MASP-1, a Promiscuous Complement Protease: Structure of Its Catalytic Region Reveals the Basis of Its Broad Specificity¹

The complement system is an integral part of the immune system. It plays an essential role in both the innate and adaptive immune responses. Activation of the complement system culminates in the destruction of invading pathogens and elimination of altered host structures (e.g., apoptotic cells) and contributes to the development of inflammation via stimulation of the cellular elements of the immune system (1). The complement system is a proteolytic cascade system that can be activated through three different routes: the classical, the lectin, and the alternative pathway. The lectin pathway of complement can directly recognize the pathogen microorganisms through lectin molecules that can bind to the carbohydrate structures on foreign surfaces. Mannose-binding lectin (MBL)⁴ and ficolins (L-/H- and M-ficolin) are pattern-recognition molecules that form the recognition subunits of the lectin activation pathway (2). When these molecules bind to the foreign structures, the associated serine protease zymogens (MBL-associated serine proteases (MASPs)) become activated. Up to now, three kinds of MASPs have been identified, denoted MASP-1, MASP-2, and MASP-3. MASP-2 is the only MBL-associated protease that can efficiently activate the complement cascade alone (3). MASP-2, like C1s in the classical pathway, can cleave C4 and C2 that form the C3-convertase complex (C4b2a) (4, 5). The function of MASP-3 is unclear. It has been suggested that MASP-3 can down-regulate the activity of MASP-2 because they are competing for the same MBL molecules (6). The most abundant MASP in the serum is MASP-1. Its serum concentration is roughly ten times higher than that of MASP-2 (∼70 and ∼5 nM, respectively) (7, 8). MASP-1, like MASP-2, can autoactivate; however, it cannot initiate the formation of the C3-convertase because it can cleave only C2 with sufficient efficiency, and it is practically inactive toward C4. Several lines of evidence suggest that MASP-1 can support the C3-convertase-generating activity of MASP-2 by increasing the rate of C2 cleavage (9, 10). This could explain why every C4b deposited by the MBL-MASPs complex can form C4b2a convertase, whereas only one of four C4b deposited by the classical pathway C1 complex can do the same (11). It was also suggested that MASP-1 can facilitate the activation of MASP-2 (12). Because MASP-2 alone can readily autoactivate (13), the relevance of this reaction is questionable. It has also been shown that MASP-1 can cleave C3 with very low efficiency (5, 14). This slow C3 cleavage could be due to the hydrolysis of C3, because MASP-1 cleaves hydrolyzed C3 (C3i) with ∼20-fold higher (but still low) efficiency than native C3 (4). The physiological relevance of this C3 cleavage is therefore questionable. Studies with synthetic peptide substrates showed that MASP-1, like thrombin, has extreme Arg preference at the P1 position (15). It was also demonstrated that MASP-1 is capable of cleaving fibrinogen and factor XIII, although less efficiently than thrombin does (16, 17). This activity of MASP-1 could represent an ancient type of innate immunity, well known in lower organisms: the localized coagulation. Molecular evolutionary studies also proved that MASP-1 is a more ancient type of protease than the other early complement proteases: C1r, C1s of the classical, and MASP-2 and MASP-3 of the lectin pathway (18). C1r, C1s, MASP-1, MASP-2, and MASP-3 form a family of mosaic serine proteases with identical domain organization. These molecules consist of six domains. At the N-terminal there is a C1r/C1s, sea urchin Uegf, and bone morphogenetic protein-1 (CUB) domain followed by an EGF-like module and a second CUB domain. This noncatalytic N-terminal region of the molecule is responsible for the dimerization and for interaction with MBL. In the C-terminal half of the molecules, a tandem repeat of complement control protein (CCP) modules precedes the trypsin-like serine protease (SP) domain.

An intact complement system is essential in preventing infections and maintaining the internal immune homeostasis; however, uncontrolled activation of the system can result in host tissue damage and inflammation (19). An important example is ischemia-reperfusion (IR) injury, in which restoration of blood flow after transient ischemic condition results in serious tissue damage (e.g., postmyocardial IR injury). Increasing evidence shows that the lectin pathway plays a critical role in IR injury and targeting the proteases of the lectin pathway might be effective in reducing the tissue damage caused by IR injury (20).

