Links between CD147 function, glycosylation, and caveolin-1

Abstract

Cell surface CD147 shows remarkable variations in size (31-65 kDa) because of heterogeneous N-glycosylation, with the most highly glycosylated forms functioning to induce matrix metalloproteinase (MMP) production. Here we show that all three CD147 N-glycosylation sites make similar contributions to both high and low glycoforms (HG- and LG-CD147). l-Phytohemagglutinin lectin binding and swainsonine inhibition experiments indicated that HG-CD147 contains N-acetylglucosaminyltransferase V-catalyzed, beta1,6-branched, polylactosamine-type sugars, which account for its excess size. Therefore, CD147, which is itself elevated on invasive tumor cells, may make a major contribution to the abundance of beta1,6-branched polylactosamine sugars that appear on invasive tumor cells. It was shown previously that caveolin-1 associates with CD147, thus inhibiting CD147 self-aggregation and MMP induction; now we show that caveolin-1 associates with LG-CD147 and restricts the biosynthetic conversion of LG-CD147 to HG-CD147. In addition, HG-CD147 (but not LG-CD147) was preferentially captured as a multimer after treatment of cells with a homobifunctional cross-linking agent and was exclusively recognized by monoclonal antibody AAA6, a reagent that selectively recognizes self-associated CD147 and inhibits CD147-mediated MMP induction. In conclusion, we have 1) determined the biochemical basis for the unusual size variation in CD147, 2) established that CD147 is a major carrier of beta1,6-branched polylactosamine sugars on tumor cells, and 3) determined that caveolin-1 can inhibit the conversion of LG-CD147 to HG-CD147. Because it is HG-CD147 that self-aggregates and stimulates MMP induction, we now have a mechanism to explain how caveolin-1 inhibits these processes. These results help explain the previously established tumor suppressor functions of caveolin-1.

Figure 1.

Variability among CD147 glycoforms. The indicated cell lines were lysed with 1% Brij 96 and then blotted with anti-CD147 polyclonal antibody B10 for endogenous CD147. FB, fibroblasts.

Figure 2.

Characterization of CD147 glycosylation. (A) HT1080 cells were treated with 0, 5, or 10 μg/ml tunicamycin for 12 h. CD147 immunoprecipitates were then blotted for CD147 (top) or associated caveolin-1 (middle). Caveolin-1 in total cell lysates was also blotted (bottom). (B) CD147 was immunoprecipitated from HT1080 cells (with mAb 8G6), and samples were then treated with 1, 5, or 10 μl of endo H (1000 U/μl), in 1.0 ml of buffer, at 37°C for 30 min before SDS-PAGE. CD147 was then detected by blotting with polyclonal antibody B10. IP, immunoprecipitated; IB, immunoblotted.

Figure 3.

Evidence for β1,6-branched polylactosamines. (A) HT1080 cells stably transfected with CD147-GFP were lysed in 1% Brij 96 detergent and then immunoprecipitated with anti-CD147 mAb 8G6 (lane 1) or the lectin l-PHA (lane 2). After depletion of l-PHA-binding material, CD147 was again immunoprecipitated (lane 3). (B) HEK293 cells were treated with 100 μM swainsonine for 2 h, incubated in cysteine- and methionine-free medium for 1 h, and then pulsed for 4 h in medium containing 0.5 mCi/ml [³⁵S]methionine/cysteine and 5.0% dialyzed fetal calf serum. Labeled cells were lysed in RIPA buffer (25 mM Tris, 2 mM EDTA, 150 mM NaCl, 0.1% NaN₃, pH 7.2, plus 1% Triton X-100, 1% deoxycholate, and 0.1% sodium dodecyl sulfate), endogenous CD147 and integrin β1 were immunoprecipitated, and labeled proteins were detected with a PhosphorImager (Storm 860; Molecular Dynamics, Sunnyvale, CA).

Figure 4.

Effects of N-glycosylation on CD147 size and association with caveolin-1. (A) HT1080 cells stably expressing mutant and wild-type GFP-CD147 were lysed in 1% Brij 96. CD147 was then immunoprecipitated with anti-GFP mAb, and immunoprecipitates were immunoblotted for CD147 (anti-GFP) (top), MCT-1 (second), or caveolin-1 (third). Caveolin-1 present in total lysates was also detected by blotting (bottom). CD147 mutants were obtained with recombinant PCRs and subcloned into pEGFP-N1 vector (Clontech, Palo Alto, CA), yielding a C-terminal GFP fusion protein. All constructions were confirmed by sequencing (1Q, N44Q; 32Q, N152Q and N186Q; 321Q, N44Q and N152Q and N186Q). (B) HT1080 cells expressing wild-type CD147-GFP were lysed in 1% Brij 96, and then endogenous caveolin-1 was immunoprecipitated with polyclonal antibody. CD147 was detected in the caveolin-1 immunoprecipitate (lane 1) or in the total lysate (lane 2) by blotting with anti-GFP mAb. It should be noted that increasing exposure time by five to tenfold longer still did not reveal the HG form of CD147 in lane 1. Similar results were observed with HEK293 cells overexpressing caveolin-1 (not shown). Cav, caveolin-1.

