General N-and O-Linked Glycosylation of Lipoproteins in Mycoplasmas and Role of Exogenous Oligosaccharide

The lack of a cell wall, flagella, fimbria, and other extracellular appendages and the possession of only a single membrane render the mycoplasmas structurally simplistic and ideal model organisms for the study of glycoconjugates. Most species have genomes of about 800 kb and code for few proteins predicted to have a role in glycobiology. The murine pathogens Mycoplasma arthritidis and Mycoplasma pulmonis have only a single gene annotated as coding for a glycosyltransferase but synthesize glycolipid, polysaccharide and glycoproteins. Previously, it was shown that M. arthritidis glycosylated surface lipoproteins through O-linkage. In the current study, O-linked glycoproteins were similarly found in M. pulmonis and both species of mycoplasma were found to also possess N-linked glycans at residues of asparagine and glutamine. Protein glycosylation occurred at numerous sites on surface-exposed lipoproteins with no apparent amino acid sequence specificity. The lipoproteins of Mycoplasma pneumoniae also are glycosylated. Glycosylation was dependent on the glycosidic linkages from host oligosaccharides. As far as we are aware, N-linked glycoproteins have not been previously described in Gram-positive bacteria, the organisms to which the mycoplasmas are phylogenetically related. The findings indicate that the mycoplasma cell surface is heavily glycosylated with implications for the modulation of mycoplasma-host interactions.


Introduction
Glycoproteins are ubiquitous in eukaryotes with the number of glycosylated proteins thought to represent over half of the proteome [1].Once thought of as rare in prokaryotes, protein glycosylation has been shown in numerous bacterial species.The glycans can serve various roles including affects on protein folding and stability, ligand binding, signaling, and cell-cell interactions [2,3].O-linked glycosylation at serine and threonine residues has been found in many species of both Gram-negative and Gram-positive bacteria [2][3][4][5][6][7][8][9].N-linked glycosylation at asparagine residues has been described for many Gram-negative but not Gram-positive bacteria [7,8].
Mycoplasmas are valuable models for the study of membrane biology and the glycosylation of membrane proteins because of the lack of a cell wall and the possession of only a single lipid bilayer.Most species have less than 1,000 genes with few annotated as having a likely role in glycobiology.Mycoplasma pulmonis and Mycoplasma arthritidis are thought to lack the ability to synthesize nucleotide sugars such as UDP-glucose, but they nevertheless synthesize a variety of glycomoieties [10][11][12].The single annotated glycosyltransferase gene identified in these species has not been disrupted in extensive transposon libraries and hence is thought to be essential [13][14][15].O-linked glycosylation at serine and threonine residues of lipoproteins of M. arthritidis has been previously reported [11].The identified glycans were always a single molecule of hexose, either glucose or mannose.
We report here that the addition of disaccharides to the growth medium increased the abundance of glucose in macromolecules of M. pulmonis and the human pathogen Mycoplasma pneumoniae.Glycostaining and GC/MS indicated that supplementation with disaccharides increased the overall level of protein glycosylation.The lipoproteins of M. pulmonis were examined by high-resolution mass spectrometry (HR-MS) and found to have both O-linked and N-linked glycosylation.An unusual aspect of the findings is that not just asparagine residues, but also glutamine residues of surface-exposed proteins are glycosylated at numerous sites.The Nand Olinked glycosites had no apparent sequence specificity.In all reported cases the glycans were a hexose.Retrospective analysis of the lipoproteins of M. arthritidis resulted in similar findings of N-linked glycosylation in addition to the previously reported O-linked glycosylation.

Staining of SDS-PAGE gels
Bacteria harvested from cultures in stationary phase were washed twice with phosphate-buffered saline (PBS).Lipoproteins were fractionated using Triton X-114 (TX-114) as previously described [18,19].Extracted lipoproteins were suspended in 20 μl SDS loading buffer and heated to 100°C.The proteins were electrophoresed on 4-15% gradient SDS-PAGE gels (Bio-Rad).Glycoproteins were stained with the Pro-Q Emerald 300 Glycoprotein Gel and Blot Stain Kit (Invitrogen) following the manufacturer's protocol.Parallel gels were analyzed for glycoproteins or total protein stained with Coomassie blue.

