surface as nearly equal molar amounts of the trypsin-digested, and untreated C1q-CLR were detected both by protein assay and by a more sensitive enzyme-linked immunosorbent assay using an anti-C1q monoclonal antibody. The data were consistent with loss of a critical sequence(s) required for the interac-  . . Amino acid sequencing demonstrated that the earliest eluting peak from the trypsin-treated sample (Fig. 4B, pool a) contained the intact NH 2 termini of the A and C chains. The lack of the B chain NH 2 terminus is consistent with the previ-ous identification of pyroglutamate at the NH 2 terminus of this chain (36). Treatment with pyroglutamate-aminopeptidase prior to NH 2 -terminal sequence analysis allowed the quantitative identification of the predicted B chain residues Leu-Ser-(Cys)-Thr and thus verified that the B chain NH 2 terminus was present but blocked by this modified glutamine. Thus, the loss of activity was not due to digestion of the amino terminus of any of the three chains of the C1q molecule. Furthermore, no evidence of any cleavage products associated with the main (NH 2 -terminal) fragment after purification was detected, nor were trypsin autodigestion fragments contaminating this main protein peak. No cleavages by trypsin occurred in the first 31 amino acids of the A or C chain as sequencing through the first NH 2 -terminal 31 residues on the A and C chains demonstrated a quantitative yield of each amino acid following the arginine and lysine residues (Fig. 6). Amino acid sequencing of the other fractions separated by the Superose column containing peptides of smaller size (Fig. 4B, pools b and c) allowed identification of multiple trypsin cleavage sites in the C1q-CLR and Data from both total digest and from purified pools after separation of total digests by Superose chromatography are included. The sequence is taken from Sellar et al. (30). Amino acids are indicated using the single letter code, with hydroxyproline and hydroxylysine (as determined by Sellar et al. (30)) being indicated by the lowercase p and k, respectively. According to Reid and Thompson (36), all hydroxylysines are glycosylated (glucose-galactose disaccharide) except at positions B 50 and B 65. The NH 2 terminus of the B chain of C1q-CLR was derived after treatment with pyroglutamate-aminopeptidase. provided evidence that the most NH 2 -terminal cleavage occurred on the A chain after Arg 38 , on the B chain after Lys 61 , and on the C chain after Arg 41 (Table I and Fig. 6). When tested for the ability to trigger superoxide production using the standard assay conditions, none of the fractions containing the smaller peptide fragments demonstrated any superoxide generating activity (data not shown).

Endoproteinase Lys-C Cleaves All Three C1q Chains but Has No Detrimental Effect on the Ability of the Molecule to Induce O 2
. Production-Digestion of C1q-CLR with lower concentrations of trypsin (E:S, 1:100) resulted in a less degraded fragment as demonstrated by both Superose chromatography and SDS-PAGE (data not shown) and variable but significant activity in the neutrophil superoxide assay. Comparison of the sequence data of the fragments generated in each case revealed an apparent inverse relationship between the amount of superoxide generating activity and the percent of C1q-CLR that was cleaved at the most internal arginine residues (Arg 38 in the A chain and Arg 41 in the C chain). Therefore, digestion with endo Lys-C, which, like trypsin, cleaves at lysine residues but, unlike trypsin, does not cleave at arginine, was assessed to determine if a population of homogeneously cleaved molecules could be generated and subsequently assessed for the ability to trigger superoxide products.
