Hyper-truncated Asn355- and Asn391-glycans modulate the activity of neutrophil granule myeloperoxidase

Myeloperoxidase (MPO) plays essential roles in neutrophil-mediated immunity via the generation of reactive oxidation products. Complex carbohydrates decorate MPO at discrete sites, but their functional relevance remains elusive. To this end, we have characterised the structure–biosynthesis–activity relationship of neutrophil MPO (nMPO). Mass spectrometry demonstrated that nMPO carries both characteristic under-processed and hyper-truncated glycans. Occlusion of the Asn355/Asn391-glycosylation sites and the Asn323-/Asn483-glycans, located in the MPO dimerisation zone, was found to affect the local glycan processing, thereby providing a molecular basis of the site-specific nMPO glycosylation. Native mass spectrometry, mass photometry and glycopeptide profiling revealed significant molecular complexity of diprotomeric nMPO arising from heterogeneous glycosylation, oxidation, chlorination and polypeptide truncation variants and a previously unreported low-abundance monoprotomer. Longitudinal profiling of maturing, mature, granule-separated and pathogen-stimulated neutrophils demonstrated that nMPO is dynamically expressed during granulopoiesis, unevenly distributed across granules and degranulated upon activation. We also show that proMPO-to-MPO maturation occurs during early/mid-stage granulopoiesis. While similar global MPO glycosylation was observed across conditions, the conserved Asn355-/Asn391-sites displayed elevated glycan hyper-truncation, which correlated with higher enzyme activities of MPO in distinct granule populations. Enzymatic trimming of the Asn355-/Asn391-glycans recapitulated the activity gain and showed that nMPO carrying hyper-truncated glycans at these positions exhibits increased thermal stability, polypeptide accessibility and ceruloplasmin-mediated inhibition potential relative to native nMPO. Finally, molecular modelling revealed that hyper-truncated Asn355-glycans positioned in the MPO-ceruloplasmin interface are critical for uninterrupted inhibition. Here, through an innovative and comprehensive approach, we report novel functional roles of MPO glycans, providing new insight into neutrophil-mediated immunity.


Abbreviations used in SI S3
Extended experimental procedures S4-S15 Supplementary Figures S16-S30 Supplementary Figure   Staphylococcus aureus from the pathogenic LS1 lab strain kindly donated by Dr Tao Jin, Department of Rheumatology and Inflammation Research, Institute of Medicine, University of Gothenburg, Sweden was used for the pathogen-based activation of resting neutrophils. Stimulation was performed at an multiplicity-of-infection (MOI) of 1:5 (bacteria:neutrophil) to simulate low-level infection sufficient to trigger neutrophil activation (and MPO degranulation) while avoiding neutrophil cell death. Approximately 10 7 neutrophils isolated from Donor g-i were inoculated with approximately 2 x 10 6 S. aureus bacteria. The inoculate was separated into four tubes and subjected to different incubation periods at 37°C: 0 min, 30 min, 60 min and 120 min. For the 0 min incubation, the samples were briefly inoculated with S. aureus, but were not incubated for any significant length of time (approximately 2-3 min). As negative infection controls, the same number of isolated neutrophils (10 7 ) from the same donors (Donor g-i) were incubated under the same conditions, but without S. aureus inoculation for 0 min and 120 min. As a positive activation control, the same number of isolated neutrophils from the same donors were incubated with 5 µg/mL aqueous cytochalasin B for 5 min at 37°C followed by 0.5 µM aqueous ionomycin (hereafter collectively referred to as CytB/I) for 10 min at 37°C. As a positive control for neutrophil cell death, the same number of isolated neutrophils from the same donors were incubated with 1% (w/v) Triton-X 100, 37°C, 5 min.
To investigate the degranulated MPO (Dg-MPO), an MPO ELISA kit (catalogue number E-80PX, ICL LAB, OR, USA) was used as per the manufacturer's instructions to measure the level of Dg-MPO secreted into the extracellular space across various study conditions. For this purpose, the supernatant fractions of all neutrophil samples were collected after spinning the samples at 2,000 x g, 4°C, 15 min. An MPO standard from the ELISA kit was used to determine the concentration of Dg-MPO in the supernatant fractions. The level of neutrophil cell death was assessed by measuring the level of cytosolic lactate dehydrogenase (LDH) released into the same supernatant fractions using the Cytotoxicity Detection KitPLUS (catalogue number 4744934001, Sigma, Sweden) as per the manufacturer's instructions.
Origin, purification, structural integrity, and activity of neutrophil-derived MPO Enzymatically active human myeloperoxidase (MPO, UniProtKB, P05164) derived from unfractionated resting neutrophils (nMPO) was isolated from pooled donor blood (product number 426-10, Lee BioSolutions Inc, MO, USA). The nMPO was isolated to greater than 95% purity as established by gel electrophoresis using preparation-scale ion exchange and concanavalin A-based chromatography (data not shown). Upon receiving nMPO, the purity of the product was confirmed using reductive SDS-PAGE, which, as expected, showed two characteristic bands corresponding to the MPO light chain (αchain, ~12.5 kDa) and heavy chain (β-chain, ~65 kDa) after Coomassie Blue staining. No other significant protein bands were observed, e.g. Supplementary Figure S7A. In addition, the enzymatic activity of nMPO was validated using a 3,3',5,5'-tetramethylbenzidine (TMB) assay (described below). The concentration of nMPO was determined with a Bradford assays (Sigma-Aldrich, Australia) using bovine serum albumin (Sigma-Aldrich, Australia) as a protein standard in the concentration range 0-2.5 mg/mL. nMPO was aliquoted and stored in a soluble form at a concentration of 1 μg/μL in 50 mM aqueous sodium acetate, 100 mM aqueous sodium chloride, pH 6.0 at -20°C until use.