GC/MS
Whole cell lysates were digested with nucleases, dialyzed, and analyzed by GC/MS as described [10].For analysis of lipoprotein fractions, TX-114 was removed from the samples with Pierce1 Detergent Removal Spin Columns.

HR-MS
Protein bands were excised and subjected to in-gel tryptic digestion.The samples were subjected to LC-HR-MS using the instrumentation and methods as described [11].The mass spectra were analyzed by using the PEAKS 7 proteomics software (Bioinformatics Solutions, Inc.).Each MS peak reported here was obtained from an obvious LC peak that spanned 4 to 5 scans.The PEAKS software manual suggests that peptide scores (-10lgP) of 20 and above are acceptable candidates for manual analysis.In this study, only spectra with scores of 50 or higher were analyzed.In the algorithm searches, asparagine, glutamine, serine, and threonine residues were searched for the presence of hexose (162.0528Da) in monomeric, dimeric and trimeric forms.A decoy dataset consisting of 8 non-lipoproteins was similarly analyzed.These proteins were MYPU_3210 (oligopeptidase F), MYPU_4280 (elongation factor G), MYPU_5110 (transketolase), MYPU_5180 (enolase), MYPU_7300 (aminopeptidase), MYPU_7610 (dihydrolipoamide dehydrogenase), MYPU_7620 (dihydrolipoamide acetyltransferase), and MYPU_7640 (pyruvate dehydrogenase E1 component).Using our search parameters and PEAKS peptide score criteria, no hexosylation was detected in this decoy data set.

Disaccharides support glycoconjugate synthesis
Using 13 C-labeled glucose and starch, it was previously shown that oligosaccharides but not monosaccharides support the synthesis of glycoconjugates in M. arthritidis, M. pneumoniae, M. pulmonis [11,12].No 13 C-labeled glycoconjugates were detected when these species of mycoplasma were grown in medium supplemented with 13 C monosaccharide glucose, but robust labeling was found when the medium was supplemented with hydrolyzed 13 C starch.Investigated here was the ability of the disaccharides maltose (α-D-glucopyranosyl-(1!4)-Dglucose) and lactose (β-D-galactopyranosyl-(1!4)-D-glucose) to support glycoconjugate synthesis in M. pulmonis and M. pneumoniae.
Lysates of mycoplasmas were digested with nucleases and extensively dialyzed to remove monosaccharides and other small molecules.The monosaccharide composition of the glycoconjugates remaining in the lysates were analyzed by GC/MS and compared to standards.The relative abundance of glucose associated with glycoconjugates significantly increased for both M. pulmonis and M. pneumoniae when the culture media was supplemented with maltose (P < 0.01 and P < 0.05, respectively) (Fig 1).Supplementation with lactose also resulted in a significant increase in the abundance of glucose in M. pulmonis glycoconjugates (P < 0.01).To control for the possibility that supplementation with disaccharides affected glycoconjugate synthesis indirectly as a result of enhanced energy metabolism because of an overabundance of glucose, the glycoconjugates were analyzed after growth of the mycoplasmas in medium supplemented with the monosaccharide glucose.Supplementation with glucose had no affect on the relative abundance of the sugars incorporated into glycoconjugates.Xylose was also present in all lysates but its abundance was unaffected by supplementation with glucose, lactose, or maltose.These results demonstrate that the mycoplasmas can utilize both alpha-and betalinked disaccharides for glycoconjugate synthesis.

Glycostaining of proteins
Proteins of M. pulmonis and M. pneumoniae were extracted with TX-114 to fractionate lipoproteins from non-lipoproteins.Proteins from the detergent phase, containing lipoproteins, were electrophoresed on SDS-PAGE gels that were stained with Pro-Q Emerald 300 to identify potential glycoproteins or Coomassie blue to identify total proteins.The banding pattern of the lipoprotein fraction was highly similar for the two stains (Fig 2A and 2B), suggesting that most of the lipoproteins were glycosylated as previously found for M. arthritidis [11].Glycostaining was increased for proteins from cultures supplemented with maltose or lactose, suggesting again that glycosynthesis in mycoplasmas is dependent on oligosaccharides.
Two doublets of proteins of M. pulmonis grown in medium supplemented with maltose were excised from the Coomassie-stained gels and analyzed by HR-MS.The major proteins were found to be MYPU_3200, MYPU_3230, and MYPU_3460.The MS coverage for these major proteins was in excess of 90%.Each of these proteins contains a signal peptide sequence indicative of a lipoprotein (see GenBank accession number NC_002771 for the predicted amino acid sequences, also available at http://services.cbib.u-bordeaux2.fr/molligen/).

