Multidomain Human Peroxidasin 1 Is a Highly Glycosylated and Stable Homotrimeric High Spin Ferric Peroxidase*

Background: Human peroxidasin 1 (hsPxd01) mediates the formation of sulfilimine cross-links within the collagen IV scaffold of basement membranes. Results: Overexpressed hsPxd01 contains covalently linked heme catalytically active for production of hypobromous acid. Conclusion: hsPxd01 has peroxidase-like active site structure but restricted substrate accessibility. Significance: Architecture of hsPxd01 facilitates product release and its interactions with the physiological substrate collagen IV. Human peroxidasin 1 (hsPxd01) is a multidomain heme peroxidase that uses bromide as a cofactor for the formation of sulfilimine cross-links. The latter confers critical structural reinforcement to collagen IV scaffolds. Here, hsPxd01 and various truncated variants lacking nonenzymatic domains were recombinantly expressed in HEK cell lines. The N-glycosylation site occupancy and disulfide pattern, the oligomeric structure, and unfolding pathway are reported. The homotrimeric iron protein contains a covalently bound ferric high spin heme per subunit with a standard reduction potential of the Fe(III)/Fe(II) couple of −233 ± 5 mV at pH 7.0. Despite sequence homology at the active site and biophysical properties similar to human peroxidases, the catalytic efficiency of bromide oxidation (kcat/KMapp) of full-length hsPxd01 is rather low but increased upon truncation. This is discussed with respect to its structure and proposed biosynthetic function in collagen IV cross-linking.

lactoperoxidase (LPO), and thyroid peroxidase (TPO) have been shown to play important roles in innate immunity as well as hormone biosynthesis (1,2). Recently, a phylogenetic study demonstrated that these vertebrate enzymes are part of a superfamily that is composed of seven subfamilies (3). In a close phylogenetic neighborhood of the above-mentioned chordata peroxidases (i.e. subfamily 1), multidomain peroxidases called peroxidasins are clustering in subfamily 2. This name was first coined in 1994 by Nelson et al. (4), who succeeded in identifying the first representative in Drosophila melanogaster. They described this peroxidasin as a multidomain peroxidase that participates in extracellular matrix consolidation. In the following 15 years, only a few papers were published reporting the occurrence of peroxidasins in other organisms like Caenorhabditis elegans (5) or Xenopus tropicalis (6) and suggesting a role in extracellular matrix biosynthesis. Finally, in 2008, the two human peroxidasins (hsPxd01 and hsPxd02) became the focus of investigations analyzing expression patterns (7)(8)(9)(10) and biochemical properties (7,8,11,12).
Great attention was devoted to human peroxidasin 1 after publication of the work by Bhave et al. (13) that demonstrated the responsibility of hsPxd01 for the formation of sulfilimine bonds in collagen IV via release of hypohalous acids. Furthermore, it was shown that this unique covalent bond in basement membrane is a primordial innovation for tissue evolution (14). Moreover, very recently it could be demonstrated that bromide oxidation to hypobromous acid mediated by hsPxd01 is essential for the assembly of collagen IV scaffolds in tissue development and architecture (15). Fig. 1 depicts the multidomain structure of human peroxidasin 1. Gene analysis shows the occurrence of a signal peptide, one leucine-rich repeat domain (LRR), four immunoglobulinlike motifs (Ig), a peroxidase domain (POX), and a C-terminal von Willebrand factor C domain (VWC). Despite the important physiological implications, no detailed structural information about human peroxidasin 1 is available. This prompted us to study the biophysical and biochemical characteristics of human peroxidasin 1 for a better understanding of its mode of function and mechanism.
Cloning, Stable Transfection, Expression, and Purification of hsPxd01-Full-length human peroxidasin 1 encoded on a pcDNA TM 3.1/V5-His TOPO vector was kindly provided by Miklós Geiszt (Department of Physiology, Semmelweis University School of Medicine, Budapest, Hungary). For the stable calcium phosphate transfection of HEK 293 cells, standard selection and cultivation procedures with minor modifications were used (17). After screening of the stably transfected clones by immunoblotting as well as peroxidase activity screening, the best producer was selected for large scale production in triple flasks (Life Technologies). During production, cells were cultivated in DMEM supplemented with 10% fetal bovine serum, 5 g/ml hematin, penicillin, and streptomycin and a protease inhibitor mixture (Sigma). The harvested media were stored at 4°C and eventually purified using immobilized metal affinity chromatography. Therefore, the harvested media were supplemented with 300 mM NaCl, 40 mM imidazole, and pH adjusted to 7.4 before application onto a chelating Sepharose gel column (GE Healthcare). The column was charged with Ni 2ϩ and equilibrated with 20 mM phosphate buffer, pH 7.4, with 300 mM NaCl and 40 mM imidazole. Protein was FIGURE 1. Arrangement and composition of domains of hsPxd01. A, models are of the following: (i) LRR domain, including N-and C-terminal capping motifs; (ii) four consecutive Ig domains; (iii) VWC domain with highlighted cysteines; and (iv) of the peroxidase domain. The latter model was constructed using homologous goat LPO (PDB code 2R5L) as template. Magnified overlay of the active site of goat LPO(gLPO) and hsPxd01 illustrates the conservation of all amino acids in the active site. Numbers in blue correspond to hsPxd01 amino acid sequence, including the signal peptide, and numbers in pink correspond to gLPO amino acid sequence with (in parentheses) and without signal and propeptide region. B, domain arrangement of human peroxidasin 1. eluted with 20 mM phosphate buffer containing 500 mM NaCl and 300 mM imidazole. Fractions with the best purity number were pooled, concentrated, and desalted using Centricon with a 100-kDa cutoff membrane (Millipore).
