N-Linked Glycosylation Regulates Human Proteinase-activated Receptor-1 Cell Surface Expression and Disarming via Neutrophil Proteinases and Thermolysin

Proteinase-activated receptor 1 (PAR1) induces activation of platelet and vascular cells after proteolytic cleavage of its extracellular N terminus by thrombin. In pathological situations, other proteinases may be generated in the circulation and might modify the responses of PAR1 by cleaving extracellular domains. In this study, epitope-tagged wild-type human PAR1 (hPAR1) and a panel of N-linked glycosylation-deficient mutant receptors were permanently expressed in epithelial cells (Kirsten murine sarcoma virus-transformed rat kidney cells and CHO cells). We have analyzed the role of N-linked glycosylation in regulating proteinase activation/disarming and cell global expression of hPAR1. We reported for the first time that glycosylation in the N terminus of hPAR1 downstream of the tethered ligand (especially Asn75) governs receptor disarming to trypsin, thermolysin, and the neutrophil proteinases elastase and proteinase 3 but not cathepsin G. In addition, hPAR1 is heavily N-linked glycosylated and sialylated in epithelial cell lines, and glycosylation occurs at all five consensus sites, namely, Asn35, Asn62, Asn75, Asn250, and Asn259. Removing these N-linked glycosylation sequons affected hPAR1 cell surface expression to varying degrees, and N-linked glycosylation at extracellular loop 2 (especially Asn250) of hPAR1 is essential for optimal receptor cell surface expression and receptor stability.

Proteinases can exert many biological effects on cells by virtue of their ability to cleave a plethora of protein substrates. Our understanding of proteinase signaling has advanced since the discovery of a small family of G protein-coupled receptors termed proteinase-activated receptors (PARs) 2 (1)(2)(3). Classically, thrombin activates PAR 1 by proteolytic cleavage of the PAR 1 N-terminal domain between Arg 41 and Ser 42 to expose a new N terminus that acts as a tethered ligand and subsequently triggers receptor activation (2). Thrombin is a serine proteinase generated at the sites of vascular injury during the coagulation cascade and activates various types of cells in the vasculature like platelets and endothelial cells (2,3). In addition to thrombin, other proteinases may be generated in the circulation under pathological situations; any proteinase that cleaves the correct peptide bond within the N terminus of PAR 1 may be able to expose the tethered ligand and then initiate intracellular signaling to provoke a cellular response (1, 2, 4 -8). Conversely, proteinases can also disarm PAR 1 by proteolytic removal of the tethered ligand domain from the receptor (4, 9 -14). Such disarmed receptors remain at the cell surface but are no longer available for activation by activating proteinases (e.g. thrombin); they can, however, be activated by the corresponding PAR 1 activating peptides (1,2,15). Moreover, some of those proteinases (e.g. trypsin and neutrophil proteinases) have been reported to activate and/or disarm PAR 1 in different cell systems (2, 4, 8, 13, 14, 16 -18). Activation and desensitization of PAR 1 must be well controlled in order to eliminate the potential pathological damage resulting from unchecked receptor signaling (14).
Our previous pharmacological studies have reported that activation of hPAR 2 by mast cell tryptase can be regulated by receptor N-terminal glycosylation (19,20). To our knowledge, hPAR 1 contains a total of five predicted sites for N-linked glycosylation; three sequons are located on the receptor N terminus (Asn 35 -Ala 36 -Thr 37 , Asn 62 -Glu 63 -Ser 64 , and Asn 75 -Lys 76 -Ser 77 ), and two are located within extracellular loop 2 (ECL2; Asn 250 -Ile 251 -Thr 254 , and Asn 259 -Glu 260 -Thr 261 ) ( Fig. 1) (21). PAR 1 , downstream of the tethered ligand, possesses two N-linked glycosylation sequons (Asn 62 and Asn 75 ), which are located along the N terminus in a domain where disarming proteinases may target. In addition, previous studies demonstrated that thermolysin and neutrophil proteinases target specific sites that are close proximally to these glycosylation sequons ( Fig. 1) (12,13,16). Hence, we have analyzed the role of N-linked glycosylation in regulating cell surface expression and proteinase activation/disarming of hPAR 1 . We have permanently expressed epitope-tagged wild-type or glycosylation-deficient mutant hPAR 1 in KNRK and CHO (Pro5 and Lec2) cell lines and sought to determine (i) the influence of N-linked glycosylation sequons on the receptor cell surface expression and signaling and (ii) the role of N-terminal glycosylation sequons, especially Asn 62 and Asn 75 , in regulating receptor signaling by neutrophil proteinases and thermolysin.
Flow Cytometry Analysis (FACS)-WT and mutant cell lines (ϳ90% confluence) with matched receptor expression were selected and used for further studies by FACS (BD Biosciences) as we described previously (19). The anti-PAR 1 ATAP-2 monoclonal antibody (1 g/ml) and anti-mouse FITC-conjugated antibody (1:1000 dilution) were applied to assess the levels of receptor cell surface expression. The data were expressed as the median fluorescence of positive minus the median fluorescence of the EV (pcDNA3.1 transfected KNRK) cells.
Confocal Microscopy Analysis-For the WT hPAR 1 eYFP and mutant cell lines, the light emitted by the eYFP (530 nm) fused to the receptor was used to assess receptor expression. Cells on coverslips were fixed by incubation in formaldehyde (3%) in PBS for 15 min at room temperature and then permeabilized in PBS containing 0.2% Triton X-100 for 10 min at room temperature. The coverslips were then mounted with anti-fade and fixed to slides. The cells were visualized by a Nikon Eclipse (TE2000-E) microscope with a Bio-Rad Radiance 2100 scanning system and lasers.