Interpretation of HR-MS spectra
The mass of the centroid for each HR-MS peak was calculated as described by Gedcke [20].As an example, an expanded view of the peak at 952.52 Da in the left panel of  presented in this paper.The estimated mass as calculated from the centroid of the MS peak for all of the peptides agrees with the expected mass calculated from the amino acid sequence to a level of accuracy much smaller than the standard deviation.

O-linked glycosylation of M. pulmonis proteins
Previously, we identified O-linked hexosylation at serine and threonine residues of the lipoproteins MARTH_455, MARTH_403, MARTH_665, and MARTH_819 of M. arthritidis [11].In the current study, the lipoproteins identified above from M. pulmonis were examined by HR-MS and fragmented by collision-induced dissociation (CID).Multiple sites of hexose addition at serine and threonine were found for each lipoprotein (see Table 1).The expected mass shift corresponding to addition of a single residue of hexose is 162.0528Da, which is the mass of the monosaccharide (180 Da) minus the mass of the water molecule that is removed by formation of the glycosidic bond.The HR-MS spectra for the peptide GTKDFLPIELQSLEVSK containing Thr64 of MYPU_3230 is shown in Fig 3 .We identified both the hexosylated and non-glycosylated forms in multiple charge states (z = 2 and z = 3), showing the glycosylated   To further validate the results, the MS/MS-CID spectrum of the glycosylated peptide shown in

N-glycosylation at asparagine in M. pulmonis and M. arthritidis
In addition to O-glycosylation, strong evidence for N-linked hexosylation was found for the lipoproteins analyzed from M. pulmonis.The HR-MS spectrum for the peptide containing Asn335 from MYPU_3200 is shown in  We previously reported widespread O-glycosylation of lipoproteins of M. arthritidis [11] but not N-glycosylation.Using recently acquired PEAKS mass spectrometry software (Bioinformatics Solutions Inc., Waterloo, ON), mass spectra for the MARTH_403 lipoprotein were reanalyzed to search for N-glycosylation.As for M. pulmonis, numerous examples of hexosylation at asparagine were found.The HR-MS spectrum for the peptide ANAKNFYGFSDAYGK N-glycosylation at glutamine in M. pulmonis and M. arthritidis Although only rare instances of N-linkage to glutamine have been reported for any organism [22,23], examples of hexosylation at glutamine were found in both M. pulmonis and M. arthritidis.The HR-MS spectrum for the peptide ITDLLSQKEVTETQK is shown in Fig 7 .We identified both the hexosylated and non-glycosylated forms showing the glycosylated form has a mass shift consistent with the addition of hexose with an accuracy of 0.0012 Da for z = 2 and the addition of a hexose with an accuracy of 0.0003 Da for z = 2.A spectrum showing glycosylation at Gln49 for the peptide ITDLLSQKEVTETQK of MYPU_3460 is shown in

