TITLE : Solvent Water Interactions Within The Active Site of the Membrane Type I Matrix Metalloproteinase

TITLE: Solvent Water Interactions Within The Active Site of the Membrane Type I Matrix Metalloproteinase AUTHORS: Elena Decaneto, Tatiana Vasilevskaya, Yuri Kutin, Hideaki Ogata, Moran Grossman, Irit Sagi, Martina Havenith, Wolfgang Lubitz, Walter Thiel, Nicholas Cox* a Max Planck Institute for Chemical Energy Conversion, Stiftstraße. 34-36, D-45470, Mülheim an der Ruhr, Germany. b Department of Physical Chemistry II, Ruhr-Universität Bochum, 44780 Bochum, Germany. c Max-Planck-Institut für Kohlenforschung, Kaiser-Wilhelm-Platz 1, D-45470, Mülheim an der Ruhr, Germany. d Department of Biological Regulation, The Weizmann Institute of Science, Rehovot, Israel. e Research School of Chemistry, Australian National University, Canberra, ACT 2601, Australia. *Email: nick.cox@anu.edu.au; nicholas.cox@cec.mpg.de, telephone: (+49)-208-306-3552.


S1)
Expression and purification of the active catalytic domain of MT1-MMP.For the crystallization of the non-inhibited Zn 2+ -MT1-MMP, the plasmid containing the catalytic domain of human MT1-MMP together with the hinge linker to the hemopexin-like domain (residues 112-292) as described in Grossman et al. 1 was transformed into Escherichia coli BL21 (DE3) (Novagen).The expression was carried out as previously described in Ogata et al. 1 For the preparation of Co 2+ -MT1-MMP, the expression plasmid for the catalytic domain of human MT1-MMP together with the hinge region (residues 112-319) was mutated in (C127S) MT1-MMP (Genscript, USA) to avoid oxidation sensitivity of the Co 2+ and cloned into the pET3a expression vector with a His-tag at the C-terminus.The plasmid was then expressed into Escherichia coli BL21(DE3) (Novagen).Expression of the apo-enzyme and washing of inclusion bodies were carried out as described previously for the wild-type MT1-MMP. 1 No significant difference in activity was observed for the mutant (C127S)-MT1-MMP and the wild type MT1-MMP, and as such only the MT1-MMP label will be hereafter used for simplicity.

S2) Molar extinction coefficient determination for wild type MT1-MMP. An accurate estimation
of the molar extinction coefficient of MT1-MMP was required to determine the correct stoichiometry, i.e. the metal/protein ratio.This value was accurately determined by means of MCD spectroscopy at room temperature, which allows the concentration of tryptophan in a protein sample to be obtained.In Figure 1-a, the blue curves represent the MCD spectra of different concentration of L-tryptophan.The MCD signal peaks at 291 nm, and it was this data points that was used for the calibration curve shown in Figure 1 The corresponding UV absorption of the protein solution at this concentration was measured by means of a UV-VIS spectrometer and corresponded to 0.64 at 274 nm in a quartz cell of 1 cm pathlength.This yieled a molar extinction coefficient of ɛ 274 =A 247 /cl =35280 M -1 cm -1 .This value is similar to the molar extinction coefficient ɛ calc = 35410 M -1 cm -1 , calculated using the protein analysis software ProtParam 2 available through the ExPASy server. 3

