CD147-Dependent Heterogeneity in Malignant and Chemoresistant Properties of Cancer Cells

Abstract

CD147 (alias emmprin or basigin), an integral plasma membrane glycoprotein and a member of the Ig superfamily, is widespread in normal tissues, but highly up-regulated in many types of malignant cancer cells. CD147 is multifunctional, with numerous binding partners. Recent studies suggest that complexes of CD147 with the hyaluronan receptor CD44 and associated transporters and receptor tyrosine kinases are enriched in the plasma membrane of cancer stem-like cells. Here, we show that subpopulations of tumor cell lines constitutively expressing high levels of cell-surface CD147 exhibit cancer stem-like cell properties; that is, they exhibit much greater invasiveness, anchorage-independent growth, spheroid formation, and drug resistance in vitro and higher tumorigenicity in vivo than those constitutively expressing low levels of cell-surface CD147. Primary CD147-rich cell subpopulations derived from mouse mammary adenocarcinomas also exhibit high levels of invasiveness and spheroid-forming capacity, whereas CD147-low cells do not. Moreover, localization at the plasma membrane of CD44, the EGF receptor, the ABCB1 and ABCG2 drug transporters, and the MCT4 monocarboxylate transporter is elevated in cells constitutively expressing high levels of cell-surface CD147. These results show that CD147 is associated with assembly of numerous pro-oncogenic proteins in the plasma membrane and may play a fundamental role in properties characteristic of cancer stem-like cells.

Numerous studies have demonstrated the presence of highly malignant and chemoresistant cell subpopulations in many types of cancers,^1–5 as well as in established cell lines.^6,7 The nature and origin of these cancer stem-like cells are still controversial, but there is a growing consensus that tumors contain varying sized subpopulations of these cells and that these cells may be largely responsible for tumor recurrence and possibly metastasis. Although several protein markers are used to identify these subpopulations, the functional basis for their role in the distinctive properties of cancer stem-like cells is still poorly understood.

CD147 (alias emmprin and basigin) is an integral plasma membrane glycoprotein and member of the Ig superfamily that is widespread in normal tissues, but highly up-regulated in remodeling tissues and in many types of cancers.^8–11 Emmprin was originally identified as a factor on the surface of tumor cells that induces matrix metalloproteinase production in fibroblasts^12,13 and was subsequently shown to be identical to basigin.¹⁴ Here we refer to it by the cluster of differentiation identifier, CD147. More recent work has shown that tumor cell CD147 induces matrix metalloproteinases in endothelial cells and tumor cells themselves, resulting in increased tumor invasiveness and angiogenesis.^15–18 However, CD147 clearly has functions other than matrix metalloproteinase induction, and most likely acts as a functional binding partner for several plasma membrane proteins, including monocarboxylate transporters,^19–21 CD98,²² drug transporters,^23,24 MT1-MMP²⁵ and the hyaluronan receptors CD44²⁶ and LYVE-1,²⁴ thus influencing activities characteristic of cancer stem-like cells, such as cell survival and drug resistance^24,27–32 and invasion and metastasis.^{15,25,33–35}

In the present study, we used flow cytometric sorting to isolate cell subpopulations with different constitutive levels of cell-surface CD147 from three different types of CD147-expressing tumor cell lines [human malignant peripheral nerve sheath tumor (MPNST) cells,³⁶ SKOV3 human ovarian carcinoma cells³⁷ and MDA MB231 human breast carcinoma cells²⁰], as well as from primary mouse mammary tumor cells. Tumor cells constitutively expressing high levels of cell-surface CD147 (CD147^high cells) exhibited much greater invasiveness, anchorage-independent growth, and drug resistance in culture and higher tumorigenicity in vivo, compared with those expressing low levels of cell-surface CD147 (CD147^low cells). Moreover, membrane localization of CD44, EGFR, the ABC-family drug transporters ABCB1/P-glycoprotein (Pgp) and ABCG2/breast cancer resistance protein (BCRP), and the monocarboxylate transporter MCT4 was higher in CD147^high cells than in CD147^low cells.

