HUVECs were stimulated with PMA, a commonly used secretagogue that induces exocytosis of endothelial vesicles
As previously reported (26), the morphology of ECs changes from spindle-shaped to round upon PMA activation, and the rod-shaped Weibel-Palade bodies, unique storage vesicles within ECs containing vWF and many other secreted proteins , fuse with the cell membrane (Fig. 1A)
In total, the secretomes of 17 primary ECs were analyzed via gel-LC-MS/MS, with or without deglycosylation
Apart from 123 secreted proteins , the conditioned medium of PMA-stimulated ECs was particularly rich in surface antigens and receptors , including many established endothelial markers (Table I)
All identified proteins and peptides are listed in supplemental Tables S1 and S2, respectively
The distribution of the frequencies and the cumulated distribution of the number of samples in which proteins were identified are shown in supplemental Fig
S1
MS datasets of three biological replicates have been deposited in PRIDE (accession numbers 26908–27003)
Immunoblots confirmed that proteins such as fibronectin and biglycan were constitutively secreted (Fig. 1B)
Others such as agrin and lymphatic vessel endothelial hyaluronic acid receptor 1 were released upon PMA stimulation, providing an explanation for why previously unidentified proteins (8, 10) were found in the present analysis (Fig. 1C)
An overlay between secreted (Cy3 and Cy 5; green and red color) and cellular (Cy 2; blue color) proteins is shown in Fig. 1D
Common spots were numbered (supplemental Fig
S2) and identified via LC-MS/MS (supplemental Table S3)
Certain proteins , such as von Willebrand antigen 2 (a propeptide of vWF , AA 23–763), were clearly more abundant in the secretome of PMA-treated HUVECs
Among the 1252 identified proteins were 253 extracellular or plasma membrane proteins (approximately 20%) related to cell adhesion, blood coagulation, hemostasis, signaling transduction, and protein transportation, of which 166 were known glycoproteins (Table I)
To further characterize this subproteome, we employed a glycoproteomics approach
Secreted proteins were precipitated and digested with trypsin, and tryptic peptides were labeled with TMT0 to increase their charge state prior to enrichment by means of zwitterionic hydrophilic interaction liquid chromatography purification (24)
For glycosite identification, an indirect and a direct strategy were pursued (Fig. 2A): (i) digestion with PNGase F in the presence of 18O water to label the conversion of asparagine to aspartic acid upon the removal of N-glycans, and (ii) alternating HCD and ETD (HCD-alt-ETD) or HCD-product-dependent ETD ( HCD-pd-ETD) fragmentation on an Orbitrap Elite MS (24)
There was little overlap in the numbers of glycopeptides (Fig. 2B) and glycosylation sites (Fig. 2C) identified via the direct ( HCD-ETD) and the indirect ( PNGase F + H218O) methods
Better agreement was observed at the protein level (Fig. 2D)
With the indirect ( PNGase F + H218O) method, 27 peptides were identified with N[+2.99] modification at non-consensus sequence , out of 1139 total identified peptides with N[+2.99]
This anomaly rate of 2.4% (27/1139) combines the rate of false identifications and the rate of chance deamidations in 18O water that were not in the consensus sequence of glycosylation (i.e. N-X(not P)-S/T)
Allglycopeptides identified are listed in Table II and supplemental Table S4
Three spectra (full MS, HCD , and ETD) from a neuronal cell adhesion molecule (UniProt accession number Q92823) (AA - 222FNHTQTIQQK231) are presented in Fig. 3
For the same samples, HCD-pd-ETD revealed 28 known, 25 potential, and 16 novel glycosylation sites based on 209 identified spectra; HCD-alt-ETD revealed 20 known, 32 potential, and 14 novel glycosylation sites from 110 identified spectra
The HCD-alt-ETD method selected mostly precursors with higher intensities, higher charge, and smaller m/z (Fig. 4A)
Several large glycopeptides were detected via only HCD-alt-ETD , and more low-abundant glycopeptides were detected via HCD-pd-ETD
There was limited overlap in the identified glycopeptides but better agreement in the protein level (Fig. 4B)
