Abstract
A unique monoclonal Ig λ light chain dimer (protein LOI) was isolated from the serum and urine of a patient with hypocomplementemic membranoproliferative glomerulonephritis. In vitro the λ light chain dimer efficiently activated the alternative pathway of complement (AP). When added to normal human serum, LOI temporarily enhanced AP hemolytic activity, but during a prolonged incubation the hemolytic activity was depleted. Protein LOI was found to bind to factor H, the main regulator molecule of AP. By binding to the short consensus repeat domain 3 of factor H, the dimer LOI blocked one of three interaction sites between H and C3b and thus inhibited the activity of H and induced an uncontrolled activation of the AP. Structural analysis showed that LOI belonged to the Vλ3a subgroup of λ light chains. The variable (V) region of LOI was most closely related to the predicted product of the Vλ3 germline gene Iglv3s2, although it contained several unique residues that in a tertiary homology model structure form an unusual ring of charged residues around a hydrophobic groove in the putative Ag binding site. This site fitted considerably well with a putative binding site in the molecular model of domain 3 of factor H containing a reciprocal ring of charged amino acids around a hydrophobic area. Apparently, functional blocking of factor H by the Ab fragment-like λ light chain dimer had initiated the development of a severe form of membranoproliferative glomerulonephritis. Thus, the λ light chain dimer LOI represents the first described pathogenic miniautoantibody in human disease.
The alternative complement pathway (AP)3 acts as a first-line defense mechanism against a wide range of targets. It contributes to elimination of foreign particles and cells through opsonization and direct lysis (1, 2). AP also serves as a bridge between the innate and acquired immunity by participating in the selection of structures that will be attacked by acquired immunity (3, 4). The initiation of AP activation is not based on any specific target-associated factor. Instead, the AP is continuously autoactivated by hydrolysis of the component C3 at a low rate in human plasma. The hydrolyzed form of C3 (C3(H2O)) has the potential to bind factor B, which, after cleavage to Bb by factor D, will cleave both C3 and C5. The generated C3b becomes covalently bound to the target structures, and the cleavage of C5 will initiate formation of the membrane attack complex (5). Under normal circumstances, activation of the AP is efficiently limited by regulatory proteins in plasma and on host cell surfaces, but not on activator surfaces, i.e., foreign targets.
Factor H (H) is a crucial fluid phase regulator of the AP, as it is an essential cofactor for factor I (C3b inactivator) in the proteolytic inactivation of C3b and C3(H2O) (6, 7). Factor H is also capable of discriminating between activator- and nonactivator-bound C3b molecules (8, 9). High affinity of H for nonactivator-associated C3b restricts activation of the AP by supporting rapid cleavage of C3b by factor I, preventing the binding of factor B to C3b and by dissociating the C3bBb convertase, all mechanisms that lead to efficient down-regulation of the AP (10). Factor H is an elongated molecule and consists of 20 short consensus repeat domains (SCR). Recently, H has been found to interact with C3b via three binding sites. The regions responsible for the binding have been localized to SCR1–4, SCR8–15, and SCR19–20, respectively (11, 12, 13).
It is known that AP dysregulation and vigorous complement (C) activation are associated with glomerulonephritis (14). Many cases of human membranoproliferative glomerulonephritis (MPGN) have been shown to be associated with C3 nephritic factors, autoantibodies that bind to and stabilize the AP C3 convertase C3bBb (14, 15). In pigs, a congenital deficiency of complement factor H has recently been shown to lead to lethal glomerulonephritis (16, 17). In man, a deficiency of factor H has been found in association with glomerulonephritis and/or the hemolytic uremic syndrome (18, 19, 20).