SDS-PAGE, Western Blot, and Multiangle Light Scattering-Protein samples were separated by gradient PAGE and analyzed either directly using Coomassie Blue staining or after immunoblotting onto a nitrocellulose membrane (GE Healthcare) using anti-His (Qiagen), anti-Strep (IBA), or anti-hsPxd01 antibodies (Sigma). Standard protocols were followed.
For determination of the molar mass, the protein samples were analyzed by SEC chromatography coupled to detectors for analysis of UV-visible, refraction index, and multiangle light scattering (MALS). For instrumentation, HPLC (Shimadzu prominence LC20) and MALS (WYATT Helios Dawn8ϩ plus QELS; software ASTRA 6) were used, and the columns were WYATT SEC Columns (7.8 ϫ 5 and 7.8 mm ϫ 300 mm; particle size, 5 m, pore size 300 Å) with corresponding guard column (WTC030S5G). SEC buffer was Dulbecco's PBS plus 200 mM NaCl; flow rate was 0.5 ml per min; injected protein was 55 g. For performance of SEC chromatography under reducing conditions, the protein was pretreated with 50 mM dithiothreitol (DTT) for 30 min at room temperature, and during chromatography, the SEC buffer contained 2 mM DTT.
Assessment of Heme to Protein Linkage by Enhanced Chemiluminescence (ECL)-MPO and hsPxd01 samples were heated for 10 min at 70°C before separation by gradient PAGE. Then the gel was blotted on nitrocellulose membrane, and the covalent heme to protein was probed by ECL detection (20). Clarity Western ECL (Bio-Rad) substrate was used as recommended, and signals were detected with a ChemiDoc XRSϩ imaging system (Bio-Rad).
Mass Spectrometric Analysis of Glycosylation Occupancy-After precipitation by 10% trichloroacetic acid, recombinant hsPxd01 pellets were dissolved in the denaturation buffer ammonium bicarbonate (100 mM) and 2,2,2-trifluoroethanol and subsequently reduced by DTT. Alkylation was performed with iodoacetamide, and the excess was quenched again by DTT. Before addition of trypsin or pepsin, the sample was diluted 10 times, and the pH was adjusted to the corresponding optimum. After overnight incubation at 37°C, the proteases were heat-inactivated. Finally, PNGase F was added and incubated overnight at 37°C.
For N-glycan site occupancy, a rapid resolution liquid chromatographic system (1200 series) coupled to a quadrupole/ time-of-flight equipped with an electrospray ionization source (ESI-QTOF 6520 series) mass spectrometer (Agilent Technologies, Palo Alto, CA) was used for analysis of the aforementioned protein digests (21). As PNGase F removes glycans and performs deamination of asparagine (Asn 3 Asp), the glycosylation occupancy was calculated by measuring the ratio between the area under the curve of the native peptide and its deaminated form. Total occupancy (100%) was obtained when the deaminated form of the peptide alone was detected. A glycosylation site free of glycans (0% of occupancy) was obtained when only the peptide with the native Asn residue was present. For prediction of glycosylation sites NetNgly was used (22).

