N-glycosylation of mannose receptor (CD206) regulates glycan binding by C-type lectin domains

The macrophage mannose receptor (MR, CD206) is a transmembrane endocytic lectin receptor, expressed in selected immune and endothelial cells, and is involved in immunity and maintaining homeostasis. Eight of the ten extracellular domains of the MR are C-type lectin domains (CTLDs) which mediate the binding of mannose, fucose, and GlcNAc in a calcium-dependent manner. Previous studies indicated that self-glycosylation of MR regulates its glycan binding. To further explore this structure–function relationship, we studied herein a recombinant version of mouse MR CTLD4-7 fused to human Fc-portion of IgG (MR-Fc). The construct was expressed in different glycosylation-mutant cell lines to study the influence of differential glycosylation on receptor glycan-binding properties. We conducted site-specific N- and O-glycosylation analysis and glycosylation site characterization using mass spectrometry by which several novel O-glycosylation sites were identified in mouse MR and confirmed in human full-length MR. This information guided experiments evaluating the receptor functionality by glycan microarray analysis in combination with glycan-modifying enzymes. Treatment of active MR-Fc with combinations of exoglycosidases, including neuraminidase and galactosidases, resulted in the loss of trans-binding (binding of MR CTLDs to non-MR glycans), due to unmasking of terminal, nonreducing GlcNAc in N-glycans of the MR CTLDs. Regalactosylation of N-glycans rescues mannose binding by MR-Fc. Our results indicate that glycans within the MR CTLDs act as a regulatory switch by masking and unmasking self-ligands, including terminal, nonreducing GlcNAc in N-glycans, which could control MR activity in a tissue- and cell-specific manner or which potentially affect bacterial pathogenesis in an immunomodulatory fashion.

The macrophage mannose receptor (MR, CD206) is a transmembrane endocytic lectin receptor, expressed in selected immune and endothelial cells, and is involved in immunity and maintaining homeostasis. Eight of the ten extracellular domains of the MR are C-type lectin domains (CTLDs) which mediate the binding of mannose, fucose, and GlcNAc in a calcium-dependent manner. Previous studies indicated that self-glycosylation of MR regulates its glycan binding. To further explore this structure-function relationship, we studied herein a recombinant version of mouse MR CTLD4-7 fused to human Fc-portion of IgG (MR-Fc). The construct was expressed in different glycosylation-mutant cell lines to study the influence of differential glycosylation on receptor glycanbinding properties. We conducted site-specific N-and O-glycosylation analysis and glycosylation site characterization using mass spectrometry by which several novel O-glycosylation sites were identified in mouse MR and confirmed in human full-length MR. This information guided experiments evaluating the receptor functionality by glycan microarray analysis in combination with glycan-modifying enzymes. Treatment of active MR-Fc with combinations of exoglycosidases, including neuraminidase and galactosidases, resulted in the loss of trans-binding (binding of MR CTLDs to non-MR glycans), due to unmasking of terminal, nonreducing GlcNAc in N-glycans of the MR CTLDs. Regalactosylation of N-glycans rescues mannose binding by MR-Fc. Our results indicate that glycans within the MR CTLDs act as a regulatory switch by masking and unmasking self-ligands, including terminal, nonreducing GlcNAc in N-glycans, which could control MR activity in a tissue-and cell-specific manner or which potentially affect bacterial pathogenesis in an immunomodulatory fashion.
The macrophage mannose receptor (MR, CD206) is a type 1 transmembrane endocytic lectin receptor and plays an essential role in immunity and homeostasis (1). It is largely expressed in macrophages, dendritic cells, as well as in some endothelial cells, where it constantly shuttles between the plasma membrane and the endosomal compartment to endocytose ligands in a clathrin-dependent manner (2).
The MR belongs to the MR family, including M-type phospholipase A2 receptor, DEC-205/gp200-MR6, and Endo180/uPARAP. The extracellular portion is composed of an N-terminal cysteine-rich domain, a fibronectin-type II domain, and 8 C-type lectin domains (CTLDs) (3). The MR cysteine-rich domain binds sulfated glycans (4)(5)(6) and the fibronectin-type II domain is known to bind collagen (7). Via the CTLDs, the receptor has been reported to bind mannose (Man), GlcNAc, fucose (Fuc), and glucose in a calciumdependent manner (8)(9)(10)(11)(12). While the MR contains eight CTLDs, only CTLD4 has carbohydrate affinity when expressed as a single recombinant unit (8,10). When expressed together with CTLD5, the CTLD4 binds mannose with the same affinity as all eight CTLDs together. For the recognition of multivalent ligands, such as yeast mannan, CTLD4-7/8 is required, indicating that the other CTLDs are important structural components (8,10). Endogenous ligands of the MR CTLDs are mainly paucimannose and oligomannose N-glycans, which are present across a variety of different tissue types in control and cancer tissues (13).