Western Blot Analysis-HA.11-tagged wild-type and glycosylation-deficient mutants of hPAR 1 E were immunoprecipitated using a ProFound HA.11 immunoprecipitation kit as per the manufacturer's protocol. Protein samples (20 g) were separated on a 10% SDS-polyacrylamide gel before transfer to Hybond C PVDF membrane. The membrane was subsequently blocked with PBS containing 5% nonfat milk before incubation overnight at 4°C with the mouse monoclonal HA.11 antibody (1:1000 dilution in PBS/Tween 20 (0.1%) containing 2% nonfat milk). Blots were washed with PBS/Tween 20 (0.1%) for 1 h, replacing the buffer every 15 min, before a final incubation with the peroxidase-conjugated goat anti-mouse IgG (1:4000 in PBS/Tween 20 (0.1%) containing 2% nonfat milk) for 1 h. Following a further wash in PBS/Tween 20 (0.1%) for 1 h, bands were visualized using ECL (Amersham Biosciences).
Calcium Signaling Assay-The calcium signaling assay was performed as we described previously (19). Functional receptor activity was assessed by challenging cells with the PAR 1 activating peptide TFLLR-NH 2 (100 M) and thrombin (5 nM). To assess whether a test proteinase was amputating the tethered ligand from the receptor (disarming), cells were first challenged by the addition of the test proteinase for 2 min. Successful amputation of the tethered ligand was then monitored by a subsequent application of thrombin (5 nM). The calcium signal in response to thrombin will be ablated if the tethered ligand has been removed from the receptor by the test proteinase. To demonstrate that the disarmed PAR 1 is still functional and present at the cell surface, 100 M TFLLR-NH 2 was applied after thrombin challenge. A calcium signal triggered by the TFLLR-NH 2 is indicative of a functional PAR 1 at the cell surface, which is unresponsive to thrombin.
Calculations and Statistical Analysis-Data were analyzed using GraphPad Prism software. Statistical tests were carried out depending on the specifics of the data sets to be compared. When two cell lines subjected to the same treatment were compared, a paired Student's t test was adopted. The comparison of a group of cell lines subjected to the same treatment was performed by one-way analysis of variance followed by Tukey's multiple comparison test. In order to assess the changes over a period of time using a specific treatment or to assess a change over a concentration range between two different cell lines, two-way analysis of variance tables were adopted.

cell clones generated for this study
In order to obtain permanent receptor-expressing cell lines, cells expressing high levels of hPAR 1 were isolated by FACS using the ATAP-2 anti-PAR 1 antibody. The glycosylation mutants were named with an N followed by a number related to the relative position of the potential glycosylation site and a Q. The alanine mutations were created at amino acid position Phe 55 -Trp 56 or Val 72 -Ser 73 of N62Q/N75Q to generate cathepsin G mutant (F55A/W56A) or proteinase 3 mutant (V72A/S73A).  lation sequons on the N terminus (N35Q/N62Q/N75Q) still resulted in nearly 50% cell surface expression compared with WT expression (percentage relative to WT Ϯ S.E.: 44.7 Ϯ 1% (n ϭ 3)). Removing all glycosylation sequons resulted in a receptor (N35-259Q) that was expressed at a level similar to that of the N250Q/N259Q mutant (percentage relative to WT Ϯ S.E.: 6 Ϯ 1% (n ϭ 3)). Analysis of eYFP-tagged hPAR 1 localization in permeabilized cells by confocal microscopy demonstrated that all hPAR 1 E types (WT and glycosylation-deficient mutant hPAR 1 ) with the exception of N35-259QhPAR 1 E were expressed on the cell surface; however, different levels of membrane localization and cytosolic retention were observed (Fig. 3). There was no detectable fluorescence (eYFP signaling) observed in EV cells (Fig. 3A). WT hPAR 1 E displayed strong cell membrane expression with little receptor observed in the cytoplasm (Fig.  3B). Removal of sequon Asn 35 in the receptor N terminus resulted in the appearance of receptor within the cytosol (Fig.  3C). Removal of both sequons (Asn 62 and Asn 75 ) after the tethered ligand of hPAR 1 resulted in a large amount of receptor being found in the cytosol (Fig. 3D). The relative intensity of the receptor membrane localization and global cellular distribution appeared to be reduced significantly after removal of ECL2 glycosylation sequons (Asn 250 and Asn 259 ) (Fig. 3E). Removal of all hPAR 1 glycosylation sequons (Asn 35 , Asn 62 , Asn 75 , Asn 250 , and Asn 259 ) resulted in a dramatic loss of cell surface expression with the highest level of receptor cytosolic retention (Fig. 3F).

