α-Macroglobulins Are Present in Some Gram-negative Bacteria

α-Macroglobulins (αMs) are large glycoproteins that have been identified in a wide range of vertebrate and invertebrate species and are mostly thiol ester containing proteinase inhibitors. A recent analysis of bacterial genomes ( Budd, A., Blandin, S., Levashina, E. A., and Gibson, T. J. (2004) Genome Biol. 5, R38 ) identified many α-macroglobulin-like sequences that appear to have been acquired by Gram-negative bacteria from their metazoan hosts. We report the first expression and characterization of such a bacterial α-macroglobulin, that from Escherichia coli. This is also the first α-macroglobulin to be characterized that is predicted to be membrane-anchored. We found that the 183-kDa protein contains an intact thiol ester, is monomeric, and is localized to the periplasmic space. Reaction with proteinase results in limited cleavage within a bait region, rapid activation of the thiol ester, cross-linking to the attacking proteinase or other available nucleophiles, and partial protection of the proteinase against macromolecular substrates. Given these properties and the co-occurrence of the αM gene with one for a repair transglycosylase, this suggests a possible role for bacterial αMs in cell defense following host attack. Such a role would make bacterial αMs appropriate novel targets for antibiotic drugs.

␣-Macroglobulins (␣Ms) are large glycoproteins that have been identified in a wide range of vertebrate and invertebrate species and are mostly thiol ester containing proteinase inhibitors. A recent analysis of bacterial genomes (Budd, A., Blandin, S., Levashina, E. A., and Gibson, T. J. (2004) Genome Biol. 5, R38) identified many ␣-macroglobulin-like sequences that appear to have been acquired by Gram-negative bacteria from their metazoan hosts. We report the first expression and characterization of such a bacterial ␣-macroglobulin, that from Escherichia coli. This is also the first ␣-macroglobulin to be characterized that is predicted to be membrane-anchored. We found that the 183-kDa protein contains an intact thiol ester, is monomeric, and is localized to the periplasmic space. Reaction with proteinase results in limited cleavage within a bait region, rapid activation of the thiol ester, crosslinking to the attacking proteinase or other available nucleophiles, and partial protection of the proteinase against macromolecular substrates. Given these properties and the co-occurrence of the ␣M gene with one for a repair transglycosylase, this suggests a possible role for bacterial ␣Ms in cell defense following host attack. Such a role would make bacterial ␣Ms appropriate novel targets for antibiotic drugs.
The ␣-macroglobulins (␣Ms) 2 are a family of large proteins (monomers of Ͼ1400 residues) present in all types of metazoans (1). Most, although not all, have been found to contain an internal thiol ester formed between the side chains of a cysteine and a glutamine residue 3 residues fur-ther C-terminal. The thiol ester is usually critical for the functioning of the ␣M. In humans the best characterized ␣M is ␣ 2 M, which is an extremely abundant tetrameric plasma glycoprotein composed of 1451 residue subunits, and which acts as a pan-proteinase inhibitor, using a unique trapping conformational change-based mechanism (2). Other human ␣Ms include pregnancy zone protein (3), CD109 (4), and CPAMD8 (5), although these have been far less well characterized. Closely related to ␣ 2 M are the complement proteins C3, C4, and C5, all three of which appear to have evolved from the same primordial gene as ␣ 2 M (6). This relationship is suggestive of a potential role for ␣ 2 M and other ␣-macroglobulins in host defense (7)(8)(9).
Although no ␣M proteins have ever been reported from bacteria, a recent data base search of bacterial genomes identified ␣M-like genes in a wide range of Gram-negative bacteria from a number of different clades. These include proteobacteria, cyanobacteria, spirochetes, and thermophillic bacteria (10). The phylogenetic distribution was, however, uneven and suggestive of acquisition of the gene from the metazoan host, perhaps as a colonization factor. Most intriguingly, the ␣M gene was almost always found in tandem with a gene for a cell wall repair transglycosylase, PBP1C (11), suggesting that the ␣M-like protein, together with the transglycosylase might function in bacterial defense subsequent to breach of the outer cell wall by the antibiotic response of the host.
To better understand the properties and potential functions of such host-acquired bacterial ␣Ms, we have carried out the first expression and characterization of a bacterial ␣M, that from Escherichia coli (which we term ECAM). This is predicted to be a mature protein of 1636 residues, following removal of a 17-residue signal peptide, with a periplasmic localization sequence and an internal thiol ester. Our experimental findings are consistent with these predictions and suggest that the protein might serve in bacterial defense by capturing host proteinases that gain access to the periplasmic space through breach of the outer cell wall. This might make such bacterial ␣Ms attractive novel antibiotic targets, particularly in instances of multidrug resistance. In addition, this characterization of a membrane-associated member of the macroglobulin family may serve as a model for two poorly studied human membrane-associated ␣Ms, CD109, which is predicted to have a C-terminal glycosylphosphatidylinositol anchor (4) and CPAMD8, which has been found to be associated with the membrane through ionic interactions (5).
