The Compact and Biologically Relevant Structure of Inter-α-inhibitor Is Maintained by the Chondroitin Sulfate Chain and Divalent Cations*

Inter-α-inhibitor is a proteoglycan of unique structure. The protein consists of three subunits, heavy chain 1, heavy chain 2, and bikunin covalently joined by a chondroitin sulfate chain originating at Ser-10 of bikunin. Inter-α-inhibitor interacts with an inflammation-associated protein, tumor necrosis factor-inducible gene 6 protein, in the extracellular matrix. This interaction leads to transfer of the heavy chains from the chondroitin sulfate of inter-α-inhibitor to hyaluronan and consequently to matrix stabilization. Divalent cations and heavy chain 2 are essential co-factors in this transfer reaction. In the present study, we have investigated how divalent cations in concert with the chondroitin sulfate chain influence the structure and stability of inter-α-inhibitor. The results showed that Mg2+ or Mn2+, but not Ca2+, induced a conformational change in inter-α-inhibitor as evidenced by a decrease in the Stokes radius and a bikunin chondroitin sulfate-dependent increase of the thermodynamic stability. This structure was shown to be essential for the ability of inter-α-inhibitor to participate in extracellular matrix stabilization. In addition, the data revealed that bikunin was positioned adjacent to both heavy chains and that the two heavy chains also were in close proximity. The chondroitin sulfate chain interacted with all protein components and inter-α-inhibitor dissociated when it was degraded. Conventional purification protocols result in the removal of the Mg2+ found in plasma and because divalent cations influence the conformation and affect function it is important to consider this when characterizing the biological activity of inter-α-inhibitor.

Inter-␣-inhibitor (I␣I) 3 is a proteoglycan composed of three separately expressed polypeptide chains called heavy chain 1 (HC1), heavy chain 2 (HC2), and bikunin (1). The protein is unique by being held together in a covalent, intracellularly assembled complex (2-5) by a chondroitin sulfate chain (6,7). The GAG is a low-sulfated chondroitin 4-sulfate (CS) chain that consists of 15 Ϯ 3 disaccharide units (5,8) attached to Ser-10 of bikunin by an O-glycosidic bond that attaches the CS chain to Ser-10 of bikunin (6). The HCs, meanwhile, are linked to the CS chain, and consequently to bikunin, through an ester bond between their C-terminal Asp residue and C-6 of an N-acetylgalactosamine in the CS chain (4,6). This type of GAG-mediated linkage is called a proteinglycosaminoglycan-protein (PGP) cross-link (6).
The physiological functions of I␣I are still mostly undefined; however, an involvement of I␣I in extracellular matrix (ECM) remodeling and stabilization and in inflammation has been recurring themes in the literature (9). I␣I is produced primarily in the liver and is considered to be a plasma protein. In the ECM, I␣I interacts with a protein called tumor necrosis factorstimulated gene-6 protein (TSG-6) (10,11). TSG-6 catalyzes the transfer of the HCs from the CS in I␣I to hyaluronan (HA), an abundant extracellular matrix GAG. The transfer reaction involves two sequential trans-esterifications and a covalent HC⅐TSG-6 intermediate (10,11). The presence of HC2 and Ca 2ϩ , Mg 2ϩ , or Mn 2ϩ is essential for the reaction with TSG-6 (12)(13)(14). The transfer reaction is believed to take place during inflammation and inflammation-like processes (e.g. ovulation) in which TSG-6 expression is induced (15).
The transfer of HC to HA alters both structural and functional characteristics of HA and can lead to changes in cell adhesion and migration in HA-rich ECMs (16 -18). Patients with rheumatoid arthritis and osteoarthritis generate high amounts of the HC⅐HA complex in inflamed synovial fluid (19,20). This has been related to an increased infiltration of leukocytes into the inflamed joints (18). The formation of the complex is, furthermore, essential for fertility in female mice. Both bikuninand TSG-6-deficient mice are unable to form a stable ECM around oocytes and ovulation consequently fails, thereby leading to infertility (21)(22)(23). Other functions of I␣I may relate to bikunin, which accounts for the protease inhibitory activity of I␣I (9).
HC1 and HC2 contain a von Willebrand factor type A (vWA) domain in which one metal ion-dependent adhesion site (MIDAS) motif is present (24). However, the tertiary structures of the HCs have never been described in details, and to this date, the only I␣I-related high-resolution structure available is the crystal structure of bikunin (25). Additionally, the structure of the heterotrimeric I␣I complex, including the CS, is not well described. An electron microscopy study has previously indicated that the overall structure of I␣I is an extended and dumbbell-like shape (26). In these electron microscopy-based analyses, the N-terminals of the HCs are observed as globular domains, with the C-terminals extending as flexible tails, and bikunin is observed as a small spherical structure (26).
The aim of the present study was to gain further insight into the structure and stability of I␣I and investigate how the CS and divalent cations influenced both. We used a combination of biochemical and biophysical methods to show that I␣I adopted a more compact conformation in the presence of Mg 2ϩ or Mn 2ϩ , but not in the presence of Ca 2ϩ . This was evident by a faster migration during native electrophoresis, a decrease in the Stokes radius and an increase in the thermodynamic stability. Furthermore, cross-linking mass spectrometry revealed that the protein components of I␣I interact directly and that these interactions depended on the presence of both the CS and/or Mg 2ϩ or Mn 2ϩ ions. It was further shown that the loose I␣I structure was unable to form the HC⅐TSG-6 complex. Without the formation of the HC⅐TSG-6 complex, the ECM stabilizing HC⅐HA complex cannot be generated. The plasma concentration of Mg 2ϩ is around 1 mM and it is thus likely that I␣I in vivo is in the compact conformation. Conventional purification protocols results in slow I␣I. Because this might not be the physiologically relevant conformation, it is important to consider this during the investigations of I␣I biology and function.

