Quantitative Characterization of the Activation Steps of Mannan-binding Lectin (MBL)-associated Serine Proteases (MASPs) Points to the Central Role of MASP-1 in the Initiation of the Complement Lectin Pathway*

Background: Autoactivation of initiator proteases of complement is a two-step process. Results: Autoactivation and possible cross-activation steps of complement lectin pathway proteases were quantified. Conclusion: Only MASP-1 can autoactivate rapidly, and MASP-2 is activated by MASP-1. Significance: The determined kinetic data are helpful to interpret activation scenarios for the lectin pathway, and the presented strategy can be used to quantify autoactivation of other proteases. Mannan-binding lectin (MBL)-associated serine proteases, MASP-1 and MASP-2, have been thought to autoactivate when MBL/ficolin·MASP complexes bind to pathogens triggering the complement lectin pathway. Autoactivation of MASPs occurs in two steps: 1) zymogen autoactivation, when one proenzyme cleaves another proenzyme molecule of the same protease, and 2) autocatalytic activation, when the activated protease cleaves its own zymogen. Using recombinant catalytic fragments, we demonstrated that a stable proenzyme MASP-1 variant (R448Q) cleaved the inactive, catalytic site Ser-to-Ala variant (S646A). The autoactivation steps of MASP-1 were separately quantified using these mutants and the wild type enzyme. Analogous mutants were made for MASP-2, and rate constants of the autoactivation steps as well as the possible cross-activation steps between MASP-1 and MASP-2 were determined. Based on the rate constants, a kinetic model of lectin pathway activation was outlined. The zymogen autoactivation rate of MASP-1 is ∼3000-fold higher, and the autocatalytic activation of MASP-1 is about 140-fold faster than those of MASP-2. Moreover, both activated and proenzyme MASP-1 can effectively cleave proenzyme MASP-2. MASP-3, which does not autoactivate, is also cleaved by MASP-1 quite efficiently. The structure of the catalytic region of proenzyme MASP-1 R448Q was solved at 2.5 Å. Proenzyme MASP-1 R448Q readily cleaves synthetic substrates, and it is inhibited by a specific canonical inhibitor developed against active MASP-1, indicating that zymogen MASP-1 fluctuates between an inactive and an active-like conformation. The determined structure provides a feasible explanation for this phenomenon. In summary, autoactivation of MASP-1 is crucial for the activation of MBL/ficolin·MASP complexes, and in the proenzymic phase zymogen MASP-1 controls the process.

Autoactivation is central to several proteolytic cascade systems, such as the coagulation cascade (1, 2), the complement system (3)(4)(5), the system of digestive enzymes (trypsinogen activation) (6), and the caspase system (7,8). "Autocatalytic activation" in these systems is quite common, where an activated protease cleaves its own zymogen form, serving as positive feedback for the activation process. There is another possibility that we term "zymogen autoactivation," where one proenzyme molecule cleaves another proenzyme of the same protease. This mechanism is less common (3,9). Prominent examples are two pathways of the complement system: the classical and the lectin pathways. In these systems, both the zymogen autoactivation and the autocatalytic activation mechanisms play a role. In the contact-kinin cascade, factor XII autoactivates on certain surfaces, where a similar scenario is possible (10).
The complement system (11) is composed of a network of serine proteases, recognition and regulatory molecules. Its main role is to eliminate invading microorganisms and altered self-structures, and in general it contributes to the immune homeostasis of the body. The complement system can be activated via three (interconnected) pathways: the classical, the lectin, and the alternative routes. The initiator complex of the classical pathway, C1, is composed of a recognition molecule, C1q, and a C1s-C1r-C1r-C1s protease heterotetramer. Upon binding to activator surfaces, C1r autoactivates, presumably in two steps, and then it activates the C1s subcomponents, which then cleave the subsequent components of the complement cascade C4 and C2. A similar mechanism occurs in the lectin pathway of complement; however, mannan-binding lectin (MBL) 2 -associated serine proteases, MASP-1 and MASP-2, are able to autoactivate separately (12,13), at least in vitro. In the lectin pathway, the picture is much more complex compared with the classical pathway because there are five known pattern recognition molecules (MBL, H-, L-, and M-ficolins, and collectin-11), three associated serine proteases (MASP-1, -2, and -3), and two other associated proteins (MAp19 and MAp44) (14 -18).
The major player in lectin pathway activation has been thought to be MASP-2 because it can autoactivate and can cleave both C4 and C2 (12,13,19). However, recent studies have shown that in normal serum MASP-1 activates MASP-2 (20 -24), and MASP-1 is responsible for 60% of the C2 cleavage (20). Both MASP-1 and -2 are therefore essential components of the lectin pathway. On the other hand, the exact role of the other associated proteins (MASP-3, MAp19, and MAp44) remains to be clarified (16,25). MASP-3 was recently shown to be activated by MASP-1 (24,26), and it might be involved in alternative pathway activation (26).
All MASPs and MBL-associated proteins (MAps) are Ca 2ϩdependent dimers. The N-terminal CUB-EGF-CUB region is responsible for the dimerization of MASPs and their association with MBL and ficolins. At present, the generally accepted model is that MASP and MAp homodimers bind to distinct recognition molecules (16,27,28); however, at least some of the MBL or ficolin molecules present in higher oligomeric state can potentially bind multiple dimers (24,29).