Among the 319 total glycopeptides identified in the conditioned media, 31 were attached with a trimannosyl core (-HexNAc2Hex3) or truncated core (-HexNAc2Hex) , 50 with high mannose (-HexNAc2Hex4–9) , and 238 with complex/hybrid glycans
Notably, HCD-pd-ETD detected almost twice as many complex/hybrid glycoforms as HCD-alt-ETD (Fig. 4C)
To validate the glycosylation status, we performed additional analysis before and after glycoprotein enrichment with affinity resins of ConA lectin (n = 4) using a Q Exactive MS (Thermo Scientific)
We then compared the number of identified spectra in the glycoprotein-enriched fraction, the flow-through, and the input (supplemental Table S5)
For most glycoproteins , a higher spectral count was observed in the glycoprotein-enriched fraction than in the original input and/or the flow-through
Representative examples ( fibronectin , neuronal cell adhesion molecule , tyrosine-protein-kinase-like 7 , and vWF ) are shown in Fig. 5A
Non-glycosylated proteins , such as annexin A2 and alpha-enolase , were more abundant in the flow-through
Glyco proteins identified in all three methods are highlighted in Fig. 5B
Section : Confirmation of Predicted Glycosylation Sites
Content :
The hemostatic protein vWF is the main protein stored within Weibel-Palade bodies (27)
After secretagogue stimulation, Weibel-Palade bodies undergo exocytosis, releasing vWF filaments
vWF is one of the few known proteins containing the ABO blood group signature, which is formed by different glycans
Although the released glycan com position of this protein has been investigated extensively (28, 29), experimental evidence for many putative glycosylation sites is still missing
The coverage obtained for vWF in our proteomics analysis is shown in Fig. 6A
The precursor protein consists of homologous units such as the VWF type A, C, and D domains and a C-terminal cystine know (CTCK)
The vWF propeptide (D1-D2, AA 23–763) is separated from the remaining domains of mature vWF (AA 764–2813) via furin-mediated proteolytic cleavage
We confirmed 6 N-glycosylation sites
Notably, three N-glycosylation sites were located within the propeptide (AA 23–763)
Comparison of direct and indirect glycopeptide detection using HCD-ETD and 18O-deamidation after PNGase F + H218O treatment, respectively: identified unique glycopeptides (B), unique glycosylation sites (C), and unique glycoproteins (D)
Section : Comparison of HCD-pd-ETD and HCD-alt-ETD
Content :
The two methods, HCD-pd-ETD (blue) and HCD-alt-ETD (red), displayed distinct distributions of the observed m/z, charge state, mass of identified peptides (M+H), and glycan mass, as well as the intensity of the precursor ions and the ByonicsTM score (all y-axes)
The x-axes represent index numbers after proteins were sorted by their corresponding y-axis value from lower to higher (A)
There was limited overlap in the identified glycopeptides (B)
C, the HCD-pd-ETD method preferentially identified complex/hybrid glycans
A, spectral count of input, glycoprotein-enriched fraction (GP), and flow-through fraction (FT) from representative glycoproteins and non-glycoproteins
B, complementarity of the different methods ( HCD-ETD, PNGase F + H218O treatment, and glycoprotein enrichment)
Only 18 glycoproteins were consistently identified
Coverage is highlighted in green, and potential glycosylation sites are shown in red
A large hexagon indicates a glycosylation site with a reference in the Uniprot database
By using the HCD-ETD (H) or PNGase F (P) method, we confirmed six N-glycosylation sites on vWF
B, ETD spectra of glycopeptides identified via HCD-ETD ( N156, N211, N666, N1574 )
The following abbreviations are used: a, y, g, k = TMT modified Ala, Tyr, Gly, and Lys , respectively; c = carboxyamidomethylation of Cys ; m = oxidation of Met
Section : This study represents a significant advance over the existing proteomics literature on ECs
Content :
Unlike other cell types, ECs do not tolerate prolonged serum starvation, and their susceptibility to cell death upon serum withdrawal poses a major challenge for proteomic workflows targeting their secretome