Experimental Procedures
Materials-Human plasma was obtained from Aarhus University Hospital, Skejby, Denmark. Chondroitinase ABC from Proteus vulgaris was purchased from AMSBIO. Bis(sulfosuccinimidyl) suberate (BS3) and sulfo-N-hydroxysulfosuccinimide (NHS) acetate were from Pierce. Polyclonal rabbit antibodies against HC1, HC2 (cross-reactivity with HC1), bikunin, and TSG-6 were produced as previously described (3,27). As described shortly, the proteins were run in SDS-PAGE and electroeluted. Antisera to the purified proteins were raised commercially in rabbits. The IgG fraction of the serum was recovered by affinity chromatography on a protein G Fast Flow column. Sequence grade trypsin was from Sigma.
SDS/Native-PAGE, Western Blotting, and Trypsin Inhibitor Counterstain-Samples were boiled in SDS sample buffer containing 30 mM DTT. SDS-PAGE was performed in 5-15% gradient gels (10 ϫ 10 ϫ 0.15 cm) using the glycine/2-amino-2methyl-1,3-propandiol HCl system as previously described (28). The same system was used for native-PAGE, but the sample was not boiled and SDS and DTT were omitted from the sample buffer and the buffer system. Bikunin was visualized using trypsin inhibitor counterstaining following non-reducing SDS-PAGE when indicated (1,29). For Western blotting, proteins were transferred to PVDF membranes as previously described (30). The membranes were blocked in 20 mM Tris-HCl, 137 mM NaCl, 0.1% Tween, pH 7.4 (TBS-T), containing 5% dry milk overnight at 4°C. Rabbit antibodies raised against HC1, HC2, or bikunin were added and the blot was incubated for 2 h at 23°C. After washing for 3 ϫ 15 min in TBS-T, the membrane was incubated in TBS-T containing the secondary anti-rabbit Cy3-labeled antibody in TBS-T and 5% dry milk. After 2 h of incubation and additional washing an image of the blot was acquired on a FluorChem Q system (Cell Biosciences).
Purification of Inter-␣-inhibitor-I␣I was purified from human plasma as previously described but with some modifications (1). Human plasma was made 5% in polyethylene glycol 8000, incubated for 1 h at 4°C, and pelleted by centrifugation at 3800 ϫ g for 15 min. The supernatant was made 16% in polyethylene glycol 8000, incubated for 1 h at 4°C, and centrifuged to collect the precipitated protein that subsequently was dissolved in 25 mM Tris-HCl, 150 mM NaCl, 10 mM EDTA, pH 7.4 (buffer A1). The sample was applied to a 5-ml Hitrap Q HP (GE Healthcare) anion exchange column connected to an ÄKTA purifier HPLC system (GE Healthcare) equilibrated in buffer A1. The column was eluted with a linear gradient of buffer B1 (buffer A1 containing 1 M NaCl) to 75% at 2% B/min with a flow rate of 3 ml/min. Fractions containing I␣I according to SDS-PAGE were pooled and dialyzed into 150 mM NaCl, 25 mM Tris-HCl, pH 7.4 (Buffer A2). The sample was applied to a 5-ml Hitrap Blue HP (GE Healthcare) equilibrated in Buffer A2. The column was eluted using a linear gradient of buffer B2 (buffer A2 containing 2 M NaCl) to 90% at 0.5% B/min with a flow rate of 5 ml/min. Fractions containing I␣I were pooled and dialyzed into 150 mM NaCl, 25 mM Tris-HCl, pH 7.4 (Buffer A3). The sample was applied to a 1-ml Hitrap Q HP (GE healthcare). The column was eluted with a linear gradient of buffer B3 (buffer A3 containing 1 M NaCl) to 75% at 2% B/min with a flow rate of 1 ml/min. The purified protein was dialyzed into 20 mM Hepes, 137 mM NaCl, pH 7.4, and frozen in aliquots.
Chondroitinase ABC Treatment of I␣I-Purified I␣I was treated with chondroitinase ABC (ChonABC) to partially or completely digest the CS chain. In the case of limited ChonABC digestion, I␣I was incubated with 1 milliunit of ChonABC/10 g of I␣I in 20 mM Hepes, 137 mM NaCl, pH 7.4, at 37°C for 2.5 h. In the case of extensive ChonABC digestion, I␣I was incubated with 1 milliunit of ChonABC/0.15 g of I␣I at 37°C for 18 h. Aliquots of the digests were analyzed by SDS-PAGE and the digests were kept at Ϫ20°C.
Chemical Cross-linking and Alkaline Hydrolysis-I␣I was titrated with BS3 at protein:cross-linker molar ratios of 1:50, 1:100, 1:500, 1:1000, 1:2500, and 1:5000. The titration was performed to find a low, but efficient cross-linker concentration. A ratio of 500 was used throughout the study. I␣I was incubated with 1 mM BS3 in 20 mM Hepes, 137 mM NaCl, pH 7.4, for 30 min at 20°C in the presence of 5 mM EDTA, 1 mM CaCl 2 , 1 mM MgCl 2 , or 1 mM MnCl 2 , as indicated. I␣I was incubated with EDTA or the specified cations for 30 min prior to cross-linking. BS3 solutions were prepared immediately before use to avoid hydrolysis. The reaction was quenched with 150 mM NH 4 HCO 3 and left for 30 min. NaOH was added to 300 mM to dissociate the protein components and the samples were left on ice for 30 min (6). Tris-HCl, pH 7.4, was added to a final concentration of 450 mM to lower the pH. The samples were separated by both SDS-PAGE and native-PAGE and visualized by either Coomassie Brilliant Blue staining or Western blotting.
Titration with Sulfo-NHS-acetate-I␣I was titrated with increasing amounts of sulfo-NHS-acetate at molar ratios from 1:65 to 1:2100 in 20 mM Hepes, 137 mM NaCl, pH 7.4, to block Lys residues. The samples were incubated for 60 min at 20°C before the reaction was quenched with 150 mM Tris-HCl, pH 7.4. Sample buffer for native-PAGE was added after 30 min and the samples were analyzed by native-PAGE.
