O- Fucosylation of ADAMTSL2 is required for secretion and is impacted by geleophysic dysplasia-causing mutations

ADAMTSL2 mutations cause an autosomal recessive connective tissue disorder, geleophysic dysplasia 1 (GPHYSD1), which is characterized by short stature, small hands and feet, and cardiac defects. ADAMTSL2 is a matricellular protein previously shown to interact with latent TGF- b binding protein 1 and influence assembly of fibrillin 1 microfibrils. ADAMTSL2 contains seven thrombospondin type-1 repeats (TSRs), six of which contain the consensus sequence for O fucosylation by protein O- fucosyltransferase 2 (POFUT2).


Introduction
ADAMTSL2 1 is a member of the A Disintegrin-like And Metalloprotease with Thrombospondin type I repeats (ADAMTS) superfamily. The mammalian ADAMTS family O-fucosylation affects ADAMTSL2 secretion contains 19 secreted metalloproteases and seven matricellular ADAMTS-Like (ADAMTSL) members (1). ADAMTS proteins consistently have one or more Thrombospondin type-1 Repeats (TSRs), a cysteine rich domain, a spacer region, and a variety of C-terminal modules in addition to the protease domain, whereas ADAMTS-like proteins lack the protease domain. Although the functions of ADAMTSL proteins are less well understood, recent findings from genetic disorders suggested that some of them may operate within the context of fibrillin microfibrils or synaptic function (2)(3)(4). ADAMTSL2 is a secreted matricellular glycoprotein containing seven TSRs (Figure 1) (5). ADAMTSL2 mutations were identified in patients with autosomal recessive geleophysic dysplasia (GPHYSD1, OMIM #231050), a syndrome characterized by short stature, short tubular bones, thick skin, and early death caused by cardiopulmonary abnormalities (6). Some of the phenotypes, such as short stature and bones are also seen in Peters Plus Syndrome (PTRPLS, OMIM #261540) patients (7), implying there might be similar molecular mechanisms contributing to GPHYSD1 and PTRPLS pathologies.
Six of sixteen documented GPHYSD1 mutations have been identified within ADAMTSL2 TSRs (6,8,9). TSRs are small protein motifs of 50-60 amino acids and are characterized by six conserved cysteines forming three disulfide bonds (10). Three types of glycosylation are predicted on TSRs including Cmannosylation, N-linked glycosylation and Ofucosylation, all of which occur in the endoplasmic reticulum (ER) (10)(11)(12). Each of the seven TSRs in mouse ADAMTSL2 contains at least one consensus sequence for glycosylation ( Figure 1). Six of the seven TSRs contain the W-X-X-(W/C) consensus sequence for potential Cmannosylation ( Figure  1) by Cmannosyltransferases (13). C-mannosylation occurs co-translationally (14) and has been suggested to promote efficient folding and secretion of target proteins, such as ADAMTSL1 (13,15,16). Disrupting C-mannosylation by eliminating C-mannosyltransferases impairs secretion of target proteins (13,17). N-linked glycosylation is also largely co-translational and occurs on asparagine residues within the consensus sequence N-X-(S/T) (X is any amino acid other than proline) by oligosaccharyl transferase complex (18). N-glycosylation is well-known to play a role in protein quality control and folding (19). ADAMTSL2 contains ten N-glycosylation sites, one of which falls within TSR6 and overlaps with a predicted Protein O-fucosyltransferase 2 (POFUT2) consensus sequence (Figure 1).
TSRs are modified with an O-fucose on a serine (S) or threonine (T) within the consensus sequence C-X-X-(S/T)-C by POFUT2 (12). POFUT2 only modifies TSRs after they are folded and has been proposed to be a folding sensor for TSRs (20,21). The O-fucose is typically elongated with a glucose (Glc) by b1,3glucosyltransferase (B3GLCT), forming a Glucoseb1-3Fucose (GlcFuc) disaccharide on TSRs (10). Elimination of Pofut2 in mice leads to embryonic lethality with defects in gastrulation (22). In contrast, mutations in B3GLCT (previously named B3GALTL) lead to PTRPLS, a developmental disorder characterized by anterior eye chamber defects (specifically a malformation termed Peters Anomaly), short stature, craniofacial abnormalities, brachydactyly, as well as variable levels of cleft palate, and a range of cardiovascular, genitourinary and cognitive defects (7,23,24). Similar craniofacial and skeletal abnormalities are observed B3glct mutant mice, although without the PTRPLS eye defects (25). Knocking down or knocking out POFUT2 in HEK293T cells reduces the secretion of target proteins, including ADAMTSL1, ADAMTS9, ADAMTS13, and ADAMTS20 (20,(25)(26)(27). In fact, secretion blockade of ADAMTS9, which is required for early development in mice, appears to be the reason for the embryonic lethality in Pofut2-null mice (28). Interestingly, siRNA knockdown or CRISPR-Cas9 knockout of B3GLCT in HEK293T cells reduces secretion of some but not all POFUT2 targets, suggesting PTRPLS and phenotypes in B3glct-null mice may result from reduced secretion of a subset of POFUT2 targets. ADAMTSL2 secretion was reduced when either POFUT2 or B3GLCT were knocked down using siRNA (20), suggesting reduction of ADAMTSL2 secretion could help explain the shared skeletal phenotypes between GPHYSD1 and PTRPLS.