S3) Preparation of apo-MT1-MMP (metal-free material) / Preparation of Co 2+ -MT1-MMP.
Since the presence of extraneous metal can arise from many sources (e.g.reagents, wares, bottles) all materials and buffers were treated to avoid metal contamination.Plastic wares were cleaned with 20% HNO 3 in triple distilled water and extensively rinsed with triple distilled water.All buffers used in experiments were prepared using chemicals of the highest purity available and made metal-free either by extraction with dithizone in CCl 4 4 or by passage over a Chelex-100 column (Bio-Rad).Solutions were made oxygen-free by blowing argon through for 1h and stored in anaerobic conditions.Dialysis membranes were demetallized by heating up to 80°C in 0.01 M EDTA for 30 min and washed with triple distilled water.This step was repeated without EDTA once again.Quartz sample cells were used and rendered metal-free by soaking overnight in 20% HNO 3 in double triple distilled water and rinsed again with triple distilled water just before the measurements.
There are several strategies for introducing a metal into a protein described in literature. 6,7Most of them involve the preparation of the apoprotein followed by metal addition 5,6 or the displacement of the original metal by exposing the native protein to excess concentration of the target metal 7,8 .Only two examples of cobalt substitution in matrix metalloproteinases have been reported in literature: i) Co 2+ -MMP-12 was prepared by exhaustive dialysis against cobalt excess in presence of AHA inhibitor 9 ; and ii) Co 2+ -MMP-1 was obtained by removal of the native metal dialyzing against o-phenanthroline and successively addition of cobalt salt 10 .We initially attempted method (ii): native MT1-MMP was dialyzed against o-phenanthroline resulting in an unstable apoprotein which subsequently aggregated and precipitated out of solution.This is because the protein is subject to autoproteolysis, and thus slow exhaustive dialysis of the zinc form against excess of cobalt (metal exchange) cannot be carried out in absence of a chemical inhibitor.To obtain uninhibited Co 2+ -MT1-MMP in the large quantities required for spectroscopic investigations, the addition of the metal was performed during the folding process of the apoprotein.This folding was carried out in anaerobic conditions through dialysis of apo-MT1-MMP (12-14 kDa molecular-weight cutoff) against 10 fold volume of 50 mM Hepes pH 8. Millipore, 10 kDa MW cut-off).The protein concentration was determined at 275 nm, using the extinction coefficient for the wild-type MT1-MMP of 35280 M -1 cm -1 .This value was calculated by using magnetic circular dichroism spectroscopy (MCD) of tryptophan at room temperature. 11,12 he quality of the samples was determined by SDS-PAGE.Zinc and cobalt content of the protein samples from three independent preparations was measured by flame atomic absorption spectroscopy (Mikrolab Kolbe GmbH, Mülheim an der Ruhr, Germany).Metals content was found to be 2.1 ± 0.2 mol Co 2+ / mol protein and 0.0 ± 0.2 mol Zn 2+ / mol protein.These values suggest that the protein contains a cobalt atom both in the catalytic and in the structural site and that the presence of residual zinc is negligible.Spectra shown in Fig. 15A were measured at low power (P = 20 μW) to minimize the saturation effects at 5 K.In the absence of saturation, the EPR signal intensity is proportional to the inverse of the temperature.We find that the ratio of signal intensities for 5 and 15 K is around 2.6 instead of 3, which indicates that at T = 5 K and P = 20 μW the spectrum is slightly saturated.This behavior is indicative of the zero-field splitting value being D>0.Fig. 15B demonstrates that the EPR spectrum at g ≈ 4.3 consists of two components with different saturation properties.The broad main feature assigned to the Co 2+ ion at the catalytic site saturates more readily than the second component, which shows additional splitting.
Fig. 16 shows the raw MCD data collected at 1.8 K in the absence and presence of the inhibitor AHA.
For both datasets, a scatter like background was subtracted from the data, which was suspected to be caused by strain.Fig. 17 and 18 show global fits of the data in terms of 8 and 4 Gaussian functions respectively, described in the main text.As described in the main text, band assignments were based on the effect of the inhibitor (AHA).
Bands 1-4 are all modified by the inhibitor and are thus assigned to the catalytic site: band 1 (17,200) band 2 (18,000) are significantly diminished or completely lost upon inhibitor binding whereas band 3 (20,000) decreases in intensity by 50% and is shifted to the red (500 cm -1 ) and band 4 doubles in intensity (its position is not well defined in the non-inhibited form).Bands 5-8 are unchanged and thus assigned to the structural site.It is important to note though that the structural site should also have some contribution in the 17,000-21,000 cm -1 region and thus the fitted bands should only be considered dominantly one species or the other.
While it is difficult to justify fitting the feature at 20,000 cm -1 to two bands on the basis of a single MCD spectrum collected at one temperature and magnetic field, changes in the exact shape of this feature as a function of temperature (see Fig. 18), make the inclusion of two bands necessary.It is noted though that the relative intensities of bands 3 and 4 are unlikely to be well constrained.
From this dataset we are unable to constrain the coordination number of the structural site in the Co 2+ substituted protein.The only feature we can assign to the structural site which does not overlap with the catalytic site is the pseudo-A type MCD feature at 24,000 cm -1 .This feature has no clear nesting behavior, suggesting a large zero-field splitting (D > 100 cm -1 ).d-d transitions of high spin Co 2+ complexes do not generally extend this far to the blue, although there are examples of 5/6 coordinate species in the literature whose highest d-d transition does appear in this region. 14,15 e hypothesize that the structural site in the Co 2+ substituted enzyme could adopt a five coordinate, square pyramidal coordination if the mono-dentate carboxylate ligand of the corresponding Zn 2+ complex favored bidentate coordination in the Co 2+ substituted protein, owing to the higher coordination number preference of Co 2+ . 15If this was the case, it could explain the blue shifted MCD signal and also the large D value inferred for the structural site. 14nally we stress that we do not observe any evidence for a 4-coordinate metal site in the Co 2+ substituted enzyme.Tetrahedral Co 2+ complexes typically display a strong pseudo-A type MCD transition, i.e. a derivative shaped signal in the 17,000 cm -1 region is completely absent in the spectra.

