Dissection of C1q Capability of Interacting with IgG

C1q-bearing immune complexes have been observed in diseases such as rheumatoid arthritis and human immunodeficiency virus infection-associated neuropathy. For the purpose of understanding better the phenomenon of C1q-bearing immune complexes, we investigated the constancy of the C1q-IgG interaction. An enzyme-linked immunosorbent assay was developed in which wells were coated with IgG to mimic antigen-complexed IgG. Serial dilutions of C1q were applied for distinct time intervals, and bound C1q was detected either directly or after exposure to one of several elution buffers. Our results show that a part of C1q attached to IgG forms a tight association that is not reversible under treatment with buffers containing usually protein-protein interaction-dissociating reagents such as 3m NaCl, 5 m urea, sodium dodecyl sulfate, or β-mercaptoethanol. The formation of the highly stable C1q-IgG complex was found to be time-, temperature-, and pH-dependent and to proceed with bound C1q even in the absence of free C1q in the supernatant. In ligand blotting experiments we demonstrate for the first time directly that all three chains of C1q can individually bind IgG. Altogether, our results provide a suitable explanation for the formation and persistence of C1q-bearing immune complexes.

C1q-bearing immune complexes have been observed in diseases such as rheumatoid arthritis and human immunodeficiency virus infection-associated neuropathy. For the purpose of understanding better the phenomenon of C1q-bearing immune complexes, we investigated the constancy of the C1q-IgG interaction. An enzyme-linked immunosorbent assay was developed in which wells were coated with IgG to mimic antigencomplexed IgG. Serial dilutions of C1q were applied for distinct time intervals, and bound C1q was detected either directly or after exposure to one of several elution buffers. Our results show that a part of C1q attached to IgG forms a tight association that is not reversible under treatment with buffers containing usually protein-protein interaction-dissociating reagents such as 3 M NaCl, 5 M urea, sodium dodecyl sulfate, or ␤-mercaptoethanol. The formation of the highly stable C1q-IgG complex was found to be time-, temperature-, and pH-dependent and to proceed with bound C1q even in the absence of free C1q in the supernatant. In ligand blotting experiments we demonstrate for the first time directly that all three chains of C1q can individually bind IgG. Altogether, our results provide a suitable explanation for the formation and persistence of C1q-bearing immune complexes.
The collagen-like C1q molecule is a subcomponent of C1, the first component of complement, and provides a link between the innate immune system, namely the classical complement pathway, and the acquired immunity and some of its most prominent actors, the immunoglobulin classes G and M.
Human C1q is a glycoprotein of about 460 kDa (1). In its electron microscopy image C1q appears as a bunch of tulips, with six globular heads, each connected by a stalk to a central bundle of fibers (2). In the model of Reid and Porter (3), one C1q molecule is composed of 18 polypeptide chains. The chains are of three different types named A, B, and C, of 29, 27, and 23 kDa, respectively. They are linked by disulfide bonds to form six A-B and three C-C dimers. Each of the six individual segments of C1q comprises one chain of each type, which acquire a triple helical structure in the fibrillar region. In contrast, the globular domains appear to be folded to a large extent into ␤-sheets (4,5).
Serum C1q is the key molecule for initiation of the classical complement cascade pathway. Its globular domains recognize the C␥2 domain of IgG or the C3 domain of IgM, especially if these antibodies are complexed with antigen and thus fixed (6 -10). However, C1q differentiates among IgG subclasses because it attaches, in terms of binding efficiency, most strongly to IgG3, followed by IgG1, but it hardly associates with IgG2 and does not react with IgG4 (11). Most studies on the interaction between C1q and IgG favor an ionic rather than a hydrophobic nature of the binding (5,8,12,13), but it has been shown that a dipeptide composed of the aromatic amino acids tryptophan and tyrosine, Trp-Tyr, strongly inhibits the association of the two proteins (14).
The attachment site or sites for IgG on the globular regions of C1q have been a subject of investigation for years. It has been reported that the IgG recognition site is mainly of an ionic nature (8) and that the arginine, histidine, and lysine residues might be essential (13,15). It has been speculated that , and Lys-202 of the C1q A chain might fit with a binding site on IgG (12). Marques and colleagues (5) have suggested that C1q provides two recognition sites for IgG. The first is thought to be located in a section ranging from positions 114 to 129 of the B chain with Arg-114 and Arg-129 being the important residues. The second site has been implicated to involve all three C1q protein chains with Arg residues at positions 162 in the A chain, 163 in the B chain, and 156 in the C chain. Histidine residues appeared again to be relevant for the C1q-IgG interaction too. However, all of these suggestions have been deduced from experiments in which distinct amino acid residues of C1q were modified chemically or where cross-linking to heterologous IgG was performed. Jiang et al. (16) suggested that the globular domain of the C chain provides the primary recognition site for IgG, based on Western blot studies. Kishore et al. (17) reported that a recombinant form of the globular region of the B chain immobilized on a microtiter plate binds IgG but not IgM.
Based on site-directed mutagenesis experiments, the binding site for C1q on IgG has been allocated to the fy2 ␤-sheet in the C␥2 domain, the charged amino acids Glu-318, Lys-320, and Lys-322 being the suggested binding motif (12). A proline residue is present in position 331 in IgG1 but is replaced by a serine in IgG4 and has previously been demonstrated to be, at least in part, responsible for the differential ability of the two isotypes to bind C1q and activate complement (18).
Immune complexes (IC) 1 bearing C1q (C1qIC) have been observed in patients suffering from diseases such as rheumatoid arthritis and HIV infection-associated neuropathy (19,20), but their physiological role is still a mystery. However, persistence of these protein complexes implies a sufficient stability of the C1q-IgG association. To understand better the phenomenon of C1qIC, we investigated the formation and constancy of the C1q association with IgG.