Glycomics
Glycan release and clean-up For glycan profiling, nMPO was reduced using 10 mM aqueous dithiothreitol (DTT), 56°C, 45 min, and alkylated using a freshly prepared solution of 25 mM aqueous iodoacetamide (IAA), 25°C, 30 min in the dark (both final concentrations) and otherwise handled as previously described (4 The N-glycan fine structures were manually characterised from the PGC-LC-MS/MS data using Xcalibur v2.2 (Thermo Scientific, Australia) based the monoisotopic precursor mass, the glycan fragmentation pattern arising from the negative-ion CID-MS/MS, and the relative and absolute PGC-LC retention time of each candidate glycan (6-9). RawMeat v2.1 (Vast Scientific, www.vastscientific.com) and GlycoMod (https://web.expasy.org/glycomod/) aided the glycan identification process. The biosynthetic relationship between the observed N-glycans served to further enhance the confidence of the reported structures. The relative abundances of the identified N-glycans were established based on the area-under-the-curve (AUC) measurements of extracted ion chromatograms (EICs) performed for all relevant charge states of the monoisotopic glycan precursors using Skyline 64-bit v20.1.0.76 (10,11).
i) Isolated nMPO (20 μg) was reduced and alkylated in technical triplicates as described above. ii) The monoprotomeric (αβ) and diprotomeric (ααββ) forms of nMPO (20 μg protein/replicate) were separated in technical triplicates using non-reductive SDS-PAGE on a pre-cast 4-12% gradient gel (Invitrogen, Australia). Separations were performed using a constant potential of 120 V, 45 min, and was carried out on ice to prevent spontaneous dissociation of diprotomeric nMPO. The gels were stained using Coomassie Blue and the protein bands corresponding to the monoprotomeric (αβ) and diprotomeric (ααββ) nMPO were excised, in-gel digested at 37°C, 16 h using 33.3 ng/μL sequencing grade modified porcine trypsin (Promega, Australia). The resulting peptide mixtures were extracted using repeated washing cycles of ACN and water, and then desalted using C18-SPE micro-columns (Merck-Millipore HF-X Hybrid Quadrupole-Orbitrap mass spectrometer (Thermo Scientific, Australia) coupled to either an UltiMate™ 3000 RSLCnano System (Thermo Scientific, Australia) or an Easy nLC1200 system (Thermo Scientific, Australia). The mass spectrometer was operated in positive ion polarity mode. The peptide mixtures were separated using reversed-phase C18 chromatography using custom-made nano-LC columns packed with either C18 HALO resin (2.7 µm particle size, 160 Å pore size, 0.1 mm inner diameter x 300 mm length, Halo, Advanced Materials Technology, Australia) or C18AQ resin (1.9 µm particle size, 120 Å pore size, 0.075 mm inner diameter x 500 mm length, Dr Maisch HPLC GmbH, Germany) or alternatively using a commercial PepMap100 C18 LC column (3 µm particle size, 100 Å pore size, 0.075 mm inner diameter x 250 mm length, Thermo Scientific, Australia). The peptide mixtures were loaded onto trap columns packed with either C18 HALO resin (2.7 µm particle size, 160 Å pore size, 0.1 mm inner diameter x 35 mm length, Halo, Advanced Materials Technology, Australia) or C18 PepMap (5 µm particle size, 100 Å pore size, 0.3 mm inner diameter x 5 mm length, Thermo Scientific, Australia) and then separated on the analytical columns (above) using different LC gradients at flow rates of 300-600 nL/min.  (12). Raw data were used as input for both search engines. The searches employed tryptic, semi-tryptic or non-specific cleavage specificity with up to two missed cleavages allowed, fixed Cys carbamidomethylation and several glycan and non-glycan variable modifications including Met and Trp mono-and di-oxidation and Tyr mono-and di-chlorination. The glycopeptide data were searched against a comprehensive library of 309 mammalian N-glycans to which several nonconventional monosaccharide compositions including phosphorylated paucimannosidic and phosphorylated oligomannosidic-type N-glycans were manually added in Byonic. The precursor and product ion mass tolerance was 4.5-10 ppm and 20 ppm, respectively, and decoy and contaminant databases were included in the searches.
