Golgi phosphoprotein 3 determines cell binding properties under dynamic flow by controlling Golgi localization of core 2 N-acetylglucosaminyltransferase 1

Abstract

Core 2 N-acetylglucosaminyltransferase 1 (C2GnT1) is a key enzyme participating in the synthesis of core 2-associated sialyl Lewis x (C2-O-sLe(x)), a ligand involved in selectin-mediated leukocyte trafficking and cancer metastasis. To accomplish that, C2GnT1 needs to be localized to the Golgi and this step requires interaction of its cytoplasmic tail (CT) with a protein that has not been identified. Employing C2GnT1 CT as the bait to perform a yeast two-hybrid screen, we have identified Golgi phosphoprotein 3 (GOLPH3) as a principal candidate protein that interacts with C2GnT1 and demonstrated that C2GnT1 binds to GOLPH3 via the LLRRR(9) sequence in the CT. Confocal fluorescence microscopic analysis shows substantial Golgi co-localization of C2GnT1 and GOLPH3. Upon GOLPH3 knockdown, C2GnT1 is found mainly in the endoplasmic reticulum and decorated with complex-type N-glycans, indicating that the enzyme has been transported to the Golgi but is not retained. Also, we have found that a recombinant protein consisting of C2GnT1 CT(1-16)-Leu(17-32)-Gly(33-42)-GFP is localized to the Golgi although the same construct with mutated CT (AAAAA(9)) is not. The data demonstrate that the C2GnT1 CT is necessary and sufficient for Golgi localization of C2GnT1. Furthermore, GOLPH3 knockdown results in reduced synthesis of C2-O-sLe(x) associated with P-selectin glycoprotein ligand-1, reduced cell tethering to and rolling on immobilized P- or E-selectin, and compromised E-selectin-induced activation of spleen tyrosine kinase and cell adhesion to intercellular adhesion molecule-1 under dynamic flow. Our results reveal that GOLPH3 can regulate cell-cell interaction by controlling Golgi retention of C2GnT1.

FIGURE 1.

Interaction of GOLPH3 with C2GnT1 CT in vitro. A, GOLPH3 Western blot of the pulldown from KG1a cell lysate with biotinylated control or C2GnT1 1–20-aa peptide. B, GOLPH3 Western blot analysis of the pulldown from the lysate of BL21 bacteria transformed with GST-GOLPH3 cDNA and from purified recombinant GOLPH3 using biotinylated control or C2GnT1(1–20 aa) peptide. C, C2GnT1 Western blot analysis of the pulldown from KG1a cell lysate using GST (control) or GOLPH3-GST immobilized on glutathione-resin.

FIGURE 2.

Identification of the amino acids in the C2GnT1 CT critical for binding to GOLPH3. A, sequences of the biotinylated C2GnT1 peptides used for the experiment; wild type (1–20) and mutants (Ala^5,6 and Ala^7–9). B, GOLPH3 Western blot analysis of the pulldown from KG1a lysates using biotin (control), biotinylated C2GnT1 1–20-aa peptide, and Ala^5,6 and Ala^7–9 mutants shown in A. C, forward yeast two-hybrid assay between Gal4(DBD)-C2GnT1(LLRRR⁹), Gal4(DBD)-C2GnT1 (AARRR⁹) or Gal4(DBD)-C2GnT1(LLAAA⁹), and Gal4(AD)-GOLPH3. Yeasts carrying the mutant plasmids did not grow (i.e. failure to interact with GOLPH3) in media lacking His but containing aureobasidin A. SV40 large T antigen and P53 are the positive controls and SV40 large T antigen and Lam are the negative controls. D, confocal fluorescence images of K562 cells transfected with various plasmid constructs showing differential Golgi localization of GFP: GFP is localized to the Golgi with the C2GnT1^(1–428)-GFP construct but not the same construct with mutated CT such as deletion of amino acids 5–9, and mutation of AA⁶ or AAA⁹. E, GFP is localized to the Golgi using C2GnT1^1–70-GFP construct, but not the same construct with mutated CT (AAAAA⁹). E, GFP was localized to the Golgi using C2GnT1CT^1–16-Leu^17–32-Gly^33–42-GFP construct but not the same construct with mutated CT (AAAAA⁹). Bar = 5 μm.