Discussion
Our findings show that M. arthritidis and M. pulmonis possess both general Nand O-protein glycosylation.The similar findings obtained for these two species suggest that general protein glycosylation may be widespread in mycoplasmas.As far as we are aware, the mycoplasmas are the first group of Gram-positive bacteria for which N-glycosylation has been described.The lack of amino acid sequence specificity at the glycosites of the mycoplasmal proteins is surprising.O-glycosylation in mycoplasmas lacked the three amino acid motif D(S/T)(A/I/L/M/T/V) found in Bacteroides [24], and N-glycosylation in mycoplasmas lacked the NX(ST) sequon found in most organisms [8,25].Glutamine glycosylation appears to be commonplace in mycoplasmas but is considered rare in other organisms [22].With glycosylation at asparagine, glutamine, serine and threonine residues and no obvious amino acid sequence specificity at the glycosites, the mycoplasmas have the most general protein glycosylation reported for any organism.No peptide was identified as having two glycosylated amino acids on the same molecule.
one hexosylation event shields the adjacent amino acids from further glycosylation.The proposed system would generate a high level of diversity with not all molecules glycosylated at the same sites and may represent a new type of antigenic variation that contributes to the chronicity of infection.
In addition to the possibility that a glycosylation event inhibits further glycosylation at neighboring glycosites, glycosylation might also inhibit cleavage by proteases at nearby sites.All of the glycopeptides listed in Table 1 have internal lysine residues that in theory could have been cleaved by the trypsin reaction that generated the glycopeptides.These lysine residues are frequently near the site of glycosylation.Similarly, most of the previously reported glycosites found in proteins of M. arthritidis were in the vicinity of a lysine residue [11].Perhaps the positively-charged amino acid is required for enzymatic recognition.In general, HR-MS analysis of the smaller peptides that were generated by cleavage at these internal residues failed to identify glycosylation events.It is probable that proteins that were non-glycosylated were cleaved by trypsin more efficiently than the glycosylated proteins because it has been shown previously that glycosylation of proteins with single residues of glucose inhibits cleavage by trypsin [21].Mycoplasmas are scavengers that do not synthesize amino acids de novo and acquire them from the host through the import of oligopeptides that are generated by the activity of proteases secreted by the mycoplasma [26,27].The general protein glycosylation system described here could have a critical role in protecting the mycoplasmal cell surface from these proteases as has been proposed for other bacterial glycosylation systems [28].
The glycosites reported here represent a small subset of the sites for which we have strong evidence for glycosylation of the lipoproteins of M. arthritidis and M. pulmonis.Dozens of additional glycosites could be identified by further analysis of the mass spectra.The sites reported are limited to the abundant lipoproteins that were examined and those sites that were glycosylated at sufficient frequency to yield a high number of MS/MS scans that clearly showed the modification.Potential minor sites for glycosylation were ignored in the current study.
Oligosaccharides, not monosaccharides, support glycoconjugate synthesis in M. pulmonis and M. arthritidis [12].Mycoplasmas are dependent on a host to provide many essential nutrients such as cholesterol, amino acids, and nucleotides.The only source of oligosaccharides available to the mycoplasma would also be the animal host.In this study we found that oligosaccharides containing alpha-and beta-1,4 linkages supported protein glycosylation in M. pulmonis and M. pneumoniae.The full range of oligosaccharide substrates that could be used by the mycoplasmas may be broad because of the complexity of the glycoconjugates encountered in the host.Current thinking is that the energy stored in the glycosidic bond of the oligosaccharides drives glycoconjugate synthesis, abrogating the need of nucleotide sugars.The oligosaccharides supplied in the culture medium or by the animal host are used to build the glycoconjugates that the mycoplasma synthesizes.The glycocalyx of the mycoplasmal cell surface might resemble closely that of the host milieu, perhaps contributing to immune avoidance.This simple strategy of decorating surface proteins with glycans obtained from host oligosaccharides is economical, as the mycoplasmas do not need to synthesize the glycans themselves or expend energy to produce nucleotide sugar substrates.This survival strategy is utilized in the human intestinal commensal, Bacteroides fragilis, which decorates surface-exposed proteins and polysaccharides with host-derived fucose [29].
Lactose is not imported by M. pulmonis [13] and yet its addition to the medium increased the level of protein glycosylation.Furthermore, there is no evidence for hexosylation of cytoplasmic proteins in mycoplasmas and hexosylation occurs at many sites in a fashion that appears to be stochastic.Hence, we propose that glycosylation of the lipoproteins occurs extracellularly.Extracellular glycosylation in mycoplasmas using host oligosaccharides as substrates would resemble oligosaccharide synthesis in some oral bacteria that use sucrose and glucans as substrates [30].The enzymes that are responsible for synthesis of these glycans are glycosidases/glycosyltransferases that transfer the sugar residue from an oligosaccharide to the bacterial substrate.In the M. pulmonis proteome there are five predicted glycosidases (MYPU_1030, 4630, 6330, 6340, 6440) that could potentially glycosylate the lipoproteins.As the membranes of M. pulmonis cells are often in intimate contact with host cell membrane [31], the possibility exists that not just mycoplasmal proteins but also host proteins are targets of the glycosyltransferase(s) produced by mycoplasmas.
Fig 3 is shown in S1 Fig.The digitized data were stored in 8 bins with the abundance of the events in a bin and the mass of the bin as indicated in the middle and right panels of the figure.The data are stored essentially as a histogram with the mass of the centroid calculated from the equation show in S1 Fig. Assuming a Gaussian distribution, the calculated mass at the centroid has a standard deviation equal to the full width of the peak at half maximum height divided by 2.35.The height and width of the peak are uncertain because of the information lost when the data were digitized, but we estimate the standard deviation to be about 0.008 Da for each of the spectra