The MR is a glycoprotein carrying seven potential N-glycosylation sites in mouse and eight in humans, and based on predictions using NetOGlyc (14), it may also contain several O-glycosylation sites, with four of them identified by sequence analysis (15). Specific MR glycosylation features have been reported to be necessary for ligand binding via the CTLDs (16) and previous studies have shown differential MR glycan signatures across different tissues (17), indicating tissue-specific functional regulation. Oddly, however, none of the N-glycosylation sites or predicted O-glycosylation sites are in CTLD4 or CTLD5, and thus, the glycosylation influence might be indirect. We recently reported that resident and elicited murine macrophages differ in sialylation and other features of N-glycosylation (18), and thus changes in glycosylation of the MR accompanying cellular differentiation could influence its function. Thus, it can be hypothesized that the biological function of the MR is cell-and tissue-dependent and could vary depending on environmental changes that influence glycosylation of the MRbearing cells, such as inflammation or exposure to microbes (19). In order to enable the analysis of structure-function relationships of the MR, and learn more about its specific role in infection and immunity, we performed a glycoproteomic analysis of the MR in different glycosylation-mutant cell lines. During the process, we found that the contribution of glycosylation towards MR activity is more complex than previously described (16).
Here, we used a recombinant fusion protein composed of the MR CTLD4-7 and the IgG Fc portion (MR-Fc) to investigate the role of MR glycosylation in receptor-glycan binding. The MR-Fc was expressed in the Chinese hamster ovary (CHO) mutant cell lines Lec2 and PIR-P3, next to CHO WT and human embryonic kidney (HEK)293 WT cell lines. Differential MR-Fc binding to glycan microarrays was correlated with their site-specific N-and O-glycosylation. These studies revealed the critical glycan features of the MR-Fc that cause its loss or gain of external or trans-binding of glycans, that is, the binding of MR CTLDs to non-MR glycans. Consequently, enzymatic deglycosylation and reglycosylation allowed us to mimic this functional switch. Our data demonstrate that the masking (gain of trans-binding) or unmasking (loss of trans-binding) of nonreducing terminal GlcNAc residues in N-glycans regulate the glycan binding of the MR CTLDs.

MR-Fc from CHO Lec2 and CHO PIR-P3 cell lines do not bind to glycan microarrays
To explore the influence of glycosylation on MR binding to glycan ligands, we used the mouse CTLD4-7, fused to the Fc-portion of human IgG (MR-Fc). MR-Fc is a well-established probe to study the glycan ligands of the CTLDs (13). We expressed the MR-Fc in CHO cell mutants with different glycan features (Fig. 1A). CHO Lec2 cells (expressing MR-Fc Lec2 ) are characterized by a defect in their CMP-sialic acid transport (20), resulting in an overall reduction in glycoconjugate sialylation. The MR-Fc was also expressed in the CHO-PIR-P3 cell line (MR-Fc PIR-P3 ), which has a "PIR" mutation (resistance to glycoprotein processing inhibitors) and a defect in N-acetylglucosaminyltransferase I activity, resulting in mainly truncated paucimannose-type Man 3 GlcNAc 2 Detection of all microarrays was performed using goat anti-human IgG Alexa-Fluor 488-conjugated at 5 μg/ml. Standard deviation is shown. Glycan linkages are indicated based on the Oxford system (38). Green circle = mannose; blue square = N-acetylglucosamine; Glycans on the array are listed in Tables S3 and S4. BSA, bovine serum albumin; CHO, Chinese hamster ovary; CTLD, C-type lectin domain; HEK, human embryonic kidney; Man, mannose; MR, mannose receptor; RFU, relative fluorescence units. N-glycans (21). As a reference, MR-Fc was also expressed in CHO WT (MR-Fc CHO-WT ) as well as HEK293T WT cells (MR-Fc HEK-WT ).
MR-Fc from all four cell lines were screened for their overall glycan binding using the Consortium for Functional Glycomics (CFG) glycan microarray (version 5.3) (Fig. S1 and Table S1). While the MR-Fc CHO-WT binds to mainly paucimannose-and oligomannose-type glycans and glycan fragments on the array (Fig. S1A), MR-Fc Lec2 and MR-Fc PIR-P3 did not bind to the glycan array (Fig. S1, C and D) at the same concentrations. As the MR-Fc CHO-WT mainly bound to mannose-terminating glycans, we also used it to probe our newly developed oligomannose array ( Fig. 1B and Table S2). MR-Fc CHO-WT bound to paucimannose and oligomannose N-glycans or truncated fragments from Man 3 to Man 8 on the oligomannose array. Consistent with the data from the CFG array, the MR-Fc CHO-Lec2 and MR-Fc PIR-P3 did not bind to the oligomannose array (Fig. 1C).