Size Exclusion Chromatography-I␣I and ChonABC-treated I␣I were applied to a Superdex 200 increase 10/300 GL column (GE healthcare) equilibrated in 20 mM Hepes, 137 mM NaCl, pH 7.4, containing 5 mM EDTA, 1 mM CaCl 2 , 1 mM MgCl 2 , or 1 mM MnCl 2 . The column was calibrated with a size exclusion chromatography markers kit (Sigma) containing carbonic anhydrase (29 kDa and R s 2.01 nm,), albumin (66 kDa and R s 3.6 nm), alcohol dehydrogenase (150 kDa and R s 4.6 nm), ␤-amylase (200 kDa and R s 4.8 nm), apoferritin (443 kDa and R s 6.1 nm), and thyroglobulin (669 kDa and R s 8.5 nm). Blue dextran was used to obtain V 0 (2000 kDa). The Stokes' radius of I␣I and the protein components were calculated using the linear relationship between V e /V 0 and Stokes radius in a semi-logarithmic plot. Elution position of protein standards (repeated for each buffer condition) were used to generate a standard curve of Stokes radius versus (Ϫlog K av ) 1/2 that was used to calculate the Stokes radii of the I␣I species (31).
Far UV Circular Dichroism Spectroscopy Thermal Scans-I␣I and ChonABC-treated I␣I were diluted to a protein concentration of 0.2 mg/ml using 20 mM Hepes, 137 mM NaCl, pH 7.4, with 5 mM EDTA, 1 mM CaCl 2 , 1 mM MgCl 2 , or 1 mM MnCl 2 and stored at 5°C. Circular dichroism (CD) was performed on a J-810 CD-spectrometer (Jasco) using a 1-mm quartz cuvette with lid (Hellma Analytics). Thermo scans were performed from 20 to 95°C at a wavelength of 222 nm with a step size of 0.2°C, bandwidth of 8 nm, response time of 8 s, and scan speed of 60°C/h. The thermal scan data were fitted, as described previously, using the software KaleidaGraph (version 4.0 Synergy Software) (32). The fitted parameters were used to calculate the fraction of folded protein at different temperatures. All samples were analyzed in triplicates.
Identification of Sulfo-NHS-acetate Differentially Modified Residues-I␣I with 1 mM MgCl 2 , I␣I with 5 mM EDTA, or Cho-nABC-treated I␣I with 1 mM MgCl 2 were titrated with increasing amounts of sulfo-NHS-acetate at molar ratios from 0 to 1:1000 in 20 mM Hepes, 137 mM NaCl, pH 7.4. The reaction was quenched after 1 h by addition of 200 mM Tris-HCl, pH 8. The sample was denatured and reduced in 6 M urea containing 5 mM DTT for 1 h, alkylated with 15 mM iodoacetamide for 1 h, and diluted to 0.8 M urea with 50 mM NH 4 HCO 3 . Samples were treated with either trypsin alone or endoproteinase GluC and subsequently trypsin. The proteases were added at a (w/w) ratio of 1:20 and the sample was incubated for 18 h at 37°C. The sample was desalted using self-packed reverse phase microcolumns containing POROS R2 (33).
Cross-linked Peptides for Mass Spectrometry Analyses-I␣I was cross-linked with BS3 in the presence of 1 mM MgCl 2 , 1 mM MnCl 2 , or 5 mM EDTA. Following quenching of the cross-linking reaction, the sample was concentrated using a 30-kDa cutoff centrifugal filter (Amicon Ultra) and the buffer was exchanged to 20 mM Hepes, 137 mM NaCl, pH 7.4, to remove the cross-linker. The sample was reduced, alkylated, treated with trypsin, and desalted as described above. The desalted peptides were dissolved in 10 mM KH 2 PO 4 , 20% acetonitrile, pH 2.8 (Buffer A), and applied to a PolySULFOETHYL A column (PolyLC) equilibrated in buffer A. The peptides were eluted using a flow rate at 0.15 ml/min and a gradient of 0 -60% buffer B (1 M KCl in buffer A) over 30 min and 60% Buffer B for 10 min. The tryptic peptides were collected and micro-purified using self-pack microcolumns containing POROS R2 and subsequently lyophilized. Samples enriched for cross-linked peptides were analyzed by mass spectrometry in 3 replicate runs.
Mass Spectrometry-Nano LC-MS/MS was performed using an EASY-nLC II system (Thermo Scientific) connected to a TripleTOF 5600 ϩ mass spectrometer (AB Sciex) equipped with a NanoSpray III source (AB Sciex) operated under Analyst TF 1.5.1 control. Peptides were dissolved in 0.1% formic acid, injected, trapped, and desalted on a ReproSil-Pur C18-AQ trap column (2 cm ϫ 100-m inner diameter packed in-house with 3 m resin; Dr. Marisch GmbH, Ammerbuch-Entringen, Germany). The peptides were eluted from the trap column and separated on a 15-cm analytical column (75 m inner diameter) packed in-house in a pulled emitter with ReproSil-Pur C18-AQ 3 m resin (Dr. Marisch GmbH, Ammerbuch-Entringen, Germany). Peptides were eluted using a flow rate of 250 nl/min and a 50-min gradient from 5 to 35% phase B (0.1% formic acid and 90% acetonitrile or 0.1% formic acid, 90% acetonitrile and 5% DMSO). The collected MS files were converted to Mascot generic format (MGF) using the AB SCIEX MS Data Converter ␤ 1.1 (AB SCIEX) and the "proteinpilot MGF" parameters.