MASP-1 and -2 as well as C1r circulate in the blood in zymogen (proenzyme) form; hence, their autoactivation must take place in two stages. First one proenzyme cleaves another proenzyme (zymogen autoactivation), and then the activated enzyme cleaves other proenzyme molecules (autocatalytic activation). In the case of C1r, the two autoactivation steps are thought to take place within the same C1 complex (30). Whether MASPs autoactivate within the same complex or they activate MASP molecules on neighboring complexes remains to be clarified.
Under physiological circumstances, MASPs associated with their recognition molecules and C1r as part of the C1 complex autoactivate when these complexes bind to activating surfaces. Autoactivation, however, is also observed on isolated C1r (3) and MASPs (27) and on recombinant fragments as well (9,13).
Upon activation, a specific Arg-Ile bond is cleaved in the serine protease (SP) domain of these enzymes. Changing the Arg to a Gln residue renders the resulting Gln-Ile bond resistant to proteolysis, making the enzyme a stable proenzyme (9,(37)(38)(39). Hence, the autoactivation process can be modeled using recombinant fragments, and a stable zymogen form of these enzymes can be obtained by mutating a specific Arg to Gln. Another mutation is commonly used for serine proteases, where the active site Ser residue is changed to Ala (40 -42). This mutation results in inactive MASPs having an unaltered scissile Arg-Ile bond. These variants serve as zymogen substrates in kinetic measurements of the autoactivation reactions.
Our aim was first to separate the two autoactivation steps of MASP-1 and MASP-2 and quantitatively measure the rate constants of each step for both enzymes using wild type (WT) and mutant recombinant catalytic fragments. We have also crystallized a stable proenzyme variant of MASP-1 encompassing the CCP1-CCP2-SP domains and solved its structure. Later, we extended our studies to the possible cross-activation steps (i.e. we measured the rate constants for MASP-2 activation by MASP-1, and vice versa). The biggest difference was observed in the zymogen autoactivation reactions; in this respect, MASP-1 is about 3000 times more efficient than MASP-2. The kinetic data suggest that the lectin pathway activation starts with the rapid autoactivation of MASP-1, which then cleaves MASP-2. At the beginning of this study, we had no preconceived assumptions regarding the mechanism of lectin pathway activation; however, in a parallel study published recently (20), we have shown, using specific inhibitors, that MASP-1 is the exclusive activator of MASP-2 in normal human serum. The conclusions of the two studies are in a perfect agreement, further emphasizing the key role of MASP-1 in the initiation of lectin pathway activation.

EXPERIMENTAL PROCEDURES
Recombinant MASPs-DNA constructs of recombinant human MASP-1 and MASP-2 catalytic fragments encoding the CCP1-CCP2-SP region (rMASP) were used for expression in Escherichia coli as described (13). The R448Q and S646A (precursor numbering) mutants of rMASP-1 were made by mutagenesis using the QuikChange Lightning multisite-directed mutagenesis kit (Agilent Technologies). The analogous R444Q and S633A mutants of rMASP-2 were made earlier (9,20). The plasmid for the expression of recombinant human MASP-3 catalytic fragment (CCP1-CCP2-SP) was prepared using the pET17b vector (Novagen) analogously to the MASP-1 construct (13). The encoded protein begins with the MASMT vector derived tag followed by the Gly 298 -Arg 728 (precursor numbering) segment of MASP-3.
WT rMASP-1 was produced as described (43) but in the absence of benzamidine. The two-step purification procedure was performed in a single day in order to minimize degradation of WT rMASP-1. The R448Q and the S646A variants of rMASP-1 were prepared just like the WT enzyme; however, in these cases, the procedure yields the proenzymic (one-chain) form.
Preparation of WT rMASP-2 was slightly modified compared with the original protocol (13). The expressed inclusion bodies were dissolved in 7 M guanidine hydrochloride, 50 mM DTT, 50 mM Tris, pH 8.0, and then diluted into a 0.75 M Arg, 50 mM Tris, 5 mM EDTA, 300 mM NaCl, 6 mM glutathione, 3 mM oxidized glutathione, pH 9.0, buffer at a final concentration of 100 mg/liter for rMASP-2. The refolding mixture was kept at 4°C for 1-2 weeks. After excessive dialysis and adjustment to pH 6.3, the rMASP-2 was purified by cation exchange chromatography on a Source 30S (GE Healthcare) column in 10 mM MES, 0.5 mM EDTA, pH 6.3, buffer using a 0 -500 mM NaCl gradient. The appropriate fractions were further purified by anion exchange chromatography on a Source 30Q (GE Healthcare) column and eluted by 0 -200 mM NaCl gradient in 5 mM Tris, 0.5 mM EDTA, pH 8.8, buffer. The R444Q and the S633A variants of rMASP-2 were prepared like the WT enzyme; however, again in these cases, the procedure yielded the proenzymic (one-chain) form.