An enzyme-linked immunosorbent assay (ELISA) was developed in which wells were coated with IgG to mimic fixed/ antigen-complexed IgG. Serial dilutions of purified and biotinylated or unlabeled C1q or serum were applied for distinct time intervals. Bound C1q was detected via its biotin tag using an avidin-peroxidase conjugate or antigenically, employing poly-or monoclonal antibodies, either directly or after exposure to one of several elution buffers. Interestingly, our results show that a part of C1q attached to IgG forms a tight association, which is not reversible under treatment with buffers containing usually protein-protein interaction-dissociating reagents such as 3 M NaCl, 5 M urea, SDS, ␤-mercaptoethanol (ME), or dithiothreitol (DTT). The formation of the highly stable C1q-IgG complex was found to be time-, temperature-, and pH-dependent and proceeded with bound C1q even in the absence of free C1q in the supernatant.
Because the C1q molecule comprises three disparate but similar protein chains, we wondered if all three contribute to the interaction with homologous IgG. Therefore, we performed ligand blotting experiments with the separated protein chains of C1q.
Altogether, our results demonstrate directly for the first time that all three chains of C1q can bind IgG individually and that the C1q-IgG interaction is strengthened upon time and thus provides a suitable explanation of the formation and persistence of C1qIC.
Human Immunoglobulins and F(ab)Ј 2 Fragments-Human (hu) IgG subclasses 1-4 and polyclonal huIgG were obtained from Sigma. F(ab)Ј 2 fragments of huIgG were purchased from Dianova, Hamburg, FRG, diluted in PBS, and preadsorbed onto protein G-Sepharose beads (Pierce, KMF Laborchemie Handels GmbH, St. Augustin, FRG) to remove any contaminating Fc fragments or complete IgG molecules. The purity was checked by SDS-PAGE and subsequent Coomassie Blue or copper staining (21) or by an ELISA as described later in this paper.
Specific Antibodies-The following murine (mur) anti-(␣)-huC1q monoclonal antibodies (mAb) are of the IgG1 subclass and were employed in the present study: 242G3, 241F11, 244G8, 4A4B11, and 12A5B7. The clones 242G3, 241F11, and 244G8 were produced in our laboratory (22); 4A4B11 and 12A5B7 were obtained from the American Type Culture Collection (Rockville, MD). The polyclonal goat (gt) ␣-huC1q antibody was also produced in our laboratory, and the polyclonal rabbit (rb) ␣-huC1q antibody was purchased from Behring AG (Marburg, FRG). Antibodies were purified from serum, ascites, or culture supernatants by fast protein liquid chromatography using protein A or G (Pharmacia Biosystems GmbH, Freiburg, FRG) according to the supplier's instructions.
The goat ␣-huC1q Ab was biotinylated using sulfosuccinimidyl-6-(biotinamido) hexanoate (SSBH; Serva) and, for some experiments, preadsorbed onto immobilized huIgG. Briefly, 11.4 mg of purified antibody (6 ml) was dialyzed against 2 liters of 0.1 M NaHCO 3 , pH 9, overnight at 4 -6°C. 1.88 mg of SSBH was dissolved in the protein solution by vortexing, and the mixture was incubated for 4 h at room temperature. Excess labeling reagent was removed by dialysis against two 5-liter volumes of PBS. To reduce background staining when detecting C1q bound to IgG in the ELISA system, 1 ml of labeled Ab was incubated twice for 1 h at 4-6°C with a 100-l packed volume of huIgG coupled to CNBr-activated Sepharose 4B (0.5 mg IgG/ml beads; IgG-S4B, see below).
Preparation and Labeling of Human C1q-Purification and labeling ( 14 C-methylation (23), biotinylation) of human serum C1q were carried out as reported previously for guinea pig serum C1q (24) with the following additional step in the C1q purification procedure. Purified C1q was applied to a protein G column (Pharmacia) equilibrated with the C1q storage buffer, to remove contaminating IgG as far as possible. The purity of the C1q preparations was assessed by SDS-PAGE and ELISA (see below), and only IgG-free preparations were used in experiments in which binding to IgG was investigated.
ELISA Systems-ELISA systems were employed to detect and quantify C1q attached to IgG and also to evaluate the effects on the antigenicity and immunodetection of IgG and C1q of the reagents that were used to detach them.
Assay for Assessment of C1q Binding to IgG-96-well microtiter plates (Maxisorb, Nunc, Wiesbaden, FRG) were coated with 30 l/well of a solution in PBS (0.01 M sodium phosphate, 0.145 M NaCl, pH 7.4; conductivity, 15 mS) containing 250 g/ml, in some experiments 125 or 62.5 g/ml, of huIgG or IgG subclasses for 2 h at room temperature or overnight at 4°C. As controls, wells were coated with BSA or in some experiments with F(ab)Ј 2 fragments instead of whole IgG molecules. The plates were then washed three times with 200 l/well PBS containing 0.3% Tween 20 (PBST) and 1 M NaCl (conductivity, 63 mS) followed by three washes with PBST of physiological ionic strength (conductivity, 15 mS) if not stated otherwise. To block potentially remaining binding sites, 100 l/well of a solution of 1% BSA in PBST was applied twice for 30 min. After two more washes with PBST, or in some experiments PBS containing 0.15 mM Ca 2ϩ and 1 mM Mg 2ϩ (PBS Ca, Mg ), serial dilutions of C1q (concentrations ranging from 0.078 to 20.0 g/ml) were applied and incubated for a certain time interval (usually 1 h) at room temperature or 37°C, as indicated. Serial dilutions of C1q applied to wells coated with goat ␣-huC1q-IgG (1:400 dilution of a stock solution containing 1.9 mg/ml Ab) instead of huIgG served as quantitative standards. Controls in which C1q was omitted or incubated in BSAcoated wells were included in each assay, and all samples were analyzed in duplicate. Respective samples were subjected in parallel to three washes or treatment for a distinct time period with PBST and/or any other elution buffer as detailed in the text. Then the plates were washed sequentially three times each with 20 mM iodoacetamide in PBST and PBST alone. Wells containing C1q standards were always treated only with PBST. Detection of bound/remaining C1q was accomplished by sequential incubations with biotinylated goat ␣-huC1q-IgG (1:800) and avidin-peroxidase conjugate (1:750); or, in the case of biotinylated C1q, with avidin-peroxidase conjugate directly; or with a murine ␣-huC1q mAb (usually 242G3, 1:400) and a goat ␣-murIgG:horseradish peroxidase (1:1,000). All dilutions of Ab or avidin-peroxidase conjugate were prepared in PBST. 20 l/well ABTS, 2 mg/ml, in 10 mM citrate, pH 4.5, containing 0.01% H 2 O 2 served as substrate; the staining reaction was terminated by the addition of 100 l/well PBS. The absorbance at 405 nm was measured in an Anthos Labtec microplate reader (Anthos Labtec Instruments, Salzburg, Austria), and the data analysis was performed using Mikrothek software version 4.0, (Mikrothek Laborsysteme, Overath, FRG).