During recent years, there has been great progress in the structural biology of the complement proteases of both the classical and the lectin pathway. The structures of the catalytic fragments of C1r (21, 22, 23), C1s (24), and MASP-2 (13, 25) have been determined. The structure of the CUB1-epidermal growth factor (EGF) fragment of C1s (26) and the CUB1-EGF-CUB2 fragment of rat MASP-2 (27) is also known. These structures provided insights into the structural background of MBL-binding, catalytic activity, and substrate specificity of these proteases and, in the case of C1r and MASP-2, they provided information about the mechanism of autoactivation. Recently, the crystal structure of the CUB1-EGF-CUB2 fragment of MASP-1/3 has been solved (28). Up to now there has not been any structural information available about the catalytic region of MASP-1. In this study we have determined the structure of the CCP1-CCP2-SP catalytic fragment of MASP-1. The structure opens up the way for designing specific inhibitors against this protease, and helped us to explain the unusual broad substrate specificity of MASP-1 providing hints about its physiological role.

MASP-1 CCP1-CCP2-SP (rMASP-1) was expressed in Escherichia coli the form of inclusion bodies. The construct contains an Ala-Ser-Met-Thr expression-enhancing tag at the N terminus, followed by the Gly298-Asn699 fragment of human MASP-1 (the numbering includes the signal peptide). The protein was expressed as described earlier (4). Details of the refolding, purification, and crystallization were published elsewhere (29). It is important to note that rMASP-1 autoactivates during the refolding procedure. The crystal, from which the best dataset was collected, was obtained as follows: 2 μl rMASP-1 at 13 mg ml⁻¹ in 50 mM NaCl, 5 mM Tris, 0.5 mM EDTA, 20 mM benzamidine-HCl, pH 8.8, was mixed with 2 μl of reservoir solution that contained 11% (w/v) PEG3350 and 0.1 M MES, pH 6.5, and crystallized by the vapor-diffusion hanging-drop method. The crystal was gradually soaked into reservoir buffer supplemented with 5, 10, 15, and 20% (v/v) glycerol, and flash-cooled in liquid nitrogen mounted on nylon loops (Hampton Research).

Data collection was conducted using synchrotron radiation (EMBL beamline X11 at DESY, Hamburg) at 100 K (wavelength 0.08148 nm, 165 mm MarCCD detector). Data reduction was conducted using the XDS package (30). The phase problem was solved by molecular replacement with the MOLREP program (31) using a SP domain of a lower resolution chimerical structure containing the SP domain of MASP-1 (E. Sebestyen, V. Harmat, J. Dobó, P. Závodszky, P. Gál, unpublished data) and CCP1 and CCP2 modules of MASP-2 (Protein Data Bank (PDB) entry 1zjk) (13) as models. Manual model building and automated search for water molecules were conducted using the Coot program (32). Refmac5 (33) of the CCP4 program suite (Collaborative Computational Project, Number 4, 1994) was used for refinement. TLS refinement was performed using two TLS groups: one for the CCP1-CCP2 part of the molecule, and the other for the SP domain. Bulk solvent scaling was applied. Hydrogen atoms were generated in the riding positions. The stereochemistry of the structure was checked with the PROCHECK program (34). Data reduction and refinement statistics are shown in Table I. The atomic coordinates and structure factors were deposited in the Protein Data Bank (http://www.rcsb.org/pdb) with accession no. 3gov.