O-fucosylation affects ADAMTSL2 secretion
Several GPHYSD1 mutations reduce secretion of ADAMTSL2, including a GPHYSD1 mutation adjacent to the Ofucosylation site in TSR6 (G811R) (6). Another GPHYSD1 mutation, S635L, falls within the POFUT2 consensus sequence in TSR3 from ADAMTSL2 (8). These results raised the possibility that altered O-fucosylation is responsible for reduced secretion of ADAMTSL2 in GHPYSD1 S635L and G811R patients. To address the relationship between GPHYSD1 mutations and glycosylation of ADAMTSL2, we used mass spectrometry to analyze glycosylation on wild type mouse ADAMTSL2, which shares 88% sequence identity with human ADAMTSL2 including complete conservation of consensus sequences for TSRs glycosylation. TSR3 and TSR6, in which these mutations occur, share sequence identity of 95% and 96%, respectively (Supporting Information Figure S1). Mouse knockouts of Adamtsl2 phenocopy many of the defects observed in GPHYSD1 patients (4,29), suggesting that mice provide an excellent model for studying this disease. In addition, we also examined the impact of GPHYSD1 mutations on glycosylation and secretion of ADAMTSL2 using cell-based secretion assays. We show that loss of O-fucosylation, such as in POFUT2 -/cells, eliminated secretion of mouse ADAMTLS2, whereas lack of extension to the GlcFuc disaccharide in B3GLCT -/cells had no effect on secretion. We also demonstrated that GPHYSD1 analogous mutations in TSR3 (S641L) and TSR6 (G817R) reduced secretion of mouse ADAMTSL2. Significantly, we showed that the reduction in secretion of the S641L mutation results from reduced O-fucosylation of TSR3, providing a molecular explanation for GPHYSD1 in patients with this mutation.

ADAMTSL2 TSRs were modified with Ofucose and C-mannose at variable stoichiometries
The seven ADAMTSL2 TSRs contain six consensus sites for O-fucosylation, eight consensus sites for C-mannosylation, and one consensus site for N-glycosylation ( Figure 1). To determine the glycoforms present on ADAMTSL2 TSRs and the stoichiometry of glycan modifications, we subjected purified mouse ADAMTSL2 (mADAMTSL2) to tryptic and chymotryptic digestions and analyzed digested fragments via nano-LC-MS/MS to identify glycopeptides containing the POFUT2 consensus sites as described in Experimental Procedures. To determine the relative amounts of O-fucose (O-Fuc) monosaccharide, disaccharide, and unmodified peptide (naked), we generated extracted ion chromatograms (EICs) for each glycoform of the selected peptides and calculated peak areas for each (30). At predicted Ofucosylation sites, the elongated GlcFuc disaccharide was the major glycoform on TSR1, TSR3 and TSR7 (Figure 2A, Supporting Information Figures S2A, S2B, S2D, S2H and S2I). TSR5 was mainly unmodified with only a small percentage of GlcFuc disaccharide, suggesting that an innate characteristic of this TSR reduces recognition by POFUT2 (Figure 2A, Supporting Information Figure  S2F). Surprisingly, TSR2 was predominantly modified with O-Fuc monosaccharide, suggesting this TSR is poorly modified by B3GLCT (Figure 2A, Supporting Information Figure S2C).
