Disease-associated mutations in human mannose-binding lectin compromise oligomerization and activity of the final protein.

Deficiency of human mannose-binding lectin (MBL) caused by mutations in the coding part of the MBL2 gene is associated with increased risk and severity of infections and autoimmunity. To study the biological consequences of MBL mutations, we expressed wild type MBL and mutated MBL in Chinese hamster ovary cells. The normal MBL cDNA (WT MBL-A) was cloned, and the three known natural and two artificial variants were expressed in Chinese hamster ovary cells. When analyzed, WT MBL-A formed covalently linked higher oligomers with a molecular mass of about 300-450 kDa, corresponding to 12-18 single chains or 4-6 structural units. By contrast, all MBL variants formed a dominant band of about 50 kDa, with increasingly weaker bands at 75, 100, and 125 kDa corresponding to two, three, four, and five chains, respectively. In contrast to WT MBL-A, variant MBL formed noncovalent oligomers containing up to six chains (two structural units). MBL variants bound ligands with a markedly reduced capacity compared with WT MBL-A. Mutations in the collagenous region of human MBL compromise assembly of higher order oligomers, resulting in reduced ligand binding capacity and thus reduced capability to activate complement.

Mannose-binding lectin (MBL) 1 has been shown to be an important component of innate immunity and is a central recognition molecule of the lectin pathway of complement (for a recent review, see Ref. 1). MBL binds to an array of carbohydrate structures on surfaces of bacteria (2)(3)(4), yeast, viruses (5,6), and parasitic protozoa (7,8). MBL functions as an opsonin (9), and the biological effect is mediated by direct killing via complement (10) through the lytic membrane attack complex or by promoting phagocytosis either by the MBL lectin pathway of complement or by direct binding to one or more cell surface receptors (11). The lectin pathway comprises at least three MBL-associated serine proteases (MASPs), namely MASP-1 (12), MASP-2 (13), and MASP-3 (14). Furthermore, the functional MBL-MASP complex contains a small MBL-associated protein (sMAP), also named MAp19, with no serine protease activity (15,16). MASP-2 is a homologue of C1s of the classical complement pathway because it activates C4 and C2 (13). When MBL associated with MASP-2 binds to sugar groups on the surface of microbes, the MBL-MASP2 proenzyme is activated and cleaves sequentially C4 and C2, thereby creating the C4b2a complex, a potent C3 convertase. The MBL-MASP-1 complex is suggested to activate C3 directly (12). Whether both MASP-1 and MASP-2 are bound on the same MBL molecule is still unclear. Moreover, the biological role of sMAP, as well as the substrate for the recently discovered MASP-3, remains unclear at this moment (for a recent review on MASPs, see Ref. 17).
MBL is a complex of six sets of homotrimers of a single polypeptide chain containing 228 amino acids (18 -21). This polypeptide consists of four domains ( Fig. 1): 1) a 20-amino acid N-terminal cysteine-rich domain involved in formation of intraand intersubunit disulfide bonds, 2) a collagen-like domain consisting of 18 -20 tandem repeats of Gly-Xaa-Yaa, 3) an ␣-helical coiled-coil neck region, and 4) A carbohydrate recognition domain capable of binding to a wide variety of carbohydrate arrays on the surface of microorganisms (3). Three polypeptides form a structural unit or subunit containing a triple helix at their collagen-like domain. Six of these units combine by interunit disulfide bonds to form the biologically active bouquet-like MBL protein (for a recent review on MBL structure, see Ref. 22). Three different genetic polymorphisms in exon 1 of the human MBL2 gene (MBL1 is a pseudogene (23)) independently lead to reduced serum concentrations of MBL. Two interrupt the tandem repeat Gly-Xaa-Yaa in the first (Gly) position, and the third introduces a cysteine residue in the second position. The designation of the MBL variant alleles is B, C, and D, whereas the normal allele is termed A. MBL-B has a G 3 A mutation in codon 54 (24), which results in a Gly 3 Asp substitution in the fifth Gly-Xaa-Yaa repeat. MBL-C has a G 3 A mutation in codon 57 (25), which translates into a Gly 3 Asp substitution of the sixth Gly-Xaa-Yaa repeat. The third mutation is MBL-D, a C 3 T mutation in codon 52 (25) that results in the introduction of a cysteine instead of an arginine in the protein (26). The presence of these mutations leads to markedly reduced MBL protein levels in the blood (27,28). The B and D alleles are seen in Eurasian and indigenous American populations with frequencies ranging from 0.1 to 0.5 and from 0.0 to 0.1, respectively (29 -31). The C allele is found most frequently in sub-Saharan African popula-tions with a frequency ranging from 0.07 to 0.3 (29 -31). The presence of these alleles is associated with increased risk of infections during childhood, particularly during the vulnerable period of infancy ranging from 6 to 18 months of age (32)(33)(34), in immunocompromised patients (35,36) and is a risk factor for critically ill patients to develop sepsis (37). Moreover, MBL variant alleles are associated with disease progression in concomitant diseases such as chronic granulomatous disease and cystic fibrosis (38,39), Additionally, the importance of MBL deficiency in autoimmunity has been emphasized in diseases like systemic lupus erythematosus and rheumatoid arthritis (40 -43).