Assessment of the Effects of Elution Buffers Employed to Remove C1q from Immobilized IgG-Wells coated with huIgG and blocked as detailed above, but not incubated with C1q, were treated in the same manner as if to liberate attached C1q. The remaining immobilized IgG was then detected with a rabbit ␣-huIgG Ab conjugated to horseradish peroxidase as described above for C1q. In some experiments, wells were coated with huIgG containing 0.5, 1, or 2% biotinylated IgG as tracer. Biotinylation of huIgG was performed as described before for the goat ␣-huC1q Ab. After treatment of the immobilized IgG with the respective elution buffers, the remaining IgG was detected using the avidin-peroxidase conjugate. Serial dilutions of IgG served as standards.
Assays for Quantification and Evaluation of Alterations in the Antigenicity of C1q-Quantification and analysis of changes of the antigenicity of C1q were accomplished using a previously described sandwich ELISA system (25) with minor modifications. In brief, wells were coated with 30 l of a 1:400 dilution of goat ␣-huC1q-IgG (1.9 mg/ml) in PBS for 1 h at room temperature or overnight at 4°C. Remaining binding sites were blocked with BSA, as described before, followed by two washes with PBS. Serial dilutions of pretreated C1q or normal human serum were prepared in PBS or PBS Ca, Mg respectively, with or without a final concentration of 20 mM EDTA. Dilutions of purified C1q served as standard. In some experiments, wells were coated with C1q directly instead of Ab. Detection of bound C1q was achieved as described before for C1q attached to huIgG, employing murine ␣-huC1q mAb 242G3 or biotinylated goat ␣-huC1q Ab or avidin-alkaline phosphatase conjugate in the case of biotinylated C1q.
Measurement of C1q Hemolytic Activity and C4 Activation-The hemolytic activity of C1q was assayed as reported earlier (26). C4 activity was determined according to a method described by Atkinson and colleagues (27) with the following minor modifications. Imidazole replaced Veronal in all buffers, and the assay was scaled down to 100 l each, a suspension of IgG-coated sheep erythrocytes (1.5 ϫ 10 8 cells/ml), dilutions of serum sample, and C4-deficient guinea pig serum containing human C2.
Affinity Adsorption-Human IgG was coupled to CNBr-activated Sepharose beads (Pharmacia), according to the supplier's instructions, at a concentration of 10 mg of protein/ml of beads. 10 g of 14 C-labeled C1q in PBS was incubated with IgG-S4B (200-l packed volume) in a total volume of 400 l. The beads were kept in suspension by end-overend rotation for 4 h at room temperature. Then the IgG-S4B was washed three times with 500 l of PBS and sequentially eluted with the following.
Step 3 was SDSsb containing 1% ME for 2 min at 100°C and 2 min at room temperature. The first two elution steps were followed each by three washes with 500 l of the respective elution buffer to prevent a carryover of labeled protein into the next sample. The eluates were subjected to SDS-PAGE and fluorographed as described previously (24). Densitometry of the fluorographs was accomplished using a Herolab Enhanced Analysis System equipment (EASY plus Revision 3.16; Herolab, Wiesloch, Germany).
Ligand Blotting-C1q A, B, and C chains were separated under reducing conditions in a 12.5% acrylamide SDS slab gel containing 1.2% bisacrylamide and transferred onto Immobilon P transfer membrane (Millipore GmbH, Eschborn, FRG) as reported earlier (24). The membrane was blocked with 2% milk powder (Frema Reform, Fink GmbH, Herrenberg, FRG) in 50 mM Tris, containing 70 or 145 mM NaCl, as indicated (pH 7.5; TBS) and 0.05% Tween 20 (TBST). After equilibration in the respective incubation buffer (TBST containing 70 or 145 mM NaCl) by two washes, each for 10 min, the membrane was incubated sequentially with 5 ml of a solution of huIgG or IgG subclasses (10 g/ml, if not stated otherwise) in the respective incubation buffer for 1 h and an goat ␣-huIgG Ab conjugated to alkaline phosphatase for 30 min (dilution 1:20,000). Between the incubation steps, the membrane was washed three times with incubation buffer, and the controls received buffer without huIgG throughout the first incubation. Bound protein was visualized as described before (24) using 5-bromo-4-chloro-3-indolyl phosphate and nitro blue tetrazolium as substrates.
Other Methods-Determination of protein concentration was accomplished as described previously (24).

