Human Neutrophils Secrete Bioactive Paucimannosidic Proteins from Azurophilic Granules into Pathogen-Infected Sputum*

Background: Protein paucimannosylation is considered an important invertebrate- and plant-specific glycoepitope. Results: Azurophilic granule-specific human neutrophil proteins from pathogen-infected sputum displayed significant core-fucosylated paucimannosylation generated by maturation- and granule-specific β-hexosaminidase A and were preferentially secreted from non-lysosomal origins into sputum upon P. aeruginosa stimulation. Conclusion: Human neutrophils produce, store, and selectively secrete bioactive paucimannosidic proteins. Significance: This work will aid in understanding the function(s) of human paucimannosylation in glycoimmunology. Unlike plants and invertebrates, mammals reportedly lack proteins displaying asparagine (N)-linked paucimannosylation (mannose1–3fucose0–1N-acetylglucosamine2Asn). Enabled by technology advancements in system-wide biomolecular characterization, we document that protein paucimannosylation is a significant host-derived molecular signature of neutrophil-rich sputum from pathogen-infected human lungs and is negligible in pathogen-free sputum. Five types of paucimannosidic N-glycans were carried by compartment-specific and inflammation-associated proteins of the azurophilic granules of human neutrophils including myeloperoxidase (MPO), azurocidin, and neutrophil elastase. The timely expressed human azurophilic granule-resident β-hexosaminidase A displayed the capacity to generate paucimannosidic N-glycans by trimming hybrid/complex type N-glycan intermediates with relative broad substrate specificity. Paucimannosidic N-glycoepitopes showed significant co-localization with β-hexosaminidase A and the azurophilic marker MPO in human neutrophils using immunocytochemistry. Furthermore, promyelocyte stage-specific expression of genes coding for paucimannosidic proteins and biosynthetic enzymes indicated a novel spatio-temporal biosynthetic route in early neutrophil maturation. The absence of bacterial exoglycosidase activities and paucimannosidic N-glycans excluded exogenous origins of paucimannosylation. Paucimannosidic proteins from isolated and sputum neutrophils were preferentially secreted upon inoculation with virulent Pseudomonas aeruginosa. Finally, paucimannosidic proteins displayed affinities to mannose-binding lectin, suggesting immune-related functions of paucimannosylation in activated human neutrophils. In conclusion, we are the first to document that human neutrophils produce, store and, upon activation, selectively secrete bioactive paucimannosidic proteins into sputum of lungs undergoing pathogen-based inflammation.

Asparagine (N)-linked glycosylation adds unprecedented structural and functional heterogeneity to polypeptide chains by the covalent attachment of oligosaccharides (hereafter called glycans) to motif-specific asparagine residues. Although advances in glycobiology and analytical glycoscience continually improve the understanding of protein N-glycosylation, the structure-function relationships of most protein glycoforms remain unknown. Dysregulation of protein N-glycosylation by the deletion or modulation of its diverse intraand extracellular functions is a common cause and/or effect of numerous pathologies (1,2). Our understanding of the conserved mammalian N-glycosylation as a structural and functional modulator of proteins is most critically built on the principles of the well described glycoprotein biosynthetic machinery (3). The defined complex and dynamic enzymatic biosynthesis of glycoproteins in the secretory pathway produces three well recognized N-glycan classes that are abundantly displayed on mammalian proteins, i.e. high mannose, hybrid, and complex type, all of which are based on a common trimannosylated chitobiose core (2).
Contrary to this dogma, several recent glycomics-based studies indicate that a fourth type of protein N-glycosylation, referred to as paucimannosylation, with monosaccharide compositions less than or equal to the N-glycan trimannosylchitobiose core, i.e. mannose(Man) 1-3 fucose(Fuc) 0 -1 N-acetylglucosamine(GlcNAc) 2 , is present in mammals. These structures do not correspond to the defined N-glycan types nor can their synthesis be described by established mammalian biosynthetic pathways. To date, mammalian protein paucimannosylation has been suggested to be present in (i) human buccal epithelial cells (4), (ii) human colorectal cancer epithelial cells and tissue (5-7), (iii) kidney tissue from mice suffering from systemic lupus erythematosus (8), (iv) mouse embryonic neural stem cells (9), and (v) rat brain (10). In addition, we recently indicated the presence of paucimannosylation in pathogen-infected sputum derived from individuals with cystic fibrosis (CF) 2 and upper respiratory tract infection (URTI) (11). Importantly, these observations were all based on molecular profiling of N-glycans released from cell/tissue-derived proteins and thus disregarded the protein carrier identities. Consequently, exogenous origin(s) of paucimannosylation could not be ruled out. Mammalian paucimannosylation was supported by immunohistochemistry and immunocytochemistry of selected human and murine tissues and cells using paucimannose-reactive antibodies (12)(13)(14). In general, however, mammals including human and mice have usually been reported to lack protein paucimannosylation (15)(16)(17)(18)(19)(20)(21). As such, human paucimannosylation remains controversial in the context of our current understanding of mammalian glycobiology.