hsPxd01 Is a Multidomain Heme Peroxidase
Mass Spectrometric Analysis of Cysteines and Cystines-Analysis of cysteine residues was carried out on hsPxd01 according to a method previously described for sequence recovery (23). The method was adjusted to allow for the recovery of free cysteines as carbamidomethylated cysteines, whereas disulfide bonds were conserved. Briefly, recombinant hsPxd01 (350 g) was incubated at 20°C in darkness for 10 h in 25 l of 100 mM ammonium bicarbonate, including 60 mM iodoacetamide. 25 l of 2,2,2-trifluoroethanol, 300 l of formic acid (2% (v/v) in water), and 300 l of pepsin (5 mg/ml in formic acid 2% (v/v) in water) were added, and hsPxd01 was digested overnight at 37°C. After pepsin heat inactivation, pH was adjusted to 7-8, and PNGase F (100 milliunits) was added to proceed to peptide deglycosylation for 7 h at 37°C. Finally, trypsin (50 l, 1 mg/ml) was added, and overnight digestion was performed at 37°C. Trypsin was inactivated by addition of 2 l of formic acid (100% (v/v)), and sample was dried down using a centrifuge vacuum. Sample was finally dissolved in 200 l of formic acid (0.1% (v/v) in water). Free cysteines are detected in an alkylated form (carbamidocysteine).
30 l of the resulting samples were injected into the LC system and analyzed as described previously (23). Briefly, analyses were performed with the rapid resolution liquid chromatographic ESI-QTOF system described above. Peptides were separated on a Poroshell-120 end-capped C18 (100 ϫ 2.1 mm inner diameter, 2.7-m particle size) from Agilent Technologies (Palo Alto, CA) using a 105-min gradient of formic acid and acetonitrile (23). Auto-MS/MS spectra were acquired in positive and high resolution acquisition mode (4 GHz) (23). Data were acquired by the Mass Hunter Acquisition for TOF and QTOF (version B.04 SP3) software and analyzed by the Mass Hunter Qualitative Analysis (version B.05 SP1) with Bioconfirm and by Spectrum Mill software (Agilent Technologies, Palo Alto, CA).
Electron Paramagnetic Resonance Spectroscopy (EPR)-Purified recombinant hsPxd01 and hsPxd01-con4 were prepared in 40 mM citrate/phosphate buffer, pH 7, and 20% glycerol (Sigma) as cryoprotectant. The sample was transferred into Wilmad quartz tubes (3 mm inner diameter) and flash-frozen in liquid N 2 . Frozen samples were kept frozen on dry ice while the headspace above the sample was flushed with argon. Oxygen-free samples were frozen back to 77 K and transferred into the resonator for 10 K measurements. Spectra were recorded on a Bruker EMX continuous wave EPR spectrometer, operating at X-band (9 GHz) frequencies, equipped with a high sensitivity resonator and an Oxford Instruments ESR900 cryostat. EPR spectra were recorded under nonsaturating conditions using 2 milliwatts microwave power, 100 kHz modulation frequency, 1 millitesla modulation amplitude, 41-ms conversion time, 41-ms time constant, and 2048 points. Simulation of high spin Fe(III) forms was carried out using the software EasySpin (24) and consisted of a weighted sum of simulations of the individual high spin compounds. The rhombicity was obtained from g x eff and g y eff (25), and the relative intensities were calculated on the basis of the simulations.
Determination of Thermal Stability-Temperature-mediated protein unfolding was followed either by electronic circular dichroism (ECD) spectroscopy or differential scanning calorimetry (DSC). For ECD, Chirascan from Applied Photophysics (Leatherhead, UK) was used, and changes of the ellipticity at 208 nm as well as at 412 nm in 20 mM phosphate buffer, pH 7.4, with increasing temperature were recorded and evaluated (29). The instrument was flushed with nitrogen at a flow rate of 5 liters min Ϫ1 and was equipped with a Peltier element for temperature control. Temperature-mediated denaturation was monitored between 20 and 95°C. Temperature was increased stepwise with 1.0°C min Ϫ1 . Single wavelength scans were performed with instrumental parameters set as follows. Visible ECD at Soret maximum was performed with 9.5 M hsPxd01 in 20 mM phosphate buffer, pH 7.4. The path length was 10 mm and spectral bandwidth 1 nm, and scan time per point was set at 10 s. Far-UV ECD at 208 nm was performed with 4.8 M hsPxd01 in 20 mM phosphate buffer, pH 7.4. The path length was 1 mm and spectral bandwidth 3 nm, and scan time per point was set at 10 s.
DSC analysis was done with a VP-DSC Capillary Cell Microcalorimeter (MicroCal). 5 M protein samples in 20 mM phosphate buffer, pH 7.4, were heated from 30 to 100°C with a heating rate of 1°C min Ϫ1 . Baselines were subtracted and data normalized for protein concentration and fitted with a nontwo-state thermal unfolding model using the software Origin 7.
