Post-translational Modification of Thrombospondin Type-1 Repeats in ADAMTS-like 1/Punctin-1 by C-Mannosylation of Tryptophan^*

Expression Plasmids, Site-directed Mutagenesis, and Cell Culture

Mammalian expression plasmids for wild-type human punctin-1 or the N-glycosylation-defective punctin-1 mutant N251Q (punctin-NQ), each having a C-terminal tandem Myc and His₆ tag, were described previously (16). Site-directed mutagenesis (QuikChange site-directed mutagenesis kit, Stratagene, La Jolla, CA) was used to substitute Trp residues of interest within likely motifs (Trp³⁶, Trp³⁹, Trp⁴², Trp³⁸⁵, and Trp⁴⁴⁵) with Ala or Phe. The primers used for mutagenesis are available on request. The CHO cell mutants Lec15.2 or Lec35.1, which lack Dol-P-Man synthase activity or the ability to utilize Dol-P-Man, respectively (23, 29), were kindly provided by Dr. Mark Lehrman, University of Texas Southwestern Medical Center. These mutants, and CHO cells, were routinely grown on tissue culture plastic in Ham's F-12 medium supplemented with 15 mm HEPES, pH 7.2, 2% fetal bovine serum, 8% calf serum, and antibiotics. Transient transfections of HEK293 cells, CHO-K1 cells (both from ATCC, Manassas, VA), Lec15.2, and Lec35.1 cells were done using FuGENE 6 (Roche Diagnostics) as per the manufacturer's directions.

Metabolic Labeling, Autoradiography, and Immunoblotting

CHO-K1 or Lec35.1 cells transiently transfected with punctin-NQ were cultured in serum-free, low glucose (0.5 mm) Dulbecco's modified Eagle's medium supplemented with 0.3 mm l-proline and 30 mCi/ml d-[2,6-³H]mannose (23). After biosynthetic labeling for 18 h, the conditioned medium was collected, and punctin-NQ was affinity-purified using Ni²⁺-agarose as described previously (4). 80% of the sample was used for reducing SDS-PAGE followed by gel autoradiography as described previously (16) The remaining 20% was used for reducing SDS-PAGE and Western blotting with anti-Myc antibody to ensure that punctin-1 was indeed affinity-isolated. For quantitative comparison of cellular protein content and secreted levels of punctin-NQ and Trp mutants, HEK293F or CHO cells were co-transfected with the appropriate punctin-1 plasmid and the plasmid HIgG-pRK5 for expression of the Fc portion of human IgG (16). Following Western blotting with anti-Myc polyclonal antibody (for punctin detection) and anti-IgG, the levels of punctin-1 and IgG were determined by densitometry. Punctin-1 levels normalized with respect to IgG were used for comparative analysis as described previously (16). Cellular levels of punctin-1 were similarly normalized to intracellular glyceraldehyde-3-phosphate dehydrogenase.

Recombinant Protein and Mass Spectrometry

Purification of His-tagged wild-type punctin-1 using Ni²⁺-agarose was described previously (4). Recombinant ADAMTS5 produced in CHO cells and encompassing residues Ser²⁶²–Glu⁷⁵³ was a kind gift from Dr. Elisabeth Morris at Wyeth Pharmaceuticals. Purified proteins were subjected to in solution digest, adapted from the method of Stone and Williams (30). Briefly, ∼1 μg of protein was precipitated with 4 volumes of acetone overnight at −20 °C. Air-dried proteins were suspended in 10 μl of 8 m urea, 10 mm tris(2-carboxyethyl)phosphine, and 0.4 m di-ammonium phosphate, pH 8.0, vortexed, and heated to 50 °C for 5 min. Samples were alkylated by adding 100 mm iodoacetamide, 50 mm Tris-HCl, pH 8.0, vortexed, and incubated in the dark for 30 min. The urea was diluted to 2 m, and samples were incubated overnight at 37 °C with 150 ng of trypsin. Digested samples were acidified (with 7 μl of 5% formic acid), and particulates were removed using a 0.22-μm spin filter.

One- to 5-μl aliquots were analyzed by liquid chromatography/tandem mass spectrometry using an Agilent 6340 ion-trap mass spectrometer coupled to a nano-flow HPLC-CHIP system. Samples were separated using an Agilent (Zorbax 300SB) nano-CHIP C18 column at a flow rate of 450 nl/min with a 25-min linear gradient from 5 to 95% acetonitrile in 0.1% formic acid. Effluent from the CHIP was sprayed directly into the mass spectrometer. The drying gas (nitrogen) flow rate was 5 liters/min with a drying gas temperature of 325 °C. The capillary voltage was maintained at 1800 V. Full MS scans (m/z) 300-2200 were performed, and the three most abundant ions in each spectrum were selected for collision-induced dissociation (MS/MS).