Name of clone
hPAR 1 Is a Heavily Glycosylated and Sialylated Receptor-All of the glycosylation-deficient mutant receptors displayed different degrees of loss in molecular mass; no bands were detected in EV cells (Fig. 4A). The molecular mass of WT hPAR 1 E migrated as a broad band ranging from ϳ37 to ϳ250 kDa, with the majority of the receptor being observed from ϳ75 to ϳ220 kDa. This is a pattern similar to that of N62QhPAR 1 E. N62Q/N75QhPAR 1 E migrated from ϳ37 to ϳ120 kDa. N35Q/ N62Q/N75QhPAR 1 E migrated with a molecular mass from ϳ37 to ϳ200 kDa, with the majority of the receptor being observed from ϳ75 to ϳ120 kDa. The predicted molecular mass of the fused hPAR 1 eYFP is approximately ϳ70 to ϳ80 kDa (hPAR 1 is ϳ47 kDa, and eYFP protein is ϳ31 kDa). We proposed that the band at ϳ75 kDa may represent an unglycosylated monomer of PAR 1 ; the bands around ϳ80 to ϳ130 kDa may represent a glycosylated monomer of PAR 1 ; and the bands around ϳ150 to ϳ220 kDa may represent glycosylated and unglycosylated dimers of PAR 1 (Fig. 4). When removing all three glycosylation sites (Asn 35 , Asn 62 , and Asn 75 ) in the N terminus of the receptor, the banding pattern changed, with only glycosylated dimer and unglycosylated monomer present. Removal of glycosylation at sites Asn 62 and Asn 75 produced a single band at ϳ120 kDa representing a glycosylated monomer of PAR 1 . Only one band migrated at ϳ37 kDa from both N250Q/N259QhPAR 1 E and N35-259QhPAR 1 E. A number of minor bands that may represent either eYFP protein or proteolytic degradation products of hPAR 1 were routinely observed around ϳ37 kDa in all lanes.
To determine the contribution of sialic acid to the molecular mass of hPAR 1 , we performed Western blot analysis for WT hPAR 1 E-transfected CHO cells (Pro5 and Lec2) (Fig. 4B). No bands were detected in the non-transfected Lec2 and Pro5. Pro5-PAR 1 E migrated as a broad high molecular weight band ranging from ϳ25 to ϳ200 kDa; the molecular mass of Lec2-PAR 1 E ranged from ϳ25 to ϳ150 kDa. Two bands were routinely observed in the ϳ25 to ϳ37 kDa range in both Pro5-PAR 1 E and Lec2-PAR 1 E.
The N-Linked Glycosylation Located after the Tethered Ligand of hPAR 1 Regulates hPAR 1 Signaling-To investigate whether N-linked glycosylation at individual residues had a role in regulating receptor function, we constructed calcium signaling concentration-effect curves for thrombin, trypsin, and TFLLR-NH 2 in WT hPAR 1 and hPAR 1 glycosylation mutant KNRK cell lines. The EC 50 values of WT hPAR 1 and mutant receptors for the calcium signaling assay are displayed in Table 2. N62Q/N75Q concentration effect curves were compared with WT-M because they displayed similar cell surface expression (p Ͼ 0.05) (Fig. 5A). TFLLR-NH 2 displayed a slight rightward shift in activating N62Q/N75Q compared with WT-M, stimulating a calcium signal at 3-100 M for WT-M and 10 -100 M for N62Q/N75Q, respectively, but achieved the same maximum response of 42 Ϯ 1.5% of A23187 at 100 M in both cell lines (Fig. 5B). For N62Q/N75Q and WT-M, thrombin induced a calcium signal from 0.05 to 5 nM (Fig. 5C). However, the magnitude of the responses to thrombin in N62Q/N75Q was smaller than those obtained for WT-M (24 Ϯ 2.6% of A23187 and 31 Ϯ 1.8% of A23187, respectively) (p Ͼ 0.05). The EC 50 value for thrombin in N62Q/N75Q was 0.3 nM, and the EC 50 value in WT-M was 0.12 nM. Interestingly, the N62Q/ N75Q displayed reduced sensitivity toward trypsin compared with WT-M, and the maximal response achieved was only 14% of A23187 compared with 32 Ϯ 3% of A23187 for the WT-M (p Ͻ 0.001) (Fig. 5D). In addition, the EC 50 value for trypsin in N62Q/N75Q was significantly greater than WT-M (42 and 26 nM, respectively). However, further analysis of N62Q and N75Q    1 Enhances the Ability of Trypsin to Disarm the Receptor-Having found that trypsin shows reduced activity toward N62Q/N75Q by a calcium signaling assay, we then explored the ability of trypsin to disarm N62Q/N75Q, N62Q, and N75Q. Fig. 6A shows the TFLLR-NH 2 (100 M) responses following trypsin (3, 10, 30, 100, and 300 nM) and then thrombin (5 nM) challenge in WT-M and N62Q/N75Q cells (i.e. trypsin and then thrombin and then TFLLR-NH 2 ). For WT-M, the TFLLR-NH 2 responses remained the same after challenging the cells with trypsin from 3 to 300 nM. Challenging the N62Q/N75Q cells with 3 nM trypsin followed by 5 nM thrombin and then 100 M TFLLR-NH 2 triggered a calcium signal that was similar to that seen for WT-M cells. However, a robust increase in the magnitude of TFLLR-NH 2 -triggered calcium signal was observed after challenging the N62Q/N75Q cells with 300 nM trypsin, which was over 4-fold greater than that observed for WT-M cells (p Ͻ 0.05).