Figure 5.

Effects of caveolin-1 removal on the CD147 HG/LG ratio. A small interfering RNA oligonucleotide (gagaagcaagtgtacgacg), previously established to reduce caveolin-1 expression (Nichols, 2002), was inserted into the vector pSilencer3.1-H1hygro (Ambion, Austin, TX) and was stably expressed in RD cells with the Oligofectamine reagent (Invitrogen, Carlsbad, CA), followed by selection with 200 μg/ml hygromycin (Sigma) for ∼2 wk. Left, results of CD147 blotting from RD cell lysates (top) and quantitative scans of CD147 blots (bottom) (scanned with GeneTools version 3.0; Syngene Laboratories, Frederick, MD). The r values refer to HG/LG ratios. Right, caveolin-1 blotting (top) and scans of caveolin-1 blots (bottom). The numbers 1 and 6 indicate the relative areas of caveolin-1 peaks. Similar results were obtained with many different film exposures and for multiple clones of RD transfectants. RNAi, RNA interference.

Figure 6.

Caveolin-1 effects on the CD147 HG/LG ratio. (A) HEK293 cells stably expressing the control vector or caveolin-GFP were lysed, and then endogenous CD147 (top) or caveolin-1 (bottom) was detected by blotting of total lysates. (B) HEK293 cells overexpressing caveolin-1 or the control vector were surface-labeled with biotin and lysed in 1% Brij 96. Immunoprecipitates of CD147 (mAb 8G6) or control IgG were then blotted for surface-labeled material with avidin-horseradish peroxidase. (C) HEK293 cells were stably cotransfected with CD147-GFP and CD151, caveolin-1-GFP, or GFP alone. CD147 was then immunoprecipitated with mAb 8G6 and analyzed by blotting. Caveolin-1-GFP was prepared as described (Tang and Hemler, 2004).

Figure 7.

Caveolin-1 effects on CD147 biosynthesis. In control HEK293 cells or cells stably expressing caveolin-1, endogenous CD147 was labeled with [³⁵S]methionine/cysteine for 1 h, immunoprecipitated, and detected as in Figure 3B. After the 1-h pulse, the ³⁵S-label was chased for the additional times indicated, in medium containing 5.0% dialyzed fetal calf serum and 25× excess unlabeled l-methionine.

Figure 8.

Selective multimerization of HG-CD147. (A) Intact HEK293 cells were treated for 2 h with the homobifunctional cross-linking agent BS3 (bis[sulfosuccinimido]suberate), at 0.5 or 1.0 μg/ml, and then CD147-GFP was immunoprecipitated from cell lysates with mAb 8G6 and blotted with anti-GFP polyclonal antibody. Because of the presence of the GFP tag, sizes are increased (∼75 and 57 kDa), compared with Figures 1 and 2 (50 and 32 kDa). Band intensities corresponding to the HG and LG forms were quantitated with GeneTools version 3.0 (Syngene Laboratories), confirming the selective decrease in the HG form after cross-linking. The slight decrease in the amount of the LG form is consistent with decreased antibody recognition, rather than decreased interchain cross-linking. (B) Endogenous CD147 was immunoprecipitated from HT1080 cells with mAb 8G6, negative control antibody, or mAb AAA6 and was then detected with CD147 immunoblotting. Even with longer exposure of the AAA6 sample (lane 4), no LG form can be observed. Ctl, control.

Figure 9.

Schematic diagram of CD147 maturation and regulation by caveolin-1. After CD147 enters the Golgi complex, it can mature via two possible pathways. First, it associates with caveolin-1, fails to acquire polylactosamines, and is escorted to the cell surface while still at least partly associated with caveolin-1, in the context of a larger, caveolin-containing complex (Tang and Hemler, 2004). This LG-CD147 does not self-aggregate and does not induce MMP production. Second, with initiation by the action of GnT-V, CD147 acquires β1,6-branched polylactosamines on its N-glycans as it matures to HG-CD147. This molecule has a tendency to self-aggregate and is active in the stimulation of MMP production by neighboring cells. ER, endoplasmic reticulum; Gly, glycosylation.