WT rMASP-3 was produced as follows. The expressed inclusion bodies were dissolved in 7 M guanidine hydrochloride, 50 mM Tris, 25 mM DTT, and then diluted into a 1 M Arg, 50 mM Tris, 5 mM EDTA, 3 mM glutathione, 1 mM oxidized glutathione, pH 8.0, refolding buffer at a final concentration of 100 mg/liter for rMASP-3 at 0°C. The refolding mixture was kept at 4°C for 1-2 weeks. After excessive dialysis against 50 mM NaCl, 10 mM MES, pH 5.7, and application to an SP Sepharose High Performance (GE Healthcare) cation exchange column, rMASP-3 was eluted using a 0 -300 mM NaCl gradient in 10 mM MES, 0.5 mM EDTA, pH 6.1, buffer. The appropriate fractions were further purified by anion exchange chromatography on a Source 30Q (GE Healthcare) column and eluted by a 0 -200 mM NaCl gradient in 20 mM Tris, 0.5 mM EDTA, pH 8.6, buffer. MASP-3 does not autoactivate, but it can be slowly cleaved by contaminating proteases (20,40). When performed in a single day, the procedure described produced about 98% zymogenic rMASP-3.
MASP Cleavage Assays-The rMASP variants, in which the catalytic serine was replaced by alanine, served as substrates (i.e. rMASP-1 S646A and rMASP-2 S633A). The substrate concentration was in the range of 1-50 M, as required by the experiment. R448Q or WT rMASP-1 and either R444Q or WT rMASP-2 served as enzymes for the cleavage assays. Typical enzyme concentrations were about 500 -1000 nM for rMASP-1 R448Q, 1-50 nM for WT rMASP-1, 20 -50 M for rMASP-2 R444Q, and 100 -500 nM for WT rMASP-2. Typically, a substrate/enzyme ratio of Ͼ10 was used, except when rMASP-2 R444Q served as the enzyme, which has a very low activity. In this case, a 1:1 ratio was used because of the limited solubility of the substrate proteins. All of the eight possible combinations were assayed. Results of the WT rMASP-1 versus MASP-2 S633A reaction and the WT rMASP-2 versus MASP-2 S633A reaction were published previously (20). As a control, rMASP-1 S646A or rMASP-2 S633A was incubated alone to detect any possible background cleavage by contaminating proteases. For the rMASP-1 S646A cleavage reaction by rMASP-2 R444Q, no detectable cleavage attributable to the action of rMASP-2 R444Q was observed.
For rMASP-3 cleavage by WT rMASP-1 (100 nM), WT proenzymic rMASP-3 (2 M) served as a substrate because MASP-3 is not capable of cleaving its proenzyme even in the active form (40). As a control, rMASP-3 was incubated alone for the same period of time to detect any possible background cleavage by potential contaminating proteases.
The enzyme and substrate were added on ice to the reaction buffer (140 mM NaCl, 50 mM Hepes, 0.1 mM EDTA, pH 7.4) and mixed, and a sample was taken for the zero time point. The reaction was started by placing the mixture at 37°C. Samples were removed periodically and heated with an equal volume of reducing SDS-PAGE sample buffer to stop the reaction. Samples were then analyzed by SDS-PAGE. The sample loaded per lane contained 1-2 g of substrate. Gels were stained with Coomassie Brilliant Blue G dye, which was followed by densitometry on a Gel Doc 1000 system using the Molecular Analyst software (Bio-Rad). Either the disappearance of the uncleaved substrate or the appearance of the B-chain ( Fig. 1) of the product was quantified.
The observed first order rate constant (k obs ) was determined by nonlinear regression using the I S ϭ I B ϩ I o ϫ exp(Ϫk obs ϫ t) equation or the I P ϭ I B ϩ I ∞ ϫ (1 Ϫ exp(Ϫk obs ϫ t)) equation, where I S is the intensity of the substrate band, I o is the intensity at the zero time point, I B is the background, I P is the intensity of the product band, and I ∞ is the hypothetical intensity at the completion of the reaction (supplemental Fig. S1). In order to make the rate constants of the different reactions comparable, T value can be considered as an approximation of the catalytic efficiency (k cat /K m ) according to the Michaelis-Menten kinetics if the substrate concentration is much less than the K m . The k obs /[E] T values are found in Table 1.
Peptide Substrate Cleavage and Inhibition by SGMI-1-R448Q or WT rMASP-1 and either R444Q or WT rMASP-2 were assayed in 140 mM NaCl, 50 mM Hepes, 0.1 mM EDTA, pH 7.4, buffer at 25°C using carboxybenzyl-Gly-Arg-thiobenzyl ester (Z-Gly-Arg-S-Bzl) as a substrate in combination with an excess of 4,4Ј-dithiodipyridine used as an accessory reagent for detection at 324 nm. At 20 M or less Z-Gly-Arg-S-Bzl, the catalytic efficiencies (k cat /K m ) for R448Q or WT rMASP-1 and for R444Q rMASP-2 were calculated from the where V 0 is the initial velocity of the reaction, [E] T is the total enzyme concentration, and [S] is the substrate concentration. V 0 was calculated from the change of the absorbance using the extinction coefficient of 19 ) equation was used, and the parameters were determined by non-linear regression. The k cat /K m values are found in Table 2. SGMI-1 is a specific, reversible, tight binding inhibitor of MASP-1 with a K I of 6.5 nM for WT rMASP-1 (21). Here, we determined the equilibrium inhibitory constant for proenzymic rMASP-1 R448Q in 140 mM NaCl, 50 mM Hepes, 0.1 mM EDTA, pH 7.4, buffer at 25°C using the same substrate as above. K I was calculated according to Kuzmic et al. (44).