As the MR-Fc CHO-WT and MR-Fc HEK-WT exhibited identical binding patterns to the oligomannose array ( Fig. 1, B and D) and CFG glycan array (Fig. S1, A and B), making them both suitable probes to study MR-Fc glycan binding, we prepared larger quantities of MR-Fc HEK-WT to further explore the sites of glycosylation and roles of terminal glycan modifications. We developed a smaller subarray (Man-BSA array) to evaluate MR-Fc glycan-binding ability under different experimental conditions. The Man-bovine serum albumin (BSA) array contains different mannose conjugates, including Man 3 -BSA, Man-BSA, Man 3 -PAA, and their corresponding controls BSA, GlcNAc-BSA, and Glu-PAA (Table S3). The array was probed with MR-Fc HEK-WT , which bound to the mannose conjugates in a calcium-dependent manner as expected (Table S3).

Desialylation of WT MR-Fc does not affect glycan binding
Proteins expressed in CHO Lec2 cells are deficient in sialic acid due to the deficient CMP-sialic acid transport. Thus, to simulate the MR-Fc Lec2 glycan phenotype, MR-Fc HEK-WT was desialylated using neuraminidase. Desialylation was confirmed by lectin blots (Fig. S2A). MR-Fc HEK-WT and the neuraminidase-treated MR-Fc HEK-WT bound in a similar fashion to the Man-BSA array without loss of activity ( Fig. 1E and Table S4). Next, both forms of MR-Fc HEK-WT were enriched using mannan-agarose beads, showing no difference in mannan-agarose binding upon desialylation (Fig. S3). Thus, these results demonstrate that removal of sialic acid from a previously active MR-Fc does not result in loss of receptor trans-binding to mannose. This strongly indicates that the lack of glycan binding by MR-Fc Lec2 is independent of its sialylation status.
N-glycan removal rescues glycan binding of MR-Fc Lec2 and MR-Fc PIR-P3 All four forms of MR-Fc were de-N-glycosylated using PNGase F (Fig. S2) to investigate the importance of N-glycans for MR-Fc mannose binding. For both MR-Fc CHO-WT and MR-Fc HEK-WT , removal of N-glycans did not affect their positive binding to the Man-BSA array ( Fig. 1F and Table S4). Unexpectedly, removal of the N-glycans from the MR-Fc Lec2 and MR-Fc PIR-P3 rescued their ability to bind to the Man-BSA array, in which case both treated receptors exhibited the same binding pattern as the WT MR-Fc ( Fig. 1G and Table S4). This result indicates that N-glycosylation is not directly necessary for glycan binding by the MR-Fc but that when present, specific features of the N-glycans can inhibit the receptor functionality.

N-glycosylation site characterization of MR-Fc reveals incomplete antennae galactosylation in MR-Fc Lec2
The MR-Fc has five predicted N-glycosylation sites: two in the linker region connecting CTLD5 and CTLD6 (Asn926 and Asn930, numbering based on complete MR sequence), two in CTLD7 (Asn1159 and Asn1204), and one N-glycan in the IgG Fc-portion. A detailed site-specific glycopeptide characterization using LC-MS analysis was performed to better understand the involvement of MR glycosylation in modulating its glycan binding (Fig. 2, Tables S5 and S6). The structural assignments were largely based on glycan composition.
Both glycosylation sites Asn926 and Asn930 are present on the same tryptic glycopeptide 925 HNSSINATAM(ox) PTTPTTPGGC 944 (underlined Asn indicates glycosylation site). Since this region also contains O-glycosylation sites, as described below, the microheterogeneity was too complex to detect and quantify the glycopeptides under the chosen conditions. In an attempt to reduce glycopeptide complexity, the protein was treated with neuraminidase and O-glycosidase prior to in-gel trypsin treatment and glycopeptide analysis. This allowed N-glycopeptide analysis for all MR-Fc expressed in CHO cells, as the O-glycosidase efficiently removes core 1 O-glycans expressed by these cells (Fig. 2, A-C). For MR-Fc HEK-WT , N-glycosylation site characterization was not achieved, as HEK cells express more complex O-glycans, such as the core 2 O-glycans, which could not be released by O-glycosidase, and they also exhibit a higher degree of N-glycan microheterogeneity in general.