We next assessed whether individual glycosylation at Asn 62 or Asn 75 is responsible for regulating trypsin disarming of hPAR 1 (Fig. 6, B and C). Challenging N62Q cells with 5 nM thrombin or 100 M TFLLR-NH 2 resulted in a calcium response of 61.4 Ϯ 6.7% of A23187 and 62 Ϯ 6.9% of A23187, respectively (Fig. 6B). Challenging N62Q cells with 100 nM trypsin resulted in an observable calcium signal (32 Ϯ 8% of A23187). The addition of thrombin following trypsin challenge resulted in a marked reduction (p Ͻ 0.001) in the calcium response to thrombin compared with the response obtained with thrombin added alone (7.8 Ϯ 1% of A23187 versus 61.4 Ϯ 6.7% of A23187). The addition of TFLLR-NH 2 following trypsin and thrombin resulted in a response that was slightly increased (p Ͼ 0.05) in magnitude compared with the response obtained to TFLLR-NH 2 following just thrombin addition (17.5 Ϯ 3.6% of A23187 versus 9 Ϯ 3% of A23187) but was significantly reduced (p Ͻ 0.001) compared with the response of TFLLR-NH 2 added alone (17.5 Ϯ 3.6% of A23187 versus 62 Ϯ 6.9% of A23187).
Challenging N75Q cells with trypsin only resulted in a response of 8 Ϯ 1% of A23187 (Fig. 6C). Challenging N75Q cells with TFLLR-NH 2 or thrombin alone resulted in a calcium response of 27 Ϯ 2% of A23187 and 12 Ϯ 0.7% of A23187, respectively. The addition of TFLLR-NH 2 following thrombin addition resulted in a calcium response of 12 Ϯ 1.7% of A23187. The addition of thrombin following the addition of trypsin resulted in no calcium response. Finally, challenging the cells with TFLLR-NH 2 following trypsin and thrombin addition triggered a calcium response of 21 Ϯ 2.7% of A23187, which was of a magnitude comparable (p Ͼ 0.05) with the response obtained with TFLLR-NH 2 alone.
Elastase, Proteinase 3, and Thermolysin but Not Cathepsin G Displayed Enhanced Ability to Disarm hPAR 1 (N62Q/N75Q)-Effective proteinase receptor disarming was assessed by thrombin activation of the cells following proteinase addition. For N62Q/N75Q and WT-M, significant differences (p Ͻ 0.0001) in responses to 5 nM thrombin following the addition of elastase at different concentrations were observed (Fig. 7A). For WT-M, no changes were observed in thrombin-triggered calcium responses after the addition of 1 nM elastase, and the thrombintriggered calcium response was totally ablated after the addition of 15 nM elastase. The thrombin response curve for N62Q/ N75Q shifted left half a log when compared with WT-M. For N62Q/N75Q, the calcium signaling response to thrombin started to decrease after challenging the cells with 1 nM elastase (ϳ80% of control thrombin response). The thrombin-triggered calcium response was totally ablated after challenging cells with 5 nM elastase.
For N62Q/N75Q and WT-M, no significant differences (p Ͼ 0.05) in responses to 5 nM thrombin following the addition of cathepsin G from 10 to 300 nM were observed (Fig. 7B). The thrombin-triggered calcium response was totally ablated after the addition of 300 nM cathepsin G for both N62Q/N75Q and WT-M.
For N62Q/N75Q and WT-M, significant differences (p Ͻ 0.0001) in responses to 5 nM thrombin following the addition of proteinase 3 were observed (Fig. 7C). For WT-M, no changes were observed in the thrombin-triggered calcium response following the addition of 1000 nM proteinase 3. In contrast, significant decreases in the response to thrombin were observed in N62Q/N75Q following the addition of proteinase 3 from 1 to 1000 nM. The thrombin-triggered calcium signal was totally abolished after the addition of 1000 nM proteinase 3.
For N62Q/N75Q and WT-M, significant differences (p Ͻ 0.0001) in responses to 5 nM thrombin following the addition of thermolysin at different concentrations were observed (Fig.  7D). The thrombin response curve after adding thermolysin at increasing concentration for N62Q/N75Q shifted 30-fold to the left when compared with WT-M. For N62Q/N75Q, the calcium signaling response to thrombin started to decrease after application of thermolysin at 3 nM (ϳ80% of control thrombin response). The thrombin-triggered calcium response was totally ablated after challenging N62Q/N75Q with 300 nM thermolysin. However, for WT-M, the thrombin-triggered calcium responses remained elevated at thermolysin concentrations up to 30 nM. The calcium signaling response to thrombin started to decrease following the 100 nM thermolysin addition (ϳ80% of thrombin control). The thrombin-triggered calcium response in WT-M was totally ablated after the addition of 1000 nM thermolysin.
Molecular Evidence That Cathepsin G and Proteinase 3 Are Disarming hPAR 1 (N62Q/N75Q)-The addition of 100 nM cathepsin G to N62Q/N75Q cells before the addition of 5 nM thrombin markedly reduced (p Ͻ 0.001) the response to thrombin compared with the response obtained with thrombin added alone (11 Ϯ 2% of A23187 and 30 Ϯ 5% of A23187, respectively) (Fig. 8A). In contrast, the prior addition of cathepsin G failed to reduce the thrombin calcium signal in cathepsin G mutant cells when compared with the control (32.5 Ϯ 8% of A23187 and 36.5 Ϯ 8% of A23187, respectively) (p Ͼ 0.05).
No observable calcium response to thrombin was observed when 300 nM proteinase 3 had been added to the N62Q/N75Q cells prior to thrombin challenge (Fig. 8B). In contrast, the prior addition of proteinase 3 failed to reduce the thrombin calcium signal in the proteinase 3 mutant cells, when compared with the response obtained with thrombin added alone (13 Ϯ 2% of A23187 and 12.33 Ϯ 2.5% of A23187, respectively) (p Ͼ 0.05).
