Two individual globular head regions (ghA and ghB) of the heterotrimeric C1q molecule (containing A, B, and C chains) were expressed in a bacterial expression system using a coproduction with the bacterial chaperone GroESL. The purified proteins were soluble and monomeric, as shown by gel-filtration analysis. No association into homotrimers was seen, which indicates that the ability to form heterotrimers is coupled with the discrimination against homotrimeric self-association. The individual globular heads retained their binding activities toward two ligands bound by the whole C1q molecule, i.e., IgG and the peptide P(601–613) derived from the HIV envelope glycoprotein gp41. The differential binding activities displayed for these ligands indicated a degree of structural independence of the binding sites from the regions responsible for heterotrimerization. It was found, using single chain recombinant anti-C1q Abs, that the binding sites on C1q for IgG and gp41 do not overlap, and this observation is also consistent with the view that specialization between the C1q polypeptide chains takes place within the C1q molecule regarding their ligand-binding activities.
The C1q molecule is part of the first complement component, C1, which also contains the proenzymes C1r and C1s. Binding of C1q to immune complexes triggers the activation of C1 as the first step in the activation of the classical pathway of the complement system. Complement can also be activated independent of Abs via the alternative pathway or via the lectin pathway (1), which play an important role in innate immune surveillance at the onset of infections by bacteria, fungi, yeasts, or viruses (2, 3). However, the classical pathway of complement can also be activated independent of Abs by direct binding to pathogens, and target organisms, including retroviruses from nonhuman sources, have been found to be lysed by human complement via the classical pathway (4). HIV-1 can directly activate complement via the classical pathway due to the binding of the transmembrane glycoprotein gp41 to C1 (5), but human retroviruses are not lysed effectively by complement (6, 7). Moreover, this Ab-independent mechanism of complement activation is thought to enhance the level of infection of complement receptor-bearing cells at low multiplicity of infection (8). All recognition functions of C1 are conducted by the C1q subcomponent; however, due to the complex molecular architecture of the C1q molecule, the identification of ligand binding sites within C1q has remained difficult. We present a recombinant approach to determine the precise sites involved.
C1q is composed of six identical subunits, each consisting of three distinct polypeptide chains, the A, B, and C chains. C1q can be thought of as a hexamer of heterotrimers. Each subunit has a collagen-like region and a globular head region (gh)3 (9, 10). The collagen-like triple helix begins close to the NH2 terminus and continues approximately 3 nm, or for 10 Gly-Xaa-Yaa triplets, before a kink in the collagenous triple helix is found at a position in which an additional threonine residue is inserted in the tenth triplet of the A chain, and an alanine for a glycine substitution occurs after the tenth triplet of the C chain. The remaining part of each of the collagenous triple helices projects the globular heads symmetrically at an angle of approximately 60° outward from the central collagenous stalk, which results in the familiar “bunch of tulips” overall structure. The remaining 136 noncollagenous amino acids in each A, B, and C chain are considered to associate with each other to form the six globular heads, also visible in electron micrographs. Upon removal of the collagenous tails of C1q with collagenase, the heterotrimeric globular heads remain associated; however, there are no disulfide bonds between the globular regions of the three chains, and they can be dissociated from each other by using strongly denaturing agents such as guanidine hydrochloride or SDS. Consistent with these observations are sequence analysis of the globular head regions of the three chains, which revealed a high content of hydrophobic residues (about 40%) (11). The sequences are about 30% identical to each other (11), and structure predictions suggest a β-sheet architecture (12) with hydrophobic residues thought to form the interface between the three chains. Studies at the genomic DNA level revealed an exon organization for the three C1q genes, with only one intron present per chain, in the collagenous region, just before the triplets at the kink position of the C1q molecule. This exon organization locates the globular head sequences of each chain, responsible for the heterotrimeric association, in the same exon as the distinct sequences of the tenth Gly-Xaa-Yaa triplets, of the A and C chains, which are thought to be involved in forming the kink in the collagen triple helix. It is reasonable to assume that the specific hydrophobic bonds within the sequences of the globular head regions facilitate the interchain recognition and alignment of the three chains at their C-terminal ends to yield the heterotrimeric (ABC) globular head structure that could in turn provide the nucleation point at which the Gly-Xaa-Yaa triplets fold into a collagenous triple helix. The nucleation point would have to align the Gly-Xaa-Yaa triplets in the correct register for the kink in the triple helical structure to form at the correct angle, some 15 triplets (or 12 nm) toward the N-terminal end of the polypeptide chains. However, in addition to conserved hydrophobic residues, which are considered to support the specific interchain recognition and heterotrimerization, there are specific charged residues important in directing the differential binding by ghA, ghB, and ghC to some of the various ligands known to interact with C1q.