For the Byonic output, only (glyco)peptides identified with high scores (low false discovery rates, FDRs) were considered (PEP-2D scores < 0.001) (13). The Byonic-identified glycopeptides were further validated by manually interrogating the precursor envelope to confirm correct monoisotopic peak picking and charge state assignment, and by manually checking the HCD-MS/MS data and C18-LC elution time to ensure the presence of the expected product ions and similar LC retention time of related glycopeptides. The confidently identified glycopeptides were quantified based on AUC measurements from narrow EICs performed for all relevant charge states of the monoisotopic glycopeptide precursors (MS1 data) using Skyline 64-bit v20.1.0.76 (10,11) or manually using the Qual Browser in Xcalibur v2.2 (Thermo Scientific, Australia). For both the Skyline-based and manual EICbased label-free quantitation (LFQ), a narrow 5 ppm precursor mass tolerance was applied, and data were manually inspected to ensure correct monoisotopic peak integration and consistent LC retention time across sample runs and an agreement with the glycopeptide identifications provided by Byonic.
For the MaxQuant output, the peptide-to-spectral matches (PSMs) and the protein identifications were filtered to FDRs below 0.01-0.03. The MaxQuant-based LFQ of the non-modified peptides and peptides modified with non-glycan modifications was based on the relative MS1 signal strength of confidently identified precursors (12). The proportion of proMPO of all MPO forms was determined for both the propeptide region and the hexapeptide based on the intensity of related peptide pairs arising from proMPO and maturely processed MPO.
The granulopoiesis data published by (14) were extracted from the ProteomeXchange consortium via the PRIDE partner (data accession number, PXD013785). The dataset consisted of LC-MS/MS (glyco)peptide data of maturing neutrophils isolated from four different maturation stages obtained by fluorescence-activated cell sorted myeloid progenitor cells derived from the bone marrow of four healthy donors including promyelocytes and myelocytes (collectively referred to as PMs), metamyelocytes (MMs), immature neutrophils with band-formed nuclei (BNs), and mature neutrophils with segmented nuclei (SNs). Circulating (mature) neutrophils (polymorphonuclear cells, PMNs) derived from blood from the same four donors were also investigated. Multiple LC-MS/MS datasets (technical replicates) were provided for each of these conditions. For our study, only a single technical replicate from each one of the four biological replicates from each maturation stage was downloaded via PRIDE and re-interrogated using Proteome Discoverer v2.4 (Thermo Scientific, Australia). Peptides and glycopeptides were identified with Byonic v3.6.0 using a human proteome database (UniProtKB, 20,416 entries, downloaded August 2019) and a custom-made neutrophil-focused N-glycan database (206 N-glycans) as the protein and glycan search space, respectively. For the granulopoiesis data sets, the quantification of (glyco)peptides was performed by Minora Feature Detector with the minimum trace length set to 5 and maximum delta retention time of isotope pattern multiplets of 0.2 min within the Proteome Discoverer v2.4 platform. For this purpose, chromatographic alignment of signals of interest was performed and the quantification based on a retention time match within a 1 min tolerance and an m/z match within a 10 ppm tolerance. Minimum S/N thresholds were enabled. Protein expression data and relative mRNA expression levels of MPO were manually retrieved (14).