In order to define the molecular mechanisms underlying the disease associations accompanying MBL variant alleles in more detail, we constructed and expressed recombinant wild type as well as variant MBL forms and investigated their structural and functional characteristics.

MATERIALS AND METHODS
Restriction enzymes and reverse transcriptase were from Amersham Biosciences. Taq polymerase was from Applied Biosystems. Cell culture utensils were from TPP (Trasadingen, Switzerland), except for triple bottom flasks that were from Nalge Nunc. Mannan, mannose, N-acetylglucosamine, trypsin (1:250), Geneticin (G418), hypoxanthine/thymidine media supplement (HT-supplement), L-glutamine solution, penicillin-streptomycin solution, methotrexate (MTX), dialyzed fetal bovine serum, RPMI 1640, and Iscove's modified Dulbecco's medium were all from Sigma. Large scale plasmid DNA isolation was performed using the Qiagen EndoFree Plasmid Maxi Kit. Small scale DNA preparations were performed using the Quantum Prep Plasmid Miniprep Kit (Bio-Rad). General methods of molecular biology were applied as described in Ref. 44.
Cloning of the MBL cDNA-mRNA was isolated from the human hepatocellular carcinoma cell line HepG2 (obtained from ATCC (Manassas, VA) and having the MBL genotype: HYPA/LYPB) using the Dynabeads mRNA DIRECT Kit (Dynal). cDNA synthesis was done using murine leukemia virus reverse transcriptase (Applied Biosystems, Foster City, CA). A PCR product was generated using Pfu DNA polymerase (Stratagene, La Jolla, CA) and 5Ј-primer (GAGATTAAC-CTTCCCTGAGT) and 3Ј-primer (GAGGGCCTGAGTGATATGAC) and cloned using the PCR-Script Amp Cloning Kit, Stratagene (La Jolla, CA). Verification of correct DNA sequence was performed on Applied Biosystems sequencing equipment. The expected DNA sequence was obtained except for the silent mutation in codon 136: AAC Asn 3 AAT Asn . This construct was used for site-directed mutagenesis.
Transfection and MBL Expression-The MBL gene was PCR-amplified out of the pPCR-Script Amp vector using the following primers: 5Ј-primer, GTCTCTTCC2ATGTCCCTGTTTCCATCAC; 3Ј-primer, TGCTCTTCC2AAGTCAGATAGGGAACTCACAGA (Eam1104I recognition sequence shown underlined, and cut site indicated with an arrow). The PCR product was digested with Eam1104I and cloned into the pDual vector (Stratagene). Attempts were made to achieve expression in Escherichia coli BL21 as well as transient expression in the human HUH-7 hepatoma cells (JCRB Cell Bank) and COS-7 (ATCC) but failed or gave low expression levels. The MBL gene was moved from the pCR-Script Amp vector to pBK-CMV (Stratagene) using NotI and XhoI and from pBK-CMV to the dicistronic vector pED using SmaI and PstI. The pEDdC vector was a kind gift from the Genetics Institute (Cambridge, MA) (45,46). This vector carries a cloning sequence for insertion of the target gene followed by the selectable and amplifiable marker gene dehydroxyfolate reductase (dhfr). The cell line used for recombinant expression of human MBL was Chinese hamster ovary (CHO) DG 44. This CHO clone is a double deletion mutant that contains no copies of the hamster dhfr gene and was a kind gift from professor Lawrence Chasin (Columbia University, New York) (47). Untransfected cells were cultivated in Iscove's modified Dulbecco's medium supplemented by 10% dialyzed fetal bovine serum, 100 units/ml penicillin, 0.1 mg/ml streptomycin, 2 mM L-glutamine, 10 mM hypoxanthine, and 1.6 mM thymidine (designated hereafter as complete Iscove's modified Dulbecco's medium) in a 37°C humidified atmosphere containing 5% CO 2 . Cells were passaged using 0.05% trypsin in PBS. Stable transfections were performed using the LipofectAMINE PLUS reagent kit from Invitrogen. Transfection was done by seeding 8 ϫ 10 5 cells in 6-cm Petri dishes on day 0. On day 1, cell medium was replaced and transfected according to the manufacturer's protocol, except that 24 l of Lipo-fectAMINE, 0.2 g of pSV 2 NEO (Clontech, Palo Alto, CA), and 20 g of the pED-MBL vector were used. On day 3, cells were moved to a T25 flask, and on day 5, cells were moved to complete Iscove's modified Eagle's medium containing 0.5 mg/ml G418 and omitting hypoxanthine and thymidine. Cells were moved to larger bottles when dense. When a G418-resistant pool of clones was obtained (after 10 -14 days), selection and gene amplification with MTX was initiated by cultivating cells in complete Iscove's modified Eagle's medium containing 50 nM MTX and omitting hypoxanthine and thymidine. After cells had regained normal growth rate and morphology, the concentration of MTX was increased to 200 nM, 4 M, 20 M, and finally 80 M. No significant amplification of MBL secretion into medium was seen at levels of MTX above 200 nM.