S8) QM/MM geometry optimizations of structural models A-K.
We employed QM/MM geometry optimizations to model the metal coordination spheres in MT1-MMP.
The coordination sphere of the structural zinc ion is tetrahedral and consists of three histidine residues (His186, His201, and His214) and one aspartic acid (Asp188).As hydrogen atoms cannot be seen in Xray experiments, we modeled both protonation states of aspartic acid: deprotonated and protonated.
When Asp188 is protonated, the hydrogen bond Asp188-O2H•••O-Thr190 appears, which is absent in the crystal structure.Hence, we propose that the structural site of MT1-MMP resembles structure A in  The catalytic zinc ion is located in the active site of the enzyme.The X-ray data show three histidine residues (His239, His243, and His249) coordinating this zinc ion, but the exact number of water molecules is not resolved.We prepared models of the catalytic zinc coordination site with 1-3 water ligands (structures C-I in Fig. 20).In the tetrahedral models the single water ligand interacts with Glu240 either directly or through an additional water molecule, and it thus represents the nucleophilic water, W n .In the penta-coordinate models, the same water ligand is present, with a second water, W b , binding on the back face of Zn 2+ .We expect back-face binding to be preferred over front-face binding because of the possible interaction with Pro259 on the back, while there are no obvious water-protein interactions on the front face.Key distances for models C-I are listed in Table 3.

Models E-G with 5-fold coordination.
In structure E two water molecules coordinate zinc and one of them forms a hydrogen bond with Glu240.In structure F two water molecules coordinate zinc and the side chain of Glu240 is negatively charged.In structure G a hydroxyl anion and a water molecule coordinate zinc and the side chain of Glu240 is neutral.The computed distances from zinc to both oxygens of the Glu240 side chain increase in the order E < F < G.