Many activating ligands for C1 bind to the globular head regions, the most important being immune complexes (13), but a number of nonimmune substances, such as DNA, C-reactive protein (CRP), serum amyloid P component (SAP), and decorin, can also bind to C1q, and the collagen-like regions have been implicated in some of these interactions. In the case of activation of C1 by immune complexes, it is believed that a conformational change occurs within the C1q molecule, transducing the activation signal to the collagenous region in which the C1r2-C1s2 proenzyme complex is bound in a Ca2+-dependent fashion. After C1 activation and removal of activated C1r2-C1s2 by C1 inhibitor, the collagenous regions are free to interact with a number of cell surface receptors for C1q. Ab-independent triggering of the complement pathway may also occur after binding of C1q to gp41, the transmembrane glycoprotein of HIV-1 (5, 14, 15, 16, 17).
The C1q binding site on IgG (18), as well as the C1q binding site on gp41 (19), both appear to be linear peptide sequences, which can be recognized by intact C1q. The precise sites within the C1q globular head sequences responsible for these interactions are not known. On the other hand, when the three chains of C1q are separated from each other on SDS-PAGE, under denaturing and reducing conditions, and subsequently transferred to nitrocellulose membranes, some binding of IgG molecules by individual chains can be observed. However, these binding reactions vary widely with conditions and are therefore difficult to quantitate.
We have designed an effective expression system generating individual globular head sequences to address the question of structural and functional independence of individual globular head sequences in their binding activities to IgG and the gp41 peptide. These studies have helped to localize some ligand-C1q interactions to specific chains and provide an experimental approach for an identification of specific binding sites within the C1q macromolecule.
Materials and Methods
Chemicals, plasmids, and reagents
All enzyme-conjugated Abs, the p-nitrophenyl phosphatase substrate system, glutathione agarose, and collagenase type VII were purchased from Sigma (Poole, Dorset, U.K.); isopropylthiogalactoside and Luria broth media were purchased from Life Technologies (Gaithersburg, MD). DNA sequencing was conducted with the Pharmacia sequencing kit using [35S]dATP, and using the Kodak X-OMAT S film for developing. The plasmids encoding GroESL (pT-GroE) and thioredoxin (pT-Trx) (20) were provided by Dr. Ishii (Riken, Tsukuba Life Science Centre, Tsukuba, Ibaraki, Japan).
The following buffers were used: STE (10 mM Tris-HCI, pH 8, 150 mM NaCI, and 1 mM EDTA), buffer A (50 mM HEPES, pH 7.4, 150 mM NaCI, 5 mM DTT, 5 mM EDTA, and 0.1% (v/v) Triton X-100), VBS (5 mM sodium barbital, 0.15 mM CaC12, 1 mM MgCl2, and 150 mM NaCl, pH 7.4), TBS-NTC (50 mM Tris-HCl, pH 7.6, 150 mM NaCl, 0.05% (w/v) NaN3, 0.05% (v/v) Tween-20, and 5 mM CaC12), and SC (sodium carbonate buffer, 0.1 M, pH 9.6).
Purified proteins and peptides
C1q was isolated from human plasma by affinity chromatography on IgG-Sepharose (21), followed by ion-exchange chromatography (22). The protein concentration was estimated by measuring the absorption at 280 nm and using an extinction coefficient at 280 nm (1%, 1 cm) of 6.82 for C1q.
