Site-specific O-glycosylation of members of the low-density lipoprotein receptor superfamily enhances ligand interactions

The low-density lipoprotein receptor (LDLR) and related receptors are important for the transport of diverse biomolecules across cell membranes and barriers. Their functions are especially relevant for cholesterol homeostasis and diseases, including neurodegenerative and kidney disorders. Members of the LDLR-related protein family share LDLR class A (LA) repeats providing binding properties for lipoproteins and other biomolecules. We previously demonstrated that short linker regions between these LA repeats contain conserved O-glycan sites. Moreover, we found that O-glycan modifications at these sites are selectively controlled by the GalNAc-transferase isoform, GalNAc-T11. However, the effects of GalNAc-T11–mediated O-glycosylation on LDLR and related receptor localization and function are unknown. Here, we characterized O-glycosylation of LDLR-related proteins and identified conserved O-glycosylation sites in the LA linker regions of VLDLR, LRP1, and LRP2 (Megalin) from both cell lines and rat organs. Using a panel of gene-edited isogenic cell line models, we demonstrate that GalNAc-T11–mediated LDLR and VLDLR O-glycosylation is not required for transport and cell-surface expression and stability of these receptors but markedly enhances LDL and VLDL binding and uptake. Direct ELISA-based binding assays with truncated LDLR constructs revealed that O-glycosylation increased affinity for LDL by ∼5-fold. The molecular basis for this observation is currently unknown, but these findings open up new avenues for exploring the roles of LDLR-related proteins in disease.

The low-density lipoprotein receptor (LDLR) 3 and related receptors are membrane-bound cell surface receptors with important endocytic functions for lipoproteins and a variety of diverse extracellular ligands (1). The founding member LDLR is important for maintaining cholesterol homeostasis, and deleterious mutations in LDLR lead to decreased LDL catabolism and elevated levels of plasma LDL-cholesterol (2). The LDLR superfamily includes the VLDL receptor (VLDLR), LDLR-related protein 1 (LRP1), LDLR-related protein 1B (LRP1B), LDLRrelated protein 2 (LRP2 or megalin), and LDLR-related protein 8 (LRP8 or ApoER2), as well as more distantly related receptors such as the sortilin-related receptor (1). These members also have important roles in cardiovascular diseases, as well as neurodegenerative and proteinuric renal diseases (3,4). The ectodomains of LDLR-related proteins share characteristic structural features including LDLR-type A repeats (LA) (complement-like cysteine-rich ligand binding repeats), epidermal growth factor-like repeats, and Tyr-Trp-Thr-Asp ␤-propeller domains (1).
The LA modules consist of ϳ40 amino acids each with three disulfide bridges (Cys 1-3 , Cys 2-5 , and Cys 4 -6 ). They are found as clusters of seven repeats in LDLR, eight in VLDLR, and multiple clusters in LRP1 and LRP2, and they constitute the dominant ligand-binding region of these receptors (see Fig. 1) (5). The functional structure of the LA module requires a calcium ion coordinated by conserved acidic residues found between Cys 4 and Cys 6 in the LA sequence, and binding to different ligands appears to require different subsets of LA modules (6 -9). The LA modules are bound by the ER-resident receptorassociated protein (RAP) early in the secretory pathway, and this interaction, believed to prevent premature intracellular binding to ligands, is lost at lower pH such as in later Golgi compartments (5,9,10). The LA modules are interspaced by a short linker sequence mostly formed by four residues ending in Thr with the sequence motif XXC 6 XXXTC 1 XX, although some linkers are longer. Recently we demonstrated that the evolutionary conserved Thr residues in these linkers of all LDLRrelated proteins carry O-glycans (11). Moreover, we found that only one of the many polypeptide GalNAc-transferase (Gal-NAc-T) isoenzymes, GalNAc-T11, which initiates O-glycosylation of proteins, was responsible for introducing O-glycans at the Thr in the XXC 6 XXXTC 1 XX motif of linkers in LDLR and presumably other receptors with this motif.
Both N-and O-glycosylation of LDLR have been reported previously (7,12). O-Glycosylation in the stem region is important for cell-surface expression and stability of this receptor as demonstrated with the CHO ldlD cell line deficient in the UDP-Glc/GlcNAc C4-epimerase (12). Previous studies have also suggested that O-glycosylation of LDLR in the N-terminal domain may be important for LDL binding and uptake; however, the nature and positions of these O-glycans were not identified (13,14). In the present study, we explored the functional role of the identified O-glycans in the short linker regions between LA modules (11). We took advantage of the finding that these O-glycans are specifically generated by GalNAc-T11, whereas the O-glycans important for stability of LDLR on the cell membrane are directed by multiple GalNAc-Ts (11). Using genetically engineered human HepG2 and HEK293 cells, as well as Chinese hamster ovary (CHO) cells, we demonstrate that the O-glycans in the LA ligand-binding region of LDLR, as well as VLDLR, are important for high-affinity lipoprotein binding and uptake and that the sialic acids carried by these O-glycans are essential for this. These findings have important implications for our understanding of the function and regulation of the LDLR-related proteins and their roles in diseases.