Cloning of Soluble ECAM-DNA encoding ECAM (residues 2-1636), and thus lacking the first 17-residue signal sequence and the N-terminal cysteine, was amplified from E. coli MG1655 genomic DNA (ATCC). The N-terminal cysteine was not included because its side chain would not be lipidated without the signal sequence. The amplified DNA was cloned into pQE-30 plasmid (Qiagen), resulting in the His tag, MRGSHH-HHHHGSACEL, being linked to the N terminus of ECAM for purification purposes. The plasmid sequence was verified by DNA sequencing. The resulting plasmid was transformed into E. coli SG13009 cells for expression.
Expression and Purification of Soluble ECAM-Cells were grown in 2YT medium (1.6% (w/v) tryptone, 1% (w/v) yeast extract, 0.5% (w/v) NaCl) to an A 600 (attenuance) of 0.8 -1.0 and induced for 5 h at 20°C with 1 mM isopropyl ␤-D-thiogalactoside. The cell lysate, obtained after sonication, was loaded onto an Ni 2ϩ -nitrilotriacetate Superflow column (Qiagen), washed with 50 mM sodium phosphate buffer, 300 mM NaCl, 14.5 mM ␤-mercaptoethanol, and 10 mM imidazole (pH 7.4) and eluted with 50 mM sodium phosphate, 300 mM NaCl, 14.5 mM ␤-mercaptoethanol, and 250 mM imidazole (pH 7.4). Iodoacetamide (20 mM) was added to the eluted protein, to specifically block the free cysteine present in the His tag, and allowed to react for 30 min. Proteins were then dialyzed overnight against 10 mM MES (pH 7.0), loaded onto a Q-Sepharose HP column equilibrated with 20 mM MES, and eluted with a 100 -500 mM NaCl gradient running at a speed of 2 ml/min. ECAM was further purified on a phenyl-Sepharose HP column equilibrated with 20 mM MES (pH 7.0), 1 M Na 2 SO 4 and eluted with a 50 to 100% buffer B (20 mM MES, pH 7.0, 0.1% Tween) gradient running at a speed of 2 ml/min. The resulting pure ECAM was dialyzed against 10 mM MES (pH 6.5), 250 mM NaCl, concentrated to 6 mg/ml, and kept frozen at Ϫ80°C.
Polyacrylamide Gel Electrophoresis-To avoid heat cleavage of ECAM during sample preparation (12), samples were denatured at 37°C by incubating for 30 min in 10 M urea containing 2% SDS, rather than by heating at 95°C. Reactions with proteinase were for 5 min, followed by inhibition with 500 M chloromethylketone or 3,4-dichloroisocoumarin inhibitors before adding loading buffer. For samples requiring breakage of the thioester bond, ECAM was incubated with 200 mM methylamine (pH 8.0) for 10 min.
MALDI-TOF Mass Spectrometry-Matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) analysis was carried out at the Research Resource Center at the University of Illinois, Chicago, IL. ZipTips (Millipore, Billerica, MA), packed with C4 resin, were used to prepare the protein sample for mass spectrometric analysis using sinapic acid as the matrix according to the manufacturer's protocol. Samples were spotted onto a MALDI-TOF target and analyzed by a positive-ion Voyager-DE PRO Mass Spectrometer (Applied Biosystems) equipped with a nitrogen laser. Spectra were externally calibrated. For proteinase-treated samples, ECAM was treated with proteinase for 5 min followed by the addition of 500 M chloromethylketone or 3,4-dichloroisocoumarin inhibitors in 100 mM Tris (pH 8.0). For ECAM and HNE reactions, 5.4 M ECAM was reacted with 5.4 M HNE in 100 mM Tris (pH 8.0) for 5 min then inhibited with 500 M AAPV-CMK. No detectable HNE activity was observed after inhibition.
DTNB Assay for Free SH Groups-Free sulfhydryl groups were assayed by reaction with DTNB (13), following the change in absorbance at 412 nm, measured on a Shimadzu UV-2101PC spectrophotometer equipped with a thermostatted cell holder maintained at 25°C. Time points were fitted to a single exponential using nonlinear least squares fit. The resulting pseudo first-order rate constant was converted to a second-order Proteinase Protection and Inhibition Assays-The ability of ECAM to protect HNE against macromolecular substrates/inhibitors was measured by reacting 10 nM HNE and ECAM (10 -200 nM) together for 5 min in 50 l, followed by addition of 50 or 100 nM ␣ 1 PI. The resulting mixture was added to cuvettes containing 200 M AAPV-p-nitoanilide in 100 mM Tris (pH 8.0) and residual proteinase activity measured spectrophotometrically, monitored at 405 nm on a Shimadzu UV-2101PC spectrophotometer. The activities of 10 nM HNE alone or in the presence of 100 nM ECAM were measured in the same way, but without addition of ␣ 1 PI. All reactions were carried out at 25°C.