rMASP-1 R448Q Crystallization, Data Collection, Structure Determination, and Refinement-Crystals of the stable zymogen rMASP-1 R448Q variant were grown as follows. 1 l of rMASP-1 R448Q at a concentration of about 9 mg ml Ϫ1 in 50 mM NaCl, 5 mM Tris, 0.5 mM EDTA, pH 8.8, buffer was mixed with 1 l of reservoir solution that contained 20% (v/v) PEG400 and 0.1 M imidazole, pH 7.0, and crystallized by the vapor diffusion hanging drop method. The crystals were gradually soaked into reservoir buffer supplemented with 25 and 30% (v/v) PEG400 and flash-cooled in liquid nitrogen. Following in-house screening on the diffractometer of the MTA-ELTE Protein Modeling Research Group, x-ray diffraction data from the best diffracting crystal were collected on beamline ID 14-1 at the European Synchrotron Radiation Facility (wavelength, 0.9334 Å) at 100 K by helical data collection.
The data were processed and scaled to a resolution of 2.5 Å using the programs XDS and XSCALE (45). The phase problem was solved by molecular replacement with the program MOLREP (46) of the Collaborative Computing Project 4 (47). Individual domains of the activated MASP-1 structure (Protein Data Bank entry 3GOV) were used as search models (31). Model building and refinement were carried out with the programs Coot (48) and Phenix (49), respectively. TLS (translation libration screw) refinement was carried out for individual domains of the structure. The model was validated using PROCHECK (50), SFCHECK (51), and Mol-Probity (52). Data collection and refinement statistics are shown in Table 3. Analysis of the contact regions was carried out with the CCP4 program and the PISA server (53). Structural figures were prepared by PyMOL (54).

Separation of the Autoactivation Steps of MASP-1 and -2-In
order to study the autoactivation process, we have used the CCP1-CCP2-SP catalytic fragments of human MASP-1 and -2. These fragments can be produced in high yield and high purity, which was essential for these studies. Domain structures of fulllength MASPs and the fragments (rMASP) used for this study are illustrated in Fig. 1A. Both WT rMASP-1 and WT rMASP-2 autoactivate during the purification process at room temperature. We note here that over 98% proenzymic WT rMASP-2 can be produced at 4°C (9), whereas WT MASP-1 fully autoactivates during dialysis even at 4°C. Hence, although the autoactivation of zymogen MASP-2 could be studied using the proenzymic WT enzyme (9), this approach is not feasible for MASP-1. Even when the autoactivation reaction can be studied using the WT zymogen form, the observed curve represents the combination of the two types of autoactivation processes running in parallel.
By mutagenesis, stable proenzymic rMASP variants were produced, in which the Arg residue in the scissile Arg-Ile bond was replaced by Gln (rMASP-1 R448Q and rMASP-2 R444Q). These mutants serve to represent the proenzymic (one-chain) form of these enzymes. Mutants of another type were also made, in which the active site Ser residue was replaced by Ala (rMASP-1 S646A and rMASP-2 S633A). These mutants serve as ideal, pure substrates for both the WT and the proenzymic variant enzymes because they still possess the scissile Arg-Ile bond; however, they are inactive even in the two-chain ("active") conformation.
Using the WT enzymes and the described mutants (Fig. 1B), the two steps of the autoactivation reaction can be examined and quantified separately.

Quantification of the Two Autoactivation
Steps of MASP-1-Measurement of the zymogen autoactivation step of MASP-1 was achieved by using the stable zymogen variant rMASP-1 R448Q, which served as the enzyme, and rMASP-1 S646A, which served as the substrate. The purification procedure yielded pure proteins essentially free from contaminating proteases that could have potentially interfered with the assay. Both rMASP-1 variants are stable when incubated alone, indicating that cleavage was due to the zymogen variant when the two variants were mixed (Fig. 2, A-C). The determined cleavage rate (k obs /[E] T ) was about 450 M Ϫ1 s Ϫ1 , which is relatively high for a zymogen serine protease. The autocatalytic activation step was modeled by a reaction where WT rMASP-1 served as the enzyme and the S646A variant as the substrate (Fig. 2D). The autocatalytic activation step (k obs /[E] T Ϸ 83,000 M Ϫ1 s Ϫ1 ), as expected, is much faster than the zymogen autoactivation step. The two constants indicate a rapid autoactivation capacity for MASP-1.

Quantification of the Zymogen Autoactivation
Step of MASP-2-Analogously to the previous case, the zymogen autoactivation step of MASP-2 was modeled by a reaction where the stable zymogen variant (rMASP-2 R444Q) served as the enzyme and rMASP-2 S633A as the substrate. Measuring the kinetic rate constant for this reaction presented a real challenge because it was too slow to be measured in the usual time frame of enzymatic reactions. Very high enzyme and substrate concentration (about 50 M) had to be used, and the total incuba-tion time was about 6 days. Data from a typical experiment are presented on Fig. 3. When the rMASP-2 S633A variant was incubated alone, no significant cleavage was detected during FIGURE 2. Separation of the two autoactivation steps of MASP-1. A, cleavage of rMASP-1 S646A by the R448Q variant modeling the "zymogen autoactivation" step. rMASP-1 S646A at 5.5 M was mixed with rMASP-1 R448Q at 550 nM final concentration and incubated at 37°C. B, as a control, rMASP-1 S646A (5.5 M) was incubated alone. No cleavage was observed during the incubation period at 37°C. C, rMASP-1 R448Q at 550 nM was incubated alone. No cleavage was observed during the incubation period at 37°C. D, cleavage of rMASP-1 S646A by WT rMASP-1 modeling the "autocatalytic activation" step. rMASP-1 S646A at 1.1 M was mixed with WT rMASP-1 at 2.2 nM final concentration and incubated at 37°C. Samples were taken at the indicated time points, and appropriate amounts (0.5-2 g/lane) were analyzed by SDS-PAGE under reducing conditions. Fitting is presented in supplemental Fig. S1, and the determined activation rate constants (k obs /[E] T ) are listed in Table 1.