Western Immunoblotting-SDS-PAGE was performed using 3-8% NuPAGE Tris acetate gels and Tris acetate running buffer both from Novex (San Diego, CA). Western blotting was done using the XCell II Mini-Cell blot apparatus and NuPAGE transfer buffer, both from Novex, and Hybond ECL nitrocellulose membranes from Amersham Biosciences (Hørsholm, Denmark). SuperSignal West Femto Maximum Sensitivity Substrate (Pierce) was used for development of immunoblots utilizing horseradish peroxidase-conjugated antibodies or horseradish peroxidase-conjugated streptavidin. As an M r standard, the Precision prestained protein standard (Bio-Rad) was used.
Quantization of MBL by ELISA-To determine expression levels of MBL in cell growth medium, in elutes from gel filtration experiments, and in fractions from SDG centrifugation, ELISA was performed as previously described (28). In brief, microtiter wells were coated overnight with a capture antibody at 4°C (1 mg/liter anti-human MBL mouse monoclonal antibody HYB 131-01 or HYB 131-11; State Serum Institute, Copenhagen, Denmark) in PBS and washed five times with a buffer of 10 mM Tris-HCl, 150 mM NaCl, 0.05% Tween 20, pH 7.4. All subsequent washes were performed with this buffer. Cell growth supernatant, eluates from gel filtration, or fractions from SDG centrifugation and, as a reference, different dilutions of a serum pool diluted in the above buffer containing 10 mM EDTA were incubated at room temperature for 3 h. Wells were washed and incubated overnight at 4°C with a mouse monoclonal biotin-labeled anti human MBL antibody (biotinylated HYB 131-01, 0.1 ng/ml, State Serum Institute). Wells were rinsed and incubated for 2 h at room temperature with a 1:2,000 dilution of a streptavidin-horseradish peroxidase conjugate (RPN 1231; Amersham Biosciences) in the above buffer, developed using OPD tablets from Dako (Glostrup, Denmark), according to the manufacturer's instructions, and the absorbance was read at 490 nm.
MBL-mediated Complement Activation-Assay for complement activation was performed by coating microtiter plates (MaxiSorp; Nalge Nunc) overnight at 4°C with mannan (0.1 g/liter) in a buffer of 15 mM Na 2 CO 3 , 35 mM NaHCO 3 , pH 9.6. Wells were washed five times in a buffer of 0.4 mM sodium barbital, 0.15 M NaCl, 2.6 mM CaCl 2 , 2.12 mM MgCl 2 , 0.05% Tween 20, pH 7.4. All subsequent incubations and washings were done with this buffer. Wells were incubated overnight at 37°C with cell growth medium containing recombinant MBL. In control experiments, MBL binding to mannan was inhibited by the addition of 10 mM mannose, 10 mM N-acetylglucosamine, or 10 mM EDTA. Wells were incubated 1 h at 37°C with serum from an MBL-deficient person diluted 1:400 (as a source of complement components, MBL genotype HYPD/HYPD (i.e. functionally defective MBL-D variant) (48)), rinsed, and incubated for 45 min at 37°C with a rabbit anti-human C4 antibody (0.3 ng/ml) (Dako). Wells were rinsed three times and incubated for 45 min at 37°C with a 1:2,000 dilution of a donkey anti-rabbit Ig horseradish peroxidase-linked F(abЈ) 2 fragment, rinsed, and developed as described above.
Quantization of MBL-mediated Complement Deposition Versus MBL Binding-Microtiter plates (MaxiSorp; Nalge Nunc) were coated with an MBL binding antibody (NImoAb001, 1.2 g/ml, a kind gift from NatImmune, in PBS) and incubated overnight at 4°C. Plates were washed three times in TBS-T plus calcium (10 mM Tris-HCl, 150 mM NaCl, 0.05% Tween 20, and 10 mM CaCl 2 ) and blocked for 2 h in TBS-T plus calcium containing 1 mg/ml human serum albumin (TBS-T plus calcium and HSA; HSA was from the State Serum Institute). Culture supernatants were diluted 1 ϩ 3, 1 ϩ 7, 1 ϩ 15, 1 ϩ 31, 1 ϩ 63, and 1 ϩ 119 in the above buffer and added to microtiter wells in triplicates. Plates were incubated overnight at 4°C. and washed three times in TBS-T plus calcium. Human purified recombinant MASP-2 (activated during purification) was added to a final concentration of 0.5 g/ml (a kind gift from NatImmune) in TBS-T plus calcium and HSA and incubated overnight at 4°C. Plates were washed and divided into two. One set of plates (A) was developed by measuring MBL binding, and the other set of plates (B) was developed by measuring deposition of C4. In plate A, MBL binding was measured by incubating wells with a mouse monoclonal biotin-labeled anti-human MBL antibody (diluted 1:3000 in TBS-T plus calcium, biotinylated HYB 131-01; State Serum Institute) for 2 h at room temperature. Wells were washed, and europium-labeled streptavidin (Wallac, Turku, Finland) was added at a concentration of 0.1 mg/liter in the above washing buffer except that calcium was omitted, and 50 M EDTA was included. Wells were incubated 1 h at room temperature, washed, and developed by adding 100 l of Delfia En-hancement Solution (PerkinElmer Life Sciences) and incubated on an orbital shaker for 5 min. at room temperature. Then wells were counted in a Wallac Victor 2 d multicounter 1420 (Wallac, Turku, Finland). In plate B, deposition of C4 was measured by incubation 1.5 h at 37°C with purified human complement component C4 (ϳ1.5-2 ng/ml) in a buffer of sodium barbital (5 mM), NaCl (181 mM), CaCl 2 (2.5 mM), MgCl 2 (1.25 mM), pH 7.4, 1 mg/ml human albumin (State Serum Institute). Wells were washed three times, and 1 mg/liter biotinylated rabbit anti-human complement component C4c was added (Dako), biotinylated according to standard procedures. Wells were incubated for 1 h at room temperature and washed. Europium-labeled streptavidin was added, and development continued as described above.
