Insights into the salivary N-glycome of Lutzomyia longipalpis, vector of visceral leishmaniasis

During Leishmania transmission sand flies inoculate parasites and saliva into the skin of vertebrates. Saliva has anti-haemostatic and anti-inflammatory activities that evolved to facilitate bloodfeeding, but also modulate the host’s immune responses. Sand fly salivary proteins have been extensively studied, but the nature and biological roles of protein-linked glycans remain overlooked. Here, we characterised the profile of N-glycans from the salivary glycoproteins of Lutzomyia longipalpis, vector of visceral leishmaniasis in the Americas. In silico predictions suggest half of Lu. longipalpis salivary proteins may be N-glycosylated. SDS-PAGE coupled to LC–MS analysis of sand fly saliva, before and after enzymatic deglycosylation, revealed several candidate glycoproteins. To determine the diversity of N-glycan structures in sand fly saliva, enzymatically released sugars were fluorescently tagged and analysed by HPLC, combined with highly sensitive LC–MS/MS, MALDI-TOF–MS, and exoglycosidase treatments. We found that the N-glycan composition of Lu. longipalpis saliva mostly consists of oligomannose sugars, with Man5GlcNAc2 being the most abundant, and a few hybrid-type species. Interestingly, some glycans appear modified with a group of 144 Da, whose identity has yet to be confirmed. Our work presents the first detailed structural analysis of sand fly salivary glycans.


Identification of Lutzomyia longipalpis salivary glycoproteins.
To determine the degree of N-glycosylation, an in silico analysis was carried out on 42 salivary proteins previously reported in Lu. longipalpis 4,20 to predict protein N-glycosylation sites using the NetNGlyc server (https ://www.cbs.dtu.dk/servi ces/NetNG lyc/). This revealed 48% of the commonly known salivary proteins contain conventional N-glycosylation sites (Supplementary Table S1). However, it is important to note this list only includes proteins available on the NCBI database as studies published to date have focused on major secreted proteins, and no deep sequencing has been carried out for salivary glands of this sand fly species.
To accompany the in silico dataset, we carried out our own analysis of the sand fly salivary proteins (Supplementary Fig. S1). First, Lu. longipalpis salivary glands were dissected and individually pierced to release saliva. Subsequent Coomassie blue SDS-PAGE analysis showed several protein bands ranging from ~ 10 to 100 kDa (Fig. 1). To identify which proteins were glycosylated, samples were analysed before and after treatment with Peptide-N-Glycosidase F (PNGase F), which cleaves high-mannose, hybrid and complex N-linked glycans. Treatment with PNGase F resulted in molecular mass shifts and migration of several protein bands, consistent with the widespread removal of N-glycans from the salivary glycoproteins (Fig. 1). De-glycosylation was also confirmed by transferring proteins to PVDF membrane and blotting with Concanavalin A (ConA) lectin, which binds specifically to terminal mannose residues on glycoproteins 21 (Supplementary Fig. S2).
For LC-MS/MS based glycoprotein identification, the major deglycosylated protein bands ( Supplementary  Fig. S3) were excised from the gel and sent to the University of Dundee Fingerprints Proteomics Facility. From the resulting list of 191 identified proteins, we excluded those without recognizable glycosylation sequons (as determined by NetNGlyc), obtaining a final list of 43 potentially N-glycosylated protein candidates (Supplementary Table S2). Fourteen of these potential glycoproteins were also identified in our initial in silico analysis (Supplementary Table S1), including LJM11, LJM111 and LJL143, which have been proposed as potential vaccine components against Leishmania infection 4 . Using the InterProScan tool to identify conserved protein domains, family distributions ( Supplementary Fig. S4) show five of the candidates belonging to the actin family, while others like tubulin, 5′ nucleotidase, peptidase M17 and the major royal jelly protein (yellow protein) are represented www.nature.com/scientificreports/ by two proteins each. After Blast2GO analysis, the "molecular function breakdown" suggested that 44% of the candidate glycoproteins are involved in binding, including 'small molecule binding' and ' carbohydrate derivative binding' (Supplementary Fig. S4). We also used the DeepLoc server to predict protein subcellular localisation and solubility of the proteins identified in Table S2. The results suggest 85% of candidate glycoproteins are soluble,  and 10 proteins are both extracellular and soluble (Supplementary Table S2).
