Recombinant production of human α2-macroglobulin variants and interaction studies with recombinant G-related α2-macroglobulin binding protein and latent transforming growth factor-β2

α2-Macroglobulins (α2Ms) regulate peptidases, hormones and cytokines. Mediated by peptidase cleavage, they transit between native, intact forms and activated, induced forms. α2Ms have been studied over decades using authentic material from primary sources, which was limited by sample heterogeneity and contaminants. Here, we developed high-yield expression systems based on transient transfection in Drosophila Schneider 2 and human Expi293F cells, which produced pure human α2M (hα2M) at ~1.0 and ~0.4 mg per liter of cell culture, respectively. In both cases, hα2M was mainly found in the induced form. Shorter hα2M variants encompassing N-/C-terminal parts were also expressed and yielded pure material at ~1.6/~1.3 and ~3.2/~4.6 mg per liter of insect or mammalian cell culture, respectively. We then analyzed the binding of recombinant and authentic hα2M to recombinant latent human transforming growth factor-β2 (pro-TGF-β2) and bacterial G-related α2M binding protein (GRAB) by surface plasmon resonance, multiple-angle laser light scattering, size-exclusion chromatography, fluorogenic labelling, gel electrophoresis and Western-blot analysis. Two GRAB molecules formed stable complexes of high affinity with native and induced authentic hα2M tetramers. The shorter recombinant hα2M variants interacted after preincubation only. In contrast, pro-TGF-β2 did not interact, probably owing to hindrance by the N-terminal latency-associated protein of the cytokine.

; and the mature growth factor moiety (TGF-β 2 ). Critical residues in LAP are C 24 , which is involved in the binding of LTBP, and R 302 , required for furin cleavage. Mature TGF-β 2 segment A 343 -Y 367 is involved in hα2M binding, important and critical residues are indicated by a grey (A 347 and A 349 ) and a black star (W 354 ), respectively. (ii) Human pro-TGF-β 2 fusion proteins produced with plasmids pIEx and pCMV-Sport6. The AKH signal peptide sequence, the Kozac (Koz), the mouse Igκ-chain leader sequence, affinity tags (His 6 and Strep) and restriction sites are graphically represented. through a currently unknown mechanism 10,12,13 . What is known is that hα 2 M positions E 753 , E 737 and D 742 within segment V 723 -T 761 (numbering according to UniProt [UP] entry P01023) are involved in TGF-β 1 binding 14 and that induced hα 2 M binds the cytokine with higher affinity than the native inhibitor 14 .
The functional and structural properties of hα 2 M are exploited by pathogens such as Streptococcus pyogenes (group A streptococci), which forms stable interactions with hα 2 M by a surface protein, the G-related α 2 M-binding protein (GRAB 15,16 ). This 23-kDa protein consists of a Gram-positive membrane anchor motif, a variable number of 28-residue repeats, and a highly-conserved N-terminal domain responsible for the interaction with hα 2 M (Fig. 1B). By recruiting native hα 2 M to the membrane, GRAB provides S. pyogenes with a mechanism to inhibit host peptidases, which protects bacterial surface structures and facilitates progressive dissemination in the infected tissue 15 .
These interactions have only been preliminary characterized 17,18 , and the mechanisms are still unknown. To shed light on them, we developed eukaryotic expression systems of hα 2 M variants and purified the authentic protein from blood. We further used these proteins to study complex formation with GRAB and pro-TGF-β 2 by several biophysical approaches.