Fig 1 .
Fig 1.The effect of substrate on glycoconjugate synthesis by M. pulmonis and M. pneumoniae.Panels A and B show the relative abundance of glucose or xylose, respectively, linked to protein as determined by GC.The values represent averages of the areas under the curves from 3 replicates of gas chromatograms that were converted to μg of sugar per mg of protein.The colors of the bars representing the different substrates used to supplement the medium are shown on the right.Plus or minus standard error bars are shown and the asterisks indicate a significant difference between a sample and control.doi:10.1371/journal.pone.0143362.g001

Fig 2 .
Fig 2. SDS PAGE of TX-114 fractions of M. pulmonis and M. pneumoniae.Panel B shows a glycostained gel, and panel A is a Coomassie-stained gel run in parallel.M. pulmonis bands excised for further analysis by HR-MS are indicated by arrows with the designation of the genes coding for the identified proteins.The unlabeled top band was analyzed by HR-MS and while there were several spectra that suggested glycosylation, none met our criteria for publication.doi:10.1371/journal.pone.0143362.g002

Fig 3 .
Fig 3. Hexosylation of the peptide GTKDFLPIELQSLEVSK of MYPU_3230.Orbitrap MS showing the doubly and triply charged ions.The 81.0262 shift for z = 2 between the non-glycosylated and glycosylated peptides equates to a mass shift of 162.0524Da, which corresponds to the addition of hexose (162.0528Da) with a mass accuracy of 0.0004 Da.The 54.0169 shift for z = 3 between non-glycosylated and glycosylated forms equates to a mass shift of 162.0507Da, which corresponds to hexosylation with a mass accuracy of 0.0021 Da.Monoisotopic values for the calculated theoretical and experimental masses of the peptide are given in bold.The images presented were obtained from an LC peak of MS scans and are expanded to show the charge states of each form.doi:10.1371/journal.pone.0143362.g003 Fig 4 was compared to that of the non-glycosylated peptide (see S3 Fig).The highly similar fragmentation patterns confirm that the spectra are from the same peptide sequence.The y 6 and b 6 ions have essentially the same mass in the non-glycosylated spectrum but differ in the glycosylated spectrum by a 162-Da.shift in b 6 .All of the b ions greater than b 6 were shifted by 162 Da, indicating hexosylation at Thr64.The O-linked glycosylation sites identified in MYPU_3200 were Ser515 and Ser519.The MS/MS spectra for these sites are shown in S4 and S5 Figs.The MS scans are shown in S6 and S7 Figs.