Salivary glycoproteins from Lu. longipalpis are mainly modified with mannosylated N-glycans. Next, we determined the N-glycome modifying the salivary proteins of Lu. longipalpis. The presence of mannosylated N-glycan structures on salivary glycoproteins was suggested by the results of a lectin blot using Concanavalin A, and to confirm these results, we next determined the N-glycome of salivary glycoproteins of Lu. longipalpis. The oligosaccharides were released by PNGase F followed by derivatization with procainamide 22 which allowed fluorescence detection following hydrophilic interaction liquid chromatography (HILIC) and provided increased signal intensity in MS and MS/MS analysis 22 . Overall, we identified 14 different structures (Table 1), elucidated from ten separate compositions due to the presence of isomeric glycans.
Most oligosaccharides are of the high mannose type (82% of the N-glycome), with the Man 5 GlcNAc 2 -Proc glycan with m/z [727.81] 2+ , being the most abundant species (21.16 min; GU 6.00, Fig. 2). In addition, few hybrid-type species (with a retention time of 15.12-17.24 min) were detected, containing either an α1-6 core fucose residue linked to the reducing GlcNAc or not fucosylated, or a single terminal LacNAc motif (Fig. 2).
All major glycan structures were characterised using positive ion MS (Fig. 3A) and MS/MS fragmentation spectra. An example of structural elucidation using MS/MS fragmentation spectrum is shown for the major glycan species Man 5 GlcNAc 2 -Proc, with m/z [727.82] 2+ (Fig. 3B) while the remaining are mainly represented by hybrid-type glycans, either a trimannosyl modified with a Fuc residue on the chitobiose core, or paucimannosidic structures containing an unknown modification of 144 Da (see below).
Although PNGase F is highly effective in cleaving N-linked glycans, its activity is blocked by the presence of core fucose residues with an α1-3 linkage found in non-mammalian glycans. Therefore, we also treated our samples with PNGase A, which cleaves all glycans between the innermost GlcNAc and the asparagine independent of core linkages 23 . No differences were observed in chromatograms yielded from both enzymes ( Supplementary  Fig. S5), indicating all core fucosylation is likely to be α1-6-linked.

MALDI-TOF-MS analysis reveals a series of sand fly salivary glycans with unidentified modifications of 144 Da.
A more detailed analysis of the saliva by MALDI-TOF MS of pyridylaminated glycans revealed not only the major oligomannosidic species, but also suggested the existence of a series of glycans containing an unidentified structure. This modification was mainly found in two isomeric glycans: one with an RP-HPLC retention time of 25.0 min and the other of 26.5 min ( Supplementary Fig. S6). The two isomers have a m/z 1,295.50, which corresponds to a pyridylaminated Man 4 GlcNAc 2 glycan carrying a modification of 144 Da. This was confirmed by treatment with Jack bean α-mannosidase, which resulted in a loss of 2 and 3 hexoses (Fig. 4) for each isomer, respectively. Interestingly, this modification seems to be located in different positions in the two structures, and in both cases this modification was lost after treatment with 48% aqueous hydrofluoric acid (aq.HF) (Fig. 4, and Table 2).
Susceptibility to aq.HF is a hallmark of phosphoester, galactofuranose and some fucose modifications, but none of these are obviously compatible with a 144 Da modification. Based on this data, a re-assessment of the data with the procainamide-labelled glycans also revealed a total of four structures carrying this modification (Peak 4, 5, 7 and 8, Table 1); however, due to the very low abundance of these glycans we were unable to determine their chemical nature. Additionally, the potential for anionic modifications of N-glycans was explored by both glycomic workflows, but limitations in spectral quality and sample amount prevented a definitive characterisation.
no O-linked glycans found in sand fly saliva. In silico predictions using the NetOGlyc 4.0 24 server suggest that 85% our 191 identified salivary proteins have putative O-glycosylation sites (Supplementary Table S3). Sand fly saliva was subjected to reductive β-elimination to release O-glycans from the de-N-glycosylated proteins. Separation using porous graphitized carbon chromatography coupled with negative ion mode ESI-MS did not detect any O-glycans in the sample ( Supplementary Fig. S8), either due to their absence, low abundance or low mass.