Here, we present unequivocal evidence that paucimannosylation is also a significant host-derived molecular signature of sputum proteins from pathogen-infected human lungs. Enabled by recent developments in system-wide biomolecular detection, we document that inflammation-associated proteins, localizing to the azurophilic granules of human neutrophils, abundantly display paucimannosylation. In line with their presence in specific micro-environments that are central to inflammation and pathogen infection, we confirm that the timely expressed human azurophilic granule-resident ␤-hexosaminidase A (Hex A) enzymatically facilitates the generation of protein paucimannosylation by trimming hybrid/complex type N-glycan intermediates using a machinery, which is formed during early myeloid maturation, and functionally associate paucimannosidic proteins with roles in innate immunity upon secretion from activated human neutrophils.

EXPERIMENTAL PROCEDURES
Sputum and Bacteria Origin/Handling-Saliva-free whole sputum (Ͼ1 ml/donor) was sampled with informed consent from individuals with (n ϭ 5) or without (n ϭ 4) CF by noninvasive expectoration at Westmead Hospital, Sydney, Australia (see Ref. 11 for donor data). Two of the non-CF individuals were diagnosed with URTI and two were diagnosed with pathogen-free pneumonia or chronic obstructive pulmonary disease. The sputum of the seven pathogen-positive individuals was infected primarily by mucoid/non-mucoid Pseudomonas aeruginosa, but also Aspergillus fumigatus, Staphylococcus aureus, and Streptococcus pneumoniae were identified. P. aeruginosa laboratory wound (PAO1) and CF sputum (PASS1-4) strains were isolated and cultured (Table 1). Sputum from all donors showed inflammation characteristics (Ͼ1 ϫ 10 10 polymorphonuclear cells/l sputum). Soluble proteins were isolated from washed sputum plugs (whole sputum) as described (11). In brief, sputum proteins were reduced, and alkylated and intact cells, cellular debris, and insoluble mucins/proteins were removed by centrifugation. The concentration of soluble sputum proteins was measured (Direct Detect, Millipore) and normalized prior to biomolecular characterization.
Origin and Isolation of Human Neutrophils-Resting human neutrophils were isolated to high purity from healthy blood donors as described (29,30). In brief, the neutrophils were isolated using dextran sedimentation (1 ϫ g), hypotonic lysis of erythrocytes, and centrifugation in a Ficoll-Paque gradient. Resting neutrophils were resuspended in a Krebs-Ringer phosphate buffer and counted. Neutrophils were pelleted (4,000 ϫ g, 5 min), and proteins were extracted by cell lysis (6 M Handling of Human Neutrophil-like Cells-Human promyelocytic leukemia cells (HL-60, ATCC CCL-240) were differentiated (5-6 days, 1.3% (v/v) DMSO, Sigma) to meta/band/segmented neutrophil-like cells. HL-60 cells were cultured (RPMI 1640, Gibco, 10% fetal bovine serum, 2 mM L-glutamine, and 50 units/ml penicillin and 50 g/ml streptomycin) at 37°C under 5% CO 2 . High differentiation efficiencies (Ͼ50%) were obtained as assessed by morphology of Wright-Giemsa stained cells (see below). Cells were washed in PBS before use.
Sequence Homology of Hex A-The sequence homology of human Hex A (␣/␤) to hexosaminidases from paucimannoserich organisms was assessed using EMBOSS Needle.