Materials and Methods
Construct preparation. Constructs spanning fragments of the gene coding for hα 2 M, namely full-length hα 2 M and its N-and C-terminal parts (N-hα 2 M and C-hα 2 M; for details on constructs, plasmids, vectors and primers, see Table 1 and Fig. 1), and the coding sequence for GRAB from Streptococcus pyogenes serotype M1 (UP Q7DAL7) were amplified with primers that introduced either restriction sites for directional cloning or overhangs for restriction-free cloning. The vectors used were pCri-8a 19 for bacterial expression, pIEx (Novagen) for expression in Drosophila melanogaster Schneider 2 embryonic cells (S2; Gibco), and pCMV-Sport 6 (Thermo Scientific) for expression in human Expi293F ™ cells (Gibco). Polymerase chain reaction (PCR) primers and DNA modifying enzymes were purchased from Sigma-Aldrich and Thermo Scientific, respectively. PCR was performed using Phusion High Fidelity DNA polymerase (Thermo Scientific) according to the manufacturer's instructions and following a standard optimization step by thermal gradient in each reaction. Mutants were generated by a modified version of the previously described procedure 20 . DNA was purified with the OMEGA Biotek Purification Kit according to the manufacturer's instructions, and all constructs were verified by DNA sequencing.
Cell-culture growth. S2 cells were cultivated in TubeSpin bioreactor tubes (TS50 for 5-to-10-mL cultures and TS600 for 100-to-200-mL cultures; Techno Plastic Products AG) as previously described 21 . Cells were passaged three times per week to a final density of 4 × 10 6 cells/mL. The cultures were incubated at 28 °C in a shaker (Brunswick Scientific Innova) under agitation at 220 rpm.
Expi293F cells were cultivated in 125-mL or 1000-mL polycarbonate Erlenmeyer flasks (FPC0125S and FPC1000S, respectively; Tri Forest Labware) for 25-to-30-mL and 100-to-250-mL cultures, respectively. Cells were subcultured three times per week to a final density of 0.3-0.5 × 10 6 cells/mL and kept in suspension at 150 rpm in a Multitron Cell Shaker Incubator (Infors HT) at 37 °C in a modified atmosphere (8% CO 2 and 85% of relative humidity). Cell densities and viability were determined by the trypan blue exclusion test 22 . Cell-culture transfection. Linear 25-kDa polyethylenimine (PEI; Polysciences Europe GmbH) was prepared in Milli-Q water at a concentration of 1 mg/mL and pH 7.0. The solution was filter-sterilized and stored at −20 °C. Plasmid DNA was produced in Escherichia coli DH5α cells, purified with the GeneJET Plasmid Maxiprep Kit (Thermo Scientific), and stored at −20 °C in sterile Milli-Q water at 1 mg/mL.
For transfection, S2 cells were centrifuged and resuspended in prewarmed fresh medium to a cell density of 15 × 10 6 cells/mL. A mixture of 0.6 μg DNA (see Fig. 1 and Tables 1) and 2 μg PEI per 1 × 10 6 cells and per prewarmed transfection volume was pre-incubated for 15-20 min at room temperature and then added dropwise to the cell cultures. These were further incubated for 1 hour at 28 °C and 220 rpm, subsequently diluted with prewarmed fresh medium to 5 × 10 6 cells/ml and harvested after seven days for protein purification.
For mammalian cultures, Expi293F cells were transfected at a cell density of 1 × 10 6 cells/mL with a mixture of 1 mg of DNA (see Table 1) and 3 mg of PEI in 20 mL of Opti-MEM Medium (Gibco) per liter of expression medium. The DNA-PEI mixture was incubated at room temperature for 15-20 min and then added dropwise to the cell cultures, which were harvested after three days for protein purification.  Mature TGF-β 2 in Expi293F cells. The coding gene extracted from the parental plasmid was inserted into the pCMV-Sport 6 vector by restriction-free cloning between the Ig κ leader sequence and the C-terminal histidine-tag. Table 1. Constructs, primers, plasmids and proteins. All constructs are for extracellular expression of the respective proteins. *Restriction-site sequences and overhangs for restriction-free cloning are underlined. **Peptide sequence of the expressed protein after fusion-tag removal. Amino acids derived from the construct are in bold. See also Fig. 1. ***Fused tags at the carboxy-terminus (C-t) or the amino-terminus (N-t). AKH, adipokinetic hormone; TEV, tobacco-etch virus peptidase; Ig κ, immunoglobulin κ. (2019) 9:9186 | https://doi.org/10.1038/s41598-019-45712-z www.nature.com/scientificreports www.nature.com/scientificreports/ 30,000 × g for 1 hour, and the supernatant containing GRAB was kept for subsequent purification steps. For the hα 2 M variants produced in S2 and Expi293F systems, cells were removed by centrifugation at 2,800 × g for 20 min and the supernatant was used for subsequent purification steps.