The wild-type enzyme was purified in the presence of benzamidine for crystallization to avoid autolysis of rMASP-1 (4). Because benzamidine could disturb the kinetic measurements, for this purpose we chose the autolysis resistant R504Q mutant that can be purified undegraded in the absence benzamidine. This mutant has the same enzymatic properties on peptide substrates as the wild-type enzyme. A one-step simplified protocol was used for the purification of R504Q rMASP-1, which was expressed and refolded as the wild-type enzyme (29). Benzamidine was omitted during purification, as it could interfere with the kinetic assays. Five hundred milliliters of the refolding mixture (200 mg L⁻¹ R504Q rMASP-1, 0.3 M Arg, 1 M NaCl, 50 mM Tris, 5 mM EDTA, 4 mM glutathione, 4 mM oxidized glutathione, pH 9.4) was dialyzed for 2 days against 2 × 10 l of 1 mM Tris, 0.5 mM EDTA, pH 8.8 with one exchange after day 1, then it was applied to a 20 ml Source 30Q (GE Healthcare) column equilibrated with 5 mM Tris, 0.5 mM EDTA, pH 8.8, buffer and eluted with a 0–200 mM NaCl gradient in 5 mM Tris, 0.5 mM EDTA, pH 8.8 buffer in 25-column volumes. Fractions were assayed for activity (see conditions in the next section) and purity by SDS-PAGE, then selected fractions were combined, concentrated to ∼3 mg ml⁻¹ on spin concentrators, and stored at −70°C. The molar concentration of R504Q rMASP-1 was calculated based on the extinction coefficient ε = 1.54 ml mg⁻¹ cm⁻¹, and a molecular mass of 45.5 kDa.

Lyophilized C1-inhibitor (C1-inh) (Berinert P; CSL Behring) was dissolved and gel-filtrated on a Sephadex G-25 column (GE Healthcare) in 50 mM HEPES, 140 mM NaCl, 0.1 mM EDTA, pH 7.4. buffer. The C1-inh containing fractions were concentrated to ∼0.9 mg ml⁻¹ and stored at −70°C. The molar concentration of C1-inh was calculated based on the extinction coefficient ε = 0.382 ml mg⁻¹ cm⁻¹, and a molecular mass of 71 kDa.

Antithrombin (AT) was prepared from pooled, outdated human plasma. Heparin Sepharose 6 Fast Flow column (GE Healthcare) was equilibrated with 20 mM HEPES, 100 mM NaCl, 0.1 mM EDTA, pH 7.4. The plasma was loaded onto the column, then washed with equilibration buffer. AT was eluted using a 20-column-volume linear salt gradient (0.1 M to 2 M NaCl). AT containing fractions were then gel-filtrated, concentrated, and stored as described for C1-inh. The molar concentration of AT was calculated based on the extinction coefficient ε = 0.65 ml mg⁻¹ cm⁻¹, and a molecular mass of 58 kDa.

The activity and inhibition of R504Q rMASP-1 was assayed in 50 mM HEPES, 140 mM NaCl, 0.1 mM EDTA, pH 7.4 buffer, using the substrate VLR-pNA (V6258; Sigma-Aldrich) at 200 μM concentration. The final enzyme concentration was 10 nM. C1-inh was used at 0 or 200-2000 nM and AT at 0 or 200-2000 nM, whereas heparin was applied at 0 or 0.1–1000 μg ml⁻¹ final concentration.

The second order rate constants of inhibition (k_ass) were determined under pseudo-first-order conditions as described (35, 36). Data were fitted by nonlinear regression using the A = a − b × e^−k_obs×t + c × t equation, where A is the absorbance; a, b, and c are fitting parameters; and k_obs is the observed association rate constant. The second order rate constants (k_a) were approximated as k_a = k_obs/[I]₀, where [I]₀ is the serpin concentration, because in the used-inhibitor range, k_obs was proportional with the inhibitor concentration (37).

The recombinant catalytic fragment of human MASP-1 (Fig. 1,A), which is composed of the first and second CCP domains and the SP domain, was produced in E. coli, refolded from inclusion bodies and crystallized. The construct (406 residues) contains an Ala-Ser-Met-Thr tag at the N terminus followed by the Gly298-Asn699 segment of MASP-1. Details of the crystallization are described elsewhere (29). The structure was solved by molecular replacement and refined to a 2.55 Å resolution (Table I). Overall, 398 of 402 MASP-1 residues could be built into the electron density map. In the N-terminal tag, four residues of the activation loop of the SP domain and 20 side chains spread in loop regions were disordered in the crystal. Interestingly, benzamidine, which was present in the crystallization liquor to protect the enzyme from autodegradation is not seen in the structure. Possibly benzamidine diffused away during a procedure when the crystal was soaked in cryo solutions containing glycerol at incrementally increasing concentrations, or the internal salt bridge (discussed later) between the S1 Asp640 (c189) and Arg677 (c224) may be responsible for this effect. It should be noted here that the numbering in the text follows the precursor MASP-1 numbering for MASP-1 and chymotrypsin numbering is indicated in parenthesis denoted with “c” when necessary for easier comparison. For other proteases chymotrypsin numbering is used throughout the text.