Terminal GlcNAc-containing N-glycans result in the loss of MR-Fc mannose binding and can be restored by regalactosylation
Interestingly, across all MR N-glycosylation sites within MR-Fc Lec2 , there is also a large portion of nongalactosylated or mono-galactosylated N-glycans present, in particular at the Asn1204 site (Fig. 2C), resulting in expression of MR-Fc Lec2 glycans with terminal, nonreducing GlcNAc. This was unexpected, as Lec2 CHO cells are only deficient in the CMP-sialic acid transport. However, GlcNAc is a reported ligand of the MR CTLDs (8,9,11,12), and we confirmed that as we observed MR-Fc HEK WT bind to the chitin-like oligosaccharide GlcNAc 3 -AEAB on the Schistosoma mansoni glycan array (13). However, binding to GlcNAc was not detected in any of our defined endogenous glycan microarrays or glycoprotein pull-down studies from mammalian cells (13). Furthermore, MR-Fc does not bind to the GlcNAc-BSA printed on the Man-BSA array (Table S3). Nevertheless, in order to more deeply evaluate if the terminal GlcNAc could present a potential self-ligand to the MR, thereby diminishing MR-Fc Lec2 mannose trans-binding, we modified the MR-Fc HEK-WT and MR-Fc CHO-WT by desialylation and degalactosylation with neuraminidase and β1,4-galactosidase, respectively, to simulate the MR-Fc LEc2 glycan phenotype. Both WT MR-Fcs were unable to bind to the Man-BSA array after removal of sialic acid and β1,4-galactose; as a control, treatment with neuraminidase and a β1,3-galactosidase had no effect ( Fig. 3A and Table S7). Thus, removal of terminal sialic acids and β1,4-galactose residues results in the inactivation of MR-Fc. In addition, incubation of MR-Fc HEK-WT with 100 mM mannose or GlcNAc, but not galactose, completely inhibited mannose binding ( Fig. S5 and Table S8). We further found that the MR-Fc HEK-WT can bind to GlcNAc-agarose beads and be specifically eluted with 100 mM GlcNAc or mannose, but not with galactose ( Fig. S6), thus validating that GlcNAc is a ligand of the MR-Fc in this experimental approach. The N-glycopeptide analysis of MR-Fc HEK WT also indicated the presence of undergalactosylated antennae (Fig. 2), which would result in a portion of inactive receptor.
Enzymatic removal of sialic acid and galactose was also performed on the commercial full-length recombinant human MR to exclude the possibility that the observed phenomenon is an artifact of the artificial presentation of the MR-Fc. Since the recombinant full-length MR was produced in murine cells, the N-glycans contain α-galactose; thus, next to treatment with β1,4-galactosidase, we did an additional treatment with α-galactosidase. These combined treatments resulted in reduced binding to the Man-BSA array ( Fig. 3B and Table S7). Our results indicate that exposure of terminal, nonreducing GlcNAc residues diminishes the activity of the full-length MR, consistent with the results seen in the MR-Fc constructs.
Since degalactosylation can abolish MR and MR-Fc glycan binding, which would be predicted to expose GlcNAc residues, we investigated whether regalactosylation could rescue mannose binding of MR-Fc. MR-Fc HEK-WT was first desialylated and degalactosylated, followed by protein A purification to remove residual glycosidases. This treatment resulted in loss of function, as demonstrated by reduced binding to the Man-BSA array (Fig. 3C). PNGase F treatment of the degalactosylated inactive degal-MR-Fc HEK-WT rescued mannose binding and restored identical binding to the Man-BSA array, as observed for the PNGase F-treated active MR-Fc HEK-WT ( Fig. 3C and Table S7). Ultimately, regalactosylation of the inactive degal-MR-Fc HEK-WT using β1,4-galactosyltransferase restored its binding to the Man-BSA array (Fig. 3D). The effectiveness of regalactosylation was also confirmed by lectin blots and binding of Ricinus communis agglutinin-I (Fig. S7). Overall, these results demonstrate that unmasked terminal GlcNAc on MR-Fc inactivates the receptor trans-binding to mannose and that regalactosylation, or the removal of all N-glycans, can restore mannose binding (summarized in Fig. 4).

MR contains O-glycans in the linker region between the CTLDs
The site-specific glycopeptide analysis of MR-Fc, containing the murine CTLD4-7, revealed the presence of O-glycosylation in the receptor portion. We further analyzed the full-length human full-length MR to evaluate the conservation of these O-glycosylation sites between human and mouse MR. Up to seven occupied O-glycosylation sites were identified in MR-Fc and up to 12 in the full-length human MR (Fig. 5; representative spectra of each site are provided in Figs. S8-S17). Interestingly, all the identified sites were located in the linker region between the individual CTLDs.
The linker region of CTLD3-4 (His627-Lys654) contains up to four occupied O-glycosylation sites at Thr633, Thr639, Thr640, and Thr641 in the full-length human MR. In both murine MR-Fc and human full-length MR, Thr790 was found to be O-glycosylated, located within the linker region CTLD4-5 (Ile779-Ile806) (Fig. 5). The region between CTLD5 and CTLD6 (Arg924-Phe951 -human, Arg924-Leu950 -murine) was the most complex one, since it also contains two N-glycosylation sites. The O-glycopeptides were analyzed after N-glycan removal, revealing Thr932 and Thr936 to be O-glycosylated in MR-Fc. Another site within the region of Ser927-Thr940 was found in MR-Fc Lec2 only. The same linker region in human full-length MR featured up to two occupied glycosylation sites at Thr933 and Thr937. The latter one corresponds to Thr936 in mouse MR-Fc. Between CTLD6 and CTLD7 (Thr1081-Lys1101human, Thr1080-Thr1100murine), two glycosylation sites were identified within the region Ser1089-Thr1093 in MR-Fc; the same region where Thr1093 was identified in the human full-length MR. Another conserved site at Thr1220 (murine)/Thr1221 (human) was found in the region between CTLD7 and CTLD8 (Arg1214-Pro1240human, Lys1213-Pro1239 -murine). The spacer between CTLD8 and the transmembrane domain of MR was present only in the human full-length MR and it contained up to three occupied O-glycosylation sites at Thr1366, Thr1371, and Thr1372 (Fig. 5).
