A Hyaluronan Receptor for Endocytosis (HARE) Link Domain N-Glycan Is Required for Extracellular Signal-regulated Kinase (ERK) and Nuclear Factor-κB (NF-κB) Signaling in Response to the Uptake of Hyaluronan but Not Heparin, Dermatan Sulfate, or Acetylated Low Density Lipoprotein (LDL)*

Background: The HARE HA binding Link domain contains N-glycans at Asn2280. Results: HARE(N2280A) cells endocytose HA or Hep normally; NF-κB signaling occurs normally during uptake of heparin, dermatan sulfate, or acetylated LDL but not hyaluronan. Conclusion: HA·HARE-mediated signaling requires a link domain N-glycan. Significance: A different mechanism for HA·HARE signaling versus other ligands indicates a physiologic need for special control of responses to circulating HA. The human hyaluronan (HA) receptor for endocytosis (HARE; the 190-kDa C terminus of Stab2) is a major clearance receptor for multiple circulating ligands including HA, heparin (Hep), acetylated LDL (AcLDL), dermatan sulfate (DS), apoptotic debris, and chondroitin sulfate types A, C, D, and E. We previously found that HARE contains an N-glycan in the HA binding Link domain (at Asn2280), and cells expressing membrane-bound HARE(N2280A) bind and endocytose HA normally (Harris, E. N., Parry, S., Sutton-Smith, M., Pandey, M. S., Panico, M., Morris, H. R., Haslam, S. M., Dell, A., and Weigel, P. H. (2010) Glycobiology 20, 991–1001). Also, NF-κB-mediated signaling is activated by HARE-mediated endocytosis of HA, Hep, AcLDL, or DS but not by chondroitin sulfates (Pandey, M. S., and Weigel, P. H. (2014) J. Biol. Chem. 289, 1756–1767). Here we investigated the role of Link N-glycans in ligand uptake and NF-κB and ERK1/2 signaling. HA·HARE-mediated ERK1/2 activation was HA size- dependent, as found for NF-κB activation. HARE(N2280A) cells internalized HA, Hep, AcLDL, and DS normally. No ERK1/2 activation occurred during HA endocytosis by HARE(N2280A) cells, but activation did occur with Hep. Dual-luciferase recorder assays showed that NF-κB-mediated gene expression occurred normally in HARE(N2280A) cells endocytosing Hep, AcLDL, or DS but did not occur with HA. Activation of NF-κB by endogenous degradation of IκB-α was observed for HARE(N2280A) cells endocytosing Hep, AcLDL, or DS but not HA. We conclude that a Link domain complex N-glycan is required specifically for HARE·HA-mediated activation of ERK1/2 and NF-κB-mediated gene expression and that this initial activation mechanism is different from and independent of the initial mechanisms for HARE-mediated signaling in response to Hep, AcLDL, or DS uptake.

tion of extracellular signal-regulated kinases 1 and 2 (ERK1/2) in a dose-and time-dependent manner (16), and anti-inflammatory cytokine release is stimulated in macrophages phagocytosing apoptotic cells via HARE binding to phosphatidylserine (10). We reported that HA⅐HARE interactions stimulate NF-B activation of gene expression and that HARE displays an HA size-sensing mechanism for detecting and responding to a narrow size range (40 -400 kDa) of HA degradation products (17). We examined seven other HARE ligands for their ability to stimulate signaling and discovered that DS, AcLDL, or Hep uptake also activates NF-B-mediated gene expression but that CS types A, C, D, and E are inactive (18). As with HA, the latter four ligands also bind to the Link domain, and all four can compete for HA binding. In contrast, DS, AcLDL and Hep are bound at a second independent site and are not competed by HA or the CS types (11). Clathrin mRNA knockdown experiments show that NF-B activation by any of the signaling ligands is dependent on the coated pit mediated endocytosis of HARE⅐ligand complexes.
We previously analyzed the N-glycans within the human 190-kDa HARE ectodomain, expressed in Flp-In 293 cells, and found that 10 Asn sites are occupied with multiple glycoforms of varying complexity. Most interesting to us was that the site with the greatest diversity of glycans was Asn 2280 in the Link domain; it contained 15 structures and was the only site with sialylated complex glycans. Cells expressing HARE(N2280A), lacking a Link domain N-glycan, endocytosed HA normally.