MBL Binding Assay-Binding of paucimannosidic proteins derived from pathogen-positive sputum and commercially sourced N-glycans (Dextra Laboratories) to agarose-conjugated mannose-binding lectin (MBL) (Thermo Scientific) was assessed in 300 l of 20 mM CaCl 2 , 1.25 M NaCl, 10 mM Tris-HCl (aq), pH 7.4, in a ratio of 1 nmol of analyte to 100 l of washed MBL-agarose slurry (12 h, 4°C, end-over-end rotation). The unbound fractions were collected for separate analysis after centrifugation. MBL-agarose beads were washed twice in 500 l of binding buffer. Bound analytes were eluted with 300 l of 2 mM EDTA, 1.25 M NaCl, 10 mM Tris-HCl (aq), pH 7.4 (8 h, 25°C, end-over-end rotation). Paucimannosidic proteins and N-glycans in the MBL unbound and bound fractions were glycoprofiled using PGC-LC-MS/MS after glycan release/ clean-up (see explicit description above).
Statistics-The significance of the individual experiments was assessed by one-or two-tailed Student's t tests where p less than 0.05 was chosen as the minimum acceptable level of confidence to support a rejection of the proposed null hypothesis, e.g. difference of two means. In general, tests fulfilling the minimum confidence level of significance were indicated by *; stronger confidence was indicated by ** and ***. The sample number (n) is given for the individual experiments. Data points are presented as a mean, and error is presented as S.D. or S.E.
Herein, we undertake a thorough investigation of this indication of human protein paucimannosylation by performing in-depth spatio-temporal analyses of the structure, function, and biosynthesis of paucimannosidic proteins in neutrophil-rich sputum from pathogen-infected individuals, isolated blood-derived human neutrophils, and neutrophil-like cells (HL-60). The detailed structures and distribution of five chromatographically pure paucimannosidic N-glycans (M1F, M2, M2F, M3, and M3F) were determined in pathogen-infected sputum (Fig. 1B). M2F (Man␣1,6Man␤1,4GlcNAc␤1,4(Fuc␣1,6)GlcNAc) was consistently the most abundant paucimannosidic N-glycan; the corresponding ␣1,3-isomer of M2F (and M2) was absent. The detailed N-glycan characterization was facilitated by de novo MS/MS sequencing and by spectral and PGC-LC retention time matching to paucimannosidic reference compounds (Fig.  1C, see supplemental Table 1 for supporting N-glycome data). Although M0F per se does not fall under our definition of pauci-  Table 1); red refers to a NaBH 4 -reduced N-glycan reducing end. EIC, extracted-ion chromatogram; Rel. intensity, relative intensity. D, the paucimannose-recognizing Mannitou antibody was highly reactive toward CF sputum proteins (panel i). CF and URTI sputum proteins (zoom of 50 -60-kDa region, short exposure) were more reactive toward Mannitou relative to proteins derived from pathogen-free sputum (panel ii). CMB, Coomassie blue; Non-infec., non-infected; Ctrl, negative control. For all data: mean Ϯ S.D. mannosylation, the presence of this paucimannosidic-related structure was confirmed in the CF sputum. Trace levels of the truncated M1, M0, (Fuc␣1,6)GlcNAc, or single GlcNAc residues may have been present below the detection limit or not released efficiently by N-glycosidase F. In support of these observations, Western blotting using the paucimannose-reactive Mannitou antibody showed high reactivity to CF and URTI sputum proteins (Fig. 1D).
Possible exogenous bacterial origins of the abundant paucimannosylation in sputum were ruled out by the absence of paucimannosidic N-glycan signatures of proteomes obtained from isolated and cultured laboratory wound (PAO1) and CF (PASS1-2) P. aeruginosa strains. In addition, no significant ␣-sialidase, ␤-galactosidase, ␤-hexosaminidase, ␣-fucosidase, and ␣-mannosidase activities were detected in any of the investigated P. aeruginosa strains using a series of digestion assays with well characterized glycoproteins displaying a spectrum of glycoepitopes and LC-MS/MS N-glycan profiling, thus confirming that sputum paucimannosylation does not result from exogenous P. aeruginosa exoglycosidase activities.