In terms of the O-glycan structural features, all CHO MR-Fc carries core 1 O-glycans or the Tn antigen, and MR-Fc HEK-WT additionally carries core 2 O-glycans. The relative O-glycan distributions of MR-Fc CHO-WT and MR-Fc PIR-P3 are highly similar, with HexNAc 1 Hex 1 NeuAc 1 as the main structure, while MR-Fc Lec2 contains the same glycans without sialic acid (Fig. S19). In contrast, MR-Fc HEK-WT contains nearly equal amounts of both sialylated and neutral O-glycans (Fig. S18).

Discussion
Here, we identified the role of MR glycosylation in regulating the glycan-binding ability of its CTLDs. A combination of glyco-enzyme treatments and site-specific glycosylation site characterization by LC-MS revealed the importance of masking terminal, nonreducing GlcNAc residues to maintain MR-Fc CTLD glycan trans-binding (Fig. 3D). The MR-Fc expressed in either Lec2 or PIR-P3 CHO cells was inactive, but the activity was restored after N-glycan removal, demonstrating the involvement of N-glycosylation in mannose binding. The active receptor loses its ability to bind mannose when being treated with neuraminidase and galactosidase, which results in N-glycans with terminal GlcNAc. Additionally, the expression of MR-Fc in Lec2 CHO cells was characterized by partially under-galactosylated N-glycan antennae, exposing terminal GlcNAc, and resulting in an inactive MR-Fc.
In return, the mannose-binding ability can be regained by masking terminal GlcNAc on N-glycan antennae with galactose or by complete N-glycan removal of the inactive receptor. Mannose-terminating N-glycans are a feature of MR-Fc expressed in PIR-P3 CHO cells, which showed no glycanbinding ability. Our results suggest that terminal mannose and GlcNAc present self-ligands to the MR CTLDs, potentially resulting in self-glycan binding or oligomerization and thus, inability for trans-glycan binding.
Our results are overall in agreement with a study by Su et al., (16) although our additional experiments lead to different conclusions. This prior study investigated the glycan-binding ability of MR expressed in CHO Lec2 (defect in CMP-sialic acid transport) and CHO Lec1 (mainly Man 5-GlcNAc 2 N-glycans) cells compared to WT CHO cells (16). In agreement with our results, it was found that Lec1 and Lec2 CHO cells expressing surface-bound MR were unable to bind mannose ligands. In addition, the ability of the MR in those cells to endocytose was hardly affected, and binding of sulfated galactose through the cysteine-rich domain was reduced but not abolished. Accordingly, the soluble-secreted full-length receptor expressed in Lec1 and Lec2 CHO cells also did not bind mannose but could bind sulfated galactose. Released N-glycan analysis of the full-length MR expressed in Lec2 CHO cells revealed Hex 5 HexNAc 4 Fuc 1 as the major N-glycan. These results led the authors to conclude that sialylation is the critical glycosylation feature that regulates the activity of MR (16).
However, in our study, we used additional approaches employing exoglycosidase treatment, such as neuraminidase and galactosidase, as well as complete N-glycan removal by PNGase F on the recombinant full-length MR and MR-Fc, to identify MR regulation by unmasking or masking of terminal GlcNAc-containing N-glycan antennae. We characterized the site-specific MR-Fc glycosylation and found that two out of four MR-Fc N-glycosylation sites carry mainly Hex 3 Hex-NAc 4 Fuc 1 , a biantennary N-glycan with unmasked terminal GlcNAc, while the other two sites carry fully galactosylated N-glycans. The presence of Hex 5 HexNAc 4 Fuc 1 as the major MR N-glycan in the previous study by Su et al. (16) is most likely reflecting an average of the overall full-length MR glycosylation. Our results clearly highlight the importance of performing site-specific glycopeptide analyses as a complementary approach for the analysis of released N-glycans.
In further support of our findings is the delayed MR activation during biosynthesis throughout the endoplasmic reticulum (ER)-Golgi compartment in monocyte-derived macrophages (22). MR activation was correlated with the development of Endo H resistance, because of N-glycan maturation throughout the ER and Golgi compartments. Thus, MR activation must take place after leaving the ER and is independent of O-glycosylation and complete maturation in the trans-Golgi, where sialylation is predicted to occur. In addition, treatment with brefeldin A, which inhibits transport to the trans-Golgi, needed for glycan maturation, resulted in reduced but not abolished MR activity. Consequently, it was concluded that activation of the MR takes place in the early Golgi (22), which correlates with the addition of galactose to hybrid and complex N-glycans.