In the present study, we used stable Flp-In 293 cell lines expressing N2280A or wild type (WT) HARE to investigate the role of the Link N-glycan at Asn 2280 on the ability of HARE⅐ligand complexes to activate the ERK1/2-and NF-Bmediated signaling pathways. The results demonstrate a surprisingly ligand-specific requirement for an Asn 2280 glycan on HA⅐HARE-mediated ERK1/2 and NF-B activation. In contrast, Hep⅐HARE-mediated activation of the ERK1/2 and NF-B signaling pathways occurred normally in the absence of a Link domain Asn 2280 glycan.
Preparation of 125 I-SA⅐b-HA or 125 I-SA⅐b-Hep Complexes-Radiolabeled complexes for HA ( 125 I-SA⅐b-HA) or Hep ( 125 I-SA⅐b-Hep) for endocytosis assays were prepared as noted (9); a 2:1 molar ratio of b-HA or b-Hep and 125 I-SA was incubated in 0.5 ml of endocytosis medium for 1 h on a rotary mixer at 22°C just before the experiment. For nonspecific binding and uptake controls, a 2:1 ratio of free biotin and 125 I-SA was used to make 125 I-SA⅐biotin. The pre-formed radiolabeled complexes were diluted in endocytosis medium to the final concentrations indicated.
Endocytosis of 125 I-SA⅐b-HA or 125 I-SA⅐b-Hep-Stable cell lines expressing WT HARE, HARE(N2280A), or EV were plated in 12-well plates and grown in complete medium to 80 -90% confluence before experiments. Cells were washed with Hanks' balanced salt solution and incubated at 37°C for 1 h with endocytosis medium to allow uptake of any bound serum glycosaminoglycans. The medium was aspirated and replaced with endocytosis medium containing 50 nM preformed complexes of 125 I-SA⅐b-HA or 125 I-SA⅐b-Hep. Nonspecific uptake was assessed in parallel samples by incubating cells with 125 I-SA⅐biotin complexes. After the indicated times at 37°C, the medium was removed by aspiration, and cells were washed 3 times (2 ml each) with ice-cold Hanks' balanced salt solution to remove unbound radioligand and solubilized in 1 ml of 0.3 N NaOH. Radioactivity was measured using a Packard Cobra II gamma counter, and lysate protein content was deter-mined by the method of Bradford (24) using bovine serum albumin as standard. Specific uptake of 125 I-SA⅐b-HA or 125 I-SA⅐b-Hep by each cell type was calculated by subtracting cellassociated 125 I-SA⅐biotin. HARE-specific binding and uptake of 125 I-SA⅐b-HA or 125 I-SA⅐b-Hep by HARE(N2280A) cells was then determined by normalizing HARE expression relative to WT using Western blot quantification of equal lysate protein samples (15). Result values are expressed as the mean Ϯ S.E. specific fmol/g of cell protein/HARE.
Cell Culture and Stimulation with HA or Hep for ERK1/2 Activation-WT, N2280A, or EV cells were grown in complete medium till confluence and then plated in 12-well tissue culture plates as described (17). The cells were grown in complete medium for at least 2 days (80 -90% confluence) before experiments. Cells were washed with sterile PBS and incubated in fresh medium without serum for 1 h at 37°C (16). Cells were washed and then incubated at 37°C in fresh serum-free medium containing 10 g/ml HA (80 kDa, 125 nM) or Hep (ϳ500 nM) for the indicated times. To determine the HA size dependence for ERK activation, cells were incubated for 20 min with 10 g/ml HA of different weight-average molar masses (14,51,178, 741, or 2600 kDa). Cells were then washed with ice-cold PBS and lysed with lysis buffer. The cell lysates were collected, stored on ice, vortexed repeatedly, and then centrifuged at 12,000 ϫ g for 10 min at 4°C to remove cell debris. Protein concentration in each cell lysate supernatant was determined by Bradford assay (24).