Previously, we reported that a monoclonal λ-chain dimer isolated from the urine of a patient LOI with MPGN activated the AP in a manner different from C3 nephritic factors (21). In the present study our goal was to elucidate the structural properties and mechanism of C activation by the protein LOI to understand its possible role in the pathogenesis of glomerulonephritis. We found that this component interfered with AP regulation and caused uncontrolled AP activation by binding to the SCR3 of factor H and preventing binding of the N-terminal binding site of H to C3b. We also determined the primary structure of the LOI λ-chain and generated a molecular model that revealed unusual features in its tertiary structure, which fitted considerably well with a molecular model of SCR3 of H. We conclude that the protein LOI functioned as a miniautoantibody against a functionally important epitope of factor H. Blocking this site probably initiated a process that resulted in the development of MPGN.
Materials and Methods
Clinical data
The patient LOI was a 57-yr-old Caucasian woman who was admitted to hospital because of renal insufficiency and hemolytic anemia (21). Histopathological and electron microscopic analyses of a percutaneous kidney biopsy revealed both subendothelial and intramembranous dense deposits in the glomerular basement membranes, changes usually seen in type I and type II MPGN, respectively (22). The alternative pathway hemolytic C activity and levels of C3 and factor B were decreased in the patient’s serum. When serum from the patient was mixed with normal human serum, a dose-dependent activation of the alternative pathway of C was observed. The C-activating property in the serum and urine of the patient was associated with a monoclonal Ig λ light chain dimer, previously called λL (21) and now designated LOI. During 6 yr of observation, the patient did not develop any manifestations of malignant plasma cell dyscrasia or amyloidosis.
Protein isolation and characterization
The monoclonal λ-chain dimer LOI was isolated from the patient’s urine by anion exchange chromatography (Mono Q HR5/5 column, Pharmacia Biotech, Uppsala, Sweden). The purity of the protein was established by SDS-PAGE. The V region subgroup was established by ELISA using anti-human Vλ subgroup-specific mAbs as previously described (23). For amino acid sequencing, 10 mg of protein LOI was reduced and pyridylethylated, then initially purified by gel filtration through a Superose 12 column containing 6 M GuHCl, followed by HPLC (ABI model 151, Applied Biosystems, Foster City, CA) using an Aquapore 300A C8 reverse phase column (210 × 4.6 mm) and a 0.1% trifluoroacetic acid to 70% acetonitrile water (v/v) linear gradient (flow rate, 1 ml/min). The single resultant protein was digested with tosyl-l-phenylamine chloromethyl ketone-treated trypsin (Worthington Biochemical, Freehold, NJ), and the tryptic peptides were separated by HPLC as described above. Ten to thirty micrograms of protein or peptide was dissolved in 10% acetic acid, loaded onto polybrene-treated glass fiber disks, and air-dried. The disks were placed in an ABI model 477A gas phase Sequenator, and the resulting phenylthiohydantoin amino acids were identified using an on-line ABI model 120A PTH amino acid analyzer. Assignment of tryptic peptides and alignment of amino acid sequences were based on published data (24). The amino acid sequence of LOI is deposited in SWISS-PROT protein sequence database (accession no. P80748).
Hemolysis assay
After incubating protein LOI (0.05–525 μg/ml; diluted in veronal-buffered saline (VBS), pH 7.35, containing 0.1% gelatin (GVB)), MgEGTA (5 mM), and normal human serum (NHS; 13%) for 5 or 30 min, rabbit RBC (1.5 × 108/ml) were added to the mixture (total volume, 100 μl). After a second incubation (15 min at 37°C), dilution with 900 μl of GVB, and spinning (5 min at 2100 × g), the release of hemoglobin was determined from the supernatants by A412. Background lysis was measured from the mixtures where MgEGTA had been replaced by EDTA, and total lysis was obtained using H2O instead of GVB. All experiments were performed in duplicate, and the data in the figures represent the mean values.
Fluorometric assay
The fluorometric assay mixture contained 40 μM 8-anilino-1-naphtalene sulfonate (ANS), 1.05 μM C3, 0.167 μM factor H, 1.54 μM LOI or HEPES-buffered saline (20 mM HEPES in 140 mM NaCl, pH 7.4), 0.54 μM factor B, 0.01 μM factor D, and 0.044 μM factor I in 1.3 ml of HEPES-buffered saline containing 20 mM MgEGTA. In the control assay the amount of factor H used was either 0.084 μM (50% H) or 0.167 μM (100% H; Fig. 2⇓B). C3b formation and inactivation were followed at 37°C by a spectrofluorometer (8000-C, SLM Instruments, Urbana, IL) with excitation and emission wavelengths of 386 and 472 nm, respectively. C3 and factors B, D, I, and H were purified from human plasma by methods previously described (25). Purity of the proteins was examined by SDS-PAGE and was >90%. No binding of ANS to protein LOI was observed.