Granule-specific Paucimannosylation of Human Neutrophil Proteins-LC-MS/MS-based proteome mapping of CF sputum revealed the protein characteristics of significant leukocytes, e.g. abundance of MPO, neutrophil elastase (NE), eosinophil peroxidase, lactoferrin, catalase, and aminopeptidase N (see supplemental Table 2 for supporting proteome data). The neutrophil-specific proteome (as predicted from a relatively unique transcriptional profile of human neutrophils among other immune cell types (36)) was significantly represented in the CF sputum proteome (63% of 672 human proteins) (34) (Fig. 2A), which was supported by morphology-based identification by microscopy of sputum neutrophils (data not shown). Exogenous bacterial proteins were negligible in the CF sputum proteome, i.e. P. aeruginosa proteins constituted Ͻ2% of the sputum proteome. Further evidence supporting strong neutrophilic molecular signatures in sputum was obtained by the near identical N-glycomes of CF sputum proteins and blood-derived  (34) shows that the putative glycoproteins (when not considering their type of N-glycosylation) localize to all four main compartments of the human neutrophil, but preferentially to the azurophilic granules (red) (see supplemental Tables 1 and 2 for more). Graphics modified and used with permission from Blausen Medical (Blausen Gallery 2014). human neutrophil proteins (R 2 ϭ 0.90); in particular, strong correlation of the paucimannosidic N-glycans (R 2 ϭ 0.96) was observed (Fig. 2B).
Utilizing sequon-based (NX(T/S), X P) prediction of N-glycosylation and published granule proteome libraries of human neutrophils (34), the putative N-glycoproteins of CF sputum were shown to localize to all four main subcellular granular compartments of the human neutrophil, i.e. azurophil, specific, gelatinase/ficolin granules, and secretory vesicles. However, significant proportions of the N-glycoproteome (ϳ40%) resided in the azurophilic granule (Fig. 2C).
Unequivocal evidence for human protein paucimannosylation in pathogen-infected sputum and its granule specificity was generated by system-wide mapping of intact glycopeptides using our recently developed glycoproteomics technology (10). Site-specific paucimannosylation was identified on 18 abundant proteins (23 N-sites, 35 unique N-glycopeptides) of the total of 30 human N-glycoproteins identified in CF sputum (36 N-sites, 115 unique N-glycopeptides), quantitatively covering ϳ20% of the CF sputum proteome ( Table 2,  see also supplemental Table 3 and supplemental Fig. 1 for supporting glycoproteome data). By overlaying these data onto granule proteome libraries of human neutrophils (34), paucimannosylation was found to be highly enriched in the azurophilic granules (p ϭ 2.3 ϫ 10 Ϫ4 ) relative to the other three main compartments of the neutrophil (Fig. 3, A and E). High mannose and complex type N-glycoproteins displaying ␤-galactosylation, Lewis type fucosylation, and ␣-sialylation localized predominantly to the other granules (Fig. 3, B-E). These observations were supported by partial co-localization of the azurophilic marker MPO and paucimannosidic epitopes in DMSO-differentiated human neutrophil-like cells using immunocytochemistry (data not shown).
Spatio-temporal Paucimannose Generation by Human ␤-Hex A-The biosynthetic mechanisms of human paucimannosylation were investigated. We have previously shown that N-glycan processing and the solvent accessibility of N-glycosylation sites on maturely folded proteins are closely correlated (37). The identified paucimannosidic N-glycosylation sites on sputum proteins were found to be significantly more accessible than the spatially hidden high mannose sites (p ϭ 8.0 ϫ 10 Ϫ4 ) showing solvent accessibilities similar to the highly processed complex sites (Fig. 4A), indicating that paucimannosylation results from significant exoglycosidase processing of the solvent-exposed N-glycan structures. No specific sequence recognition motifs for the paucimannosidic N-glycosylation sites were evident as assessed by a frequency plot.