Western Blotting Analysis-Western blotting was performed as noted (16) with some modifications. Briefly, 25 g of cell lysate protein was resolved in 10% SDS-PAGE and electrotransferred to nitrocellulose using a mini-transfer unit (Bio-Rad) for 1 h at 4°C at 110 V in 25 mM Tris-HCl, pH 7.4, 190 mM glycine, and 20% (v/v) methanol. Membranes were washed in TBST for 5 min and incubated with 5% nonfat dry milk in TBST for 1 h at 22°C. For phospho-ERK1/2 (pERK1/2) detection, the preblocked membranes were incubated overnight at 4°C with the Ab to pERK1/2 (1:1000 dilution in blocking buffer). After washing 3 times for 5 min each in TBST, the membranes were incubated with anti-rabbit IgG-HRP conjugate (1:2500 dilution in blocking buffer) at 22°C for 1 h. The membranes were washed 3 times for 5 min each with TBST, developed using enhanced chemiluminescence substrate, and exposed to autoradiography film. For total-ERK1/2 (tERK1/2) detection, the same membranes were stripped to remove bound the Ab by washing with water for 5 min and then incubating in stripping buffer at 55°C for 30 min. After five washings with TBST for 10 -15 min each, the blots were reprobed with tERK1/2 (1:1000). For normalizing protein load in each well, the membranes were further stripped and reprobed with anti-actin Ab. Similarly, for determination of HARE expression by WT and N2280A cells, preblocked membranes were incubated with 20 ng/ml anti-V5 Ab (Bethyl Laboratories), washed 3 times for 5 min each in TBST, incubated with the appropriate anti-goat IgG-HRP (1:2000 dilution), and processed as above. Autoradiographic band densities were scanned into digital files and quantified as integrated densitometry values (i.e. the sum of all pixel values minus background correction) using an Alpha Innotech FluoroChem 8000 imaging system (Alpha Innotech Corp., San Leandro, CA). ERK1/2 activation in response to HA, Hep, or the control (buffer) treatment is expressed as a ratio (pERK: tERK) and compared with the mock treatment as 100%.
Transient Transfection with Firefly and Renilla LUC-WT, N2280A, or EV cells were grown in complete medium till confluence, plated in 12-well tissue culture plates, and grown in complete medium for at least 2 days until cells reached 60 -70% confluence. The medium was replaced with fresh transfection medium 10 min before transfection, transfection complexes were generated, and cells were transiently transfected as described (17).
Stimulation of Cell Lines with HARE Ligands and LUC Assays-WT, N2280A, or EV cells transiently transfected with firefly and Renilla LUC vectors for 18 h were washed 1 time each with sterile PBS and then with DMEM without serum. Cells were preincubated with fresh serum-free DMEM for 1 h at 37°C. To assess ligand specificity for HARE-mediated NF-B-activated reporter gene expression, cells were stimulated for 4 h with low endotoxin HA, Hep, DS, AcLDL, or CS types A, C, D, or E (100 nM) in DMEM. After the indicated times, the medium was aspirated, and cells were processed to determine NF-B-activated gene expression (17).
IB-␣ Degradation Assay-Cells expressing HARE or N2280A were grown to confluence in 6-well plates, washed with sterile PBS, and incubated with serum-free DMEM for 1 h. Cells were then incubated with 1 ng/ml TNF-␣, 100 nM HA, Similarly, the three cell types were incubated with 125 I-SA⅐biotin and processed as described under "Experimental Procedures" to determine nonspecific (not shown) and specific endocytosis of HA or Hep. Based on three independent experiments, values are the means Ϯ S.E. (n ϭ 9) of specific cell-associated ligand (fmol/g of protein/HARE); HARE expression in N2280A cells was normalized relative to WT HARE expression, which was 100%.
Hep, DS, or AcLDL or 250 nM CS-A, CS-C, CS-D, or CS-E for 45 min and processed as described previously (18). Equal amounts of cell lysate protein were resolved on 10% SDS-PAGE, transferred to nitrocellulose, and probed with anti-IB-␣ Ab. The same membrane was reprobed for actin, and the ratio of IB-␣: actin in each sample was determined from the digital images.
Statistical Analysis-Western blot data are presented as the mean Ϯ S.E. based on three independent (n ϭ 3) experiments unless noted otherwise, and differences were assessed by analysis of variance as noted below. LUC reporter assay data are presented as the mean Ϯ S.E. based on three independent experiments, each performed in triplicate (n ϭ 9). Data were first analyzed by a one-way analysis of variance, and any significant difference in the group was then assessed by individual pair-wise post hoc Tukey's HSD tests using GraphPad Prism v5 statistical software (GraphPad Software, Inc., San Diego). Comparisons were made for none treatment and treatment with the same ligand concentration. Values of p Ͻ 0.05 were considered statistically significant. Only those samples found to be statistically significant compared with both the no treatment and same ligand or cell type are indicated with a symbol (*, p Ͻ 0.05; **, p Ͻ 0.005; ***, p Ͻ 0.001).