Recombinant MASP-1 CCP1-CCP2-SP (rMASP-1) has essentially the same fold and overall shape as the C1s/C1r/MASP-2 structures containing the two CCP domains concatenated in a rod-like structure attached to the globular SP domain (Fig. 1 B). The domain orientation at the CCP2/SP junction shows minor differences from that of other structures, which is in line with the limited flexibility of this region among these proteases. Comparison of the CCP2/SP interface of MASP-1, MASP-2, C1r, and C1s revealed conserved contacts between the Pro-Tyr-Tyr motif (400–402) of the CCP2 domain, the Pro-Val-Cys linker region (434–436), and the SP domain, although in the contact region of the SP domain Pro570 is the only conserved residue. A novel element of the MASP-1 contact region is the salt bridge between the C-terminal Asn699 carboxyl group and Lys403 of the CCP2 domain. The C terminus is stabilized in an extended conformation by hydrogen bonds of Asn699 and hydrophobic contacts of Val697 to the body of the SP domain. Comparison of the CCP1/CCP2 domain junction of known structures MASP-1, MASP-2, and C1r reveals many structurally conserved main chain hydrogen bonds and hydrophobic contacts, allowing a minor bend at the domain border. After superimposing the CCP2 domains of MASP-1, MASP-2 (PDB ID: 1zjk), and C1r (PDB ID: 1qy0), the distance of the N-terminal Cys301 Cα in the CCP1 domain is 6.4 and 9.8 Å apart from the Cys in equivalent position in MASP-2, and C1r, respectively.

The overall fold of the SP domain resembles that of other proteases of the chymotrypsin family. The residues of the active site (Fig. 1 C), as well as the S1 binding site, are in canonical conformations. Unique features of the MASP-1 SP domain are described in detail below.

Comparison of the SP domain with the structures of human MASP-2, C1r, C1s, thrombin, and bovine trypsin reveals that, except for the variable loops surrounding the substrate binding region, there are two interesting differences in the conformation. First, the region 575–580 (c107–112) is extended, similarly to that of trypsin, though it lacks the disulfide bridge. In contrast, MASP-2, C1r, C1s, and thrombin show helical conformation in this region. Second, the helix of the C-terminal region is about one turn shorter than that of related complement proteases, caused by the unusual interaction of the C terminus with the CCP2 domain.

To explore the structural basis of the substrate selectivity of MASP-1, which is unusually wide compared with related complement proteases, we compared the substrate binding region of the structure with other selected serine proteases. These enzymes have common substrates with MASP-1 (C1r, C1s, MASP-2, and thrombin), and additionally bovine trypsin was included as a reference structure with wide substrate selectivity. The surface topology of MASP-1 resembles that of trypsin, rather than those of enzymes with restricted substrate specificity (C1r, C1s, MASP-2, and thrombin). The substrate-binding groove is wide and exposed. Surface representation near the S1 site (Fig. 2) shows bulky loop regions bending above the substrate binding site in the case of C1r, C1s, and MASP-2, while the enzyme surface of MASP-1 and trypsin is more exposed and smoother at this region. At the border of the S1-S1′ sites, the bulky c192 residue narrows the substrate binding groove in the compared enzymes (in Fig. 2, the middle of the groove, lower side). In contrast, Ala643 in MASP-1 makes the substrate binding site more accessible.