The peptide corresponding to the amino acid sequence 601–613 of gp41 of HIV-1 was provided by Dr. G. Arlaud (Institut de Biologie Structurale, Grenoble, France). The peptide was chemically synthesized by Merrifield’s stepwise solid-phase method.
Clones secreting single chain recombinant anti-C1q Abs (scFv/N7 and scFv/N10) and scFv-specific mAb 9E10 were provided by Dr. R. Kontermann with kind permission of Dr. G. Winter (MRC Centre for Protein Engineering, Cambridge, U.K.). The scFv/N7 and scFv/N10 were obtained from the “Vagon” library of human recombinant scFvs and recognize different epitopes within the globular head regions of human C1q (23).
Construction of the plasmids encoding the globular head fragments of C1q
PCR primers were designed to introduce a SmaI restriction site at the 5′ end and an EcoRI restriction site at the 3′ end of the globular head sequences, which encoded the entire globular head region of each chain, beginning with the last glycine of the collagenous region. (SmaI introducing oligonucleotides: A chain, 5′-AAGGGCAGCCCCGGGAACCTCAAG-3′; B-chain, 5′-AAAGGTGAACCCGGGGACTACAAG-3′; and C chain, 5′-CCAGGTGAGCCCGGGAGATACAAG-3′.) (Introducing an EcoRI restriction site after the stop codon at the 3′ end of the PCR fragment: A chain, 5′-AAAGAATTCTCAGGCAGATGGGAA-3′; B chain, 5′-AAAGAATTCTCAGGCCTCCATATC-3′; and C chain, 5′-AAAGAATTCCTAGTCGGGGAAGAG-3′.)
The cosmid 4xB1 (11) containing 24 kbp of genomic DNA with the genes for the three C1q chains was used as a PCR template, with 25 cycles of 60°C annealing temperature, and 72°C and 94°C for extension and denaturing, respectively.
The PCR products had the expected size of the globular head coding regions of the three chains (ghA, 435 bp; ghB, 444 bp; ghC, 426 bp). After digestion with SmaI and EcoRI, the DNA fragments were cloned into the plasmid form of the M13 mp18 and sequenced manually using the Pharmacia dedeoxy-sequencing kit.
The sequences did not show any alterations to the published globular head sequences, and the DNA fragments were transferred to the pGEX-3X bacterial expression plasmid (24). The junction between the GST fusion partner and the globular head sequences included a linker sequence that resembled the collagen-like triplet Gly-Ile-Pro-Gly- in all three constructs.
The Escherichia coli BL21 strain was transformed with the recombinant vectors. After induction of protein expression with 0.2 mM isopropyl β-d-thiogalactoside (final concentration), 20-μl aliquots were analyzed on 12.5% (w/v) SDS-PAGE gels.
Coexpression of the globular head GST fusion proteins with GroESL/Trx chaperone and purification of the fusion protein
Analytical scale preparation of the GST fusion protein.
For coexpression of ghA or ghB with the bacterial chaperones thioredoxin and GroESL, a previously described expression system was used (25). Briefly, to produce ghA or ghB, along with GroESL or thioredoxin, the E. coli BL21 (DE3), harboring pT-GroESL/pT-Trx, were transformed with pGEX-3X-ghA or pGEX-3X-ghB. The transformed cells were grown in 2 ml Luria-Bertani medium with 100 μg/ml ampicillin and 35 μg/ml chloramphenicol. A total of 200 μl of this culture was added to 10 ml LB containing 100 μg/ml ampicillin and 35 μg/ml chloramphenicol, and cultivated to an OD600 = 0.8. Production of proteins was induced by 0.5 mM isopropyl β-d-thiogalactoside for 3 h at 32°C. Bacteria were pelleted, washed with STE, resuspended in STE buffer containing 100 μg/ml lysozyme, and incubated on ice for 15 min. Then DTT and PMSF, to a final concentration of 5 mM and 0.1 mM, respectively, were added.