O-Glycosites in the linker regions of class A repeats of LDLR-related proteins
We have previously identified a total of 17 O-glycosylation sites in evolutionary conserved linker regions in between LDLR LA repeats within the sequence motif XXC 6 XXXTC 1 XX of most of the LDLR-related proteins in a large number of engineered human SimpleCells (SCs) (Fig. 1, labeled 1) (15)(16)(17). SCs are genetically engineering cell lines with knockout of the COSMC gene, which is an essential chaperone for the core1 synthase, C1GALT1, that elongates the initial O-GalNAc glycans in most cell types. Thus, the O-glycan structures in SCs have the rather homogeneous GalNAc or in some cells NeuAc␣2,6GalNAc structures facilitating sensitive identification by MS using lectin enrichment of glycopeptides in cell digests (15). The identified O-glycosites in LDLR-related proteins include two of the four XXC 6 XXXTC 1 XX motifs in LDLR, two of five in VLDLR, two of two in LRP8, five of eight in LRP1, one of six in LRP1B, and seven of thirteen in LRP2. Additional O-glycosites in linker regions with slightly different sequences were also identified ( Fig. 1, linkers shown in red). Analysis of LDLR in HEK293 and CHO cells suggested that recombinant shed LDLR had partial occupancy of O-glycans in the LA regions, whereas O-glycosylation of the juxtamembrane region was more complete (11).
The SimpleCell glycoproteomics strategy involves genetic truncation of O-glycosylation for efficient enrichment and sensitive MS sequencing, and here we first sought to confirm O-glycosylation of the linker regions in LDLR-related proteins in WT cells and rat organs. We previously did confirm some of the LDLR, VLDLR, and LRP1 sites in WT CHO cells (18), as well as in rat liver and human plasma (19) (Fig. 1, labeled 2 and 3, and Data Set S1). We extended these studies to include comprehensive analysis of LDLR-related proteins in rat liver, brain, and kidney ( Fig. 1, labeled 4, and Data Set S2), and we were able to confirm four of the four predicted glycosites in the XXC 6 XXXTC 1 XX linker regions of LDLR, three of five in VLDLR, seven of eight in LRP1, and eight of ten in LRP2 (rat LRP2 has three fewer predicted glycosites than human between LA1-2, 23-24, and 30 -31).
To further confirm these O-glycans, we analyzed LRP2 purified from rat kidney. The neuraminidase-treated and trypsindigested LRP2 yielded coverage of ϳ57% of the ectodomain of LRP2 (Data Set S4). We identified 11 O-glycosites with six found in the 10 XXC 6 XXXTC 1 XX motifs ( Fig. 1 and Data Set S4). In most cases the corresponding peptides without O-glycans were also identified, which may suggest incomplete stoichiometry in agreement with our previous studies of secreted LDLR, although the purified LRP2 may include intracellular immature glycoforms (11).

Trafficking and cell-surface expression of LDLR is not dependent on GalNAc-T11
It is well established that LDLR requires O-glycosylation in the stem region for stability at the cell surface (12), so we first established that surface expression of LDLR in human liver HepG2 cells was unaffected by the loss of GalNAc-T11mediated O-glycosylation of the linker regions. Knockout of GALNT11 or the COSMC chaperone required for the core 1 synthase C1GALT1 to truncate O-glycans to the initial GalNAc residue in HepG2 cells did not affect expression, trafficking, or shedding of the endogenous LDLR, as evaluated by SDS-PAGE Western blotting analysis of total cell lysates ( Fig. 2A) and 35 Slabeled pulse-chase experiments (Fig. 2B). Flow-cytometric analysis of nonpermeabilized HepG2 cells further confirmed that the cell-surface expression of endogenous LDLR was unaffected by the loss of GalNAc-T11-mediated O-glycosylation or capacity for elongation of O-glycans (Fig. 2C).

Site-specific O-glycosylation of LDLR and ligand interactions
Interestingly, the general truncation of O-glycans to the unsialylated GalNAc residue by COSMC knockout in HepG2 cells (20) did not appear to affect cell surface stability or shedding of the LDLR (Fig. 2C). This is, however, at least partly in agreement with previous studies of LDLR shedding in the CHO ldlD cell line, where complete loss of O-glycosylation results in rapid shedding and degradation of LDLR, whereas allowing only GalNAc glycosylation by supplementing growth medium with GalNAc alone at least partly retained cell-surface expression (20).