Purification of rMBL from Tissue Culture Supernatant-MBL was purified from tissue culture medium using both affinity and ion exchange chromatography. 200 ml of tissue culture medium was centrifuged (15,000 ϫ g, 20 min) and filtered (0.2 m) before incubation end-over-end at 4°C with anti-MBL-Sepharose (HYB 131-11; State Serum Institute). The beads were packed into a chromatography column (Econo-Column (25 mm); Bio-Rad) and washed with 10 -30 ml of PBS (1 ml/min on a BioLogic LP chromatography system; Bio-Rad) until the A 280 base line was reached. Bound rMBL was eluted with 1 ml/min of 4.5 M MgCl 2 , 50 mM Tris-HCl, pH 7.5, and ELISA-positive fractions were pooled and dialyzed against 20 mM Tris-HCl, pH 8.0. Samples were concentrated using Centricon Plus-20 (regenerated cellulose, M r cut-off 10,000; Millipore Corp., Bedford, MA). The concentrated samples were loaded on an ion exchange column (Resource Q (1 ml), flow rate 1 ml/min; Amersham Biosciences) mounted in a BioLogic LP chromatography system (Bio-Rad). The column was washed with 5 ml of 20 mM Tris-HCl, pH 8.0, and eluted using a 10-ml linear gradient of 0 -100% 1 M NaCl in 20 mM Tris-HCl, pH 8.0. ELISA-positive fractions were pooled and dialyzed against PBS, and purified recombinant MBL was stored at Ϫ20°C.
Gel Filtration-Purified MBL or tissue culture medium containing MBL was centrifuged (16,000 ϫ g for 5 min) before being fractionated on a Superose 6 column (300 ϫ 10 mm) equilibrated in 20 mM HEPES, 130 mM NaCl, 0.5 mM EDTA, using the fast protein liquid chromatography system (Amersham Biosciences) operated at a flow rate of 0.7 ml/min. The void volume of the column was 8 ml. The column was equilibrated using a range of standards with known Stokes radii: human C1q (10.7 nm), human factor H (7.2 nm), human IgG (5.4 nm), chicken ovalbumin (2.9 nm). Standard protein Stokes radii were calculated as described in Ref. 49 where a represents the Stokes radius, K is the Bolzman constant, T is the temperature in kelvin, Z is the viscosity of water at 20°C, and D is the diffusion coefficient of the protein in water at 20°C. The Stokes radii of the different MBL mutations were determined by their elution from the gel filtration column relative to the protein standards (49). Sucrose Density Gradient Centrifugation-Two aliquots of sucrose solution (20 and 10% sucrose in 20 mM HEPES, 130 mM NaCl, 0.5 mM EDTA) were layered into centrifuge tubes, and the linear gradient was generated by centrifugation at 35,000 rpm for 17 h at 4°C (Beckman SW40Ti rotor) to form a linear sucrose density gradient. Protein samples (500 l) were loaded onto the gradient and centrifuged at 35,000 rpm for 16 h at 4°C. Gradients were fractionated into ϳ20 fractions by peristaltic pumping from the base of the gradient. Individual fractions were analyzed by SDS-PAGE and by measuring A 280 . MBL was detected in fractions by ELISA as described above. Sedimentation coefficients (s 20,w ) were estimated by comparison of their mobilities with those of the following standard proteins: thyroglobulin (19.2 S), bovine liver catalase (11.2 S), bovine serum albumin (4.2 S), and equine skeletal muscle myoglobin (2.0 S). All proteins were from Sigma, and sedimentation coefficients were taken from Ref. 50 where M r represents the relative molecular mass, Z is the viscosity of water at 20°C, N is Avogadro's number, a is the Stokes radius, s is the sedimentation coefficient in Svedberg units, r is the density of water at 20°C, and V p is the partial specific volume of the protein (taken as 0.712 cm 3 /g for MBL, calculated from the known amino acid sequence (52)).

Determination of Molecular Weight by Surface-enhanced Laser Desorption/Ionization
Mass Spectrometry-Mass spectra were obtained using a surface-enhanced laser desorption/ionization time-of-flight (SELDI-TOF) mass spectrometer (PBS2; Ciphergen Biosystems). Proteins were captured on a 24 spot reverse-phase (H4) ProteinChip® Array (C16 carbon backbone) and analyzed as described below.