Partial co-localization of human Hex A and paucimannosidic glycoepitopes as evaluated by immunocytochemistry of DMSOdifferentiated human neutrophil-like cells supported the involvement of Hex A in paucimannose production (Fig. 4F). Azurophilic granule residence of human Hex A was indicated by moderate/strong co-localization of Hex A with the azurophilic marker MPO (Fig. 4G). This was supported by proteomics-based identifications of ␣ and ␤ subunits of Hex A in isolated azurophilic granules of human neutrophils (34,38). High expression of the genes coding for the putative paucimannosidic enzymes, e.g. HEXA and HEXB (coding for ␤-hexosaminidase subunit ␣ and ␤, respectively, which together hydrolyze ␤-GlcNAcand ␣-GalNAc-terminating glycoconjugates), and for paucimannosidic proteins, e.g. MPO and AZU1 (coding for MPO and azurocidin, respectively) in promyelocytes, relative to levels in myelocytes and mature (resting) neutrophils, indicated assembly of the synthetic machinery for paucimannosylation (but not necessarily the complete biosynthetic generation of paucimannosidic proteins), early in the bone marrow maturation (Fig. 4H). M3(F) truncation to M2(F), M1(F), and M0(F) may be facilitated by human ␣and ␤-mannosidases previously identified in the azurophilic granules of human neutrophils (34,38); promyelocytic stage-specific expression of the corresponding mannosidase genes, i.e. MAN2A2/MAN2B2 (coding for ␣-mannosidases, which hydrolyze terminal ␣1,3/6-linked mannosides) and MANBA (coding for ␤-mannosidase, which hydrolyzes terminal ␤-linked mannosides), was indeed observed. Taken together, we propose a new granule-and maturation-specific assembly of the biosynthetic machinery for human protein paucimannosylation in the azurophilic granules during early myeloid maturation of neutrophil precursors in the bone marrow (Fig. 4I).

Binding of Paucimannosidic Proteins to Mannose Receptors-
Potential involvement of protein paucimannosylation in complement activation was assessed by evaluating the binding capacity of paucimannosidic glycoepitopes to MBL. All paucimannosidic glycoforms carried by the CF sputum proteins showed affinities to MBL (Fig. 5C). Isolated paucimannosidic N-glycans from commercial sources showed similar binding behavior. However, the significant presence of paucimannosidic glycoforms in the MBL unbound fractions indicated low binding affinities or MBL saturation under the assayed condi-tions. High mannose-containing bovine ribonuclease B did not bind to MBL, whereas free high mannose N-glycans showed significant affinities (data not shown).

DISCUSSION
Augmenting established glycobiology (15)(16)(17)(18)(19)(20)(21), we here, for the first time, demonstrate that humans also produce bioactive paucimannosidic proteins similar to lower organisms such as nematodes, insects, and plants (15,16,26,28) (see Fig. 6 for overview of our findings). However, in contrast to lower organ-  isms where paucimannosylation is generated in the classical ER-Golgi secretory pathway and constitute a ubiquitous class of structures in the N-glycome repertoire, humans appear to use a non-classical ER-Golgi-granule pathway to produce paucimannosidic proteins. Irrespective of this diverted (organelle-specific) and time-demanding hydrolytic route, we argue that the abundance of paucimannosylation in infected sputum and the presentation of these structures on intact and functionally bioactive proteins support their classification as a fourth N-glycan type (in addition to high mannose, hybrid, and complex type) in neutrophil-rich environments central to inflammation and infection.
The discovery of this alternate type of human N-glycosylation, as exemplified in pathogen-infected sputum, was enabled by recent analytical developments in system-wide characterization of protein glycosylation (10,31). Protein paucimannosylation was found to be abundant in sputum from inflamed pathogen-infected lungs irrespective of lung disease/condition, infecting microorganism, gender, age, and antibiotic treatment. Pathogen-free sputum, although derived from neutrophil-rich lungs (39), displayed negligible amounts of protein paucimannosylation. This implies that human protein paucimannosylation is neither genotype-specific, microbe-specific, nor diseasespecific, but rather a general molecular feature common to inflamed micro-environments of hosts undergoing pathogenic attack.