Bacteria were lysed by addition of 1.5% (w/v) N-laurylsarcosine; cells were disrupted by sonication; the lysates were clarified by centrifugation for 10,000 × g for 10 min, 4°C; and Triton X-100 (2% v/v) was added to the supernatant, which was spun at 10,000 × g for 15 min, 4°C, to give a soluble fraction and a pellet. The pellet was suspended in 200 μl of SDS sample buffer, boiled for 3 min, and centrifuged. The SDS-soluble material, derived from the pellet, was defined as the insoluble fraction.
The recombinant fusion protein was purified by affinity chromatography on glutathione-agarose beads (batch procedure). A total of 30 μl of washed and swollen matrix was added, and the lysate was incubated at 4°C for 25 min on a shaking platform. The beads were washed seven to eight times with ice-cold PBS and resuspended in buffer A. The fusion protein was eluted using the same buffer with 10 mM reduced glutathione.
Large scale preparation of the GST fusion protein.
Cells were treated identically to the analytical scale preparation, except that 10 ml of an overnight culture was used to inoculate a 1L culture, which was then induced as described above.
Purification of the ghA or ghB fragments
The soluble fraction from the sonicated cell lysate was allowed to pass twice through a glutathione-agarose affinity column (10 mm × 150 mm); unbound material was washed off the column until the A280 = 0.02. The ghA and ghB proteins were released by collagenase digestion directly on the agarose beads for 16 h at 37°C. The agarose beads were pelleted, and the supernatant, containing the ghA or ghB fragments, was collected. The recombinant ghA and ghB proteins were subsequently separated from the collagenase by size-exclusion chromatography on a Superose 12 column.
Identification of the ghA and ghB fragments
SDS-PAGE and Western blot analysis.
The recombinant proteins were analyzed by standard SDS-PAGE (26). For Western blot analysis, the proteins were transferred to a nitrocellulose membrane using Phast Transfer. The membrane was blocked with 2% (w/v) BSA in TBS for 2 h. The blot was washed extensively with TBS-T/0.5% (v/v) Tween-20, and the proteins were probed with predetermined dilutions of rabbit anti-human C1q Ab for 4 h. Ab was cross-adsorbed with E. coli proteins before use. The bound Ab was visualized by horseradish peroxidase-conjugated goat anti-rabbit IgG, followed by reaction with 4-chloro-l-naphtol substrate.
Protein sequencing. The purified recombinant ghA was run under reduced conditions on 10% (w/v) SDS-PAGE and transferred to a nitrocellulose membrane, and the single protein band was cut out and subjected to NH2-terminal peptide sequencing. Automated Edman degradation was performed in an Applied Biosystems (Foster City, CA) model 477A gas-phase protein sequencer, and amino acids were identified according to the manufacturer’s instructions on a model 120A-HPLC system.
Gel-filtration chromatography. The recombinant proteins in buffer A with 0.05% (w/v) NaN3 were loaded onto Superose 12 column. The molecular size markers employed were thyroglobulin (667 kDa), alcohol dehydrogenase (150 kDa), BSA (67 kDa), and soybean trypsin inhibitor (20.3 kDa).
Solid-phase binding assays
ELISA for the detection of interactions between C1q, ghA, or ghB with IgG.
Recombinant proteins ghA/ghB (at different concentrations 0–4 μg/well) and C1q (0–1 μg/well) in SC buffer were coated on poly(vinyl chloride) microtiter plates overnight at 4°C. Any residual binding sites were blocked by 2% (v/v) BSA in the same buffer for 2 h. After washing four times with SC/0.05% (v/v) Tween-20, heat-aggregated human IgG (HAIgG) (20 μg/ml) in TBS-NTC was added. The plates were incubated 2 h at 37°C, and the amount of bound IgG was detected by alkaline phosphatase-labeled goat anti-human IgG conjugate. After 2-h incubation, the wells were developed using the p-nitrophenyl phosphate substrate system. The data points are shown as the mean of triplicate experiments.