Characterisation of intact nMPO using native MS and mass photometry
Top down and native MS of intact nMPO Intact mass analyses were performed of i) intact α-chain of nMPO and ii) intact nMPO.
i) The nMPO was reduced using 10 mM aqueous DTT, 56°C, 45 min. The dissociated α-and β-chains of nMPO were then desalted using C4 SPE ZipTips (Millipore, Australia), dried and resuspended in 0.1% (v/v) aqueous FA. Approximately 1 μg (6-7 pmol) protein material was injected on a reversedphase C4 LC column (Proteocol C4Q, 3 μm particle size, 300 Å pore size, 0.3 mm inner diameter × 100 mm length, SGE Analytical Science, Australia) kept at a constant temperature of 25°C and operated at a constant flow rate of 5 μL/min using an Agilent 1100 HPLC system connected directly to a highresolution Agilent 6538 quadrupole-time-of-flight mass spectrometer (Agilent Technologies, Australia). A multi-step LC gradient was applied comprising the following linear steps: a) 100% solvent A containing 0.1% (v/v) aqueous FA for 5 min, b) 0-60% solvent B containing 99% ACN in 0.1% (v/v) aqueous FA ramp over 30 min, c) immediate increase to 100% solvent B, which was kept for 5 min and d) re-equilibration in 100% solvent A for 5 min. The mass spectrometer was operated in high-resolution (4 GHz) positive ion polarity mode using the following settings: The fragmentor potential was +200 V, the MS1 scan range was m/z 400-2,500, the nitrogen drying gas flow rate was 8 L/min at 300°C, the nitrogen nebuliser pressure was 10 psi, the capillary potential was +4.3 kV, and the skimmer potential was +65 V. Mass spectra were viewed, deconvoluted and analysed with MassHunter work station vB.06 (Agilent Technologies, Australia). The intact α-chain of nMPO was profiled by deconvoluting the multi-charged signals into the expected mass range for the MPO α-chain (10-20 kDa) using mass stepping of 0.5. Signals corresponding to the α-chain of nMPO were assigned based on accurate matches better than 2 Da (see Supplementary Table S3A ii) nMPO was desalted and buffer exchanged into 1 M aqueous ammonium acetate, pH 7.0 using Bio-Spin 6 Columns (BioRad, UK). Further, three washes in 1 M aqueous ammonium acetate were performed using 3,000 molecular weight cut-off centrifugal filters (Amicon Ultra-0.5, Merck, UK). The S10 concentration of nMPO was measured using a UV-VIS spectrophotometer (DeNovix, DE, USA) and adjusted to a final concentration of 10 μM with 1 M aqueous ammonium acetate immediately prior to analysis. A modified Q-Exactive (Thermo Scientific, UK) was used for the native MS profiling of intact nMPO (15). Ions were generated by static nano-electrospray using gold-coated capillaries prepared inhouse as described previously (16). The capillary voltage was +1.3 kV, the source was 200°C, the Slens RF was 200, the maximum injection time was 50 ms, the HCD potential was off with in-source trapping enabled, and the desolvation potential was -10 V (positive ion polarity mode). MS1 data were acquired in the range m/z 1,000-15,000 with 10 microscans. Data were processed with Xcalibur v2.2 (Thermo Scientific, UK). Resulting spectra were deconvoluted with UniDec (17), and the signals were annotated using software produced in-house.