Analysis by SDS-PAGE demonstrated that specific peptide bonds of the C1q-CLR molecules were quantitatively and homogeneously cleaved by endo Lys-C at an E:S of 1:200 (Fig. 7). No additional cleavage occurred when higher amounts (E:S, 1:20) of endo Lys-C were added (data not shown) or when C1q-CLR was preheated to 50°C prior to digestion (Fig. 7, lane  4). The major proteolytic components were then separated and purified by Superose size exclusion chromatography. NH 2 -terminal sequencing of the first eluting peak (Fig. 8B, pool a, corresponding to Fig. 7, lane 1) demonstrated that as in the trypsin digest, this peak contained an unaltered NH 2 terminus with no cleavage fragments remaining associated with the major C1q-CLR fragment generated. Sequencing of the later eluting peak areas (Fig. 8B, pools b and c) as well as the total TABLE I Cleavage site and peptide sequence identification Cleavage site identification and sequences were based on enzyme site specificity and quantitative sequence data from analyses of both total digests (not shown) and pooled fractions derived from chromatography. All amino acid derivatives observed in each cycle were accounted for by the deduced sequences, no predicted amino acids derivatives were missing, and no other cleavages could be accounted for by the observed amino acid derivatives. The ratios of the amino acid derivatives observed were always those expected from the deduced sequences.   2 and 6) and C1q-CLR after digestion with endo Lys-C (lanes 1, 4, and 5) and before (lanes 4, 5, and 6) and after (lanes 1 and 2) purification via Superose gel filtration. Samples in lanes 1 and 2 are from the peak fractions designated by the arrows in Fig. 8 (B and A, respectively) digest identified all of the endo Lys-C fragments generated (Table I and Fig. 6). The most NH 2 -terminal cleavage sites that identify the COOH terminus of the major fragment generated by endo Lys-C and purified by gel chromatography were at Lys 59 of the A chain, Lys 61 of the B chain, and Lys 58 of the C chain. The NH 2 -terminal C1q-CLR fragment generated by endo Lys-C retained complete activity in the neutrophil superoxide assay (Fig. 9), in contrast to the observed loss of activity of the fragment generated by trypsin digestion (Fig. 5).
Enzymatic Digestion of C1q-CLR Does Not Lead to a Loss of Secondary Structure of the Collagen Fibril-The loss of ability of C1q-CLR to trigger neutrophil superoxide production after trypsin treatment could be related to the loss of specific amino acid residues (positions A 38 -59 and/or C 41-58). Alternatively, because the loss of collagen helical structure by heat denaturation abrogated all superoxide generation by C1q-CLR (Fig. 2), it is possible that digestion by trypsin also caused the denaturation or unwinding of the collagen helix leading to a loss of activity. Therefore, the CD spectra of the inactive fragment derived from C1q-CLR by digestion with trypsin was recorded and compared with mock digested C1q-CLR. In addition, C1q-CLR digested with endo Lys-C was also analyzed by circular dichroism for comparison. No differences in the positive peak at 223 nm were seen after enzymatic digestion with either trypsin (Fig. 10A) or endo Lys-C (Fig. 10B) relative to the undigested material. Although small differences are detected in the CD in the 197-200 nm region, such differences are also detected between untreated samples. Therefore, the data suggest that the fragments remaining after enzymatic activity retained collagen helical secondary structure. DISCUSSION Structural and functional studies presented here demonstrate that the region of the C1q collagen-like domain that interacts with a neutrophil surface receptor to initiate superoxide production is present on the fragment generated by endo Lys-C digestion of C1q-CLR but is lost upon digestion with trypsin. Amino acid sequencing of the fragments generated by each of these enzymes localized the most NH 2 -terminal cleavage sites for endo Lys-C to the COOH terminus of Lys 59 of the A chain, Lys 61 of the B chain, and Lys 58 in the C chain, whereas trypsin cleaved on the COOH-terminal of Arg 38 of the A chain, Lys 61 of the B chain, and Arg 41 of the C chain. CD analyses of the fragments after protease digestion demonstrated that peptide cleavage under these conditions, even close to the kink region, does not alter, to a major extent, the remainder of the collagen-like secondary structure. Thus, it appears that the C1qR O 2 . interacts with the collagen helix of C1q in the area of amino acid residues 38 -60 of either the A or C chain. Alternatively, either two or all three correctly folded chains in this region may be required for effective receptor interaction.
