Cytoplasmic Domain of P-selectin Glycoprotein Ligand-1 Facilitates Dimerization and Export from the Endoplasmic Reticulum^*

Antibodies

Rabbit polyclonal IgG antibody against a 19-residue peptide corresponding to the N-terminal sequence of mature murine PSGL-1 (20) was a gift from Richard Cummings (Emory University School of Medicine, Atlanta, GA). All other antibodies were from BD Biosciences unless noted otherwise.

Mice

Knock-in ΔCD mice expressing PSGL-1 without the cytoplasmic domain (20) and knock-out mice lacking core 1 β1–3-galactosyltransferase in endothelial and hematopoietic cells (EHC T-syn^−/−) have been described (11, 23). Mice of both genotypes were backcrossed at least 10 generations into the C57BL/6J background. WT C57BL/6J mice were used as controls. ΔCD mice were crossed with C57BL/6J mice to generate heterozygotes bearing one allele encoding WT PSGL-1 and one allele encoding ΔCD PSGL-1. All experiments were performed in compliance with protocols approved by the Institutional Animal Care and Use Committee of the Oklahoma Medical Research Foundation.

Leukocytes

Leukocytes from murine bone marrow or peripheral blood were isolated as described previously (20). CFU-GM colonies from ΔCD bone marrow cells were cultured in MethoCult^TM medium GF M3434 (Stem Cell Technologies, Vancouver, British Columbia, Canada). Splenocytes were isolated by mincing murine spleens and flushing cells through a 100-μm strainer (BD Biosciences) with Hanks' balanced salt solution (HBSS) without Ca²⁺ or Mg²⁺. Red blood cells were lysed as described previously (20).

Transfected CHO Cells

Stably transfected CHO cells expressing matched surface densities of human WT or ΔCD PSGL-1 have been described previously (20). The cDNA encoding murine WT or ΔCD PSGL-1 was used as template to mutate Cys-307 to Ser (C307S). All constructs were confirmed by DNA sequencing. Amplified products were inserted into the expression vector pZeoSV (Invitrogen) and transfected into CHO-K1 cells with FuGENE 6 (Roche Applied Science). Stably transfected clones were selected and sorted for matched PSGL-1 surface expression. C307S and ΔCD C307S constructs were cloned into the pDsRed-monomer-N1 and pEGFP-N1 vectors (Clontech) to fuse DsRed or GFP to the C terminus of each PSGL-1 protein. These constructs were transiently transfected into CHO-K1 cells.

Cell-surface Cross-linking

Stably transfected CHO cells expressing matched surface densities of murine C307S or ΔCD C307S were incubated with BS³ as described previously (18). The cells were then lysed and subjected to Western blotting.

Flow Cytometry

Leukocytes from peripheral blood or bone marrow or CFU-GM colonies were suspended in HBSS containing 0.5% human serum albumin. Cells were incubated with phycoerythrin-labeled rat anti-murine PSGL-1 mAb 2PH1 (PE-2PH1). To identify neutrophils, CFU-GM colonies were also incubated with FITC-labeled rat RG6–8C5 mAb to the murine myeloid differentiation antigen Gr-1 (FITC-Gr-1). In some experiments, 5 × 10⁶ bone marrow leukocytes in 0.5 ml of HBSS were incubated for 30 min at 37 °C with or without 250 μg/ml O-sialoglycoprotein endopeptidase (OSGE, Cedarlane Laboratories Ltd., Hornby, Ontario, Canada). The cells were then fixed in 3.7% paraformaldehyde for 10 min at 4 °C and, in some cases, permeabilized for 15 min at 4 °C with BD Perm/Wash Buffer from the Cytofix/Cytoperm kit (BD Biosciences). For all analyses, leukocytes were pretreated with 20 μg/ml mouse Fc block (BD Biosciences) for 20 min at 4 °C. Neutrophils were gated and analyzed on a FACScan using CellQuest Pro software (BD Biosciences).