LDL ligand binding and uptake by LDLR is affected by GalNAc-T11-mediated O-glycosylation
Fresh fluorescently labeled DiI-LDL was used in binding and uptake studies with isogenic HepG2 cells with isogenic cells with knockout of COSMC (SC), GALNT11, or GALNT2 (Fig.  (19), and in rat liver, kidney, and brain tissues (labeled 4) and purified rat LRP2 (labeled 5) in the present study are shown. Detailed information about O-glycosites is presented in Data Set S1, O-glycopeptides identified in the present study are described in Data Set S2, and (glyco)peptides identified in purified LRP2 are described in Data Set S4. Glycosites specifically regulated by GalNAc-T11 (yellow and red squares) were identified by comparative analysis of isolated shed LDLR expressed in HEK293 SC cells with and without KO of GALNT11 as previously reported (11) or by differential dimethyl-labeling O-glycoproteomics (17) of CHO and HEK293 SC cell lines (indicated) with and without KO of GALNT11 in unpublished studies. Note that some LA modules are not conserved between rat and human (Table S1). 3A). Initial assays were performed by exposing cells to 20 g/ml DiI-LDL for 4 h at 4°C for binding and 37°C for uptake, followed by washing, air drying, fixation, and analysis by direct fluorescent microscopy. Both binding and uptake of DiI-LDL were substantially reduced in HepG2 cells with knockout of COSMC and GALNT11, whereas knockout of the isoenzyme control GALNT2 had no effect (Fig. 3A). Interestingly, knockout of COSMC or GALNT11 produced the same reduction, suggesting that the elongated O-glycans with sialic acids are required for the function.

Site-specific O-glycosylation of LDLR and ligand interactions
We next performed a dose study (10 -60 g/ml) using FITC-LDL and including HepG2 cells with knockout of LDLR ( Fig. 3B and Fig. S2). A strong dose-dependent reduction of LDL binding and uptake was found in HepG2 cells without GANT11 and COSMC, which was intermediary to the reduction found upon knockout of the LDLR (Fig. 3B). Interestingly, both binding The only immunoreactive band corresponds to the expected migration of the mature form of LDLR (molecular mass, 160 kDa). A representative experiment of at least three is shown. B, 35 S-pulse-chase analysis of shed LDLR from HepG2 WT and GALNT11 KO cells using immunoprecipitation with a polyclonal anti-LDLR antibody from culture media. The signals were exposed in a PhosphorImager, and the only immunoreactive band corresponds to the expected migration of shed LDLR (molecular mass, 97 kDa). The cells were pulsed for 30 min and chased as indicated. A representative experiment of two is shown. C, flow cytometry analysis of LDLR surface expression on nonpermeabilized HepG2 WT and mutant cells as indicated. A representative experiment of at least three is shown.

Site-specific O-glycosylation of LDLR and ligand interactions
and uptake of FITC-LDL appeared to saturate at 40 g/ml in HepG2 cells with deficiency in O-glycosylation, as well as in LDLR, whereas saturation was not evident at 60 g/ml with WT cells (Fig. 3B). We confirmed that these observations were not cell type-specific by testing isogenic HEK293 and CHO cells with similar knockouts (Fig. 4). Interestingly, knockout of Ldlr in CHO cells produced a particularly pronounced reduction of binding and uptake of LDL compared with knockout of Galnt11 or Cosmc (Fig. 4, C and D). Because truncation of O-glycans by knockout of COSMC produced similar effects as knockout of GALNT11, we tested the effect of sialic acid capping alone by pretreatment of cells with neuraminidase to remove sialic acids on WT CHO and HepG2 cells and found similar reduction in binding and uptake of FITC-LDL (Fig. 5).
In summary, these data indicate that sialic acid residues on O-glycans in the LA linker regions directed by GalNAc-T11 enhance the capacity of LDLR for binding and uptake of LDL.

Recombinant LDLR ectodomains require O-glycans for efficient inhibition of LDL binding and uptake by WT CHO cells
We expressed truncated secreted constructs of human LDLR in CHO cells with and without knockout of Galnt11 and isolated them to apparent homogeneity (Fig. 6, A and B). The full secreted ectodomains of LDLR, sLDLR, expressed in WT CHO cells exhibited substantially better inhibition of FITC-LDL uptake by HepG2 cells compared with the same construct expressed in CHO cells without Galnt11 ( Fig. 6C and Fig. S3). Moreover, direct binding assays with LDL and sLDLR, as well

Site-specific O-glycosylation of LDLR and ligand interactions
as a further truncated construct LA1-4, showed that the constructs expressed in CHO without Galnt11 exhibited an ϳ5-fold lower affinity compared with the same constructs expressed in WT CHO cells (Table 1), emphasizing that loss of O-glycosylation by GalNAc-T11 results in an LDLR that binds LDL with much lower affinity.

VLDLR also require O-glycans for VLDL binding and uptake
VLDLR contains eight LA modules with five of the linker regions predicted to carry O-glycans, although we have only identified three of these yet with our shotgun glycoproteomics strategy (Fig. 1). However, we have found in preliminary studies that the two identified sites are dependent on GALNT11 in CHO and HEK293 cells. We therefore tested binding and uptake of FITC-VLDL by isogenic CHO cells with and without knockout of Cosmc (SC) or Galnt11 (1C8 clone) and found that loss of either gene resulted in substantial loss of binding and uptake of VLDL, similar to what was found for LDL (Fig. 7). We were unable to confirm cell-surface expression of VLDLR in mutant cells because of a lack of antibodies, and we predict that expression similar to LDLR is not affected by loss of GalNAc-T11-mediated glycosylation (Fig. 2).