ECAM Cellular Localization-For identification of the compartmental localization of ECAM, the full-length protein, including both the signal sequence and the N-terminal cysteine of the mature protein, was expressed. Full-length ECAM cDNA in plasmid pNTR-SD (mobile plasmid collection, ORF:b2520) was obtained from the Japanese National BioResource Project (www.shigen.nig.ac.jp/ecoli/strain/top/top.jsp). The pNTR-SD plasmid has the addition of three amino acids, Met-Arg-Ala, to the N terminus, and two amino acids, Gly-Leu, to the C terminus of ECAM. Expression of ECAM from pNTR-SD plasmid was tested in the E. coli SG13009 cell and found to be inducible by 1 mM isopropyl ␤-D-thiogalactoside at 20°C. Furthermore, ECAM was found to be associated with the membrane by a fractionation study. For this study, E. coli SG13009 cells con-taining pNTR-SD plasmid were grown to A 600 0.5-0.8 and ECAM expression induced at 20°C for 5 h. Expressing cells were sonicated and the supernatant isolated by centrifugation at 20,000 ϫ g for 1 h. The supernatant was further fractionated by ultracentrifugation at 138,000 ϫ g for 1 h to separate the membrane and cytoplasmic fractions. The membrane fraction was dissolved in 2% SDS buffer. All fractions were then subjected to SDS-PAGE analysis.
Confocal Microscopy-Cells, with and without induction of ECAM, were washed 3 times with 10 mM sodium phosphate (pH 8.0) before and after the addition of 5 mM iodoacetamide and reacted for 30 min, to block all free sulfhydryl groups. Cells were then treated with both 1 M IAF (or iodoacetamide to prevent IAF incorporation) and 200 mM methylamine for 10 min to label free sulfhydryl group specifically released by methylamine. Next, cells were washed 3 times with 10 mM sodium phosphate (pH 8.0) to remove excess IAF, and fixed with 4% formaldehyde and washed again with 10 mM sodium phosphate (pH 8.0). Slides were prepared from cells mixed with agarose (0.3%) to keep cells static for microscopy. Cell imaging experiments were done on a Zeiss LSM 510 confocal microscope with excitation filter 488 nm and emission filter 505 nm.
Ultracentrifugation-Samples for analytical ultracentrifugation were prepared in 100 mM Tris (pH 8.0). For the reaction of HNE and ECAM, ECAM was reacted with equimolar HNE in 100 mM Tris (pH 8.0) for 5 min then inhibited with 500 M AAPV-CMK. No detectable HNE activity was observed after inhibition. Native ECAM was treated in the same way. For multisignal experiments, absorbance data at 280 and 250 nm from 3 samples (4 M HNE alone, 2.7 M ECAM alone, and 5.4 M ECAM reacted with 5.4 M HNE) were obtained simultaneously. These multiwavelength data were used for global multisignal analysis to permit location of the component proteins in the sample of the reaction mixture (14). Methylamine-reacted samples were treated the same way for global multisignal analysis. A Beckman Proteomelab XL-I analytical ultracentrifuge, equipped with a Ti-60 rotor was used for sedimentation velocity experiments, with a rotor speed of 50,000 rpm. Radial absorbance scans were collected at 4-min intervals. 5.4 SEDFIT was used to calculate c(s) distributions and SEDPHAT was used to normalize the obtained sedimentation coefficient values to the corresponding values in water at 20°C.
Fluorescence Spectroscopy-Fluorescence experiments were performed on a PTI Quantamaster instrument equipped with double monochromators on both the excitation and emission sides. Cuvettes containing 500 nM ECAM in 100 mM Tris buffer (pH 8.0), either alone or with 200 mM methylamine or 500 nM HNE, were excited at 295 nm and emission spectra recorded from 300 to 400 nm in 4-nm steps at room temperature. Slits of 1 and 8 nm were used for excitation and emission, respectively. For the reaction of ECAM with HNE, correction was made for the fluorescence of HNE recorded on a separate sample under identical conditions. CD Spectroscopy-CD spectra were recorded on a Jasco J-710 spectropolarimeter at 25°C. Samples of 0.1 M ECAM or 0.025 M plasma ␣ 2 M in 20 mM potassium phosphate buffer were run in 2-mm cuvettes.
Differential Scanning Calorimetry-Differential scanning calorimetry measurements were performed on a VP-DSC calorimeter (Microcal). Melting temperatures were measured in 20 mM sodium phosphate buffer (pH 7.4) with 7.1 M ECAM. Data were analyzed with Origin software (Microcal).