Supernatants containing the proteins of interest were incubated for 20 min (expression in insect cells) or 1 hour (expression in mammalian cells) with nickel-nitrilotriacetic acid resin (Ni-NTA; Invitrogen), which was subsequently loaded onto an open column for batch purification (Bio-Rad), washed extensively with buffer A plus 20 mM imidazole, and eluted with buffer A plus 300 mM imidazole (direct Ni-NTA). For GRAB, eluted samples were then dialyzed overnight against buffer A plus 1 mM 1,4-dithio-DL-threitol (DTT) in the presence of His 6 -tagged TEV at a peptidase:protein weight ratio of 1:100 and 1 mM DTT. The resulting cleavage left additional residues (glycine-alanine-methionine) at the N-terminus of the target proteins due to the cloning strategy (see Table 1). Digested samples were passed several times through Ni-NTA resin previously equilibrated with buffer A plus 20 mM imidazole to remove His 6 -tagged molecules and the flow-through containing untagged GRAB was collected (reverse Ni-NTA). www.nature.com/scientificreports www.nature.com/scientificreports/ In all cases, proteins eluted from direct and reverse Ni-NTA chromatographies were dialyzed overnight against buffer B (20 mM Tris-HCl, 5 mM sodium chloride, pH 7.5) and further purified by ionic-exchange chromatography (IEC) on a TSKgel DEAE-2SW column (TOSOH Bioscience) equilibrated with buffer B. A gradient of 2-30% buffer C (20 mM Tris-HCl, 1 M sodium chloride, pH 7.5) was applied over 30 mL, and samples were collected and pooled. Subsequently, each pool was concentrated by ultrafiltration and subjected to size-exclusion chromatography (SEC) in Superdex 75 10/300 (GRAB and pro-TGF-β 2 ), Superdex 200 10/300 (N-hα 2 M and C-hα 2 M) or Superose 6 10/300 (full-length recombinant hα 2 M) columns (GE Healthcare Life Sciences) in buffer D (20 mM Tris-HCl, 150 mM sodium chloride, pH 7.5). Strep-tagged GRAB was purified by affinity chromatography with Streptactin ® XT Superflow Suspension resin (IBA Life Sciences) and eluted with buffer E (100 mM Tris·HCl, 150 mM sodium chloride, pH 8.0) at a further 50 mM in biotin. IEC and SEC purification steps followed as above.
Authentic full-length hα 2 M was isolated from blood plasma from individual donors and purified essentially as described previously 17,24,25 . Briefly, plasma was subjected to sequential precipitation steps with 4-12% PEG 4,000, and the final precipitate containing hα 2 M was reconstituted in 20 mM sodium phosphate at pH 6.4. Partially purified hα 2 M was captured with a zinc-chelating resin (G-Biosciences), washed with buffer F (50 mM sodium phosphate, 250 mM sodium chloride, pH 7.2) plus 10 mM imidazole and eluted in the same buffer plus 250 mM imidazole and 100 mM EDTA. The protein was first passed through a PD10 desalting column (GE Healthcare Life Sciences) previously equilibrated with 20 mM HEPES, pH 7.5 and then subjected to an IEC step in a Q Sepharose column (2.5 × 10 cm; GE Healthcare Life Sciences), previously equilibrated with 15% buffer G (20 mM HEPES, 1 M sodium chloride, pH 7.5). A gradient of 20-30% buffer G was applied for 150 min and fractions were collected. Collected samples were dialyzed overnight against buffer H (20 mM sodium phosphate, 5 mM sodium chloride, pH 7.4) and further purified by IEC in a TSKgel DEAE-2SW column, previously equilibrated with buffer H. A gradient of 7-20% buffer I (20 mM sodium phosphate, 1 M sodium chloride, pH 7.4) was applied over 30 mL, and samples were collected and pooled. Subsequently, each pool was concentrated and subjected to a final polishing step by SEC in a Superose 6 10/300 column in buffer J (20 mM sodium phosphate, 150 mM sodium chloride, pH 7.4).