Conservation of GalNAc-T11-directed O-glycans in LA linkers in Drosophila
We previously demonstrated that the Drosophila ortholog of GalNAc-T11, dGalNAc-T1 (dT1), or l(2)35Aa, is essential for viability (21,22). Drosophila members of the LDLR-related protein family consist of the lipophorin receptors (Lpr), which share a similar organization of domains including seven or eight LA modules (23). Lprs are essential for the efficient uptake of neutral lipids in Drosophila and include Lpr1 and Lpr2, of which Lpr2-E is the major receptor involved in the uptake by nurse cells and oocytes (24). We expressed a C-terminal HAtagged full coding construct of Lpr2-E in WT CHO cells with and without KO of Galnt11, as well as the combination of KO of Galnt11 and co-expression of dGalNAc-T1 (Fig. S4). The naturally shed ectodomain of Lpr2-E without the HA tag was

Site-specific O-glycosylation of LDLR and ligand interactions
enriched by Jacalin chromatography of glycoproteins with sialylated ST (NeuAc␣2-3Gal␤1-3GalNAc␣1-O-Ser/Thr) O-glycans followed by trypsin digestion and direct LC-MS/MS analysis without prior neuraminidase treatment. We obtained an overall sequence coverage for Lpr2-E of 45% with the sample from CHO WT and CHO Galnt11 KO and 35% from CHO Galnt11 KO with the co-expression of dGalNAc-T1. The linker region (LA2-3) with the consensus motif C 6 RSHTC 1 was identified as glycopeptide only in CHO cells with expression of human GalNAc-T11 (WT) or co-expression of Drosophila dGalNAc-T1, whereas in CHO Galnt11 KO, we only identified the corresponding unglycosylated peptide ( Fig. 8 and Data Set S3). These results suggest that the mammalian and Drosophila orthologs share the unique capabilities in glycosylating LDLRrelated protein linker regions.

Discussion
Protein O-glycosylation of the GalNAc-type serves diverse and highly specific roles in fine-tuning protein functions (25), and here we provide conclusive evidence for a novel role in modulating the binding properties of the large family of LDLRrelated proteins. Other types of protein O-glycosylation including O-Fuc, O-Glc, and O-GlcNAc modulate the binding properties and signaling of the Notch receptors, and these types of glycosylation serve important roles in development and diseases (26). GalNAc-type O-glycosylation is unique and orches-trated by a large family of isoenzymes that enables differential and perhaps dynamic regulation of single glycosites. The specific modification of the LA linkers in LDLR-related proteins by the GalNAc-T11 isoform presented here is perhaps the most exclusive function of a GalNAc-T isoenzyme identified to date. We show that GalNAc-T11 directed O-glycosylation of the LA linkers in the ligand-binding domains of LDLR and VLDLR markedly affect the binding affinity and uptake of their respective lipoprotein targets, and it is likely that this is a general function for all LDLR-related proteins.
Our results are in agreement with and extend early studies suggesting the existence of O-glycans in addition to the well characterized glycans in the juxtamembrane region of LDLR (27), and they may provide an explanation for the otherwise puzzling finding that a Monensin-resistant CHO line, MonR-31, with apparent loss of these glycans showed reduced binding and uptake of LDL (13,14,28). The MonR-31 cell line is unfortunately no longer available, but our results suggest that the lectin selection used to establish the cell line resulted in deletion of the GALNT11 gene. Studies over decades with lectin-resistant CHO mutants demonstrate that the O-glycans in the juxtamembrane region of LDLR are essential for stability of the receptor at the cell surface (12), but importantly O-glycosylation in the juxtamembrane "mucin-like" domain with high density of O-glycans is redundantly achieved by multiple GalNAc-T iso-

