The X-ray Structure of Human Mannan-binding Lectin-associated Protein 19 (MAp19) and Its Interaction Site with Mannan-binding Lectin and L-ficolin*

MAp19 is an alternative splicing product of the MASP-2 gene comprising the N-terminal CUB1-epider-mal growth factor (EGF) segment of MASP-2, plus four additional residues at its C-terminal end. Like full-length MASP-2, it forms Ca 2 (cid:1) -dependent complexes with mannan-binding lectin (MBL) and L-ficolin. The x-ray structure of human MAp19 was solved to a resolution of 2.5 Å. It shows a head to tail homodimer held together by interactions between the CUB1 module of one monomer and the EGF module of its counterpart. A Ca 2 (cid:1) ion bound to each EGF module stabilizes the dimer inter-faces. A second Ca 2 (cid:1) ion is bound to the distal end of each CUB1 module, through six ligands contributed by Glu 52 , Asp 60 , Asp 105 , Ser 107 , Asn 108 , and a water

served for wild-type MAp19. All variants were purified from the cell culture supernatants by anion-exchange chromatography on a Q-Sepharose Fast Flow column followed by gel filtration in the presence of Ca 2ϩ ions on a TSK G3000 SWG column, as described previously for wild-type MAp19 (16). The concentrations of the recombinant proteins were determined using absorption coefficients (A 1 cm 1% at 280 nm) of 11.4 (Y59A and Y106A mutants) and 12.0 (wild-type MAp19 and all other mutants), calculated by the method of Gill and von Hippel (23). Molecular masses, calculated from the amino acid sequences, were as follows: wild-type MAp19, 19 Site-directed Mutagenesis-The expression plasmids coding for all MAp19 mutants were generated using the QuikChange TM XL sitedirected mutagenesis kit (Stratagene, La Jolla, CA) according to the manufacturer's protocol. The pFastBac1/MAp19 expression plasmid coding for wild-type MAp19 (16) was used as a template. Mutagenic oligonucleotides were purchased from MWG-BIOTECH (Courtaboeuf, France). The sequences of all mutants were confirmed by doublestranded DNA sequencing (Genome Express, Grenoble, France).
Chemical Characterization of the Recombinant Proteins-Mass spectrometry analyses were performed using the matrix-assisted laser desorption ionization technique on a Voyager Elite XL instrument (Perseptive Biosystems, Cambridge, MA), under conditions described previously (24).
Circular Dichroism-CD measurements were performed on a JASCO J-810 spectropolarimeter using a 1.0-mm path length quartz cell. Spectra were recorded over the 200 -320 nm range at 0.5-nm intervals at a scan speed of 50 nm/min and corrected for the contribution of the buffer (145 mM NaCl, 1 mM CaCl 2 , 50 mM triethanolamine hydrochloride, pH 7.4). The concentrations of the MAp19 variants were as follows: wildtype, 17 Crystallization and Data Collection-Recombinant MAp19 was concentrated to 2.5 mg/ml in 145 mM NaCl, 1 mM CaCl 2 , 50 mM triethanolamine HCl, pH 7.4. Crystals suitable for x-ray data collection were obtained at 20°C by the hanging drop vapor diffusion method by mixing equal volumes of the protein solution and a reservoir solution composed of 3% (w/v) PEG 8000 and 0.1 M Tris-HCl, pH 8.5. They were transferred to a cryoprotectant solution containing 15% (w/v) PEG 8000, 21.75% glycerol, and 0.1 M Tris-HCl, and then flash cooled in liquid nitrogen. A native data set indexed in the space group P4 3 2 1 2 was measured at the European Synchrotron Radiation Facility (ESRF) beamline BM14 to a resolution of 2.50 Å. The images were processed and the reflections scaled using the program XDS (25). Crystallographic statistics for the native data set are given in Table I.
Structure Determination and Refinement-The structure of MAp19 was determined using the molecular replacement method. The rotational and translational searches were carried out using the program AMORE (26). Whereas initial searches using the human C1s CUB1-EGF structure (22) were unsuccessful, a molecular replacement solution was obtained using the CUB1-EGF fragment of rat MASP-2 in the dimeric form (Protein Data Bank accession code 1NT0) (21). Rigid body refinement with the program CNS (27) further improved the orientation and position of each of the domains of the search model. Additional refinement of the model was carried out with CNS and model rebuilding was performed using the graphics program O (28). The quality of the map allowed construction of 333 residues of a total of 340 in the asymmetric unit. Residues Thr 1 (molecule A), Thr 1 -Leu 2 (molecule B), and Ser 169 -Leu 170 (molecules A and B) are disordered. Residue Pro 21 is in the cis configuration. The atomic coordinates have been deposited in the Protein Data Bank under the code 1SZB.