Analysis of MS Data-For cross-linked samples data analyses were performed using the MassAI software package (MassAI Bioinformatics) (34). The MGF files were preprocessed with the MGF-filter tool using default values and keeping the 125 most intense MS/MS peaks. The files were searched against the I␣I sequences using the following settings: MS error tolerance of 0.02 Da for cross-linked peptides, MS/MS error tolerance of 0.1 Da, three missed cleavages, iodoacetamide as fixed modification, oxidized methionine variable modification (M-ox), one modification per peptide, and BS3 cross-linking between lysine residues. In the case of multiple assignments to the same MS/MS scan, the highest scoring match was selected. For every unique cross-link, the MS/MS scan of the highest scoring match was manually inspected. A list of the identified crosslinks can be found in the supplemental material. Cross-linked peptides were only accepted if the precision on MS was below 10 ppm and the score above 25. For the sulfo-NHS-acetate titration the generated peak-list (MGF) was searched against an in-house database containing mature HC1, HC2, and Bikunin using the Mascot search engine (Matrix Science) with the following search parameters: MS tolerance of 10 ppm, MS/MS tolerance of 0.1 Da, trypsin with 2 miscleavages, carbamidomethyl as fixed modifications, and oxidized methionine and acetyl as variable modification. The search result from Mascot was used in Skyline for MS1 filtering to extract the ion intensity of all identified peptides (35,36). The individual peptide intensities were normalized to the total peptide intensity. The average normalized intensity and standard deviation were calculated based on three LC-MS/MS analysis.

HC⅐TSG-6 Complex Formation after Sulfo-NHS-acetate
Treatment of I␣I-I␣I was titrated with increasing amounts of sulfo-NHS-acetate at molar ratios from 0 to 1:1000 as described above in the presence of either 1 mM MgCl 2 or 5 mM EDTA. After acetylation, MgCl 2 was added to a final concentration of 10 mM. The acetylated I␣I (0.8 g) was incubated with a fixed amount of TSG-6 (0.8 g) for 1 h at 37°C in a total volume of 15 l. The total samples were separated by SDS-PAGE and visualized by Western blotting.

I␣I Adopts a Compact Structure in the Presence of Mg 2ϩ and
Mn 2ϩ -BS3 is a homobifunctional chemical cross-linker that reacts with the ⑀-amino group of lysine residues and the N termini. The maximal length of the BS3 spacer is 11.4 Å and may thus be used to cross-link primary amines that are within spatial proximity (37). I␣I was incubated with BS3 in the presence of Ca 2ϩ , Mg 2ϩ , and Mn 2ϩ . These divalent cations are essential for the TSG-6-HC2 mediated transfer of HCs to HA (12). The cross-linked samples were analyzed by SDS-PAGE ( Fig. 1A) and native-PAGE (Fig. 1B). Prior to SDS-PAGE, the PGP cross-link was cleaved by alkaline hydrolysis, thereby releasing the BS3 cross-linked proteins from the CS. Incubation of I␣I with BS3 results in the formation of an alkaline hydrolysis insensitive protein complex, which migrated approximately as a 200-kDa protein during reduced SDS-PAGE (Fig. 1A). Hence, the protein components were cross-linked and indicate that they interacted in the native complex.
There were subtle qualitative differences in the outcome of the cross-linking reaction when it was performed in the pres-ence of Mg 2ϩ or Mn 2ϩ compared with EDTA or Ca 2ϩ . In the presence of Mg 2ϩ or Mn 2ϩ , cross-linking resulted in the formation of a band of higher molecular weight and more homogenous migration (Fig. 1A, lanes 5 and 6). The molecular weight of this cross-linked product was similar to that of I␣I (Fig. 1A, I␣I in lane 1 compared with lanes 5 and 6). This suggested that the presence of Mg 2ϩ or Mn 2ϩ is essential for the efficient cross-linking of the three protein components and that the protein-protein interactions were divalent cation-dependent. Protein bands of lower molecular weight could, as judged by their size, arise from crosslinking of HC1 and HC2, HC1 and bikunin, or HC2 and bikunin. Chemical cross-linking analysis indicated that the protein components of I␣I interacted as opposed to functioning as separate entities bound to the CS chain.
The analysis of cross-linked I␣I by native-PAGE revealed further divalent cation-dependent differences (Fig. 1B). Addition of Mg 2ϩ or Mn 2ϩ , instead of EDTA or Ca 2ϩ , prior to the cross-linking reaction resulted in a faster migrating I␣I molecule. Apparently, Mg 2ϩ or Mn 2ϩ ions induce conformational changes in the molecule that cause the protein to obtain a more compact structure, which is maintained by the covalent cross-linking and thus remain evident after native-PAGE. Without prior cross-linking, divalent cation-dependent differences in the migration of I␣I could not be observed, indicating that the metal ion-dependent interactions are disrupted during electrophoresis (data not shown). The cross-linking of I␣I gave rise to two bands during native-PAGE (Fig. 1B, lanes 2-6). The upper band was present in all cross-linking conditions and was likely a less compact I␣I species. We suspect that the upper band represents an I␣I species with disrupted protein or protein-CS interactions. This could be caused by the presence of BS3-modified Lys residues. Hydrolysis of one of the reactive groups in BS3 prior to the formation of a cross-link will result in a modified Lys residue rather than cross-linking. Titration of I␣I with sulfo-NHS-acetate, which reacted with primary amines, gave rise to a band of similar size, thus supporting this hypothesis (Fig.  2). Moreover, the reduction in the amount of the upper band spe-   FEBRUARY 26, 2016 • VOLUME 291 • NUMBER 9 cies in the presence of Mg 2ϩ and Mn 2ϩ indicated that these two ions induce a more compact I␣I structure in which cross-linking of protein components is efficiently achieved.