Site-specific O-glycosylation of LDLR and ligand interactions
forms and presumably not regulated (11). In contrast, the single O-glycosites in the LA modules are selectively controlled by the GalNAc-T11 isoform and therefore amenable to differential and perhaps dynamic site-specific regulation in cells and organs.
It is increasingly becoming apparent that site-specific regulation of selected O-glycosites by the large family of GalNAc-T isoenzymes are involved in fine-tuning protein function with important roles in health and diseases (25). The most illustrative example to date is the unique role of the GalNAc-T3 isoform in phosphate homeostasis, where O-glycosylation co-regulates proprotein processing of FGF23 in a complex interplay with proprotein convertases and the Golgi protein kinase Fam20C, within the short sequence element 176 RHTR2S, where O-glycosylation of Thr 178 directed by GalNAc-T3 inhibits processing and phosphorylation of Ser 181 by Fam20C inhibits O-glycosylation (29,30). Congenital deficiency in GALNT3 cause hyperphosphatemia and ectopic ossifications, whereas deficiency in Fam20C causes osteosclerotic dysplasia and hypophosphatemia (31,32), and GALNT3 is a GWAS candidate for low mineral bone density (33). Many of the other 20 GALNT genes have been identified as candidate genes for dispositions to common diseases (25,34), and we were recently able to provide extensive validation for the role of GALNT2 in Figure 6. Analysis of truncated LDLR constructs and inhibition of LDL uptake. A, graphic depiction of truncated LDLR constructs sLDLR and LA1-4 used. B, SDS-PAGE Coomassie and Western blotting analyses of purified sLDLR and LA1-4 using anti-LDLR (Abcam, Ab30532) or anti-His. sLDLR and LA1-4 were expressed in CHO cells and purified using Jacalin lectin chromatography or nickel-nitrilotriacetic acid, respectively, and further purified by ion exchange with a yield of ϳ30 g from 100 ml of growth medium. The presence of two molecular weight forms of sLDLR is likely due to N-terminal truncation as reported previously (57). Note minor differences in migration of sLDLR and LA1-4 derived from cells with and without Galnt11, presumably because of a loss of O-glycans at the three glycosites in the linker region (Thr 67 , Thr 108 , and Thr 147 ). Loading was based on BCA quantification of purified proteins, but the yield of sLDLR isolated from cells with knockout of Galnt11 appeared higher based on the Coomassie staining and WB. C, quantification of FITC-LDL uptake by HepG2 WT cells in the presence of sLDLR with and without O-glycans introduced by GalNAc-T11. 20 g/ml FITC-LDL was preincubated with different concentrations of sLDLR as indicated (based on the BCA quantification) and subsequently incubated with HepG2 WT cells for 1 h at 37°C. The values represent the means of triplicate determinations (n ϭ 3); error bars represent Ϯ S.D.

Site-specific O-glycosylation of LDLR and ligand interactions
dyslipidemia in man, primates, and rodents and identified several site-specifically regulated O-glycoprotein targets (19). The substrate specificities and functions of GalNAc-Ts have mainly been studied by in vitro enzyme assays using short peptides as acceptors in the past (35), and this approach has provided important insights into specific functions of many individual isoenzymes including GalNAc-T11. However, although Gal-NAc-T11 does function with a number of short peptide substrates (21,22,35), it only functions in vitro with the LA linker regions in the context of the entire folded LA-modules (11). This may suggest that GalNAc-T11 is unique among the Gal-NAc-T isoforms in recognizing folded domains as substrates.

Other glycosyltransferases initiating O-Fuc, O-Glc, and O-
GlcNAc recognize small folded domains such as epidermal growth factor-like and thrombospondin type 1 repeats (26,36), and these domains share a design with three conserved disulfide bridges with the LA modules (37,38). However, in all cases the acceptor sites are located in the folded domain, whereas the substrate site for GalNAc-T11 is located in the short linker between the folded domains. We only discovered the unique function of GalNAc-T11 through O-glycoproteomics with glycoengineered cell lines (11). Availability of isogenic cell lines with and without individual GALNT genes are important discovery platforms for the isoform-specific functions as shown here and in combination with our quantitative differential glycoproteomics approach enable unbiased discovery of nonredundant biological functions of the large family of GalNAc-T isoenzymes (17). We also used genetic dissection to demonstrate the functional importance of the structure of the O-glycans. Thus, knockout of COSMC in CHO and HepG2 cells, which results in truncation of the normal core1 O-glycans to the simple GalNAc residue, produced the same marked change in the function of LDLR, demonstrating the importance of the O-glycan structure and the terminal sialic acid (Figs. [3][4][5]. Cellular trafficking of LDLR-related proteins is complex and essential for their functions. These receptors are transported through the ER-Golgi secretory pathway bound to the receptor-associated protein RAP (LRPAP1 or ␣2-macroglobulin receptor-associated protein) that acts as a chaperone and presumably protect receptors from early binding to ligands until dissociation occurs at the lower pH in later Golgi compartments (39). Studies with Escherichia coli produced LA modules without glycans established that RAP binds to multiple LA modules through electrostatic interaction between three calcium-coordinating acidic residues in each LA and lysine residues (9,40,41). In LRP1 the most juxtamembrane cluster of LA modules LA3-6 that contains linker O-glycans ( Fig. 1 and Table S1) appears to be the most important (9,41). In a recent study of the complete LRP1 protein expressed in HEK293 cells, presumably with GalNAc-T11 directed O-glycosylation, strong nanomolar interaction with RAP was demonstrated (42), suggesting that O-glycosylation of the linkers in LA modules may not play a role in the interaction with RAP, which is also in line with our finding that surface expression and shedding of LDLR was unaffected by LA module O-glycosylation (Fig. 2, A and C). The co-crystal structure of RAP with the LA modules 3 and 4 in a complex shows that the acceptor Thr in the LA linker region is facing toward RAP at a distance of ϳ10 Å from the closest amino acid of RAP (9), suggesting that O-glycosylation may occur only after release from RAP.
The ligand binding of LDLR-related proteins resembles the interaction with RAP and depends on the conserved acidic residues in the LA modules (9). Many studies of the ligand binding properties of LDLR and VLDLR have been performed with E. coli produced truncated receptor constructs (43)(44)(45), which would exclude analysis of the O-glycans found in the LA linkers. Similarly, studies with LDLR expressed in insect cells are likely to exclude these O-glycans or at least completely lack sialic acids (9). However, the O-glycans are not essential for binding, but they induce a marked enhancement in extracellular binding and uptake of LDL and VLDL by LDLR and VLDLR in cell assays. This was clearly shown for LDLR in CHO cells, where deficiency in O-glycans produced intermediary LDL binding and uptake compared with complete LDLR deficiency and further supported by direct binding assays showing detectable but ϳ5-fold lower binding affinity of sLDLR to LDL in the absence of O-glycans (Fig. 4, C and D, and Table 1). How O-glycans exert these effects requires structural studies with appropriately glycosylated receptor constructs, but we envision that the sialylated O-glycans may participate directly in the ligand interaction by, for example, adding to the negative charges of the conserved aspartic acids in the LA module or perhaps more likely induce local conformational effects that organize consecutive LA modules in more favorable binding mode.
The implications of our findings for the many diseases and conditions associated with the diverse functions of LDLR-related proteins require further studies. Genetic deficiency in GALNT11 has not been identified so far in man, and it is predicted to be extremely rare if occurring (34), but GALNT11 has been identified as a GWAS candidate gene for chronic kidney decline (46). The GWAS signal for chronic kidney decline resides in intron one of the GALNT11 gene, suggesting that the genetic predisposition is a result of altered gene regulation, similar to what has been established for the role of GALNT2 in dyslipidemia (19,47). Our study did not directly address the functional role of O-glycans on other LDLR-related proteins including LRP2; however, GalNAc-T11 is highly expressed in proximal tubules of the kidney, and LRP2 is the major endocytic receptor responsible for reabsorbtion of proteins from the glomerular filtrate (21,48). Given the common structure of the ligand binding LA modules and conservation of LA linker O-glycan sites directed by GalNAc-T11, we propose that the O-glycans on all the LDLR-related proteins serve similar roles in modulating ligand binding. Moreover, it seems plausible that the identification of GALNT11 as a GWAS candidate for chronic kidney disease may relate to altered glycosylation and function of LRP2, and further studies into this are clearly warranted. Given the important roles of LDLR and VLDLR in lipoprotein metabolism and hypercholesterolemic conditions (49) and the dramatic effect GalNAc-T11 directed O-glycosylation has on these receptors, it may be surprising that GALNT11 has not been associated with cholesterol disorders. We are in progress with such studies, but we did provide evidence that the unique substrate specificity of GalNAc-T11 is at least partially conserved by the Drosophila ortholog dGalNAc-T1 also designated l(2)35Aa, which is essential for viability (21,22). In sum-