Materials and Methods

Animals

The FVB/NTg(MMTV-PyVT)634Mul transgenic mice were obtained from Dennis Watson (Medical University of South Carolina) and NOD/SCID mice were obtained from The Jackson Laboratory (Bar Hatbor, ME). Animal housing and all experimental procedures were conducted in facilities for laboratory animals provided by the Division of Laboratory Animal Resources and the studies were approved by the Institutional Animal Care and Use Committee of Medical University of South Carolina.

Reagents

Fetal bovine serum was purchased from Atlanta Biologicals (Lawrenceville, GA), and RPMI 1640, Dulbecco’s modified Eagle’s medium, and Ham’s F12 media were purchased from Sigma-Aldrich. The following antibodies were obtained for these studies: BCRP/ABCG2 clone BXP-21 (Kamiya Biomedical, Seattle, WA); Alexa Fluor 647-conjugated CD44/HCAM (SC-7297), CD44/HCAM (DF1485), MCT4 (H-90) (Santa Cruz Biotechnology, Santa Cruz, CA); Pgp/ABCB1 (Calbiochem C219; EMD Millipore, La Jolla CA); mouse anti-human CD147 (HIM6; BD Biosciences, San Jose, CA); EGFR (2232; Cell Signaling Technology, Danvers, MA); β-actin clone AC-15 (Sigma-Aldrich), Alexa Fluor 647-tagged chicken anti-mouse IgG, Alexa Fluor 647-tagged chicken anti-rabbit IgG, Alexa Fluor 555-tagged goat anti-mouse IgG, Alexa Fluor 555-tagged goat anti-rabbit IgG (Invitrogen; Life Technologies, Carlsbad, CA); and goat anti-mouse horseradish peroxidase and goat anti-rabbit horseradish peroxidase (Chemicon; EMD Millipore, Temecula, CA). Fluorescence-activated cell sorting (FACS) was performed using fluorescein isothiocyanate mouse anti-human CD147 (HIM6), rat anti-mouse CD147 (AbD Serotec, Raleigh, NC), chicken anti-rat IgG-Alexa Fluor 488 (Invitrogen), and fluorescein isothiocyanate mouse IgG1κ isotype control (BD Biosciences). Pierce Western blotting detection reagent (enhanced chemiluminescence) was purchased from Thermo Fisher Scientific (Rockford, IL). All other chemicals were of reagent or higher grade.

Cell Culture

SKOV3 human ovarian adenocarcinoma and MDA MB231 human breast carcinoma cell lines were obtained from ATCC (Manassas, VA). The ST88-14 MPNST cell line was obtained from Larry Sherman (Oregon Health and Science University, Portland, OR). The carcinoma and MPNST cells were maintained in RPMI 1640 and in Dulbecco’s modified Eagle’s medium-Ham’s F12 medium, respectively, with 2.38 g/L HEPES, 2 g/L sodium bicarbonate, and 10% fetal bovine serum, pH 7.4, at 37°C in a humidified incubator with 5% CO₂-enriched air.

Primary mouse mammary carcinoma cells (MMTV-PyMT cells) were isolated from adenocarcinomas that grow spontaneously in FVB/NTg(MMTV-PyVT)634Mul female transgenic mice. The tumors were harvested after reaching 0.5 cm diameter (8- to 11-week-old mice), washed with 2% fetal bovine serum in PBS, and minced with razor blades. Tissue was shaken at low speed for 6 hours at 37°C in a solution containing 1 part 10× collagenase/hyaluronidase (Stemcell Technologies, Vancouver, BC, Canada), 9 parts complete mouse EpiCult B medium (Stemcell Technologies), 5% fetal bovine serum, and antibiotic-antimycotic solution (HyClone Laboratories; Thermo Scientific, South Logan, UT). Dissociated tissue was vortexed, centrifuged at 450 × g for 5 minutes, treated with ammonium chloride in Hank’s balanced salt solution to remove the red blood cells, and then treated with 0.25% trypsin-EDTA (HyClone Laboratories), followed by 5 mg/mL dispase (Stemcell Technologies) and 200 U/mL DNase I (Stemcell Technologies) in Hank’s balanced salt solution. The cell suspension was filtered through a 40-μm nylon mesh before plating in complete mouse EpiCult B medium supplemented with 5% fetal bovine serum, 4 μg/mL heparin (Sigma-Aldrich, St. Louis, MO), 10 ng/mL recombinant human basic fibroblast growth factor (rhFGF-basic; PeproTech, Rocky Hill, NJ), 10 ng/mL recombinant human epidermal growth factor (rhEGF; PeproTech), and antibiotic-antimycotic solution. Incubation at 37°C for 1 hour in a humidified incubator with 5% CO₂-enriched air was performed to allow the attachment of stromal cells (mainly macrophages and fibroblasts) to the plate. Nonadherent cells were collected and replated overnight in the same medium, after which the medium was replaced with serum-free medium.