Ligand blotting
Radiolabeling of λ-chain LOI (and factor H) was performed as described previously (12). To perform the ligand blot, the recombinant proteins were electrotransferred to a nitrocellulose membrane (0.45 μm; Schleicher & Schuell, Dassel, Germany) after SDS-PAGE under nonreducing conditions. After blocking the membrane (17 h at 22°C in 3% BSA-PBS) [125I]LOI (1.8 μg/ml in 3% BSA-PBS containing 0.02% NaN3) was incubated with the membrane for 110 h at 4°C. Bound radioactivity was detected with a bioimaging analyzer (BAS-1500) using a BAS-IIIs imaging plate (Fuji Photo, Tokyo, Japan).
Radioimmunoassay
RIA was performed using plastic microtiter plates (Falcon, Becton Dickinson, Oxnard, CA) coated with LOI (10 μg/ml) for 17 h at 22°C. After washing the wells (three times using VBS-T, i.e., VBS containing 0.1% Tween-20) increasing amounts of either factor H or protein LOI and a constant amount of [125I]H in GVB-T were added. After a 60-min incubation at 22°C the wells were washed for four times using VBS-T, and radioactivity in each well was counted.
Surface plasmon resonance assay
A real-time monitored surface plasmon resonance assay was performed using a Biacore 2000 instrument (Biacore, Uppsala, Sweden). Flow cells of the carboxylated dextran chip (Sensor Chip CM5 from Pharmacia Biosensor, Uppsala, Sweden) were used for coupling with C3b via an amine-coupling procedure or for a control flow cell (same procedure without coupled protein). C3b was prepared as described previously (26). After activating the flow cell with a solution containing N-hydroxysuccinimide and N-ethyl-N′(dimethylaminopropyl)carbodiimide, 30 μl of C3b (200 μg/ml in 20 mM acetate, pH 4.5) was injected (flow rate, 5 μl/min) to reach an appropriate level of bound C3b (14,000 resonance units). Afterward the flow cells were inactivated by a standard ethanolamine-HCl injection. The binding assay was performed using 1/3 VBS and a flow rate of 5 μl/min throughout, and all the samples used were dialyzed against 1/3 VBS. For the binding assays 10 μl of a solution containing 120 μg/ml H (or H recombinant fragments SCR1–6 or SCR8–20) and either λ-chain LOI (360 μg/ml) or a control λ-chain ÅHB (360 μg/ml; the same λ subclass as LOI) or buffer (1/3 VBS) was injected after an equilibration (15 min at 22°C). The bound material was detached from the chip surface using 3 M NaCl in acetate buffer (pH 4.2).
Baculovirus expression
Cloning, expression, and purification of the various recombinant proteins have been described previously (11, 27, 28). Several fragments (SCR1–7, SCR1–6, SCR1–5, SCR1–4, SCR1–3, SCR1–2, SCR1, SCR8–20, SCR6–20, and HΔ6–10 (containing SCR1–5 and SCR11–20)) of human factor H were expressed. The recombinant proteins (0.5 μg) were separated by SDS-PAGE using either 12 or 15% gel under nonreducing conditions and electrotransferred to a nitrocellulose membrane. The binding of [125I]LOI to the fragments was analyzed by ligand blotting (described above). The recombinant fragments were examined by SDS-PAGE and were >98% pure.