Determination of Bromination Activity-Bromination activity of recombinant hsPxd01 and hsPxd01-con4 was tested spectrophotometrically by measuring the halogenation of NADH. The produced hypobromous acid reacts with NADH to stable bromohydrin. This assay detects the initial rate of bromohydrin production at ϭ 275 nm using ⑀ 275 of 11,800 M Ϫ1 cm Ϫ1 (30 -32). Bromide was chosen because it was shown to be an essential trace element for assembly of collagen IV scaffolds in tissue development and architecture (15). Briefly, 100 nM hsPxd01 or hsPxd01-con4 was incubated at 25°C in 20 mM phosphate buffer, pH 7.4, containing 100 M NADH and varying concentrations of either hydrogen peroxide (with a constant NaBr concentration of 5 mM) or NaBr (with a constant hydrogen peroxide concentration of 50 M) (32). Reactions were started upon addition of H 2 O 2 . Initial rates were measured over the 1st min of reaction to minimize possible inhibition by hydrogen peroxide. Note that Michaelis-Menten parameters (K m and V max ) derived from steady-state rate measurements do not conform to reactions catalyzed by peroxidases, which follow an irreversible ping-pong mechanism (where equilibria between reactants and products are not established). Nevertheless, to allow the comparison of the bromination activity of hsPxd01 with data from peroxidases investigated under the same assay conditions (32) and to keep the kinetic model as simple as possible, we present K M app and V max app values, which were determined using nonlinear regression (Sigma Plot, Jandel Scientific). Here, the hsPxd01 Is a Multidomain Heme Peroxidase APRIL 24, 2015 • VOLUME 290 • NUMBER 17 app value simply represents the substrate concentration where the overall rate is half of its maximum value, and the V max app value represents the limiting step of the overall reaction at certain enzyme and substrate concentrations. Importantly, no reaction was observed upon mixing of only NADH, hydrogen peroxide, and hsPxd01 or hsPxd01-con4.

Properties of Recombinant Protein and Oligomeric
Structure-The average yield of recombinant hsPxd01 produced in HEK 293 cells by stable transfection was 0.5-0.9 mg per liter of harvested media. Recombinant human peroxidasin 1 (hsPxd01) either produced by stable or transient transfection applied to SDS-PAGE had a molar mass of Ͼ500 kDa ( Fig. 2A,  lanes 1-3). The theoretical molar mass of the protein (1445 amino acids without signal peptide, see also Fig. 5) is 163 kDa. Under reducing conditions, the protein band is found at ϳ180 kDa ( Fig. 2A, lanes 4 -6), suggesting that hsPxd01 is an oligomeric protein with monomers being linked by DTT-sensitive covalent bonds. The gels were stained with anti-hsPxd01 antibody ( Fig. 2A, lanes 1 and 4), anti-His antibody (lanes 2 and 5), and Coomassie Blue (lanes 3 and 6).

hsPxd01 Is a Multidomain Heme Peroxidase
The far-UV ECD spectrum of recombinant hsPxd01 showed a clear minimum at 208 nm as well as a shoulder at 220 nm and a maximum around 195 nm corresponding to an overall secondary structure composition of 41% ␣-helices and 30% ␤-sheets (Fig. 2B). Purified recombinant hsPxd01 had a Soret band at 410 nm and additional bands at 529, 563, and 627 nm. Very similar spectra were obtained for the construct hsPxd01-con4 (data not shown). The average Reinheitszahl (purity number, i.e. A 410 nm /A 280 nm ) of hsPxd01 varied between 0.41 and 0.47 corresponding to 72-81% heme occupancy. Molar extinction coefficients at 280 and 410 nm using the Bradford protein assay were determined to be 163,000 and 89,000 M Ϫ1 cm Ϫ1 , respectively. The spectrum of a representative preparation is depicted in Fig. 2C. The ECD spectrum in the visible region shows a minimum at 396 nm and a maximum at 412 nm (inset to Fig. 2C). The heme in hsPxd01 (as well as in the constructs) is covalently linked as demonstrated in Fig. 2D. Continuous wave low temperature EPR spectra of hsPxd01 at pH 7 demonstrates the presence of mostly high spin Fe(III). Fig. 2E depicts a representative experimental spectrum of 20 M hsPxd01 as well as the simulated spectrum (red line) as sum of individual high spin species clearly indicating ligand coordination of axial symmetry. A similar high spin Fe(III) spectrum was also obtained with hsPxd01-con4 (data not shown). Depending on the protein batch, a minor contribution of low spin species was observed in both proteins.
To obtain the exact molar mass and oligomeric composition, purified recombinant hsPxd01 was analyzed by HPLC combined with UV-visible, refraction index, and MALS detectors. Fig. 3A shows that the protein eluted as a single peak at a retention time of 14.35 min. With SEC-MALS, an average size of 504 kDa was obtained, which is in accordance with a homotrimeric structure (Fig. 3B). When SEC was performed under reducing conditions, the most prominent peak eluted at 18.54 min with an average size of 175 kDa. These data clearly demonstrate that this human multidomain peroxidase is a homotrimeric and glycosylated protein. Glycosylation contributes with about 12 kDa to the total molar mass of hsPxd01.