Predicted Structural Features of E. coli ␣-Macroglobulin-
The E. coli YfhM gene encodes a 1653-residue pre-protein, containing a 17-residue signal peptide (Fig. 1). The presence of a "lipobox" consensus sequence of LAG-C at the junction between the signal peptide and the cysteine that is destined to become the N terminus of the mature protein indicates that the protein should be post-translationally modified with the addition of a diacylglycerol moiety to the side chain of the cysteine, and of a palmitoyl group in amide linkage to the N terminus (16). In addition, the presence of aspartic acid at position 2, following the cysteine, indicates that the mature lipoprotein should be retained by the inner membrane and so be localized to the periplasmic space (17)(18)(19). A sequence of CLEQ at positions 1170 -1173 fits the consensus requirement of CXEQ for the presence of an internal ␤-cysteinyl-␥-glutamyl thiol ester between the Cys and Gln side chains (1). No disulfide bonds are possible for ECAM, because the only cysteines are the one at the N terminus that becomes lipidated and the one destined to form the thiol ester. This is in marked contrast to mammalian ␣-macroglobulins and complement proteins, which contain many disulfides, both within and between domains (20,21).
Secondary structure prediction was carried out using the PhD algorithm (22,23). Most of the structure was predicted to be stretches of ␤ sheet, with short loop regions between each (Fig. 1). This is consistent with the expected presence of multiple copies of the fibronectin type III domain and one copy of a CUB domain (24), as were found in the x-ray structures of C3 (25)(26)(27) and of selected regions of human ␣ 2 M (28). The fibronectin type III domain, also termed MG (for macroglobulin) in the context of ␣M proteins, is ϳ105 residues containing seven strands in the ␤ conformation, which form a two-sheet sandwich, whereas the CUB domain is a different type of ␤ sandwich (24). The main exception to the all-␤ prediction for ECAM is in the region from 1149 to 1424, which is predicted to be composed of 12 ␣-helices (Fig. 1). Because this region also contains the thiol ester-forming CLEQ sequence, this is exactly as expected for a domain that should be homologous to the thiol ester-containing domains of C3 and C4, each of which is composed of 12 ␣-helices (29,30). In C3 and C4 the sequence CXEQ lies in the turn between the first and second helices. This is the same relationship that is predicted here for ECAM (Fig.  1). In addition, a region of 60 residues between residues 883 and 943 has low prediction for either ␣ or ␤ secondary structure and may serve as the equivalent to the "bait region" that is the primary site of proteolytic attack in other ␣Ms. It contains both acidic (nine) and basic (seven) residues, as well as large and small hydrophobic side chains, thereby providing it with E. coli ␣ 2 -Macroglobulin OCTOBER 17, 2008 • VOLUME 283 • NUMBER 42 potential cleavage sites for proteinases with the most usual specificities.
A significant difference between C3 and ECAM is that 1148 residues precede the predicted thiol ester domain in ECAM, whereas only 962 residues, containing 7 MG domains and half of the CUB domain, precede the thiol ester domain in C3. The latter also includes an ϳ100-residue long region that generates the excisable, ␣-helical C3a fragment. Thus, given the ϳ105 residue size of an MG domain, ECAM may either possess additional MG domains before the thiol ester domain compared with C3, or else some additional type of ␤-sheet-containing domain that is not present in C3. Here it should be noted that the very low sequence homology between equivalent strands in the eight MG domains of C3, which made it impossible to predict the secondary/tertiary structure of C3 prior to its x-ray structure determination (27), makes it impossible here to unambiguously identify the exact number and boundaries of MG domains expected in this region of ECAM. Similarly, the region following the thiol ester domain in ECAM is 213 residues long, whereas in C3 there are 145 residues that cover the second part of the CUB domain and the last fibronectin type III domain (MG8). If a CUB domain, interrupted as it is in C3 and human ␣ 2 M by the thiol ester domain, as well as a homolog of MG8 are also present in ECAM, there may be an additional small domain of about 70 residues in this region in ECAM that accounts for the extra length. Consistent with this, the secondary structure prediction for the last 100 residues of ECAM only poorly fits the pattern expected for an MG domain, with no ␤-sheet predicted for the 30 residues between 1538 and 1569, which is suggestive of a different domain following the last MG domain (Fig. 1).
ECAM Is a Monomeric Multidomain Protein-To produce a soluble form of ECAM for biochemical and biophysical characterization, a modified form of the protein was expressed in which the 17-residue N-terminal lipoprotein signal sequence and the following cysteine were deleted, and a His tag substituted to help with purification. This protein should contain all of the domains present in membrane-localized ECAM, but lack the lipid modifications to the N-terminal cysteine. Overexpression of this construct in E. coli resulted in high level expression of an ϳ180-kDa protein that could be isolated from the soluble fraction of the harvested cells (supplemental Fig. S1). Mass spectrometry confirmed the expected mass of 183 kDa (Fig. 2).