Endocytosis of HA or Hep Is Not Affected by Loss of the HARE
Link Domain N-Glycan-We previously found that Hep and HA bind to distinct, non-overlapping binding sites within the HARE ectodomain (11). Four different CS types also bind to the HA binding site in a competitive manner with HA, whereas DS and AcLDL bind within or near the Hep binding site and compete among themselves to varying extents. Cells expressing EV, WT, or N2280A HARE were assessed for their ability to endocytose complexes of b-HA or b-Hep with 125 I-SA (Fig. 1). Flp-In 293 cells were chosen for studies on HARE function because these cells do not express significant levels of other HA recep-tors, such as CD44 or RHAAM (7). Thus, the level of specific 125 I-SA⅐b-HA uptake by EV cells was very low, whereas the uptake of these complexes by either WT or N2280A cells was much greater and identical (Fig. 1A). The presence or absence of a Link domain N-glycan had no effect on the ability of HARE to bind and endocytose HA. Similar results were observed for the uptake of 125 I-SA⅐b-Hep by the three cell types (Fig. 1B), except that the background level of uptake by EV cells was , or N2280A (E and F) HARE were grown to confluence and washed. After 1 h of incubation in serum-free medium at 37°C, the cells were incubated with 10 g/ml HA (80 kDa) for the indicated times and processed as in "Experimental Procedures." Sample cell lysates (25 g) were resolved in 10% SDS-PAGE and transferred to nitrocellulose. Western blot analysis was performed with the Ab against pERK1/2 (top row, top panels). The same membranes were stripped and reprobed with Ab against tERK1/2 (middle row, top panels) and then with anti-actin Ab (bottom row, top panels). Blots from three independent experiments (n ϭ 3) were scanned and digitized, and densitometry analysis was performed to determine pERK:tERK ratios at each time point; **, p Ͻ 0.005; ***, p Ͻ 0.001.  Fig. 2 after incubation for 20 min with 10 g/ml HA using preparations of different mass as indicated (weight-average molar masses were determined by light scattering as described previously (17,22)). Normalized data are presented as the mean Ϯ S.E. percent of the pERK:tERK ratio compared with the time 0 (no addition) value as 100%. Values for p are based on one-way analysis of variance and pair-wise Tukey's tests as noted under "Experimental Procedures" using pairwise comparisons of WT or N2280A cells at each time or with each other at the same time. Only the sample sets with significant differences in each case are marked: **, p Ͻ 0.005.
greater than for HA, as reported previously (9). The rates of specific 125 I-SA⅐b-Hep endocytosis by WT or N2280A cells were identical and much greater than EV cells. Thus, loss of the N-glycan in the HA-binding Link domain does not disrupt the endocytosis of either HA or Hep.
Activation of ERK1/2 in Response to HA⅐HARE Uptake Does Not Occur When the Link Domain N-Glycan Is Missing-We previously found that HARE-mediated uptake of HA causes activation of ERK1/2 (16), although the downstream components activated via pERK1/2 are not known yet. To assess the effect of Link domain N-glycan loss on the ability of HARE to activate ERK in response to HA, we followed tERK1/2 and pERK1/2 in EV, WT, and N2280A cells treated with HA. EV cells showed no changes in pERK1/2 levels over 60 min when treated with HA (Fig. 2, A and B), whereas WT cells showed a significant biphasic increase in pERK1/2 that peaked at 15 min (Fig. 2, C and D). However, N2280A cells lacking a Link domain glycan showed no significant changes in pERK1/2 levels in the presence of HA (Fig. 2, E and F), similar to EV cells.
HARE-mediated Activation of ERK1/2 Is Dependent on HA Size-There is a very pronounced HA size dependence for the ability of HA⅐HARE complexes to stimulate NF-B signaling and gene activation (17). The active HA size range is 40 -400 kDa, and smaller or larger HA is unable to activate NF-B even though all HA molecules Ͼϳ2 kDa are bound and internalized. Because the relationship between the ERK1/2 and NF-B pathways activated by HA⅐HARE complexes is not known, we assessed whether the activation of ERK1/2 is also dependent on HA size (Fig. 3). Because HA-stimulated ERK1/2 activation was maximal at 15-30 min, we assessed pERK1/2 levels after 20 min in the presence of HA sizes below (14 kDa), within (51 and 178 kDa) and above (741 and 2600 kDa) the activating range. The results showed the same HA size dependence as for NF-B activation. Only WT cells treated with 51-and 178-kDa HA showed significant stimulation of ERK1/2 (p Ͻ 0.005). In contrast, treatment of WT cells with smaller (14 kDa) or larger HA (741 and 2600 kDa) did not stimulate ERK1/2 activation (Fig. 3,  A and B). As expected, Asn 2280 cells treated with any size HA showed no stimulation of ERK1/2 activation (Fig. 3, C and D). ERK1/2 Activation in Response to Hep⅐HARE Uptake Occurs with or without the Link Domain N-Glycan-We also examined the possibility that Hep endocytosis by WT HARE cells stimulates ERK phosphorylation (Fig. 4). WT cells endocytosing Hep showed a similar level of pERK1/2 activation (Fig. 4, C and D), significantly greater from 5-60 min than EV control cells (Fig.  4, A and B); interestingly, the response to Hep was sustained longer than the HA response. Unlike the results with HA, ERK1/2 activation by N2280A cells endocytosing Hep (Fig. 4, E and F) was not altered by Link glycan loss; it was identical with that of WT cells.