Protein identity and purity were assessed by 10-15% Tricine sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE 26 ) stained with Coomassie Brilliant Blue, peptide mass fingerprinting of tryptic protein digests, N-terminal sequencing through Edman degradation, and mass spectrometry. The latter three were carried out at the Protein Chemistry Service and the Proteomics Facilities of the Centro de Investigaciones Biológicas (Madrid, Spain). Ultrafiltration steps were performed with Vivaspin 15 and Vivaspin 500 filter devices of 10-to 50-kDa cut-off (Sartorius Stedim Biotech). Protein concentrations were estimated by measuring the absorbance at 280 nm in a spectrophotometer (NanoDrop) and applying the respective theoretical extinction coefficients. Concentrations were also measured by the BCA Protein Assay Kit (Thermo Scientific) with bovine serum albumin fraction V (BSA; Sigma-Aldrich) as a standard. Induced hα 2 M was obtained by treating native hα 2 M in buffer D with 200 mM methylamine hydrochloride for one hour at room temperature. Subsequently, the sample was dialyzed against the same buffer D.
Human pro-TGF-β 2 (UP P61812) constructs (see Table 1 and Fig. 1C) were produced in S2 and Expi293F cells and purified as reported elsewhere 8 . Production of mature TGF-β 2 with a C-terminal His 6 -tag (see Table 1) was assayed with the insect and human systems, which included harvesting periods of seven and three days, respectively. Supernatants were collected after the centrifugation at 2,800 × g for 20 min and the purification steps were, first a direct Ni-NTA in buffer A plus 20 mM imidazole for the wash step, and plus 300 mM imidazole for the elution; and finally purified by SEC with a Superdex 75 10/300 column in buffer D. protein labeling. GRAB and pro-TGF-β 2 were labelled with fluorogenic sulfosuccinimidyl-7-amino -4-methylcoumarin-3-acetate (Sulfo-NHS-AMCA; Thermo Scientific) according to the manufacturer's instructions with a 10-15 molar excess of reagent over protein in buffer J for 1 hour at room temperature. Thereafter, the proteins were extensively dialyzed against buffer J to remove non-reacted dye. To assess binding, labelled GRAB or pro-TGF-β 2 were mixed with authentic hα 2 M (native and induced) or recombinant fragments N-hα 2 M and C-hα 2 M at a 4:1 molar ratio, incubated in buffer J for two hours at 37 °C, and analyzed by 10% native PAGE 27 . Gel fluorescence was visualized in a gel reader (G:BOX F3 Gel Doc System, Syngene) and the fluorescence was measured (λ ex = 345-350 nm and λ em = 440-460 nm). Negative controls (unlabeled proteins) were included in each experiment. After fluorescence detection, native gels were stained with Coomassie Brilliant Blue (Thermo Scientific) to detect the negative controls.