GlcNAc has been reported as a lower affinity ligand of the MR CTLDs (8,9,11,12), since it contains a pair of equatorial 3-and 4-OH groups in the same stereochemistry as mannose (23). While recombinantly expressed CTLD1-3 or CTLD6-8 do not bind GlcNAc, the CTLD4-5 binds to GlcNAc and also to mannose (8). Recently, we investigated endogenous MR-Fc ligands using glycan microarrays, and no binding was observed to GlcNAc-terminating structures representing glycans or glycan fragments of the mammalian glycome. However, we also observed that the MR-Fc bound to GlcNAc 3 -AEAB on the S. mansoni glycan microarray (13) as well as to high-density GlcNAc-agarose beads (Fig. S6), indicating that these low-affinity interactions are enhanced by multimeric or multivalent presentation.
MR has been reported to aggregate in the presence of Ca 2+ , which can be reduced by treating the receptor with Endo H and/or adding α-methylmannose (24). In that report, sedimentation velocity experiments suggest that if the receptor binds another glycan on the receptor, it might be most likely due to the binding of another receptor molecule (oligomerization) but not due to internal receptor binding and bending of the structure. Similarly, in the presence of Ca 2+ , oligomerization of the receptor was detected, which is not the case in the presence of EDTA and added mannose using gel filtration chromatography (25). It can be speculated that part of the receptor in the previous study carried oligomannose N-glycans, resulting in receptor oligomerization. Also, the presence of EDTA or mannose did not change the migration of MR in gel filtration, whereas migration was changed upon change in pH, suggesting a conformational change (25). All of these studies indicate that the presentation of receptor self-ligands results in dimerization or oligomerization rather than selfbinding; however, this needs to be further investigated.
Differential amounts of monomeric and dimeric MR have been extracted from a rat liver and lung (26). While the monomeric form is specific to mannose binding, the dimeric form was needed for stable receptor binding with sulfated GalNAc via the cysteine-rich domain (26). It can be speculated that the presentation of receptor self-ligands increases dimerization, which leads to differential glycan recognition and functional properties of the receptor in a tissue-specific manner.
We confirmed seven MR N-glycosylation sites that are conserved between humans and mice. In addition, seven O-glycosylation sites were identified in MR-Fc and 12 in the human full-length MR, all localized in the linker regions between the CTLDs. An early sequencing study of the human MR revealed the presence of O-glycosylation in the linker region between CTLD3-4, corresponding to position Thr639/ Thr640/Thr641 (15). We also found up to four O-glycosylation sites in this linker region, confirming Thr639/Thr640/ Thr641. Another O-glycosylation site previously reported was Thr1093 in the linker region between CTLD6 and CTLD7, which was also confirmed in our mouse and human versions by mass spectrometric analysis.
Hydrodynamic analyses in combination with proteolysis experiments revealed some further structural information about MR (24). The linker regions on both sites of CTLD3 and CTLD6 are flexible and exposed to proteases. In contrast, close contact was reported between CTLD1-2, CTLD4-5, and CTLD7-8 in proteolytic experiments (24). The tight link between CTLD4-5 has been also confirmed by a separate proteolysis study using CTLD4-7 (10). Interestingly, these results correlate with our O-glycosylation site identification. The linker in between tightly connected CTLDs carries either no identified O-glycosylation site (CTLD1-2) or a single conserved O-glycosylation site as found between CTLD4-5 (Thr790) and CTLD7-8 (Thr1220 murine/1221 human). The linker regions that were found to be more exposed to protease tend to carry more O-glycosylation sites.
Overall, a very rigid structure of the receptor was found in modeling studies in combination with experimental studies, which indicated a mainly extended structure of the receptor (24) and it was speculated that larger flexibility in some of the linker region might be favorable for adjusting to binding a larger variety of glycans. However, EM-based studies were used to elucidate the three-dimensional structure of the receptor, indicating that it can be present in the bend and extended conformation, which is pH-dependent (25,27). Considering that the receptor is shuttling between the plasma membrane and endosomal compartment where the ligand is released, this conformational change could be a regulatory mechanism. In agreement with the previous study, the threedimensional structure also revealed tightly packed CTLD4-5 and CTLD7-8 (25).
Our study revealed that the function of the MR CTLDs can be modulated by their own glycosylation. Differential MR glycosylation has been reported across a variety of tissues (17) and it can be speculated that this might be a regulatory mechanism to control the activity of MR CTLDs in a tissue-or cell-specific manner.
