Structure of homo- and hetero-oligomeric meprin metalloproteases. Dimers, tetramers, and high molecular mass multimers.

Meprin A and B, metalloproteases consisting of evolutionarily related alpha and/or beta subunits, are membrane-bound and secreted enzymes expressed by kidney and intestinal epithelial cells, leukocytes, and cancer cells. Previous work established that the multidomain meprin subunits (each approximately 80 kDa) form disulfide-bridged homo- and heterodimers, and differ in substrate and peptide bond specificities. The work herein clearly demonstrates that meprin dimers differ markedly in their ability to oligomerize. Electrophoresis, light scattering, size exclusion chromatography, and electron microscopy were used to characterize quaternary structures of recombinant rat meprins. Meprin B, consisting of meprin beta subunits only, was dimeric under a wide range of conditions. By contrast, meprin alpha homodimers formed heterogeneous multimers (ring-, circle-, spiral-, and tube-like structures) containing up to 100 subunits, with molecular masses at protein peaks ranging from approximately 1.0 to 6.0 MDa. The size of the meprin alpha homo-oligomers was dependent on protein concentration, ionic strength, and activation state. Meprin alphabeta heterodimers tended to form tetramers but not higher oligomers. Thus, the presence of meprin beta, which has a transmembrane domain in vivo, restricts the oligomerization potential of meprin molecules and localizes meprins to the plasma membrane. By contrast, the propensity of secreted meprin alpha homodimers to self-associate concentrates proteolytic potential into high molecular mass multimers and thus allows for autocompartmentalization. The work indicates that different mechanisms exist to localize and concentrate the proteolytic activity of membrane-bound and secreted meprin metalloproteinases.


INTRODUCTION
Proteolytic enzymes are essential components of many cellular and extracellular processes from maturation of proteins to cell death (1). Their activities and localities, however, must be highly regulated because of their destructive potential. Regulation of proteases is accomplished through several mechanisms including zymogen formation, inhibition, localization to specific compartments and transcriptional regulation. The structures of proteases themselves have revealed mechanisms to regulate proteolytic activity, and this has been amply demonstrated in the high molecular mass oligomeric structures of the proteasome, tripeptidyl peptidase II, and the tricorn protease (2)(3)(4).
These multimeric serine and threonine proteolytic complexes are homo-or heterooligomeric, have molecular masses of 0.7 to 9 MDa, and are found intracellularly. The structures serve to restrict, localize and concentrate proteolytic activity. Proteases at the cell surface are known to form transient oligomeric complexes, such as those between MT1-MMP (membrane-type 1 matrix metalloproteinase), TIMP-2 (tissue inhibitor of metalloproteinase 2) and MMP-2 (matrix metalloproteinase 2) that lead to the activation of MMP-2 (5). However, stable secreted, multimeric proteolytic complexes were not described until recently, when homooligomers of meprin A were found to form multimers of approximately 0.9 MDa (6). These observations established meprin A as one of the largest known secreted proteases.
Meprins, zinc-dependent metalloendopeptidases of the 'astacin family' and 'metzincin superfamily', consist of multidomain, evolutionarily-related α and β subunits (7,8). The subunits are 42% identical at the amino acid level, highly glycosylated, and form disulfide-linked homo-or hetero-dimers (7,9). Meprin A (EC 3.4.24.18) is a heterooligomer of α and β subunits or a homooligomer of α subunits; meprin B (EC 3.4.24.63) is a homooligomer of β subunits (10). The nascent subunits each have a signal peptide that directs the protein to the lumen of the endoplasmic reticulum during biosynthesis, and a propeptide that inhibits activity (11). Each subunit has an astacinlike catalytic domain, and several protein interaction domains including a MAM (meprin, A5 protein and protein-tyrosine phosphatase µ), a MATH (meprin and tumor necrosis factor receptor-associated factor (TRAF) homology), and an AM (after MATH) domain (12). The MAM and MATH domains are found in cell-adhesion superfamily proteins and in adapter proteins in signal transduction, respectively, and permit homo-and hetero-philic selective associations with self and other proteins (13,14). The MAM, MATH, and AM interaction domains are essential for the biosynthesis of active, stable meprins (12). The nascent meprin subunits are both synthesized with COOH-terminal EGF (epidermal growth factor)-like domains, putative transmembrane spanning domains and short cytoplasmic tails. However, the nascent meprin α subunit, but not the β subunit, has a 56 amino acid inserted (I) domain between the AM and EGF-like domains, and the presence of this domain allows for a proteolytic event during maturation that liberates the meprin α subunit from the membrane (15). Because of this COOH-terminal processing, homooligomers of meprin A are secreted proteins, whereas heterooligomers of meprin A and homooligomers of meprin B are membrane-bound.