Cell Sorting of CD147^high and CD147^low Subpopulations

Cells were trypsinized into a single-cell suspension using 0.25% trypsin-EDTA (HyClone Laboratories), counted, blocked with 3% bovine serum albumin in PBS, and treated with antibodies in culture medium. SKOV3, MDA MB231, and MPNST cells were incubated with anti–CD147-fluorescein isothiocyanate, or with IgG-fluorescein isothiocyanate as a negative control, for 30 minutes on ice. Mouse MMTV-PyMT cells were incubated with rat anti-mouse CD147 (1:50 dilution) for 1 hour, followed by incubation in chicken anti-rat IgG-Alexa Fluor 488 (1:200) for 30 minutes. Unbound antibody was washed off, and cells were then suspended in medium containing 1:1000 propidium iodide, filtered, and sorted on a FACSAria II system (BD Biosciences). Cells expressing the highest levels of cell-surface CD147 (the top 10% of SKOV3, MDA MB231, and MMTV-PyMT cells and the top 20% of MPNST cells) were selected as the CD147^high fraction. Conversely, those cells expressing the lowest levels of cell-surface CD147 (the bottom 10% of SKOV3, MDA MB231, and MMTV-PyMT cells and the bottom 20% of MPNST cells) were selected as the CD147^low fraction.

Recombinant Adenovirus Infections

Recombinant human CD147 adenovirus and control β-galactosidase (β-gal) adenovirus were constructed and used as described previously.^38,39 CD147 adenovirus dose was adjusted to cause moderate (two- to fivefold) increases in expression.

Matrigel Invasion Assay

Matrigel invasion chambers (BD Biosciences) were hydrated for 2 hours at 37°C in culture medium. After hydration, the medium in the bottom of the well was replaced with fresh culture medium, and 5 × 10⁴ cells were plated in the top chamber. After 24 hours of incubation, the cells were fixed with 3% paraformaldehyde for 15 minutes at room temperature; the chambers were then rinsed in PBS and stained with 0.2% crystal violet for 10 minutes. After the chambers had been washed five times with water, the cells at the top of the Matrigel membrane were removed with cotton swabs. The invaded cells on the bottom side of the membrane were subsequently counted under a phase-contrast microscope (DFC320; Leica Microsystems, Wetzlar, Germany).

Anchorage-Independent Growth of Colonies in Soft Agar

Six-well plates were prepared with a layer of 0.5% agar in culture medium. Cells (2500 per well) were suspended in 0.35% agarose in culture medium and layered on top of the 0.5% agar. Additional culture medium was added to the top layer of soft agar twice a week. After 28 days, colonies were stained with 0.005% crystal violet and photographed under a phase-contrast microscope (Leica DFC320).

Spheroid Formation

Cells were plated in ultralow attachment plates at a density of 20,000 cells/mL and were cultured in complete MammoCult medium (Stemcell Technologies) supplemented with 0.5 μg/mL hydrocortisone and 4 μg/mL heparin. Cells were incubated at 37°C and in 5% C0₂-enriched air for 7 to 12 days, with the medium being changed every 3 or 4 days. Cell were then photographed under a phase-contrast microscope (Leica DFC320).