Using mass spectrometry, we were unable to detect peptides containing the TSR6 POFUT2 consensus site, even with multiple different protease and glycosidase digestions. TSR6 is unusual in that the POFUT2 consensus site overlaps with an N-glycosylation consensus site and therefore could be modified with either or both modifications ( Figure 1). As an alternative approach to determine whether TSR6 was O-fucosylated, we tested whether ADAMTSL2 TSR6 (TSL2-TSR6) expressed in HEK293T cells incorporated the bio-orthogonal probe 6-alkynylfucose (6AF). Previously we have shown that 6AF is efficiently transferred onto O-fucose sites in TSRs by POFUT2 and can be modified with azido-biotin using "click" chemistry, allowing for detection with streptavidin (31). 6AF was efficiently incorporated onto TSP1-TSR3 which is known to be O-fucosylated (31) but was not incorporated on to TSL2-TSR6 ( Figure 2B), demonstrating that TSR6 from ADAMTSL2 is not Ofucosylated. To determine whether TLS2-TSR6 was N-glycosylated, we tested for presence of Nglycans by treatment with PNGase F ( Figure 2B). TSL2-TSR6 size-shifted significantly following O-fucosylation affects ADAMTSL2 secretion PNGase F treatment, suggesting that it is Nglycosylated. The complete shift of TSL2-TSR6 after PNGase F treatment suggested the Nglycosylation in TSR6 was stoichiometric. In contrast, TSR3 from human thrombospondin-1 (TSP1-TSR3), which lacks an N-glycosylation site, did not shift ( Figure 2B).
In contrast to POFUT2 consensus sites, C-mannosylation sites were modified at lower stoichiometries ( Figure 2C). For ADAMTSL2 TSR1, the most abundant C-mannosylation species were detected on the second and third tryptophans in the W-X-X-W-X-X-W-X-X-C motif, with very little modification of the first W (Supporting Information Figure S2B). In contrast, TSRs 3-7 each have a single W-X-X-C Cmannosylation consensus motif ( Figure 1). Among these TSRs, significant levels of Cmannosylation were only detected on TSR3 and TSR7, with no modification detected on TSRs 4, 5, or 6 (Supporting Information Figures S2D-2G, and S2I). The glycan modifications are summarized in Figure 2D.

ADAMTSL2 secretion requires TSR Ofucosylation but not extension to the GlcFuc disaccharide
Previous work demonstrated that siRNA knockdown of POFUT2 in HEK293T cells significantly reduced ADAMTSL2 secretion (20). To confirm the importance of TSR Ofucosylation for ADAMTSL2 secretion, we examined ADAMTSL2 secretion in Lec13 CHO cells grown in the presence or absence of fucose ( Figure 3A). Lec13 CHO cells have a mutation in GDP-mannose 4,6-dehydratase (GMD) (32), one of the enzymes in the de novo GDP-fucose biosynthetic pathway, the predominant pathway for GDP-fucose synthesis and utilization in humans. The mutation can be rescued through the GDP-fucose salvage pathway by addition of fucose to the medium (20). In the absence of fucose, ADAMTSL2 was trapped in the cell lysate and was not secreted to the medium from Lec13 CHO cells, but secretion was restored when medium was supplemented with fucose ( Figure 3A), suggesting that loss of either TSR Ofucosylation or fucosylation of N-glycans impaired ADAMTSL2 secretion. To distinguish between these possibilities, we compared secretion of ADAMTSL2 in wild type (Pro5) and mutant (Lec1) CHO cells ( Figure 3B). Lec1 cells have a mutation in Mgat1 that blocks processing of N-glycans to hybrid or complex types that could contain fucose (33). ADAMTSL2 was secreted from both Pro5 and Lec1 cells ( Figure  3B). Moreover, ADAMTSL2 secreted from the Lec1 cells migrated faster than that from the Pro5 cells, consistent with the lack of N-glycan processing. The efficient secretion of ADAMTSL2 lacking fucosylation on N-glycans suggested that loss of O-fucosylation on ADAMTSL2 TSR motifs was responsible for reduced secretion of ADAMTSL2 secretion from Lec13 CHO cells.