In mammals, meprins A and B are highly concentrated in kidney and intestinal brush border membranes; for example, these proteins are estimated to compose 5% of the mouse brush border membrane of proximal tubule juxtamedullary nephrons (16,17).
Further, membrane-bound and secreted meprins are expressed in leukocytes and cancer cells, implicating these enzymes in inflammation and tumor biology (17)(18)(19). The homooligomeric form of meprin A is found in rodent urine and the media of transfected cell lines and colon cancer cells (18,20,21). While meprin B and heterooligomeric meprin A are predominantly membrane-bound proteins in vivo, there is some evidence that the membrane-bound form of human and rat meprin β can be shed from the cell surface (9,22,23).
Meprins can cleave diverse polypeptides including cytokines, basement membrane proteins, growth factors, protein kinases, gastrointestinal peptides and peptide hormones (10,(24)(25)(26)(27). Recent studies demonstrated that the individual subunits have marked differences in their peptide bond and substrate specificities and conditions for optimal activity (26,28). The meprin β subunit has an acidic pH optimum, prefers low ionic strength and has a distinct preference for acidic residues flanking the scissile bond in substrates. In contrast, the meprin α subunit has a neutral to alkaline pH optimum, prefers small or hydrophobic residues flanking the scissile bond and has a stronger preference for proline residues proximal but not flanking the scissile bond. These substrate and activity differences imply different functions for the meprin isoforms.
Although it is known that dimerization of meprin A is important for stability and activity toward protein substrates, little is known about the propensities of meprin subunits to oligomerize (21). The recent finding that secreted mouse homooligomeric meprin A was primarily decameric, contrasted with previous observations that indicated heterooligomeric meprin A isolated from mouse kidney was primarily tetrameric (6,9).
Little is known about the structure of meprin B, partially because only small amounts have been available. In addition, to date there has not been an accurate determination of the molecular masses of meprin subunits, or oligomeric forms. The work herein was conducted to determine definitively meprin subunit masses and to characterize the oligomeric forms of meprin A and B. For these studies, truncated and histagged recombinant rat meprin subunits were prepared, allowing for the secretion and subsequent large-scale production of meprins. The recombinant proteins were found to have similar properties to the native enzymes (28).

EXPERIMENTAL PROCEDURES
Expression and Purification of Meprins -Meprins were stably expressed in human embryonic kidney 293 cells by transfecting cells with pcDNA 3.1 (+) expression vectors containing meprin sequence using the calcium phosphate precipitation method.
Recombinant rat meprin α was truncated at the putative mature carboxy terminus (R603 of the AM domain) and a histidine tag was added to the carboxy terminus. Meprin β was truncated at K648 which is at the EGF-transmembrane border and a GGGS linker and histidine tag were added to the carboxy terminus (28). The truncation of meprin proteins at R603 and K648 resulted in the secretion of the subunits into the media and the addition of histidine tags allowed for a facile purification scheme. Nickel nitrilotriacetic acid affinity chromatography was used to purify meprins. Homooligomeric meprin A and B were produced by transfecting cells with meprin α or β cDNA alone. A third cell line was transfected with both cDNAs to allow for the production of heterooligomeric meprin A. All proteins were secreted into the media as proenzymes. Active forms of meprins were produced by limited digestion of the purified latent meprin with trypsin as described (28). Meprin was treated with a 1:20 w/w ratio trypsin to meprin and meprin activity was monitored over time. Trypsin was inhibited with a 20-fold excess of soybean trypsin inhibitor when meprin activity no longer increased. Trypsin and soybean trypsin inhibitor were subsequently removed by size exclusion chromatography (SEC) using a Superose 6 column. Meprin protein was subjected to SDS-polyacrylamide gel electrophoresis (PAGE) and gels were stained with Coomassie blue to assess purity and verify complete activation.