To determine the ability of ghA or ghB to inhibit binding of immune complexes to C1q (10 μg/ml), HAIgG (20 μg/ml) was preincubated at 37°C for 2 h with various concentrations of ghA or ghB (0, 10, 20, 30, 40, 50 μg/ml), before transferring to C1q-coated microtiter plates. The amount of bound aggregated IgG was detected as described above.
ELISA for interactions between ghA, ghB, or C1q with P(601–613).
The gp41 peptide P(601–613) was immobilized on microtiter plates (2 μg/well), as in Stoiber et al. (27). Nonspecific binding was blocked by two incubation steps with VBS containing 1% (w/v) BSA for 30 min each. After washing twice with the same buffer, the C1q, ghA, or ghB fragments were added at different concentrations in 100 μl VBS/BSA and incubated for 4 h. The plates were washed with VBS/0.1% (v/v) Tween-20. Antiserum to C1q (50 μl) was added at a 1/250 dilution in VBS/BSA. After the next washing step, the amount of bound proteins was detected as described above using alkaline phosphatase-labeled anti-rabbit IgG.
To determine the competition of ghA/or ghB for C1q binding to P(601–613), biotinylated C1q (10 μg/ml) was preincubated with various concentrations of ghA and ghB (0, 10, 20, 30, 40, 50 μg/ml), transferred to gp41 P(601–613)-coated microtiter wells, and incubated for 3 h at room temperature. The amount of bound C1q was detected using biotin-streptavidin system, as described (27).
ELISAs for detection of inhibitory activity of anti-C1q/scFv on the interaction between C1q, or ghA, and P(601–613) of gp41.
The ability of the anti-C1q scFv/N7 and scFv/N10 to inhibit interaction between C1q and P(601–613) or between ghA and P(601–613), was tested as follows. Recombinant anti-C1q/scFv/N7 or N10 (10 μg/ml) were transfered to C1q- or ghA-coated ELISA plates (2 μg/well) in the presence of 500 ng of P(601–613). Plates were incubated 3 h at room temperature and extensively washed, and mAb 9E10 (20 μg/ml) was added. After 3 h of incubation, the amount of bound scFv was detected by peroxidase-labeled anti-mouse IgG Ab, as described above.
Expression, purification, and identification of the ghA and ghB fragments of C1q
We attempted to express the globular head sequences of all three chains of C1q in E. coli as fusion proteins with GST. To assess the level of production and solubility of the expressed protein, the bacteria expressing the fusion proteins were harvested by centrifugation and the bacterial pellet was resuspended and lysed, as described in Materials and Methods, and the resulting lysate was separated into soluble and insoluble fraction by centrifugation. Protein bands of the expected size were detected by SDS-PAGE for the A and B chain globular heads, but a truncated fusion protein was detected for the C chain globular head, and therefore no further work was conducted on the C chain material. In initial experiments, most of the recombinant proteins for all three chains remained insoluble (Fig. 1). The GST-ghA fusion protein, when applied to the GST, could not be eluted from the resin by 50 mM reduced glutathione, and attempts to recover the insoluble protein by solubilization with sarcosyl were unsuccessful. The induction at 32°C and shortening of the time of induction slightly increased the amount of soluble fusion protein. We then tested two bacterial chaperones, thioredoxin and GroESL, for their ability to increase the solubility of the ghA and ghB of C1q. The appearance of recombinant protein in the soluble fraction was demonstrated by SDS-PAGE, followed by Coomassie staining (Fig. 1). The coproduction of GroESL with the ghA had better effect on the solubility of the GST-ghA than thioredoxin. The soluble fraction containing the fusion protein and GroESL was used for purification of both GST-ghA and ghB. GST-ghA was purified by a two-step purification procedure: affinity chromatography on glutathione-agarose and size-exclusion chromatography. The ghA was obtained as follows: the fusion protein was applied on glutathione-agarose beads, and collagenase digestion was conducted directly on the immobilized GST-ghA to release the ghA from its fusion partner. The ghA was separated from collagenase by size-exclusion chromatography on a Superose 12 column. Similar procedures were used to purify ghB. Under the nondissociating conditions used, both ghA and ghB ran at the expected m.w. for a monomer (i.e., 12 kDa). The yield of purified ghA and ghB was approximately 20 μg/L of bacterial culture. The purified ghA and GST-ghA were analyzed by SDS-PAGE and Western blotting (Fig. 2), and shown to be immunologically reactive with polyclonal anti-C1q antisera. N-terminal sequencing of the protein band, obtained after electrophoresis and transfer onto a nitrocellulose membrane, confirmed that the protein was the carboxyl-terminal part of the A chain of C1q. An N-terminal sequence of GNIKDQPRPAFSA was obtained that corresponds exactly to residues 85 to 97 of the A chain of human C1q.