Drug Resistance Assay

Cells were seeded in 96-well plates and incubated for approximately 48 hours before the experiment to reach a final density of approximately 70% confluency. Cells were then treated with doxorubicin (0, 0.001, 0.01, 0.1, 1, 10, 100, or 1000 μmol/L) for 96 hours at 37°C in an incubator with 5% CO₂-enriched air. After three washes with PBS, adherent cells were trypsinized and counted in a Z1 particle counter (Beckman Coulter; Fullerton, CA).

FURA-2 AM Assay

Efflux of Fura-2-acetoxymethyl ester (FURA-2 AM) was measured as described previously.⁴⁰ Cells were incubated in 96-well plates to reach a final density of approximately 70% confluency. Cells were treated in feed medium containing 2.5 μmol/L FURA-2 AM. After 1 hour of incubation, plates were read (excitation 340 nm, emission 500 nm) from the bottom in an FLx800 microplate fluorescence reader (BioTek Instruments, Winooski, VT) to determine FURA-2 AM levels in the cell layer.

Immunoblot Analysis

Whole-cell lysates were prepared for immunoblotting using a radioimmunoprecipitation assay buffer modified to contain 1 mmol/L phenylmethylsulfonyl fluoride, 10 μg/mL aprotinin and leupeptin, 2 mmol/L sodium orthovanadate, and 10 mmol/L sodium fluoride. Protein content was determined by BCA assay (Pierce; Thermo Fisher Scientific, Rockford, IL), and aliquots containing 25 to 50 μg of protein were solubilized in SDS sample buffer, resolved on Pierce 4% to 20% reducing polyacrylamide gels (Thermo Fisher Scientific), transferred to nitrocellulose (GE Osmonics-Micron Separations, Westborough, MA), and stained with antibody.

Flow Cytometry Analyses

Cells (1 × 10⁵) were trypsinized to achieve single-cell suspensions, counted, blocked with 3% bovine serum albumin in PBS, and incubated with antibody for 30 minutes on ice. Unbound antibody was removed by subsequent washes, and cells were analyzed no later than 1 hour after staining on a FACSCalibur system (BD Biosciences).

Xenografts

After trypsinization and a PBS wash, FACS-sorted SKOV3 cells were suspended in a 1:4 ratio of PBS to Matrigel (0.2 mL), and then 10⁴ to 10⁵ CD147^low or 10³ to 10⁴ CD147^high cells were injected subcutaneously into the flank of NOD/SCID mice with a 23-gauge needle. Tumors were excised and weighed after 12 weeks.

Results

CD147^high Tumor Cell Subpopulations Exhibit Elevated Invasiveness, Anchorage-Independent Growth, Spheroid Formation, and Drug Resistance Compared with CD147^low Subpopulations

CD147^high and CD147^low subpopulations were prepared by FACS from ST88-14 human MPNST cells and SKOV3 human ovarian carcinoma cells as described in Materials and Methods. Results from flow cytometry analyses of CD147 distribution and Western blot analyses of the levels of total CD147 in the CD147^high and CD147^low subpopulations are shown in Figure 1. The CD147^high cells typically expressed more CD147 than the CD147^low cells (Figure 1); however, in some preparations the difference in total expression was insignificant, despite a clear-cut difference in the level of cell-surface CD147. Very similar subpopulations of cells were obtained from MDA MB231 human mammary carcinoma cells (data not shown).

Figure 1.

Flow sorting of CD147^low and CD147^high subpopulations from MPNST cells and SKOV3 ovarian carcinoma cells. MPNST cells (A) and SKOV3 cells (B) were sorted with FACS for CD147. Those cells expressing the highest levels of cell-surface CD147 (the top 10% of SKOV3 cells and 20% of MPNST cells) were selected as the CD147^high fraction. Conversely, those cells expressing the lowest levels of cell-surface CD147 (the bottom 10% of SKOV3 cells and 20% of MPNST cells) were selected as the CD147^low fraction. Sorting of MDA MB231 cells was performed in a similar fashion as the SKOV3 cells (data not shown). Western blot analysis was performed of whole-cell lysates of CD147^low and CD147^high fractions from MPNST cells (25 μg protein) (A) and SKOV cells (50 μg protein) (B). Only the CD147 band is shown (A and B, upper right corner), at approximately 55kDa; a faint band at approximately 35 kDa was also present in each case (not shown).