Molecular modeling
Computer-aided molecular modeling was performed with the Insight II program package (Biosym Technologies, San Diego, CA) using an Iris Indigo XZ 4000 work station (Silicon Graphics, Mountain View, CA). The λ light chain LOI was modeled as follows. After screening of all Ig structures available in the Brookhaven Protein Data Bank (PDB, release 68, Brookhaven National Laboratory, Upton, NY), the most homologous crystallographically determined structure of a human Ig λ light chain (PDB access code 8FAB) (29) was chosen as the template for a single variable domain of protein LOI. The Biopolymer module of Insight II was used to build a preliminary model of LOI. Two individual variable domains were superimposed onto a highly homologous λ light chain dimer structure (3 MCG) (30), to achieve a preliminary model structure for the LOI light chain variable (V) domain dimer. The preliminary model structure was soaked in a waterbox of 4274 water molecules (dimensions of the box were 64 × 55 × 47 Å) to achieve an ∼6-Å-thick layer of water around the whole protein. The energy minimizations (using both the steepest descents and the conjugate gradient algorithms) and dynamics simulation were performed as described previously (31). The target was simulated for 10 ps at 100 K and 20.0 ps at 300 K using throughout a 1.0-fs time step. Five structures were selected from the last 5 ps (at 300 K) according to their low potential energy and were subjected to energy minimization in their waterboxes using the conjugate gradient algorithm until the maximum derivative was below 4.2 J/Å. One of the five structures is shown in Fig. 7⇓ (coordinates are deposited in the Brookhaven Protein Data Bank; entry code 2LOI).
A molecular model of the factor H domain SCR3 was constructed using the published nuclear magnetic resonance structure of H SCR15 as a template (pdb entry 1HFH) (32). As the number of residues differ in SCR3 and SCR15, two loops were needed to complete the preliminary model structure. The best available loop structures were searched from protein structures deposited in PDB using anchors of five residues. The criteria in choosing the most suitable loops were amino acid sequence, distance between the first Cα atoms of the anchors, root mean square deviation (RMSd), and estimation of suitability to the overall tertiary structure. Model loop 1 (residues 28–32) was from 2AZA, and model loop 2 (residues 54–57) was from 1BBP. The preliminary model structure of SCR3 was subjected to energy minimizations using steepest descents and conjugate gradient algorithms with gradually relaxed model until the maximum derivative was <4.2 J/Å.
To determine the residues on SCR3 of H responsible for binding to the protein LOI, a manual docking procedure was performed. Using surface chart of charged and hydrophobic areas on the whole SCR3 and on the putative Ag binding site of LOI, the surface area of SCR3 that fits best the surface characteristics of LOI was chosen for docking as displayed in Fig. 7⇓, B and C.
Results
Activation of complement by λ-chain dimer LOI
Using a hemolysis assay we found that the molecule LOI had a dual effect on AP activation. When the protein was incubated for 5 min in NHS, the hemolytic activity of serum AP was increased, while after a 30-min incubation serum AP hemolytic activity was markedly diminished (Fig. 1⇓). We have previously shown that the λ-chain dimer LOI activated both C3 and factor B when added to NHS (21).
Effect of protein LOI on the hemolytic activity of serum. Hemolysis of rabbit RBC by the alternative pathway of complement was examined after incubation of NHS with either λ-chain LOI or control λ light chain (both at 0.05–525 μg/ml) for 5 min (A) or 30 min (B). The hemolytic activity of human serum was enhanced when LOI had been present for 5 min, but was inhibited or abolished when protein LOI had been present for 30 min. The control λ light chain did not significantly alter the AP hemolytic activity. The y-axis (percent hemolysis) describes the relative hemolytic activity of the serum when hemolysis without any added light chains is 100%.
To analyze in more detail the effect of protein LOI on C3b formation and inactivation we next used an isolated mixture of AP proteins (C3, B, D, H, I) and a fluorometric assay. In this assay C3 conversion to C3b began immediately after all the required AP proteins (C3, B, D) were present (Fig. 2⇓). The newly generated C3b started simultaneously to become inactivated to iC3b by factors H and I, seen as a rapid decline in the C3b peak. However, in the presence of protein LOI (76.9 μg/ml) the AP C3 conversion was strikingly enhanced, and the inactivation occurred considerably more slowly (Fig. 2⇓A). The effect of LOI on the C3 conversion was comparable to the effect of reducing the amount of factor H by 50% (Fig. 2⇓B), leading to both enhanced formation of C3b (lack of H competing with B on binding to C3b) and slower inactivation of the formed C3b by factor I (lack of H as a cofactor). Together with the finding that nearly all C3 had become converted to C3a and C3b in the patient’s serum, these in vitro results indicated that the λ-chain LOI caused C depletion by time-dependent overconsumption.