Elastase, Cathepsin G, Proteinase 3, and Thermolysin Mainly Disarm hPAR 1 (N75Q)-The ability of thermolysin and neutrophil serine proteinases to activate/disarm N62Q and N75Q was investigated (Fig. 9). Challenging N62Q cells with 5 nM thrombin or 100 M TFLLR-NH 2 resulted in a calcium response of 64.7 Ϯ 7% of A23187 and 67.8 Ϯ 5% of A23187, respectively (Fig. 9A). In addition, challenge with TFLLR-NH 2 following the thrombin addition resulted in a calcium response of 9 Ϯ 3% of A23187. Thermolysin (50 nM) had no observable effect on calcium signaling in N62Q. Challenge with thrombin following the thermolysin addition resulted in a calcium response at 53.5 Ϯ 4.6% of A23187, which is comparable with that obtained with thrombin added alone (64.7 Ϯ 7% of A23187). The addition of TFLLR-NH 2 following thermolysin and thrombin resulted in a similar calcium response to the response obtained with TFLLR-NH 2 following just thrombin challenge (13.6 Ϯ 4% of A23187 and 9 Ϯ 3% of A23187, respectively). The addition of 1.5 nM elastase had no observable effect on calcium signaling in N62Q cells. The addition of thrombin following elastase addition resulted in a calcium response of 49 Ϯ 3.5% of A23187. The addition of TFLLR-NH 2 following the disarming with elastase and thrombin resulted in a similar calcium response to the TFLLR-NH 2 response obtained following just thrombin alone 12 Ϯ 1% of A23187 and 9 Ϯ 3% of A23187, respectively). The addition of 100 nM cathepsin G had no observable effect on calcium signaling in N62Q cells. After challenging N62Q cells with cathepsin G, the addition of thrombin resulted in a small decrease in the calcium response compared with that obtained with thrombin added alone (39 Ϯ 3.8% of A23187 and 64.7 Ϯ 7% of A23187, respectively). The addition of TFLLR-NH 2 following the addition of thermolysin and thrombin triggered a calcium response at 16 Ϯ 6% of A23187 that was slightly larger than that obtained with TFLLR-NH 2 following thrombin alone (9 Ϯ 3% of A23187). The addition of 300 nM proteinase 3 had no observable effect on calcium signaling in N62Q cells. Challenge with thrombin following proteinase 3 addition resulted in a calcium response of 67 Ϯ 6.7% of A23187, which is comparable with that obtained when thrombin was added alone (64.7 Ϯ 7% of A23187). The addition of TFLLR-NH 2 following the addition of  proteinase 3 and thrombin resulted in a calcium response similar to that obtained with TFLLR-NH 2 following the thrombin addition (11.6 Ϯ 4% of A23187 and 9 Ϯ 3% of A23187, respectively).
Challenging N75Q cells with thrombin or TFLLR-NH 2 resulted in a calcium response of 12 Ϯ 0.7% of A23187 and 27 Ϯ 2% of A23187, respectively. The addition of TFLLR-NH 2 following thrombin challenge resulted in a calcium response of 13.6 Ϯ 1% of A23187 (Fig. 9B). No observable calcium response was detected after challenging N75Q cells with thermolysin. After challenging N75Q cells with thermolysin, the addition of thrombin resulted in a calcium response that was comparable with that obtained to thrombin alone (10 Ϯ 1% of A23187 and 12 Ϯ 0.7% of A23187, respectively), and finally challenging the cells with TFLLR-NH 2 triggered a calcium response of 21 Ϯ 2% of A23187. No observable calcium response was detected after challenging N75Q cells with elastase or with thrombin following the addition of elastase. In addition, the response to TFLLR-NH 2 following elastase and thrombin challenge was the same as that obtained with TFLLR-NH 2 alone (28 Ϯ 1% of A23187 versus 27 Ϯ 2% of A23187).
The addition of cathepsin G had no observable effect on calcium signaling in N75Q cells. Challenging cells with thrombin following the addition of cathepsin G resulted in a calcium response of 10 Ϯ 1% of A23187, which was comparable with that obtained with thrombin alone (12 Ϯ 0.7% of A23187). The addition of TFLLR-NH 2 following the addition of cathepsin G and thrombin triggered a calcium response, which was of a magnitude similar to that obtained with TFLLR-NH 2 following thrombin alone (17 Ϯ 1 and 13.6 Ϯ 1% of A23187, respectively). The addition of proteinase 3 had no observable effect on calcium signaling in N75Q cells. The addition of thrombin after challenging N75Q cells with proteinase 3 resulted in a very small decrease in calcium response compared with that obtained with thrombin alone (10 Ϯ 0.5% of A23187 and 12 Ϯ 0.7% of A23187, respectively). Finally, challenging the cells with TFLLR-NH 2 following the proteinase 3 and thrombin addition triggered a calcium response of 20.5 Ϯ 0.5% of A23187.

DISCUSSION
In the present study, we have demonstrated for the first time that N-linked glycans (Asn 62 and Asn 75 ) after the tethered ligand on the N terminus of hPAR 1 regulate receptor signaling to trypsin, thermolysin, and the neutrophil proteinases elastase and proteinase 3 but not cathepsin G. In addition, we reported that hPAR 1 is heavily N-linked glycosylated and sialylated in epithelial cell lines (KNRK and Pro5 cells), and glycosylation occurs at all five consensus sites, namely Asn 35 , Asn 62 , Asn 75 , Asn 250 , and Asn 259 . Removing N-linked glycosylation sequons at these positions affected hPAR 1 cell surface expression to varying degrees, and N-linked glycosylation at ECL2 (especially Asn 250 ) of hPAR 1 might be essential for optimal receptor cell surface expression and receptor stability.