NF-B Activation in HARE(N2280A) Cells Occurs during Endocytosis of Hep but Not HA-Because HARE-mediated endocytosis of either HA or Hep activates NF-B, leading to expression of a LUC recorder gene (18), we used this system to assess the effect of Link domain glycan loss on cell signaling. EV cells treated with Hep or two different HA preparations, a narrow-size range 107-kDa HA and essentially mono-disperse 133-kDa Select HA TM , showed no change in expression of inducible LUC driven by an NF-B promoter (Fig. 5A, gray  bars). WT cells (Fig. 5A, black bars) or N2280A cells (Fig. 5A, white bars) treated with Hep showed essentially identical and significant activation of LUC gene expression (p Ͻ 0.005). Similarly, WT cells treated with either HA preparation showed significant gene activation (p Ͻ 0.001). In contrast, the N2280A cells showed no NF-B-mediated gene activation in the pres- A mixing control with the two ligands was performed to verify that HA did not interfere with Hep signaling in N2280A cells; in WT cells, signaling responses with both HA and Hep are additive (18). The activation of 5 nM Hep was not affected by the presence of 5 nM HA, consistent with the lack of competition for binding by the two ligands (Fig. 5B), indicating that the lack of a Link domain N-glycan does not affect the independence of the two ligand binding sites.
We also examined DS and AcLDL, the two other signaling ligands that do not compete with HA for binding. DS and AcLDL were both able to activate LUC gene expression in WT or N2280A cells but not EV cells (Fig. 5C), demonstrating that these ligands are endocytosed normally and neither ligand⅐HARE complex requires a Link domain N-glycan for signaling leading to NF-B activation and gene expression. An additional control was performed to verify that HA⅐HARE-mediated signaling was completely lost and that the size range for activating HA had not been altered in cells expressing the N2280A HARE variant. We tested HA preparations that were smaller (14 kDa) or larger (741 kDa) than HA in the normal 40 -400-kDa signaling range (Fig. 5D). None of these HA preparations activated NF-B signaling in N2280A cells, and only the small-intermediate size HA (e.g. 51, 107, and 133 kDa) used here was able to activate signaling in WT cells.

Loss of the HARE Link Domain or the Link Domain N-Glycan Differentially Affects HA Signaling Compared with Other
Ligands-We next sought to further define the possible interactions between the Hep and HA binding sites and to determine if N-glycan loss alters the signaling potential when CS types A, C, D, and E bind to the Link domain. N2280A cells incubated with 250 nM CS types A, C, D, and E, alone or in combination without or with HA, did not show activation of NF-B mediated gene expression (Fig. 6A). In addition, treatment of these cells with Hep led to significant activation of LUC gene expression (p Ͻ 0.001), and this was not altered by the presence of the four CS types. To define the requirement of the Link domain for signaling responses to the uptake of HA, DS,  Hep, and AcLDL, we treated HARE(⌬Link) cells, in which the entire Link domain was deleted, with each of the four signaling ligands (Fig. 6B). As reported previously (17), HA was unable to initiate signaling, because HA as well as the four CS types (not shown) is not bound and internalized by HARE(⌬Link) cells. In contrast, DS (p Ͻ 0.001), Hep (p Ͻ 0.005), and AcLDL (p Ͻ 0.001) all activated significant HARE-mediated NF-B signaling, identical to that observed in WT cells. The results also demonstrate that the ability of HARE complexes with DS, Hep, or AcLDL to activate signaling does not require the Link domain. Thus, the results confirm that the binding sites for DS, Hep, and AcLDL are distinct from the HA binding site.
In addition, the above results indicate that the early mechanism by which endocytosed (or endocytosing) HA⅐HARE complexes initiate activation of the ERK1/2 and NF-B signaling pathways is different than the early signal activation mechanisms for HARE complexes with Hep, DS, and AcLDL.