Northern Blot

Total RNA from splenocytes was isolated using TRIzol (Invitrogen) according to the manufacturer's instructions. RNA integrity was verified by ethidium bromide staining after electrophoresis and quantified by absorbance at 260 nm. Northern blotting was performed as described previously (24). A murine PSGL-1 cDNA labeled with ³²P (PerkinElmer Life Sciences) was used as probe.

Real Time RT-PCR

To measure the expression of PSGL-1, real time PCR was conducted as described previously (25). Briefly, total RNA was isolated from transfected CHO cells expressing matched surface densities of human WT or ΔCD PSGL-1. RNA (5 μg) in a 20-μl volume was reverse-transcribed using random hexamers with the SuperScript II RT system (Invitrogen) as specified by the manufacturer. Quantitative PCR was performed on an ABI Prism 7000 spectrofluorometric thermal cycler (Applied Biosystems, Foster City, CA) using SYBR Green I as a double strand DNA-specific binding dye. PCR primers were chosen using Primer Express 1.7 software. The primer sequences are as follows: 5′-TCCTCCTGTTGCTGATCCTACTG-3′ and 5′-TACTCATATTCGGTGGCCTGTCT-3′ for detection of human PSGL-1; 5′-GCTCTGGCTCCTAGCACCAT-3′ and 5′-CCACCGATCCACACAGAGTACT-3′ for detection of β-actin as an endogenous control. The relative starting concentration of the target in each reaction was quantified according to the relative standard curve method. Data are expressed as an n-fold difference relative to the WT.

Immunofluorescence

All incubations were performed at 4 °C. Bone marrow leukocytes were settled on poly-l-lysine-coated coverslips and fixed in 3.5% paraformaldehyde for 10 min. In some experiments, fixed cells were permeabilized with 0.05% Triton X-100 in HBSS for 10 min. After blocking with 1% BSA in HBSS for 1 h, cells were incubated with 15 μg/ml biotinylated anti-murine PSGL-1 mAb 2PH1 (rat IgG1) in HBSS plus 0.5% BSA for 45 min, followed by streptavidin-Alexa 488 (Molecular Probes, Eugene, OR) for 45 min. Biotinylated rat IgG1 was used as an isotype control. Immunofluorescence was detected with a Leica TCS NT confocal microscope (Leica Microsystems, Deerfield, IL) equipped with a krypton/argon laser. For colocalization studies, leukocytes were stained with 15 μg/ml rabbit anti-calnexin IgG (Santa Cruz Biotechnology, Santa Cruz, CA) followed by a 1:500 dilution of Alexa 546-conjugated goat anti-rabbit IgG (Molecular Probes, Eugene, OR). As controls for the specificity of staining, cells were stained with rabbit anti-calnexin antibody and streptavidin-Alexa 488 or with 2PH1 and goat anti-rabbit IgG. Irrelevant rat and rabbit IgG were also used as controls. Immunofluorescence was detected with the Leica TCS NT confocal microscope. A series of 0.5-μm optical sections in the z axis was analyzed with confocal assistant software (University of Minnesota, Minneapolis).