Mass photometry of intact nMPO
Intact nMPO was also analysed using mass photometry as described (18). The concentration of nMPO was measured using a UV-VIS spectrophotometer (DeNovix, DE, USA) and adjusted to a final concentration of 10 nM with phosphate buffered saline (PBS). After dilution, nMPO was equilibrated for 15 min prior to analysis. A clean coverslip was assembled for sample delivery using silicone gaskets (CultureWell™ gasket, 3 mm diameter x 1 mm depth, Grace Bio-Labs, OR, USA). For data acquisition, 36 µL sample was injected and data were collected for 60 s after a 10 s initial delay. Data processing and analysis were performed using custom software written in Python (19).

Sequence alignment
The

Visualisation, modelling, and molecular dynamics
Amongst the 18 three-dimensional (3D) protein structures currently available of human MPO from the Protein Data Bank (PDB, http://www.rcsb.org/pdb), the structure of the intact diprotomer obtained with the highest resolution by X-ray crystallography containing both the heme, bromide substrates and crystallisable fragments of Asn-conjugated glycans (PDBID, 1D2V, 1.7 Å) was used for visualisation, modelling and structural assessments. Importantly, the MPO used to generate this crystal structure was purified from human neutrophils and covered nearly the entire α-and β-chains of the mature diprotomer (20). Visualisation and the initial structural assessments were performed using PyMOL Molecular Graphic System v2.3 (Schrödinger, LLC, MA, USA). Signature N-glycans (typically the most abundant N-glycans identified based on glycan and glycopeptide data, see Figure 1B, Figure 1D, Figure 3D, Supplementary Table S2 and Supplementary Table S6) were added in-silico to the four highly occupied sequons of each β chain of diprotomeric MPO (Asn323, Asn355, Asn391, Asn483) to mimic key MPO glycoforms of interest to this study including i) nMPO (referred to as WT in the MD simulations), ii) Endo H-treated nMPO (P1), iii) the highly truncated Asn355-/Asn391-glycophenotype elevated in Se/Pl-MPO (P2) and iv) a semi-truncated Asn355-/Asn391-glycophenotype (P3) included to also explore the structure of MPO when intermediate length glycans were positioned at Asn355 and Asn391, see Supplementary Table S10 for details. The Asn729 was consistently left unoccupied based on the glycoprofiling data. The 3D structures of the glycans were generated using the Carbohydrate Builder functionality available at http://glycam.org. The glycans were conjugated to the Asn side chains using in-house code. The side chain angles were initially set to common values (χ1 = 192°, χ2 = 178°, S11 ψ = 177°, φ = 261°) (21). The code relieved any atomic overlaps between the conjugated glycan and the underlying protein by adjusting these initial values within the observed ranges (21,22). Energy minimisation of the resulting glycoproteins was performed by placing structures a periodic box of approximately 45,000 TIP5P waters with a 10 Å buffer between the glycoprotein and the edges of the box. Energy minimisation of all atoms was performed for 20,000 steps (10,000 steepest decent, followed by 10,000 conjugant gradient).
All molecular dynamics (MD) simulations were performed with the CUDA implementation of the PMEMD (23,24) simulation code, as present in the Amber14 software suite (25). The GLYCAM06j force field (26) and Amber14SB force field (25) were employed for the carbohydrate and protein moieties, respectively. The GAFF forcefield was employed for the heme molecule and the side chains of the residues directly conjugated to the heme (27). A Berendsen barostat with a time constant of 1 ps was employed for pressure regulation, while a Langevin thermostat with two collisions per ps was employed for temperature regulation. A nonbonded interaction cut-off of 8 Å was employed. Longrange electrostatics were treated with the particle-mesh Ewald method (28). Covalent bonds involving hydrogen were constrained with the SHAKE algorithm, allowing an integration time step of 2 fs to be employed (29). All simulations were performed under nPT conditions and the restraints employed were 5 kcal/mol-Å 2 Cartesian. The energy-minimised coordinates were equilibrated at 300 K over 400 ps with restraints on the solute heavy atoms. The system was then equilibrated with restraints on the Cα atoms of the protein for 1 ns, prior to performing five unique 500 ns production MD simulations of WT and P1-3 MPO during which no restraints were employed.
The per-residue solvent accessibilities of each of the MD simulated structures were calculated using the Cpptraj program available in Ambertools 18 and averaged across the five production MD simulations of WT and P1-3 MPO. The values from each protomer in the diprotomeric MPO were combined. Values from WT MPO were statistically compared to those from the P1-P3 simulations using p < 0.05 as a significance threshold. The results were plotted with Gnuplot 5.2 and visualised on the 3D structure of MPO using VMD 1.9.3. Further, the secondary structure, root mean squared deviation (RMSD) and root mean squared fluctuation (RMSF) were calculated over the trajectory using Cpptraj in AmberTools 18.
The cpptraj program was used to extract 1,000 snapshots from the 500 ns MD simulation of the WT and P1 MPO. Each snapshot was aligned with human ceruloplasmin by employing a 3D structure of the ceruloplasmin-MPO complex (PDBID, 4EJX, 4.7 Å) (30) using VMD 1.9.3 (31). The degree to which each MPO glycan would overlap with the complexed ceruloplasmin was visualised by displaying each shape of the MPO glycan from the 1,000 snapshots in VMD 1.9.3.