the same incubation period (data not shown). The determined rate constant (k obs /[E] T Ϸ 0.14 M Ϫ1 s Ϫ1 ) indicates an extremely slow initial proenzymic phase for MASP-2 autoactivation.
The rate constant for the autocatalytic activation step (k obs / [E] T Ϸ 600 M Ϫ1 s Ϫ1 ) published earlier (20), combined with the newly determined rate constant for the zymogen autoactivation step, indicates that although MASP-2 can autoactivate in vitro, it is several orders of magnitude slower than the autoactivation of MASP-1, and MASP-2 probably requires external cleavage, especially for the initial generation of active MASP-2 during the lectin pathway activation.
"Cross-activation" between MASP-1 and -2-If zymogen MASP-2 cannot be cleaved efficiently by its own zymogen, what is the enzyme that causes the initial cleavage when all enzymes are present as zymogens? A logical answer is that MASP-1 autoactivates first, and then active MASP-1 cleaves zymogen MASP-2. Another possibility is that zymogen MASP-1 cleaves its own zymogen and zymogen MASP-2 simultaneously, and then the generated active enzymes take over.
In order to explore all possibilities, we have attempted to measure the activation rates for all combinations. The cleavage of rMASP-2 S633A by WT rMASP-1 (k obs /[E] T Ϸ 12,000 M Ϫ1 s Ϫ1 ) has been determined earlier (20). The reversal of this reaction ( Fig. 4A) (i.e. the cleavage of rMASP-1 S646A by WT rMASP-2) gave a cleavage rate about 3 times lower (k obs /[E] T Ϸ 4100 M Ϫ1 s Ϫ1 ). These values indicate that the active enzymes can potentially activate each other's zymogen form.
Interestingly, zymogen MASP-1 is quite an efficient MASP-2 activator. This reaction was modeled by the cleavage of rMASP-2 S633A by rMASP-1 R448Q (Fig. 4B), and the determined cleavage rate (k obs /[E] T Ϸ 300 M Ϫ1 s Ϫ1 ) is very close to the zymogen autoactivation rate of MASP-1. For the reciprocal reaction (i.e. the cleavage of rMASP-1 S646A by rMASP-2 R444Q), we could not determine the cleavage rate, because the reaction (if any) was too slow. The values determined for the activity of zymogen MASP-1 and MASP-2 imply that for the initial phase of activation, when all enzymes are present as zymogens, the activation of lectin pathway proteases strictly depends on zymogen MASP-1.
A Kinetic Model of Lectin Pathway Activation-We have summarized all the kinetic data in Table 1 regarding MASP autoactivation and cross-activation. Also, in order to graphi-cally illustrate how these constant relate to each other, we have put them into a MASP-1/2 (auto)activation scheme (Fig. 5). Although the measured rate constants were determined using recombinant fragments, their relative values should be similar to those of the full-length enzymes and enzymes bound to MBL or ficolins. A kinetic model of the activation of the lectin pathway immediately emerges from Fig. 5, which is discussed below.
Activation of MASP-3 by MASP-1-Following MASP-1, MASP-3 is the second most abundant MASP or MAp (25,55). Recent publications regarding the potential role of MASP-3 in the alternative pathway and its possible cleavage by MASP-1 (24,26) prompted us to quantitatively characterize the latter reaction. In this case, WT zymogen rMASP-3 served as a substrate because MASP-3 does not autoactivate (40). Data from a typical experiment is presented in Fig. 6. As a control zymogen, WT rMASP-3 was incubated alone for the same period of time, and no activation was observed (data not shown). The observed cleavage rate of rMASP-3 by rMASP-1 (k obs /[E] T Ϸ 1200 M Ϫ1 s Ϫ1 ) is somewhat lower than the active-zymogen cross-activation reactions between MASP-1 and MASP-2 (Table 1) but still reveals an efficient process. This reaction is not included in the lectin pathway activation scheme (Fig. 5) because we did not explore all the combinations for MASP-3 yet, and the subsequent events (i.e. the roles of active MASP-3) are not well established.
Cleavage of Small Substrates and Inhibition of Proenzyme MASP-1-Both WT rMASP-1 and WT rMASP-2 are very active on a synthetic thiobenzyl ester substrate ( Table 2). On amide substrates, MASP-1 is usually more active than MASP-2 (56). Nevertheless, both active enzymes have readily measurable activity. We attempted to measure the activity of the proenzyme variants using the most sensitive substrate available, Z-Gly-Arg-S-Bzl. Previously, it was reported that rMASP-2 R444Q does not cleave this substrate at all (9), but here we could detect a very weak activity at high enzyme concentration. On the other hand, the zymogen variant rMASP-1 (R448Q) had a significant, easily measurable activity toward this substrate ( Table 2).