Western Blots

Bone marrow leukocytes (5 × 10⁶) were lysed in 50 μl of 50 mm Tris-HCl, pH 7.4, 150 mm NaCl, 1 mm PMSF, 1 mm EDTA, 5 μg/ml aprotinin, 5 μg/ml leupeptin, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS. Before lysis, some leukocytes were treated with Pronase (Calbiochem) as described previously (26) or with 10 μg/ml brefeldin A (Sigma) for 90 min. In some experiments, 5 × 10⁶ splenocytes were incubated with 100 μg/ml cycloheximide (Sigma), 8 μm MG132 (Sigma), or an equivalent volume of DMSO in Iscove's modified Dulbecco's medium containing 10% FBS. After various intervals, the splenocytes were lysed in the same buffer used for bone marrow leukocytes. Lysates were centrifuged at 12,000 × g for 10 min to remove insoluble material. Some supernatants (50 μl) were incubated at 37 °C with 20 milliunits of endoglycosidase H (New England Biolabs, Beverly, MA) for 45 min at 37 °C or 250 μg/ml OSGE for 30 min. All supernatants were alkylated in 20 mm iodoacetamide or 30 mm N-ethylmaleimide, boiled in Laemmli sample buffer with or without 3% β-mercaptoethanol, and fractionated by SDS-PAGE. All samples were electrophoresed under reducing conditions except for the data shown in Fig. 7E, where nonreducing conditions were employed. Gels were blotted onto PVDF membranes (Bio-Rad) using a Trans-Blot SD system (Bio-Rad). After blocking in 5% nonfat dry milk, the blots were incubated with rabbit anti-murine PSGL-1 IgG or control rabbit IgG. Peroxidase-labeled anti-rabbit IgG (1:5000, Pierce) and ECL reagents (GE Healthcare) were used to visualize protein bands.

Nondenaturing PAGE

Blue native PAGE was performed according to a protocol supplied by the manufacturer (Invitrogen). Briefly, cells were lysed with 1% n-dodecyl β-d-maltoside on ice. The lysate was centrifuged at 15,000 × g for 30 min at 4 °C. The supernatant was mixed with Native PAGE sample buffer and Coomassie G-250 at a final concentration of 0.125% w/v. Electrophoresis was performed with running buffer containing Coomassie G-250. Gels were transferred to a PVDF membrane. The membranes were incubated in 8% acetic acid to fix the proteins, blocked with 5% milk, and incubated with anti-murine PSGL-1 mAb 4RA10 (27). Peroxidase-labeled anti-murine IgG (1:5000, Pierce) and ECL reagents (GE Healthcare) were used to visualize protein bands.

Pulse-Chase

Pulse-chase experiments were performed as described previously (28), with slight modifications. After starvation in methionine/cysteine-free Dulbecco's modified Eagle's medium for 30 min, murine splenocytes were metabolically labeled with 250 μCi/ml of [³⁵S]methionine/cysteine (1200 Ci/mmol, PerkinElmer Life Sciences) at 37 °C for 30 min. After labeling, cells were washed three times and incubated in medium containing nonradiolabeled cysteine (500 μg/ml) and methionine (100 μg/ml). Cells were lysed at the indicated time in 1 ml of 50 mm Tris-HCl, pH 7.4, 150 mm NaCl, 1 mm PMSF, 1 mm EDTA, 5 μg/ml aprotinin, 5 μg/ml leupeptin, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS. After preclearing with protein A/G-agarose (Santa Cruz Biotechnology), lysates were incubated overnight at 4 °C with polyclonal anti-murine PSGL-1 antibody and then with protein A/G-agarose for 2 h. The immunoprecipitates were washed seven times in 50 mm Tris-HCl, pH 7.4, 150 mm NaCl, 1 mm PMSF, 1 mm EDTA, 5 μg/ml aprotinin, 5 μg/ml leupeptin, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS. They were then analyzed by SDS-PAGE under reducing conditions, followed by fluorography.

FRET

CHO-K1 cells were transiently transfected with constructs encoding C307S-PSGL-1-GFP and C307S-PSGL-1-DsRed or with constructs encoding ΔCD-C307S-PSGL-1-GFP and ΔCD-C307S-PSGL-1-DsRed. After 48 h, FRET between GFP and DsRed was measured by the sensitized acceptor emission method using a Zeiss LSM 510 confocal microscope. Excitation and emission wavelengths were 488 and 505–530 nm, respectively, for GFP, and 543 and 560–615 nm, respectively, for DsRed. For the FRET channel, excitation and emission were 488 and 560–615 nm, respectively. Cells expressing either GFP alone or DsRed alone were imaged to determine the bleed through in the FRET, donor, and acceptor channels. For all samples, images were acquired in three separate channels as follows: acceptor channel (488 excitation/505–530 emission), donor channel (543 excitation/560–615 emission), and FRET channel (488 excitation/560–615 emission). FRET efficiency was calculated as described previously (29), using fluorescence intensities measured in the FRET, donor, and acceptor channels.