Enhancement of fluid phase AP C3 conversion by protein LOI. Cleavage of fluid phase C3 to C3b by the AP C3 convertase C3bBb was assayed fluorometrically using ANS as a probe. Conversion of C3 to C3b and inactivation of C3b to iC3b can be followed as a function of time, as ANS gives fluorescence only when bound to C3b. Purified proteins of the AP (C3, B, D, H, I) and Mg2+ ions were added sequentially as indicated (for details, see Materials and Methods). A, The relative amount of C3b is seen either in the presence (upper curve) or absence (lower curve) of λ-chain LOI (76.9 μg/ml) in the mixture. B, The amount of C3b in the presence of 50% H (0.083 μM) or 100% H (0.167 μM) is shown as a control for the effect of LOI.
λ-Chain dimer LOI binds to factor H
We next investigated the mechanism by which protein LOI interfered with the inactivation of C3b. We have previously shown that this molecule did not act like a C3 nephritic factor by binding to and stabilizing the C3bBb convertase. Instead, we found that LOI bound to factor H. In a ligand blotting assay using human serum proteins as targets radiolabeled LOI bound only to IgG and a molecule with a mobility similar to that of factor H (data not shown). The target molecule for protein LOI was confirmed as factor H in a RIA in which [125I]H was shown to interact with solid phase protein LOI (Fig. 3⇓A). This interaction could be blocked by competition with unlabeled H or λ-chain LOI (Fig. 3⇓B).
Binding of λ-chain LOI to factor H. A, RIA where [125I]H binding to solid phase bound LOI was measured. Factor H bound to λ-chain LOI but not to the three control λ3 light chains (□, ○, ▵). B, RIA where [125I]H binding to solid phase LOI was measured in the presence of increasing amounts of either unlabeled H or protein LOI. Both H and LOI can competitively block the interaction between [125I]H and λ-chain LOI.
Binding site of LOI on factor H
Using several recombinant fragments of factor H we mapped the binding site involved in its interaction with LOI. Factor H is an elongated molecule consisting of 20 SCRs, each containing about 60 residues (33). Protein LOI was found to bind to all constructs containing SCR3 (i.e., SCR1–7, SCR1–6, SCR1–5, SCR1–4, and SCR1–3 and a construct HΔ6–10 containing SCR1–5 and SCR11–20), but not to the other constructs tested (SCR1–2, SCR1, and SCR6–20; Fig. 4⇓). Thus, the primary binding site of LOI was localized to the SCR3 domain of factor H.
Localization of λ-chain LOI binding site on factor H by ligand blotting. Recombinant fragments of factor H were expressed using the baculovirus system. The purified fragments (0.5 μg) were electrotransferred from an SDS-PAGE gel onto a nitrocellulose sheet that was incubated with 125I-labeled protein LOI. The SCR domains of each fragment are shown above the corresponding lane. Protein LOI bound to fragments SCR1–7, SCR1–6, SCR1–5, SCR1–4, and SCR1–3 and HΔ6–10 (containing SCR1–5 and SCR11–20), but not to SCR1–2, SCR1, or SCR6–20. Binding of λ-chain LOI to the constructs seemed to be slightly enhanced when both SCR3 and SCR6 were present (SCR1–6 and SCR1–7).