O-Fucosylation

Peptides with O-fucose glycans were identified by neutral loss searches for the loss of glucose (hexose, 162 Da), fucose (deoxyhexose, 146 Da), or the sequential loss of glucose-fucose (308 Da) as described previously (16). Peptides identified as glycosylated by neutral loss were manually selected in subsequent runs for MS/MS/MS, in which the most abundant ion from MS/MS (usually the unglycosylated parent ion) is fragmented again, yielding high intensity b and y ions to match the predicted fragmentation of unglycosylated parent peptides. The mass of unglycosylated peptides was matched to predicted tryptic peptides which contain the consensus sequence C¹X_2–3(S/T)C²XXG.

C-Mannosylation

Peptides that differed in mass by 162 Da (for each mannose) from predicted tryptic fragments that contained a WXXW (WXXWXXW) motif, with or without the consensus sequence for O-fucosylation, were subjected to MS/MS and/or MS/MS/MS. C-Mannosylation of Trp is highly stable, so modification was confirmed by both identification of b and y ions where Trp residues may have the additional 162 Da, but also by loss of 120 Da in MS/MS spectra, a characteristic cross-ring fragmentation product of aromatic C-glycosides, as well as loss of water molecules (31).

Incorporation of Mannose during Biosynthesis of Punctin-1

On examination of the sequences of the four TSRs present in the short form of punctin-1, we noticed that TSR1 contained the classic consensus sequence WXXW for C-mannosylation in tandem (Fig. 1 A). TSR3 and TSR4 each contain a Trp residue with a Cys residue at the +3 position (Fig. 1 A). TSR2 does not contain either the WXXW consensus sequence for C-mannosylation nor the WXXC variant motif present in TSR3 and TSR4. Analysis of human punctin-1 (GenBankTM accession number AF176313) at the NetCGLyc 1.0 server predicted that Trp³⁹ and Trp⁴² in TSR1 and Trp⁴⁴⁵ in TSR4 were likely to be modified (see Table 1). Surprisingly, in contrast to the predicted modification of ⁴⁴⁵WSPC, modification of the ³⁸⁵WTAC motif in TSR3 was not predicted. TSR1 contains tandem consensus C-mannosylation sequences, i.e. Trp³⁶-Asp-Trp³⁹-Gly-Pro-Trp⁴²-Ser-Glu-Cys, such that Trp³⁶, Trp³⁹, or Trp⁴² could each serve as the target residue for C-mannosylation, and Trp³⁹ or Trp⁴² could be the +3 residue for modification of Trp³⁶ and Trp³⁹, respectively.

N251Q punctin-1 (punctin-NQ) was used for biosynthetic analysis because this mutation eliminates the possibility of incorporation of mannose into the single N-linked oligosaccharide present in punctin-1 (4) (Fig. 1 A), and thus reports exclusively the incorporation of radiolabeled mannose at other sites. This N-glycan-deficient mutant is secreted at a lower level than wild-type punctin-1 but is nevertheless stable and is readily detected by Western blot in conditioned medium of transfected cells using anti-Myc monoclonal antibody (16). Affinity purification of punctin-NQ was done using the medium of metabolically labeled control CHO-K1 or mutant CHO cells that cannot utilize Dol-P-Man (Lec35.1) (29). This demonstrated incorporation of d-[2,6-³H]mannose in punctin-NQ secreted from CHO-K1 cells but not from Lec35.1 cells, despite successful isolation of punctin-NQ from the medium of both cell types (Fig. 1 B). These data strongly suggested that mannose is incorporated into punctin-1 at one or more sites independent of N-glycosylation.

Identification of the Mannose Attachment Sites and Linkages on Punctin-1 and ADAMTS5 Using MS

For analysis by mass spectrometry, punctin-1 was purified by nickel-Sepharose chromatography, digested with trypsin, and subjected to liquid chromatography/tandem mass spectrometry. Analysis of tryptic peptides from recombinant human punctin-1 by MS identified an ion with m/z = 892.5 (Fig. 2 A, MS). This ion was consistent with the mass of the triply charged form of the peptide ²⁹EEDRDGLWDAWGPWSECSR⁴⁷ derived from TSR1 and modified by two mannose residues. MS² fragmentation of this peptide showed characteristic cross-ring cleavage of C-mannose, resulting in sequential losses of 120 Da (40 Da for a triply charged peptide), supporting the presence of two C-mannose residues on this peptide (Fig. 2 A, MS/MS). Multiple losses of water molecules (indicated by *) are also consistent with the presence of C-mannose (31). In addition, a series of y ions clearly reveals the presence of C-mannose on Trp³⁹ and Trp⁴². In contrast, the b10 ion demonstrates Trp³⁶ is unmodified. These data support C-mannosylation of Trp³⁹ and Trp⁴² but not Trp³⁶.