Site-specific O-glycosylation of LDLR and ligand interactions
mary, our study presents conclusive evidence for an important functional role of O-glycans in modulating the ligand binding properties of the large family of LDLR-related proteins, which provides a novel understanding of the function and potentially regulation of these receptors with wide implications for common diseases.

O-Glycoproteomic analysis of rat organs
Samples of ϳ100 mg of tissue from rat liver, brain, or kidney were homogenized in 300 l of lysis buffer containing 1% RapiGest (Waters) in 50 mM ammonium bicarbonate using an IKA Ultra Turbax blender at maximum speed for 20 s followed by 30 s of sonication using a Sonic Dismembrator (Fischer Scientific). Homogenates were boiled for 5 min, diluted with 50 mM ammonium bicarbonate to 0.1% RapiGest, reduced with 20 mM DTT in 50 mM ammonium bicarbonate at 60°C for 45 min, alkylated with 40 mM iodoacetamide in 50 mM ammonium bicarbonate at room temperature for 35 min, followed by addition of 20 mM DTT to terminate alkylation, and digestion with trypsin (50 g/sample) (Roche) at 37°C, ON. Digests were treated with concentrated 6 l TFA at 37°C for 30 min, cleared by centrifugation, and purified on C18 Sep-Pak (Waters). The digests were desialylated with 1 unit of neuraminidase (N3001, Sigma) in 50 mM sodium citrate (pH 5.0) at 37°C overnight followed by SepPak purification and lyophilization. The samples were dissolved in lectin-binding buffer (175 mM Tris, pH 7.4) and subjected to Jacalin-agarose (AL1153, Vector Laboratories) lectin weak affinity chromatography (LWAC) performed as described previously with PNA (19), with minor modifications. The glycopeptides were eluted with 3 ϫ 1-column volume 0.7 M galactose in lectin buffer (175 mM Tris, pH 7.4). Lectin weak affinity chromatography fractions were screened by LC-MS for glycopeptide content, and fractions enriched in glycopeptides were pooled and submitted to LC-MS analysis using EASY-nLC 1000 UHPLC (Thermo Scientific) interfaced via NanoSpray Flex ion source to an LTQ-Orbitrap Fusion Pro spectrometer (Thermo Scientific), as previously described (17). Data processing was performed using Proteome Discoverer 1.4 software (Thermo Scientific).