Interaction of ghA and ghB fragments with IgG
A number of solid-phase binding assays were performed to detect the IgG-binding abilities of the recombinant ghA and ghB fragments. The functional activities of the recombinant ghA and ghB were compared. The native C1q, and the ghA and ghB fragments were tested for their IgG-binding properties. HAIgG was added to the microtiter plates coated with the examined proteins. Insulin and GST were used as negative controls. After 3 h at room temperature, the amount of bound IgG was quantified by ELISA. The results (Fig. 3) indicated that IgG binds C1q in a saturable manner, and that there was no binding of IgG to insulin or GST. As expected, HAIgG was bound to both ghA and ghB in a dose-dependent manner, but the IgG-binding activity of ghB was much higher than that of ghA. These results were confirmed by the use of an inhibitory ELISA (Fig. 4), in which HAIgG was incubated with different concentrations of inhibitors (ghA, ghB, ghA + ghB, insulin, and GST). IgG binding to C1q was found to be competitively inhibited either by ghA or by ghB, but the inhibitory activity of ghA was less than that of ghB, as shown on Figure 4.
Interaction of ghA chain with P(601–613)
To analyze the gp41-binding properties of the expressed ghA and ghB fragments, we used a synthetic peptide, corresponding to residues 601–613 of gp41 previously shown to contain the binding site for C1q on gp41. Different concentrations of ghA or ghB were incubated with P(601–613)-coated microtiter plates for 4 h, and their ability to bind immobilized gp41 peptide was determined. The results are shown in Figure 5. The ghA and C1q bind in a saturable manner to P(601–613), whereas only background levels of binding were seen with GST (OD490 < 0.1) or ghB (OD490 < 0.2).
The interaction of immobilized P(601–613) with ghA was found to be approximately fivefold stronger than that with ghB. To correlate the observed binding of both native C1q and recombinant globular heads to P(601–613), ghA or ghB was used to compete with C1q for binding to P(601–613) in an inhibitory ELISA. GhA can inhibit the binding of C1q to P(601–613) in a dose-dependent fashion, whereas no inhibitory effect was observed when ghB was used to try and inhibit the binding of C1q to the immobilized gp41 peptide (Fig. 6).
In a complementary set of experiments, we tested the inhibitory effect of P(601–613) on the interaction of C1q or ghA with the single chain Abs scFv/N7 and scFv/N10, both of which are specific against distinct, but unknown antigenic sequences within the globular heads of C1q. ScFv/N10 recognizes native C1q and inhibits the C1q-IgG interaction; however, it does not interact with any of the isolated polypeptide chains of C1q. ScFv/N7 also binds to native C1q and its globular heads, but it also recognizes denatured C1q and the disulfide-bonded A-B fragment. It does not inhibit the interaction between C1q and IgG. The results demonstrated that in the presence of 500 ng of P(601–613), the scFv/N7 binding to C1q was decreased (Table I), whereas the binding of the single chain Ab scFv/N10 was not affected. The peptide P(601–613) also reduced the binding of scFv/N7 to ghA (Table II).