The I␣I Structure Is Affected by Cations and the CS Chain
Mg 2ϩ and Mn 2ϩ Decrease the Stokes Radius of I␣I-To ensure that the compact structure, produced in the presence of Mg 2ϩ and Mn 2ϩ , was not an artifact of the cross-linking reaction, the I␣I structural changes induced by divalent cations was also investigated using size exclusion chromatography (Fig. 3A). The retention time of I␣I was determined in buffers containing a physiological salt concentration and EDTA, Ca 2ϩ , Mg 2ϩ , or Mn 2ϩ . From these experiments, it was evident that the retention of I␣I changed in the presence of Mg 2ϩ and Mn 2ϩ , even without prior crosslinking. The size of I␣I was estimated using a size exclusion chromatography standard calibration kit ( Fig. 3 and Table 1). In the presence of EDTA or Ca 2ϩ , the Stokes radius of I␣I was calculated to be 60 Å while in the presence of Mg 2ϩ or Mn 2ϩ , the Stokes radius of I␣I changed to 55 and 56 Å, respectively. Hence, I␣I seems to change its conformation in the presence of Mg 2ϩ or Mn 2ϩ . This conformational change resulted in a reduction of the Stokes radius, which is consistent with the altered migration of cross-linked I␣I during native-PAGE.
Because the CS chain is highly negatively charged, it may interact with residues in I␣I and with the added divalent cations. We wanted to investigate if the removal of CS affected the protein structure. I␣I was therefore treated with ChonABC prior to size exclusion chromatography, to produce either incomplete or complete degradation of the CS chain. Partial degradation released bikunin, whereas a short stretch of the CS chain still connects HC1 and HC2. This covalent complex had been named HC1⅐HC2 (5). There was no apparent cation-dependent difference in the retention of HC1⅐HC2 or bikunin when these I␣I species were analyzed by size exclusion chromatography in the presence of EDTA, Ca 2ϩ , Mg 2ϩ , or Mn 2ϩ ( Fig.  3B and Table 1). This indicates that the CS is essential for the structural change observed in the presence of MG 2ϩ or Mn 2ϩ .
Extensive ChonABC digestion results in complete digestion of the CS and causes the release of all protein components from the covalent complex (5). After extensive ChonABC digestion, I␣I gives rise to multiple peaks (Fig. 3C). Western blot analysis of the eluting peaks revealed that HC2 elutes in the first three peaks, followed by HC1 and finally bikunin ( Fig. 3C and Table  1). This indicated that protein interactions were lost when the CS was digested by ChonABC. Furthermore, released free HC2 appeared to polymerize. Additionally, the presence of Ca 2ϩ , Mg 2ϩ , or Mn 2ϩ did not result in changes of the retention time during size exclusion chromatography of the released protein components (Fig. 3C and Table 1). In summary, size exclusion chromatography revealed that Mg 2ϩ or Mn 2ϩ and the CS chain in concert cause I␣I to adopt a more compact structure.
The Thermal Stability of I␣I Is Affected by Divalent Cations and the CS-To further substantiate the effects of divalent cations we employed a thermal shift assay by following the thermal unfolding of I␣I using far-UV CD. CD data were recorded at 222 nm during a thermal scan from 20 to 95°C in the presence of EDTA, Ca 2ϩ , Mg 2ϩ , or Mn 2ϩ (Fig. 4A). The CD data (Fig. 4) Table 1). *, peak eluting at V c (total column volume) and containing EDTA added to the sample prior to loading on the column. were fitted to a two-state transition to obtain the midpoint of the transition (T m ) (Fig. 5A). In the presence of Mg 2ϩ and Mn 2ϩ the T m was increased to 55.7 and 57.3°C, respectively, from 53.8°C in the presence of EDTA, whereas the addition of Ca 2ϩ did not induce a T m shift. The data demonstrated that both Mg 2ϩ and Mn 2ϩ are able to stabilize the structure of I␣I during thermal unfolding, with the largest effect being observed in the presence of Mn 2ϩ . In parallel to the size exclusion chromatography analysis, we investigated if CS was involved in the metal ion-mediated stabilization (Fig. 4, B and C). The T m of the HC1⅐HC2 complex was unchanged in the presence of EDTA and Ca 2ϩ compared with I␣I (Fig. 5A). The thermal shift for Mg 2ϩ and Mn 2ϩ were reduced, compared with I␣I, and only Mn 2ϩ was significantly thermal shifted. A similar tendency was observed after extensive digestion of I␣I with ChonABC (Fig. 4C). Only the addition of Mn 2ϩ altered the T m (Fig. 5A). However, the thermal scan of dissociated I␣I did not fit a two-state unfolding scheme. This suggests that the process monitored during the thermal scan involves several unfolding steps. It is likely that this can be    FEBRUARY 26, 2016 • VOLUME 291 • NUMBER 9

The I␣I Structure Is Affected by Cations and the CS Chain
attributed to complete dissociation of I␣I leading to an independent thermal unfolding of HC1, HC2, and bikunin. Although we do not observe distinct unfolding steps, the three proteins may unfold close to each other in temperature so that the process becomes a composite of different events. This is consistent with HC dissociation observed by SEC. The enthalpy of unfolding at the midpoint of denaturation (⌬H Tm ) was likewise obtained by fitting the CD data (see Fig. 4) to a two-state transition (Fig. 5B). ⌬H Tm is a measure of the cooperativity of protein unfolding. A highly cooperative unfolding indicated that I␣I exists as a compact structure in which unfolding of one part of the structure affects the unfolding of other parts (38). Low cooperativity, on the other hand, indicates that the complex consists of individually folded structures that unfold independently (39). Addition of Mg 2ϩ or Mn 2ϩ led to an increase in ⌬H Tm , thus indicating that these cations stabilize protein interactions in I␣I or induce higher cooperativity of unfolding. Similarly to the thermal stability, the effect of Mg 2ϩ and Mn 2ϩ on ⌬H Tm also depends on the presence of the CS.