N-Linked glycosylation, which adds oligosaccharides to the nitrogen in the side-chain amide of asparagine residues, requires the asparagines in a consensus sequence of Asn-X-Ser/ Thr (where X is any amino acid except proline) (22,23). However, not all consensus sites are actually glycosylated because the oligosaccharide might be trimmed and elaborated during transit through the endoplasmic reticulum and the Golgi apparatus (23). Disrupting the glycosylation sites by mutagenesis, individually and/or in combination, allowed us to investigate the status and even the function of each specific putative N-linked glycosylation site.
We first examined the role of N-linked glycosylation in regulating the receptor cell surface expression and the receptor global distribution in KNRK cells. Interestingly, unlike in previous studies in HeLa (24), there is still a very high level of cell surface expression of hPAR 1 when removing all glycosylation sites on the N terminus. The ECL2 single glycosylation mutant N250Q displayed the lowest level of cell surface expression of hPAR 1 among all of the other single glycosylation mutants; furthermore, removing all glycosylation sequons on ECL2 resulted in similar low levels of receptor cell surface expression to the non-glycosylation mutant, where all glycosylation sequons had been deleted. N-Linked glycosylation of hPAR 1 on ECL2 (especially Asn 250 ) versus glycosylation in the N terminus might therefore be more important for receptor optimal cell surface expression. Glycosylation on ECL2 of PAR 2 is also important for enhancing receptor expression but not regulating receptor signaling; more importantly, the ECL2 region is extremely sensitive to mutations, and mutating residues in and around the glycosylation sequon, even neutral mutations such as alanine (Ala) in the ECL2 glycosylation sequon, alters receptor structure and function (19,24). The PAR 1 glycosylation study on HeLa cells (24) reported that N-terminal N-linked glycosylation of PAR 1 versus ECL2 N-linked glycosylation is critical for the regulation of PAR 1 signaling and trafficking. In this study, mutants were generated by substituting Ala for Glu. It is therefore hard to determine whether the altered hPAR 1 cell surface expression and signaling in HeLa cells are a result of the loss of N-linked glycosylation or due to the Ala substitution altering receptor structure and function. We have generated our N-linked glycosylation mutants by substituting Asn for Glu at all five potential N-linked glycosylation sites; Glu is chemically and structurally more similar to Asn. For this reason, we propose that this substitution itself will not lead to structural and functional differences between the mutants and wild-type hPAR 1 , and any differences detected should be due to the loss of glycosylation.
Using confocal microscopy, we demonstrated that the glycosylation-deficient mutants were abundant in the cytosol. Removal of a single N-terminal glycan resulted in observable receptor cytosolic retention, but removal of all glycosylation sites produced a receptor that could mainly be seen in the cytosol with minimal cell surface expression. Therefore, we conclude that N-linked glycosylation of hPAR 1 at all sites might be required for the optimal receptor cell surface expression (24,25).
Similar to the study of Soto and Trejo (24), we have demonstrated that in KNRK cells, hPAR 1 is heavily glycosylated, and the glycosylation occurs at all five consensus sites. Removal of N-linked glycosylation from ECL2 or removal of all N-linked glycans from the receptor resulted in only a single Western blot band visualized at around 37 kDa compared with wild type or the N terminus mutants, which produced a second band at a lower molecular weight. It is well known that N-linked glycosylation can regulate receptor conformation, folding, or stability and even protect receptors from proteolytic degradation. The bands shown around ϳ37 kDa may represent either eYFP protein or deglycosylation and/or proteolytic degradation products of hPAR 1 (5, 14, 24, 26 -28). We thus propose that N-linked glycosylation, especially on ECL2 (Asn 250 and Asn 259 ), of hPAR 1 might be vital for receptor stability and even protect hPAR 1 from endogenous proteolytic degradation. Furthermore, we expressed the hPAR 1 in CHO cells (Pro5 and Lec2) in order to assess the sialylation status of hPAR 1 . The Pro5 cell line is the parent clone for the Lec2 cell line, which has a substantial loss in the ability to attach sialic acid to the terminal positions on oligosaccharides. Western blot analysis of hPAR 1 in Lec2 cells revealed a protein that had a molecular mass that was up to ϳ47 kDa lower than that observed for hPAR 1 expressed in the Pro5 cells. Presumably, this significant loss of ϳ47 kDa is due to the loss of receptor-associated sialic acid, and we can therefore conclude that hPAR 1 in Pro5 cells, at least, is a heavily sialylated receptor.
In order to establish the role of N-linked glycosylation on the N terminus of hPAR 1 in regulating receptor function, agonist concentration effect curves were produced for the glycosylation-deficient hPAR 1 cell lines and their respective WT hPAR 1 cell line to evaluate the coupling of the mutant receptors to calcium. Receptor cell surface expression for each mutant cell line was compared and matched by FACS to WT hPAR 1 cells in order to ensure that any functional differences detected were not due to differences in receptor cell surface expression (19). Except for the N62Q/N75Q mutant, no significant changes in receptor function were detected when compared with their respective matching WT hPAR 1 for all mutant cell lines. This N62Q/N75Q mutant displayed a slight reduction in sensitivity toward lower concentrations of TFLLR-NH 2 and thrombin, although statistical analysis showed no significant differences in EC 50 value and receptor maximum calcium signaling response toward 100 M TFLLR-NH 2 or 5 nM thrombin. The biggest reduction in agonist sensitivity was observed with trypsin activation. PAR 1 coupling to calcium signaling appears not to be affected by the lack of glycosylation on the receptor. Even the N35-259Q mutant cell line still displayed a detectable calcium signal toward TFLLR-NH 2 and thrombin (data not shown), thus adding strength to this conclusion.