SGMI-1 is a specific canonical inhibitor developed against active rMASP-1 using phage display (21). We have measured the equilibrium inhibitory constant for the zymogen variant rMASP-1 R448Q and SGMI-1. The observed value (K I ϭ 5.5 Ϯ  Table 1. 1.4 nM, n ϭ 3, ϮS.D.) is practically the same as the one determined for the WT activated enzyme (6.5 nM). This is somewhat surprising and shows that this inhibitor can be effective at shutting down the lectin pathway at a very early zymogen stage and implies that proenzyme MASP-1 can easily adopt an active-like conformation.
Structure of the Proenzyme Form of rMASP-1-We solved the crystal structure of rMASP-1 R448Q consisting of the CCP1-CCP2-SP segment of the enzyme and refined it to a 2.5 Å resolution ( Table 3). The overall structure is shown in Fig. 7A. The relative orientation of the three domains is essentially the same as in the activated structure (Protein Data Bank entry 3GOV; root mean square deviation of 352 fitted C␣ atoms is 0.754 Å), showing the CCP1-CCP2 moiety in a very similar conformation in both structures (the only difference is in the 406 -409 region).
The uncleaved serine protease domain shows characteristics typical to zymogens in the chymotrypsin family; the residues of the catalytic triad are in active conformation, but the oxyanion site is distorted, and Asp 640 (Asp c189 in chymotrypsin numbering) is rotated (Fig. 7B). The substrate specificity (S1) pocket is blocked sterically and electrostatically by a positively charged residue (Lys 675 (Lys c222 )), similarly to that seen in zymogen MASP-2, where Arg 630 (Arg c192 ) covers the S1 site. Similarly to several other zymogen structures, in the rMASP-1 R448Q structure, the Ser 667 -Trp 668 (c214-c215) peptide bond is  Table 1. rotated, which results in a segment of loop 2 (residues 669 -670) blocking the S2 and partially the S1Ј substrate binding pockets, thus being in a collapsed conformation in the zymogen enzyme. It was proposed that in solution such collapsed forms are in dynamic equilibrium with open conformations of both the zymogen and activated enzymes of the chymotrypsin family; moreover, this equilibrium can be shifted by interactions with the substrate or other molecules, and even crystal contacts can stabilize either the open or the collapsed forms (57). In zymogen MASP-1, the loop 2 region contains mainly polar amino acids (6 of 10 residues are charged; see Table 4 and Fig. 7C); therefore, solvation would favor the flexible transformation between its open and collapsed forms. In contrast, loop 2 of zymogen MASP-2 appears in open form in its crystal structure (Protein Data Bank entry 1ZJK; Fig. 7D). However, loop 2 in this latter consists overwhelmingly of hydrophobic residues (10 of 11 residues; Table 4), and the open conformation of the loop is stabilized by hydrophobic crystal contacts, suggesting that for MASP-2, solvation does not promote flexible loop 2 transformation, but hydrophobic interactions with macromolecular substrates may play an important role.
Restoring the S1 and oxyanion sites is essential for enzyme activity. It can be realized through either conformational fluctuation between different forms, or it can be induced during substrate binding. In both cases, liberating closely packed part of the activation domain is required. A closely packed region is formed by loops D and 1, containing an important hinge point of the activation cascade, which is c194 in loop 1. In zymogen MASP-2, this residue (Asp 632 ; accessible surface area, 0.60 Å 2 ) is buried by a hydrophobic segment of loop D, whereas in zymogen MASP-1, it is more accessible (Asp 645 ; accessible surface area, 23.30 Å 2 ). Restoring the active conformation of the c194 Asp has been proven to be possible by forming a salt bridge with positively charged moieties other than the new N terminus of the activated enzyme (58). In zymogen MASP-1, two positively charged residues (Lys 591 and Arg 596 ) of loop D reaching close are candidates for that role (Fig. 7). In contrast, in MASP-2 where c194 Asp is buried, positively charged residues are farther from the ideal interacting position.
In summary, comparison of the zymogen and activated structures suggest that the higher basal activity of zymogen MASP-1 compared with zymogen MASP-2 can be explained by looser packing and by the more polar character of the loop regions of the activation domain in critical position for the inactive to active conformation change (Table 4), which is indicated by the smaller buried surface area of zymogen MASP-1 and the higher extent of its collapse during activation, as well as the different neighborhood of the important hinge point c194 residue.