Five human paucimannosidic N-glycans were found to be carried by 18 abundant human sputum proteins that localized specifically to the azurophilic granules of the multi-compartmentalized human neutrophil (40). Neutrophilia is well established in CF and other respiratory conditions including URTI, featuring high proportion (Ͼ95%), counts (Ͼ10 7 cells/g of sputum), and viability (Ͼ70%) of neutrophils in sputum (41). Strong support of paucimannosidic subcellular-specific localization in azurophilic granules and the proposed association between sputum protein paucimannosylation and neutrophil activation by pathogens comes from the observation that purified neutrophil proteins, including human MPO (42,43), proteinase 3 (PR3) (44), azurocidin (45), Hex B (46), and bovine ␣-mannosidase (47), which localize to azurophilic granules (34), were previously shown to carry monosaccharide compositions corresponding to paucimannosidic N-glycans. Our glycoproteomics data indicate that sputum glycoproteins localizing to the specific and gelatinase granules and secretory vesicles in human neutrophils carried preferentially complex and high mannose N-glycans, suggesting a compartment-specific production and storage of paucimannosidic proteins in azurophilic granules. Highly similar N-glycosylation profiles of pathogeninfected sputum and neutrophil proteins, including the paucimannosidic profiles, further supported the neutrophilic origin of paucimannosylation and are congruent with studies indicating mammalian paucimannosylation in cancer and systemic lupus erythematosus (5,7,8,12), which are neutrophil-rich pathologies. The neutrophil N-glycosylation profile in our study resembles a previously reported human neutrophil N-glycan profile (48) in which the low mass paucimannosidic structures however were not reported. The biosynthetically intriguing mono-antennary sialo-N-glycans (NeuAc 1 Gal 1 Man 3 -GlcNAc 3 Fuc 0 -1 ) were abundant in the neutrophil and CF sputum N-glycomes, but did not appear to be directly related to the azurophilic granule-specific paucimannosylation based on their attachment to proteins localizing to other neutrophil granules (supplemental Fig. 1).
The fact that paucimannosidic N-glycans are carried by highly solvent-accessible sites on sputum proteins suggests that they are derived from extensive exoglycosidase processing (37). The high solvent accessibilities also explain the prevalence (Ͼ85%) of the accessibility-dependent ␣1,6-(core) fucosylation on the paucimannosidic N-glycans. High core fucosylation, in turn, implies that the paucimannosidic biosynthetic route involves N-glycan intermediates displaying terminal ␤1,2-GlcNAcylation, a substrate requirement for fucosyltransferase 8 (49). This implies again that paucimannose generation follows the initial synthesis of cis-Golgi-localized fucosylated hybrid/complex glycan intermediates.
Hexosaminidases are highly expressed in paucimannose-rich organisms including C. elegans, D. melanogaster, and plants (21). We observed high sequence homology of the ␣ and ␤ subunits of the heterodimeric human Hex A to these hexosaminidases in line with a previous study (20). In addition, the specific identification of human Hex A ␣/␤ subunits in azurophilic granules of neutrophils (34,38) and our immunocyto- chemistry data showing partial co-localization of both human Hex A and paucimannosidic glycoepitopes with an azurophilic granule marker (MPO) in differentiated neutrophil-like HL-60 cells together support the Hex A-driven compartment-specific paucimannosylation pathway proposed herein. The capacity of Hex A to generate paucimannosidic N-glycans in vitro from biosynthetic intermediates at realistic physiological conditions, albeit at low enzymatic rates, also confirms this relationship. The similarities of primary (66.5% sequence similarity) (data not shown) and higher structural levels of ␣ and ␤ subunits of Hex A (50,51) suggest that the homodimeric Hex B (␤␤) and Hex S (␣␣) isoenzymes may also be able to catalyze paucimannosylation. The relative broad substrate specificity of hexosaminidases to both ␤-GlcNAc-terminating and ␣-GalNActerminating glycoconjugates (52), together with the low activity observed here, may evolutionarily be more beneficial than high enzyme activity considering the prolonged storage of Hex A and glycoprotein substrates in the azurophilic granules: from the compartment assembly early in the neutrophil maturation in the bone marrow (40) over blood circulation averaging 5 days (53), to transmigration and mobilization via degranulation mechanisms at the inflammatory site. The relative efficiency of Hex A-driven paucimannosylation of membrane, e.g. lysosome-associated membrane protein 2 (LAMP2) and soluble e.g. NE proteins and any co-factor requirements for optimal paucimannosidic protein generation as reported for GM2 ganglioside degradation by Hex A (54), await further investigation.