We performed secretome analysis after 45 min of PMA stimulation combined with enrichment strategies for glycoproteins and glycopeptides
Glycopeptides were analyzed via three complementary MS techniques: the detection of 18O asparagine deamidation after digestion with PNGase F in H218O, HCD-alt-ETD , and HCD-pd-ETD using an Orbitrap Elite MS
The secretagogue PMA minimized EC death by allowing a shorter incubation period under serum-free conditions while increasing coverage in the proteomic analysis by inducing the exocytosis of intracellular storage vesicles (14) such as Weibel-Palade bodies
These unique storage vesicles in ECs play a major role in hemostasis and cell-to-cell communication
Using this approach, many more proteins were identified than in any previous proteomics study on ECs, including known endothelial surface markers such as endoglin ( CD105 ), integrin beta-1 ( CD29 ), tyrosine-protein kinase receptor Tie-1, and junctional adhesion molecule A ; secreted growth factors (i.e. C-type lectin domain family 11 member A ); co-receptors (i.e. neuropilin-1 ( co-receptor for VEGF-A )); proteases(i.e. furin ); and inflammatory mediators (i.e. macrophage migration inhibitory factor ), to name just a few
Short-term PMA treatment does not release microparticles (30), as shedding events make it difficult to discern intracellular from secreted/membrane proteins
In a direct comparison of the cellular proteome and the secretome utilizing difference gel electrophoresis, 70 out of 96 proteins analyzed were present in both samples, representing <10% of the visible protein spots in the secretome
Glycosylation is key for the stability and solubility of secreted and membrane proteins
It is the most complex post-translational modification (31) and mediates extracellular matrix network assembly, cell–cell interactions, and cell–matrix interactions
Unlike polynucleotides and polypeptides , which have a linear structure, sugars tend to be arranged in branched polymers, resulting in an exponential increase of possible polysaccharide combinations
Theoretically, just six monosaccharides can give rise to 1012 different glycan structures
This high diversity of protein-bound glycans requires a combination of different techniques
For example, new MS-based methods were developed to profile the cell surface N-glycoproteome as a differentiation marker for stem cells (32)
We applied a combination of different glycoproteomics techniques to further enrich for secreted and shed membrane proteins and reveal potential glycosylation sites within the endothelial secretome
Glyco proteins play important roles in many biological processes related to ECs, such as angiogenesis, in which the structural change of the glycans will determine the attachment property of cells and influence cell-to-cell interactions (33)
Interestingly, vWF is a glycoprotein produced uniquely by ECs and megakaryocytes
Previous publications investigating vWF isolated from plasma failed to identify glycosylation sites within the propeptide (29)
In plasma, the concentration of the propeptide is about one-tenth of the concentration of mature vWF (34, 35)
In the conditioned medium of ECs, however, we observed several glycopeptides of the propeptide
Thus, the endothelial secretome allowed us to interrogate the glycosylation sites of von Willebrand antigen 2, the N-terminal cleavage product of vWF that aids N-terminal multimerization and protein compartmentalization of mature vWF in storage granules
Section : Conventional Methods for Glycoproteomics
Content :
As reviewed elsewhere (36), conventional glycoproteomic methods involve the enrichment of glycoproteins (typically with lectins like ConA and wheat germ agglutinin ), cleavage of the glycans, and identification of the remaining peptide sequence
The most widely used method for detecting N-glycopeptides is digestion by PNGase F . PNGase F cleaves the GlcNAc molecule closest to the peptide (37)
After PNGase F treatment, formerly N-linked glycosylated peptides are identified based on the conversion of Asn to Asp (deamidation) in the consensus motif for N-linked glycosylation (sequence N-X(not P)-S/T)
This method has two major caveats