Biosynthesis-related solvent accessibility of MPO
To assess the relative solvent accessibility to each of the glycosylation sites of MPO and their conjugated N-glycans in the context of MPO biosynthesis and processing (see Figure 2), five highresolution 3D structures of diprotomeric human MPO were selected from the Protein Data Bank including 1D2V (1.75 Å), 1CXP (1.8 Å), 1DNU (1.85 Å), 1DNW (1.9 Å), and 5FIW (1.7 Å). Two separate assessments were then performed using these five structures of human MPO.
Firstly, to determine the relative solvent accessibilities to Asn323, Asn355, Asn391, Asn483, and Asn729 of monoprotomeric MPO (αβ) as the trafficking protein would appear in the early-stage biosynthesis within the cis-Golgi, the five PDB files were manually edited to create an artificial monoprotomeric MPO comprising a single αand β-chain. The five monoprotomeric MPO structures were deposited and processed using the Glycoprotein Builder functionality of GLYCAM web (http://glycam.org), which stripped the structures of all hetero-atoms, water, ions, heme and glycans, and energy-minimised the resulting glycoprotein structures. The relative solvent accessibilities of the side chains of non-glycosylated Asn323, Asn355, Asn391, Asn483, and Asn729 were then established using NACCESS, which rolls a spherical probe (5 Å radii) on the surface of MPO to determine the atomic accessible areas (Van der Waal's interaction) by providing unit-less output values that are comparable across sites and protein structures (32).
Secondly, to determine the relative solvent accessibilities of N-glycans proximal to the dimerisationinterface of diprotomeric (ααββ) MPO in the context of late-stage MPO biosynthesis/processing in the late Golgi, or shortly before reaching or upon arrival in the granules, the five selected PDB structures were deposited and processed in the Glycoprotein Builder in their mono-and diprotomeric forms as detailed above. However, for these experiments FA1-glycans (see Figure 1 for structure) were attached to Asn323, Asn483 and Asn729 carrying a significant level of processed N-glycans as determined using glycopeptide analysis. Of these three sites, Asn323 and Asn483 were proximal while Asn729 was distal to the dimerisation interface. The under-processed Asn355 and Asn391 largely carrying oligomannosidic N-glycans were not considered in this assessment. The PDB output files were applied to NACCESS using a 5 Å radii probe as detailed above, but the accessibility measurements differed slightly by this time considering the hetero-atoms enabling the determination of the relative solvent accessibilities to the conjugated N-glycans rather than the underlying Asn residues. The relative solvent accessibilities to the entire monosaccharide residue of the terminal α1,3-arm-positioned 1,2-N-acetylβ-D-glucosamine (GlcNAc) moiety of the FA1-glycans attached to each site of the mono-and diprotomeric MPO as it would be experienced by the processing glycoside hydrolases i.e. the N-acetylβ-hexosaminidase isoenzymes were used for ratiometric analyses.

Endoglycosidase H-treatment of nMPO
Intact nMPO (0.5-15 μg) was incubated with or without 1 μL recombinant Streptomyces plicatus endoglycosidase H (Endo H) expressed in E. coli, 500 U/μL (Promega, Australia) in a total volume of 60-150 μL buffer containing PBS or 5 mM aqueous sodium acetate and 10 mM aqueous sodium chloride, 37°C, 16 h, under native conditions. Blanks (water) and samples containing only Endo H and ceruloplasmin (see below) were included as controls. The untreated nMPO, Endo H-treated nMPO and controls were used immediately for functional assays and characterised using circular dichroism (CD), SDS-PAGE and LC-MS/MS.

Chlorination and oxidation activity of MPO (activity assay 1-3)
Three independent enzyme activity assays designated activity assay 1-3 were performed to establish the chlorination and oxidation activity of the granule-separated MPO in crude protein mixtures, and the isolated Endo H-treated and untreated nMPO.