Activation of Endogenous IB-␣ Degradation Does Not Occur during HA⅐HARE(N2280A) Endocytosis-Part of the cell signaling cascade leading to NF-B promoter-driven gene expression is phosphorylation of NF-B⅐IB-␣/␤ complexes; IB-␣/␤ is a potent intracellular inhibitor of NF-B (25). The localization of these inactive NF-B complexes with IB-␣/␤ is also shifted from the nucleus to the cytoplasm. We previously found that in the absence of the LUC recorder gene plasmids, HARE-mediated endocytosis of HA, Hep, DS, or AcLDL leads to degradation of ϳ50% of the cellular IB-␣ protein in Ͻ1 h (18). To determine whether this endogenous signaling cascade is altered by the loss of the Link domain N-glycan, we assessed degradation of IB-␣ in WT (Fig. 7, A and B) and N2280A (Fig. 7, C and D) cells endocytosing the different ligands. When both cell types were incubated with TNF-␣ as a positive control, cellular IB-␣ was significantly decreased (Fig. 7, B and D) by 45-50% within 45 min (p Ͻ 0.005). When both cell types were incubated with a mixture of the four nonsignaling CS-types, A, C, D, and E (250 nM each), no IB-␣ degradation occurred. In contrast, WT cells responded significantly to the uptake of the three signaling ligands Hep, DS, or AcLDL by degrading 35-51% of the endogenous IB-␣ protein ( Fig. 7B; p Ͻ 0.005). Similarly, N2280A cells showed significant reduction of cellular IB-␣ during the endocytosis of Hep, DS, or AcLDL ( Fig. 7D; 35-51% degradation, p Ͻ 0.005-0.05). During HA uptake in WT cells (Fig. 7B), endogenous IB-␣ levels decreased to 64.4 Ϯ 5.2% of the no addition (buffer only) control (p Ͻ 0.05). However, the level of endogenous IB-␣ did not change in N2280A cells endocytosing HA (Fig. 7D). These results confirm that the endogenous NF-B signaling pathway in the Flp-In 293 cell lines used here is functional and that activation of this pathway during HAREmediated endocytosis of HA requires a Link domain N-glycan. The results also validate the above findings using LUC recorder gene assays.

DISCUSSION
HARE/Stab2 are constitutively recycling clathrin-coated pittargeted receptors. By definition, constitutive recycling means that receptor internalization occurs continuously and is not coupled to or dependent on ligand binding. The pathways by which these receptors function were well studied 3-4 decades ago and were understood and described by kinetic parameters just as for a typical metabolic pathway (26 -31). Because these receptor pathways are continuous, the kinetic constants for the various steps define the fraction of receptors present in the multiple participating surface and intracellular locations (compartments). Thus, HARE molecules continuously move along a fixed surface-intracellular spatial and temporal route designed for efficient large scale interactions with ligands, often at high concentrations. In contrast, most signaling receptors are resident on the surface for much longer times and designed to respond to relatively small amounts of a special ligand whose concentration is normally low and only increases at relatively long intervals to levels that allow binding and signaling.
The primary function of systemic clearance receptors is the continuous removal of ligands from the lymphatic and vascular circulation systems, so that they can be degraded within cells. As with the asialoglycoprotein, mannose, LDL, and other clearance receptors, HARE/Stab2 do not localize to any plasma membrane domain other than coated pits, which also undergo a dynamic parallel recycling pathway. Consequently, the surface residence time of these receptors is short, and usually Յ30% of total receptors are on the surface at any time. The majority of total receptors (Ն70%) travel through many intracellular compartments along the endocytic and receptor recycling pathways. Receptors move to the cell surface, diffuse to a coated pit, are internalized with or without bound ligand, and are recycled back to the cell surface. If ligand is present or not, the same level of receptor recycling occurs, and there is no difference in the time receptor spends on the surface nor its ability to interact with adaptor proteins and be incorporated into a coated pit.
The expression levels and the surface-internal distribution of N2280A and WT HARE are similar, with ϳ20% of the total on the cell surface (21). More importantly, both receptors are equally functional in being able to mediate continuous ligand uptake, which demonstrates identical times for receptor movement along the constitutive receptor recycling pathway. During HA (Fig. 1A) or Hep (Fig. 1B) uptake, the surface receptor cohort is recycled and replaced by internal recycling receptors about every 7 min (7).