PAGE and Immunoblotting -Protein samples, boiled in sample buffer with SDS and 2-mercaptoethanol, were routinely subjected to electrophoresis on 7.5% Ready gels (Bio-Rad) unless indicated. For native PAGE, 3-8% NuPAGE Tris-acetate gels in the absence of SDS and reducing agent were used (Invitrogen). Individual meprin α and β subunits were detected using the subunit specific polyclonal antibodies, HMC52 and PSU56 respectively that were developed by our laboratory. HMC52 and PSU56 antibodies did not cross-react with the incorrect subunit with the amounts used in these studies.
Collection of Rat Urine Two female rats were placed in metabolic cages and urine was collected for 6 h. Urine was kept on ice. After collection, urine was filtered through a 0.2 µm cellulose acetate filter to remove particles. The filtered urine (6 ml) was buffer exchanged into 8 ml of 20 mM Tris-HCl, 150 mM NaCl, pH 7.5 using Econo-Pac 10 DG chromatography columns (Bio-Rad) and then concentrated to 600 µl using a Centriplus YM-50 concentrator (Millipore). weight average molecular masses (M W ) of rat meprins were calculated at peak maxima using three independent analyses, the two and three detector method and ASTRA analysis. A second order Berry fit was used for latent and active homooligomeric meprin A and a zero order Debye fit was used for the other meprins. The M W was estimated throughout the entire eluting peak at 5 µl intervals using ASTRA software. Computations were performed as described (29,30). For analysis of meprin molecular masses under different conditions a Superose 6 10/30 HR column was calibrated using the molecular mass data obtained from the SEC-LS analyses. The column was equilibrated in 20 mM Tris-HCl, 150 mM NaCl, pH 7.5 and the flow rate was 0.3 ml/min. Composites of typical fields and galleries of individual images were prepared using Adobe Photoshop. Statistics of individual particle measurements were compiled from enlarged prints on a digitizing tablet using SigmaScan (Jandel).

Masses of Meprin
All image analysis was performed using the SPIDER/WEB software package (31). For two-dimensional averaging, each data set of untilted images was processed by reference-free alignment and hierarchical ascendant classification using principal component analysis. Three-dimensional volumes were calculated by iterative back projection. Those for activated meprin B and latent heterooligomeric A were constructed de novo from tilt pairs of micrographs. These volumes were used as references for determining reconstruction angles by projection mapping for the volumes of latent meprin B and activated heterooligomeric A, respectively. The numbers of images in each data set were: 7382 for latent meprin B, 5850 for active meprin B, 4366 for latent heterooligomeric meprin A, and 4470 for active heterooligomeric meprin A. In all instances, the number of images was limited so that angular coverage would be as even as possible. There were no significant areas of missing information for any of the volumes.
Resolution limits were determined from the 50% cutoff of the Fourier shell coefficient between volumes of half data sets. Volume surfaces were created using IRIS Explorer (Numerical Algorithms Group). They are shown after filtering to their resolution limits and at the threshold for 100% mass as calculated using the molecular masses shown in Table I and a partial specific volume of 0.71 g/cm 3 (Dr Faoud Ishmael, personal communication).

Purification and Initial Characterization of Recombinant Histidine Tagged Rat
Meprins -All forms of meprin proteins were purified to homogeneity ( Fig. 1; left panel).