RESULTS
Binding of C1q to Immobilized IgG and Probing of the Stability of the Interaction-Serial dilutions of purified C1q in concentrations ranging from 10 to 0.078 g/ml in PBST of physiological ionic strength (conductivity, 15 mS) were applied to immobilized IgG in 96-well ELISA plates and incubated for 1 h at room temperature as described under "Experimental Procedures." Then the wells were washed with incubation buffer, and the amount of bound C1q was determined using the ␣-huC1q mAb 242G3. The percentage of applied C1q that bound to IgG and thus the binding efficiency correlated inversely to the C1q concentration (Fig. 1A), ranging from about 10% with a C1q concentration of 10 g/ml to about 40% with 0.078 g/ml of C1q. To assess the stability of the C1q attachment to IgG, the wells were washed with PBST containing 1 M NaCl (conductivity, 63 mS) instead of incubation buffer before detection of C1q. Surprisingly, about 40 -65% of the C1q amount that was bound in buffer of physiological ionic strength remained attached to IgG (Fig. 1B). The addition of 5% (710 mM) ME to the washing buffers further reduced the amount of detectable bound C1q but was insufficient to dissociate C1q from IgG completely (Fig. 1, C and D). An interaction with C1q and formation of the high salt buffer-resistant association were only observed for the IgG subclasses 1 and 3, but not with IgG of the subclasses 2 or 4 ( Fig. 2, A-C). In our assay, C1q bound specifically to immobilized IgG and thereby to the Fc domain, since no binding to immobilized BSA or F(ab)Ј 2 fragments occurred (Fig. 2, D and E). Furthermore, C1q bound minimally if the immobilized IgG was reduced with 710 mM ME or 20 mM DTT and alkylated, using iodoacetamide, before incubation ( Fig. 2, F and G). Increasing the NaCl concentration in the elution buffer up to 3 M or prolonged incubation did not result in a further release of tightly bound C1q (Fig. 3, A and B). An extension of the incubation time with the eluent to 30 min diminished the amount of bound C1q only if the reducing agent ME was present. A treatment for 60 min with the same buffer showed no additional effect (Fig. 3C). C1q did not interact with IgG in the presence of high salt buffers (Fig. 4). Similar obser-FIG. 1 Binding of C1q to immobilized huIgG. The C1q binding assay was performed in 96-well microtiter plates as described under "Experimental Procedures." After a 1-h incubation and respective treatments, bound C1q was immunochemically detected using the murine ␣-huC1q mAb 242G3. C1q bound to wells coated with the goat ␣-huC1q Ab provided the quantitative standard. Percentages (mean Ϯ S.E. of three experiments) are shown of applied C1q that was bound to immobilized IgG in PBST containing 0.145 M NaCl (conductivity, 15 mS) and after washes with the same buffer (panel A); of bound C1q that resisted elution with buffer containing 1 M NaCl (conductivity, 63 mS) (panel B); of bound C1q not removed with 1 M NaCl buffer containing 710 mM ME (panel C); and of bound C1q remaining attached after exposure to incubation buffer in the presence of 710 mM ME (panel D).
vations on the interaction of C1q and IgG were obtained if C1q was detected with other ␣-huC1q mAbs or the polyclonal goat ␣-huC1q Ab or if biotinylated C1q was employed (data not shown). Comparable results were also observed if the concentration of immobilized IgG/well was reduced so that the wells were coated with IgG in concentrations of 125 g/ml or 62.5 g/ml instead of 250 g/ml (see Fig. 2, A-C and below; not all data are shown). Measurement of IgG suggested that, under the conditions of our assay, coating wells with 125 or 250 g/ml IgG resulted in nearly the same amount of immobilized IgG (data not shown).
Additional Attempts to Detach C1q from IgG-To analyze further the nature of the association, the complex of immobilized IgG and C1q was subjected to treatment with PBST at pH 5.5 or PBST containing various agents such as 20 mM DTT instead of ME, arginine, lysine, and urea, sometimes in combination with ME. Alternatively, the associated proteins were incubated in SDSsb at 60°C for 30 min, in the presence or FIG. 2. Specificity of C1q binding to immobilized IgG and IgG subclasses. 96-well microtiter plates coated with huIgG subclasses 1, 2, 3, or 4 were incubated with C1q. Bound C1q was detected after washes with incubation buffer using mAb 242G3 (panels A and D-G) or goat ␣-huC1q Ab (panels B and C; C1q bound to IgG1 and IgG3, respectively) after treatment in parallel with PBST containing 0.145 M NaCl (K15) or 1 M NaCl (K63) or 1 M NaCl and 710 mM ME (K63 ME). Alternatively, C1q was incubated in wells coated with huIgG or BSA (panel D; a control in all assays) or F(ab)Ј 2 fragments of huIgG (panel E) or reduced huIgG (panels F and G). Reduction of IgG was performed for 1 h in PBST containing 1 M NaCl and 710 mM ME (panel F) or 20 mM DTT (panel G) after the initial coating of the wells and before blocking with BSA. Free thiol groups were subsequently inactivated by two washes and incubation for 1 h with 200 l/well each time of 20 mM iodoacetamide in PBST. Representative results of one out of three experiments are shown, respectively. K followed by a number indicates conductivity in mS.
FIG. 3. C1q remaining bound to immobilized IgG throughout treatment with buffers containing NaCl in the absence or presence of ME. Immobilized IgG was incubated with C1q in PBST (0.145 M NaCl) for 1 h at room temperature and then treated with PBST containing 0.58, 1, 2, or 3 M NaCl (panel A, one representative experiment out of three). Alternatively, the proteins were subjected to treatment with 1 M NaCl in PBST, without (panel B) or with 710 mM ME (panel C), for 3 min (three washes), 30 min, or 60 min. C1q resisting elution was detected with mAb 242G3 and is expressed as a percentage of the quantity remaining attached under exposure to PBST of physiological ionic strength (0.145 M NaCl, conductivity, 15 mS). Mean values Ϯ S.E. of five experiments are shown in panels B and C. absence of ME. The various reagents employed in this study to dissociate the complex of immobilized IgG and C1q are listed in Table I.
There were three major considerations in the dissociation of C1q and IgG: the efficiency of the applied reagents in liberating C1q compared with 1 or 3 M NaCl; the effects of the reagent on the immobilized IgG; and the properties of C1q, in particular its antigenicity.
The experiments were performed as outlined under "Experimental Procedures" and in the figure legends. The results are displayed in Figs. 5 and 6 and show that C1q bound to IgG was still detectable after all of the different treatments except that with SDSsb containing 710 mM ME.
Arginine displayed some dissociating activity at a concentration of 100 mM or higher, whereas lysine showed no specific effect (data not shown). The effect of arginine at concentrations of 0.1 and 0.2 M was not the result of increased ionic strength because the conductivity of the buffers was adjusted to 15 mS. 0.5 M arginine displayed a conductivity of 22 mS and was a slightly more effective eluent than PBST of the same conductivity. However, 0.2 and 0.5 M arginine in PBST eluted amounts of C1q similar to those eluded by 1 M NaCl (Fig. 5A). Interestingly, 1 and 3 M NaCl in PBST removed about the same amount of C1q from IgG (Fig. 3A) and were as effective in liberating C1q as was incubation buffer (PBST, 0.145 M NaCl, pH 7.4) adjusted to pH 5.5 (Fig. 5B). 1 and 2 M urea in PBST also removed about the same quantity of C1q as 1 M NaCl (Fig. 5C). 0.5 M urea was a slightly less effective eluent than 1 M urea, but 5 M urea caused the strongest reduction of the detection signal for C1q.