Initial studies of heat-denatured C1q-CLR demonstrated that the ability of C1q to interact with the neutrophil C1q receptor and trigger superoxide production required some de-  Table I. Flow rate of the column was 0.5 ml/min. gree of collagen helical secondary structure (Figs. 1 and 2). Brodsky and colleagues previously presented data in which different synthetic collagen-like peptides were analyzed by both CD and equilibrium sedimentation (37). Comparison of our CD data with their more extensive characterization of peptide structure would suggest that the decrease in the magnitude of both the positive (223 nm) and the negative peaks (197 nm) seen after heating C1q-CLR to 70°C or higher is due to a loss of triple helical content within the C1q-CLR subunits. There is no evidence for a significant population of single chain molecules, as there is no discernible shift in the position of the peaks detected by CD. These results suggest that the C1qR O 2 . interaction site on C1q requires some specific secondary structure rather than merely a linear sequence of amino acids for activity. A model for the spatial structure of this region of C1q based on the recently determined crystal structure of a synthetic collagen peptide (38) and the alignment of C1q chains proposed by Kilchherr et al. (39) is presented in Fig. 11. One feature of the model is that the three hydroxylysines that are known to be glycosylated (31) appear to be situated on one side of the helix nearly in the same plane. Although there is no direct evidence that these residues are important in the C1q receptor interaction, data of others demonstrate that carbohydrates often are of critical relevance to pattern recognition and thus host defense (40 -42).
The region of the C1q molecule shown in this report to be of functional importance for triggering superoxide production in the neutrophil is different from that proposed by Malhotra et al. (43) to be the site of interaction of C1q with the calreticulinlike 56,000 M r C1q-binding protein, designated as cC1qR (C1q receptor interacting with the collagen-like region of C1q). These investigators deduced that the region on the NH 2 -terminal side of the kink in the C1q-CLR was the interaction site for cC1qR based on the comparisons of amino acid sequences and charge distributions among members of the family of molecules referred to as "defense collagens" (44) or "collectins" (45). This family of molecules, including MBP, SP-A, conglutinin, and C1q, have collagen-like sequence domains and usually display lectin-like activity (46). However, MBP and SP-A have not yet been shown to independently trigger O 2 . production as does C1q, and there has been no evidence reported that cC1qR is the receptor through which immobilized C1q triggers superoxide production. Therefore, it is likely that C1qR O 2 . is distinct from the calreticulin-like cC1qR and that the functionally relevant interaction site for triggering superoxide production is COOHterminal to the kink region of the C1q molecule.
MBP and SP-A, like C1q, have been shown to stimulate the ingestion of pathogens (47)(48)(49) and enhance the phagocytic capacity of monocytes and macrophages (4,9). Monoclonal antibodies to a 126,000 M r cell surface molecule (distinct from the calreticulin-like cC1qR) have been isolated that inhibit the MBP-(16), SPA-, 2 and C1q-mediated enhancement of phagocytosis (18,19). These antibodies do not inhibit the production of superoxide by neutrophils (19), providing further evidence that the C1q receptor that modulates phagocytic activity on myeloid cells and C1qR O 2 . are distinct receptor complexes on these cells. Although the C1q binding component of C1qR O 2 .
has not yet been identified, a recent study has used immunological and biochemical analysis to rule out several candidate surface molecules and additionally has obtained evidence implicating a multicomponent receptor complex, including CR3, in the transmembrane signaling involved in this response (50).
In 1983, Schumaker and co-workers (51) postulated a model of C1 macromolecular structure in which the tetramer C1r 2 C1s 2 associates with the central portion of the C1q molecule within the cone defined by the C1q arms with the ends of C1r 2 C1s 2 wrapping around the arms of the C1q hexamer at the region just above the kink region. This model was based on electron microscopy, molecular volume calculations, and symmetry considerations. The data presented here that localizes the neutrophil receptor interaction site above the kink support this model of Schumaker and co-workers (51) because our previous data demonstrated that the addition of C1r and C1s to C1q to form the complete C1 complex blocks the ability of C1q to stimulate superoxide production (6). These data then are consistent with the close association of C1r 2 C1s 2 with the region between residues 40 -60 rather than with the region closer to the globular heads.