The latent meprin β used in these studies is predicted to be larger than meprin α due to the presence of the EGF-like and a portion of the AM domain in meprin β and indeed it appeared to be larger by SDS-PAGE. Meprin bands were diffuse as previously published and as expected for highly glycosylated proteins (9). Rat meprin β has eight potential asparagine-linked glycosylation sites compared to six in the rat meprin α sequence; additional glycosylation in the meprin β subunit could contribute to the higher molecular mass compared to meprin α as well as the occurrence of the more diffuse band.
The band for the latent heterooligomeric meprin A migrated between that of homooligomeric meprin A and B. In some instances two bands were visible under reducing conditions corresponding to each subunit. However, the subunits in the heterooligomeric protein did not resolve well on the gels ( Fig. 1; left panel). In order to obtain more accurate molecular masses, MALDI-TOF was employed (Table I). Using MALDI-TOF the molecular masses of latent meprin β and α subunits were found to be 85.5 and 77.7 kDa, respectively.
Quantitative Western analysis was employed to determine the ratio of the meprin α and β subunits in the purified rat heterooligomeric meprin A (21). Subunit specific antibodies were used to quantify the amount of each subunit present. Known amounts of purified homooligomers were used as standards, run on the same gels and calibration curves were constructed. The amount of each standard and meprin subunit in each sample was determined by densitometry. The protein consisted of an approximately equal amount of subunits; a 1 : 1.2 ratio of the meprin β and α subunits was calculated.
The activation of meprin subunits by trypsin resulted in mobility shifts by SDS-PAGE ( Fig. 1; left panel). The molecular mass losses were 3.4 and 9.4 kDa for the meprin α and β subunits, respectively, as determined by MALDI-TOF (Table I).
Trypsin treatment removes the propeptide of both subunits (11,22,32). The greater loss of molecular mass in the meprin β subunit is probably due to the additional loss of amino acids within the EGF domain. Indeed, trypsin is used to remove meprin β subunits from brush border membranes (32).
All six forms of recombinant meprins were subjected to PAGE in the presence of SDS and absence of 2-mercaptoethanol to ensure that histidine-tagged proteins were also able to covalently dimerize in a similar manner to wild-type meprins (9,33). All forms of meprins migrated in a manner consistent with the formation of disulfide-linked dimers ( Fig. 1; right panel). The molecular masses of homooligomeric latent meprin A and B were 156 and 171 kDa respectively (Table I). The heterooligomeric form of latent meprin A had a molecular mass of 166 kDa. The molecular mass and densitometry data are consistent with a disulfide-linked heterodimer of meprin α and β subunits as expected (33). Molecular masses of 148, 154 and 152 kDa were determined for the active forms of homooligomeric meprin A and B and heterooligomeric meprin A respectively (Table I). A had a molecular mass considerably greater than 669 kDa (Fig. 2). Therefore the activated homooligomer of recombinant rat meprin A forms high molecular mass complexes, analogous to the mouse homologue (6). The latent homooligomeric meprin A was not able to enter the gel indicating that it was larger than the active counterpart.

Native PAGE Demonstrates Evidence of Meprin Oligomers
PAGE in the presence of SDS and absence of 2-mercaptoethanol yielded dimers ( Fig. 1; right panel). Therefore, the formation of the large complexes was dependent on noncovalent interactions. In contrast, the latent and activate forms of homooligomeric meprin B had electrophoretic mobilities that corresponded to molecular masses of approximately 200 kDa, consistent with the formation of dimers (Fig. 2). Analytical ultracentrifugation studies also demonstrate that rat meprin B exists as a dimer (Dr Faoud Ishmael, personal communication). Dimers were also formed under denaturing conditions ( Fig. 1; right panel). Therefore, rat meprin β subunits do not interact noncovalently to form larger complexes under these conditions. The active form of meprin B was slightly smaller than the latent form as assessed by native PAGE as expected. For heterooligomeric meprin A two bands were visible for both the latent and active forms. The mobilities of the bands corresponded to molecular masses of approximately 200 and 400 kDa. Thus, latent and active heterooligomeric meprin A existed as a mixture of dimers and tetramers under these conditions (Fig. 2).