The individual spots of an H4 ProteinChip Array were equilibrated with 5 l 50% (v/v) acetonitrile and incubated for 5 min prior to sample application. All incubations took place at room temperature in a humidity chamber. Acetonitrile was removed, and 2 l sample plus 2 l 20% (v/v) acetonitrile were added and mixed on the spot. Samples were incubated for 30 min and then removed. The spot surface was washed three times with 10% (v/v) acetonitrile (3 ϫ 5 l) and then twice with H 2 O (2 ϫ 5 l), after which the surface was air-dried. This was followed by two additions of 0.5 l of saturated EAM1 (matrix molecule (Ciphergen Biosystems), applied in solvent containing 50% (v/v) acetonitrile and 0.5% (v/v) trifluoroacetic acid), and the surface was air-dried. After insertion of the ProteinChip Array into the mass spectrometer (PBS2; Ciphergen Biosystems), a laser beam was focused on the sample in vacuo. This caused the proteins to become ionized and, simultaneously, to be desorbed from the ProteinChip surface. The ionized proteins were detected and their molecular masses were determined using TOF analysis. The arrays were read at a suitable laser intensity and sensitivity, and data from 65 laser shots were acquired and averaged to produce the spectra. Prior to the measurements, the instrument was externally calibrated using a mixture of protein standards (Ciphergen Biosystems) of bovine ubiquitin (8564.8 Da), bovine ␤-lactoglobulin (18,363.3 Da), and chicken conalbumin (77,490.0 Da).

Variant Forms of MBL Do Not Form Higher Oligomers-To
understand the molecular basis underlying the MBL deficiency observed in humans with variant forms of MBL, the variant and wild type human MBL were produced in CHO cells. Expression levels of 1-2 mg of MBL/liter of cell growth medium were typically obtained corresponding to median serum levels in humans (27). The oligomeric composition of recombinant MBL was analyzed on nonreduced and reduced SDS-PAGE followed by Western blotting (Fig. 2). As can be seen in Fig. 2A, the recombinant wild type form of MBL forms covalent oligomers containing two polypeptide chains (50 kDa) and smaller amounts of oligomers containing three, four, and five chains (75, 100 and 125 kDa, respectively). Strong bands are seen corresponding to two globular heads (equal to six chains with three chains in each "head") and three, four, five, and six heads. Oligomers containing intermediate amounts of chains are also visible (e.g. two heads and one chain (seven polypeptides)), as are oligomers containing seven, eight, and nine and even more heads. This pattern is in agreement with the pattern seen in a serum pool where the dominant bands are four chains, two heads, three, four, and five heads as well as intermediate amounts of chains ( Fig. 2A, right panel). The variant forms clearly differ in that none of them form covalent oligomers of higher M r . They all form a dominant band with a molecular mass of ϳ50 kDa, corresponding to two single chains. Progressively weaker bands are seen at 75, 100, and 125 kDa, corresponding to three, four, and five chains. The extra cysteine in MBL-D evidently does not contribute to interchain disulfide bonding. All forms of MBL can be reduced to a single chain with an apparent molecular mass of ϳ32 kDa (Fig. 2B).
Variant Forms of MBL Form Noncovalent Oligomers Containing Two Subunits ("Heads")-To determine the M r of variant MBL in a nondenaturing condition, gel filtration and sucrose density gradient (SDG) centrifugation were carried out as described under "Materials and Methods." Gel filtration makes it possible to calculate the Stokes radius (a) of a native conformation of the protein. SDG centrifugation enables a calculation of the sedimentation coefficients (s 20,w ). Having determined the two physical constants a and s 20,w , calculation of the molecular mass, as described under "Materials and Methods," is straightforward (Table I) Previous studies have revealed that collectins show anoma-

FIG. 2. Western blot of a nonreduced (A) and reduced (B) SDS-PAGE of recombinant MBL variants (A-F) from tissue culture supernatant of transfected cells and the State Serum Institute MBL (clinical grade MBL purified from pooled donor plasma).
Shown also is the empty vector construct (pED) and a pool of human serum (NHS). Indicated on the left of A is the number of polypeptide chains in oligomers. The number of polypeptides above the molecular weight standard is extrapolated using the pattern seen below the molecular weight standard; bands that contain a multiple of three polypeptides are more intense. The band having two polypeptides deviates from that rule. Also shown is the molecular weight standard. lous behavior in gel filtration because of the highly extended conformation of the molecule, which in turn is due to the presence of a triple helical collagen-like domain (54,55). This means that the molecular weight cannot be calculated from gel filtration data alone, unless standards of the same shape are used. This makes SDG centrifugation essential for determination of M r of native MBL. Selected fractions containing MBL from gel filtrations were analyzed by SDG centrifugation. Sedimentation coefficients (s 20,w ) were calculated as described under "Materials and Methods." The results of calculated Stokes radii, sedimentation coefficients, and M r are presented in Table  I. The Stokes radii and sedimentation coefficients are in good agreement with previously published values for MBL (56).