Next, we were interested in which domains were involved in the covalent linkage between the three monomers. Different (truncated) constructs were designed (summarized in Fig. 4A) and transiently expressed in HEK 293 6E cells. Briefly, besides the complete protein (hsPxd01) variants without the VWC domain (hsPxd01-con3), without LRR and VWC domains (hsPxd01-con4), without LRR, Ig, and VWC domains (hsPxd01-con5), and without LRR, Ig and VWC domains and the propeptide N-terminal of the peroxidase domain (hsPxd01-con6) were produced (Fig. 4B). The media of small scale experiments were harvested and directly applied to SDS-PAGE without any further purification step. Under nonreducing conditions the presence of trimeric hsPxd01 was visible, whereas hsPxd01-con3, hsPxd01-con4, and hsPxd01-con5 are monomeric (compare upper panels (Coomassie Blue staining) with lower panels (immunoblotting) in Fig. 4B). Under reducing conditions a similar pattern was obtained except for the fulllength protein, which now gave a band at 180 kDa. In the case of hsPxd01-con6, various bands at a higher molar mass range appeared indicating instability and aggregation. Because the yield of this construct was poor compared with the other constructs, this might indicate that the homologous propeptide region contributed to the conformational stability. However, the main conclusion from these data is that in the absence of the VWC domain the constructs were monomeric, clearly suggesting that hsPxd01 is a homotrimeric protein having the monomers covalently linked by disulfide bridges between their C-terminal VWC domains.
N-Glycosylation Site Occupancy and Disulfide Pattern of Human Peroxidasin 1-Computationally, 11 N-glycosylation sites were predicted for hsPxd01 (Fig. 5). The first one (Asn-390) was located at the second Ig domain on the loop between the C-and D-strand (similar to CH2 glycosylation in IgG1-Fc). In the peroxidase domain seven NX(S/T) motifs were found, four in propeptide region (Asn-640, Asn-699, Asn-719, and Asn-731) and four in the core of the peroxidase domain (Asn-865, Asn-964, Asn-1178, and Asn-1280). Two further N-glycosylation sites were predicted to be at the linker between the peroxidase domain and the VWC domain (Asn-1368) as well as on the VWC module itself (Asn-1425) (Fig. 5).
For experimental verification, hsPxd01 was proteolytically cleaved by either pepsin or trypsin, deglycosylated using PNGase F, and analyzed by mass spectrometry. Table 1 presents a detailed list of identified peptides and their glycosylation status. Glycosylation occupancy was assessed by deamination of asparagines (Asn3 Asp) due to glycan removal by PNGase F. Data demonstrate that except for Asn-964, all other sites were N-glycosylated, four completely (Asn-640, Asn-731, Asn-865, and Asn-1425), whereas the others were found partially glycosylated. Table 1 depicts the relative occupancy in both digests. For example, more than 70% of the recovered peptides at Asn-390, Asn-699, Asn-1178, and Asn-1280 were glycosylated, whereas Asn-719 was mostly nonglycosylated (peptic digest, 100%; tryptic digest, 86.59%).
Furthermore, we analyzed which cysteines are involved in disulfide bridges. Mass spectrometric analysis of recombinant hsPxd01 demonstrated that Cys-12 and Cys-20 (signal peptide), Cys-736 and Cys-970 (peroxidase domain), and Cys-1315, Cys-1316, and Cys-1319 (VWC domain) are free cysteines as demonstrated by the presence of alkylation (Table 2). This suggests that the other cysteines are involved in disulfide bond formation as has been proposed by Uniprot (33). From the 20 predicted bound cysteines, 12 could be confirmed by MS.  Table 2). Eight predicted disulfide bridges could not be found, two in the IgG-like domains, one in the peroxidase domain, and five in the VWC domain. This does not mean that they are not present because during sample preparation bond breakage and subsequent alkylation might occur (this was also partially observed with the verified disulfide bridges). Additionally, the sequence alignment of the four IgGlike domains suggests the existence of one disulfide bridge per hsPxd01 Is a Multidomain Heme Peroxidase APRIL 24, 2015 • VOLUME 290 • NUMBER 17 ␤-barrel structure. Regarding the proposed cystines in the VWC domain it has to be mentioned that its sequence recovery was poor, which might explain that the respective peptides were not found. In general, due to the complex sequence of hsPxd01, attaining a good sequence mapping is quite challenging because the obtained peptides were either rather big or very small. This was the reason for using both pepsin and trypsin in a sequential manner.