The functional equivalence of WT and N2280A HARE shows that the protein expression and distribution levels along the recycling pathway are the same, and their functionalities are also the same in terms of their rates of mediating endocytosis, ligand dissociation, and receptor recycling back to the surface. Thus, loss of the Link glycan does not alter receptor localization, trafficking, or endocytic function. This conclusion is further supported by the finding that NF-B signaling, mediated by the uptake of glycan-deficient HARE complexes with Hep, AcLDL, or DS, is identical to WT receptor. Because NF-B signaling is dependent on clathrin-mediated endocytosis (18), the overall localization, trafficking, and function of WT and N2280A HARE must be identical; the only effect of not having a Link glycan (or domain) is the loss of HA-HARE-mediated signal transduction.
Specific N-glycans on receptors have been linked to cell signaling functions, including a salt-sensing mucin that activates a mitogen-activated protein kinase (32), M3 muscarinic receptor trafficking (33), and the dendritic cell immunoreceptor (34). Several groups reported that N-or O-glycans can influence the HA binding function of proteins containing a Link module (35)(36)(37)(38)(39)(40)(41). It is now generally accepted that N-glycans within the Link module of the HA receptor CD44 influence HA binding. In some cases the presence of an N-glycan inhibits HA binding. The HA binding ability of the lymphatic receptor LYVE-1 is inhibited by sialylation of O-glycans (42), which are not on the Link module, and neuraminidase treatment of sialylated LYVE-1 unmasks HA binding activity. The present results show a clear relationship between the presence of an N-glycan in the HARE Link domain and the ability of HA⅐HARE com- plexes to activate ERK1/2 and NF-B. Loss of a Link domain N-glycan eliminates the 40 -400-kDa HA size-dependent NF-B signaling; it does not shift the response to a smaller or larger size (Fig. 5D).
Early molecular details of how the four signaling ligand⅐HARE complexes interact with intracellular signaling components must be different for HA versus Hep, DS, and AcLDL. The latter three ligands bind at a different site than HA (11), and their ability to initiate ERK1/2 and NF-B signaling is different and independent of the HA signaling pathway, as demonstrated by their normal signaling in HARE(⌬Link) cells that are unable to endocytosis HA or initiate HA-mediated signaling (Fig. 5D). We conclude that, due to the involvement of two different ligand binding regions, there are at least two ligand⅐HARE complex conformational states during coated pitmediated endocytosis that are capable of interacting with, and activating, the initial molecular events leading to downstream signaling. One activating conformation is associated with the formation of HARE complexes with Hep, DS, or AcLDL, and the other conformation is associated with HA⅐HARE complex formation that is somehow dependent on the presence of a Link domain N-glycan.
We previously found that all N2280A clones express a major HARE species of WT size and a minor slightly smaller HARE species; both proteins are delivered to the cell surface and bind HA (21). Endoglycosidase H sensitivity shows that WT HARE and the major N2280A HARE have mature N-glycans, but the minor N2280A HARE species is shifted to smaller size, indicating multiple immature high mannose N-glycans. Both N2280A proteins are shifted to the same smaller size after peptide N-glycosidase F treatment. Because half of the 10 glycan sites are downstream of the Link glycan (Fig. 8E), these data indicate that loss of the Link glycan slightly increases the chance that some of these five C-terminal N-glycans are processed abnormally. Thus, two N2280A HARE isoforms are expressed that are as functional as WT HARE in their ability to bind and endocytose HA, although we cannot test this directly. The minor species is not altered in binding, endocytosis, or signaling, as no differences are apparent with the three other NF-B signaling ligands between WT and N2280A cells (Figs. 6 and 7), and both the HA and Hep uptake kinetics are identical to WT (Fig. 1).
A plausible mechanism to explain a critical role for an oligosaccharide is that it could enable an HA⅐HARE complex to create or stabilize a favorable conformation (e.g. an allosteric-like response after HA-HARE binding) for interacting with a key initial protein responsible for starting the activation cascade. Support for such a mechanism in the ability of HA⅐HARE complexes to initiate signaling may be evident in the earlier finding that the transmembrane domain also influences the binding of HA to the Link domain. Purified HARE(N2280A) ectodomain binds little or no HA in vitro in ELISA-like, pulldown, or surface plasmon resonance assays (21), whereas WT HARE ectodomain shows robust binding to HA in all assays, with a K d ϳ 5 nM. This indicates that an Asn 2280 glycan stabilizes a conformation needed for the Link domain to bind HA in the absence of the transmembrane region. In contrast, membrane-bound HARE binds and endocytoses HA normally with or without an Asn 2280 glycan. Although a Link domain N-glycan is not required for HA binding and endocytosis, the glycan might, nonetheless, interact with HA directly. Alternatively, the glycan might interact with another region of HARE (in or outside of the Link domain) made available by a conformational change associated with the binding of HA. Either scenario could create a glycandependent HA⅐HARE conformation that is required to initiate signal transduction.