Modeling of the MBL Collagen-like Triple Helix-The human MBL collagen-like segment (positions 36 -63) was modeled using the graphics program O (28). Regional differences based on the helical propensity of residues, as established from statistical analysis of collagen-like structures (30), were included in the model. The template Protein Data Bank file corresponding to imino-rich parameters (30) was used for modeling most of the triple helix, except for segments 45-50 and 54 -56, where the alternative template file designed for imino-deficient segments (30) was used. The graphics program O (28) was then used to search for a plausible model of MAp19 interaction with two modeled MBL triple helices. Although different sets of MAp19 orientations were tried initially, only one major hypothetical mode of interaction was found to be fully compatible with both the present mutagenesis data and previously published experimental information (see "Discussion").
Surface Plasmon Resonance Spectroscopy and Data Evaluation-Sur-face plasmon resonance spectroscopy analyses were performed using a BIAcore 3000 instrument (BIAcore AB, Uppsala, Sweden). MBL and L-ficolin/P35 were immobilized on the surface of a CM5 sensor chip (BIAcore AB) using the amine coupling chemistry as described previously (17). Binding of the wild-type and mutant MAp19 species was measured over 16,000 resonance units of immobilized L-ficolin or 9,000 resonance units of immobilized MBL, at a flow rate of 20 l/min in 145 mM NaCl, 1 mM CaCl 2 , 50 mM triethanolamine hydrochloride, pH 7.4, containing 0.005% surfactant P20 (BIAcore AB). Equivalent volumes of each MAp19 sample were injected in parallel over a surface with immobilized bovine serum albumin to serve as blank sensorgrams for subtraction of the bulk refractive index background. Regeneration of the surfaces was achieved by injection of 10 l of 1 M NaCl, 20 mM EDTA. Data were analyzed by global fitting to a 1:1 Langmuir binding model of both the association and dissociation phases for several concentrations simultaneously, using the BIAevaluation 3.1 software (BIAcore AB). The apparent equilibrium dissociation constants (K D ) were calculated from the ratio of the dissociation and association rate constants (k off /k on ). Each MAp19 variant was analyzed at six different concentrations, ranging from 5 to 60 nM for wild-type MAp19 and all mutants except D60A and E83A (25-250 nM), E109A (50 -500 nM), Y59A, D105G, and Y106A (100 -1200 nM).

RESULTS
Overall Structure-Human MAp19 was produced in a baculovirus/insect cell expression system as described previously (16). Mass spectrometry analysis of the recombinant protein yielded a value of 19,088 Ϯ 9 Da, in full agreement with the amino acid sequence (calculated mass: 19,087.1 Da). The crystal structure of MAp19 was solved by molecular replacement using the rat MASP-2 CUB1-EGF structure (21) as a search model, and refined to 2.5-Å resolution. The final R work and R free factors are 0.261 and 0.316, respectively, and the refined model has satisfactory stereochemistry ( Table I). As expected from previous sedimentation velocity analyses (16), MAp19 associates as a Ca 2ϩ -dependent homodimer (Fig. 1). The two monomers interact in a head to tail manner, through major contacts between the CUB1 module of one monomer and the EGF module of its counterpart. The dimer contains four Ca 2ϩ ions, one at each CUB1-EGF interface (site I) and one bound to the distal end of each CUB1 module (site II) (Fig. 1A). The structure has a length of 95 Å and a width of 20 -45 Å.
Overall, the assembly is reminiscent of those observed for the CUB1-EGF segments of human C1s and rat MASP-2 (21,22). However, a detailed comparison of the three structures reveals differences in the relative positioning of their CUB1 and EGF modules. Thus, superimposition of the MAp19 and C1s monomeric structures on their CUB1 domains reveals extensive displacement of the EGF modules, resulting in a distance of ϳ10 Å between their distal ends, corresponding to a rotation of ϳ25°about the long axis of the CUB domains (Fig. 1C). Similarly, a displacement of EGF modules is observed when the human MAp19 and rat MASP-2 monomers are compared, although the rotation angle is only about 14°in that case. In both instances, the flexibility originates from a hinge located near Glu 122 , at the boundary between the CUB1 and EGF modules (Fig. 1C). Additional differences are observed when the dimeric structures are compared. Thus, there are significant variations in the relative positioning of the CUB1 module of one monomer and the EGF module of its counterpart, with rotation angles of 15-19°between MAp19 and the other two dimers. Interestingly, in contrast to the C1s structure (22), MAp19 does not feature a groove in the region where the four modules meet, the corresponding space being partly filled by loops 10 of both EGF modules (Fig. 1B). A comparison in this respect with the rat MASP-2 CUB1-EGF dimer is not possible, because both loops 10 are disordered in the structure (21).