Restriction of factor H function by LOI
Although we showed that λ-chain LOI caused AP activation and bound to H, the results did not necessarily prove that the two phenomena were linked. It remained possible that protein LOI caused AP activation by restricting H binding to C3b or by disturbing H cofactor function without inhibiting binding of H to C3b. To test whether LOI restricted H binding to C3b, we analyzed the binding of H to C3b in the presence of protein LOI or a control λ light chain dimer using a surface plasmon resonance technique. In contrast to the control light chain, LOI inhibited efficiently the binding of H to C3b (Fig. 5⇓A). As factor H contains three binding sites for C3b, we next analyzed which of the binding sites was blocked by LOI. Using two recombinant fragments of H (SCR1–6 and SCR8–20), it was found that the inhibition of H binding to C3b by LOI was due to blockage of the N-terminal binding site within the fragment SCR1–6 (Fig. 5⇓B). On the contrary, the middle or C-terminal binding sites were not influenced by LOI, as it had no effect on binding of H fragment SCR8–20 to C3b (Fig. 5⇓C).
Binding of factor H to C3b in the presence of λ-chains. Binding of H (A) or two recombinant fragments of H, SCR1–6 (B) or SCR8–20 (C), to C3b was analyzed in the absence or the presence of λ-chain LOI or a control λ light chain dimer in Biacore instrument. The flow cells of a CM5 sensor chip were either coupled with purified C3b or used as an uncoupled flow cell (blank channel). Light chain dimers were equilibrated with H before injecting them into the flow cells. Binding of H to C3b is seen as an increase in the surface plasmon resonance units as a function of time until the injection is completed and the dissociation begins. Signal obtained from the blank channel was subtracted from the C3b binding signal, explaining the short sharp peaks in the beginning and the end of some of the injections and the values below zero of some of the curves. The λ-chains were not found to bind to C3b (data not shown).
Structure of λ-chain dimer LOI
Serologic analyses with anti-human Vλ subgroup-specific mAbs (23) revealed that protein LOI was a member of the Vλ3a subgroup. SDS-PAGE analysis under reducing and nonreducing conditions showed that LOI was a disulfide-linked light chain dimer with an apparent Mr of 45.3 kDa.
Amino acid sequence analysis revealed that the primary structure of the V domain of protein LOI was most highly homologous (∼93%) to the predicted protein sequence encoded by the λ3 germline gene Iglv3s2 (Fig. 6⇓) (34). By comparing with the product of the germline gene and the published sequences of other light chains, protein LOI contained several unique or rarely occurring residues. Most of these amino acids are negatively charged and would account for the exceptionally low isoelectric point of the LOI (pKa = 5.6).
Comparison of the primary structure of λ-chain LOI with the deduced amino acid sequence of the most homologous Vλ3 germline gene Iglv3s2. CDR, complementarity determining region.
A tertiary model structure of the V domains of protein LOI was constructed by computer-aided homology modeling, using energy minimization and a total of 30 ps of molecular dynamics simulation. The RMSd of the atoms during the dynamics simulation reached a plateau level of about 2.1 Å after 15-ps simulation at 300 K (data not shown). In the tertiary model structures the unique and rare residues of protein LOI were located mainly in an area that corresponded to the Ag-binding Fab region (Fig. 7⇓). This area showed a distinct ring-like distribution of negatively charged residues.
Molecular modeling of protein LOI and docking analysis. Two variable domains of the Ig λ-chain dimer LOI were constructed in silico, and the structure was energy minimized and subjected to dynamics simulations at 100 and 300 K. A plateau of RMSd was reached after about a 15-ps simulation at 300 K. Solvent accessible surface of one of the structures gained after the RMSd plateau was reached (Brookhaven PDB entry 2LOI) is shown in a lateral (A) and a top (B) view. The negatively and positively charged solvent accessible surfaces are shown in red and blue, respectively. To see the three-dimensional structure of the domain where LOI binds on H we constructed a model of the SCR3 of H. Using manual docking, the most probable binding site of LOI was elucidated, and the docking result is shown when the proteins are apart (B, a top view) or interacting (C, a lateral view). Labels on LOI indicate either some of the unique and rarely occurring residues (Glu15, Asp25, Glu29, Phe48, Arg50, and Glu94; see Fig. 6⇑) or the suggested interacting residues (Phe48, Trp57, Arg50, Asp51, and Glu94). Labels of one of the two chains of LOI are marked with a b. Labels on H domain SCR3 indicate most of the suggested interacting amino acids.