Similarly, sialic acid-binding immunoglobulin-like lectins (Siglecs) bind sialylated selfglycans or glycans on neighboring cell surface proteins, which can inhibit trans-binding of other cells, and enzymatic desialylation/resialylation could regulate Siglec trans-binding (28). On the contrary, certain pathogens may use this regulatory switch of unmasking MR CTLD self-ligands on the receptor and thus, inactivating receptor trans-binding to their advantage to escape from the immune system. A variety of bacteria secrete enzymes, including β1,4-galactosidases, in order to hydrolase glycan epitopes, which can have immunomodulatory effects for pathogenesis (29,30). Our data sheds new light on the structure-function relationship of MR glycosylation, and further studies will be needed to determine to what extent this switch is implemented in biological processes.

MR-Fc expression and purification
HEK293 WT, CHO WT, CHO Lec2, and CHO PIR-P3 cells were transfected with MR-Fc DNA, comprising the murine CTLD4-7, fused to the Fc-portion of human IgG (31) (kind gift from L. Martinez-Pomares). CHO PIR-P3 cells were kindly provided by Dr Mark Lehrman (UT Southwestern Medical Center) (21) and all other lines were obtained from ATCC. Transfection was performed with Lipofectamine LTX with Plus Reagent (Thermo Fisher Scientific), according to the manufacturer's guidelines. The cells were incubated with Lipofectamine complex at 37 C in a CO 2 incubator for 24 h, followed by replacing the medium. HEK293 WT cells were transfected with MR-Fc and were cultured in RPMI culture media with 10% heat-inactivated fetal bovine serum (HI FBS), 1% penicillin-streptomycin (PS), and 10 mM Hepes containing nonessential amino acids; CHO WT cells were grown in MEM alpha (no nucleosides) Gibco 12561049, with 10% fetal calf serum, and PS (100 IU/100 μg/ml) finally added. CHO PIR-P3 cells were cultured in RPMI with 10% HI FBS and 1% PSglutamine; CHO Lec2 cells were cultured in αMEM with ribonucleosides and deoxyribonucleosides in 10% HI FBS. The supernatant, containing the MR-Fc, was collected after 9 days. MR-Fc was purified from cell culture supernatant using Protein A-agarose beads. The supernatant was incubated with the beads at 4 C on a rotator overnight, followed by three washes with PBS and MR-Fc elution using 0.1 M Glycine-HCl, pH 2.7 and immediate neutralization with 1 M Tris-HCl, pH 8.7. MR-Fc was stored at −20 C until further use.

Microarray printing Man-BSA array
We developed a customized glycan microarray to test MR-Fc glycan-binding activity. Different presentations of mannose were used, such as Man3-BSA, Man3-PAA, Man-BSA and the corresponding controls Glu-PAA, biotinylated Glu-PAA, GlcNAc-BSA, BSA, MR-Fc HEK-WT , and PBS. All probes were printed onto Nexterion Slide H NHS Slides (Schott AG), using the sciFLEXARRAYER S11 (Scienion US, Inc) equipped with piezo-dispensing capillary nozzles with a modified Type 3 coating (PDC 70 Type 3) as sold by the manufacturer. Careful adjustments were made to the voltage and pulse parameters so as to dispense 330 ± 10 pl per spot, and four spots per probe were printed per subarray on the slide. In total, there were 16 subarrays per slide. In some cases, additional probes may have been printed which were irrelevant to the current study so as to minimize costs. After the print, the slides were incubated overnight at 70% relative humidity and room temperature (RT). The next day, any unreacted surface of the slides was blocked by placing the slides in a solution of 50 mM ethanolamine in borate buffer (100 mM sodium tetraborate buffer pH 8.5) for 1 h. Slides were then removed from this solution one-by-one, dip-washed 10 times in PBS containing 0.05% Tween-20 and then ten times in water, following which they were dried using a centrifuge. Slides were stored at −20 C prior to use. Note that, since the probes were very different in nature (e.g., protein linker versus polyacrylamide linker with variations in the stoichiometry of conjugation to the glycan), their concentrations were kept proportional to that provided by the manufacturer and diluted with buffer to one or more dilutions so that the majority of the buffer was PBS. Final concentrations of all probes are listed in Table S3. As a control for slide printing, appropriate lectin controls were assayed and analyzed.

Microarray screening
Glycan microarrays were screened using a standard protocol as published elsewhere (32). Briefly, MR-Fc binding was performed in tris salts metals buffer (TSM) binding buffer [TSM (20 mM Tris-HCl, pH7.4, 150 mM NaCl, 2 mM CaCl 2 , and 2 mM MgCl 2 ) + 0.05% Tween 20 + 1% BSA] for 1 h at RT on a shaker, followed by four washes with TSM + 0.05% Tween 20 and incubation with a secondary antibody (goat anti-human IgG Alexa-Fluor 488-conjugated, 5 μg/ml) in TSM-binding buffer for 1 h at RT on a shaker. The slides were washed four times with TSM + 0.05% Tween 20, followed by TSM and H 2 O. As a control, the binding was performed in the presence of 5 mM EDTA instead of CaCl 2 . The arrays were scanned using a GenePix 4300A scanner (Molecular Devices) at 488 nm, and data readout was performed using the GenePix Pro software package.