Sedimentation velocity measurements indicated that the protein was a single species, with sedimentation coefficient (s 20,w ) of 5.9 S, consistent with it being a 183-kDa monomer (Fig. 3A). Differential scanning calorimetry showed that, even without stabilizing disulfides, ECAM was a quite robust protein, with a total unfolding enthalpy of ϳ1700 kcal mol Ϫ1 , which gave three distinct unfolding transitions with T m values of 49, 64, and 72°C (Fig. 4). This compares with a single, unresolved transition for human ␣ 2 M with T m of 69°C and unfolding enthalpy of ϳ2300 kcal mol Ϫ1 (31). Each of the three ECAM transitions could be fitted to one or more Gaussians, allowing analysis in terms of the possible number of domains present, by comparing the van't Hoff and calorimetric enthalpies for each transition (32). The ratio ⌬H cal /⌬H van't Hoff corresponds to the number of domains unfolding during a given transition. The results in Table 1 indicate that ECAM might be composed of 11-14 domains. Although this type of prediction is not likely to be completely accurate, it should give a rough estimate of the number of domains present. In this regard, a comparable analysis of human ␣ 2 M carried out well before any structural infor- mation on C3 or C4 was available, suggested that about 12 domains are present in the fast form (31). The expectation now for human ␣ 2 M, based on homology with C3, is that it contains 11 domains, if a putative bait region domain is included.
Additional support for the correct folding of ECAM was provided by the CD spectrum, which was very similar to that of human ␣ 2 M (supplemental Fig. S2), and thus also consistent with the secondary structure prediction (Fig. 1), which indicated similar extents of ␣ and ␤ secondary structure as in human ␣ 2 M.
ECAM Contains an Intact Thiol Ester-A hallmark of members of the ␣-macroglobulin and C3/4/5 family that also contain the CXEQ motif is that they contain an intact, ␤-cysteinyl-␥-glutamyl thiol ester in the native state that is relatively stable. From the structure of C3 it is clear that the reason for the stability of the thiol ester is that, although it occurs at the surface of the thiol ester domain, it is protected from hydrolysis or other nucleophilic attack by domain-domain contacts with the MG8 domain. In particular a pair of tyrosines in MG8 (residues 1439 and 1440 in mature human ␣ 2 M), which are highly conserved in the family, form a hydrophobic pocket around the thiol ester, although this YY motif is absent from ECAM. In the absence of proteolytic activation, and consequent conformational disturbance around the thiol ester, cleavage of the thiol ester requires a high concentration of a potent, small, nucleophile such as methylamine or hydroxylamine. Such cleavage results in generation of a free thiol (33), and in many cases a large conformational change in the ␣-macroglobulin (34).
Reaction of native ECAM with DTNB detected no free SH groups, consistent with the only cysteine in the soluble form of ECAM being in a thiol ester linkage. To test this, ECAM was reacted with 0.2 M methylamine at pH 8.0 in the presence of DTNB and the change in absorbance monitored as a function of time (supplemental Fig. S3). By analogy with other ␣ 2 Ms and complement components, the methylamine should cleave the Panel B, ECAM after reaction with 1 eq HNE. The likely assignment of the proteolysis fragments is indicated (C ϭ C-terminal fragment; N ϭ N-terminal fragment; C ϩ P ϭ C-terminal fragment linked to proteinase). The small peak at 183 kDa is residual unreacted ECAM, whereas that at 130 kDa may derive from heat fragmentation of intact ECAM.  . Thermal denaturation profile of ECAM. Unfolding of 7.1 M ECAM was followed by differential scanning calorimetry. Three distinct peaks were observed (black line), which could be fitted to four Gaussian lines (dashed lines). Thermodynamic analysis of these Gaussians is given in Table 1. OCTOBER 17, 2008 • VOLUME 283 • NUMBER 42 thiol ester and generate a new free SH group. This was found to be the case with ECAM, resulting in a final concentration of SH that was consistent with the presence of one thiol ester per chain (0.9 Ϯ 0.10 from 3 determinations). The second-order rate constant calculated for the reaction of methylamine with the thiol ester was 11.5 M Ϫ1 s Ϫ1 . This compares with a value of 12.5 M Ϫ1 s Ϫ1 determined here for the thiol ester in human ␣ 2 M reacted with methylamine, which is closely similar to values determined elsewhere under similar conditions (35)(36)(37). Thus ECAM appears to contain an intact thiol ester that is protected from nucleophilic attack, even though the YY motif present in other ␣-macroglobulins does not seem to be present. Possibly a comparably hydrophobic surface patch is generated in ECAM by different residues.