PBST of pH 5.5 or of pH 7.4 containing substances such as arginine, lysine, 0.145-3 M NaCl, or up to 5 M urea affected immobilized IgG only to a small extent, if at all. In contrast, SDSsb and elution buffers containing a reducing agent (ME or DTT) removed large amounts of immobilized IgG (see Table I). 0.145 and 1 M NaCl buffers containing ME detached about 50%, SDSsb and SDSsb with ME about 82 and 97%, respectively, of the immobilized IgG.
Potential effects of the elution buffers/reagents on the antigenicity and function of C1q were assessed by two approaches. In one approach C1q was immobilized in wells of a microtiter plate, subsequently treated with the respective buffers, and then detected in the same way as if bound to immobilized IgG (see "Experimental Procedures"). In the other approach soluble C1q (33-68 g) was incubated in a total volume of 100 -110 l of PBS for 1 h at room temperature with the respective molar concentration of urea. After dialysis against three 100-ml volumes of PBS, antigenicity and hemolytic activity of C1q were analyzed as outlined under "Experimental Procedures." The first approach showed that arginine, lysine, pH 5.5, up to 3 M NaCl and up to 2 M urea had no effect on the antigenic detection of C1q (data not shown). Treatment with 1 M NaCl in the presence of ME (or DTT) reduced the amount of detected C1q as the C1q concentration increased (Fig. 6). Exposure to 5  Alternatively, the complex of immobilized IgG and C1q was incubated with SDSsb without or with 710 mM ME for 30 min at 60°C (panel E). Respective treatments with PBST containing 0.145 or 1 M NaCl were performed for comparison in experiments shown in panels A-D. C1q was detected employing the mAb 242G3 (panels A-D) or the goat ␣-huC1q Ab (panel C). Biotinylated C1q was used in experiments in which SDSsb served the eluent and was detected with avidin-peroxidase. One representative experiment out of three is shown. M urea diminished the detection signal obtained with the polyclonal goat ␣-huC1q Ab, the effect being more pronounced if ME was present (Fig. 6), and it completely abolished recognition of C1q by the ␣-huC1q mAb 242G3 (data not shown). After treatment with SDSsb at 60°C C1q could still be detected with the goat ␣-huC1q Ab. If SDSsb containing ME was applied, the detected C1q amount was reduced to about 12% of that assessed after treatment with SDSsb alone. Exactly the same effect of SDSsb Ϯ ME was observed with biotinylated C1q (data not shown).
The second approach confirmed that treatment with 0.5-2 M urea did not change the antigenicity of C1q. Furthermore, the hemolytic activity of C1q was recovered after removal of urea by dialysis (data not shown). In contrast, pretreatment with 5 M urea destroyed completely and irreversibly the hemolytic function of C1q as well as its capability of binding to Fc of IgG and depressed the antigenic recognition in the C1q sandwich ELISA to about 25%, compared on the basis of equal protein amounts. (The biotinylated goat ␣-huC1q Ab had to be used for detection since C1q treated with 5 M urea retained no reactivity with mAb 242G3; data not shown.) Kinetics of C1q Binding to IgG and Formation of the High Salt-resistant Association-Unlabeled or biotinylated C1q was incubated with immobilized IgG for different time periods and subsequently washed with PBST containing 0.145 M NaCl (incubation buffer) or 1 M NaCl with or without ME. Bound C1q was detected immunochemically or with avidin-alkaline phosphatase conjugate, respectively, as described under "Experimental Procedures." The results (Fig. 7, A and B) showed that the amount of C1q resisting elution with 1 M NaCl increased with time. After 4 h, more than 80% of the bound C1q remained associated with IgG throughout exposure to 1 M NaCl. Although the detected amount of bound C1q was always lower after treatment with buffers containing 710 mM ME, the quantity of C1q withstanding the dissociating effect of the reducing agent increased simultaneously (data not shown).
Binding of soluble C1q to immobilized IgG occurred faster than formation of the tight association between bound C1q and immobilized IgG. This finding suggested that the interaction of C1q and IgG proceeds in two steps, and we wondered if the increasing strength of the association depended on free C1q in the supernatant. Therefore we performed experiments in which the supernatant containing soluble C1q was replaced by PBST after 15 min, and the incubation was continued for up to 4 h. Bound C1q was determined after 15 min (at the time when soluble C1q was removed), 1 h and 4 h (Fig. 7, C and D). The amount of associated C1q barely changed over time in the absence of soluble C1q, but the quantity of C1q resisting elution with 1 M NaCl (ϮME) increased remarkably, as in the presence of soluble C1q. Thus, formation of the tight attachment of C1q and IgG occurs with time and independently of free soluble C1q in the supernatant.
Incubation of C1q with IgG Coupled to Sepharose Beads-In an alternative approach to the ELISA system, 14 C-labeled C1q was incubated with IgG-S4B for 4 h at room temperature. Subsequent treatment with incubation buffer, 0.58 M NaCl, SDSsb, and SDSsb containing ME was performed as described under "Experimental Procedures" to elute C1q. Analysis of the eluates by SDS-PAGE and fluorography revealed that C1q could only be liberated under the influence of SDSsb. SDSsb alone removed only a part of the bound protein since subsequent reduction with ME caused a further release of C1q (Fig.  8). Densitometric estimation showed that SDSsb removed 48.1 Ϯ 9.9%, and SDSsb containing ME removed 51.9 Ϯ 9.9% of all eluted C1q (mean Ϯ S.D. of three experiments). The observed binding of C1q was specific for IgG since no adsorption onto Sepharose-coupled BSA was detected (data not shown, but see Ref. 29).