We used a Transwell Matrigel invasion assay (BD Biosciences) to determine the constitutive differences in invasiveness of the CD147^high and CD147^low subpopulations. The CD147^high subfractions of both MPNST and SKOV3 cells were highly invasive, but their CD147^low counterparts were not (Figure 2, A and B). Virtually identical results were obtained for MDA MB231 cells (data not shown). To confirm that differences in invasiveness were directly related to expression of cell-surface CD147, we infected the CD147^low subfraction of SKOV3 cells with an adenovirus driving expression of CD147. This led to a dramatic increase in invasion, to a level similar to that of the CD147^high subfraction (Figure 2C).

Figure 2.

Increased invasiveness of CD147^high versus CD147^low cells. CD147 flow-sorted subpopulations of MPNST (A) or SKOV3 (B) cells were plated on Matrigel-coated invasion chambers and analyzed. C: CD147^high and CD147^low subpopulations of SKOV3 cells were compared with CD147^low cells infected with CD147 or β-gal adenovirus. Data are expressed as means of three or more independent experiments. *P < 0.05 between CD147 subpopulations. Micrographs are representative of three or more independent experiments. Original magnification, ×50.

We also tested the CD147^high and CD147^low subpopulations of MPNST and SKOV3 cells for their ability to grow as colonies in soft agar, a measure of the ability of cells to grow in an anchorage-independent manner, which is characteristic of transformed cells.⁴¹ Strikingly, the CD147^low subpopulations of both cell types repeatedly failed to grow colonies, whereas the CD147^high subpopulation of both cell lines demonstrated a strong constitutive ability to form colonies in soft agar (Figure 3A). The ability to form spheroids under nonadhesive culture conditions is a common feature of cancer stem-like cells isolated from carcinoma cell lines or primary carcinomas.^6,7,42,43 We therefore tested the ability of CD147^high and CD147^low subpopulations of SKOV3 ovarian and MDA MB231 breast carcinoma cells to form spheroids and found that the CD147^high cells readily form spheroids, whereas CD147^low cells do not (Figure 3B). We also examined cell proliferation under routine cell culture conditions; however, despite the differences in anchorage-independent proliferation observed (Figure 3, A and B), we found no significant differences in anchorage-dependent proliferation between CD147^high and CD147^low subpopulations for SKOV3 ovarian carcinoma or MPNST cells (Figure 3C). This is in accord with previous results showing that experimental elevation of CD147 expression leads to increased anchorage-independent but not anchorage-dependent growth.³²

Figure 3.

Increased anchorage-independent growth and spheroid formation in CD147^high versus CD147^low cells. A: CD147 flow-sorted subpopulations of MPNST or SKOV3 cells were cultured in soft agar for analysis of colony formation. B: CD147 flow-sorted subpopulations of MDA MB231 or SKOV3 cells were cultured in nonadhesive wells for analysis of spheroid formation. Colonies (A) or spheroids (B) formed from CD147^high but not CD147^low subpopulations of cells. C: Proliferation under routine anchorage-dependent cell culture conditions was also compared over a 96-hour period, but no significant differences were observed. Original magnification, ×10.

We then measured the effects of doxorubicin on cell survival of the CD147^high and CD147^low subpopulations prepared from MPNST and SKOV3 cells. The CD147^high subpopulations were approximately 100 to 1000 times more resistant to doxorubicin than the corresponding CD147^low subpopulations (Figure 4, A and B). The approximate half-maximal inhibitory concentration (IC₅₀) values obtained in this series of experiments were 10 μmol/L for the MPNST-CD147^high cells, compared with 0.01 μmol/L for the MPNST-CD147^low cells, and 5 μmol/L for the SKOV3-CD147^high cells, compared with 0.05 μmol/L for the SKOV3-CD147^low cells.

Figure 4.