To further confirm the importance of TSR O-fucosylation for ADAMTSL2 secretion and determine the role of the B3GLCT-mediated extension to the disaccharide, we evaluated secretion of ADAMTSL2 using recently developed CRISPR-Cas9 knockouts of POFUT2 or B3GLCT in HEK293T cells (28,34). ADAMTSL2 was efficiently secreted in wild type HEK293T cells. In contrast, secretion was eliminated in POFUT2 -/cells ( Figure 4A, B). Co-transfection with POFUT2 restored secretion of ADAMTSL2 in POFUT2 -/cells, demonstrating the specificity of the mutation. In contrast, knocking out B3GLCT did not affect ADAMTSL2 secretion ( Figure 4A, B). Secretion of hIgG was not affected in either POFUT2 -/or B3GLCT -/cell lines, suggesting that reduced secretion of ADAMTSL2 in POFUT2 -/cells did not result from a global defect in trafficking. Moreover, the loss of the upper band in the POFUT2 knockout cell lysate provided evidence that ADAMTSL2 lacking O-fucosylation was trapped in the endoplasmic reticulum and did not move to the Golgi where terminal processing of the N-glycans occurs ( Figure 4A, right-hand panel). Taken together (Figures 3 and 4), Ofucosylation by POFUT2 on TSR motifs was essential for ADAMTSL2 secretion, but elongation of the O-fucose on ADAMTSL2 TSRs by B3GLCT was dispensable for proper secretion.

The S641L and G817R GPHYSD1 mutations impair ADAMTSL2 secretion, and the S641L mutation blocks O-fucosylation of ADAMTSL2 TSR3
O-fucosylation affects ADAMTSL2 secretion Previous work by Le Goff et al. (6) showed that the human ADAMTSL2 GPHYSD1 G811R substitution reduced protein secretion. We examined whether introducing GPHYSD1 S635L or G811R equivalent mutations into mADAMTSL2 (S641L or G817R, respectively) similarly affected secretion. mADAMTSL2 S641L was secreted at less than 30% and G817R at less than 10% of wild type mADAMTSL2 ( Figure 5). These data suggested that the GPHYSD1 phenotype caused by these mutations likely resulted from inappropriately low levels of ADAMTSL2 in the extracellular matrix.
The GPHYSD1 S641L and G817R mutations are located within and adjacent to the TSR3 and TSR6 POFUT2 consensus sites, respectively ( Figure 1). Given their location, reduced ADAMTSL2 secretion could result from defects in glycosylation. To test this hypothesis, we evaluated the glycosylation on mADAMTSL2 S641L using mass spectrometry. Full-length WT and S641L mutated mADAMTSL2 were subjected to chymotrypsin digestion and analysis by nano-LC-MS/MS. The most abundant form of the peptide from TSR3 of WT mADAMTSL2 was fully modified with one C-mannose and the GlcFuc disaccharide ( Figure  6A). In contrast, the most abundant form of the equivalent peptide containing the S641L mutation only had the C-mannose with no Ofucose glycan ( Figure 6A, Supporting Information Figure S3). The mutation only affected O-fucosylation of TSR3 of ADAMTSL2; O-fucosylation and C-mannosylation of all other TSRs remained the same as WT ADAMTSL2 (Supporting Information Figure S4). Hence, the S641L mutation only caused local loss of Ofucosylation in TSR3 without global effects on O-fucosylation or C-mannosylation on other TSRs in mADAMTSL2.

Elimination of the O-fucose site on TSR3 reduces ADAMTSL2 secretion
To determine whether loss of Ofucosylation on TSR3 was responsible for the secretion defect, we mutated the O-fucosylated Thr to Ala (T643A) in full-length mADAMTSL2 and evaluated the effect of the mutation on secretion. The T643A mutated form was secreted approximately 50% less than WT ( Figure 6B, C), suggesting that much of the secretion defect of the S641L mutant was caused by loss of Ofucosylation on TSR3 rather than the S641L mutation itself.
Since the glycosylation sites in TSR6 could not be detected using mass spectrometry, we analyzed whether the G817R mutation affected N-glycosylation and secretion by introducing G817R, as well as mutations to inhibit N-glycosylation (N813Q, T815V, N813Q/T815V) into mADAMTSL2-TSR6 (TSL2-TSR6) and labeled the proteins with 6AF as in Figure 2B. TSL2-TSR6-G817R was secreted, although poorly compared to WT, and migrated at the same molecular weight with WT TSL2-TSR6 in both medium and cell lysate ( Figure 7A, B). In contrast, disrupting the Nglycan site (N813Q, T815V, and N813Q/T815V) prevented secretion and resulted in a molecular weight shift downwards in cell lysates compared to WT TSL2-TSR6 and G817R ( Figure 7B). The reduction of secreted G817R protein ( Figure 7A) was consistent with our result that the G817R mutation also reduced full-length mADAMTSL2 secretion ( Figure 5). Interestingly, removal of Nglycosylation with N813Q did not restore Ofucosylation at threonine 815 ( Figure 7A). Taken together these results show that the TSR6 G817R mutation did not disrupt N-glycosylation but reduced secretion of mADAMTSL2 without affecting the N-glycosylation status of TSR6.