Heme Cavity Architecture-To gain further insight into the heme cavity architecture of hsPxd01, we probed its redox properties by spectroelectrochemistry. This allowed the comparison of peroxidasin 1 with other mammalian peroxidases (MPO, LPO, and EPO) and a prediction of the catalytic ability of the enzyme especially regarding the oxidation of halides (Cl Ϫ Ͻ Br Ϫ Ͻ I Ϫ ). Fig. 6A shows the spectral transition upon electrochemical reduction of ferric to ferrous hsPxd01 at pH 7.4 and 25°C. There was a clear direct transition to ferrous hsPxd01 with a Soret maximum at 420 nm and additional bands at 523 and 553 nm and isosbestic points at 413 and 510 nm. The slope of the corresponding Nernst plot (Fig. 6B) is close to the theoretical value for a redox equilibrium involving the exchange of a single electron. The reduction potential of the Fe(III)/Fe(II) couple at pH 7.4 and 25°C was calculated to be Ϫ233 Ϯ 5 mV, which is more negative than that reported for bovine LPO
Thermal Stability and Bromination Activity of Human Peroxidasin 1-Both ECD and DSC were used to probe the thermal stability of the full-length protein as well as hsPxd01-con4 (Fig.  7). These complementary methods allow determination of the unfolding pathway of this multidomain protein. They provide crucial information about stability of and cross-talk between the individual domains of hsPxd01, which might be important for understanding the interaction with its physiological reaction partner(s).
The loss of secondary structure at 208 nm by ECD shows that significant loss of ellipticity already starts around 40°C and ends around 70°C suggesting a multistep noncooperative unfolding pathway (Fig. 7A). By contrast, the ellipticity at the heme Soret band at 412 nm (reflecting unfolding events in the heme cavity of the peroxidase domain) changed within 45 and 60°C (estimated T m at 52.2°C) (Fig. 7B).
To get more detailed information about the unfolding pathway of hsPxd01 and hsPxd01-con4 (Ig-Ig-Ig-Ig-POX), DSC experiments were performed (Fig. 7, C and D). The thermogram of fulllength hsPxd01 is complex and is best fitted by the assumption of a noncooperative (non-two-state or multiple state) transition that includes four distinct unfolding steps. Best fitting is obtained with four endotherms with T m values of 51.5, 61.1, 69.1, and 77.5°C, respectively. Because ellipticity in the heme region is lost between 46 and 60°C (Fig. 7B), it is reasonable to assume that the peroxidase domain has the lowest thermal stability of this multidomain protein with a T m value at ϳ51-52°C. Finally, the enzymatic activity of hsPxd01 and hsPxd01-con4 was probed (Fig. 8). The few available publications about the FIGURE 4. SDS-PAGE of human peroxidasin 1 and four (truncated) constructs under nonreducing and reducing conditions. A, hsPxd01 is full-length protein; hsPxd01-con3 lacks the VWC domain; hsPxd01-con4 lacks LRR and VWC domains; hsPxd01-con5 represents the peroxidase domain only including the "propeptide region"; hsPxd01-con6 represents the peroxidase domain without the "propeptide region." Additionally, the molar masses of the respective (unglycosylated) proteins in the monomeric and (putative) trimeric states are presented. B, electrophoresis of recombinant proteins produced in HEK 293E cells under denaturating and either nonreducing (left panels) or reducing (right panels) conditions. Gels were stained with Coomassie Blue (upper panels) as well as immunoblotted and detected with anti-Strep antibody (lower panels).
halogenation activity of human peroxidasin 1 (12,13,15) are puzzling, because the reported activities are very low compared with chordata peroxidases, including LPO, EPO, and MPO. This observation is somehow contradictory to the high sequence similarity and the presence of all essential amino acids in the active site of subfamily 1 and 2 proteins (see also Figs. 1 and 5). Thus, to examine the kinetics of oxidation of bromide, the effect on the initial rate of bromination of NADH was measured at varying concentrations of bromide and hydrogen peroxide. This assay is very sensitive, and it allowed the detection of apparent kinetic constants of both hsPxd01 and hsPxd01-con4 (Fig. 8). Saturation kinetics behavior was observed in the presence of 5 mM bromide and varying hydrogen peroxide concentrations as well as in the presence of 50 M hydrogen per-  N-Glycosylation and disulfide pattern of human peroxidasin 1. A, schematic presentation of glycosylation sites of hsPxd01. B, multiple sequence alignment (Clustal X, version 1.81) of hsPxd01 with human myeloperoxidase (hMPO) and lactoperoxidase from goat (gLPO). The respective propeptides of hMPO and goat LPO are underlined. Important catalytic residues are highlighted in blue, and amino acid residues forming the covalent links with the prosthetic group are depicted in red. Cysteine residues are marked in gray, and cysteines involved in disulfide bonds (confirmed by MS spectrometry) are connected. Glycosylation sites are highlighted in black, and partially glycosylated residues of POX are shown in black and marked with *. One putative not glycosylated site on POX is boxed in black. Additionally, the distinct domains are marked by different coloring (signal peptide, gray; LRR, yellow; Ig, green; POX, blue; VWC, light gray). Please note that the background highlight colors for different sections of the sequence alignment match the domain coloring indicated in the upper panel.