Heterooligomeric meprin A existed as dimers in the presence of SDS and absence of 2mercaptoethanol ( Fig. 1; right panel). Thus noncovalent interactions were involved in the putative dimer to tetramer transition. Interestingly, the active tetramer had a larger apparent molecular mass than the latent form by this technique.
The Oligomeric Size of Meprins in Solution -SEC-LS was used to obtain additional oligomeric state information. SEC separates polydisperse mixtures of proteins before the determination of M W . A single peak with a large tail was evident after SEC of the latent form of homooligomeric meprin A (Fig. 3A). The peak contained macromolecules with a M W range between 1.5 to 8.0 MDa (Table II). The signal saturated the LS detector at the peak maximum (7.9 ml). Nevertheless, it was clear from the M W distribution that homooligomeric meprin A formed a polydisperse pool of various molecular mass macromolecules. The large complexes were composed of up to more than 100 monomers based on a monomeric molecular mass of 77.7 kDa (Table I and II). The M W at 7.8 ml, near the peak maximum (7.9 ml) was 6.1 Mda (Table II).
This value is consistent with this form of meprin existing as oligomers composed of approximately 39 dimers or 78 subunits (Table I and II). Using the two and three detector approach, a M W of 5.5 MDa was predicted for the peptide portion of the protein, indicating that on average 82 monomers are involved in the macromolecular structures.
Berry analysis revealed that the macromolecules had root mean square (rms) radii ranging from 20 to 55 nm and an average of 26 nm (data not shown).
The SEC UV trace of the active form of homooligomeric meprin A indicated that this protein was less heterogeneous than the latent protein (Fig. 3B). The active protein peak contained macromolecules with M W values ranging from 1.0 to 1.7 MDa as compared to 1.5 to 8.0 MDa for the latent protein (Table II) Single, well-resolved peaks were evident when the latent and active forms of homooligomeric meprin B were subjected to SEC. The peak maxima of latent and active meprin B were at 14.7 and 14.9 ml respectively indicating that meprin B was much smaller than homooligomeric meprin A (Fig. 3C and D). After SEC of the latent form of heterooligomeric meprin A, two peaks were evident with maxima at 13.6 and 14.5 ml (Fig. 3E). The 13.6 ml peak contained macromolecules with M W in the range of 300 and 360 kDa. The maximum of the peak had a M W 333 kDa by Debye analysis. The two and three detector approaches predicted molecular masses of 268 and 269 kDa respectively. Based on the dimeric molecular mass of a heterodimer (166 kDa; Table I) and the polypeptide sequence-predicted molecular mass for the equimolar mixture of monomeric forms of the two subunits (140 kDa), this peak contained a tetramer, a noncovalent dimer of disulfide-linked heterodimers (Table II). Debye analysis indicated that the second peak, which eluted at  (Table II). Based on the dimeric molecular mass of a heterodimer (152 kDa; Table I) and the polypeptide sequence-predicted M W for the equimolar mixture of monomeric forms of the two subunits (127 kDa), this peak contained a tetramer, presumably a noncovalent dimer of disulfide-linked heterodimers. The second peak eluted with a maximum at 14.8 ml and contained macromolecules with a range of M W values between 140 to 200 kDa (Table II). The observed M W distribution indicated that this peak contained a heterodimer of meprin α and β subunits. The peak did not resolve well from the major peak that eluted at 13.8 ml. The maximum of the 14.8 ml peak had M W values of 182, 145 and 146 kDa by the Debye, two and three detector approaches (Table II). Although these values are higher than expected for dimers, the observed M W distribution indicates that this peak contained a heterodimer of meprin α and β subunits.
The estimated values are probably higher than the actual values due to poor resolution 20 by guest on March 23, 2020 http://www.jbc.org/ Downloaded from between the dimer and tetramer in the size exclusion chromatography step.