The CUB1 Module and Ca 2ϩ -binding Site II-The MAp19 CUB1 module is organized in two four-stranded ␤-sheets, each made of anti-parallel strands, providing further evidence that this topology, also observed in the corresponding module of human C1s and rat MASP-2, defines a particular CUB domain subset distinct from the plasma spermadhesins (31). As expected from sequence homologies, the MAp19 CUB1 module is structurally closer to its counterpart in rat MASP-2 (root mean square deviation ϭ 0.56 Å, based on 106 C␣ atoms) than to the C1s CUB1 module (root mean square deviation ϭ 0.94 Å, based on 99 C␣ atoms). Compared with C1s, MAp19 contains a 7-residue extension at its N-terminal end (Figs. 1C and 2). Except for Thr 1 of molecule A and residues Thr 1 and Leu 2 of molecule B, the corresponding segments are well defined in the structure and closely interact with each other. As illustrated in Fig. 1B, the interactions mainly involve van der Waals contacts between residues Pro 5 , Trp 7 , Pro 8 , Glu 9 , and Pro 10 of molecule A and residues Leu 3 , Gly 4 , Pro 8 , Glu 9 , Pro 10 , and Val 11 of molecule B. Other interactions are mediated by Trp 7 of molecule A, which forms a hydrogen bond with the main chain carbonyl group of Glu 9 of molecule B, and additional van der Waals contacts with residues Ala 36 , Pro 37 , and Ala 121 of molecule B. Remarkably, these interactions are asymmetrical and the two N-terminal segments exhibit quite different conformations in the structure (Fig. 1B).
A Ca 2ϩ ion (site II) is bound to the distal end of each CUB1 module ( Fig. 1). In molecule A, the Ca 2ϩ ion is coordinated by six oxygen ligands, namely one of the side chain oxygens of Glu 52 , Asp 60 , and Asp 105 , the main chain carbonyl oxygen of Ser 107 , the side chain carbonyl oxygen of Asn 108 , and a water molecule (Fig. 3). The bond distances are in average 2.4 Å, consistent with the value determined for known Ca 2ϩ -binding sites (32). The coordination involves the same protein ligands in molecule B, but the water molecule is missing. A comparison with the homologous site previously observed in C1s (22), reveals several significant differences. (i) The residues equivalent to Ser 107 and Asn 108 of MAp19 are not Ca 2ϩ ligands in C1s. (ii) The C1s residue Asp 53 contributes two ligands instead of one for the homologous residue Asp 60 of MAp19. (iii) Whereas Asp 105 of MAp19 binds Ca 2ϩ through its side chain, its counterpart Asp 98 in C1s binds through its main chain carbonyl. (iv) Two water molecules are present in the C1s Ca 2ϩ binding site. On the other hand, as observed in C1s, the Ca 2ϩ ion is the central element of a network of interactions that connect together loops L3, L5, and L9, thereby extensively stabilizing the distal end of the MAp19 CUB1 module. In this respect, it is interesting to note that Tyr 24 is H-bonded to the side chain carboxyl of Asp 60 and the main chain nitrogen of Glu 52 , and thereby stabilizes loop L3 in the same way as does its counterpart Tyr 17 in C1s (22).
Ca 2ϩ -binding Site I-The MAp19 EGF module has a -fold similar to that described for other EGF-like modules (12), with one major and one minor anti-parallel double-stranded ␤-sheets (Fig. 1, A and B). Loop 10, which is disordered in the rat MASP-2 structure (21), is structurally well defined in MAp19 and exhibits a rather extended conformation (Fig. 1B),  different from that of the corresponding loop of C1s (22). The remainder of the MAp19, C1s, and rat MASP-2 EGF modules show root mean square deviation values of 0.61-0.69 Å, indicative of high structural homology. The Ca 2ϩ ion bound to each EGF module (site I) is coordinated by seven oxygen ligands, including a water molecule and six ligands provided by the EGF module, namely one of the side chain oxygens of Asp 123 and Glu 126 , the side chain carbonyl of Asn 143 , and the main chain carbonyl of Ile 124 , His 144 , and Gly 147 . These residues are strictly homologous to those providing Ca 2ϩ ligands in C1s. In contrast, as discussed previously (22), the Ca 2ϩ -binding site observed in the rat MASP-2 structure only involves five coordination ligands (21). As observed in the C1s structure (22), Asn 143 lacks ␤-hydroxylation, providing further evidence that post-translational modification of this residue to erythro-␤-hydroxyasparagine is not achieved in baculovirus/insect cell systems and is not essential for Ca 2ϩ binding.