hsPxd01 Is a Multidomain Heme Peroxidase
oxide and different bromide concentrations. The bromination rate of NADH did not depend on NADH concentration and was entirely related to the rate of hypobromous acid production by hsPxd01 and hsPxd01-con4 as reported in the literature (30,31). Table 3 summarizes the apparent kinetic constants K M app , k cat , and k cat /K M app obtained from these curves. Interestingly, the K M app values for both H 2 O 2 and Br Ϫ for hsPxd01 and hsPxd01-con4 were similar (Table 3), whereas the k cat values of hsPxd01-con4 were more than 15 times higher.

DISCUSSION
The interest in human peroxidasin 1 has increased rapidly over the last years especially since the discovery of its essential role in the formation of collagen IV cross-links (13, 15). It has  APRIL 24, 2015 • VOLUME 290 • NUMBER 17 been shown that the absence of the enzyme has severe and fatal impacts in model organisms (5,6,15), but detrimental mutations in humans have also been reported (35,36). Therefore, a better understanding of the biochemical function and biophysical properties can be of crucial importance.

hsPxd01 Is a Multidomain Heme Peroxidase
Sequence analysis of the multidomain subunit structure of hsPxd01 suggested the occurrence of one enzymatic domain with homology to mammalian peroxidases (3). Similar to the latter, the heme in hsPxd01 is post-translationally modified and covalently linked to acidic amino acids (Asp-826 and Glu-980) (3,37). The proximal heme ligand histidine is likely hydrogenbonded to an asparagine, and at the distal heme cavity all catalytic residues (His-827, Gln-823, and Arg-977) are found (3). The His/Arg pair is involved in the heterolytic cleavage of hydrogen peroxide during compound I formation, whereas the glutamine residue is involved in the binding of halides (3). As a consequence, the observed spectral (Soret band at 410 nm, high spin ferric state) and redox properties of hsPxd01 are comparable with LPO. The reduction potential of the Fe(III)/Fe(II) couple of Ϫ233 Ϯ 5 mV is more negative compared with LPO (i.e. Ϫ176 to Ϫ190 mM) (28,34), EPO (Ϫ126 mV) (28), or MPO (5 mV) (27).
N-Glycosylation occupancy of the peroxidase domain of hsPxd01 could be verified at seven positions. This matches with five N-glycans in mature homodimeric MPO and pro-MPO produced in CHO cell lines (21). Monomeric bovine lactoperoxidase has four N-glycosylation sites (38). Furthermore, the peroxidase domain of hsPxd01 was shown to be stabilized by seven confirmed disulfide bridges, including the propeptide region. One predicted disulfide bond in the peroxidase domain (Cys-959 -Cys-970) could not be confirmed; however, there was also no free Cys-959 detected. In hsPxd01, this region is part of the polypeptide chain, whereas in mammalian peroxidases (which have no additional domains), it is proteolytically cleaved off during maturation. In MPO there are six intra-chain disulfide bridges (and one between the two subunits), whereas monomeric LPO is stabilized by seven cystines. Together with the recently published phylogenetic studies (39), these findings

hsPxd01 Is a Multidomain Heme Peroxidase
suggest a similar overall (mainly ␣-helical) structure as well as heme cavity architecture of the peroxidase domain of hsPxd01 (Fig. 5).
Besides the heavily glycosylated peroxidase domain, three additional glycosylation positions could be verified at the second immunoglobulin domain, the linker between the peroxidase and VWC domain as well as at the von Willebrand factor C domain. Four intradomain cystines could be identified by MS analysis in the LRR region and two in the Ig domains.
Furthermore, we could demonstrate that hsPxd01 is a homotrimeric oxidoreductase. This corresponds to the overall structure of peroxidasin from D. melanogaster reported by Nelson et al. in 1994 (4). Analyses of molar masses of the full-length protein (504 kDa) and truncated versions under oxidized and reduced conditions clearly demonstrated that hsPxd01 is a homotrimeric protein having the protomers covalently bound via (redox-sensitive) cystine bridges at their C-terminal VWC domains. One might speculate whether this trimeric structure of hsPxd01 is related to the fact that type IV collagen before post-translational modification is also a (hetero)trimeric protein. Peroxidasin 1 has been reported to cross-link these triplehelical protomers into oligomers. In detail, during this process hsPxd01 should be responsible for cross-linking two opposing C-terminal NC1 trimers of collagen IV to form a hexameric dimer (13). This collagen "knot" (covalent sulfilimine bonds between Met-93 and Lys-211 at the interface of adjoining protomers) is synthesized by hsPxd01 by the release of hypobromous acid (15). To allow such a specific reaction, a close interaction between the three-armed hsPxd01 and the C termini of the collagen IV protomers must occur. Furthermore, a close proximity between the enzyme and NC1 protomers reduces the oxidative damage by hindering accidental scavenging of hypobromous acid by other residues. One might speculate whether the additional domains (LRR and Ig domains) mediate a suitable distance and orientation by protein-protein interactions.