The Divalent Cation-dependent Interactions within I␣I Depend on the Presence of the CS-Chemical cross-linking, size exclusion chromatography, and thermal denaturation showed that Mg 2ϩ or Mn 2ϩ induced conformational changes in I␣I. This resulted in the formation of a more compact protein structure. To decipher which protein interactions contributed to the formation of this structure, the outcome of the cross-linking reaction was further characterized by Western blotting. I␣I was cross-linked with BS3 in the presence of EDTA, Ca 2ϩ , Mg 2ϩ , or Mn 2ϩ . The cross-linked protein components were dissociated from CS by mild hydrolysis and separated by SDS-PAGE. We used antibodies against HC1, HC2, and bikunin to visualize the products of the cross-linking reaction (Fig. 6). Hydrolysis of the PGP cross-link would usually result in the appearance of three bands representing HC2, HC1, and bikunin. However, cross-linking generated protein bands of higher molecular weight showing molecular weights and antigenicity corresponding to HC1⅐HC2, HC2⅐bikunin, HC1⅐HC2, and HC1⅐HC2⅐bikunin suggesting that all of the protein components interacted with each other (Fig. 6). The double band observed migrating similar to the HCs appears due to intramolecular cross-linking within a single HC generating two different versions of the same HC (Fig. 6, A and B).
There was a qualitative and a quantitative difference between cross-linking that has been performed in the presence of EDTA or Ca 2ϩ compared with Mg 2ϩ or Mn 2ϩ (Fig. 6, lanes 1 and 2,  compared with 3 and 4). In the presence of Mg 2ϩ or Mn 2ϩ the cross-linked products migrating in the region of HC1⅐HC2 and I␣I were more intense, had a more homogenous migration, and migrated at higher mass more similar to I␣I. These data suggest that the inter-molecular cross-linking of bikunin is more efficient in the presence of Mg 2ϩ or Mn 2ϩ . This consequently increases the amounts of cross-linked material containing all of the protein components: HC1⅐HC2⅐bikunin. Mg 2ϩ and Mn 2ϩ , thus, seem to stabilize or induce interactions between the protein components in I␣I.
To study the interplay between Mg 2ϩ and CS, three variants of I␣I were studied by BS3 cross-linking with and without Mg 2ϩ (Fig. 7). The three types of I␣I were: (a) I␣I with no treatment (Fig. 7, A-C, lane 1); (b) limited ChonABC digestion producing free bikunin and the HC1⅐HC2 complex (Fig. 7, A-C, lane 2); and (c) extensive ChonABC digestion generating dissociated I␣I where all three protein components are released from the CS (Fig. 7, A-C, lane 3). The cross-linked samples were separated by SDS-PAGE. Products containing HCs were visualized by Western blotting using antibodies toward HC1 (Fig. 7A) or HC2 (Fig. 7B). After BS3 treatment free bikunin lost the antibody reactivity and was alternatively visualized by trypsin  It is evident that all protein components can be cross-linked in all conditions. However, the presence of Mg 2ϩ or Mn 2ϩ produces a band containing HC1 and HC2 of higher molecular weight. Cross-linking of bikunin seems to be more efficient in the presence of Mg 2ϩ or Mn 2ϩ . Evidently, there are qualitative and quantitative differences in the cross-linking of I␣I when adding Mg 2ϩ or Mn 2ϩ , compared with the addition of Ca 2ϩ or EDTA. The data represent more than 3 independent technical replicates.
inhibitor counterstain (Fig. 7C) (1). Limited CS digestion did not alter cross-linking of the two HCs (Fig. 7, A and B, lanes 5  and 8). Cross-linking of bikunin to the HCs is both dependent on Mg 2ϩ (Fig. 7C, lanes 4 and 7) and CS (Fig. 7C, lanes 6 and 8). Extensive ChonABC digestion completely abolished BS3-mediated cross-linking between HC1, HC2, and/or bikunin (Fig. 7,  A-C, lanes 6 and 9). Cross-linking still resulted in the formation of several high molecular weight HC2 species (Fig. 7B, lane 9). The high molecular weight HC2 species supported the evidence of HC2 polymerization, which was observed during size exclusion chromatography. These data shows that the interaction of bikunin and the HCs are both Mg 2ϩ and CS dependent and that HC1 and HC2 interact in a CS-dependent manner.
Chemical Footprint Reveals CS and Mg 2ϩ -mediated Interaction Sites in I␣I-In the sulfo-NHS-acetate acetylation experiment a destabilization/unfolding of the I␣I structure was observed indicating that at least one lysine residue is essential for the correct fold (Fig. 2). Chemical footprinting was applied tostudytheinvolvementoflysineresiduesintheCS-andMg 2ϩ -dependent interactions as well as probing the overall structure. Basically, I␣I was incubated with increasing amounts of sulfo-NHS-acetate. After acetylation, the samples were treated with either trypsin or a combination of trypsin and endoproteinase GluC. The acetylated residues were identified by MS and the amount of each peptide was quantified based on MS1 intensities (XIC) (Fig. 8). I␣I was studied under three different conditions: with MgCl 2 , EDTA, or after enzymatic removal of the CS. The degree of acetylation can be used to measure how exposed particular Lys residues are during labeling. Bikunin had four residues (Lys-22, Lys-70, Lys-106, and Lys-126), which showed increased acetylation when either Mg 2ϩ or the CS chain was removed. The largest effect was seen for Lys-70 and Lys-126. The increased acetylation, upon removal of Mg 2ϩ or CS shows that Lys-70 and Lys-126 either directly binds the CS or is shielded by the CS. These two residues are located in close proximity within the bikunin structure in a region with several basic amino acid residues (25). Proteins often bind GAGs by clusters of basic residues. Our data thus indicates the presence of an Mg 2ϩ -dependent CS binding site in bikunin. A single residue in HC1 (Lys-278) showed increased acetylation upon removal of Mg 2ϩ , which further increased when the CS was removed. HC2 showed the most diverse pattern of acetylation. Lys-96 and Lys-325 showed a minor increase in acetylation by EDTA treatment and additional acetylation without CS, a pattern similar to Lys-278 in HC1. The sequential increased acetylation by EDTA and ChonABC treatment indicate the presence of an Mg 2ϩ -dependent interaction with CS in both heavy chains. HC2 Lys-192 became highly acetylated when the CS was removed, but was unaffected to the absence of Mg 2ϩ . The acetylation profile of Lys-192 indicates that the residue is either directly involved in binding to CS or highly shielded by the CS. In addition, HC2 contains two sites (Lys-273 and -433) only affected by EDTA, indicating the presence of an Mg 2ϩ -mediated interaction independent of the CS. Finally, in HC2 there are two sites (Lys-136 and Lys-198) where the acetylations are decreased by removal of CS. The lower signal from Lys-198 can be explained by the high degree of acetylation in Lys-192, however, the lower amount of acetylation can also be caused by the multimerization of HC2 seen when CS is removed.