In molecule A, the water molecule involved in Ca 2ϩ coordination also forms hydrogen bonds with three of the other Ca 2ϩ ligands (Asp 123 , Glu 126 , and Gly 147 ) and with Gly 39 in loop 4 of the CUB1 module. Thus, as observed in the C1s and rat MASP-2 structures (21,22), this molecule provides a link between the CUB1 and EGF modules, thereby stabilizing the inter-modular interface.
The Inter-monomer Interface-The interface between the two monomers involves a series of hydrophobic interactions, with in particular three major hydrophobic pockets (Fig. 1A). (i) A central pocket around the 2-fold symmetry axis of the dimer, involving Tyr 40 , and Leu 145 from each monomer. Pro 8 of molecule B is inserted in this pocket. (ii) An aromatic triad at each inter-monomer interface, comprising Phe 12 (from the CUB1 module), His 144 , and Tyr 149 (from the EGF module). His 144 also coordinates Ca 2ϩ , and thereby provides a link between Ca 2ϩbinding site I and the inter-monomer interface. (iii) A distal pocket at both ends of the interface, involving Phe 20 , Pro 21 , and His 48 from the CUB1 module and Ala 154 from the EGF module. These three pockets are conserved in the C1s structure and, with a few exceptions, they involve homologous residues (22). An additional hydrophobic pocket not present in C1s, involving Tyr 45 , Thr 47 , Phe 118 , His 140 , and His 142 , is found between the distal pocket and the aromatic triad. As a result, water molecules are excluded from the mid-to-distal region of the dimer interface, and hence the indirect hydrogen bond network found in this area in C1s (22) is not conserved in MAp19. Additional interactions between the CUB1 and EGF modules are provided by direct hydrogen bonds between the side chains of Tyr 45 and His 142 , and of Arg 43 and Asn 143 (Fig. 1A). As illustrated in Fig.  2, most of the residues involved in hydrophobic interactions and hydrogen bonds at the inter-monomer interface are either conserved or substituted by similar residues within the C1r/ C1s/MASP family.
Mapping of the Interaction Site with MBL and L-ficolin-To locate the residues of MAp19 involved in the interaction with MBL and L-ficolin, a series of recombinant point mutants were produced and their binding properties were analyzed by surface plasmon resonance spectroscopy. We initially targeted various charged amino acid residues, of which some appeared to be conserved in the whole MASP family and absent in C1r and C1s. As summarized in Table II, most of these mutations (R41A, H55A, E58A, E78Q, D81N, and E133Q) had very little effect, if any, on the interaction with either MBL or L-ficolin. However, mutation of Glu 83 to Ala significantly increased the K D values for both MBL and L-ficolin (Table II). A second series of mutations targeted residues of the CUB1 module either providing Ca 2ϩ ligands (D60A and D105G) or located in the vicinity of Ca 2ϩ -binding site II (Y59A, Y106A, and E109A). All of these mutations markedly decreased the ability of MAp19 to associate with MBL and L-ficolin (Fig. 4, Table II). The Y59A, D105G, and Y106A mutations virtually abolished interaction with either protein. In contrast, the E109A and D60A mutations had more discriminating effects, the latter abolishing binding to L-ficolin, but only increasing 7-fold the K D value for MBL. In most cases, the increases in K D resulted from a decrease in the k on value (Table II).
Circular dichroism was used to assess the structural integrity of most of the key mutants. As shown in Fig. 5A, wild-type MAp19 and its variants Y59A, E83A, D105G, Y106A, and E109A yielded comparable spectra in the far UV region, with in each case a minimum molar ellipticity value at about 218 nm, characteristic of a ␤-sheet structure. In the same way, similar spectra were obtained in the near UV region (Fig. 5B), providing unambiguous evidence that these mutants all retained a native conformation. Further support to this conclusion came from the observation that all mutants were purified by gel filtration in the presence of Ca 2ϩ (see "Experimental Procedures") and had a chromatographic behavior indistinguishable from that of wild-type MAp19, indicating that they all retained the ability to associate as Ca 2ϩ -dependent dimers.