Multi-angle laser light scattering. Multi-angle laser light scattering in a Dawn Helios II apparatus (Wyatt
Technologies) coupled to a SEC Superose 6 10/300 column (SEC-MALLS) equilibrated in buffer J at 25 °C was performed at the joint IBMB/IRB Crystallography Platform, Barcelona Science Park (Catalonia, Spain) to analyze binding of GRAB or pro-TGF-β 2 to native or induced authentic hα 2 M at a molar ration of 4:1. ASTRA 7 software www.nature.com/scientificreports www.nature.com/scientificreports/ (Wyatt Technologies) was used for data processing and analysis, for which a dn/dc value typical for proteins (0.185 mL/g) was assumed. All experiments were performed in triplicate.  www.nature.com/scientificreports www.nature.com/scientificreports/ BSA. His 6 -tagged proteins were detected by immunoblot analysis using the monoclonal His-HRP Conjugated Antibody (Santa Cruz Biotechnology) diluted 1:5,000 in PBS plus 0.1% Tween 20. Strep-tagged proteins were detected with the Streptavidin-Peroxidase Conjugated Antibody from Streptomyces avidinii (Sigma-Aldrich) diluted 1:1,000 in PBS plus 0.1% Tween 20 and 1% BSA. Complexes were detected using an enhanced chemiluminescence system (Super Signal West Pico Chemiluminescent; Pierce) according to the manufacturer's instructions. Membranes were exposed to Hyperfilm ECL films (GE Healthcare Life Sciences). proteolytic inhibition assays. Inhibition assays against protein substrates were performed in a microplate fluorimeter (Infinite M200, TECAN) in 200 μL reaction volumes with the fluorescence-based EnzCheck Assay Kit containing BODIPY FL-casein (λ ex = 505 nm and λ em = 513 nm) as fluorescein conjugate (Invitrogen) at 10 μg/ mL in buffer D. Inhibition was measured after preincubation of a two-fold molar excess of authentic or recombinant hα 2 M with trypsin (0.25 μg) for 15 min at room temperature. The substrate was added to the reaction mixture and the residual tryptic activity was measured over a period of two hours.

Thiol quantification. Detection of free sulfhydryl groups was performed with the Fluorometric Thiol Assay
Kit (ab112158 assay; Abcam) following the manufacturer's instructions and using glutathione as a standard for the dose response curve. The fluorescent signal was measured in a microplate fluorimeter (Infinite M200, TECAN) at λ ex = 490 nm and λ em = 520 nm in 96-well plates containing 100 μL reaction volumes (50 μL of assay reaction mixture plus 50 μL of glutathione-standard or test samples) in duplicate. Fluorescence was measured after preincubation of authentic hα 2 M (0.39 μM) or C-hα 2 M obtained from human cells (1.6 μM), with or without treatment with methylamine for 10, 20, 30, 45 and 60 min, at room temperature.
surface plasmon resonance and kinetic data analysis. The binding kinetics (association and dissociation) and affinity (complex formation at the equilibrium) of GRAB or pro-TGF-β 2 (ligands) with native authentic hα 2 Table 4. Equilibrium constants of the interaction between native or induced authentic hα 2 M and GRAB. Values were derived from the corresponding plot of steady state response against concentration assuming a 1:1 model (one GRAB molecule per hα 2 M dimer), see Fig. 3B. www.nature.com/scientificreports www.nature.com/scientificreports/ described previously 22 . Subsequently, Strep-tagged GRAB (at 9.7 nM) or pro-TGF-β 2 (at 19.0 nM) in HBNS buffer (10 mM HEPES, 150 mM sodium chloride, pH 7.4) were immobilized at low RU density on different flow cells of the chip by virtue of the strong interaction between the Strep-tag and streptactin at 5 μL/min for 24 sec at 37 °C. To monitor association, the immobilized ligands were then exposed to the analytes at different concentrations in HBNS (4-600 nM for native and induced authentic hα 2 M; 75-1,200 nM for N-hα 2 M and C-hα 2 M), which were injected at 30 μL/min for 120-240 sec at 37 °C. Thereafter, HBNS was injected for analyte dissociation from the immobilized ligands for 90-300 sec. To dissociate bound ligands and regenerate the chip surface, 3 M guanidine hydrochloride was injected at 30 μL/min for 30 sec after each cycle. These experiments were double referenced by keeping the first flow cell without ligand, and by an injection step at analyte concentration zero. The affinity analysis was performed by plotting binding responses in the steady-state region of the sensorgrams (R eq ) against analyte concentrations to determine the overall equilibrium dissociation constant (K D ). Sensorgrams were analyzed with the BIAEVALUATION program v. 3.0 (GE Healthcare Life Sciences) and fitted to a 1:1 Langmuir interaction model. The likelihood of fitting was assessed through the χ 2 statistical parameter 28 .