E. coli ␣ 2 -Macroglobulin
Thiol Ester Cleavage in ECAM Does Not Cause a Major Conformational Change-Cleavage of the thiol ester in human, although not bovine, ␣ 2 M causes a major change in conformation that results in a 45% enhancement of the endogenous protein fluorescence, a change in sedimentation coefficient from 17.2 to 18.4 (38), and an increase in electrophoretic mobility on non-denaturing PAGE (39). Although ECAM has many more tryptophans than human ␣ 2 M (30 versus 11), no change in tryptophan fluorescence was found between the native and methylamine-treated states of ECAM (supplemental Fig. S4). The sedimentation coefficient also did not change (data not shown) and native PAGE showed no detectable difference in mobility (Fig. 5B). Together these results suggest that, unlike human ␣ 2 M, thiol ester cleavage alone does not greatly alter either the conformation of individual domains or their relative arrangement. In this respect ECAM behaves more like bovine ␣ 2 M (40).
Reaction with Proteinase Is Stoichiometric-␣ 2 M and other metazoan ␣-macroglobulins are efficient proteinase inhibitors that function by physically sequestering the attacking proteinase as a result of massive conformational changes that are initiated by proteolytic cleavage of the macroglobulin within a so-called bait region. The bait region is located approximately in the middle of the polypeptide and, in the case of human ␣ 2 M, contains a sufficient range of different amino acids to make it a suitable "bait" for a wide spectrum of proteinase specificities (41). The efficiency of inhibition is extremely high, such that there is often a stoichiometric relationship between pairs of bait regions cleaved and molecules of proteinase inhibited (42).
Reaction of ECAM with HNE resulted in specific cleavage to give a pair of bands that ran as ϳ80and 100-kDa species, with the 100-kDa species being more intense (Fig. 5A). On non-denaturing PAGE this cleavage corresponded to a complete con-version of the ECAM band into one of much lower mobility, consistent with association with the very basic HNE (Fig. 5B). Approximately 1 eq HNE was required to convert all of the ECAM to the slower moving band on the native gel and to complete the cleavage of the 180-kDa band on the SDS gel. MALDI-TOF of the reaction products showed conversion of the 183-kDa ECAM into two poorly separated species with masses of 102-103 kDa and a third with mass of 79 kDa (Fig.  2B). This is consistent with cleavage within the 883-943 region, which was suggested above to constitute the bait region, to produce an N-terminal fragment of ϳ102 kDa and a C-terminal fragment of ϳ80 kDa. The latter, if cross-linked to the proteinase HNE, would give a fragment of mass ϳ103 kDa, and could thus account for the shoulder at 103 kDa in Fig. 2B. This near coincidence in size of two species would also explain the higher intensity of the ϳ100-kDa band on SDS-PAGE (Fig. 5A), because it would contain all of the N-terminal fragment as well as that fraction of the C-terminal fragment that was covalently cross-linked to HNE, whereas the 79-kDa band would be the remainder of the C-terminal fragment. Consistent with this analysis, a Western blot of the reaction products confirmed that the band at ϳ100 kDa was a composite of two different species with closely similar masses. Thus it reacted with both anti-HNE antibody (consistent with the presence of the C-terminal fragment plus HNE) and with an anti-His tag antibody (N-terminal fragment) (Fig. 6). Furthermore, pretreatment of ECAM with methylamine prior to reactions with HNE abolished the staining of the ϳ100-kDa band with the anti-HNE antibody, as a result of the methylamine reacting with the thiol ester such that it was no longer available for cross-linking to HNE, whereas the anti-His tag antibody still stained, suggesting this band now contained only the N-terminal fragment (Fig. 6, lane 4).  a Peak number corresponds to labels in Fig. 5. b Unfolding temperatures for deconvoluted spectrum in which peak 2 is decomposed into two separate transitions. c Ratio of ⌬H cal /⌬H van't Hoff is a measure of the number of domains unfolding within the transition.
Together, these observations imply that HNE cross-linked to ECAM by reacting to, and therefore requiring, an intact thiol ester bond.
To test that the region of cleavage by HNE represented a true, wide specificity, bait region, other proteinases that the enteric bacterium E. coli might encounter (trypsin, chymotrypsin and porcine pancreatic elastase) were reacted with either methylamine-treated ECAM, to avoid complications arising from cross-linking to the thiol ester (chymotrypsin and PPE), or native ECAM (trypsin reaction). Consistent with analogous cleavage within the 883-943 region, reactions with chymotrypsin and porcine pancreatic elastase produced 2 bands, similar to those that resulted from the reaction of ECAM with HNE, which migrated as ϳ80and 100-kDa species on SDS-PAGE (supplemental Fig. S5, a-c). The trypsin reaction was carried out on native ECAM that still contained an intact thiol ester. In addition to the new lower molecular mass band at ϳ100 kDa, there was also a higher mass band at about 250 kDa that may be a crosslinked species containing only ECAM, generated by reaction of the activated thiol ester of one ECAM with another ECAM molecule. Consistent with this, Western blotting carried out on the trypsinreacted ECAM sample, using antitrypsin antibody, showed that the 100-kDa band again contained proteinase, whereas the ϳ250-kDa band did not (supplemental Fig. S5d).