In our experiments, trypsin triggered calcium signaling in wild-type hPAR 1 and its mutants to varying degrees. However, the TFLLR-NH 2 calcium signaling response after the addition of trypsin and thrombin is much greater in N62Q/N75Q mutants compared with wild-type hPAR 1 . This just confirmed that trypsin disarms the mutant receptor, and the disarmed mutant receptors still remained on the cell surface but could not be activated by thrombin anymore because the tethered ligand had been removed; however, TFLLR will still activate those mutant receptors by binding to receptor ECL2. Trypsin can activate PAR 1 by cleaving at the activating cleavage site Arg 41 -Ser 42 to expose the tethered ligand (2,4,29). Moreover, trypsin has also been reported to cleave PAR 1 at residues Arg 70 -Leu 71 and Lys 82 -Gln 83 , which amputates the tethered ligand from the receptor (4) (Fig. 1). Furthermore, these residues are suggested to be cleaved by trypsin much faster than the residues Arg 41 -Ser 42 (9). Given that these two potential trypsin cleavage sites are located close to Asn 75 , it is not surprising that in the N62Q/N75Q and N75Q mutants, trypsin can gain easy access to these cleavage sites and therefore disarm these mutant receptors. In contrast, when PAR 1 is fully glycosylated at Asn 75 , these cleavage sites are subsequently shielded by the attached oligosaccharide, and thus trypsin disarms WT hPAR 1 less efficiently. Interestingly, the previous study by Nakayama et al. (9) reported that 100 nM trypsin cleaved PAR 1 but did not activate it in HUVEC cells, thus suggesting receptor disarming. We therefore assume that the glycosylation status of PAR 1 in HUVEC cells must be different from our PAR 1 -KNRK system. Indeed, previous studies reported that PAR 1 expressed in HUVECs and platelets migrated as a homogenous species with an apparent mass of 66 kDa (27). This molecular mass is considerably below that reported in other cell systems, including ours, and suggests that PAR 1 is not so heavily glycosylated in these cell types. Therefore, we suggest that the Asn 62 and Asn 75 might not be fully glycosylated in HUVECs, and therefore these potential trypsin cleavage sites around Asn 75 are more readily available for trypsin cleavage and subsequent disarms.
We have also shown that the neutrophil proteinases elastase, cathepsin G, and proteinase 3 do not activate PAR 1 in our KNRK cell system but rather disarm the receptor. Like us, Renesto et al. (13) reported that elastase, cathepsin G, and proteinase 3 can cleave and inactivate PAR 1 in platelets and human endothelial cells. In contrast, Suzuki et al. (8) reported that elastase can activate PAR 1 in human lung epithelial cells and also mentioned that proteolytic cleavage of PAR 1 by elastase might activate or inactivate the receptor depending on the site of cleavage. The authors also suggested that the extent of glycosylation of the receptor could influence the site and extent of cleavage of the receptor (8).
In our current study, we note that elastase disarms N62Q/ N75Q better than WT hPAR 1 at the same concentration, and 1.5 nM elastase disarms N75Q, but not N62Q. The mass spectrometry study by Renesto et al. (13) indicated that elastase cleaves the N terminus of hPAR 1 at Val 72 -Ser 73 and Ile 74 -Asn 75 , which are located after the tethered ligand. Furthermore, Loew et al. (4) reported that elastase cleaved the PAR 1 exodomain at Ala 36 -Thr 37 , Val 72 -Ser 73 , and Ala 86 -Phe 87 sites with preferential cleavage at Val 72 , which would disable the receptor. Therefore, after amputating the Asn 75 glycosylation, the potential cleavage sites Val 72 -Ser 73 might be more easily accessible for elastase cleavage.
Because cathepsin G has been reported to have both activating and disarming actions on PARs (12,13,30,31), we tested the role of N-terminal glycosylation of hPAR 1 in regulating this proteinase. We have demonstrated that cathepsin G could clearly disarm N62Q/N75Q and WT hPAR 1 to the same efficiency. In addition, our data also revealed that cathepsin G at 100 nM disarms N62Q and N75Q. Like us, others have also reported that cathepsin G at 30 nM could disarm PAR 1 (30). Molino et al. (12) reported that in addition to the Arg 41 -Ser 42 site, cathepsin G cleaves PAR 1 at Phe 43 -Leu 44 and Phe 55 -Trp 56 , removing the tethered ligand and rendering the receptor unresponsive to thrombin. In 1997, the mass spectrometry study of Renesto et al. (13) indicated that cathepsin G cleaves PAR 1 on platelets and endothelial cells downstream of the thrombin cleavage site at Phe 55 -Trp 56 and Tyr 69 -Arg 70 . Thus, we then concluded that cathepsin G might have four potential cleavage sites in the N terminus of hPAR 1 (Arg 41 -Ser 42 , Phe 43 -Leu 44 , Phe 55 -Trp 56 , and Tyr 69 -Arg 70 ). Furthermore, the kinetic analyses of Loew et al. (4) showed that the preferential cathepsin G cleavage site for PAR 1 is the Phe 55 -Trp 56 site. Moreover, our finding with the cathepsin G mutant is in accord with a previous report that has demonstrated that Phe 55 -Trp 56 is the main region cleaved within the receptor N terminus by cathepsin G (12). Also, our Western blot data revealed that Asn 62 is not heavily glycosylated in our KNRK system. Therefore, we suggest that cathepsin G sensitivity toward hPAR 1 is not regulated by N-linked glycosylation.