Increased drug resistance and ABC transporter activity in CD147^high versus CD147^low cells. A and B: Cell survival after treatment with doxorubicin (Dox) (left panels) and FURA-2 AM efflux (right panels) was analyzed with CD147 flow-sorted subpopulations of MPNST (A) and SKOV3 (B) cells. C: Unsorted control, uninfected SKOV3 cells were infected with CD147 or β-gal adenovirus. Data are expressed as means ± SEM of triplicate wells. Cell survival results are representative of three or more independent experiments; FURA-2 AM efflux results are representative of four or more independent experiments. *P < 0.05 between CD147 subpopulations (A and B) or between β-gal and CD147 adenovirus-infected SKOV3 cells (C).

Drug resistance is often due to enhanced drug export by ATP-dependent efflux pumps such as Pgp. We therefore measured the levels of efflux of FURA-2 AM, a fluorescent substrate for Pgp. In cells expressing high levels of Pgp, intact FURA-2 AM is rapidly transported out of the cell, whereas in the absence of transporter activity it is cleaved to FURA 2 and accumulates in the cytosol. Thus, transporter activity is inversely proportional to accumulation of cytosolic FURA 2 fluorescence.⁴⁰ Consistent with our data on doxorubicin resistance, the CD147^high subpopulation from both cell types exhibited increased efflux, compared with the CD147^low subpopulations (Figure 4, A and B). To show that differences in drug resistance were directly influenced by CD147 levels, we also compared SKOV3 cells infected with adenovirus driving expression of CD147 with SKOV3 cells infected with control β-gal adenovirus. Increased CD147 caused the cells to be at least 1000 times more resistant to doxorubicin than the controls (Figure 4C) and to efflux higher amounts of FURA 2 (Figure 4C).

Primary CD147^high Cell Subpopulations from MMTV-PyMT Mouse Mammary Carcinomas Exhibit Elevated Invasiveness and Spheroid Formation

To determine whether the differences in properties of CD147^high and CD147^low subpopulations described above were restricted to cell lines, we separated these subpopulations from primary mouse tumor cells isolated from MMTV-PyMT mouse tumors (Figure 5A). Mouse CD147 is typically more heterogeneous in size than human CD147.³⁹ In a similar manner to the cell lines, primary CD147^high cells from these tumors were much more invasive than the corresponding CD147^low cells (Figure 5B). Furthermore, the primary CD147^high cells readily formed spheroids under nonadhesive culture conditions, whereas CD147^low cells did not (Figure 5C).

Figure 5.

Increased invasiveness and spheroid formation in primary CD147^high versus CD147^low cells from MMTV-PyMT mouse mammary tumors. A: Tumor cells were isolated from MMTV-PyMT adenocarcinomas and sorted with FACS for CD147. The top 10% of cells expressing the highest levels of cell-surface CD147 were selected as the CD147^high fraction and the bottom 10% were selected as the CD147^low fraction. Western blot analysis of whole-cell lysates of CD147^low and CD147^high fractions is shown in the upper right corner. B and C: CD147 flow-sorted subpopulations were plated on Matrigel-coated chambers for analysis of invasiveness (B) and were cultured in nonadhesive wells for analysis of spheroid formation (C). Original magnification, ×10. *P < 0.05.

CD147^high Cells Have Elevated Levels of Membrane-Localized CD44, EGFR, and Transporters

Previous studies from this and other laboratories have shown that hyaluronan-CD44 interactions promote formation of signaling complexes containing receptor tyrosine kinases (RTKs) and transporters in the plasma membrane of cancer cells^44,45 and that CD147 is associated with these complexes.^26,31,46 We therefore examined the levels of RTK, EGFR, and the transporters Pgp, BCRP, and MCT4 at the surface of CD147^high and CD147^low subpopulations from MPNST and SKOV3 cells by FACS analysis. The CD147^high subpopulations expressed higher levels of CD44, EGFR, Pgp, BCRP, and MCT4 in their membranes than did the CD147^low subpopulations (Figure 6). Furthermore, CD147^high cells produced higher amounts of hyaluronan, the major ligand for CD44 (data not shown). However, Western blot analyses did not show consistent differences in total levels of expression of CD44, EGFR, Pgp, BCRP, or MCT4 (data not shown). Thus, cells with relatively high levels of cell-surface CD147 exhibit correspondingly high levels of these other proteins at the cell surface.