The CS and Mg 2ϩ Mediates a Contraction of Bikunin-Identification of chemically cross-linked residues by mass spectrometry may provide additional insight to the structure of I␣I. To identify interfacing protein regions I␣I was cross-linked with BS3 in the presence of Mg 2ϩ , Mn 2ϩ , or EDTA and digested with trypsin. After enrichment of cross-linked peptides by strong cation exchange, an LC-MS/MS analysis identified the cross- linked peptides ( Fig. 9 and supplemental material). The LC-MS/MS analysis resulted in the identification of both intraprotein and inter-protein cross-links (Fig. 9). In this context, "intra" refers to cross-links within the same subunit and "inter" refers to cross-links between different protein components within the same I␣I molecule. Bikunin became cross-linked to the N-terminal region of both HC1 and HC2 when Mg 2 was included during BS3 cross-linking. In contrast, in the presence of Mn 2ϩ bikunin was only cross-linked to HC1. Cross-linking in the presence of either Mg 2ϩ or Mn 2ϩ resulted in a similar pattern of inter-protein and intra-protein cross-links of the HCs. Intra-protein cross-links that connect the N-and C-terminal were found in both HCs. This corresponds well with the known long-range disulfide bond found in HC1 (41). The identified intra-protein cross-links suggest that the N-and C-terminals of the HC are in close proximity in the native folded proteins. Inter-protein cross-links are present along the entire amino acid sequences of the HCs. Most cross-linked residues have multiple contact points at the opposing HC supporting an overall compact fold of the HCs. The combined pattern of inter-protein and intra-protein cross-links support that the Nand C-terminal are in close proximity in the native I␣I complex (Fig. 9). Cross-linking in the presences of EDTA produced a more random cross-linking pattern especially evident by the multiple contact points between bikunin and the HCs, but also by the increased number of unique cross-links between the HCs. In addition, EDTA results in less intra-protein cross-links of the HCs indicating that divalent cations are involved in maintaining a more compact and rigid fold.
The Compact Structure Stabilized by Mg 2ϩ and CS Is Essential for HC⅐TSG-6 Complex Formation-The biological relevance of the loose and compact structures were investigated by preventing the formation of compact I␣I using sulfo-NHS-acetate (see Fig. 2) and analyzing the ability to form the HC⅐TSG-6 complex. The results of the sulfo-NHS-acetate titration and the chemical footprinting experiments supported that Mg 2ϩ -and CS-dependent interactions exist and that these are responsible for inducing the observed compact structure. This was underscored by the ability to prevent the formation of compact I␣I after altering the charge of lysine residues including those involved in the compact structure (see Figs. 2 and 8). In addition, we have previously shown that formation of a HC⅐TSG-6 complex is dependent on the presence of divalent cations (13). To test if TSG-6 complex formation depended on the compact I␣I structure we incubated TSG-6 and sulfo-NHS-acetate pretreated I␣I in the presence of either MgCl 2 or EDTA (Fig. 10). The ability to form a HC⅐TSG-6 complex was detected by Western blotting using a TSG-6 antibody. In the presence of Mg 2ϩ TSG-6 formed complexes with the I␣I HCs (lane 1), however, increasing the sulfo-NHS-acetate concentration abolished complex formation (lanes 2-8) suggesting that the loose I␣I conformation is unable to form the complex. Furthermore, when I␣I was treated with sulfo-NHS-acetate in the presence of EDTA, a lower concentration of sulfo-NHS-acetate was required to prevent complex formation. These data suggest that the loose I␣I conformation is unable to form the HC⅐6TSG-6 complex.

Discussion
In the present study, we have used a range of biochemical and biophysical methods to show that I␣I changes its conformation and adopts a more compact structure in the presence of Mg 2ϩ and Mn 2ϩ . Additionally, we have shown that CS is essential for the overall fold and stability of I␣I. Previous studies have shown that Mg 2ϩ , Mn 2ϩ , or Ca 2ϩ are able to act as cofactors in the TSG-6-mediated transfer of HCs from I␣I to HA (13). Here we show that Mg 2ϩ or Mn 2ϩ , but not Ca 2ϩ , cause I␣I to adopt a more compact conformation and that this compact conformation is essential for the interaction with TSG-6.
The presence of either Mg 2ϩ or Mn 2ϩ shifts the retention of I␣I during size exclusion chromatography. The altered retention time corresponds to a decrease in the calculated size of I␣I, from a Stokes radius of 60 Å in the absence of cations to 55 and 56 Å in the presence of Mg 2ϩ or Mn 2ϩ , respectively. Although I␣I is compacted by Mg 2ϩ and Mn 2ϩ , the calculated size is still larger than expected. The molecular mass of I␣I including all known modifications is ϳ179 kDa. If I␣I were to be a fully globular protein it would result in a Stokes radius of 37 Å (42).