Assembly of the paucimannose generating azurophilic granules and its molecular components early during myeloid maturation in the bone marrow was supported by the temporal gene expression of paucimannosidic proteins and putative paucimannose biosynthetic enzymes in promyelocytes in excellent agreement with previous studies (55,56). Gene set enrichment analysis (57) revealed a high gene expression of paucimannosidic proteins in isolated blood leukocytes from S. pneumoniaeinfected (but not S. aureus-, Escherichia coli-, and influenza A virus-infected) individuals (9-fold enrichment, p ϭ 3.4 ϫ 10 Ϫ3 ) relative to healthy individuals, suggesting that the assembly of the paucimannose biosynthetic machinery in neutrophil precursors can be shifted from the bone marrow to the blood circulation via infection-dependent "left shifts." The "targeted-by-timing biosynthesis" hypothesis mechanistically explaining the formation of the granule-specific proteomes in neutrophils (34,40,58) is congruent with our observation of compartment-specific N-glycosylation; the majority of all glycoproteins trafficking through the N-glycosylation machinery at the promyelocytic stage of the neutrophil development are directed to the azurophilic granules by vesicles budding from the cis-Golgi without reaching the late N-glycan maturation stage, e.g. ␤-galactosylation and ␣-sialylation in the trans-Golgi network (59). As expected, proteins localizing to the specific (and other) granules, which are synthesized exclusively in the myelocyte and more mature stages of the neutrophil development by vesicles budding off from the trans-Golgi network (58,59), displayed complex type N-glycosylation. This subcellular-specific N-glycosylation is further supported by the absence of human ␣-sialidases and ␤-galactosidases in azurophilic granules (34,38), which suggest that Hex A does not function in concert with other outer-arm exoglycosidases and may explain why the sterically protected GlcNAc-terminating N-glycans are unacceptable substrates for human Hex A. This fascinating feature of compartment-specific N-glycosylation is not unique to neutrophils (60).
The further trimming of the paucimannosidic glycoforms by human ␣-mannosidases appears to be linkage-specific as shown by the specific generation of the ␣1,6-linked mannoseterminating M2F and M2 isomers (see supplemental Table 1). This agrees well with the exclusive presence of the ␣1,6-mannose isomer of M2F and M2 previously reported on neutrophilderived human proteins (45) and the preferential hydrolysis of ␣1,3-linked mannosides by human ␣-mannosidase (61). Human ␤-mannosidase may be responsible for yielding amannosylated di-(M0) or tri-(M0F) saccharides. The detection of ␣and ␤-mannosidases in isolated azurophilic granules (34,38) and the promyelocyte-specific expression of genes coding for the corresponding mannosidases support their association with protein paucimannosylation. By overlaying our observations on the existing map of the mammalian N-glycosylation machinery (3), we propose a new spatio-temporal restricted biosynthetic route enabling paucimannosylation of human neutrophil proteins (Fig. 4I).
Azurophilic granules of human neutrophils share some functional and molecular commonalities with the glycoprotein-degrading lysosome, yet the two compartments are separate entities as displayed by their unique proteomes as well as by the mobile characteristics of azurophilic granules as secretory components (59,62). In addition, bidirectional lysosomal glycoprotein degradation is facilitated by a suite of exoglycosidases and proteases, leaving the released and partially degraded N-glycans without the reducing-end ␤-GlcNAc and ␣1,6-fucose residues (61), contrary to the paucimannosidic N-glycans and indeed the N-glycoproteins identified in this study. Together this indicates that protein paucimannosylation in pathogen-infected sputum is of non-lysosomal origin. The absence of significant co-localization of paucimannosidic glycoepitopes and several established lysosomal and ER/Golgi markers in other paucimannose-positive human cells further supported a non-lysosomal/ER/Golgi-based paucimannose synthesis. 3 As such, we argue that protein paucimannosylation in the azurophilic compartment of the human neutrophil should not be perceived as a degradation product aimed to salvage monosaccharides and amino acids, but rather as a cellular mechanism to generate an arsenal of releasable biomolecules displaying unique glycoepitopes and activities.
Sorting of proteins to storage granules is not unique to neutrophils, but seems to occur by mechanisms common to all cells (59). Intrigued by the temporal and compartment-specific biosynthetic route for paucimannosylation in neutrophils proposed here, we are investigating whether paucimannosylation is unique to neutrophils or common across cell types. We have recently reported that cultured human colon and breast cancer epithelial cells produce paucimannosidic epitopes (6,60), albeit in lower quantities (typically less than 15-20% of the total N-glycome), supporting a possible oncofetal antigen potential of paucimannosylation (12) and the possible molecular and functional similarity of neutrophils and epithelial cells (58). However, the lack of azurophilic granules in epithelial cells implies that production, storage, and secretion of paucimannosidic proteins may be facilitated by other mechanisms in these systems.