In a separate qualitative experiment, ligands GRAB (at 120 nM) and pro-TGF-β 2 (at 950 nM) were premixed with the analytes at different concentrations (2-150 nM for native and induced authentic hα 2 M; 38-600 nM for N-hα 2 M and C-hα 2 M) and incubated for one hour at 37 °C. Subsequently, the mixtures were injected at 15 μL/ min at 37 °C according to a published multicycle method 29 . The binding was measured through the increase in RU after injection of the premixes and the stability of the resultant complexes through their elution with buffer HBNS at a flow rate of 30 μL/min. Ligand solutions without analyte were used as negative controls of complex formation and the sensor surface was regenerated after each sample injection.

Results and Discussion
Biochemical characterization of the recombinant proteins. Authentic hα 2 M has been routinely isolated from blood serum, where it is found at an excess of 2-4 mg/mL but is rather heterogeneous as to conformational state, glycosylation and presence of contaminants 17,18 . Native recombinant hα 2 M was obtained from immortalized myelogenous leukemia cell line K-562 but the yield was not reported 14 . Therefore, efforts were made here to develop a system for heterologous expression of the protein with high yield, purity and homogeneity, as well as the necessary flexibility to engineer the protein at will. Full-length hα 2 M with a C-terminal His 6 -tag was expressed in S2 insect cells using a standard transfection protocol 30 and the signal peptide of the adipokinetic hormone (AKH) for secretion to the extracellular environment. After seven days of expression and harvesting of the supernatant, the protein was purified by affinity chromatography, IEC and SEC steps with yields of up to ~1.0 mg of pure protein per liter of expression medium (Fig. 2A). The protein migrated as a tetramer of ~690 kDa according to SEC (data not shown). Its electrophoretic mobility in native-PAGE was similar to that of induced authentic hα 2 M (Fig. 2B), which migrates faster than the native protein 31 . Chemical treatment with methylamine, which mimics the transition from native to induced hα 2 M by opening the reactive thioester bond to produce a free cysteine 31 , did not have any effect on protein mobility. Consistently, the protein could not inhibit trypsin activity against a fluorogenic protein substrate, even at 10-fold molar excess. We conclude that recombinant hα 2 M produced in insect cells was in the induced form, which does not permit the physiological entrance and entrapment of attacking peptidases 4 , similarly to a previous report of a baculovirus expression system 32 . Moreover, the thioester bond was either not formed or it was opened after secretion into the extracellular environment by nucleophiles from the expression medium. Unfortunately, we could not evaluate this possibility as the composition of the commercial medium that was used is not available. However, the latter hypothesis seems more plausible given that it is reported that insects can produce thioester-containing proteins 33 .
We next developed a transient expression system based on Expi293F cells, which derived from the HEK293 human embryonic kidney cell line and were cultured and harvested at 37 °C for three days. The protein was furnished with the leader sequence of mouse immunoglobulin κ (Ig κ) for secretion and produced ~0.4 mg of pure protein per liter of expression medium ( Fig. 2A). The protein migrated as a tetramer in SEC and showed electrophoretic mobility in native-PAGE between native and induced authentic hα 2 M (Fig. 2B). Consistently, its capacity to inhibit trypsin was 35% of native authentic hα 2 M. Together, these data indicate that the recombinant protein is partly native but mainly induced. Previous studies had indicated that thioester formation is a spontaneous process triggered by the packing energy of the polypeptide chain during folding in mammals 34 . Therefore, the limited ability of the Expi293F system to produce native protein was attributed, as in the insect cell system above, to the expression medium rather than to the lack of crucial cell machinery for proper thioester bond formation.