DISCUSSION
Autoactivation is the intrinsic property of the initiator proteases of complement, namely C1r, MASP-1, and MASP-2; however, it turned out from this and parallel studies (20,24) that autoactivation of MASP-2 is probably not relevant physiologically. Proenzyme forms of WT full-length C1r and its cat-FIGURE 5. A kinetic model of lectin pathway activation. This figure graphically illustrates the kinetic rate constant obtained for the "zymogen autoactivation" and "autocatalytic activation" steps for both MASP-1 and MASP-2 as well as those of the possible "cross-activation" steps. Colored arrows point from the enzyme to the substrate. The model suggests that the autoactivation of MASP-1 is rapid, whereas the autoactivation of MASP-2 is very slow due to the extremely slow "zymogen autoactivation" step. The model also suggests that MASP-2 is activated primarily by MASP-1, and at the initial phase, when all enzymes are zymogens, zymogen MASP-1 plays a pivotal role in initiating the activation process. FIGURE 6. Activation rMASP-3 by rMASP-1. Cleavage of WT zymogen rMASP-3 by WT (active) rMASP-1 modeling a possible cross-activation step. rMASP-3 at 2 M was mixed with rMASP-1 at 100 nM final concentration and incubated at 37°C for 300 min. Samples were taken at the indicated time points, and appropriate amounts (about 1.5 g of total protein/lane) were analyzed by SDS-PAGE under reducing conditions. Fitting is presented in supplemental Fig. S1, and the determined activation rate constant (k obs /[E] T ) is shown in Table 1.   (n ϭ 3). b Very low activity close to the detection limit. alytic region (3,59) and rMASP-2 (9) were produced earlier, and the kinetics of their autoactivation was studied. The observed sigmoid-shaped activation curve is a result of two reactions: zymogen autoactivation (Reaction 1) and autocatalytic activation (Reaction 2). The two-step autoactivation reaction can be described by the following scheme, where E is the zymogen and E* is the active enzyme.

REACTIONS 1 and 2
The autocatalytic activation can be considerably faster than the zymogen autoactivation step, and preparations of a wild type proenzyme can contain trace amounts of active enzyme; therefore, characterization of the zymogen autoactivation step is particularly challenging. Using stable proenzyme variants of rMASP-1 and -2 (R448Q and R444Q, respectively) and variants in which the active site Ser was replaced by Ala (S646A and S633A), this problem has been overcome. For the first time, we have provided quantitative data for the zymogen autoactivation step of a complement protease. The determined zymogen autoactivation rate of rMASP-1 is about 3000-fold faster than that of rMASP-2. Based on this observation, it seems very unlikely that MASP-2 zymogen autoactivation plays a significant role in the initiation of the lectin pathway.
Using the wild type forms of rMASP-1 and -2 and the S646A and S633A variants, the autocatalytic activation steps of rMASPs were quantified. Again, the autocatalytic activation rate of rMASP-1 was higher, about 140-fold, than that of rMASP-2. Although MASP-2 can autoactivate in vitro, this is so slow that MASP-2 probably requires external cleavage under a physiological scenario, especially for the initial generation of active MASP-2. On the other hand, kinetic data suggest that MASP-1 has a propensity to autoactivate very rapidly.
Using the wild type, the stable zymogen (R448Q or R444Q), and the inactive forms (S646A or S633A) of rMASPs, the possible cross-activation steps were also quantified. We have shown previously that rMASP-2 S633A is cleaved by active rMASP-1 about 20-fold faster than by its own active form (20). Now we added the missing rate constants whenever the cleavage was detectable.
Taking into account all of the kinetic data of the two autoactivation steps for both proteases and also those of the possible cross-activation steps, we have set up a kinetic model of lectin pathway activation (Fig. 5). In the initial phase, where all MASPs are zymogens, the activity of zymogen MASP-1 controls the activation process. The zymogen autoactivation of MASP-1 is considerably fast, but also, proenzyme MASP-1 can potentially activate proenzyme MASP-2. Whether the latter happens under physiological circumstances depends on the exact composition of MBL/ficolin⅐MASP complexes and also on the nature of activation (i.e. intramolecular (within the same complex) or intermolecular (between neighboring deposited complexes)).
In the later phase of activation, when enough active MASP-1 and potentially some active MASP-2 is generated, the major player in the activation process is still the more efficient MASP-1. Although active MASP-2 can cleave proenzyme MASP-2 slowly, and it can cleave proenzyme MASP-1 at a considerable rate, one must bear in mind that the concentration of MASP-1 is about 25-fold higher than that of MASP-2 (55). The concentration difference is particularly relevant if the mode of activation is intermolecular (i.e. MASP-1 or -2 in one complex activates MASPs in neighboring complexes). In this scenario, clearly, MASP-1 is the major activator in every phase of the activation process.
If we assume that the activation process is intramolecular and some of the complexes contain both MASP-1 and -2 (24, 29), then the concentration of MASP-1 and -2 becomes irrelevant for the activation. Nevertheless, even in this scenario, our model predicts that in the initial phase, the activity of zymogen MASP-1 still controls the activation process. It is important to note that in the case of an intracomplex activation scenario, the orientation of the MASP molecules can have a major effect on the effectiveness of the activation process. This and dimerization of MASP-1 and MASP-2 can greatly influence their activation because of the local concentration effect. This study alone does not support either the intramolecular or the intermolecular (i.e. intercomplex) activation mechanism; however, at least for MASP-2 activation, an intercomplex activation is more likely, which is discussed below. Hence, there are certain limitations of our proposed kinetic model. The listed rate constants were determined using recombinant CCP1-CCP2-SP fragment and not the whole dimeric molecules. We believe that the relative values of these constants are similar for the full-length molecules and probably molecules incorporated in MBL⅐MASP and ficolin⅐MASP complexes and serve as good starting points for the interpretation of previous and upcoming experimental data. One must also bear in mind that another MASP, namely MASP-3, and two other MAps (MAp19 and MAp44) are also associated with MBL and ficolins. These components may have regulatory functions and modulate the activation process of the lectin pathway. MASP-3 can be particularly important because it has a high concentration, almost half of that of MASP-1, and it was shown to be associated with a developmental disorder, the 3MC syndrome (60,61). MASP-3 and MASP-1 were also suggested to be required for factor D activation and therefore for the activity of the alternative pathway in mice (26,62); however, at least in human serum, this scenario seems unlikely (24). However, MASP-3 might cleave factor B (26), and it was shown to be activated by MASP-1 (24,26). Hence, as a last addition, we have measured the rate of cleavage of rMASP-3 by rMASP-1. We found that rMASP-1 cleaves rMASP-3 at a rate about 10-fold slower than rMASP-2 (S633A). This is still a significant rate; therefore, it is likely that MASP-3 can be slowly activated by MASP-1.