Deleting the Cytoplasmic Domain Impairs Expression of PSGL-1 on Surfaces of Leukocytes and Transfected Cells

ΔCD mice express a truncated form of PSGL-1 that lacks the entire 67-residue cytoplasmic domain, except for two arginines that anchor the protein in the membrane (Fig. 1A) (20). Peripheral blood neutrophils from ΔCD mice express 10-fold less PSGL-1 on their surfaces than peripheral blood neutrophils from WT mice (20). We observed a similar 10-fold reduction in surface expression of ΔCD PSGL-1 on neutrophils from bone marrow (Fig. 1B) and from CFU-GM grown in methylcellulose (Fig. 1C). Other leukocyte subsets from both peripheral blood and bone marrow exhibited a comparable reduction in surface expression of ΔCD PSGL-1 (data not shown). These data demonstrate that extraction and/or shedding of ΔCD PSGL-1 from circulating cells is not a major contributor to lower surface expression. Northern blots revealed equivalent levels of mRNA for PSGL-1 from both WT and ΔCD leukocytes (Fig. 1D). We also prepared constructs encoding human WT and ΔCD PSGL-1 (Fig. 1E) and selected stably transfected CHO cell clones expressing matched surface densities of each protein (Fig. 1F). Quantitative RT-PCR revealed nearly 10-fold higher transcripts for PSGL-1 in cells expressing ΔCD PSGL-1 than in cells expressing WT PSGL-1 (Fig. 1G). Thus, much higher mRNA levels were required to match ΔCD PSGL-1 surface densities to those of WT PSGL-1 on transfected cells.

Deleting the Cytoplasmic Domain Causes PSGL-1 to Accumulate in an Intracellular Compartment

We reasoned that redistribution of ΔCD PSGL-1 to an intracellular compartment might account for its reduced surface expression. To test this hypothesis, we incubated leukocytes with or without OSGE, a protease that cleaves proteins with clustered, sialylated O-glycans such as PSGL-1 (30). The cells were then fixed, washed, incubated with or without buffer that permeabilized the plasma membrane, stained with fluorescent mAb to an N-terminal epitope on PSGL-1, and analyzed by flow cytometry. Almost all of the PSGL-1 was expressed on the surface of WT murine leukocytes (Fig. 2A), consistent with the predominant surface expression of PSGL-1 on human leukocytes (17). Much less PSGL-1 was expressed on the surfaces of ΔCD leukocytes (Fig. 2A). This surface pool was quantitatively removed by treating cells with OSGE. After permeabilization, comparable total levels of PSGL-1 were detected in WT and ΔCD leukocytes. This was due to a large intracellular pool of PSGL-1 in ΔCD cells that was not removed by treating the cell surface with OSGE. Immunofluorescence microscopy confirmed that most PSGL-1 was on the surface of WT leukocytes, whereas most PSGL-1 was inside ΔCD leukocytes (Fig. 2B). These data demonstrate that deleting the cytoplasmic domain of PSGL-1 causes accumulation of intracellular PSGL-1.

Deleting the Cytoplasmic Domain Causes Retention of a PSGL-1 Precursor in the ER of Leukocytes

As observed previously (20), Western blotting with a polyclonal antibody to the N terminus of PSGL-1 detected 115- and 105-kDa forms of PSGL-1 in lysates from WT and ΔCD PSGL-1 leukocytes, respectively, following electrophoresis under reducing conditions (Fig. 3A). The faster migration of the ΔCD PSGL-1 species is consistent with deletion of its cytoplasmic domain. The anti-PSGL-1 antibody detected the 65-kDa species in ΔCD but not in WT leukocytes (Fig. 3A). Treating cell lysates with OSGE eliminated the 115- and 105-kDa bands but did not affect the 65-kDa band (Fig. 3A). These data suggest that the 65-kDa species is a precursor without the clustered, sialylated O-glycans required for recognition and cleavage by OSGE. Treating intact leukocytes with Pronase, a mixture of proteases, eliminated most of the 105-kDa form of ΔCD PSGL-1 (Fig. 3B). In contrast, Pronase did not decrease the 65-kDa form, suggesting an intracellular location that the proteases could not access.