Determination of Molecular Weight by Surface-enhanced Laser Desorption/Ionization Mass Spectrometry-Mass spectra
were obtained using a SELDI-TOF mass spectrometer (PBS2; Ciphergen Biosystems) as described under "Materials and Methods." As can be seen from Fig. 3 (uppermost panel), the unreduced WT MBL-A variant shows peaks at 25,548, 38,176, 50,830, 76,218, 101,324, 152,077, and 227,255 Da. The 25,548and 38,176-Da peaks represent the doubly charged ion of the 50,830-and 76,218-Da peak, respectively, since no material of that size is present when analyzed in Western blot (Fig. 2). Other peaks can all be assigned to represent the covalent dimer, trimer, tetramer, hexamer, and nonamer of chains (Table II). Since larger oligomers are not easily desorbed from the surface of the chip, the M r of these is not determined by this technique. As can be seen from Fig. 3 (second panel), the A variant can be reduced to a single chain of 25,545 Da. As a reference, MBL purified from pooled donor plasma (53) (State Serum Institute) was also analyzed. As can be seen from Fig. 3  (third panel), the pattern of peaks is nearly the same. Extra peaks at 60,762 and 113,869 Da are seen. The 113,869-Da peak is the doubly charged ion of the 226,639-Da peak, whereas the 60,762-Da peak probably represents MASPs co-purifying with serum MBL (an extra peak, not shown, at 19,123 Da is also present, corresponding to sMAP (15,16)). The M r of reduced MBL purified from plasma comes very close to that of recombinant MBL (25,545 versus 25,546 Da). The peaks at 51 kDa probably represent nonreduced dimers often seen in SDS-PAGE of MBL.
As can be seen from Fig. 4 (first panel), the B variant forms oligomers of double (51,172 Da) and triple chains (76,259 Da). This preparation of MBL is contaminated with human serum albumin that is seen as a peak at 66,940 Da. Again, the 25,713-Da peak and the 33,485-Da peak represent doubly charged ions of the two-chain MBL and of human serum albumin, respectively. No peaks of higher M r MBL are seen for the B variant, and it can be reduced to a single chain of 25,690 Da ( Fig. 4, second panel). MBL D forms a similar pattern of two chains (50,797 Da) and three chains (76,125 Da (Fig. 4, fourth panel). Mass spectrometry results are summarized in Table II.
Variant Forms of MBL Bind Sugar Motifs with a Reduced Capacity-MBL binds to an array of carbohydrate structures on surfaces of bacteria, yeast, viruses, and parasitic protozoa. Since variant forms of MBL do not form covalently higher oligomers, it is reasonable to expect a reduced avidity toward sugar motifs. We therefore coated microwells with mannan, incubated with recombinant WT MBL-A or variant MBL in cell growth medium, and added the competing sugars mannose and N-acetylglucosamine. As can be seen from Fig. 5, the variant forms all bound mannan, although their binding capacity was reduced 8 -25-fold. MBL binding could be inhibited by the addition of both mannose and N-acetylglucosamine in WT MBL-A as well as the MBL variants, artificial and native. As should be expected of a member of the group of C-type collectins, the binding of MBL to mannan was dependent on Ca 2ϩ , since the addition of EDTA completely eliminated binding.
MBL Activation of the Lectin Pathway of Complement-To establish whether the recombinant MBL could activate complement, mannan was coated on microwells, incubated with recombinant WT MBL-A or variant MBL cell growth medium, washed, and, as a source of complement components, incubated with serum from an individual deficient of functional MBL (genotype HYPD/HYPD). Activation of complement was measured as deposition of C4 assayed by ELISA as described under "Materials and Methods." As can be seen in Fig. 6, the WT MBL-A form of MBL could activate complement. However, in none of the variants could complement activation be detected when measured as C4 deposition on a mannan surface, even when we added an increased concentration of variant MBL or increased the time of development of the ELISA (data not shown).
Variant Forms of MBL Can Bind MASP-2 and Deposit C4 on an Anti-MBL Antibody Surface-To be able to quantify MBL/ MASP-2 interaction independently of MBL binding to mannan, a new assay for complement activation versus MBL binding was designed. In this assay, MBL is bound to an anti-MBL antibody-coated surface, and recombinant purified MASP-2 is allowed to bind. Then C4 is added, and the deposition of C4 is determined. A plot of bound MBL versus C4 deposited for the Each sample was run on gel filtration to determine Stokes radius, and fractions were analyzed by sucrose density gradient centrifigugation to determine the sedimentation coefficient (s 20,w ), The molecular weight of each gel filtration peak was calculated using Stokes radius and the sedimentation coefficient as described under ''Materials and Methods.'' same culture supernatant dilutions will give an indication of the capability of the different MBL variants to bind MASP-2 and catalyze deposition of C4 independent of the capacity to bind to the mannan surface. As can be seen from Fig. 7, deposition of C4 is found in WT MBL-A variant and to a smaller extent in the B and C variants (ϳ60% of WT MBL-A) and to a lesser extent in the D variant (ϳ40% of WT MBL-A).