As shown in Fig. 6A, the mutations that significantly decrease or abolish interaction of MAp19 with MBL and L-ficolin pinpoint residues gathered at the distal end of the CUB1 module, providing strong support for the implication of this particular area in the binding. A detailed view of the proposed binding site (Fig. 6C) shows that the side chains of Tyr 59 , Glu 83 , Tyr 106 , and Glu 109 all protrude from the CUB1 module, and are therefore available for mediating protein-protein interactions.

DISCUSSION
The x-ray structure of human MAp19 determined in the present study provides a third example of a CUB-EGF domain structure, allowing a comparison with those determined recently for human C1s (22) and rat MASP-2 (21). The three proteins clearly form similar head to tail homodimers mainly stabilized by hydrophobic interactions and share many common structural characteristics, further supporting the hypothesis that this type of assembly is conserved in the proteins of the C1r/C1s/MASP family. Nevertheless, a superimposition of the C1s, MAp19, and rat MASP-2 structures reveals significant differences in the relative positioning of their modules, both at the intra-and inter-monomer levels, because of the presence of a restricted hinge at the boundary between the CUB1 and EGF modules. Although it cannot be excluded that these differences may arise in part from packing constraints, this observation provides evidence for a limited degree of freedom at the CUB1-EGF interfaces, a feature that possibly facilitates assembly of the dimeric structures.   Value not measurable, due to the weakness of the binding (see Fig. 4).
An interesting characteristic of the MAp19 homodimer lies in the peculiar topology of its two N-terminal segments, which, despite their identical amino acid sequences, exhibit strikingly different conformations and therefore interact with each other in an asymmetrical manner (Fig. 1B). Whether this structural feature is shared by rat MASP-2 remains to be determined, because the CUB1-EGF-CUB2 fragment crystallized by Feinberg et al. (21) starts at the Ser residue equivalent to Pro 5 of human MAp19 (Fig. 2), the N-terminal segment Ser-Lys-Trp-Pro being disordered in the structure. It appears likely that the tight network of hydrophobic interactions contributed by the N-terminal extensions of MAp19 provides further stabilization of the dimeric structure. Nevertheless, as judged from the rat MASP-2 structure, it may be inferred that these interactions are not essential to the dimerization process. Interestingly, the N-terminal extension is strikingly different in the MASP-2/ MAp19 and MASP-1/MASP-3 lineages (Fig. 2), suggesting that this region may contribute to the exclusive formation of homodimers through exclusion of unlike partners.
As previously observed in the human C1s CUB1-EGF structure (22), a Ca 2ϩ ion is bound to the distal end of each CUB1 module of MAp19. In both proteins, Ca 2ϩ is coordinated by six oxygen ligands, of which three (Glu 52 , Asp 60 , and Asp 105 in MAp19; Glu 45 , Asp 53 , and Asp 98 in C1s) are common to both sites. This observation strengthens our initial proposal (22) that these three acidic residues, which are conserved in about twothirds of the CUB module repertoire, define a particular CUB module subset with the ability to bind Ca 2ϩ . This conservation applies in particular to the CUB1 modules of the C1r/C1s/MASP family, which in addition also contain the two residues (Ser 107 and Asn 108 ) that provide additional Ca 2ϩ ligands through their main chain carbonyl in MAp19. The fact that these latter two residues are present in C1s, but are not Ca 2ϩ ligands, provides a further indication that Ca 2ϩ binding sites have the ability to slightly adapt their coordination mode depending on their particular protein (and/or crystal) context (32).
As discussed previously in light of the C1s CUB1-EGF structure (22), the above considerations raise questions about the fact that no Ca 2ϩ ion was seen in the CUB1 module of rat MASP-2 (21). To further investigate this question, we have superimposed the corresponding regions of human MAp19 and rat MASP-2. This analysis indicates that, whereas Glu 52 and Asp 60 are positioned similarly in both proteins, the side chain of Asp 105 has a different orientation in rat MASP-2, its residues Ser 107 and Asn 108 being disordered (21). It appears very likely from these observations that Ca 2ϩ -binding site II is indeed present in the CUB1 module of rat MASP-2, but was not occupied to a significant extent in the x-ray structure solved by Feinberg et al. (21).