FIGURE 3.

Analysis of the overlap of C2GnT1 with GOLPH3 in the Golgi of KG1a cells. A, overlap images of C2GnT1 (green) and GOLPH3 (red) on each Golgi z-section. The Mander's coefficient for green (C2GnT1) and red (GOLPH3) at each z-section is shown. The z-sections were obtained in 0.35-μm increments along the z axis (microscope optical axis) in a basal to apical direction. The Mander's coefficient for each z-section across the entire Golgi obtained from one representative cell is shown. B, average Mander's coefficients (mean ± S.E.) for C2GnT1 and GOLPH3 were analyzed for an average of 10 z-sections/cell in 30 different cells from three independent experiments. Bar = 2 μm.

FIGURE 4.

Colocalization of C2GnT1 and GOLPH3 with markers of various Golgi compartments in KG1a cells. A, the Golgi line scan graphs showing the immunofluorescence intensity of C2GnT1 (left panel) and GOLPH3 (right panel) with cis-Golgi GM130, medial Golgi ManII, and trans-Golgi TGN46 along randomly positioned arrowed lines spanning the Golgi. The intensity profiles shown are representatives of 30 cells from three independent experiments. B, three-dimensional rendering of confocal fluorescence images showing the Golgi colocalization of C2GnT1 (top panel) and GOLPH3 (bottom panel) with GM130, ManII, and TGN46. C, Pearson's correlation coefficients for the colocalization of C2GnT1 or GOLPH3 with Golgi markers; *, p < 0.01; **, p < 0.001. Bar = 5 μm.

FIGURE 5.

GOLPH3 knockdown prevents Golgi localization of C2GnT1 without affecting the Golgi morphology, and analysis of the N-glycan types on C2GnT1 in KG1a cells treated with vanadate, water control, non-targeting or GOLPH3-specific siRNAs. A, GOLPH3 Western blot showing about 80–90% depletion of GOLPH3 in KG1a cells after transfection with GOLPH3 siRNAs as compared with cells transfected with non-targeting siRNAs (Mock). B and C, confocal fluorescence images of KG1a cells showing localization of C2GnT1 to the Golgi (Giantin) but distribution of C2GnT1 to the ER (protein-disulfide isomerase, PDI) after treatment with GOLPH3 siRNAs. D, quantification of the C2GnT1 fluorescence intensity in the ER relative to that of the Golgi in non-targeting or GOLPH3 siRNA-treated cells (n ≥ 25, bar = 5 μm). *, p < 0.01. E, ConA and PSA blots of the KG1a cell lysates with and without endoglycosidase H (Endo H) treatment. These KG1a cells have been treated with vanadate (inhibitor of ER-to-Golgi transport), water, non-targeting siRNAs (Mock), or GOLPH3 siRNAs. F, FACS analysis of KG1a cells after treated with fluorescein-labeled ConA or PSA lectin. Cells transfected with non-targeting siRNA (Mock), dotted line; or GOLPH3 siRNA, solid line. ConA binds high mannose-type N-glycans preferentially, PSA binds bi- and triantennary complex type N-glycans terminated with GlcNAc.

FIGURE 6.

GOLPH3 knockdown decreases PSGL-1-associated C2-O-sLe^x. A, KG1a cells were transfected with either mock (non-targeting) siRNA, GOLPH3 siRNAs, or C2GnT1 siRNA and cultured for 48–72 h. Cell surface C2-O-sLe^x and PSGL-1 were analyzed by flow cytometry using mouse anti-C2-O-sLe^x (CHO-131) IgM and F(ab)₂ fragment of goat anti-mouse IgM fluorescein isothiocyanate-conjugated secondary Abs, and rabbit anti-PSGL-1 Abs and F(ab)₂ fragment of goat anti-rabbit phycoerythrin-conjugated secondary Abs. GOLPH3 or C2GnT1 siRNA treatment reduced PSGL-1-associated C2-O-sLe^x from 40.6 to 10.3 and 7.3%, respectively. B, GOLPH3 or C2GnT1 knockdown did not affect the PSGL-1 protein level. C and D, real-time quantitative PCR analysis of C2GnT1 and GOLPH3 in KG1a cells transfected with C2GnT1 or GOLPH3 siRNA, respectively. Knockdown of the mRNA of either gene did not affect the expression level of the other.