Models H-I with octahedral coordination.
In structure H three water molecules coordinate zinc and one of them forms a hydrogen bond with Glu240.In structure I three water molecules coordinate zinc, two water molecules form H-bonds with a hydroxyl anion, and the side chain of Glu240 is neutral.
Models J-K of the AHA inhibited catalytic site.The optimized structures are shown in Fig. 21.
Structure J corresponds to the inhibited complex in the absence of the first coordination sphere water.
Structure K represents the 6-coordinated complex with one water molecule.
-b.The spectrum of the wild type MT1-MMP was acquired and the MCD absorption at 295 was 8•10 4 deg.This value corresponded to a concentration of L-tryptophan of 73 µM, determined by means of the calibration curve.Since only four tryptophan are present in the sequence of wild type MT1-MMP, the concentration of the enzyme that generated the red spectrum in Figure 1-a was readily calculated from [MT1-MMP] = [Trp] / 4 = 18 µM.
Co 2+ -MT1-MMP preparation results in a pure enzyme as demonstrated by SDS page (Figure2).The protein has a molecular weight of ~ 21 kDa, which corresponds to the size of the catalytic domain after autoproteolytic cleavage of the hinge region (residues 112-285). 120 µL of protein solution was mixed with 10 µL of reducing sample buffer and heated for 5 min at 70 °C before application to the gel.The composition of the buffer is 20 mM Tris-HCl pH 6.8, 1 mM Na 2 EDTA, 20% (v/v) glycerol, 0.025 (w/v) bromophenol blue, 5% (v/v) 2-mercaptoethanol.Dual Color Protein Standard III (Serva Electrophoresis) was used as marker (m).The purity of the Co 2+ -MT1-MMP construct was checked by SDS page.Figure2shows the detected band of Co-MT1-MMP one day before and right after the last purification dialysis step.The enzyme was characterized by a molecular weight of ≈ 21 kDa, which corresponds to the size of the catalytic domain after autoproteolytic cleavage of the hinge region (residues 112-285).1

Figure 4 :
Figure 4: The structural and catalytic site of the catalytic domain of MT1-MMP.

Figure 5 :
Figure 5: The interaction of the His-tag the catalytic zinc ion in our MT1-MMP construct showing the electron density (2Fo─Fc map).

Figure 6 :
Figure 6: Modeled location of the His-tag of our MT1-MMP construct showing the Fo─Fc omit electron density map (blue mesh).

Figure 7 :
Figure 7: Additional electron density (Fo─Fc omit map, blue mesh) identified in our MT1-MMP crystal structure assigned to two cryoprotectant (glycerol) molecules (See Figure 4 and Supporting information Figure 5).The identified electron density was similar to that seen in the earlier Bertini structures which included the cleaved peptide (N-fragment of Ile-Ala-Gly), suggesting the glycerol molecules have the same surface protein interactions as the substrate and the electron density associated with these small molecules can be used to model the likely conformation of the N-fragment product would take in our crystallographic model.

Figure 8 :
Figure 8: Comparison of the catalytic site of MT1-MMP, with the His-tag bound to the Zn 2+ ion (PDB entry 5H0U, this study: green) with the resting state of MMP-12 (PDB entry 2OXU: pink) i.e. three number of waters interacting with the Zn.

Figure 9 :
Figure 9: Comparison of the catalytic site of MT1-MMP, with the His-tag bound to the Zn 2+ ion (PDB entry 5H0U, this study: green) with the resting state of MMP-8 (PDB entry 2OY4: pink) i.e. three number of waters interacting with the Zn.

Figure 10 :
Figure 10: Comparison of the catalytic site of MT1-MMP, with the His-tag bound to the Zn 2+ ion (PDB entry 5H0U, this study: green) and the modeled peptide (IAG model : light blue), with the cleaved peptide (IAG 2oxz : pink) of MMP-12 (PDB entry 2OXZ: pink).

Figure 11 :
Figure 11: Comparison of the catalytic site of MT1-MMP, with the His-tag bound to the Zn 2+ ion (PDB entry 5H0U, this study: green) and the modeled peptide (IAG model : light blue), with the cleaved peptide (IAG 2oxw : pink) of MMP-12 (PDB entry 2OXW: pink).

Figure 12 :
Figure 12: Comparison of the catalytic site of MT1-MMP, with the His-tag bound to the Zn 2+ ion (PDB entry 5H0U, this study: green) and the modeled peptide (IAG model : light blue), with the cleaved peptide (IAG 2oxw : pink) of MMP-8 (PDB entry 2OY2: pink).