|Ab .||Inhibitor .||A490 (nm) .||SD (nm) .||Binding to C1q (%) .|
|Ab .||Inhibitor .||A490 (nm) .||SD (nm) .||Binding to C1q (%) .|
ELISA assay for detection of inhibitory effect of P(601-613) on interaction between C1q with C1q-specific single chain Abs N7 and N10 (triplicates). Recombinant anti-C1q scFv/N7 or scFv/N10 were transfered to C1q-coated ELISA plates in the presence of 500 ng of inhibitor. Insulin as an inhibitor and the unrelated mAb 16/1 were used as controls. P(601-613) dramatically reduced the binding of scFv/N7 to C1q, whereas the binding of scFv/N10 remained unaffected.
|Ab .||Inhibitor .||A490 (nm) .||SD (nm) .||Binding to ghA (%) .|
|Ab .||Inhibitor .||A490 (nm) .||SD (nm) .||Binding to ghA (%) .|
ELISA assay for detection of inhibitory effect of P(601-613) on interaction between ghA with C1q-specific single chain Ab N7 (triplicates). Recombinant anti-C1q scFv/N7 was transfered to ghA-coated ELISA plates in the presence of 500 ng of inhibitor. The single chain anti-C1q Ab scFv/N10 does not bind to ghA. Insulin as an inhibitor and the unrelated mAb 16/1 were used as controls. P(601-613) greatly reduced the binding of scFv/N7 to ghA, indicating the localization of the HIV gp41 binding site on the globular head sequence of the A-chain of human C1q.
We have successfully expressed the ghA fragment of C1q in a soluble form using a coproduction of GST-ghA with the bacterial chaperone GroESL. The recombinant ghA protein represented approximately 25 to 30% of all proteins in the total E. coli cell extract. The GST-ghA fusion protein was partially purified by affinity chromatography on a glutathione-agarose column. When a standard procedure for elution with reduced glutathione was used, some E. coli proteins coeluted with the GST-ghA. When the recombinant ghA was released from its fusion partner with collagenase while still bound to the glutathione agarose, and subsequently separated from collagenase by size-exclusion chromatography, a pure preparation of ghA was obtained. SDS-PAGE and Western blotting confirmed the predicted m.w. of 12 kDa in dissociating conditions, and it was shown, in gel-filtration analysis under nondissociating conditions, that the protein exists in solution in a monomeric form. Finally, protein sequencing confirmed the N-terminal sequence of the globular head fragment of the A chain of C1q. The ghB fragment was prepared and characterized in the same manner as for the ghA fragment.
It is important to realize from the formation of individual globular heads that the interchain recognition process, which normally forms a heterotrimer from three different chains, does not allow the formation of homotrimeric assemblies of only a single chain of Clq. The mechanism of trimerization involving an inhibition of homotrimerization while allowing heterotrimerization seems to be encoded in the globular head sequence itself and might be sufficient to direct the folding and assembly of Clq during biosynthesis.
The functional activity of the expressed proteins was analyzed in binding studies involving two known ligands of the C1q globular heads: IgG and the peptide P(601–613) derived from gp41. We found that both of the recombinant proteins, ghA and ghB, bound to IgG, but that the affinity, and thus the relative contribution of each individual globular head to the interaction, was different. This finding suggests that both ghA and ghB can bind to IgG, and this is in agreement with results obtained by others (28) showing that both chains possess exposed arginine residues (A 162, B 114, 129, 163) that are thought to be good candidates for the IgG binding site on C1q (29). Our results are consistent with the view that the B chain might play a leading role in the C1q-IgG interaction.
A differential binding activity was also found in the binding to the gp41-derived peptide; however, the binding activity of the ghA was significantly higher than that of the ghB. This observation was emphasized by the results of the inhibition assay of C1q binding to P(601–613) using ghA as an inhibitor (Fig. 6). The single globular head of the A chain of C1q was able to inhibit approximately 50% of the C1q-gp41 binding in our assay. Although a complete inhibition was not observed, this finding is still significant since the maximal concentration of ghA used was 50 μg, representing a molar excess of about 100 times over the amount of C1q present. Considering the multivalent avidity effect that the six A chain globular head sequences present in intact C1q could produce by simultaneously binding to the P(601–613)-coated surface, a 100-fold excess might not be sufficient to produce 100% inhibition. We cannot rule out, however, that other regions of C1q are involved in the gp41-binding reaction.