Compared with myeloperoxidase (40) and lactoperoxidase (41), human peroxidasin 1 shows several intermediate states during temperature-mediated unfolding. This reflects the multidomain architecture of this peculiar peroxidase but also suggests a noncooperative unfolding pathway that might indicate loose interaction between the individual domains. This flexibility could support the interaction of hsPxd01 with the protomers of trimeric collagen IV as described above.
Based on the homology with the human peroxidases and the fact that those efficiently mediate the two-electron oxidation of halides, it was speculated that hsPxd01 also catalyzes the formation of hypohalous acids. The available data from the literature are puzzling and mostly based on end-point determinations after long incubation of hsPxd01 with halides and FIGURE 7. Thermal stability of human peroxidasin 1. A, temperature-mediated unfolding of hsPxd01 followed by ECD spectroscopy at 208 nm. The corresponding fit is depicted in gray. B, temperature-mediated unfolding of hsPxd01 followed by ECD spectroscopy at 412 nm. The corresponding fit is depicted in gray. Temperature-mediated unfolding of hsPxd01 (C) and hsPxd01-con4 (D) followed by differential scanning calorimetry. Original DSC data are depicted in black. Fitting of data (red lines) suggests the occurrence of four consecutive unfolding events. The corresponding midterm transition temperatures (T m values) are depicted. APRIL 24, 2015 • VOLUME 290 • NUMBER 17 hydrogen peroxide (12,13,15). In this study, for the first time the steady-state kinetics of bromide oxidation at physiological pH 7.4 was examined. Oxidation of chloride (140 mM) at pH 7.4 was negligible (data not shown), which is closely related to the hierarchy of the reduction potential of the Fe(III)/Fe(II) couple (i.e. hsPxd01 Ͻ LPO Ͻ EPO Ͻ Ͻ MPO) and the fact that hsPxd01 as well as LPO or EPO are not able to form the electron-withdrawing sulfonium ion heme to protein linkage. The latter is typical for MPO and its strong oxidation capacity (42)(43)(44). However, bromide oxidation by hsPxd01 and hsPxd01-con4 could be monitored continuously and shown to depend on both the hydrogen peroxide as well as the bromide concentration (Fig. 8). This (for the first time) allowed us to determine apparent enzymatic parameters. Interestingly, the K M app values for bromide were similar for both hsPxd01 (4.1 mM) and hsPxd01-con4 (4.4 mM), which reflects the similar spectral and redox properties (and thus heme cavity architecture) for both proteins. Nevertheless, compared with mammalian peroxidases, the apparent catalytic efficiency of bromide oxidation by hsPxd01 is relatively low. Based on the same assay, the apparent catalytic efficiency of bromide oxidation of hsPxd01-con4 is only about 1% compared with MPO (32). But most interestingly

hsPxd01 Is a Multidomain Heme Peroxidase
(despite the fact that it binds bromide with the same affinity), the full-length enzyme is even less active. This discrepancy is surprising at first sight but could be related to the physiological role of hsPxd01 described above. In contrast to the human peroxidases, which are involved in the innate immune system and release antimicrobial hypohalous acids at high flux, such an uncontrolled reactivity would be deleterious for a biosynthetic oxidation of distinct (hydroxyl)lysine and methionine residues. Because there is no x-ray structure of neither full-length hsPxd01 nor a truncated version so far available, one can only speculate about necessary conformational changes that occur upon binding of hsPxd01 to collagen IV thereby increasing the accessibility of the heme cavity. Most probably the structural differences between monomeric hsPxd01-con4 and trimeric full-length hsPxd01 are related to substrate accessibility rather than differences in the heme cavity.
Clearly more studies are necessary to probe this hypothesis. Together with TPO (45), peroxidasins represent peroxidases that are not primarily related with defense of the mature organism (like LPO, EPO, and MPO that release freely diffusing hypohalous acids) but with biosynthetic tasks. In both cases, the target proteins (thyroglobulin and collagen IV) are not able to enter the active site but will be modified specifically by mobile oxidants (hypoiodous acid and hypobromous acid). Understanding theses reactions at a molecular level will be a great challenge in the future.