Discussion
Here we examined the effects of two GPHYSD1 mutations in ADAMTSL2 occurring within POFUT2 consensus sites, one in TSR3 (S641L) and the other in TSR6 (G817R). The corresponding human mutations (S635L and G811R, respectively) were both identified in GPHYSD1 patients who showed compound heterozygosity, i.e., the other ADAMTSL2 allele in each affected individual harbored a different mutation, W862X (6) and c.
[1219C>T] (8), respectively. The W862X is predicted to produce a truncated peptide lacking TSR7 and C-terminal residues but is likely to lead to nonsensemediated mRNA decay (6). The c.
[1219C>T] mutation is a splice-site mutation that would block splicing of the exon encoding the N-glycan rich domain, although this has not been verified experimentally (8). These two mutations, W862X O-fucosylation affects ADAMTSL2 secretion and c. [1219C>T], are likely to be quite severe, resulting in loss of function from the respective alleles. Our data shows that the corresponding GPHYSD1 substitutions in mouse ADAMTSL2, S641L and G817R significantly reduced secretion of ADAMTSL2, suggesting that low levels of ADAMTL2 in the extracellular matrix lead to GPHYSD1. The fact that the S641L mutation eliminated O-fucosylation of TSR3, and that elimination of O-fucosylation on TSR3 significantly reduced secretion of ADAMTSL2, suggests that loss of O-fucose on TSR3 is a major contributing factor to the defects observed in patients with the S635L mutation. In contrast, since the G817R mutation had no effect on the glycosylation status of TSR6, the reduction of mADAMTSL2 secretion of G817R appeared to be caused directly by the mutation.
PTRPLS patients display facies and skeletal defects similar to those seen in GPHYSD1. Similar skeletal defects were also observed in global knockouts of Adamtsl2 and B3glct in mice such as shortened front and hind limbs and irregular shape of the skull (4,25). Due to these similar resemblances, we hypothesized that ADAMTSL2 may be a prime B3GLCT target relevant to the skeletal defects observed in PTRPLS. Several GPHYSD1 missense mutations in addition to those described above have been reported to reduce secretion (6,8,9) suggesting this may be a major mechanism causing disease. Previous work from our laboratory showed that ADAMTSL2 secretion was lost when B3GLCT was knocked down using siRNA in HEK293T cells, supporting our hypothesis that ADAMTSL2 is a key factor for the overlapping phenotypes between PTRPLS and GPHYSD1 (20). In contrast, our data here indicated that the secretion of ADAMTSL2 was not affected by CRISPR-Cas9 deletion of B3GLCT in HEK293T cells. Comparing these two gene manipulation methods, siRNA has stronger off-target effects by triggering the translational repression and/or degradation of non-target genes as well as nonsequence specific off-target effects (35). This can overwhelm the endogenous interaction between microRNA and the RNA induced silencing complex (RISC) (35). We attribute the loss of ADAMTSL2 secretion to non-specific repression of global cellular networks in B3GLCT siRNA knockdown experiments. These data suggest that reduced secretion of ADAMTSL2 may not cause the overlapping phenotypes between GPHYSD1 and PTRPLS.
Our mass spectral analyses showed that ADAMTSL2 is O-fucosylated at most of the predicted sites in each of its TSR domains, further validating the reliability of the POFUT2 consensus sequence: C-X-X-(S/T)-C. TSRs 1, 3 and 7 were all modified and elongated with the O-fucose disaccharide at high stoichiometries, whereas TSR2 was predominantly monosaccharide, and TSRs 5 and 6 were unfucosylated (Figure 2A). The S641L mutation completely eliminated O-fucosylation of TSR3, suggesting the replacement of S by L in the POFUT2 consensus sequence (C-L-R-T-C) severely reduced recognition of this site. Prior work in our laboratory analyzed the effects of mutations in the consensus sequence (C-X a -X b -(S/T)-C) using a model TSR (TSR3 from human thrombospondin 1, TSP1-TSR3) (36). Consistent with our S641L results, an L in the X a position significantly reduced O-fucosylation of TSP1-TSR3 (36). TSR6 has overlapping Nglycosylation and O-fucosylation consensus sequences: C-N-T-T-C. Our results showed that TSR6 is N-glycosylated but not O-fucosylated, but that eliminating the N-glycosylation site by mutating the N to a Q failed to restore Ofucosylation. This was surprising, as we assumed that the addition of the N-glycan, which occurs prior to protein folding, was blocking Ofucosylation by POFUT2, which occurs after TSR folding (20). This result suggested that mutating the N to Q in the TSR6 POFUT2 consensus sequence (C-Q-T-T-C) reduced modification by POFUT2. A Gln in the X a position also significantly reduced Ofucosylation in TSP1-TSR3, which may explain why the N813Q mutant was not O-fucosylated in TSR6. Our results here confirm the importance of the amino acid in the X a position in the consensus in determining the efficiency of O-fucosylation by POFUT2. Alternatively, the lack of Ofucosylation on TSR5 or 6 could be the result of overexpression of ADAMTSL2 in a cell-culture system or that TSR6 was analyzed as a single TSR instead of in the context of the full-length protein.