Incubation of Serum with Immobilized IgG-Because C1q in serum usually acts in concert with C1r and C1s and thus exerts its function as a part of the C1 complex, we wondered whether the tight binding of C1q to IgG could also occur in serum. Serial dilutions of a pooled serum containing 250 g C1q/ml in PB-S Ca, Mg were applied to 96-well plates coated with IgG and incubated for 1 h at 37°C. Then the serum dilutions were removed, and IgG-bound C1q was immunochemically assessed after washes with PBST or PBST containing 1 M NaCl (ϮME) as described before. Additionally, the residual C4 activity in the incubated serum dilutions was determined as a measure of classical complement pathway activation. Serum dilutions applied to wells coated with BSA or containing 20 mM EDTA and purified serum C1q served as controls. The result, shown in Fig. 9A, revealed that tight binding of C1q to IgG developed in serum in the presence and absence of EDTA. Purified C1q and C1q in serum containing EDTA behaved similar to each other with respect to binding to IgG but were distinct from C1q engaged in the C1 complex in normal serum. Binding of C1rand C1s-associated C1q (C1) was less effective at higher, but more efficient at lower concentrations, compared with free C1q in serum containing EDTA. Activation of serum C4 occurred only throughout incubation with immobilized IgG and in the absence of EDTA. Furthermore, C4 consumption correlated well with the observed binding of C1q to IgG.

Incubation of a Treated, Immobilized C1q-IgG Complex with C1q-deficient Serum-To investigate if the tightly IgG-bound
C1q is capable of activating the classical pathway of complement, the following experiment was performed. IgG-associated C1q was sequentially treated with PBS Ca, Mg containing 0.145 or 1 M NaCl or 1 M NaCl containing 710 mM ME and 20 mM iodoacetamide in PBS Ca, Mg as outlined under "Experimental Procedures." After equilibration by three washes with PBS Ca, Mg , C1q-deficient serum was added, and the microtiter plates were incubated for 1 h at 37°C. Thereafter, the serum was removed and assayed for remaining C4 activity as described above. C1q attached to IgG was detected immunochemically after three washes with PBST (not shown). C1qdeficient serum incubated with immobilized IgG alone or applied to wells coated with BSA and preincubated with or without C1q served as controls. Compared with the controls, IgG-attached C1q exerted a remarkable ability to activate C4, also after treatment with 1 M NaCl and even to some degree after exposure to 1 M NaCl with 710 mM ME (Fig. 9B). FIG. 6. Immunodetection of C1q after exposure to buffers containing 1 M NaCl or 5 M urea with or without reducing agent. C1q adsorbed onto microtiter plates was washed three times with PBST containing 1 M NaCl or 5 M urea in the presence or absence of 710 M ME as described above for C1q bound to immobilized IgG. C1q washed with PBST served as standard, and the goat ␣-huC1q Ab was used for detection.

Approaches to Influence the Formation of the Highly Stable
C1q-IgG Complex-To analyze the formation of the tight association of C1q and IgG the potential influence of Ca 2ϩ , Mg 2ϩ , EDTA, temperature, pH, a reducing agent (NAC) and the thiolblocking substance iodoacetamide was investigated.
The addition of Ca 2ϩ and Mg 2ϩ ions (0.15 and 1 mM, respectively) or of their chelator EDTA (20 mM) showed no effect on the tight binding of C1q to IgG (data not shown).
Performance of the binding assays at 4, ϳ22, and 37°C (ϳ22°C was room temperature) revealed that binding of C1q to IgG was achieved at all temperatures. Binding was most pronounced at ϳ22°C, slightly less effective at 4°C, and to the lowest extent at 37°C (Fig. 10A). But formation of the high salt-resistant association occurred markedly only at ϳ22 and 37°C (Fig. 10B). The reduced binding of C1q at 37°C compared with ϳ22°C might be caused by a slight increase of conductivity from 15 to ϳ17 mS at a constant salt concentration. However, the tight attachment of C1q to IgG is influenced by the temperature and strongly impaired in the cold. If C1q and immobilized IgG were incubated in PBST of different pH, ranging from 5.5 to 8.5, the observed binding was the best at pH 7.4 (Fig. 10C). At least for C1q concentrations below 2.5 g/ml the attachment was inhibited at any pH higher than 7.4. Impairment of the C1q-IgG interaction was even more pronounced in buffers of any pH below 7.4. The development of the tight, 1 M NaCl-resistant association was also strongly diminished at pH below 7.4 but appeared to be not changed at pH 8.0 and even increased slightly for C1q concentrations below 2.5 g/ml at pH 8.5 (Fig. 10D).
The association of C1q and IgG depends strongly on intact disulfide bridges of both proteins since prior reduction of one of both is sufficient to prevent the interaction (see Fig. 2 and Ref. 30). To assess a potential role of disulfides in the formation of the tight attachment, we sought a reducing agent that would not abrogate the interaction but allow us to regulate it. Out of four substances considered in pilot experiments, ME, DTT, glutathione, and NAC, the last one appeared to be most suitable. Immobilized IgG was treated with 0.2, 2, or 20 mM NAC in PBST followed by iodoacetamide for 1 h before incubation with C1q. No difference in C1q attachment was observed after exposure of IgG to 0.2 and 2 mM NAC, but less C1q bound to IgG preincubated with 20 mM NAC (Fig. 11). This finding was not due to a loss of IgG induced by 20 mM NAC (data not shown). Incubation of C1q with immobilized IgG in the presence of 2 or 20 mM NAC and subsequent determination of the bound amount revealed a dose-dependent reduction not only of binding in 0.145 M NaCl but also an impaired formation of the 1 M NaCl-resistant association (Fig. 11, B and C). About 30% of the bound C1q resisted elution with 1 M NaCl after incubation without NAC, but only about 20% and about 15% and less, depending on the C1q concentration applied, after incubation in the presence of 2 and 20 mM NAC, respectively. NAC did not inhibit the antigenic detection of C1q (data not shown).
To address the question of whether free and accessible sulfhydryl groups are involved in the C1q-IgG interaction, we performed C1q binding assays in the presence of 0.2, 2, or 20 mM iodoacetamide. However, neither the C1q-IgG interaction nor the antigenic detection of C1q was affected by the thiolblocking reagent (data not shown). Furthermore, iodoacetamide did not influence the hemolytic activity of C1q if present in concentrations ranging from 10 to 250 mM during the incubation of C1q with IgG-coated erythrocytes (data not shown).