Proteolysis of ECAM Results in Rapid Cleavage of the Thiol Ester and Conformational
Change-Reaction of ECAM with stoichiometric amounts of proteinase resulted in rapid appearance of free SH groups (one per molecule) as a result of thiol ester hydrolysis (supplemental Fig. S6). Although the reaction was too fast to follow the first 80% of the cleavage, both fitting of the remainder of the time course to a second-order process and an estimate of the rate constant from use of half-lives gave values of the rate constant of about 5 ϫ 10 4 M Ϫ1 s Ϫ1 . This serves to show that proteolysis results in large activation of the thiol ester, analogous to the activation observed for various ␣ 2 Ms and complement components, because it is of the order of 4000-fold faster than methylamine-induced cleavage. This in turn suggests that proteolysis results in a conformational change in ECAM in the vicinity of the thiol ester that renders the latter much more reactive to ambient nucleophiles, as with C3, C4, and other ␣ 2 Ms. This was examined by fluorescence spectroscopy, which showed, however, that cleavage within the bait region led to only ϳ5% change in fluorescence (supplemental Fig. S4). This compares with an enhancement of over 60% for bait region cleavage of human ␣ 2 M by proteinase (38).
Sedimentation velocity measurements, however, showed much clearer evidence for conformational change as a result of reaction of ECAM with proteinase. Thus the monomer increased its sedimentation coefficient from 5.9 to 7.7 S (Fig. 3). In addition, more rapidly sedimenting species were present that may reflect the ability of cleaved ECAM to associate to form dimers and trimers. This appears to be concentration dependent, with more of the larger species present at higher protein concentration (Fig. 3A). Multisignal analysis of the sedimentation data (14), which enabled identification of the location of HNE in the ECAM:HNE sample (see "Experimental Procedures"), revealed that the HNE associated almost completely with the larger species (Fig. 3B).
Although HNE is mostly cross-linked to ECAM, such covalent cross-linking was not necessary for association with ECAM, because the same localization of HNE with the faster sedimenting species was obtained for ECAM preincubated with methylamine and then reacted with HNE (data not shown). In addition, the same large mobility shift on native PAGE was seen upon HNE reaction of ECAM that had been pre-treated with methylamine to cleave the thiol ester (Fig. 5B, lanes 15-19).
Here it is worth noting that such pretreatment, which resulted in cleavage of the thiol ester, left the bait region accessible to proteolysis (Fig. 5A, lanes 6 -9). This independently implies no major conformational change is induced by thiol ester cleavage alone and so makes ECAM more like bovine ␣ 2 M than human ␣ 2 M in this regard (40).
ECAM Inhibits Proteinase against Reaction with Macromolecular Substrates-To determine its ability to inhibit proteinase, ECAM was incubated with HNE and the residual proteolytic activity against both low and high molecular weight substrates measured as described under "Experimental Procedures." No significant inhibition of the proteinase against low molecular weight substrates was found (Fig. 7). To determine whether ECAM could prevent HNE from reacting with macromolecular substrates, HNE:ECAM reaction products were incubated with the irreversible macromolecular proteinase inhibitor ␣ 1 PI (5-fold excess of inhibitor over proteinase), and residual HNE activity toward chromogenic substrate measured (Fig. 7). Here residual catalytic activity quantitatively reflects the extent of protection of the HNE from being able to react with ␣ 1 PI. Reaction of a fixed concentration of HNE (10 nM) with ECAM resulted in partial protection against ␣ 1 PI, with the extent of protection depending on the amount of ECAM used. At a stoichiometric concentration of ECAM only 25% of the HNE was protected, but this increased to 50% when 200 nM ECAM was used (Fig. 7). These values were unaltered when a 10-fold excess of ␣ 1 PI rather than a 5-fold excess was used, indicating that the residual activity of HNE did not trivially result from consumption of the serpin as a consequence of a greatly increased stoichiometry of inhibition (Fig. 7). These results suggest that ECAM can sequester proteinase from access to other macromolecular species, but that this may require formation of higher order ECAM species to effect protection (e.g. the dimers and trimers seen to be exclusively associated with HNE in the analytical ultracentrifugation experiment (Fig. 3B)). Analogously, human ␣ 2 M requires two subunits for inhibition of 1 eq proteinase to give a maximum inhibitory stoichiometry of 2 mol of proteinase per ␣ 2 M tetramer (42).