Expression and purification of secreted LDLR
A construct (LA1-4) containing residues 1-196 of the coding region of human LDLR was generated by PCR using a plasmid containing the entire coding region and cloned into pcDNA3.1/myc-His (C-terminal tags) (Invitrogen). The following primers were used: LA-1-F, 5Ј-CCCAAGCTTGCCAC-CATGGGGCCCTGGGGCTGGAA-3Ј; and LA-4-R, 5Ј-CCG-CTCGAGGGGGCTACTGTCCCCTTGGA-3Ј. The plasmid encoding LA1-4 and a bicistronic enhanced GFP (EGFP) plasmid encoding human full-length LDLR with a C-terminal V5 tag (53) were expressed in CHO WT and CHO Galnt11 KO cells (1C8 clone). The sLDLR was purified as follows: after 48 h of culture the medium (300 ml) was collected and dialyzed twice against 50 mM Tris (pH 7.4) at 4°C, diluted into 2ϫ binding buffer (350 mM Tris, pH 7.4), and applied to a 1-ml packed Jacalin-agarose column (Vector Laboratories). Following washing by 40 -50 column volumes (CV) of binding buffer O-glycoproteins were eluted with 0.8 M galactose in binding buffer and detected by Western blotting using anti-LDLR (ab30532, Abcam). The secreted His-tagged LA1-4 construct was purified on nickel affinity chromatography (Invitrogen) by mixing the culture media 3:1 (v/v) with 4ϫ binding buffer (200 mM Tris, pH 8.0, 1.2 M NaCl) and applied to 0.3 ml of packed nickelnitrilotriacetic acid-agarose (Invitrogen) pre-equilibrated in binding buffer (50 mM Tris, pH 8.0, 300 mM NaCl). The column was washed with binding buffer and eluted with binding buffer containing 250 mM imidazole. Fractions containing LA1-4 were analyzed by SDS-PAGE and pooled. Further purification of sLDLR and LA1-4 were achieved by anion-exchange MonoQ (5/50 GL, GE Healthcare) chromatography on ÄKTA FPLC interfaced by UNICORN 4.12 control software. Enriched pooled fractions were centrifuged at 14,000 rpm at 4°C for 30 min and loaded on the MonoQ pre-equilibrated in 25 mM Bis-Tris (pH 6.5) with 10 mM NaCl. The column was washed with 15 CV and eluted using a linear gradient of 10 mM-1 M NaCl. Fractions were analyzed by SDS-PAGE Coomassie and Western blotting with anti-LDLR (ab30532, Abcam) or anti-His antibody (A00186, GenScript). Purified proteins were quantified by BCA protein assay kit (Thermo Scientific).

SDS-PAGE Western blotting analysis
NuPAGE Novex Bis-Tris 4 -12% gels were used with transfer to nitrocellulose membranes for 90 min (0.45 m, Bio-Rad) 200 V. The blots were blocked in 5% skimmed milk in TBS-T

Site-specific O-glycosylation of LDLR and ligand interactions
and washed three times in TBS-T prior to overnight incubation at 4°C with rabbit polyclonal anti-LDLR antibody (ab30532, Abcam) or rabbit polyclonal anti-calnexin antibody (1:2,000) (Enzo Life Sciences). The blots were developed with ECL (Pierce catalog no. 32106, Thermo Scientific) after incubation with horseradish peroxidase-conjugated secondary antibody (Dako) for 1 h. Calnexin was used to normalize.

Metabolic labeling and immunoprecipitation
HepG2 WT and GALNT11 KO cells were cultured in methionine and cysteine-free RPMI 1640 (Gibco) media containing 2 mmol/liter L-glutamine for 30 min, before replacement of the same medium containing 0.1 mCi/ml Promix ([ 35 S]methionine/[ 35 S]cysteine) (PerkinElmer Life Sciences). For pulsechase, cells were labeled for 30 min and chased for 0 -6 h in complete RPMI 1640 medium containing 5 mM methionine and cysteine. For co-immunoprecipitation the LDL receptor was immunoprecipitated overnight at 4°C using a polyclonal anti-LDLR antibody (ab30532, Abcam) and protein G-Sepharose (Amersham Biosciences). The immunoprecipitate was analyzed by NuPAGE after incubation in LDS sample buffer (Invitrogen) for 10 min at 95°C. The bands were quantified using a Phos-phorImager (STORM 840, Molecular Dynamics) and Image-Quant software (Molecular Dynamics).