Ligand Blotting-C1q consists of three similar but distinguishable protein chains. To analyze if all of these can contribute to the interaction with IgG, the A, B, and C chains were FIG. 9. C1q binding and C4 activation. Panel A, serial dilutions of a human serum pool in PBS Ca, Mg were incubated in the presence or absence of 20 mM EDTA with immobilized IgG (125 g/ml used for coating of wells) or, as a control, with immobilized BSA, for 1 h at 37°C in microtiter plates. Then the serum dilutions were removed, and the residual C4 activity was determined as described under "Experimental Procedures." Samples incubated with BSA or containing EDTA served both as negative controls. C4 consumption in serum samples incubated with IgG in the absence of EDTA is shown and indicated on the right side of the graph. The microtiter plates were treated as described for the C1q binding assay with PBST or PBST and 1 M NaCl, and bound C1q was subsequently detected using the mAb 242G3. Panel B, after incubation of C1q (5 g/ml) with immobilized IgG for 1 h in a microtiter plate, the wells were washed with PBS or PBS containing 1 M NaCl without or with 710 mM ME as described before. Subsequent incubation with C1q-deficient serum and assessment of C4 consumption were performed as described (27). The mean percentage Ϯ S.D. of C4 consumption in C1q-deficient serum observed in three experiments is shown for the incubation with C1q bound after washes with PBS (K15), PBS and 1 M NaCl (K63), and K63 with ME (K63ME), where K indicates conductivity in mS. separated by SDS-PAGE under reducing conditions and transferred onto a polyvinylidene difluoride membrane. Subsequent to saturation of the blotting membrane, the separated C1q chains were incubated with huIgG or IgG subclasses as detailed under "Experimental Procedures." The ligand blotting revealed that all three chains were able to bind IgG in the presence of 70 or 145 mM NaCl (Fig. 12). Binding to the B chain could hardly be detected after incubation with IgG at a concentration of 10 g/ml but clearly after applying 50 g/ml. Interestingly, the separated C1q chains bound IgG1 and 3 but not IgG4. IgG attachment to C1q was specific since no binding to BSA occurred. Furthermore, 1 M NaCl or 710 mM ME failed to remove bound IgG from the C1q chains. In contrast, no IgG remained attached through the washes with SDSsb. DISCUSSION The present study addresses two currently open questions concerning the understanding of the interaction of complement with immunoglobulins: how constant the C1q-IgG association is and how the three protein chains of C1q contribute to this association.
The binding of C1q to immobilized IgG was highly specific under the conditions of our ELISA system since, first, C1q bound selectively to IgG subclasses 1 and 3; second, no association with F(ab)Ј 2 fragments was observed, which indicates that the interaction is mediated by the Fc region; third, C1q barely attached to reduced IgG; and finally, no binding to immobilized BSA occurred (Fig. 2). However, the percentage of applied C1q that bound to IgG, which can be taken as a measure of the binding efficiency, varied among different C1q preparations but was always correlated inversely with the C1q concentration ( Fig. 1).
Direct fixation of IgG in the wells was preferred over antigen-bound antibody because we intended to expose the complex of C1q and immobilized IgG to reagents that would probably also interfere with the antigen-antibody interaction. However, some of the buffers employed to detach C1q from immobilized IgG, such as SDSsb or buffers containing reducing agents, removed also certain amounts of immobilized IgG (Table I).
Because eluted IgG might have carried away C1q, an underestimation of the C1q quantity remaining attached to IgG is likely under these conditions.
To investigate how constant the interaction is, we treated the complex of C1q and immobilized IgG with buffers that were slightly acidic (pH 5.5) or buffers containing arginine, lysine, or high concentrations of salt or denaturing substances, the last two in the presence or absence of reducing agents, respectively (Table I). Subsequent to each of the various elution procedures, which left behind at least about 12% of the immobilized IgG, IgG-bound C1q was still detectable. Arginine and Lysine have been reported to inhibit the hemolytic activity and binding of C1q to IgG (14,31). C1q hardly bound to IgG in the presence of 1-3 M NaCl (Fig. 3) or at pH 5.5 (Fig. 10) urea abrogated irreversibly the IgG binding ability of C1q (data not shown, but see Ref. 30). But once C1q was attached to IgG the respective agents or conditions failed to reverse the binding completely. Furthermore, the reagents that eluted a certain amount of C1q but which did not affect C1q antigenicity (arginine, 1-3 M NaCl, or 1 or 2 M urea) appeared to be equally effective in releasing C1q (Fig. 5). Altogether, our results indicate for the first time that under certain conditions the binding of C1q to IgG is not reversible, in contrast to earlier suggestions by others (32). A part of C1q remained tightly attached to IgG even during exposure to usually strongly dissociating or denaturing agents such as 1 or 3 M NaCl, up to 5 M urea or SDS and heating, indicating that a change in the quality of the C1q-IgG interaction had occurred.
C1q that had been incubated in suspension with IgG coupled to Sepharose beads also resisted elution with high salt buffer and SDSsb, indicating that the tight association could be a general feature of the C1q-IgG interaction. A closer investigation revealed that the binding of soluble C1q to immobilized IgG occurred faster than the formation of the high salt-resistant association between both proteins. The strengthening of the interaction of bound C1q and IgG was found to be time-, temperature-and pH-dependent, even in the absence of free C1q in the supernatant (Figs. 7 and 10). This demonstrated first that the attachment of C1q to IgG proceeds in two steps; second, that two qualitatively distinct states exist in the association of C1q and IgG; and third, that the C1q-IgG interaction cannot only be described in terms of an equilibrium process as suggested earlier (32). The first stage of the C1q-IgG association is characterized by its reversibility in the presence of, for example, high salt concentrations, and the following second stage by its constant nature under conditions that usually dissociate the protein-protein interaction, such as the presence of high salt buffer or denaturing agents.