Another peptide containing these C-mannosylated Trp residues resulting from incomplete trypsin digestion was also detected (Fig. 2 B, MS). This peptide (m/z = 1008.2) also contains the predicted O-fucosylation site in TSR1. MS/MS fragmentation of this peptide showed sequential loss of glucose (m/z 967.5) and fucose (m/z 931.0) (Fig. 2 B, MS²), confirming the presence of the glucose-fucose disaccharide. A major ion characteristic of the cross-ring cleavage of a C-mannose on Trp was also observed (−30, MS/MS, Fig. 2 B). MS³ fragmentation of the m/z 931.0 ion provided sufficient sequence information to confirm the C-mannose modifications on Trp³⁹ and Trp⁴², as shown above (MS/MS/MS, Fig. 2 B), although the fragmentation pattern is more complicated because of the larger size of the parent ion. The same peptide modified with a single mannose on Trp³⁹ and the glucose-fucose disaccharide was also identified (Table 2 and Fig. 2 C), but no forms of this peptide lacking either mannose or O-fucose were found, suggesting that both modifications on this peptide occur at high levels of stoichiometry.

Identification of C-Mannosylation and O-Fucosylation in ADAMTS5

Analysis of the entire ADAMTS superfamily using the NetCGlyc 1.0 server predicted that each ADAMTS protease and each ADAMTS-like protein was likely to contain at least one modified Trp (Table 1). The majority of predicted sites were in TSR1 of the respective family members (Table 1). Accordingly, to extend the analysis of punctin-1 to additional members, we undertook MS analysis of recombinant ADAMTS5 as for punctin-1. We specifically asked whether like punctin-1, ADAMTS5 also contained C-mannosylation and O-fucosylation in TSR1. These experiments identified a peptide from TSR1 of ADAMTS5 modified with two C-mannoses and the glucose-fucose disaccharide (Table 2 and Fig. 3). Although the primary structure of ADAMTS5 predicts two TSRs (Table 1), recombinant ADAMTS5 frequently undergoes C-terminal processing that results in loss of TSR2, which was not present in the recombinant preparation used for this analysis.

Effect of Lack of C-Mannosylation on Punctin-1 Secretion

We investigated the consequence of inhibiting C-mannosylation of punctin-NQ using CHO-K1 variants, Lec15.2 and Lec35.1, and compared the levels of protein in the conditioned medium to that in parental CHO cells. Quantitation of secreted punctin-NQ was done in relation to the level of secreted co-transfected IgG. The punctin-NQ levels so normalized showed a statistically significant reduction of levels in medium from Lec35.1 cells but not in medium from Lec15.2 cells (Fig. 4 A). Quantitative comparison also suggested accumulation of punctin-NQ intracellularly in Lec15.2 and Lec35.1 cells, but this was statistically significant only in Lec35.1 cells (Fig. 4 B). Because we had previously shown that O-fucosylation led to decreased secretion, and because the O-fucosylation consensus sequences were intact in the expressed punctin-NQ, we asked whether supplementation with exogenous l-fucose (to ensure no deficiencies in GDP-fucose levels exist in these cells) would modify the secretion defect in Lec15.2 and Lec35.1 cells. However, the relative levels of punctin-NQ in the medium of CHO, Lec15.2, and Lec35.1 cells were unaffected by l-fucose supplementation (Fig. 4 C).

Site-directed Mutagenesis Suggests a Possible Dual Role for Trp Residues in Punctin-1

In a previously reported analysis of O-fucosylation of punctin-1, mutagenesis of the modified Ser or Thr residues demonstrated that O-fucosylation was essential for secretion (16). A similar approach for analysis of C-mannosylation required consideration of the three-dimensional structure of TSR2 and TSR3 from thrombospondin-1, which had suggested that Trp residues at the N terminus of TSRs have a structural role through stacking with arginine residues present in the anti-parallel β-strand (32). The mutagenesis strategy used was therefore designed to minimize the potential structural consequences of mutating the Trp residues in the TSRs. In initial analysis, we replaced Trp³⁶, Trp³⁹, Trp⁴², Trp³⁸⁵, and Trp⁴⁴⁵ in punctin-NQ individually with Ala. Each is the target Trp residue within a possible (Trp³⁶, Trp³⁸⁵, and Trp⁴⁴⁵) or experimentally determined (Trp³⁹ and Trp⁴²) C-mannosylation consensus sequence. In our previous analysis of punctin-1 O-fucosylation, we demonstrated that the peptides containing Trp³⁸⁵ and Trp⁴⁴⁵ are O-fucosylated (16), but we did not obtain any experimental evidence supporting C-mannosylation of these Trp residues.