Finally, NH 2 -terminal amino acid sequencing of the intact C1q molecule and fragments produced by enzymatic digestion (Table I and Fig. 6) allowed the confirmation of several previously reported sequences throughout the collagen-like region of C1q (31). 3 The sites of pepsin cleavage on each chain were clearly defined as in these preparations the COOH-terminal sequences were obtained in high yield without evidence of partial cleavages. In addition, we established the presence of a hydroxylated proline at position 75 in the A chain. Proline had been predicted by the cDNA sequence (30) but differed from the reported residue determined by amino acid sequencing (31). In addition, the lysines at amino acid 10 of chain A and at residue 19 of chain C reported by Reid (31) were obtained in a lower yield than expected and thus can be assumed to be partially hydroxylated. Interestingly, the peptide bond after the hydroxylysine at position B 65 was only partially cleaved by trypsin but was quantitatively cleaved by endo Lys-C. Although this was the only hydroxylysine that was cleaved by these proteases, its susceptibility was not due to heterogeneity 2 S. Ruiz and A. J. Tenner, unpublished data. 3 In the course of our analyses we noticed that the post-translational hydroxylation at residue 39 of the A chain precursor protein (corresponding to amino acid 17 in the mature protein) and residues 71 and 72 in the C chain precursor protein sequence (corresponding to residues 43 and 44 in the mature protein) were incorrectly entered in the Protein Identification Resource data base (accession number, PO 2747); our data clearly show that proline at position A 17 is hydroxylated but at position C 43 it is not hydroxylated, whereas the lysine at position C 44 must be hydroxylated, all in agreement with the protein sequence published by Reid and colleagues (30).  (39)) was used to place the C1q A, B, and C chains in register. The C1q A, B, and C chain residues were superimposed on coordinates of the threedimensional structure of a collagen triple helix peptide consisting of proline-hydroxyproline-glycine repeats (38) using the Biopolymer module in the Biosym Insight molecular modeling package. The amino acid side chains were rotated to match the well established favored rotamers for side chains. of the post-translational hydroxylation, because there was no evidence of a lysine detected upon amino acid sequencing. Whereas, unlike most other hydroxylysines in C1q-CLR, this residue is unglycosylated (31), carbohydrate-dependent steric hindrance cannot be the only factor determining cleavage, because the hydroxylysine at position B 50 is also not glycosylated but was never seen to be cleaved by either trypsin or endo Lys-C.
In summary, C1q functions both as the recognition molecule of C1, the initiator of the classical complement pathway, and as an element in the first line of defense (52,53) by triggering the activation of the NADPH oxidase in neutrophils (17), resulting in the production of microbicidal oxygen radicals, and by enhancing the efficiency of phagocytic cells to clear pathogenic material. Because, as discussed above, the C1q receptor that modulates phagocytic activity on myeloid cells and C1qR O 2 . differ in some critical parameter(s), a potential for selective modulation of these functions exists. Generation of O 2 . , while an effective and critical antimicrobial defense mechanism, is well known to be a major contributor to tissue destruction in many situations including myocardial infarction, pulmonary infection, and joint inflammation. If the C1q interaction with neutrophils contributes to the detrimental production of this toxic oxygen radical in inflammation, it would be clinically advantageous to inhibit this response. Once the interaction site involved in the C1q-mediated enchancement of phagocytosis is identified, the data presented here can be used to initiate the development of strategies for inhibiting the generation of the toxic O 2 . without compromising the ability of C1q to enhance phagocytic capacity.