Conformation and length of the loops surrounding the substrate binding region are different even in the case of enzymes with similar substrate specificity (Fig. 3 A). The most striking feature of the S1-S3 side of the substrate binding region of MASP-1 is the exceptionally long loop B (38), which is also called the 60-loop. This loop contributes highly in the specificity of thrombin containing a cluster of aromatic residues folding over the P2 residue of the substrate. Loop B of MASP-1 is even longer and is loosely packed in a hydrophobic patch of the enzyme surface with seven hydrophobic residues turned inside, while its seven charged residues are solvent exposed. The distribution of stabilizing hydrogen bonds in Loop B suggest that it has kink regions, making possible its bending toward the substrate peptide and interaction of Leu496, Pro498, and Pro501 side chains and a small hydrophobic P2 residue. On the other side of the S1-S3 region, backbone conformation of Loop 2 is similar among the enzymes compared, but this loop holds Arg677 (c224) of MASP-1, which is discussed later in detail. The flexible Loop C closes the substrate binding groove on this side.

On the S1′-S3′ side of the substrate binding region, even higher conformation variability was detected among the complement proteases. This region is significantly more accessible in MASP-1 and trypsin than in other complement proteases or thrombin, contributing to the less restricted substrate specificity. Loop A shields the substrate binding groove at its C-terminal end in all related complement proteases and thrombin. Moreover, Loop D and Loop B restrict accessibility in the case of MASP-2 and thrombin, respectively.

The most surprising feature of the SP domain of MASP-1 is that the side chain of Arg677 (c224) is rotated toward the Asp640 (c189, which is responsible for primary substrate specificity in the chymotrypsin family) and forms a salt bridge (Fig. 3 B). This interaction competes with the S1-P1 interaction upon substrate binding. Comparison with the structures of related complement proteases, thrombin, and trypsin reveals a possible way of concerted or induced rotation of side chains, resulting in releasing Asp640 from its salt bridge with Arg677. Loop 2 holding Arg677 is of equal length and similar conformation in MASP-1, MASP-2, thrombin, and trypsin. Loop 2 of C1r and C1s is shorter and it is impossible that any side chain of Loop 2 blocks the S1 Asp. In thrombin and trypsin, a positively charged residue Lys224 is in equal position to that of Arg677. Lys224 side chain of thrombin is stabilized in a noninhibiting position by Glu217, which is in equal position with Asp670 of MASP-1. A c224-c217 salt bridge can also be formed in MASP-1 by rotation of Asp670 (c217) side chain and breaking its salt bridge with Loop 3 (Lys623). Upon substrate binding and formation of the S1-P1 salt bridge, breakage of the salt bridge Arg677-Asp640 is induced, presumably followed by a domino effect, i.e., the re-organization of the ion pair network in a thrombin-like manner. The domino effect and breakage of the salt bridge with Lys623 can possibly be allosterically promoted by changing the conformation of Loop 3 and turning the Lys623 side chain out by interaction with a protein substrate.

In Factor D, the specificity-determinant Asp189 is blocked by a salt bridge (Arg218). This enzyme is in functionally inactive conformation with several distortions in the active site and substrate binding region (39). Comparison of MASP-1 and factor D structures is shown in Fig. 3 C. In factor D, the S1 pocket is narrowed by Loop 2 (self-inhibitory loop) holding Arg218 and a major rearrangement of the S1 region is needed to make it accessible for the substrate. Loop 2 is involved in the conformational switch, ensuring the high specificity of factor D toward its only macromolecular substrate, C3b-bound factor B. In contrast, in MASP-1 the S1 pocket and the c189 Asp is freely accessible suggesting, that the role of the c189-c224 salt bridge is only the modulation of activity and does not restrict the specificity. In summary, the unusual conformation of Arg677 is likely to play a role in the lower activity of MASP-1 compared with trypsin.

Many studies confirmed that C1-inh is a physiological inhibitor of MASP-1 (4, 5, 14, 40). In contrast, only one report showed that AT inhibits MASP-1, but only in the presence of heparin (15). Because of the discussed dual role of MASP-1, both in complement and coagulation, we undertook to quantitatively examine the reaction between rMASP-1 and AT in the absence and presence of different concentrations of heparin, and compared it with the rMASP-1/C1-inh reaction and data presented in the literature (35, 41, 42) regarding the thrombin/AT and rMASP-2/C1-inh reaction (Table II). We found that AT was the fastest inhibitor of rMASP-1 in the presence of heparin, and in contrast to Presanis et al. (15), the reaction took place in the absence of heparin as well, although without heparin AT inhibited rMASP-1 slower than C1-inh did. The second order association rate constant (k_ass) of the rMASP-1/AT reaction was enhanced by ∼30-fold in the presence of 50 μg ml⁻¹ heparin. The observed rate is the fastest rate reported so far for MASP-1 inhibition by a natural inhibitor. Our results can be summarized by the following scheme:

The heparin concentration dependence of the rate constant of the rMASP-1/AT reaction was examined at 200 nM AT and in the 0.1–1000 μg ml⁻¹ unfractionated heparin range (Fig. 4). The lack of a bell-shaped curve is similar to the one observed to the fXa/AT/heparin reaction (42), and indicates that the heparin activation is likely to be mainly allosteric. Heparin accelerates the thrombin/AT reaction differently. The so-called bridging or template mechanism is characterized by a bell-shaped curve (43), indicating a binding site on both the protease and the serpin for heparin. Calculated surface charge map of MASP-1 does not show any highly positive regions similar to the heparin binding site of thrombin (Fig. 5), which is in line with the kinetic data.

Because MASP-1 is associated with the pattern recognition molecules (MBL, ficolins) of the lectin pathway, initially it seemed evident that it has complement substrates. However, the only complement component that is cleaved efficiently by MASP-1 is C2 (k_cat/K_m = 3 × 10⁴ M⁻¹ s⁻¹) (4). The catalytic efficiency of C1s and MASP-2 toward C2 is higher by one order of magnitude (4, 5). MASP-1 also cleaves C3, but the catalytic efficiency of this reaction is very low (300 M⁻¹ s⁻¹), raising doubt on the physiological relevance of this reaction (4, 14). In contrast, MASP-1 cleaves hydrolyzed C3 more efficiently. MASP-1 itself is capable of autoactivation and further autolysis in the SP domain after Arg504 (4). The fact that the catalytic power of MASP-1 is low toward complement substrates (except for C2), and that MASP-1 alone may not initiate the complement cascade, indicates that MASP-1 could have substrates outside the complement system. Indeed, MASP-1 has thrombin-like properties, because it can cleave fibrinogen (α- and β-chains), factor XIII (16, 17), and the protease activated receptor-4 (M. Megyeri, V. Mako, L. Beinrohr, Z. Doleschall, Z. Prohaszka, L. Cervenak, P. Závodszky, P. Gál, unpublished data), and it can be inhibited by AT in the presence of heparin (15).

In this study, quantitative characterization revealed that AT inhibits rMASP-1 both in the presence and absence of heparin. Heparin accelerated the reaction by ∼30-fold, which makes AT in the presence of heparin a more potent inhibitor of MASP-1 than C1-inh. In our experiments, the catalytic fragment of MASP-1 was used. Full length MASP-1 in blood is present in a dimeric form associated with MBL or ficolins, which may influence the reactions of MASP-1. The extent of this effect, however, is possibly similar to different serpins. Both serpins (AT and C1-inh) are found in blood at similarly high concentration (2–3 μM), therefore the determined inhibition rates serve as a good indication of their relative contribution to MASP-1 regulation. The fact that AT in the presence of heparin is the most potent inhibitor of MASP-1 found so far reinforces its analogous role with coagulation proteases.

The structure of the ternary complex of thrombin, AT, and heparin (44, 45) confirmed that bridging plays the pivotal role in the heparin acceleration of the thrombin/AT reaction. Thrombin possesses a heparin-binding exosite that is characterized by a cluster of positively charged side chains. The surface electrostatics of MASP-1 do not show any marked, positively charged patches that could constitute a similar, strong heparin binding site. The effect of heparin on C1-inh is characterized by a charge neutralization, or sandwich mechanism (46, 47). Depending on the charge of the protease near the active site, heparin can increase or decrease the rate of inhibition. MASP-1 seems to be neutral in this respect, i.e., heparin does not enhance or reduce the rate of inhibition by C1-inh. These data, along with the kinetic results and the lack of influence of heparin on rMASP-1’s activity (data not shown), suggest that heparin does not bind to MASP-1 and acceleration of the MASP-1/AT reaction by heparin is caused by allosteric effects on AT.