The Effect of Meprin Concentration and Ionic Conditions on the Formation of
Higher Order Oligomers -SEC data indicated that the oligomeric states of homo-and hetero-oligomeric meprin A are dependent on the concentration of meprin (Summarized in Table III). The molecular masses and therefore oligomeric states of meprins were estimated using the LS-SEC data rather than traditional protein standards to avoid erroneous results due to shape effects, interaction with the resin and other problems associated with calibrations of this type.
The multimeric state assigned to homooligomeric meprin A was based on the elution volume of the peak maximum. As the concentration of homooligomeric meprin A was increased, the elution volume at which the peak maxima appeared was lowered, therefore the apparent molecular mass of the protein complex increased. The peak maxima had apparent molecular masses that indicated on average, oligomers formed which had between 16 and 78 subunits in 150 mM NaCl. However it is clear that the samples had a broad range of molecular masses and oligomeric states. Thus, some oligomers exist that are composed of less than 16 monomers and some exist that have more than 78 monomers in the multimer. The range of elution volumes seen with latent homooligomeric meprin A was much larger than that for active protein (Table III) (Table III). In addition, the oligomeric state appeared to be independent of the concentration of NaCl from 150 mM to 1 M (data not shown).
Therefore, dimeric species existed under all conditions studied.
It was evident that heterooligomeric meprin A formed two oligomeric species by SEC. These forms were dimers and tetramers based on the SEC-LS data ( Fig. 3E and F;  Fig. 4A and B). These included rings, crescents and spiral chains that are similar to those seen for the mouse homologue (6).
However much longer chain lengths were observed for the latent form of the rat enzyme.
In low salt conditions the latent form of homooligomeric meprin A existed as chains typically about 90-100 nm in length but they could extend up to 400 nm (Fig. 5A). In the presence of 150 mM NaCl, fewer of the extremely long polymers were seen and the majority measured 50-75 nm consistent with the SEC-LS data (Fig. 5B). As presented at higher magnification, the latent protein not only existed as circles and crescents but also as novel tube-like and long spiral-like structures ( Fig. 6; rows 1-3). The tubes measured 30-40 nm in length and approximately 30 nm in width ( Fig. 5C and D). The tube width corresponded to the diameter of the circular forms, and indicated that these are stacks of 3 to 6 rings or highly condensed spirals. The occurrence of tube-like structures was dependent on the presence of 150 mM NaCl.
The latent and active forms of meprin B were observed to be much smaller particles than homooligomeric meprin A ( Fig. 4C and D). The predominant class of averaged images of the latent form of meprin B measured approximately 11 by 12 nm, while that of the active form was slightly smaller at 10 by 10 nm (Fig. 7; A and B (Fig. 4E and F). The most populated averaged images of these particles clearly show a distinctly barrel shape with a symmetric arrangement of two protomers connected by thin bridges (Fig. 7; rows B  There is evidence that membrane-bound meprins associate with other proteins at the membrane, including other proteases (e.g., angiotensin converting enzyme and leucine aminopeptidase) and amino acid transporters (24,32,39). The data thus far indicate that meprin β is the subunit involved in heterophilic interactions, in contrast to meprin α that is clearly involved with homophilic interactions. Heterophilic proteolytic complexes at the plasma membrane could coordinately degrade proteins to produce small peptides, dipeptides and amino acids that can then directly feed to the associated amino acid transporters. This would be an efficient link for proteolysis and recycling of protein building blocks.
It has become increasingly apparent that several key proteolytic processes depend on the formation of very large multiprotein complexes. For example, in apoptosis a central scaffold protein oligomerizes to form the apoptosome (700 to 1,400 kDa) and recruits and activates caspases (40). Also the tricorn protease, an icosahedral capsid of 5 The meprins have emerged as unique proteinases with features/domains common to many other proteins. The differences between the evolutionarily-related meprin α and β subunits will provide insights into elements or motifs that drive oligomerization and this will no doubt have relevance to many protein-protein interactions and physiologic and pathological processes.