LDLR functional assays
The cells were cultured in 24-well plates and used at ϳ60% confluence. CHO cells were cultured with EX CELL CD CHO fusion medium (Sigma) with 2% glutamine. HepG2 and HEK293 cells were cultured in Dulbecco's modified Eagle's medium containing 10% FBS, and 1 day before FBS was substituted with 10% lipoprotein-deficient serum (LPDS) (Alfa Aesar). The cells were incubated for 4 h at 4°C for binding and 37°C for uptake with different concentrations of FITC-LDL (54) or DiI-LDL (Alfa Aesar), followed by washing twice in PBS with 1% BSA, fixation in 4% formaldehyde for 10 min, and washing twice with PBS with 1% BSA. FITC-LDL was analyzed by FACS in a LSR-II flow cytometer (Beckman Coulter) with 10,000 events acquired for data analysis, and the results were expressed as the mean fluorescence of activated gated cells, selected in a forward versus side scatter window. DiI-LDL was analyzed with cells mounted in ProLong Gold antifade reagent (4Ј,6Ј-diamino-2-phenylindole) (Invitrogen) using a Zeiss Axioskop 2 plus with an AxioCam MR3. Bit depth and pixel dimensions were 36 bits and 1,388 ϫ 1,040 pixels, respectively. Analysis of FITC-VLDL was performed as described for LDL (54). All measurements have been performed in triplicate. Competition assays with secreted LDLR constructs were performed with 20 g/ml FITC-LDL or DiI-LDL preincubated with the indicated concentrations of secreted LDLR constructs at 37°C before addition to cells for 4 h.

Neuraminidase pretreatment
The cells for binding and uptake assays were incubated with Clostridium perfringens neuraminidase (Sigma, N3001) diluted in PBS to a final concentration of 0.1 unit/ml at 37°C for 4 h, followed by washing twice with in medium containing LPDS. Controls for neuraminidase included staining with biotinylated PNA (B-1073) and biotinylated Maackia amurensis lectin II (B-1265) lectins followed by Alexa Fluor 488 -conjugated streptavidin (Life Technologies, S32354).

Direct LDL-binding assays
A modified ELISA binding assay with purified sLDLR and LA1-4 coated in 96-well plates in 10 mM Tris-HCl (pH 7.4) and 50 mM NaCl, 2 mM CaCl 2 (buffer A) was used (55). The plates were blocked with 5% BSA and incubated with freshly purified human LDL in buffer A for 2 h at room temperature, before washing four times with buffer A and once with buffer A containing 0.1% Tween 20, followed by incubation with goat polyclonal anti-apolipoprotein B (Abcam, UK) for 1 h, peroxidaseconjugated mouse anti-goat IgG (Thermo Scientific) for 1 h, and development with 2,2Ј-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid) substrate (Sigma-Aldrich) and photometric quantification at 405 nm. All data were corrected for unspecific binding and relativized to maximum, and EC 50 values were extracted from binding curves after fitting the data to a fiveparameter logistic equation (SigmaPlot 13.0; Systat Software Inc.). In control experiments, mouse monoclonal anti-LDL receptor (PROGEN) and horse anti-mouse horseradish peroxidase-linked secondary antibody (Cell Signaling Technology) were used to detect the amount of LDLR fragments bound to plate and calculate the surface coverage percentage. Experiments were carried out at two surface subsaturating coverage percentages and [L] Ͼ Ͼ [R] conditions, obtaining the same binding values.

Recombinant expression of Drosophila pgant35A and Lpr2-E
The full coding region of Drosophila melanogaster pgant35A was derived from pUAST-pgant35A (56) and subcloned into pcDNA3 (Invitrogen) to generate pcDNA3-pgant35A. The ORF of D. melanogaster lpr2E with a C-terminal 3ϫ HA tag in pUAST-lpr2E was kindly provided by Joaquim Culi (Sevilla, Spain; Ref. 23) and used to generate pcDNA3-lpr2E. The pcDNA3-pgant35A construct was expressed in CHO Galnt11 KO cells (1C8 clone) by electroporation and G418 selection, and stable clones were identified by PCR with primers PCDNAFOR and dT1-KPS (5Ј-CGATTCGCCAGTGGTAAT-GCTGGC-3Ј). The pcDNA3-lpr2E construct was stably expressed in CHO WT, CHO Galnt11 KO, and CHO Galnt11 KO with expression of pgant35A, using electroporation and Zeocin selection. Stable clones were screened by immunocytochemistry and Western blotting analysis using a monoclonal anti-HA antibody (Santa Cruz).

Analysis of recombinant shed Drosophila Lpr2-E
Culture medium from CHO cells stably expressing full coding C-terminal HA-tagged Drosophila Lpr2-E construct was

Site-specific O-glycosylation of LDLR and ligand interactions
diluted twice in 175 mM Tris (pH 7.5) and loaded twice onto a 0.8-ml Jacalin lectin column, followed by 20ϫ CV washing in the same buffer and 2ϫ CV washing with 50 mM ammonium bicarbonate added. Bound O-glycoproteins were eluted by heating the lectin slurry in 0.1% RapiGest. The eluate was heated for 10 min at 80°C, reduced with 20 mM DTT at 60°C for 45 min, alkylated with 40 mM iodoacetamide in 50 mM ammonium bicarbonate at room temperature for 35 min followed by addition of 20 mM DTT to terminate alkylation, and finally digested with trypsin (1 g/sample) (Roche) at 37°C overnight. Digests were treated with concentrated TFA at 37°C for 30 min, cleared by centrifugation, purified on C18 Sep-Pak (Waters), and lyophilized. Each sample was dissolved in 0.1% TFA and submitted to LC-MS and high-energy collision dissociation/electron transfer dissociation-MS/MS.