It has been shown that the proteolytic activity of MASP-1 can be detected on gelatin zymography, which also indicates broader substrate specificity than a typical complement protease (48). The broad substrate specificity of MASP-1 is reflected in the wide substrate-binding groove in the surface representation of SP domain structure. The opening readily accommodates peptides and hence the surface loops of protein substrates can easily gain access to the active site. In this respect, MASP-1 mostly resembles trypsin among the examined proteases (C1r, C1s, MASP-2, thrombin, and trypsin).

Several lines of evidence indicate that the SP domain of MASP-1 has different evolutionary status than that of the remaining members of the C1r/C1s/MASP family. The members of the S1 peptidase family usually contain three highly conserved disulfide bridges. However, in the cases of C1r, C1s, MASP-2, and MASP-3, the Cys42-Cys58 (chymotrypsin numbering) disulfide bridge, the so-called histidine loop, is missing. Because MASP-1 SP contains the histidine loop, MASP-1 can represent an earlier evolutionary stage of complement SP domains (18). Another important fact that suggests an evolutionary different status is that the SP domain of MASP-1 is encoded by six exons, whereas the SP domain of C1r, C1s, MASP-2, and MASP-3 is encoded by one. One of the evolutionary markers of serine proteases introduced by Krem and Di Cera (49) also show that MASP-1 has a more ancient configuration than the other early complement proteases. MASP-1 has a TCN-encoded active Ser, whereas in the related proteases an AGY codon is found. The evolutionary markers place the SP domain of MASP-1 somewhere between trypsin and thrombin and further apart from C1r, C1s, MASP-2, and MASP-3.

MASP-1’s unique features among complement proteases with the same domain organization are reflected in the crystal structure of its catalytic region. The SP domain resembles that of trypsin more than any of those in the MASP-2, C1r, C1s group, hence MASP-1 can be considered as a relatively unspecific protease domain joined to five complement regulatory domains. Of course, MASP-1 is less active and has a more restricted specificity than trypsin, because a protease like that would be unfavorable in blood. In the structure of the SP domain of MASP-1, there are two features that can counteract the impact of the wide substrate-binding groove. One is the unusually large Loop B (also called 60-loop), which might restrict the specificity, and the other is a salt bridge between the crucial Asp residue at the bottom of the S1 pocket and an Arg side chain, which may contribute to the moderate activity of MASP-1 compared with trypsin. A similar salt bridge is involved in the self-inhibitory mechanism that reduces the activity of factor D, the initiator protease of the alternative pathway activation (39, 50, 51). The residue involved in factor D (Arg218) is in the vicinity, but not in equivalent position with the residue (Arg677, c224) in MASP-1. In contrast, factor D has numerous other structural deviations, which render it a practically inactive protease on small substrates. Interestingly, in triple mutant factor D, which has increased activity compared with wild type, Arg223 replaces Arg218 as the partner of Asp189 (50). In this mutant factor D, as well as in MASP-1, the rotation of only a few torsion angles is enough to allow substrate binding. Therefore, the effect of Arg223 in factor D, or Arg677 (c224) in MASP-1 is to reduce, but not to prevent the activity of the protease.

MASP-1 has several substrates outside the complement system, which might contribute to mounting an even more powerful response against infection. In this respect, it is intriguing that MASP-1 is able to cleave protease activated receptor-4, which triggers proinflammatory reactions in leukocytes and endothelial cells (M. Megyeri, V. Mako, L. Beinrohr, Z. Doleschall, Z. Prohaszka, L. Cervenak, P. Závodszky, P. Gál, unpublished data). Our structure explains why MASP-1 behaves as a promiscuous protease compared with the other complement proteases. Its substrate binding groove is quite accessible, resembling that of trypsin; however its reactivity could be modulated by an unusually large loop B (60-loop), and a salt bridge between the S1 Asp and an Arg side chain.

We thank István Szabó for his help with the purification of R504Q rMASP-1 and Júlia Balczer for her technical assistance. We thank Dr. Michele Cianti (EMBL, Hamburg, Germany) for help during data collection. Access to EMBL beamline X11 at the DORIS storage ring, DESY, Hamburg is gratefully acknowledged.

J.D., V.H., P.Z., and P.G. have a pending patent regarding the structure. The other authors have no financial conflict of interest.