Ser/Thr-linked core 1-derived O-glycans are major post-translational modifications of PSGL-1 (11, 31–33). The 65-kDa species in ΔCD leukocytes lacked at least some of the O-glycan structures required for recognition by OSGE. Many polypeptide GalNAc transferases add the initial GalNAc to Ser/Thr residues (34), whereas a single core 1 β1–3-galactosyltransferase adds Gal to GalNAc to form the core 1 structure, Galβ1–3-GalNAcα1-Ser/Thr, also known as the T antigen (35). The core 1 backbone is required for further extension or branching of this major class of O-glycans (36). Leukocytes from mice lacking core 1 β1–3-galactosyltransferase in endothelial and hematopoietic cells (EHC T-syn^−/− mice) express truncated O-glycans with the structure GalNAcα1-Ser/Thr, also known as the Tn antigen (11, 23). These cells expressed a 105-kDa form of PSGL-1 that was recognized by the plant lectin, Helix pomatia agglutinin, which binds to terminal GalNAc residues (Fig. 3C). In contrast, H. pomatia agglutinin did not bind to the 65-kDa species, suggesting that it lacked even the initial Ser/Thr-linked GalNAc residues. Because this modification and all subsequent O-glycosylation steps occur in the Golgi apparatus (34), we reasoned that the 65-kDa protein remained in the ER.

High mannose N-glycans are attached to asparagines in the ER and then modified into complex forms after transport to the Golgi apparatus (37). The enzyme endoglycosidase H removes high mannose but not complex N-glycans from proteins. Murine PSGL-1 has two potential sites for attaching N-glycans (38). Treating lysates of ΔCD leukocytes with endoglycosidase H increased the mobility of the 65-kDa species, consistent with removal of one or two high mannose N-glycans that were added in the ER (Fig. 3D). Endoglycosidase H did not affect the mobility of the mature 105-kDa protein (Fig. 3D), consistent with conversion of its high mannose glycans to enzyme-resistant complex forms in the Golgi apparatus. Confocal immunofluorescence microscopy of ΔCD leukocytes demonstrated that PSGL-1 colocalized with the ER marker calnexin (Fig. 3E). These combined data indicate that the 65-kDa form of ΔCD PSGL-1 is a precursor that accumulates in the ER.

Deleting the Cytoplasmic Domain Delays Export of PSGL-1 from the ER to the Golgi Apparatus

We hypothesized that the PSGL-1 precursor in ΔCD leukocytes accumulated in the ER because it moved very slowly to the Golgi, whereas the precursor in WT leukocytes moved rapidly to the Golgi. We detected a putative 75-kDa precursor in Western blots of WT leukocytes only after prolonged exposure of the blots (Fig. 4A). The 75-kDa species increased markedly after treating cells with brefeldin A, which collapses the cis-Golgi and causes biosynthetic intermediates to accumulate in the ER (Fig. 4A) (39). This observation provides strong evidence that the 75-kDa species is an ER-based PSGL-1 precursor. The 75-kDa precursor in WT cells was larger than the 65-kDa precursor in ΔCD cells, consistent with retention of its cytoplasmic domain. To measure the kinetics of conversion of precursor to mature forms of PSGL-1, we pulsed WT or ΔCD splenocytes with [³⁵S]methionine/cysteine and chased them in unlabeled medium for 2 or 4 h. Cell lysates were immunoprecipitated with polyclonal anti-PSGL-1 antibody. The immunoprecipitates were subjected to SDS-PAGE under reducing conditions and analyzed by fluorography. In WT splenocytes, most of the 75-kDa precursor was converted to the 115-kDa mature form after a 2-h chase (Fig. 4B). We did not detect other biosynthetic intermediates, suggesting that ER-to-Golgi transport is a rate-limiting step. In sharp contrast, very little of the 65-kDa precursor in ΔCD splenocytes was converted to the 105-kDa mature form after a 4-h chase (Fig. 4C). These data demonstrate that impaired ER-to-Golgi transport is a primary cause of the low surface expression of PSGL-1 in ΔCD leukocytes.