Structural analysis of a putative interaction site
To analyze binding interaction between λ-chain LOI and SCR3 of H, a tertiary molecular model of SCR3 was constructed. Subsequently, models of protein LOI and SCR3 were subjected to manual docking using a surface chart of charged and hydrophobic areas on the whole SCR3 and on the putative Ag binding site of LOI. The ring-like distribution of the negatively charged residues around a more hydrophobic groove on protein LOI fitted to dock only to one area of the model of SCR3. The residues involved in the interaction of the docked models were either charged or hydrophobic. The main interacting charged residues between LOI and SCR3 of H were (residues of LOI shown in italic, one of the two chains of LOI is marked with a b): Arg50→Glu60, Asp51→Lys61, Glu94b →Lys63, Arg50b→Glu22/Asp24, Asp51b/Glu94→Lys4, and Glu29b/Asp49b→Lys15. The main residues in the hydrophobic groove of LOI were Phe48, Trp89, and Tyr47. The reciprocal hydrophobic amino acids of SCR3 of H were Phe56, Trp57, Leu6, and Ala10. The surface structures of the putative binding sites are displayed in Fig. 7⇑, B and C.
Discussion
Published reports of Ag-binding miniantibodies include hybridoma cell lines producing light chain dimers, experimentally expressed light chains with Ag-binding properties, and engineered Ab fragments (35, 36, 37, 38, 39). None of these represents naturally occurring autoantibodies in association with human disease. As some Ig light chains have been found to bind Ag and usually the same Ag as the original native Ig molecule, it is highly likely that the light chains devoid of heavy chains bind to the Ag by using the same structures, i.e., complementarity determining regions within the light chain sequences. The same has been found to occur with isolated heavy chains, heavy chain fragments, or single-chain equivalents of the Ig Ag binding region (i.e., Fv chains) (40, 41). Even small changes within the Ag binding region may alter the binding affinity and specificity of an Ab, and the variability within different Abs is likely to be exceptionally broad. Thus, it is not surprising that protein LOI as a λ light chain dimer had the potential to bind to an Ag.
The results of our in vitro studies were concordant with the in vivo findings. Our experimental demonstration that protein LOI inactivated AP by overconsumption would explain the markedly reduced serum AP hemolytic activity, low levels of C3 and factor B, and extensive deposition of complement activation products in the renal glomeruli (21). The disease mediated by functional blocking of H by the λ-chain LOI is comparable to that found in glomerulonephritis associated with human C3 nephritic factor and in the lethal glomerulonephritis associated with porcine congenital factor H deficiency (14, 16, 17, 19). The mechanism by which AP activation leads to glomerulonephritis is not yet known. The glomeruli are probably susceptible to C attack because the glomerular basement membrane (BM) is readily accessible to the activation products of complement proteins, and it is not protected by the cell surface-expressed complement control proteins decay-accelerating factor (CD55), membrane cofactor protein (CD46), protectin (CD59), or CR1 (CD35). Under normal circumstances, restriction of AP amplification on the glomerular BM is probably dependent on the interaction between factor H and the anionic BM structures (9). As patient LOI had in addition hemolytic anemia, it is also possible that other bystander structures may be attacked by the inappropriately controlled C system.
We propose that LOI served as a miniautoantibody directed against factor H and was responsible for AP activation. The λ-chain LOI binds to the SCR3 of H and restricts binding of H to C3b resulting in uncontrolled AP activation. Although LOI does not act directly on the C3bBb complex, it can have an indirect enhancement effect on the fluid phase AP convertases as it restricts down-regulation of the alternative pathway amplification cascade by H. To date, the detailed nature of interactions between H and C3b and especially the recognition mechanisms of the AP are not thoroughly understood. Recently, a total of three binding sites on H for C3b have been reported (11, 12, 13). The most N-terminal site for C3b is located in SCR1–4 (42, 43). The cofactor activity and decay-accelerating activity of H appear to remain within the same domains (28, 44). We have now shown that LOI blocks H binding to C3b by binding to SCR3, thus preventing the interaction of the binding site in factor H SCR1–4 with C3b.