Because a Link domain N-glycan is required for HA⅐HARE signaling, the diversity of glycans at Asn 2280 was unexpected; especially interesting is that Asn 2280 has the greatest diversity of glycans and the only sialylated structures of the 10 glycan sites (21). Fifteen glycoforms were identified at Asn 2280 : 12 are fucosylated in the core chitobiosyl region (Fig. 8, A-C), 12 are biantennate (Fig. 8, A, C, and D), and 3 are tri-antennate (Fig. 8B), 7 are sialylated at the same bi-antennate position (Fig. 8, C and  D), and 10 contain an additional GlcNAc with an ambiguous site of attachment (Fig. 8, A-D). This great diversity of glycan structures at a site requiring a glycan for signaling competence indicates either (i) that all the structures possess the basic features needed to activate signaling mediated by HA⅐HARE complexes or (ii) that there are subpopulations of glycan structures that are not signaling-competent and others that are signalingcompetent (perhaps with different levels of signaling responsiveness to HA). We favor the latter possibility because it fits the complexity of HA⅐HARE-mediated signaling. If the former possibility is correct, however, then one candidate for an activating structure that is common to all 15 glycans is the [GlcNAc 2 Man 3 GlcNAc 2 ] core region (red box; Fig. 8D).
In addition to the HA⅐HARE signal-initiating pathway being distinct from that of other ligand⅐HARE complexes, this pathway shows a remarkable selectivity for the size of the bound HA that activates signaling. Although HARE binds and endocytoses HA varying in size between ϳ2 kDa (17) and Ͼ8 MDa (43), only HA between 40 and 400 kDa can activate signaling; smaller or larger HA is endocytosed normally but does not activate ERK1/2 or NF-B (17). Considering the broad HA size ranges in most mammalian sources, the steady-state HA size in the vascular circulation of healthy people (ϳ100 kDa; Ref. 44) is within the activating size range and close to the optimum size (ϳ140 kDa) for HA⅐HARE signaling (17).
Since its discovery as an HA clearance receptor in 1981 by Laurent and co-workers (45), HARE/Stab2 have been found to clear at least 14 ligands (including bacteria and apoptotic cells). Clearance is achieved by tissue-resident sinusoidal endothelial cells and migratory macrophages. HARE on sinusoidal endothelial cells in lymph node and liver rapidly removes ligands (e.g. half-lives of Ͻ5 min) from the lymphatic and vascular systemic circulations, respectively. Macrophages clear HARE ligands they encounter in their travels through the circulatory systems and tissues. During HARE/Stab2-mediated phagocytosis, macrophages activate NF-B leading to new gene expression, in particular TGF-␤ (10, 16 -18).
The discovery that ERK1/2 and NF-B signaling pathways are activated by a subset of HARE ligands was surprising (18), as this additional function must be coupled to ligand uptake for a different purpose than the housekeeping clearance function (5). The realization that essentially all HARE ligands are derived from tissue biomatrices or cells led us to propose a Tissue Stress Sensor Model in which the HARE/Stab2 receptor signaling system, operating in parallel with its clearance-turnover function, monitors the status of tissue biomatrix turnover and stress throughout the body (5). The signaling activity and responses of the HARE signaling system during continuous clearance of key indicator ligands (e.g. HA, Hep, DS, AcLDL, and phosphatidylserine) could monitor and respond to the pattern of tissuederived ligands and thus the homeostasis of tissue-biomatrix turnover throughout the body.
The complexity of how the HARE signaling mechanism senses HA supports a growing body of evidence that HA turnover and homeostasis is a more fundamentally important process than just simply the systemic or localized removal of degraded HA. We can understand why this is the case based on the importance of HA size in determining the nature of cellular interactions with, and responses to, HA in the environment. Since the first report in 1985 that small HA oligosaccharides are angiogenic (46), many groups have confirmed the basic finding that size matters; that cellular behaviors are different in the presence of different size ranges of HA (14). The physiologic importance of the ability to monitor tissue HA degradation and homeostasis is underscored by the present finding that the early HA⅐HARE signaling pathway is also different from the HAREmediated signaling pathway for other activating ligands.