DISCUSSION
It has become evident that MBL plays an important role in innate immunity (57). The most important tool to dissect this role has been disease association studies showing that structural as well as promoter alleles in the MBL2 gene confer increased risk of a variety of infectious and autoimmune dis-eases as well as poor prognosis (58). Although it is well established that the structural MBL alleles dramatically decrease the serum level of functional MBL in the blood, the molecular basis for this phenomenon is still enigmatic.
In this report, we demonstrate that introduction of naturally occurring MBL mutations in recombinant MBL constructs prevents the correct assembly of higher order oligomers, resulting in reduced capacity of binding to sugar ligands. The ability of variant MBL to interact with MASP-2 is only reduced and is not completely destroyed.
In the last decade, recombinant MBL has been produced in a variety of different cell lines ranging from E. coli (59), insect cells (60)

TABLE II Summary of results from mass spectrometry
The difference between calculated and analyzed M r of a single chain is shown in parentheses. This number can be interpreted as the weight of possible post-translational modifications. NA, not applicable. Determination of molecular mass Ͻ100 kDa is performed with an accuracy of 0.05-0.1%.  (65). Formation of higher M r oligomers has been postulated to be necessary for proper biological activity of MBL (66). Many of the studies conducting recombinant synthesis of MBL suffer the disadvantage that they have resulted in a predominant formation of lower M r oligomers due to 1) failure in complete disulfide bridge formation; 2) proper hydroxylation of proline residues, which is believed to stabilize the triple helix; and 3) hydroxylation and/or glycosylation of lysine residues (67). The recombinant WT MBL-A variant produced in this study has an exceptionally high M r due to extensive disulfide bridge formation, a normal post-translational modifica-tion, and shows full biological functionality in ligand binding, complement activation, and MASP interaction. The MBL-B variant has previously been reported to lack complement-activating capacity and to have a reduced number of disulfide bonds (61,62). Despite the low covalent oligomerization, the MBL-B variant has been reported to have wild type activity of the recently discovered MBL-dependent cell-mediated cytotoxicity (68) as well as wild type opsonic and phagocytic function, since they both mediate the uptake of Salmonella montevideo by human neutrophils (61) and enhance the H 2 O 2 production (6). We find in this report that variant MBL can interact with MASP. This raises the possibility that vari- ant MBL can activate complement if a surface exists that will bind low oligomerized MBL variant.
In this study, we find that the unpurified recombinant B variant forms a quaternary structure that resembles that of the A variant with an M r of ϳ450 kDa but with an altered arrangement of disulfide bonds so that the protein is composed of predominantly two and three covalently joined polypeptide chains. Upon purification of MBL-B, it seems that the high M r molecule might dissociate generating low M r MBL-B, probably due to instability of higher oligomeric structures because of decreased disulfide bonding. All recombinant MBL variants have a reduced capacity to bind carbohydrate surfaces, which explains why we do not observe complement deposition on a mannan surface although we find deposition of C4 that is 40 -60% of that of WT MBL-A when MBL is captured on an anti-MBL surface. The finding that low oligomerized MBL binds MASP is in concordance with observations from rat MBL. Chen and colleagues (69) showed that dimers of rat MBL subunits bind to a single MASP-2 dimer, whereas trimers and tetramers can bind up to two MASP-2 dimers. We show here that variant MBL form a quaternary structure containing two heads (dimer) and that they can bind MASP-2 and deposit C4 on an MBL antibody surface. To explore this in greater detail, kinetic measurement of the interaction of variant MBL and MASP would be required.
In contrast to the previous notion, we demonstrated in a recent publication that quite a significant amount of variant MBL might be found in the circulation of MBL variant individuals (48). Defective MBL from these individuals reveals a molecular structure compatible with recombinant MBL with a predominant form of a double head chain structure. The existence of two molecular forms of MBL was recently confirmed by Terai et al. (70).
Several cases of C1q deficiency due to homozygotic mutations in the collagen-like motif Gly-Xaa-Yaa have been described (71). The resulting C1q molecule has a low M r and contains only two structural subunits (globular heads) compared with the normal C1q that contains six. The structural consequences of MBL mutation are very similar; molecules containing two globular heads are formed.
The point mutations in codons 52, 54, and 57 of exon 1 of the MBL2 gene are frequently described as being associated with decreased MBL plasma concentrations. This view is highly simplified. In our opinion, the mutations cause expression of a dysfunctional protein rather than a repressed level of expression. The dysfunctional MBL has an altered molecular structure and explains the reduced MBL level that is typically measured by ELISA (48). In addition, the perturbation of the collagen-like structure of MBL by the natural mutation has been shown to render MBL more susceptible to matrix metalloproteinase proteolysis (72) and explains the observation that the serum half-life of recombinant MBL-B is about half that of wild type MBL when assessed in a mouse model (73).