Delayed ER export of ΔCD PSGL-1 could result from the loss of an export signal in the cytoplasmic domain. Alternatively, loss of the cytoplasmic domain might cause the luminal domain of PSGL-1 to misfold. In the latter case, ER chaperones could initiate transport of ΔCD PSGL-1 from the ER to proteasomes, resulting in its rapid degradation. The pulse-chase data in Fig. 4C argue against rapid diminution of the ΔCD PSGL-1 precursor in the ER. To explore this issue further, we incubated splenocytes with cycloheximide for various intervals to block new protein synthesis. We then lysed the cells and measured turnover of PSGL-1 by Western blotting. The amount of ΔCD PSGL-1 precursor declined very slowly. Indeed, its decline was slower than that of mature WT or ΔCD PSGL-1 (Fig. 4D). Furthermore, incubating splenocytes with the proteasome inhibitor MG132 did not increase the amount of ΔCD PSGL-1 precursor (data not shown). These data indicate that the ΔCD PSGL-1 precursor is not prematurely degraded in proteasomes.

Deleting the Cytoplasmic Domain Impairs Dimerization of PSGL-1

When analyzed by SDS-PAGE under nonreducing conditions, very little of the accumulated 65-kDa precursor in ΔCD leukocytes forms disulfide-linked dimers (20). This suggests that lack of the cytoplasmic domain impairs the noncovalent interactions that are required to form the disulfide bond in the ER. To test this hypothesis, we prepared constructs of WT and ΔCD PSGL-1 that retained the single extracellular cysteine (residue 307 in the murine sequence) or replaced it with serine to prevent formation of disulfide bonds (Fig. 5A). These constructs were expressed at matched surface densities on stably transfected CHO cells. Western blots revealed that both the ΔCD and the ΔCD C307S cells expressed large amounts of the 65-kDa precursor (Fig. 5B). This result indicates that chaperone recognition of an unpaired cysteine in the lumen of the ER is not responsible for retention of the precursor. To test whether C307S or ΔCD C307S expressed mature noncovalent dimers on the cell surface, we treated intact cells with the membrane-impermeable cross-linker BS³. The cells were then lysed, fractionated by SDS-PAGE under reducing conditions, and analyzed by Western blotting. BS³ readily cross-linked the mature C307S PSGL-1 into dimers but cross-linked almost none of the mature ΔCD C307S PSGL-1 (Fig. 5C). In an alternative assay, we expressed constructs of C307S or ΔCD C307S PSGL-1 fused to GFP or DsRed (Fig. 5D). For each form of PSGL-1, we transfected cells with both the GFP and DsRed fusion constructs and measured close proximity of the donor/acceptor pairs at the cell surface by FRET. By measuring sensitized emission of DsRed following excitation of GFP, we detected significantly more FRET for C307S than for ΔCD C307S donor/acceptor pairs (Fig. 5E). The complementary cross-linking and FRET experiments demonstrate that the cytoplasmic domain contributes significantly to noncovalent dimerization of PSGL-1.