Of the TSRs where we have analyzed stoichiometry of modification with the GlcFuc disaccharide (ADAMTS9 (37), ADAMTS20 (25), ADAMTS17 (34)), TSR2 of ADAMTSL2 is the first TSR we have seen that is only modified with O-fucose monosaccharide. The only other TSR that is partially modified with the glucose elongation is TSR6 from ADAMTS20 (25). The consensus sequence for TSR2 of ADAMTSL2 is C-S-A-T-C, while that for TSR6 of ADAMTS20 is C-T-A-T-C. It will be interesting to see if this similarity contributes to the reduced modification of O-fucose by B3GLCT at these sites.
All of the TSRs in ADAMTSL2 besides TSR2 have the consensus sequence W-X-X-C for C-mannosylation, and TSR1 contains the extended W-X-X-W-X-X-W-X-X-C consensus sequence ( Figure 1). TSRs 1, 3, and 7 were modified at 1 or multiple W-X-X-(W/C) tryptophans while TSRs 4-6 lacked any Cmannose modifications. The W-X-X-W sites have been reported to be modified by the Cmannosyltransferase DPY19L1, and the W-X-X-C sites by DPY19L3 (13). Since TSRs 4-6 were not modified, additional factors must be controlling recognition of these W-X-X-C sites by DPY19L3. Recent research has shown that Cmannosylated tryptophans with the W-X-X-W-X-X-W have a stabilizing effect on a TSR from UNC-5, with the mannose on the first W having a stronger effect than the mannose on the second W (13,16). Since the first W of the W-X-X-W-X-X-W-X-X-C site on TSR1 of ADAMTSL2 is poorly modified, we anticipate the stabilizing effects from C-mannosylation would be low for ADAMTSL2 TSRs. In contrast, four out of six TSRs in mADAMTSL2 had high stoichiometry of O-fucosylation, adding weight to the importance of O-fucosylation for stabilization of ADAMTSL2 TSRs.
Our analysis of ADAMTSL2 GPHYSD1-related mouse mutations provided evidence that disruption of TSR specific Ofucosylation may contribute to protein secretion defects that lead to potential disease mechanisms and pathology. Sub-stoichiometric Ofucosylation of highly conserved TSRs suggested that modulation of O-fucosylation was important for efficient trafficking of TSR-containing proteins. In light of the current interest in development of therapeutics with TSR motifs as targets (38), it is essential to advance our knowledge of TSR motif function by further investigating the functional role of Ofucosylation and the mechanistic role of glucosefucose elongation.

Experimental Procedures
Plasmids and Mutagenesis pcDNA3.1-mADAMTSL2-Myc-His6 was described previously (5) and used for expression of mouse ADAMTSL2 and its analysis. The plasmid expressing TSR6 was created by PCR amplification using primers: 5'CGTACGAAGCTTCCCACTGGCTGGCTC AAG3'and 5'GGGCCCTCCTCGAGAGCAGTGCTCTC 3', incorporating Hind III and Xho I restriction sites (underlined). Amplicons were digested with Hind III and Xho I and ligated into pSecTag2C/hygroC (Invitrogen). The plasmid encoding TSR3 of human thrombospondin 1 (hTSP1-TSR3) with a C-terminal Myc-His6 tag (pSecTag2-TSP1-TSR3-Myc-His6 ) was described previously (39). Mouse ADAMTSL2 substitutions S641L (for cell-based secretion assay), and N813Q, T815V, G817R, and N813Q+T815V in TSR6 were made by PCR-mediated site-directed mutagenesis using primers listed in Supporting Information Table S1 and Herculase II (Agilent), following manufacturer's instructions. PCR products were digested with Dpn I at 37 ℃ overnight then transformed into DH5a competent cells (Invitrogen). cDNA sequences were verified by sequencing the entire plasmid using primers listed in Supporting Information Table S2. Mouse ADAMTSL2 S641L (for protein purification and mass spectral analysis) was generated by PCRmediated site-directed mutagenesis using primers: 5'GTGAGTGCCTGCGTACCTGTGGTGAGG GCCATCAGTTCCG3'and 5'CAGGTACGCAGGCACTCACTCCAGCTG CTGGTCTCCCAC3'. This PCR reaction was performed with CloneAmp HiFi Premix (Takara Bio USA) following manufacturer's instructions. PCR products digestion and transformation were the same as described above. cDNA sequence was verified by sequencing the entire insert using primers in Supporting Information Table S2.