Discussion
Sand fly saliva has important implications both for the insect and the vertebrate host 4 . Lu. longipalpis salivary proteins and their biological roles have been well studied 4,20 ; however, the sugars that modify these proteins have not been characterised in detail. Most work on sand fly salivary glycans comes from in silico analyses [13][14][15]17,18,25 and lectin blotting. They were first reported by Volf et al 19 , who used lectins to detect mannosylated N-type glycans. Mejia et al 16 reported high mannose glycans in Lu. longipalpis saliva, with some potential hybrid-type structures (also based on lectin specificity). However, results from lectin-based methods should be interpreted with care as detection controls have not always been included in these studies, and results can be highly dependent on glycan abundance in samples and specific protocols. Our work is the first time that a mass spectrometry approach has been used to study the salivary N-linked glycans of Lu. longipalpis, providing detailed information about their structures and relative abundances. We found that sand fly salivary glycoproteins consist mainly of oligomannose glycans (ranging from the core Man 3 GlcNAc 2 to Man 9 GlcNAc 2 ), with some hybrid-type (e.g. fucosylated) structures. Additionally, this is the first report of a 144 Da (unknown) modification present in some salivary glycans. Our results provide new insights into how these structures could be recognised by vertebrate host cells.  10,26,27 . It is generally accepted that N-linked type glycoproteins in arthropods are mainly of the high-mannose or paucimannose type, and account for over 90% of glycan complexity in Drosophila 10,28 . One of the first indications of the capacity of insects to produce complex type N-glycans came from bee venom phospholipase A2, which contains the core α1,3-fucose (an IgE epitope allergenic to humans). Anionic and zwitterionic N-glycans with up to three antennae have more recently been found in a range of insects [29][30][31][32] . Furthermore, Vandenborre et al. 33  www.nature.com/scientificreports/ several economically important insects, and found glycoproteins to be involved in a broad range of biological processes such as cellular adhesion, homeostasis, communication and stress response. Some researchers have predicted the presence of mucins in the mouthparts of bloodfeeders 34,35 , proposing their possible role as lubricants to facilitate bloodmeals. Even though O-linked glycans have been widely documented in invertebrates, we were unable to detect these sugars in sand fly saliva after reductive β-elimination. This was surprising given that our bioinformatic analysis (NetOGlyc server) predicted the presence of putative O-glycosylation sites. The presence of O-linked glycans in Lu. longipalpis saliva has been suggested through peanut agglutinin and Vicia villosa lectin detection 16 ; however, it is worth noting that the experiment does not include positive controls or binding inhibition by competitive sugars, so non-specific binding cannot be ruled out. Interestingly, Lu. longipalpis midgut mucin-like glycoprotein has been described 36 Table 1    www.nature.com/scientificreports/ suggesting that these dipterans may not be able to O-glycosylate proteins in salivary tissues, or they are below the level of mass spectrometry detection. A surprising finding in this work were the 144 Da structures modifying some of the salivary glycans (i.e. Man 4 GlcNAc 3 , and two Man 4 GlcNAc 2 isomers). They were present in very low abundance (< 1%), were located on different mannose residues (as shown by jack bean α-mannosidase digestion), and appeared susceptible to aqueous HF. However, we have yet to confirm the identity and biological role of this modification. A literature search revealed that structures of a 144 Da mass have been found on glycans from other organisms, including bacteria, viruses and sea algae [40][41][42] , but were not further addressed by the authors. One possibility is that these correspond to an anhydrosugar, like 3,6-anhydrogalactose (of 144 Da mass) 43 . Interestingly, work on mosquitoes has shown that these insects are able to produce anionic glycans with sulphate and/or glucuronic modifications that can be tissue specific 29,44 . The glycans identified here carrying this rare 144 Da residue may be another example of such modifications and could play a role specific to their location in sand fly saliva.
Even though every effort was made during salivary gland dissections to obtain saliva with minimal tissue contamination, this cannot be completely avoided. Analysis with the DeepLoc server suggested that although most protein candidates are 'soluble' , only some are predicted to be 'extracellular' . Furthermore, some proteins without signal peptide can still be secreted through a non-classical or "unconventional" secretory pathway 47,48 . An alternative way of saliva extraction would be to induce salivation by chemical means like pilocarpine [49][50][51] ; however, this carries its own logistical difficulties considering the amount of saliva needed to detect glycans in such low abundances (even with the highly sensitive techniques we have used here). Another limitation of this work is the low protein profile resolution provided by 1D gel electrophoresis, where we may have missed weaker bands during our selection of proteins for sequencing. Higher protein concentrations and analysis through 2D gel electrophoresis could help us address this issue; nevertheless, we believe our work includes the major proteins in Lu. longipalpis saliva, providing a good overview of glycan abundance and composition in this bloodfeeding insect.
The biological role of protein glycosylation in the saliva of sand flies (and other bloodfeeding arthropods) is uncertain. One possibility is that glycans affect salivary protein half-life in the blood once they enter vertebrate host. Another possibility is that these glycans influence other in vivo processes like the interactions between saliva and cell surface carbohydrate recognition domains. For instance, the mannose receptor and DC-SIGN are c-type lectins that recognize mannosylated structures (uncommon in vertebrate cells); they are present on macrophages and dendritic cells, playing a role in both innate and adaptive immune systems 52 , making glycans highly relevant in parasitic infection processes. Additionally, the mannose-binding lectin activates the 'lectin pathway' of complement, and has an important role in protection against various pathogens 53 . An example of this was reported in tick saliva, which contains a mannose-binding lectin inhibitor whose activity was shown to be glycosylation-dependent 54 .