Figure 13 :
Figure 13: Comparison of the catalytic site of MMPs.(A) Structural alignment with MMP3s with the carboxylate ligands.(B) MMP3 (PDB id 4dpe) with W b .(C) Structural alignment with MMP12s with the carboxylate ligands.(D) MMP12 (PDB id 3ts4) with W f .

Figure 14 :
Figure 14: Interaction of the His-tag with the protein cleft.According to the nomenclature of Schechter and Berger (Biochem Biophys Res Commun, 1967), the MMPs active site is characterized by three unprimed pockets (to the left side of the catalytic zinc: S1 to S3) and three primed pockets (to the right side: S1´ to S3´) which drive the binding of the substrates.The main subsite for substrate recognition inMMPs is the specificity pocket S1´13 which varies considerably in terms of amino acid sequence, size and shape.The primed pockets thus interact with the N-terminus fragment, while the C-terminus fragment instead interacts with the unprimed pockets.

Figure 16 :
Figure 16: Raw MCD data Co 2+ substituted MMP-12 in the presence (B) and absence (A) of the inhibitor AHA (100 mM) collected at 1.8 K.A baseline was subtracted from all data to account for a field dependent baseline.

Figure 17 :
Figure 17: Global MCD fitting of Co 2+ substituted MT1-MMP in the presence (B) and absence (A) of the inhibitor AHA.Left and middle) Magnetic field dependence measured at 1.8 K and 5K; Right) Temperature dependence measured at 7 T.

Figure 18 :
Figure 18: MCD fitting of the low frequency (red) region in terms of four Gaussian functions.(A) without AHA, (B) with AHA.Top) Magnetic Field dependence measured at 1.8 K (LHS) and 5K (RHS); Bottom Temperature dependence measured at 7 T (LHS) and the entire dataset (RHS).A slightly different baseline was subtracted from the raw data in this analysis.These fittings were used to construct the magnetization curves in the main text.

Fig. 19 .
Fig. 19.Key distances for models A and B are listed in Table 3.They confirm the proposed assignment since the computed Asp188-O2•••O-Thr190 distance is in the range of the experimental values for model A, but much smaller for model B.

Figure 19 :
Figure 19: Coordination of zinc at the structural site: models A and B.

Figure 20 :
Figure 20: Coordination of zinc at the catalytic site of an uninhibited MT1-MMP: models C-I.

Figure 21 :
Figure 21: Coordination spheres of zinc at the AHA inhibited catalytic site: models J and K.

Table 1 :
1-ray data-collection statistics for the catalytic domain of MT1-MMP.The values in parentheses are for the outermost resolution shell.1

Table 2 :
Geometry of the catalytic Zn 2+ ion seen in literature MMP structures which have an exogenous carboxylate ligands bound to the Zn.

Table 3 :
Key distances in crystal structures and QM/MM-optimized coordination spheres of zinc at the structural site of MT1-MMP, in Å.
a) a dimer; b) 3.26 Å with Thr190 in the QM part; c) 2.44 Å with Thr190 in the QM part.

Table 4 :
Key distances in the QM/MM-optimized coordination spheres of zinc at the catalytic site of an

Table 5 :
Key distances in the QM/MM-optimized coordination spheres of cobalt and zinc at the structural site of MT1-MMP, in Å.

Table 6 :
Key distances in the QM/MM-optimized coordination spheres of cobalt and zinc at the catalytic site of uninhibited MT1-MMP, in Å.

Table 7 :
Key distances in the QM/MM-optimized coordination spheres of cobalt and zinc at the catalytic site of uninhibited MT1-MMP, in Å.

Table 8 :
Key distances in the QM/MM-optimized coordination spheres of cobalt and zinc at the catalytic site of uninhibited MT1-MMP, in Å.

Table 9 :
Key distances in the QM/MM-optimized coordination spheres of cobalt and zinc at the catalytic site of inhibited MT1-MMP, in Å.

Table 10 :
Optimized Cartesian coordinates in xyz format.

Optimized model geometries used for EPR parameter calculations.Table 11 :
Optimized Cartesian coordinates in xyz format