The following glycan microarrays were screened for binding under the same conditions as mentioned above: MR-Fc binding was analyzed on the CFG glycan array (32) (version 5.3) at 16 μg/ml MR-Fc HEK-WT , 14 μg/ml MR-Fc CHO-WT , 24 μg/ml MR-Fc Lec2 , 32 μg/ml MR-Fc PIR-P3 ; oligomannose array (33) at 10 μg/ml MR-Fc HEK-WT and 15 μg/ml for all CHO-based MR-Fc; on the Man-BSA array at 1 μg/ml. The human full-length MR was analyzed for binding on the Man-BSA array at 16 μg/ml with an anti-penta-His Alexa-Fluor 488 conjugate at 5 μg/ml. Details about the preparation of these arrays can be found in the corresponding references.

MR-Fc enzymatic treatment before mass spectrometry analysis
For glycopeptide analysis, 2 μg MR-Fc from CHO cells, 8 μg MR-Fc HEK-WT , and 1 μg human full-length MR were treated with PNGase F, or mock treated, according to the manufacturer's instructions. Alternatively, MR-Fc was treated with O-glycosidase according to the manufacturer's instruction, including the addition of 2 μl neuraminidase. A portion of the human full-length MR PNGase F-treated samples was also incubated with 1 μl neuraminidase and further incubation at 37 C overnight. The samples were subjected to SDS-PAGE and subsequent in-gel trypsin digestion as described elsewhere (34). Protein bands were digested with 20 μl of 0.0025 μg/μl sequencing grade trypsin in 25 mM ammonium bicarbonate overnight at 37 C. The next day, the supernatant was collected and 50 μl of 50% acetonitrile was added and incubated for 10 min on a shaker. Both supernatants were combined and dried down in a speed vac concentrator. The samples were stored at −20 C until further use.

LC-MS glycopeptide analysis and data analysis
LC-MS was performed on an Ultimate 3000 nano LC coupled to an Orbitrap Fusion Lumos mass spectrometer (both Thermo Fisher Scientific), as described elsewhere (35). MR-Fc samples were diluted in 0.1% formic acid (FA) in H 2 O and loaded onto a C18 precolumn (C18 PepMap 100, 300 μm × 5 mm, 5 μm, 100 Å, Thermo Fisher Scientific) with 15 μl/min solvent A (0.1% FA in H 2 O) for 3 min and then separated on a C18 analytical column (PicoFrit 75 μm ID × 150 mm, 3 μm, New Objective) using a linear gradient of 2% to 45% solvent B (80% acetonitrile, 0.1% FA) over 29 min, followed by 45% to 90% B over 5 min at 400 nl/min. The mass spectrometer was operated under the following conditions: the ion source parameters were 1750 to 1850 V spray voltage and 200 C ion transfer tube temperature. MS scans were performed in the orbitrap at a resolution of 60,000 within a scan range of m/z 400 to m/z 1600, a RF lens of 30%, and AGC target of 1e5 for a maximum injection time of 50 ms. The top five precursors were selected for MS 2 in a data-dependent manner, within a mass range of m/z 400 to m/z 1600, a minimum intensity threshold of 5e4, and an isolation width of 2 m/z. Higher-energy collisional dissociation is performed in stepped collision energy mode of 25% or 28% or 30% (±5%) and detected in the orbitrap with a resolution of 30,000 with the first mass at m/z 120, an AGC target of 1e5, and a maximum injection of 200 ms. EThcD was performed in a product ion-dependent manner (m/z 204.0867) with 20% supplemental activation energy and detected in the orbitrap with a resolution of 30,000 with the first mass at m/z 120, an AGC target of 3e5, and a maximum injection of 250 ms. Detailed LC-MS settings for the analysis of the human fulllength MR is provided in the supplementary information.
Relative quantitation of all glycopeptides was performed in an automated manner as described previously (36). The glycopeptide reference list contained all glycopeptides that were identified based on MS 2 fragmentation but also lower abundant glycopeptides based on their exact mass, corresponding retention time, isotopic pattern, and biosynthetic-related glycan composition. Relative intensities were determined based on triplicate analysis, and standard deviation was calculated.

Data availability
The MS data have been deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE (37) partner repository with the dataset identifier PXD034781. The glycan microarray data will be available at the National Center for Functional Glycomics website, https://ncfg.hms.harvard.edu/ncfg-data/microarray-data.
Any remaining data is either presented in the article or freely available upon request to: Dr Richard D. Cummings, rcummin1@bidmc.harvard.edu.