This, in turn, could be of importance within the context of Leishmania infection as both macrophages and dendritic cells have been shown to have critical roles in the initial stages of infection and subsequent dissemination of the parasite inside the vertebrate host 55 . In order for Leishmania to survive and multiply inside the host, it must be internalized by macrophages; however, promastigotes appear to avoid the MR receptor during invasion, as it promotes inflammation and can be detrimental to their survival 55 . The saliva of Lu. longipalpis can prevent macrophages from presenting Leishmania antigens to T cells 56 , but these effects are species-specific; in the case of other sand flies like Phlebotomus papatasi, saliva inhibits the activation of these cells 57 . Work on a patient-isolated L. major strain that causes nonhealing lesions in C57BL/6 mice found that its uptake by dermal-macrophages is MR-mediated 58 . Even though the MR does not play a role in the healing strain, it is an indication that sand fly saliva may be involved in other parasite-macrophage interactions. Leishmania also interacts with DC-SIGN (particularly amastigotes and metacyclic promastigotes) and this varies depending on species 59 . It remains to be seen whether mannosylated glycoproteins in saliva impair or facilitate these interactions and their outcomes.
Many sand fly salivary proteins are currently being explored as potential vaccine candidates against Leishmania, and knowing the nature of their post-translational modifications is relevant to their activity and efficacy. Several salivary proteins from Lu. longipalpis that are being researched as vaccine candidates (e.g. LJM11, LJM17 and LJL143 4 ) have potential glycosylation sites (as indicated in the results of our in silico analysis). As recombinant versions of these proteins are normally expressed in non-insect cells 60 , care should be taken to ensure the glycoprotein's profile and activity remains the same.
Finally, it is also worth considering the role salivary glycoproteins could play inside the sand flies themselves. Both male and female sand flies rely on plant sugars to survive, and Cavalcante et al. showed that Lu. longipalpis ingest saliva while sugar feeding 61 . Lectins (which bind to glycans) represent a major part of a plant's defence system 62 , and can cause damage to an insect's midgut when ingested 63 . Salivary glycoconjugates may be potentially recognized by these plant lectins, helping to decrease the damage they can cause. Moreover, the ingestion of saliva during the bloodmeal may impact parasite differentiation in the fly's gut 64 . Furthermore, sand fly-borne viruses use the host cell machinery for replication, which includes the insect glycosylation pathways, before it is transmitted to the vertebrate host. In this context, understanding the glycosylation of insect salivary glands is also relevant to understand their pathogenicity.  enzymatic release of N-linked glycans. The N-glycans from sand fly saliva were released by in-gel deglycosylation using PNGase F as described by Royle et al. 70 . For deglycosylation using PNGase A, peptides were released from gel pieces by overnight incubation at 37 °C with trypsin in 25 mM ammonium bicarbonate. The supernatant was dried, re-suspended in water and heated at 100 °C for 10 min to deactivate the trypsin. Samples were dried by vacuum centrifugation and the tryptic peptide mixture was incubated with PNGase A in 100 mM citrate/phosphate buffer (pH 5.0) for 16 h at 37°C 71 . Samples were separated from protein and salts using LudgerClean Protein Binding Plate (Ludger Ltd., Oxfordshire, UK). All wells were flushed with extra water to ensure full recovery and then dried by vacuum centrifugation prior to fluorescent labelling.

Fluorescent labelling and purification of released N-glycans.
Released N-glycans were fluorescently labelled via reductive amination reaction with procainamide using a Ludger Procainamide Glycan Labelling Kit containing 2-picoline borane (Ludger Ltd.). The released glycans were incubated with labelling reagents for 1 h at 65 °C. The procainamide labelled glycans were cleaned up using LudgerClean S Cartridges (Ludger Ltd) and eluted with water (1 mL). Samples were evaporated under high vacuum and re-suspended in water prior to use. MALDI-TOF analysis of aminopyridine-labelled glycans. Sand fly salivary glycans were released according to previous procedures and labelled with PA (aminopyridine) as described elsewhere 75 , prior to RP-HPLC and analysis by MALDI-TOF MS using a Bruker Daltonics Autoflex Speed instrument (Hykollari). Aliquots of samples were treated with Jack bean α-mannosidase (Sigma), α-1,3 mannosidase and 48% aqueous