Based on x-ray crystallographic data on the light chain dimer MCG (30), we have constructed a tertiary model structure of protein LOI. The model showed that the Ag binding site of LOI included a ring of charged residues and a rare central doublet of hydrophobic Phe48 residues. To analyze structure of the interacting molecular surfaces of LOI and factor H, we constructed also a molecular model of SCR3 of H. In a manual docking analysis only one area on SCR3 of H was found to fit considerably well with the putative Ag binding site of LOI. On the basis of this docking result it seems that the binding site on SCR3 consists of negatively and positively charged residues and a hydrophobic ridge. The fitting between LOI and SCR3 of H was good, but not optimal, as some negatively charged groups on SCR3 did not have a counterpart on LOI (e.g. Asp25). The missing positively charged counterparts could be located on the SCRs of H next to SCR3, but as the angles between SCRs are not rigid, it was considered not to be informative to model the adjacent SCRs of H. However, the model showed that the residues of LOI that were unique or only rarely occurring in other λ light chains were clustered to the area corresponding to the Ag binding site of Fab. In addition, the docking showed that most of these residues (Phe48, Arg50, Glu94, Glu29b, Phe48b, Arg50b, and Glu94b) are likely to be involved in the binding interaction with the SCR3 of H.
The results of the structural analysis showed that the fitting of SCR3 of H to LOI binding site was not optimal. This suits to the possibility that the λ-chain LOI was not primarily emerged to bind to H by an Ag-driven process, but that the binding of LOI to H was a harmful bystander phenomenon. However, it remains possible that the heavy chain corresponding to the λ-chain LOI would have been completed the binding site of the native Ig to fit even better to the SCR3 of H. In that case the existence of only the light chain may explain the apparent evasion of apoptosis by the B cell line producing LOI, although the product was a part of an autoantibody.
In summary, we have described that the λ-chain LOI causes dysregulation of the alternative complement pathway leading to the development of MPGN. The present study is the first detailed description of a pathogenic miniautoantibody in human disease. By binding to complement factor H and blocking its activity the LOI dimer causes vigorous AP activation and subsequent overconsumption. Apparently, factor H dysfunction and consequent activation of AP have a critical role in the development of MPGN type II.
Acknowledgments
We thank J. Hellwage and S. Kühn from the Bernhard Nocht Institute and A. Sharma from the University of Texas Health Science Center for providing purified recombinant fragments of factor H. We thank D. Wolfenbarger for the serologic and C. M. Murphy for the sequence analyses of protein LOI. We thank Mrs. M. Ahonen for excellent technical assistance.
Footnotes
-
↵1 This work was supported by the Academy of Finland, the Sigrid Juselius Foundation, the Foundation of the 350th Anniversary of the University of Helsinki, the Finnish Kidney Foundation, and U.S. Public Health Service Research Grant CA10056 from the National Cancer Institute. A.S. is an American Cancer Society Clinical Research Professor.
-
↵2 Address correspondence and reprint requests to Dr. T. Sakari Jokiranta, Haartman Institute/HD Diagnostics, Haartmaninkatu 3, FIN-00290 Helsinki, Finland. E-mail address: sakari.jokiranta{at}helsinki.fi
-
3 Abbreviations used in this paper: AP, alternative pathway of complement; H, factor H; SCR, short consensus repeat domain; MPGN, membranoproliferative glomerulonephritis; LOI, λ light chain dimer purified from patient LOI; VBS, veronal-buffered saline; NHS, normal human serum; GVB, veronal-buffered saline containing 0.1% gelatin; ANS, 8-anilino-1-naphtalene sulfonate; BM, basement membrane; RMSd, root mean square deviation.
- Received July 20, 1998.
- Accepted July 30, 1999.
- Copyright © 1999 by The American Association of Immunologists