Then we expressed shorter variants of hα 2 M in the insect and mammalian systems ( Fig. 2A). N-hα 2 M spanned from macroglobulin-like (MG) domain 1 (MG1) to MG7 in the insect cell system and from MG1 to MG6 in the mammalian system. C-hα 2 M ranged from MG7 to the C-terminal receptor binding domain in both systems ( Fig. 1A and Table 1). Expression of N-hα 2 M yielded ~1.6/~3.2 mg per liter of insect and mammalian cell culture, respectively, while the values for C-hα 2 M were ~1.3/~4.6 mg. N-hα 2 M formed a dimer of ~170 kDa due to the presence of an intermolecular disulfide bond (C 278 -C 431 ), which is also required for dimerization of the authentic full-length protein (Fig. 1A). Consistently, the protein migrated as a monomer of ~85 kDa in the presence of reducing agents. In turn, C-hα 2 M was monomeric (~75 kDa) and treatment with methylamine did not affect the content of free cysteines or electrophoretic mobility in native-PAGE (Fig. 2B). To follow this up, we qualitatively assayed the content of free sulfhydryl groups by a fluorometric thiol assay kit, which gave a strong fluorescent signal for both the untreated and methylamine-treated C-hα 2 M samples. This contrasted with native full-length authentic hα 2 M, which gave no significant signal, and was similar to methylamine-induced authentic hα 2 M, which likewise gave a strong signal. These assays indicated that the thioester bond was opened in C-hα 2 M as mentioned above for the full-length recombinant variant, possibly owing to a nucleophilic component of the undisclosed cell-growth medium. (2019) 9:9186 | https://doi.org/10.1038/s41598-019-45712-z www.nature.com/scientificreports www.nature.com/scientificreports/ The insect and mammalian systems were also assayed for expression of mature human His 6 -tagged TGF-β 2 ( Table 1), but without noticeable yields. Therefore, full-length pro-TGF-β 2 encompassing LAP and mature TGF-β 2 (see Fig. 1C) was expressed and purified in Expi293F cells as described elsewhere 8 , with a final yield of ~2.7 mg and ~2.3 mg of N-terminally octahistidine-tagged and Strep-tagged forms, respectively, per liter of mammalian cell culture ( Fig. 2A). The protein migrated as a dimer of ~110 kDa in SEC, which indicates that the characteristic disulfide bonds were formed between the LAP and the mature TGF-β 2 moieties. The purified protein was partially cleaved before residue A 303 by host peptidases. Subsequent treatment with the physiological activating endopeptidase furin produced a homogenously cleaved species consisting of LAP associated with the mature cytokine ( Fig. 2A). Under physiological conditions, TGF-β 2 maturation is a complex process that involves a cascade of events under participation of several proteins that interact with the initial complex of pro-TGF-β 2 and the latent TGF-β binding protein (LTBP). LTBP participates as a localizer of pro-TGF-β 2 to the extracellular matrix, whereas LAP senses the changes and releases mature TGF-β 2 11 . Previous studies with a Chinese hamster ovary cell expression system benefited from the sensitivity of the LAP domain towards denaturing conditions at very low pH to separate it from mature TGF-β 2 11,35 . In our case, this was unsuccessful, probably due to different post-translation modifications introduced by Expi293F cells in the highly glycosylated LAP 8,36 .
Finally, full-length GRAB was expressed without the cell-wall anchoring region (Fig. 1B) in a bacterial system yielding ~4 mg of pure protein per liter of expression medium after affinity chromatography, IEC and SEC steps ( Fig. 2A). The protein migrated as a ~55-kDa species in SEC and as a ~33-kDa species in SDS-PAGE, but the values determined by SEC-MALLS (15.5 kDa; Table 2) were closer to the theoretical mass (15.8 kDa). We attribute this abnormal migration, which was described previously 11 , to the highly unstructured character of the protein.