Protein expression and purification HEK293T cells (ATCC) were cultured in Dulbecco's Modified Eagle Medium (DMEM,
O-fucosylation affects ADAMTSL2 secretion GE Healthcare Life Sciences) supplemented with 10% bovine calf serum (BCS, Hyclone) and 1% penicillin plus streptomycin antibiotic (pen+strep, Sigma-Aldrich). Cells were seeded in DMEM with 10% BCS into 10 cm dishes and grown to 80% confluency before transfection. Before transfection, medium was changed to 6 ml Opti-MEM (Gibco) for each 10 cm dish. The cells were transiently transfected with mouse ADAMTSL2 wild-type (WT) or S641L plasmid using 5 µg plasmid, 30 µl of 1 mg/mL polyethylenimine (PEI, (40)), and 500 µL Opti-MEM. The plasmid, PEI and Opti-MEM mixture was incubated at room temperature for 15 min before addition to the cells. The culture medium was collected by centrifuging at 3900 rpm, 4 ℃ for 10 min after a 3-day incubation at 37 ℃. The supernatant was filtered through a 0.45 µm filter and purified using Ni-NTA (Qiagen) affinity chromatography at 4 ℃. The proteins were eluted in Tris-buffered saline (TBS) containing 250 mM imidazole and stored in -20 ℃ until use.

Cell-based Secretion Assays
Wild type and mutated ADAMTSL2 plasmids were transiently transfected into Lec1 or Pro5 cells using 0.8 µg plasmid with 0.2 µg hIgG control plasmid DNA with 6 µL PEI transfection reagent in 100 µL Opti-MEM and incubated at room temperature for 20 min. Cells at 80% confluence in 35 mm wells were washed with 2 ml PBS, then incubated in 1ml Opti-MEM with transfection mixture added drop-wise and cultured for 4 hrs before a medium change with 1 ml of fresh Opti-MEM. CRISPR-Cas9 HEK293T knockouts of POFUT2 and B3GLCT were generated as described previously by Benz et al. (28) and by Hubmacher et al. (34), respectively. Cells were seeded in 2 ml (4.25x10 6 cells/ml) of DMEM supplemented with 10% BCS in 6-well plates, incubated overnight and transfected when 80% confluent. Transient co-transfection of POFUT2 -/-, B3GLCT -/and wild type (WT) HEK293T cells used 1.2 µg of mouse ADAMTSL2 plasmid or pSecTag2/hygroC empty vector as a control, 0.1µg of hIgG as a control for normalization of secreted protein, 0.24 µg of POFUT2 or B3GLCT plasmid for rescue of secretion, or empty vector as a control with 9.84 µL of PEI in 154 µL of Opti-MEM incubated at room temperature for 15 min. Before transfection, the old media was changed to 700 µL Opti-MEM and the transfection mixture was added drop-wise and cultured for 48 hrs at 37 ℃ before collection for western blot analysis.

Labeling with 6-alkynylfucose
To label proteins with 6-alkynylfucose (6AF), transfected cells were cultured in 200 µM alkynyl fucose (Invitrogen). Cycloaddition "click" reactions were performed to label fucosylated proteins as previously described (31). Medium and lysates were incubated with 1 mM copper sulfate and 2 mM sodium ascorbate in PBS with 0.1 mM azido-biotin and 0.1 mM Tris(benzyltriazolymethyl) amine for 1 hr at room temperature. Precipitant in the reaction mixture was removed by centrifugation, and samples were analyzed by immunoblot. For the PNGase F digest in Figure 2B, plasmids encoding hTSP1-TSR3 or mADAMTSL2-TSR6 were transfected into HEK293T cells using Lipofectamine 2000 as described by the manufacturer (ThermoFisher), and the cells were cultured in 200 µM alkynyl fucose for 48 hours. The proteins were purified from the medium using Ni-NTA-agarose and digested with PNGaseF in 100 mM Tris-HCl, pH 8.0, 2% NP-40, for 4 h at 37 o C. The samples were then subjected to click chemistry with final concentrations of 20 µM azido-biotin, 500 µM THPTA, 100 µM CuSO4, and 5 mM sodium ascorbate for 20 min at RT. The reaction was stopped by heating to 100 o C in 1X Laemmli sample buffer.