Activity assay 1
The chlorination activity of MPO was determined by the formation of hypochlorous acid (HOCl) captured via taurine per time as described (33). Firstly, a HOCl standard curve was generated by adding known concentrations of HOCl freshly made from an aqueous sodium hypochlorite solution (catalogue number 425044, Sigma, Australia) to a final concentration of 5 mM aqueous taurine (Sigma, Australia) in PBS, which was kept on ice for 30 min prior to measurements. For the standard curve, the concentrations of hypochlorite (OCl -) were determined by measuring its absorbance at 292 nm at pH 12 (ε292 = 350 M -1 cm -1 ), and solutions were subsequently neutralised before use. The relative and absolute chlorination activities of granule-fractionated MPO were established using 2 μg total protein extract from each of the Az, Sp/Ge and Se/Pl fractions from Donor a-b in technical triplicates and triplicate experiments using 0.5 μg Endo H-treated and untreated nMPO per replicate, respectively. The protein samples were mixed with 5 mM aqueous taurine (final concentration) and firstly incubated for 30 min at room temperature. Reactions were then initiated by the addition of fresh hydrogen peroxide (H2O2) (catalogue number 516813, Sigma, Australia) to a final concentration of 20 μM. The reactions ran for 20 min at room temperature, before being stopped by the addition of 20 μg/mL aqueous bovine catalase (final concentration, Sigma, Australia). In total, 200 μL of each reaction and the HOCl standards were mixed with 50 μL developing reagent containing 2 mM aqueous TMB in 400 mM aqueous acetate buffer, 10% (v/v) aqueous dimethylformamide and 100 μM aqueous sodium iodide, pH 5.4 in a 96-well plate. After lightly shaking the plate for 5 min, the absorbance of each sample was measured using a plate reader at 650 nm (SPECTROstar Nano, BMG Labtech, Germany). The absolute amount of HOCl produced was determined for Endo H-treated and untreated nMPO based on a pre-established standard curve after baseline subtraction using water blanks and after subtracting readings of the Endo H only samples from the Endo H-treated nMPO (called "nMPO + Endo H^"). The relative chlorination activities of the granule-fractionated MPO were determined after blank correction of absorbance measurements and normalisation to the relative MPO protein levels in each granule fraction determined using LC-MS/MS peptide-based quantitation, see Figure 3B for relative MPO protein levels across granules.
Activity assay 2 The 3,3',5,5'-tetramethylbenzidine (TMB) reagent is a chromogenic substrate that can report on the MPO oxidation activity. This assay was used to measure the relative oxidation activities of granulefractionated MPO in 2 μg total protein extract from each of the Az, Sp/Ge and Se/Pl fractions from Donor a-b (each in technical triplicates), and of 0.05 μg Endo H-treated and untreated nMPO per replicate in technical triplicates. The protein samples were diluted in PBS to a total sample volume of 50 µL and subsequently mixed with 150 µL TMB solution (catalogue number T4444, Sigma-Aldrich, Australia) in a 96-well plate and incubated for 10 min at room temperature. The reaction was quenched by the addition of 100 µL 1 M aqueous sulphuric acid (Sigma, Australia) and the colour intensity was measured on a plate reader at 450 nm absorbance (SPECTROstar Nano, BMG Labtech, Germany). The relatively oxidation activities of Endo H-treated and untreated nMPO were established after baseline subtraction using water blanks and after subtracting absorbance values of an Endo H only sample from the absorbance values measured of Endo H-treated nMPO (called "nMPO + Endo H^"). The relatively oxidation activities of the granule-fractionated MPO were determined after blank correction of absorbance measurements and normalisation to the relative MPO protein levels in each granule fraction determined using LC-MS/MS peptide quantitation, see Figure 3B.
Activity assay 3 The 1,2-phenylenediamine reagent is another chromogenic peroxidase substrate that can be utilised to measure the oxidation activity of MPO. This assay was used to measure the relative oxidation activities of 0.05 μg Endo H-treated and untreated nMPO per replicate in technical triplicates. Insufficient material from the granule-separated MPO in crude protein mixtures from Donor a-b was available to probe these samples with this activity assay. The protein samples were diluted in PBS to a total volume of 50 μL which was mixed with 150 μL aqueous 1,2-phenylenediamine (0.4 mg/mL final concentration, Sigma, Australia) and 0.015% (v/v) aqueous H2O2 (Sigma, Australia) solution and incubated for 10 min at room temperature in the dark. The reaction was quenched by the addition of 100 µL 1 M aqueous sulphuric acid (Sigma, Australia) and the colour intensity was measured on a plate reader at 492 nm absorbance (SPECTROstar Nano, BMG Labtech, Germany). The relatively oxidation activities of Endo H-treated and untreated nMPO were established after baseline subtraction using water blanks and after subtracting absorbance values of an Endo H only sample from the absorbance values measured of the Endo H-treated nMPO (called "nMPO + Endo H^").

Ceruloplasmin inhibition of MPO
Glycoform-dependent inhibition of MPO by ceruloplasmin (UniProtKB, P00450) was investigated using Endo H-treated and untreated nMPO. Three enzyme activity assays (activity assay 1-3) carried out as described above were employed, but the Endo H-treated and untreated nMPO were incubated in technical triplicates in the presence or absence of 100 nM ceruloplasmin purified from human serum (final concentration, catalogue number 187-51, Lee BioSolutions Inc, MO, USA) for 30 min at room temperature, prior to the commencement of the MPO enzyme activity measurements. Samples containing only ceruloplasmin were added to the set of other controls included for the three enzyme activity assays (see above). The absolute chlorination activities were determined for Endo H-treated and untreated nMPO with and without ceruloplasmin inhibition by activity assay 1 and displayed as HOCl produced per min as determined by the pre-established standard curve. The relative oxidation activities were determined by activity assays 2 and 3 as established from relative absorbance values. For all assays, baseline subtraction using water blanks and corrections for the absorbance of ceruloplasmin and Endo H were applied called "nMPO + Cp^" and "nMPO + Endo H + Cp^" to enable direct comparisons to nMPO and nMPO + Endo H^, respectively. nm and data were plotted as the average of the spectra from 20-50°C before any of the protein samples melted. All wavelength scan spectra were baseline-corrected by subtracting a water spectrum collected in the same cuvettes used for the protein samples. The spectra obtained from the two independent sample preparations were of the same shape and very similar magnitudes. Thus, only one set of data have been presented. To enable direct comparisons between the nMPO glycoforms for the CD spectral profile and the temperature stability experiments, the readings from the Endo H only sample were subtracted from the measurements performed for the Endo H-treated nMPO (denoted "nMPO + Endo H^"), plotted and compared to the CD data of untreated nMPO.