Based on our results and the following facts, we suppose that the strengthening of the C1q-IgG association is a consequence of a time-dependent arrangement of ␤-sheet interactions between both proteins. First, the globular regions of C1q appear to be to a large extent folded as ␤-sheets (4). Second, IgG domains form ␤-sheets. Third, the binding site for C1q on the C␥2 domain is located on the fy2 ␤-sheet (32). Fourth, C1q enhances ␤-structure formation in amyloid ␤-peptide aggregates (33). Finally, aggregates consisting of ␤-structure are highly resistant to solubilization or dissociation (33).
Interesting questions are whether the two stages of the C1q-IgG association are reflected by changes in the antigenicity of C1q and IgG and whether the stages display functional differences. In the present study, we did not assay for antigenic changes in IgG. We preferred to use unlabeled C1q and its antigenic detection to keep the manipulation of the protein to a minimum. Furthermore, a targeted labeling of C1q in whole serum would have been impossible. However, the results we obtained using a mAb or a polyclonal Ab to detect IgG-bound C1q did not indicate the appearance of neoantigenicity throughout the formation of the high salt buffer-resistant C1q-IgG complex, but neither do they permit exclusion of it. Further investigations employing a reasonable, but not yet available set of mAbs may be required to address this point in more detail.
Some of the elution buffers influenced the immunodetection of C1q per se, in particular 5 M urea, SDSsb, and the buffers including a reducing agent. Therefore, the reduction of the amount of C1q observed after exposure to one of these buffers could be caused by detachment from immobilized IgG or destruction of the C1q structure or by impaired reactivity from disappearance or alteration of epitopes. To overcome at least the disadvantage of epitope loss or change, we employed biotinylated or 14 C-labeled C1q in some experiments. The biotin residue and the radioactive carbon are both linked covalently to the C1q molecule, and their detection is therefore independent of the antigenicity of the protein.
Tight binding of C1q to IgG occurred in serum among C4 activation, which lends additional support to the assumption that it is a physiological event. But C1q engaged in the C1 complex and free C1q released from C1 in the presence of EDTA differed in their behavior (Fig. 9A). C1r 2 /C1s 2 -associated C1q interacted most efficiently with IgG at lower concentrations compared with its free counterpart. This indicated that C1r and C1s not only mediate the activation of C4 by C1 but also influence the C1q-IgG interaction and therefore probably the recognition of IC by the classical pathway of complement. Tightly attached C1q was readily capable of mediating C4 activation after exposure to the elution buffers ( Fig. 9 B). However, our results did not permit identification or exclusion of difference(s) between the amounts of reversibly and irreversibly IgG-bound C1q in terms of a biological function such as the extent of C4 activation. The strengthening of the C1q-IgG interaction could possibly prolong the availability of the C1q-IgG complex for initiation of the classical pathway of complement without changing the quality or extent of the activation event itself.
Biotinylation and 14 C-labeling both involve preferentially free amino groups provided by lysine residues. But we and others have observed that they do not interfere with the IgG binding capability or the hemolytic activity of C1q (23,34). The behavior of biotinylated and unlabeled C1q toward immobilized IgG was identical in our binding assays. In contrast, biotinylation of IgG has been shown to abrogate the binding of C1q and activation of the classical pathway of complement (35). Therefore, it appears that (free and accessible) lysine residues of IgG but not of C1q play a role in the interaction of both proteins.
Sufficient association of C1q and IgG depends strongly on intact disulfide bridges in both proteins since prior reduction of one of the molecules prevents their association (Fig. 2 and Refs.  30 and 36). Our observation that the reducing agent NAC led to a dose-dependent reduction not only of C1q binding to IgG in 0.145 M NaCl but also to an impaired formation of the 1 M NaCl-resistant association (Fig. 11) indicated that disulfide bridges play a prominent role in the strengthening of the C1q-IgG association. This view is supported by the finding that at pH 8.5 compared with pH 7.4 the relative amount of C1q resisting high salt buffer elution increases (Fig. 10), although the binding of C1q to IgG is in general impaired. Disulfide interchange reactions are favored at pH 8.5 (37), and C1q has been observed to form disulfide-linked oligomers under these conditions (38). Interestingly, a half-cystine is located within the binding site for C1q on IgG in position 321 between Lys-320 and Lys-322 on the fy2 ␤-sheet (see Fig. 2 in Ref. 32). Although we do not have any direct experimental evidence, one could speculate that this half-cystine of IgG could be involved in a disulfide interchange reaction with a matching counterpart in C1q. The results of our experiments would be consistent with this hypothesis. However, free and accessible sulfhydryls are probably not required for the C1q-IgG interaction, since iodoacetamide, which blocks thiol groups and has the potential to interfere with disulfide interchange (37), did not influence the C1q-IgG interaction. But C1q possesses cryptic sulfhydryls (30,39), and therefore, disulfide interchange reactions could occur in a concealed molecular microenvironment, inaccessible to iodoacetamide.
Our ligand blotting experiments demonstrated for the first time that each of the three C1q protein chains contributes to the interaction with IgG, and thus we have extended the findings of others that a recombinant fragment of the B chain globular region (17) and the C chain (16) are both capable of binding IgG. However, we employed 5 g, but Jiang and colleagues (16) only 1 g of C1q/lane in the ligand blotting study, which may account for our success in detecting IgG binding to the A and B chains. Additionally, our study shows for the first time that separated C1q chains bind selectively to IgG subclasses 1 and 3 but not to IgG4. Furthermore, the association of the A, B, and C chains with IgG was resistant to elution with 1 M NaCl or ME. This demonstrates that each chain of C1q behaves toward IgG in a way that is similar to the intact, complex C1q molecule under physiological conditions. However, SDSsb removed IgG completely from the separated C1q chains. The association of IgG with the complete C1q molecule might be more stable compared with that of the separated C1q chains since the intact C1q molecule may provide a sterical combination and thus an accumulation of the binding sites and affinities of all three chains.
In summary, the results of our study provide a plausible explanation of the occurrence and persistence of C1qIC, which have been observed in patients suffering from diseases such as rheumatoid arthritis and HIV infection-associated neuropathy (19,20). Therefore, it will be an important issue of future investigations to isolate and analyze C1qIC from sera of such patients to assess the biological relevance of our in vitro observation of a hitherto unrecognized highly stable C1q-IgG association.