Containing an ensemble of bioactive molecules, including antimicrobial peptides and proteases, the azurophilic granule is the microbicidal compartment of the neutrophil (62,63). Thus, the azurophil-specific localization of protein paucimannosylation in neutrophils is potentially of high biological significance. It has been established that azurophilic granules are mobilized as the last compartment upon phagocytosis (40,62,63), emptying their soluble content into the phagolysosome and the extracellular environment to combat invading pathogens (40). The virulence-specific release of paucimannosidic proteins into sputum upon P. aeruginosa stimulation indicates an infectiondependent mobilization of azurophilic granules, aspects we are currently investigating by bacterial genome sequencing and proteomics. The importance of granule mobilization for innate immunity is well illustrated in the Chediak-Higashi syndrome where immobile azurophilic granules reduce the host response to pathogens (64). The induced secretion of paucimannosidic proteins from neutrophils presented here indicates that, as reported (40,62), granules fuse not only with the phagolysosome, but also with the plasma membrane upon activation.
The strong neutrophilic association with the paucimannosidic proteins observed in sputum from pathogen-infected inflamed lungs prompted us to investigate possible functional aspects of human paucimannosylation, i.e. lectin-based recognition by the immune system via mannose-receptor interactions. Mannose-terminating glycoconjugates are infrequent in the extracellular environment of healthy human cells/tissues (60), but such determinants from internal membranes or granules may be exposed under specific cellular conditions, e.g. immature ER-resident glycoepitopes were shown to be exposed in apoptotic cells serving as "eat me" signals for macrophagebased clearance (65). Exposure of ␣and ␤-mannose determinants on solvent-accessible glycosylation sites such as those presented by paucimannosidic proteins may be a unique feature to enable molecular and cellular communication of activated neutrophils via mannose receptors in the micro-environment by mechanisms of active secretion (degranulation) of paucimannosidic proteins or by release upon cell death (39). The abundant paucimannosidic determinants found in infected sputum may also arise from the release of granular contents into neutrophil extracellular traps (NETs) via the activation and NETosis of polymorphonuclear cells (66). We have demonstrated binding of paucimannosidic proteins and N-glycans to MBL; however, the exact roles of paucimannosylation in the downstream complement activation clearly need further investigation. In addition, paucimannosidic binding to other lectins including the macrophage mannose receptors (CD206 and CD280) and dendritic C type lectins (e.g. CD209, CD299, and CD303) may provide avenues for the neutrophils to communicate with other immune cells (67). Opportunistic pathogens may also recognize exposed mannosidic determinants as an avenue for host adherence, e.g. E. coli fimbrial FimH adhesin shows high affinity for high mannose type as well as paucimannosidic N-glycans (68,69).
In addition, bacteriostatic and bactericidal effects of the paucimannose-rich neutrophilic proteins NE, azurocidin, and cathepsin G have previously been reported (70). We are currently investigating the functional role of the carbohydrate moieties on these human paucimannosidic proteins in the context of bacterial killing and growth inhibition.
Finally, we speculate that paucimannosylation may carry out modulatory roles in the generation and/or recognition of antineutrophil cytoplasmic autoantibodies by masking or presenting immunogenic epitopes on the anti-neutrophil cytoplasmic autoantibody-typic and paucimannose-rich LAMP2, PR3, MPO, and NE (71). As such, it becomes clear that paucimannosylation may be linked to multiple diverse functional roles in the micro-environments where paucimannosidic proteins appear to be enriched following neutrophil activation, i.e. in phagolysosomes (39,62), microvesicles/ectosomes (72) and neutrophil extracellular traps (73).
In conclusion, we document that human neutrophils produce, store, and selectively secrete bioactive paucimannosidic proteins into sputum of lungs undergoing pathogen-based inflammation and infection. We show that the azurophilic granules of neutrophils are the biosynthetic "venues" and that human hexosaminidases are the enzymatic "facilitators" of this fourth type of human protein N-glycosylation. In line with established neutrophil biology, we propose that paucimannosidic proteins are targeted to the azurophilic granule "by timing" rather than by selective sorting in the early bone marrow maturation of developing myeloid cells. The rather narrow temporal and spatial nature of paucimannosidic proteins in micro-environments surrounding inflammation, and in response to pathogen infection, suggests specialized biomolecular immune functions and may explain how protein paucimannosylation to date has remained under the radar in human glycobiology.