We have also assessed proteinase 3 inhibition of thrombininduced Ca 2ϩ mobilization in WT hPAR 1 , N62Q/N75Q, N62Q, and N75Q. Our data revealed that 300 nM proteinase 3 disarms N62Q/N75Q but not WT hPAR 1 , N62Q, or N75Q. It is well established that proteinase 3 has an elastase-like specificity for Ala, Ser, and Val at the P1 site (32). One previous study on human oral epithelial cells (33) also pointed out that proteinase 3 can rapidly cleave PAR 2 between Arg and Ser and relatively inefficiently cleave between Lys and Val. Furthermore, proteinase 3 cleaved the peptide corresponding to the N terminus of PAR 2 at Arg 36 -Ser 37 , indicating that the site of the PAR 2 is structurally accessible by proteinase 3 (33). Interestingly, one previous study on synthetic peptides (13) reported that the cleavage site for proteinase 3 in the N terminus of PAR 1 is Val 72 -Ser 73 . In addition, the kinetic analysis of Loew et al. (4) on the N terminus domain of PAR 1 showed that proteinase 3 early cleavage sites for PAR 1 are Ala 36 -Thr 37 , Pro 48 -Asn 49 , Val 72 -Ser 73 , and Ala 92 -Ser 93 , and the late cleavage site is Pro 54 -Phe 55 . Furthermore, Sokolova and Reiser (29) suggested in a recent review that proteinase 3 cleaves human PAR 1 at Ala 36 -Thr 37 and Val 72 -Ser 73 . Although there are differences, previous studies have agreed that the Val 72 -Ser 73 might be the earliest cleavage site for proteinase 3. We therefore altered that site by creating a mutation, Ala 72 -Ala 73 , in our N62Q/N75Q (named the proteinase 3 mutant), and indeed the proteinase 3 almost immediately lost the ability to disarm the mutant receptor. As mentioned by Loew et al. (4), there exists a late cleavage site between Pro 54 and Phe 55 ; we therefore suggest that apart from the Val 72 -Ser 73 cleavage site, there might exist relatively inefficient later cleavage events between Pro 54 and Phe 55 and possibly at Pro 48 -Asn 49 . Thus, there might be a possibility that if we expose PAR 1 to proteinase 3 for a greater length of time, proteinase 3 might disarm WT hPAR 1 . Indeed, we did notice that this phenomenon happened when we preincubated PAR 1 with proteinase 3 for more than 3 min (data not shown). We have shown that in WT hPAR 1 cells, proteinase 3 does not disarm the receptor immediately (within 1 min); however, after the removal of the glycans in Asn 62 and Asn 75 , proteinase 3 immediately disarmed the glycosylation mutant receptor and inhibited thrombin-induced calcium mobilization. We therefore suggest that glycosylation of Asn 62 and Asn 75 together regulate hPAR 1 signaling to proteinase 3.
Our data also revealed that thermolysin disarms N62Q/ N75Q with 30-fold better efficiency than WT hPAR 1 . One previous study in insect SF9 cells (34) reported that thermolysin may represent an important mechanism of rapid receptor deactivation of the human thrombin receptor. Furthermore, like us, one previous study on PAR 1 in astrocytes (35) found that treatment with thermolysin generated a thrombin-insensitive receptor, whereas the response to the activating peptide was not affected. In the literature, it is well established that the predicted thermolysin cleavage sites in PAR 1 are within the tethered ligand domain (SFLLR) at Phe 43 -Leu 44 and Leu 44 -Leu 45 , suggesting that thermolysin would be expected to remove the exodomain of PAR 1 and destroy the tethered ligand and thus disarm the receptor (16,36). Interestingly, in our study system, removing both of the glycosylation sequons (Asn 62 and Asn 75 ) resulted in a receptor that was more efficiently disarmed by thermolysin, whereas removal of glycosylation at the Asn 62 sequon alone from the receptor had no observable effect on thermolysin disarming. However, the N75Q did display increased disarming compared with WT hPAR 1 , thus suggesting that this sequon regulates PAR 1 disarming by thermolysin. Therefore, we predict that, apart from the well established cleavage sites, there might be some other thermolysin cleavage sites in the N terminus of PAR 1 , especially in the region of the glycosylation site Asn 75 . In theory, bacterial thermolysin cleaves peptide bonds N-terminally to the hydrophobic residues leucine and isoleucine, with some specificity for phenylalanine and valine (36). Therefore, we suggest that Leu 71 -Val 72 in PAR 1 may also be a potential thermolysin cleavage site, and it is very close to the Asn 75 site. When amputating the Asn 75 glycan, this potential thermolysin cleavage site (Leu 71 -Val 72 ) will be more readily available for thermolysin cleavage.
In summary, hPAR 1 is a heavily glycosylated receptor and is important for cell surface expression, receptor stability, and function. We show here for the first time that glycosylation in the N terminus of hPAR 1 downstream of the tethered ligand (especially Asn 75 ) governs receptor disarming to trypsin, thermolysin, and the neutrophil proteinases elastase and proteinase 3 but not cathepsin G. Thus, we concluded that N-linked glycosylation at the N terminus of hPAR 1 plays a vital role in regulating the susceptibility to proteinase disarming of the receptor.