FIGURE 7.

C2GnT1 or GOLPH3 knockdown reduces interaction of KG1a cells with immobilized E- or P- selectin under dynamic flow. C2GnT1 siRNA, GOLPH3 siRNA, or non-targeting siRNA (mock) treated cells (2.5 × 10⁵ cells/ml) were perfused through a parallel-plate flow chamber containing a coverslip coated with E- or P-selectin at a constant wall shear stress of 1.0 dyne/cm². Tethering or rolling was visualized with a phase-contrast microscope at ×40 magnification and videotaped. A, compared with mock treated cells, C2GnT1 or GOLPH3 siRNA-treated cells that were tethered to immobilized E-selectin were reduced by 68.9 (145 ± 8 versus 45 ± 5 cells, p < 0.001) and 90.3% (145 ± 8 versus 14 ± 4 cells, p < 0.001), respectively. B, the C2GnT1 or GOLPH3 siRNA-treated cells that rolled on immobilized E-selectin were reduced by 81.1 (180 ± 12 versus 34 ± 5 cells, p < 0.001) and 73.8% (180 ± 12 versus 47 ± 6 cells, p < 0.001), respectively. C, C2GnT1 or GOLPH3 siRNA-treated cells that were tethered to immobilized P-selectin were reduced by 67.2 (113 ± 14 versus 37 ± 5 cells, p < 0.001) and 60.1% (113 ± 14 versus 45 ± 15 cells, p < 0.01), respectively. D, C2GnT1 or GOLPH3 siRNA-treated cells that rolled on immobilized P-selectin were reduced by 77.1 (214 ± 9 versus 49 ± 7 cells, p < 0.001) and 72.9% (214 ± 9 versus 58 ± 9 cells, p < 0.001), respectively.

FIGURE 8.

GOLPH3 or C2GnT1 knockdown prevents activation of SYK and reduces the adhesion of E-selectin-treated KG1a cells to immobilized ICAM-1 under dynamic flow. A, EDTA, C2GnT1 siRNA, GOLPH3 siRNA, or non-targeting siRNA (mock)-treated cells were incubated with E-selectin for 20 min. Whole cell lysates were analyzed by Western blotting for phosphorylated SYK. EDTA inhibits phosphorylation of SYK and BSA does not induce SYK phosphorylation. B, C2GnT1 siRNA, GOLPH3 siRNA, or non-targeting siRNA (mock)-treated cells (2.5 × 10⁵ cells/ml) were incubated with E-selectin for 20 min before being perfused through a parallel plate flow chamber containing a coverslip coated with ICAM-1 at a constant wall shear stress of 1.0 dyne/cm². Compared with mock treated cells, GOLPH3 or C2GnT1-depleted and E-selectin-activated KG1a cells exhibited a decrease in their adherence to immobilized ICAM-1 by 41.9 (694 ± 86 versus 403 ± 56, p < 0.05) and 59.2% (694 ± 86 versus 283 ± 30, p < 0.01), respectively.

FIGURE 9.

A diagram depicting the function of GOLPH3 in retaining C2GnT1 in the Golgi, and promoting the formation of C2-O-sLe^x and subsequent interaction with P/E-selectin and ICAM1. GOLPH3 binds to the CT of C2GnT1 to retain this enzyme in the Golgi to synthesize the core 2 structure, which enables the synthesis of selectin ligand (C2-O-sLe^x) on PSGL-1 to facilitate the interaction of the KG1a cells with P/E-selectin. This interaction results in phosphorylation of SYK, which in turn activates β2-integrin to induce a firm adhesion of the activated cells to the endothelial cells via ICAM-1.