The first of these is a high false positive rate due to spontaneous deamidation
Asn-Gly sites , in particular, are prone to spontaneous deamidation (38–40)
To reduce false positives, PNGase F treatment is performed in 18O water, adding a larger tag of 2.99 Da
Importantly, all known glycosyltransferases that mediate N-linked glycosylation are supposed to recognize a consensus motif , and this consensus sequence for N-linked glycosylation must be taken into consideration (41)
2) The second caveat is that after PNGase F cleavage, the released sugars can be analyzed separately, but the link to the identified peptides with deamidated amino acids is lost (42, 43)
Ideally, intact glycopeptides are analyzed directly via MS/MS even in complex biological samples
HCD fragmentation mostly breaks glycosidic bonds, whereas ETD preserves the glycan attachment and fragments the peptide backbone, providing more complete peptide sequence information
Current MS/MS acquisition strategies for glycopeptide analysis rely on the acquisition of MS/MS spectra for all precursor ions
In this study, HCD was employed to generate glycan oxonium ions and trigger an ETD spectrum in a data-dependent manner
HCD presents the sugar signatures within the low m/z range, which are otherwise lost as a result of the one-third rule of ion trap fragmentation (44)
Glycopeptides with terminal HexNAc generate typically an m/z 204.0864 oxonium ion and its fragments at m/z 168.0653 and 138.0550
The oxonium ion and its fragments are measured with the high mass accuracy of the Orbitrap analyzer, and the unambiguous identification of the glycan oxonium ion generated by the HCD scan serves as a diagnostic marker for glycopeptides
This approach was compared against conventional HCD-alt-ETD scans using a complex biological sample
The HCD-alt-ETD preferentially detects higher charged and higher intensity precursor ions than HCD-pd-ETD
This might be because (i) a higher charge increases ETD fragmentation efficiency, resulting in more identified glycopeptides ; (ii) high-charged precursors did not produce HCD spectra of sufficient quality to trigger ETD based on the diagnostic oxonium ions; or (iii) more abundant peptides were selected in HCD-alt-ETD because the instrument duty cycle is less efficient than in HCD-pd-ETD
Overall, the combination of multiple MS methods used in our study provides greater confidence in the identification of glycopeptides than studies relying on a single approach and offers complementary advantages in the assessment of the glycoproteome, notably, the simultaneous identification of the peptide sequence , the glycosylation site , and the glycan com position
N-linked and O-linked glycosylation are the two most common forms of glycosylation in mammals (45)
Only N-linked glycosylation was analyzed in the present study
Unlike N-linked glycosylation, O-linked glycosylation has no consensus site (46)
This makes the analysis of O-linked glycopeptides a more daunting task (47)
Lectins are widely used for glycoprotein enrichment
There are many types of lectins binding to different sugars, such as ConA (binds to α-d-mannosyl and α-d-glucosyl residues ) and wheat germ agglutinin (binds to GlcNAcβ1–4GlcNAcβ1–4GlcNAc- and N-acetylneuraminic acid)
Here we used only ConA as a proof of principle to demonstrate the complementary results of multiple glycoprotein identification methods
ConA is known to display nonspecific avidity for hydrophobic ligands such as certain domains of tropomyosin (48)
Furthermore, the standard protocol for the ConA glycoprotein enrichment kit is not optimized for cleanliness, and several known non-glycoproteins were also detected in the eluate samples
Sequential washes with low- and high-ionic-strength buffers before elution might have reduced this contamination (49)
Also, mixing different lectins would increase the coverage of the glycoproteome in biological samples (39)
Additional efforts are needed for a complete structural characterization of protein glycosylation; in particular, the quantitation of the occupancy rates and the identification of the glycan structure as complex/hybrid glycans cannot be discerned via our current MS approach