Dimerization of PSGL-1 Is Not Required to Exit the ER

Deleting the cytoplasmic domain impaired dimerization of PSGL-1 and caused precursors to accumulate in the ER. One interpretation of these data is that PSGL-1 must dimerize to efficiently move from the ER to the Golgi. The transmembrane domain also mediates noncovalent dimerization of PSGL-1. A human PSGL-1 construct containing the transmembrane domain of CD43 with no juxtamembrane cysteine is expressed on CHO cells as a monomer that BS³ does not cross-link (18). Accumulation of a 75-kDa precursor of this monomeric PSGL-1 was not observed. To readdress whether PSGL-1 must dimerize to exit the ER, we designed a murine PSGL-1 construct with the transmembrane domain of murine CD43 and an alanine replacing the extracellular cysteine, called CD43 TMD PSGL-1 (Fig. 6A). We also swapped the transmembrane domain of PSGL-1 with the transmembrane domain of glycophorin A, which is known to dimerize (40). Finally, we made ΔCD versions of CD43 TMD, glycophorin A TMD, and WT murine PSGL-1 by deleting the cytoplasmic domain. All constructs were expressed at matched surface densities on stably transfected CHO cells. Western blots of cell lysates electrophoresed under nondenaturing conditions confirmed that WT and glycophorin A TMD PSGL-1 migrated as dimers, whereas CD43 TMD PSGL-1 migrated as monomers (Fig. 6B). Western blots of lysates resolved by SDS-PAGE under reducing conditions revealed that 65-kDa precursors of PSGL-1 accumulated only in constructs lacking the cytoplasmic domain (Fig. 6C). Thus, the cytoplasmic domain promotes ER export of PSGL-1 even when it contains a heterologous transmembrane domain. Furthermore, dimerization of PSGL-1 is not required to exit the ER. A single cytoplasmic domain was sufficient for ER export of the monomeric CD43 TMD PSGL-1 chimera.

We asked whether a single cytoplasmic domain on a dimer pairing a WT subunit with a ΔCD subunit could “rescue” the ΔCD subunit from the ER and transport the heterodimer to the cell surface. We compared expression of PSGL-1 on leukocytes from WT mice, ΔCD mice, and heterozygous mice containing one allele for WT PSGL-1 and one allele for ΔCD PSGL-1. If interactions between cytoplasmic domains were irrelevant for dimerization, then ∼50% of the surface PSGL-1 on heterozygous leukocytes should be WT/ΔCD heterodimers and another 25% should be WT/WT homodimers, with a total expression of ∼75% of that on WT leukocytes given the very low surface expression of ΔCD/ΔCD homodimers. The export of WT/ΔCD heterodimers from the ER to the cell surface should decrease the 65-kDa ΔCD precursor. However, flow cytometry revealed that surface expression of PSGL-1 on heterozygous leukocytes reached only ∼50% of that on WT leukocytes (Fig. 7, A and B). Furthermore, Western blots showed that heterozygous leukocytes still expressed high levels of the 65-kDa precursor (Fig. 7C). To clearly resolve the mobilities of mature PSGL-1 with or without the cytoplasmic domain, we fractionated cell lysates by SDS-PAGE under reducing conditions for a longer period in a gel with a low polyacrylamide concentration. The mature PSGL-1 in heterozygous leukocytes co-migrated with PSGL-1 in WT leukocytes and migrated slower than mature PSGL-1 in ΔCD leukocytes (Fig. 7D). Thus, very little ΔCD PSGL-1 reached the surface of heterozygous leukocytes. These data confirm that interactions between cytoplasmic domains promote dimerization of PSGL-1. Furthermore, dimerization of two WT subunits takes precedence over dimerization of a WT and ΔCD subunit. Almost all of the mature WT PSGL-1 in WT or heterozygous leukocytes migrated as disulfide-linked dimers under nonreducing conditions (Fig. 7E). This could occur if dimerization of two WT subunits in the ER is faster than export of a monomeric WT protein, if two cytoplasmic domains in a WT homodimer cause faster ER export than one cytoplasmic domain in a WT monomer, and/or if some WT monomers form disulfide-linked dimers after they reach the cell surface.