Circular dichroism profiling, temperature stability and secondary structure prediction
For structure fitting, all spectra data were converted to  (mol -1 cm -1 dm 3 ) per residue based on the Bradford assay concentrations. CD spectra were analysed using a graphical user interface version [GUI to Implement SOMSpec, a CD Secondary Structure Fitting Approach 2018, gitlab.com] of our validated self-organising map approach to secondary structure fitting (34,35). In this approach a reference set, in our case SP175 (36) (truncated to 190 nm) augmented with a constructed 100% helix and 100% random coil peptide data (37) was organised into a map with spectra of similar shape in the same neighbourhood. Another map, which overlays the first, contains the secondary structure contents of the reference proteins. The secondary structures of the intervening nodes were determined by interpolation from five neighbouring references spectra. An unknown spectrum is placed on the map where it minimised the distance (difference) between it and the spectral shape and magnitude of a nodal spectrum. The unknown protein's secondary structure was then assigned to be that of the node. We ran SOMSpec with a 50 × 50 map, five best matching units, and a wavelength range from 240-190 nm. The goodness of fit of the secondary structure estimates were assessed by considering the overlay of the experimental and predicted spectrum visually and by a normalised RMSD (NRMSD). Visual inspection weights maxima, minima and where a spectrum crosses the zero-line more than other points in the spectrum, whereas the NRMSD approach equally weights all wavelengths. In general, for spectra from 240-190 nm an NRMSD < 0.02 was used as a threshold to indicate a very good fit whereas higher NRMSD indicate more uncertainty in the prediction. Higher values usually require user-interference to choose the best fit.
It was visually apparent from the room temperature spectra that the ratios of the 190 nm:208 nm intensities were smaller than expected for highly helical or sheet proteins. In addition, our measured  values were small, suggesting systematic concentration errors due to the use a non-glycosylated reference protein. In accord with this, attempts to use SOMSpec to predict structures from the spectral data were not satisfactory (NRMSD > 0.08). We therefore removed fractions of a random coil spectrum (in 0.025 steps) for the SufI-KK peptide MSLSKKQFIQASGIALCAGAVPLKASA (37), and also scaled the spectra to adjust for overestimates of the protein concentrations (in 10% steps) (38). We then reran SOMSpec to get the best possible fit. The structure-fitting results for best fit derandomised and scaled spectra are provided in Supplementary Table S11C as are the secondary structure percentages when the removed random coil was added back in. There were two options for best fit depending whether one considers the whole wavelength region or emphasises from 190-210 nm as the 222 nm region is unusually small. Consideration of the crystal structures led us to select the rows where the Bradford-determined concentration was twice the protein residue concentration and 12.5-15% random coil removed for the fitting process. Overall the CD data suggest that both Endo H-treated and untreated nMPO in solution are ~34% helix, ~11% sheet, ~16% bonded turns, ~8% bends, and ~31% other structures (including random coil) with each number being indicative not absolute.

Data representation and statistics
Statistical significance was tested using one-or two-tailed paired or unpaired Student's t-tests as indicated for each experiment. p ≥ 0.05 was chosen as the universal threshold to reject various null hypotheses (e.g. glycosylation sites are equally solvent accessible or MPO glycoforms display the same enzyme activity). For each test, the statistical confidence has consistently been indicated by *p < 0.05, **p < 0.01, ***p < 0.005, ****p < 0.001 and *****p < 0.0005. ns signifies a non-significant test (p ≥ 0.05). The n values representing either biological replicates or technical (sample handling or analytical) replicates have been indicated for each experiment. Data points have been plotted as the mean, and the error bars consistently represent their standard deviation (SD). The N-glycans were depicted according to the established symbol nomenclature (39), see also Figure 1 for key.

Supplementary Figures
Supplementary Figure S1 Supplementary Figure   i a nc c a n an i ic n            Residues with reduced (blue) or increased (red) accessibilities (p < 0.05) are mapped with an intensity shading reflecting the proportional duration of the differential accessibility, see key. Darker residue intensity signifies a longer duration of differential accessibility during the simulations. The glycans are shown as 3D-SNFG symbols. The heme moiety is represented in yellow sticks.