Interaction analysis of hα 2 M and GRAB. Interaction of streptococci with hα 2 M has been reported to be highly specific 15,16 . Group A, G and C streptococci all bind the native form, whereas only the latter interact with the induced form. This result was attributed to the types of surface proteins, which are specific for each strain. GRAB is found in group A streptococci, and we studied its interaction with native authentic hα 2 M by surface plasmon resonance. GRAB was immobilized as a ligand through a Strep-tag on a chip with covalently bound streptavidin. In a multicycle experiment, saturation of the ligand was reached with the two highest analyte concentrations, which gave on-and off-rate kinetic constants and results from affinity analysis (Fig. 3A,B). From the sensorgrams during the sequential injections of different analyte concentrations, we observed fast association and slow dissociation of hα 2 M from GRAB, which indicated stable complex formation. Therefore, the ligand was removed in a regeneration step to make sure that all bound hα 2 M was eliminated between injections with different analyte concentrations. The group of curves in Fig. 3A,B were fitted to a 1:1 Langmuir interaction model. These calculations revealed a χ 2 value < 10% of R max , which is indicative of a good fit. Consistently with the sensorgrams, the association rate constant (k a ) and the dissociation rate constant (k d ) were 1.32 × 10 5 M −1 s −1 and 1.90 × 10 -3 s −1 , respectively, with an estimated dissociation halftime (t 1/2 = ln2/k off ) of 365 sec. The equilibrium dissociation constants (K D ) from the kinetic and affinity analysis were 1.43 × 10 −8 M and 3.45 × 10 −8 M (Tables 3  and 4), respectively, which indicates high affinity and stable complex formation. The complex was also detected by SDS-PAGE and native-PAGE employing fluorophore-labelled GRAB (Fig. 4). Finally, SEC-MALLS analysis ( Fig. 3D and Table 2) showed a molecular mass difference of 27.3 kDa over free hα 2 M, which corresponds to 1.7 molecules of GRAB. Hence, we assume that two molecules of GRAB bind one hα 2 M tetramer.
Under a similar experimental setup, methylamine-induced authentic hα 2 M was injected over immobilized GRAB to reach equilibrium and saturation, which enabled analysis by affinity. The affinity data permitted calculation with confidence (χ 2 < 10% of R max ) of the equilibrium dissociation constant (9.46 × 10 −8 M), which was three times higher than that of native hα 2 M (Fig. 3A,B and Table 4). This is consistent with published results, which indicated that GRAB shows preference for native over protease-induced hα 2 M 16 . The complex was likewise analyzed by SDS-PAGE and native-PAGE with fluorophore-labelled GRAB (Fig. 4). The results showed an increase in the molecular mass of 26.5 kDa over noncomplexed induced hα 2 M, which is equivalent to the results for native hα 2 M.
To map down the region of hα 2 M engaged in GRAB binding, we repeated the above experiments with N-hα 2 M and C-hα 2 M. In a similar multicycle experimental setup, we could not detect any interaction. However, previous incubation of the proteins at 37 °C for one hour apparently enabled complex formation. Protein remained complexed over time after injection and washing of the chip (Fig. 3B), but in this case we could not determine the affinity constants due to the experimental setup. The complexes were subsequently evaluated in native-PAGE using fluorophore-labelled GRAB (Fig. 3D). In this case, we detected the interaction of GRAB with C-hα 2 M but not with N-hα 2 M.
Interaction analysis of hα 2 M and pro-tGF-β 2 . Previous biochemical data had revealed that hα 2 M binds TGF-β 2 mainly through a mature cytokine segment spanning residues A 343 -Y 367 , in which W 354 plays a major role 10 . No data have been reported on the role of LAP. However, inspection of the crystal structure of homologous pro-TGF-β 1 (see Protein Data Bank code 3RJR 37 ) reveals that the interacting segment is partially shielded by LAP. Other studies employing a library of overlapping glutathione S-transferase fusion proteins ascribed the potential binding site for TGF-β 1 to segment V 723 -T 761 of hα 2 M 38 , which was subsequently narrowed down to E 737 -V 756 employing synthetic peptides 39 . However, further details on the mechanism are unknown. To further shed light, we set out to characterize binding of pro-TGF-β 2 to hα 2 M. We checked the interaction by surface plasmon resonance in multicycle experiments with immobilized pro-TGF-β 2 as ligand but could not detect complex formation. Only after analysis by native-PAGE using fluorophore-labelled pro-TGF-β 2 we observed interaction with native authentic hα 2 M but not with the induced form or the short variants (Fig. 4). Given that the pro-TGF-β 2 sample contained a mixture of cleaved and intact protein, we assayed N-terminally His 6 -tagged pro-TGF-β 2 with native hα 2 M in native-PAGE followed by Western blotting. The two proteins were not co-migrating (data not shown). Thus, we conclude that LAP prevents hα 2 M from binding mature TGF-β 2 as suggested by structural studies on pro-TGF-β 1 .