It has been proposed (19) that the gp41 binding site on C1q lies at the junction between the collagen-like region and the globular heads. In our experiments, the sequence of the recombinant ghA lacks the amino acid residues from 79 to 84. This absence could be the reason for the incomplete inhibitory effect of recombinant ghA compared with that of intact C1q, although recent data suggest that A chain residues 88–94 play the predominant role in the interaction of C1q with gp41 of HIV-1 (unpublished observation). Our results seem to indicate a prominent role for ghA regions in the interaction of C1q with gp41; however, a more detailed comparative study involving all three globular head sequences will be required to verify and quantitate this apparent specificity.
Additional information was obtained by analysis of the inhibitory effect of P(601–613) on the interaction of scFv with C1q or ghA. The scFv/N7 and scFv/N10 Abs are specific against different antigenic determinants on the globular heads of C1q: as was indicated above, scFv/N10 interacts with native C1q, whereas scFv/N7 recognizes denatured C1q and A-B dimer; scFv/N10 inhibits the C1q-IgG interaction, but does not inhibit the C1q-P(601–613) interaction. In addition, scFv/N10 does not interact with individual C1q chains; therefore, it is probably directed specifically against conformationally restricted antigenic determinants at or near the IgG binding site on C1q. This binding site is different from gp41 binding site. Most likely, the scFv/N7 is directed specifically against sequential antigenic determinant on the globular head region of C1q. The remarkable decrease of scFv/N7 binding to C1q (Table I) or to ghA (Table II) in the presence of P(601–613) supports the view that at least part of the gp41 binding site on C1q is composed of a portion of the A chain globular head.
The three chains involved in the assembly of an intact Clq trimeric globular head are held together by very strong, probably hydrophobic, forces. In addition, since the globular heads once isolated, e.g., by collagenase digestion of intact Clq, are difficult to separate into the individual ghA, ghB, and ghC fragments, it appears likely that the single globular head sequences expressed in this study exhibit conformational alterations to avoid exposing the hydrophobic contact area to the solvent. The use of molecular chaperones during the expression might have eased the folding pathway required to generate individual globular heads in a soluble and monomeric form.
However, since the ligand binding sites localized on individual chains do not seem to be dependent on this part of the molecule to be correctly formed, the sequences forming the binding sites must show a considerable degree of structural independence.
The ligand-binding activities mapped to individual globular heads also indicate that the degree of self-structuring observed within the globular head sequences of each chain is independent of a prior or simultaneous trimer formation, a process that might also help in the interchain recognition process required to form intact Clq.
Since globular head sequences of a single chain of Clq do display the IgG and gp41 binding sites, it seems likely that specific binding sites for IgM or other activating ligands of Clq may be found on predominantly only one of the ghA, ghB, or ghC regions. Although binding sites can be localized to individual chains, Ab-style combining sites, i.e., binding sites made up of more than one chain, cannot yet be ruled out for all ligands.
Future studies will include site-specific mutagenesis to obtain a more complete picture of the different ligand binding sites within the globular heads of C1q and to begin to identify the regions involved in trimerization.
We are grateful to Dr. G. Arlaud for his gift of peptide P(601–613), to Dr. G. Winter and Dr. R. Kontermann for clones secreting C1q-specific scFv Abs, and to Dr. Ishii for the plasmids encoding the chaperones GroESL and thioredoxin.
This investigation was supported by Grant K23/95 from the Bulgarian National Scientific Foundation, and by Grants ERC-CIPA-CT-92-2215 (to M.S.K.) and TEMPUS JEP 08043/96 (to M.S.K. and I.D.P.). H.-J.H. is a Postdoctoral Research Fellow of the Arthritis and Rheumatism Council of U.K. and a Winthrop Research Fellow at Green College (Oxford, U.K.).
Abbreviations used in this paper: gh, globular head; GST, glutathione S-transferase; HAIgG, heat-aggregated human immunoglobulin G; SC, sodium carbonate buffer.