Full-length ECAM Localizes to the Periplasmic Space of E. coli-To test the prediction that ECAM localizes to the inner membrane of the periplasm, fluorescence microscopy and ultracentrifugation of cellular components were carried out. Fractionation of cellular components of E. coli overexpressing full-length ECAM (containing the leader sequence and N-terminal cysteine and hence presumably lipidated) by ultracentrifugation, as described under "Experimental Procedures," showed that full-length ECAM was inducible and only present in the membrane fraction (supplemental Fig. S7). Fluorescence microscopy of E. coli overexpressing full-length ECAM and that had been treated with IAF alone showed minimal nonspecific fluorescence (Fig. 8A). Treatment of the cells with methylamine prior to reaction with IAF, however, resulted in intense labeling with the fluorophore, with a pattern consistent with localization to the periplasm (Fig. 8B). To confirm that this resulted from reaction with the free SH derived from the thiol ester of ECAM, treatment of the cells with iodoacetamide following reaction with methylamine and prior to reaction with IAF, blocked fluorophore incorporation (Fig. 8C). Similarly, treatment of E. coli that only expressed endogenous levels of ECAM showed relatively low levels of fluorescein incorporation following methylamine and IAF treatment (Fig. 8D).
Significance of ECAM for Bacteria-It is clear from the sequence analyses carried out by Budd et al. (10) that ␣-macroglobulins present in metazoans have not evolved from an ancestral gene present in bacteria, but rather that bacterial ␣-macroglobulins represent instances of opportunistic acquisition of a gene from metazoan hosts. As such, elucidation of the function of bacterial ␣-macroglobulins is unlikely to shed light on the physiological role of human members of the C3/␣ 2 M family. However, the acquisition of such a gene on multiple occasions likely attests to a beneficial function for the protein in bacteria. In addition, the distribution of ␣Ms is restricted to Gram-negative bacteria, and then nearly always in colonizing, rather than free-living species, which suggests a role that enhances their ability to survive in their host environment (10). There are even differences in the presence of an ␣-macroglobulin between closely related species. For example, the liver bacterium Helicobacter hepaticus possesses one, whereas the stomach bacterium Helicobacter pylori does not (10). The question then is what properties do such bacterial ␣-macroglobulins possess that may confer such an advantage?
We have demonstrated that ECAM behaves in many respects very similarly to metazoan ␣Ms. It contains an intact thiol ester that has low reactivity to ambient nucleophiles unless there is stoichiometric reaction with proteinase. Proteolysis occurs for a variety of proteinases within the bait region of ECAM, is relatively rapid, and results in association of a large fraction of the proteinase with dimeric or trimeric ECAM. Given the demonstration that ECAM is localized to the periplasmic space, probably attached to the inner membrane through lipid anchoring of its N terminus, as predicted from its primary structure, this suggests that ECAM could efficiently function to trap foreign proteinases that gain access to the periplasmic space following host-mediated breach of the outer wall of the bacterium. Although the trapping represents incomplete protection from access to macromolecular substrates, it should still serve to localize the proteinase to the immediate vicinity of the membrane-associated ECAM, and hence may serve to limit the proteolytic potential of otherwise freely diffusing proteinase. An indication of the responsiveness of the protein to environmental conditions is that it has been shown to be up-regulated over 3-fold under anaerobic compared with aerobic conditions (43). A second type of bacterial use of ␣Ms involves a very different type of "acquisition" of an ␣-macroglobulin from the host. It has recently been shown that a Gram-positive bacterium, Streptococcus pyogenes, possesses a surface protein, GRAB, that can bind to human ␣ 2 M (44). One use of such a surface-bound host ␣ 2 M could then serve to inhibit host proteinases in the immediate vicinity of the outer cell wall. In keeping with this, GRAB has been found to be an importance virulence factor for group A streptococci (45,46). An alternative, rather intriguing, use for the GRAB-host ␣ 2 M-proteinase complex could be as a means of gaining entry into host cells by using the recognition of the proteinase-complexed host ␣ 2 M to bind to, and be internalized by, the clearance receptors very low density lipoprotein receptor or low density lipoprotein receptor-related protein (LRP) of the host (47). However, for ECAM, the likely periplasmic location of the protein, and the absence of the necessary lysine motif that is required for binding to LRP (there being no lysines in the region from residue 1535 to 1569 (Fig. 1)) (48), makes such a use for ECAM less likely.
The frequent occurrence of a bacterial ␣M gene in tandem with that of a cell-wall repair transglycosylase, PBP1C, is consistent with such a role in defense against host attack in Gramnegative bacteria (10). As a result, bacterial ␣Ms such as ECAM could be useful novel targets for anti-microbial drug development, particularly for multidrug-resistant Gram-negative bacteria such as pathogenic strains of E. coli and Pseudomonas aeruginosa, where resistance has been developed against drugs that target more orthodox bacterial proteins (49).