Fig 5 .
We identified both the hexosylated and non-glycosylated forms showing the glycosylated form has a mass shift consistent with the addition of hexose with an accuracy of 0.0004 Da for z = 3 and the hexosylated form z = 2 for the peptide STLEYTINNSQELQNILKQTYEEFTK.A MS/MS spectrum showing N-glycosylation at Asn335 of MYPU_3200 is shown in Fig 6.The b 13 and b 14 ions clearly delineate the Gln334 residue.The b 18 ion to b 13 ion separation gives a clear view of the N+162ILK portion of the sequence.The y 9 and y 10 ions as assigned delineate the Leu337 residue.The y 12 ion shows a

Fig 4 .
Fig 4. LC MS/MS-CID showing hexosylation at Thr64 of the peptide Gt 64 KDFLPIELQSLEVSK of MYPU_3230.The assigned b and y ions are shown in blue and red, respectively.Glycosylation of Q and S glycosites is absent in this spectrum as illustrated.The PEAKS peptide score (-10lgP) for this spectrum was 87.The charge state of the parental ion was z = 3. doi:10.1371/journal.pone.0143362.g004

Fig 5 .
Fig 5. Hexosylation of the peptide STLEYTINNSQELQNILKQTYEEFTK of MYPU_3200.Orbitrap MS showing the doubly and triply charged ions.The monoisotopic mass of the doubly charged species at 1648.8077 is consistent with the hexosylated peptide at z = 2 with a mass accuracy of 0.0012 Da.The 54.0175 shift for z = 3 between non-glycosylated and hexose forms equates to a mass shift of 162.0525Da with a mass accuracy of 0.0003 Da.Monoisotopic values for the calculated theoretical and experimental masses of the peptide are given in bold.The images presented were obtained from an LC peak of MS scans and are expanded to show the charge states of each form.doi:10.1371/journal.pone.0143362.g005 Fig 8.The assigned peaks for b 5 (556.3),b 6 -NH3 (626.0),b 7 -H 2 O (915.4), and b 8 (1061.6)provide a clear view of the residues Ser48, Gln49+162.05,and Lys50.The assigned peaks at y 7 (834.3),y 8 (962.3) and y 10 (1339.5)provide delineation in the other direction.The assigned peaks for this spectrum are shown in S4 Table.The MS/MS-CID spectrum for the non-glycosylated peptide ITDLLSQK is shown in panel A of S12 Fig, which is a tryptic truncation of the peptide shown in Fig 8.The glycosylated peptide ITDLLSQKEVTETQK showing hexosylation at Gln49 is shown in panel B for comparison.As for S8 Fig, nonapplicable y ions and extraneous peaks have been removed for clarity and the unaltered spectra are available in Fig 8 and S13 Fig.The similarities between the spectra are obvious.The major difference between the spectra is the increase in the b 2 ion in panel A and the shift at the b 7 ion by 162 Da., indicative of glycosylation at Gln49.

Fig 6 .
Fig 6.LC MS/MS-CID showing hexosylation at Asn335 of the peptide STLEYTINNSQELQn 335 ILKQTYEEFTK of MYPU_3200.The assigned b and y ions are shown in blue and red, respectively.Glycosylation of N, Q, T, S, and Y glycosites is absent in this spectrum as illustrated.The PEAKS peptide score (-10lgP) for this spectrum was 60.The charge state of the parental ion was z = 3. doi:10.1371/journal.pone.0143362.g006

Fig 7 .
Fig 7. Hexosylation of the peptide ITDLLSQKEVTETQK of MYPU_3460.Orbitrap MS showing the doubly and triply charged ions.The 81.027 shift for z = 2 between the non-glycosylated and glycosylated peptides equates to a mass shift of 162.054Da, which corresponds to the addition of hexose (162.0528Da) with a mass accuracy of 0.0012 Da.The 54.0177 shift for z = 3 between non-glycosylated and glycosylated forms equates to a mass shift of 162.0531Da, which corresponds to hexosylation with a mass accuracy of 0.0003 Da.The calculated theoretical and experimental values for m/z are given in bold.The images presented were obtained from an LC peak of MS scans and are expanded to show the charge states of each form.doi:10.1371/journal.pone.0143362.g007

Fig 8 .
Fig 8. LC MS/MS-CID showing hexosylation at Gln49 of the peptide ITDLLSq 49 KEVTETQK of MYPU_3460.The assigned b and y ions are shown in blue and red, respectively.Glycosylation of Q, S, and T glycosites is absent in this spectrum as illustrated.The PEAKS peptide score (-10lgP) for this spectrum was 74.The charge state of the parental ion was z = 3. doi:10.1371/journal.pone.0143362.g008

Table 1 .
Glycosites identified in this study.