The other naturally occurring variants (MBL-C and -D) or artificial variants (MBL-E and -F) produced in this study form only low covalent M r oligomers containing two and three covalent, joint polypeptide chains. As shown for the MBL B variant, the MBL-C, -D, -E, and -F variants all have a reduced ligand binding capacity and fail to activate complement when measured as C4 deposition on a mannan surface. MBL-C and -D variants interact with MASP-2, but among the variants, MBL-D has the lowest interaction.
In the literature, the M r of a single polypeptide chain of MBL is often reported to be between 25 and 32 kDa when analyzed by SDS-PAGE. All of our recombinants, as well as serum MBL, show a single band of ϳ32 kDa when run in reduced SDS-PAGE. In the nonreduced SDS-PAGE, MBL migrates as molecules with an M r that is a multiple of 25 kDa. We suspect this discrepancy in the literature to arise from two characteristics of MBL: 1) the high content of glycine in MBL compared with other proteins (i.e. when run in SDS-PAGE, the M r will be overestimated, because SDS-PAGE determines the length of a protein rather than the mass), and 2) the fact that collagenous proteins can be difficult to denature in SDS due to their rigid triple helix structure and thus bind fewer SDS molecules per amino acid. In both cases, the true M r is overestimated. The true M r of reduced MBL was determined by SELDI-MS to be ϳ25,500. This value is in agreement with the M r calculated on the basis of the cDNA sequence with the addition of 1-2 kDa due to post-translational modifications such as O-linked glycosylation of lysine residues and hydroxylation of proline residue (74). MBL is O-glycosylated on lysine in the Yaa position of Gly-Xaa-Yaa. The characteristic glycosylation is a disaccharide consisting of galactose and glucose (-O-Gal-Glu). The resolution offered by the SELDI-MS instrument used in this study is not sufficient to distinguish between the different glycoforms of MBL shown to be present in human MBL purified from plasma (78). Our claim that the post-translational modification of our recombinant proteins is homogenous is supported by the very sharp bands obtained in Western blot of recombinant human MBL (Fig. 2) and by the sharp peaks obtained in SELDI-MS.
In humans, the gene that is the homologue to rat MBL-A is a pseudogene (MBL1), whereas the functional MBL gene in humans (MBL2) may rather be equivalent to the rat MBL-C gene (rodent liver form). This fact complicates the comparison of our results with the results of Wallis and Cheng obtained with rat MBL-A (75). Wallis and Cheng (75) characterized recombinant rat MBL-A expressed in CHO cells artificially mutated in analogous codons as seen in human MBL. They found a substantially reduced complement fixation activity in the variant MBL, although the effect of the different mutations varied. The human D variant in the rat MBL was found to contain mainly single trimeric subunits with a small amount of dimer, and the authors concluded that the deficiency in complement fixation caused by this mutation results from its failure to assemble into higher M r oligomers. The human MBL-B and -C mutants made in the rat MBL had a reduced complement fixation activity but formed nearly normal M r oligomers but with an arrangement of disulfide bonds that were severely corrupted (containing two and four polypeptide chains). In concordance with the findings in rats, we find that the B and C variants have a better capacity to interact with MASP-2 compared with the D variant. In contrast to the findings in the rat MBL, we show in this study with mutations in human MBL that the presence of an extra cysteine in codon 52 of the human MBL-D variant does not alter the formation of interchain disulfide bonds compared with the other human mutants. Extrapolation of structural consequences of human MBL2 mutations made in rat serum MBL-A, which is homologous to the human MBL1 pseudogene, might thus be problematic, especially in view of the fact that rat liver MBL-C (homologous to the human MBL2 gene) forms three chain oligomers (55) and the very homologous rat serum MBL-A (homologous to human MBL1 pseudogene) forms predominantly dimers and trimers of three-chain subunits (76,77). Another important structural difference is that human MBL probably can form larger covalent structures than seen in its rodent counterpart.
Wallis and Drickamer (55) reported that in rat liver MBL (which contains two cysteines in the N-terminal region), the polypeptide chains are linked by disulfide bonds between two cysteine residues at the N-terminal junction of the collagen-like domain in a heterogeneous and asymmetrical manner. In a later paper (76), the same group reported that for the rat serum MBL (which contains an extra cysteine compared with human WT MBL-A), polypeptides within each trimeric structural unit are mostly linked by disulfide bonds between cysteine residues at positions 13 and 18 arranged in an asymmetrical configuration and that disulfide bonds involving Cys 6 connect polypeptides within separate trimers. Thus, Cys 13 and Cys 18 are involved in intratrimeric disulfide bridge formation, and Cys 6 is indispensable for the formation of oligomers of trimers. Based on the observations in rat MBL (55,75,76) and the findings in this paper, we propose a pattern of disulfide bonding and a quartinary structure of the variants as described in the legend to Fig. 8. It is important to stress that this is a purely speculative structure and that other arrangements of disulfide bond patterns could be possible.
We conclude that in human MBL, the N-terminal region of the collagen region is very vulnerable and that mutations in this region cause a structural perturbation making the molecule unstable and decreasing the binding capacity to ligands. Moreover, interaction with the MASPs could be detected by variant MBL. Thus, lack of complement activation by the variant forms of MBL is probably mainly due to reduced capacity to bind ligand, because of an incorrect assembly of the mature molecule.