Immunoblots were visualized and band intensities were quantified on an Odyssey Imager (LI-COR) software.

Mass Spectral Analysis
Recombinant full-length mouse ADAMTSL2-Myc-6xHis WT or S641L mutant were expressed and purified as described above. Approximately 1µg of purified protein was denatured in 8 M urea and 0.4 M ammonium bicarbonate. Samples were reduced with 10 mM Tris(2-carboxyethyl) phosphine (Thermo Fisher) for 5 min at 50 ℃, alkylated with 30 mM iodoacetamide in the dark for 30 min at room temperature and diluted with mass spec grade H2O to a final urea concentration of 2 M. Samples were then digested with 0.5 µg of trypsin (cleaves C-terminal to lysines or arginines, Sigma-Aldrich) or chymotrypsin (cleaves C-terminal to tryptophan, phenylalanine, tyrosine, leucine, or isoleucine, Sigma-Aldrich) for 4 hrs at 37 ℃, acidified with 5% formic acid, sonicated for 20 min, then desalted using C18 ZipTips (EMD Millipore), eluted in 50% acetonitrile/0.1% acetic acid mixture. Samples were diluted to 2 ng/µl in 20% acetonitrile with 0.1% formic acid prior to injection into the mass spectrometer. 10 ng of each sample was injected and detected on a Q-Exactive Orbitrap mass spectrometer (Thermo Fisher) equipped with an Easy nano-LC HPLC system with a C18 EasySpray PepMap RSLC C18 column (50 µm × 15 cm, Thermo Fisher Scientific). Separation of glycopeptides was carried out using a 30 min binary gradient consisting of solvent A (0.1% formic acid in water) and solvent B (90% acetonitrile and 0.1% formic acid in water) with a constant flow rate of 300 nl/min. The resulting spectra were acquired in the positive polarity/ion mode over a range of 350-2000 m/z at a resolution of 35,000 with an automatic gain control target value of 1x10 6 . The top 10 most abundant precursor ions in each full MS scan were isolated and subjected to higher energy collision induced dissociation-tandem mass spectrometry (HCD-MS/MS) and fragmented with a normalized collision energy of 27%, an automatic gain control target value of 2x10 5 with an isolation window of 3 m/z at a fragment resolution of 17,500 and dynamic exclusion enabled. Peak lists and .raw data files were generated using Xcalibur software set to its default settings. Raw data files were analyzed using Proteome Discoverer 2.1.0.81 (Thermo Fisher) and were searched against a mouse ADAMTSL2 database (Accession No. Q7TSK7 v1 (October 1, 2003)). Byonic software v.2.10.5 (Protein Metrics) was used as a module inside Proteome Discoverer for identifying peptides with glycan modifications. Fixed modification was carbamidomethyl on cysteines; variable modifications were Hex on tryptophan, oxidation on methionine, dHex on serine and threonine, dHexHex on serine and threonine. Five missed cleavages were permitted. Mass tolerance for precursor ions was set to 10 ppm and mass tolerance for fragment ions was set to 20 ppm. Due to the lability of the fucose-peptide bond in HCD experiments, Byonic is frequently unable to correctly assign the O-fucosylated Ser/Thr residue in a peptide. All assignments are based on the well-documented consensus sequence for Ofucosylation of Group 1 TSRs: C 1 -X-X-(S/T)-C 2 (21). Note that all of the peptides identified by Byonic with an O-fucose modification contained this consensus sequence. Protein and peptide false discovery rates were set to a threshold of 1% and calculated in Byonic software v.2.10.5 (Protein Metrics) using the 2-dimensional target decoy strategy as described (41). Extracted ion chromatograms (EIC) for all peptides were generated using Xcalibur Qual Browser 4.0.27.19 (Thermo Fisher). The glycoform distribution on each TSR was quantified based on area under the curve of each EIC for all biological and technical replicates.

Experimental Design and Statistical Rationale
See Table 1.