The structure of zymogen MASP-1 catalytic fragment (rMASP-1 R448Q) was determined and refined to a 2.5 Å resolution. The structure shows that rMASP-1 R448Q is in a typical zymogen-like conformation in the crystal. In order to gain catalytic activity, it must adopt an active-like conformation, at least temporarily. rMASP-1 R448Q has about 1 ⁄240 the activity on a small synthetic substrate compared with the wild type enzyme, and the zymogen autoactivation rate of MASP-1 is about 1 ⁄190 of the autocatalytic autoactivation rate. The zymogenicity (63) of MASP-1 therefore shows little substrate dependence. Using the conformational selection (fluctuation fit) model, this might indicate that zymogen MASP-1 spends roughly 1 ⁄200 of its time in an active-like conformation. Interestingly, MASP-2 shows much higher variation of zymogenicity, depending on the substrate. The activity of zymogen MASP-2 is 1 ⁄83,000, 1 ⁄4300, and 1 ⁄8 that of the WT active enzyme on a small synthetic substrate, its own zymogen, and C4 (9), respectively. This might indicate that zymogen MASP-2 cannot effectively adopt an active-like conformation without the aid of an appropriate substrate (induced fit). Comparison of the structures of zymogen and activated MASP-1 and MASP-2 supports this hypothesis. In zymogen MASP-1, the hinge point determining the oxyanion site and loop 1 conformation (Asp 645 (Asp c194 )) is accessible, suggesting that its rotation is easier; furthermore, the highly polar character of the activation domain facilitates hydration and dynamic conformational behavior of the loops of the activation domain. In contrast, in zymogen MASP-2, the c194 aspartate (Asp 632 ) is buried, and the activation domain contains several hydrophobic residues, suggesting that the hydration of these loops is less favorable and enhancing the relative importance of a suitable binding partner.
We have also determined that SGMI-1, a MASP-1-specific inhibitor, which was developed against the activated WT enzyme (21), also effectively blocks the activity of zymogen MASP-1. When this inhibitor was added to human serum containing zymogen MASPs, no lectin pathway activity was detected in a C4 deposition assay (20). Inhibition of zymogen MASP-1 by SGMI-1 explains why there was no MASP-2 activation (and hence C4 deposition) observed under these circumstances. It is important to mention that zymogen MASP-2 bound to MBL does not cleave C4, only its isolated fragment (9,19), so MASP-2 activation is required for C4 cleavage. In mice deficient in MASP-1 and MASP-3, there is a significant reduction of lectin pathway activation (23,64); however, there is still some residual activity. Studies with a complex reconstituted from recombinant full-length MASP-2 and isolated MBL showed delayed activation of MASP-2 compared with the natural complexes (12), but still MASP-2 could activate. All of these data together suggest that on surfaces where MBL⅐ MASP-2 complexes bind in close proximity, MASP-2 can still slowly autoactivate, but when the more abundant MBL⅐MASP-1 complexes are present, then MASP-2 is exclusively activated by MASP-1 very rapidly. Inhibition of both proenzymic and active MASP-1 by SGMI-1 under these circumstances completely prevents activation of MASP-2. In all, a proximity-induced intermolecular (i.e. intercomplex) activation mechanism can reconcile all of the previously mentioned experimental data.
Using recombinant mutant enzymes, the two autoactivation steps of an appropriate protease can be studied separately. Particularly challenging is to characterize the "zymogen autoactivation" step, where one zymogen protease cleaves another zymogen molecule. This can be achieved by using a stable proenzyme variant, in which the scissile bond is altered to be resistant to proteolysis, and another mutant, in which the active site residue is mutated. These mutations have to be designed so that the folding of the enzyme is unaffected. It is also important to prepare very pure mutant enzymes, because the activity of the zymogen enzyme is usually low, and possible protease contaminants can interfere with the reaction. Measuring the "autocatalytic activation" step is somewhat simpler. It can be studied using the wild type enzyme and the active site mutant. This concept can theoretically be applied to any class of proteases.
We have used this approach to serine proteases of the complement lectin pathway, MASP-1 and MASP-2. The two autoactivation steps of both enzymes as well as the possible "crossactivation" steps were quantitatively characterized. Using the rate constant, we have outlined a kinetic model of lectin pathway activation. According to this model, MASP-1 autoactivates rapidly, whereas MASP-2 does not, and MASP-2 is activated dominantly by MASP-1, which is in perfect agreement with recent studies (20 -22). Our model also adds that in the initial phase, when all MASPs are zymogens, the activity of proenzyme MASP-1 is crucial to the activation of the lectin pathway. Our results also strengthen the view that MASP-1 is a prime target for developing drugs to prevent pathological activation of the lectin pathway.