The increased Stokes radius may be a result of a moderate elongation and/or be a result of a large hydration shell. In solution, proteoglycans typically obtain a large hydration shell due to the presence of the negatively charged GAG chains (43). Mg 2ϩ and Mn 2ϩ also increase the thermal stability and cooperativity of unfolding of I␣I. This indicates that the conformational changes induced by Mg 2ϩ and Mn 2ϩ lead to a stabilization of the protein interactions in I␣I. The increased cross-linking efficiency in the presence of Mg 2ϩ and Mn 2ϩ supports this observation. In plasma, the concentration of Ca 2ϩ and Mg 2ϩ is around 1 mM, whereas Mn 2ϩ only is present in trace amounts (44). Based on the physiological concentrations, Mg 2ϩ , rather than Mn 2ϩ , is most likely the physiologically relevant ion in relationship to I␣I structure and functions.
The structure of the native I␣I complex is not well described. Two studies have examined the structure of I␣I, however, in light of our new data showing that the structure of I␣I involves both Mg 2ϩ and CS it seems that these studies are based on either the EDTA or ChonABC-treated variant of I␣I (26,45). In the present study, we have used chemical footprinting and cross-linking to identify potential CS and Mg 2ϩ -dependent interaction sites as well as potential interfacing protein regions. The chemical footprinting revealed that bikunin has an Mg 2ϩ -  The identified cross-links are mapped to the sequences of bikunin, HC1, and HC2. Intra-protein cross-links (red) and inter-protein cross-links (blue) are shown. The cross-linking pattern of I␣I indicates that the HCs have a C-to N-terminal packing where bikunin is located close to the N-terminal (lower panel). The graphic representation of I␣I shows the transition from a loose and dynamic structure to a more rigid structure with interactions between the proteins subunits as well as the CS in the presence of MgCl 2 . In the compact conformation, bikunin is via CS-and Mg 2ϩ -dependent interactions located in vicinity of the N-terminal of the HCs. FEBRUARY 26, 2016 • VOLUME 291 • NUMBER 9

JOURNAL OF BIOLOGICAL CHEMISTRY 4667
and CS-dependent interaction, most likely directly to the CS. The Lys-70 and Lys-126 in bikunin are located in proximity in a region containing most of the charged residues in bikunin (25). This region is located on the opposing side of the protease inhibitor sites and could form a lysine-and Mg 2ϩ -dependent CS binding region, which would position bikunin for optimal interaction with targeted proteases. Lys-106 are found in the outskirts of this region and binds a sulfate group in the crystal structure, which could have substituted for the CS. Chemical footprinting also reveal that both HCs have Mg 2ϩ -and CS-dependent interactions. HC2 had what seems to be a direct interaction with the CS at Lys-193. Furthermore, HC2 had two lysine residues only affected by EDTA treatment, indicating an Mg 2ϩ -dependent interaction site. One of the residues, Lys-273, is located close to the MIDAS found in both HCs (residues 262-266 in HC2). An Mg 2ϩ -dependent interaction between HC1 and HC2 via their MIDAS sites could explain our observation. The biological significance of these interactions is emphasized by the observation that complex formation between TSG-6 and HCs is disrupted when the lysine-dependent interactions are prevented by acetylation.
The HC1-HC2 interaction is supported by chemical crosslinking data. Chemical cross-linkers are commonly used to characterize the protein structure and protein complexes by identifying residues that are spatially close (37). The maximal length of the BS3 spacer is 11.4 Å, thus allowing C ␣ -atoms of cross-linked lysine residues to be 26 -30 Å apart in a proposed structure (46). We have identified both intra-protein and interprotein cross-links in I␣I. In both HCs, intra-protein crosslinks that span from the N-terminal region to the C-terminal region are present, thus indicating that the N-and C-terminal regions are within close proximity in the native structure. Although the similar pattern of intra-protein cross-links in HC1 and HC2 suggests that there are similarities in the overall structure, the chemical footprinting and the fact that HC2 polymerize also reveal a clear difference between the two proteins. Inter-protein cross-links were identified between all protein components of I␣I. Removal of divalent cations by EDTA resulted in a more loose I␣I structure with a more dynamic distribution of bikunin. This falls in line with other results, showing that the packing of bikunin is strongly affected by divalent cations.
The HCs contain a vWA domain and within this domain the previously mentioned MIDAS motif. Our cross-linking data agrees with the common fold of the vWA domain in which the N and C termini of the domain are located at the same site of the domain and close to each other (47). The vWA domain is commonly associated with protein-protein interactions and with the formation of multiprotein complexes (24). Interactions between vWA containing proteins and their ligands often depend on the binding of divalent cations to a MIDAS motif and can be associated with conformational changes within the vWA domain. This is at least the case for vWA domain-containing integrins. However, among the vWA domain-containing proteins the mechanism of ligand binding is diverse (47). Our results show that divalent cation binding is associated with conformational changes in I␣I. A possible site of cation binding is the MIDAS motif of the HCs. The coordination of Mg 2ϩ ions at the MIDAS motifs in I␣I could, similarly to Mg 2ϩ ions in integrins, mediate interactions to other proteins or mediate interactions between the I␣I protein components. Interestingly, interaction between I␣I and another ECM protein, pentraxin 3, is also Mg 2ϩ dependent (48). Pentraxin 3 is, similarly to I␣I, essential for fertility in female mice, and it seems that the interactions between I␣I and pentraxin 3 are important for organizing the ECM around oocytes. Interactions between I␣I and complement components have similarly been suggested to depend on divalent cations and interactions with the vWA domain in I␣I (40,49).
In the present study, we have shown that interactions between the protein components exist within the native structure of I␣I. These interactions are GAG-dependent and/or divalent cation-dependent. Furthermore, we have shown that divalent cations induce a more compact conformation of I␣I in which the protein interactions are stabilized and the structural stability is increased. The cation-induced compact conformation is essential for the HC⅐TSG-6 complex formation. The compact structure is most likely the native fold of I␣I. Because I␣I is normally purified from citrated or EDTA-treated human plasma, this might have been overlooked in the past.