Analysis of the interaction properties of the various point mutants produced in this study provides unambiguous evi- dence that the MAp19-binding site for both MBL and L-ficolin lies at the distal end of the CUB1 module, in close vicinity of Ca 2ϩ -binding site II. As depicted in Fig. 6A, this site is different from those proposed by Feinberg et al. (21) on the basis of shape and symmetry considerations. On the other hand, our proposition is fully consistent with the observations that: (i) interaction of human MAp19 with either MBL or L-ficolin is abolished in the presence of EDTA (16,17). This characteristic also applies to MASP-1, MASP-2, and MASP-3, although in the case of L-ficolin their binding ability is only partly sensitive to EDTA (17). (ii) A naturally occurring mutation D105G has been described recently in human MASP-2 and shown to result in a complete loss of interaction with MBL (33). As this same mutation, when introduced in MAp19, also abolishes its interaction with MBL, it may be inferred that, although the CUB2 domain of MASP-2 significantly stabilizes its interaction with MBL (17,19,20), the site identified here in the CUB1 module is the primary binding site.
The fact that Asp 60 and Asp 105 each provide a Ca 2ϩ ligand raises the question whether these residues are directly involved in the interaction with MBL and L-ficolin, or are required solely because they stabilize, through Ca 2ϩ binding, the distal end of the CUB1 module in the conformation that is appropriate for binding. A comparative analysis of the elution profiles of the MAp19 mutants upon purification by anion-exchange chromatography (see "Experimental Procedures") provides interesting information in this respect. As anticipated from the loss of a negative charge, each of the mutants E58A, E78Q, D81N, E83A, E109A, and E133Q eluted significantly earlier than wild-type MAp19. In contrast, the D105G mutant eluted at a position similar to that of wild-type MAp19, and D60A was significantly delayed, indicating an increase in the overall negative charge of the molecule. A plausible explanation for these observations is that these latter two mutations destabilize, possibly to different extents, the Ca 2ϩ -binding site, hence releasing the Ca 2ϩ ion and therefore unmasking the carboxyl group of the other two acidic residues engaged in Ca 2ϩ binding. Based on these considerations, it appears likely that Asp 60 and Asp 105 are not directly involved in the interaction with MBL and L-ficolin but, like the other Ca 2ϩ ligands, are essential to stabilize the protein scaffold that maintains the binding site.
On the other hand, it may be postulated that Tyr 59 , Tyr 106 , and Glu 109 , on the distal extremity of the Ca 2ϩ -binding site, and Glu 83 , located in the more distant loop 7 (Fig. 6C), each directly participate in the interaction with MBL and L-ficolin. Whereas Tyr 106 is specific of the MASP-2/MAp19 lineage, Tyr 59 , Glu 83 , and Glu 109 as well as the residues involved in Ca 2ϩ -binding site II are conserved in the CUB1 module of the MASPs identified so far in different species. This strongly supports the hypothesis that the MBL/L-ficolin-binding site identified in the present study applies not only to MAp19 and MASP-2, but also to MASP-1 and MASP-3. This hypothesis appears consistent with the recent report (29) that rat MBL associates with all MASPs through similar sites located in the same area of the collagen-like domain. Our finding that MAp19 shares a common binding site for MBL and L-ficolin is in agreement with the observation that most of the residues that form the core of the proposed MASP-binding site in MBL are also present in the ficolins (29). Thus, in the current state of our knowledge, it appears likely that MBL and L-ficolin associate with the MASPs through similar mechanisms involving homologous sites in either partner.
A three-dimensional model of the homotrimeric triple-helical segment of human MBL containing the MASP-binding site proposed by Wallis et al. (29) was constructed as described under "Experimental Procedures" and docked onto the MBL-binding site identified in the present study. As illustrated in Fig. 6B, the spacing between residues Glu 83 and Glu 109 in MAp19 is similar to that between residues Arg 47 and Lys 55 in the MBL triple helix model. As Arg 47 and Lys 55 are both likely candidates for the interaction (29), a tempting hypothesis is that these make salt bridges with Glu 83 and Glu 109 , respectively (Fig. 6B). If this model is correct, additional hydrophobic interactions may involve Tyr 59 and Tyr 106 of MAp19 and Leu 56 of MBL. In the proposed interaction, the two opposite collagen helices of MBL are oriented in such a way that they form an angle of 80 -90°at the hinge, a value that appears consistent with electron microscopy pictures of MBL (34). We believe that this model provides a sound basis for further investigation, e.g. by site-directed mutagenesis, of the assembly of MBL-and ficolin-MASP complexes.