Punctin-NQ with and without the specific Trp mutations was expressed in two cell lines, CHO and HEK293, to determine the consequence of the mutation for protein secretion and to elucidate the possible influence of cell- and species-specific effects. In HEK293F cells, W36A punctin-NQ showed a small but statistically significant reduction of secretion (Fig. 5, A and B), whereas W39A punctin-NQ and W42A punctin-NQ were not detectable in the medium, and therefore were not formally quantitated (Fig. 5 A). The levels of both W385A punctin-NQ and W445A punctin-NQ in conditioned medium were reduced (Fig. 5, A and B). Analysis of cellular levels of punctin-NQ and the Trp mutants demonstrated that there was an elevated cellular content of each mutant (Fig. 4 C). In addition, lysates of cells expressing W36A, W39A, and W42A, and to a lesser extent W385A but not W445A or pNQ, showed the presence of molecular species of higher than the expected mass (Fig. 5 C), suggesting the formation of aberrant complexes. The molecular mass observed for the putative complexes, 120 and 180 kDa, corresponds to the expected mass of punctin dimers and trimers, although we cannot exclude the possibility that these molecular species could represent punctin-1 complexed with another protein. Similar results were observed when the same mutants were transfected in CHO-K1 cells. Specifically, in CHO-K1 cells, there was a modest reduction of secretion of W36A and W42A, whereas there was statistically significant reduction of secretion of W39A and W385A with corresponding increases in cellular levels (data not shown). Overall, secretion of the mutants in CHO cells showed a similar trend as in HEK293F cells, but with a milder effect, indicating that effects of mutagenesis were dependent on the cell type used for the experiments. Because HEK293F cells showed a more severe effect on the punctin-NQ Trp substitutions, we used this cell line for further analysis.

We interpreted the results of this set of Ala for Trp substitutions as suggesting that mutation of Trp³⁹ and Trp⁴² had severe consequences because these residues were themselves mannosylated, with Trp⁴² also serving as the +3 residue for mannosylation at Trp³⁹, constituting in effect a double knock-out of C-mannosylation. However, an alternative interpretation was that structural consequences of W39A and W42A, i.e. lack of structural stability resulting from absence of the Trp aromatic side chain, could underlie the observed effect. Therefore, Ala substitution might be potentially unable to distinguish between the effects of C-mannosylation deficiency or the effects of structural perturbation.

It was previously shown that substitution of Ala for Trp at the +3 position in the WXXW consensus position of RNase 2 abolished C-mannosylation, whereas substitution of Phe for this Trp residue partially allowed the C-mannosylation of RNase 2 (up to 23%) (22). Phe has not, however, been reported to be a substrate for the putative mannosyltransferase. In view of this and because Phe substitution for Trp constitutes a structurally conservative substitution for Trp (both residues have an aromatic side chain), we undertook substitution with Phe. We substituted Trp³⁹ and Trp⁴² with Phe because Ala substitution of these residues, which were shown to be modified by C-mannosylation, abolished punctin-1 secretion. We also substituted Phe for Trp³⁸⁵, because Ala substitution had a more severe effect on this residue than Trp⁴⁴⁵ (Fig. 5). Phe for Trp at position 385 (i.e. ³⁸⁵FTAC, which contains no Trp residues, and cannot be C-mannosylated) rescued the secretion defect observed in W385A punctin-NQ, whereas W39F punctin-NQ and W42F punctin-NQ were not detected in the medium (Fig. 6, A and B). However, we did not detect significant differences in the cellular levels (Fig. 6 C). These results suggested that the role of Trp at position 385 was predominantly structural, whereas the mutations of Trp³⁹ and Trp⁴² possibly reported the effects of both a structural perturbation as well as lack of C-mannosylation. Interestingly, the higher molecular mass species observed intracellularly with Ala substitutions (Fig. 5 C) were not seen in any of the Phe substitutions (Fig. 6 C), suggesting less of a structural perturbation by Phe.