Figure 6.

Increased membrane localization of CD44, EGFR, and transporters in CD147^high versus CD147^low cells. CD147 flow-sorted subpopulations of MPNST (A) and SKOV3 (B) cells were analyzed by flow cytometry for membrane localization of CD44, Pgp, BCRP, EGFR, and MCT4.

CD147^high Tumor Cell Subpopulations Are More Tumorigenic than CD147^low Subpopulations

Finally, we compared the capacity of constitutively CD147^high and CD147^low subpopulations of cells to form tumors in vivo. The relative weights of tumors formed by CD147^high and CD147^low subpopulations of SKOV3 cells grown as xenografts in NOD/SCID mice are shown in Figure 7. Significantly larger tumors were formed by the CD147^high cells than the CD147^low cells after injection of equivalent cell numbers (10⁴ cells per mouse) (Figure 7). Injection of 10 times lower numbers of CD147^high cells (10³ cells per mouse) resulted in variably sized tumors of higher average size than those formed from 10⁴ CD147^low cells; injection of higher numbers of CD147^low cells (10⁵ cells per mouse) produced smaller tumors than either 10³ or 10⁴ CD147^high cells (data not shown). In these cases, however, the differences were not statistically significant.

Figure 7.

CD147^high SKOV3 cells (CD147-lo) are more tumorigenic than CD147^low cells (CD147-hi) in vivo. CD147^high and CD147^low subpopulations were isolated from SKOV3 human ovarian carcinoma cells by flow sorting and were injected into the flanks of NOD/SCID mice (10⁴ cells per mouse). Mice were sacrificed after 12 weeks and tumor weight was measured. Data are expressed as means ± SEM. n = 5 mice per group. *P < 0.05.

Discussion

Previous studies from our laboratory and those of other investigators have demonstrated that experimental manipulation of CD147 levels influences tumor cell invasiveness,^17,25 anchorage-independent growth and anoikis,^32,47 drug resistance,^24,27 lactate transport,^20,26 and tumor progression.^15,33,34

In the present study, we examined the properties of cell subpopulations, with different constitutive levels of cell-surface CD147, that were derived directly from malignant human cell lines or primary mouse tumor cells. Subpopulations of cells that exhibit constitutively high levels of cell-surface CD147 were found to be highly invasive, anchorage-independent, drug resistant, spheroid-forming, and tumorigenic, whereas cells that exhibit constitutively low levels of cell-surface CD147 were found to be noninvasive, anchorage-dependent, drug sensitive, non-spheroid-forming, and less tumorigenic. Our flow cytometry results (Figures 1 and 5A) indicate that each of the cell types used in the present study exhibits a continuum of cell-surface CD147 expression, rather than a bimodal distribution. Thus, our selection of CD147^high and CD147^low cells is arbitrary, and it is to be expected that subpopulations of cells with intermediate properties would also be present. Likewise, it is unknown whether CD147^high and CD147^low cells arise stochastically or represent hierarchical cell subpopulations.

It is striking that these dramatic differences in cellular properties correspond with different levels of cell-surface CD147, but not necessarily with different levels of total CD147, because differences in the total expression of CD147 in CD147^high versus CD147^low cells were often small or insignificant. These observations suggest that there are considerable differences in transit of CD147 to or from the cell surface in these two subpopulations. A previous study of trafficking of CD147 and MCT4 to the cell surface showed that the levels of cell-surface CD147 and MCT4 were mutually dependent on their interaction with one another.²⁰ Because the CD147^high cells studied here also exhibited higher levels of cell-surface MCT4, it is possible that similar effects on trafficking occurred in these cells. Interestingly, CD44, EGFR, Pgp, and BCRP were also higher in the membranes of CD147^high cells than CD147^low cells. In previous studies, we had shown that CD147 associates with CD44, Pgp, and BCRP at the cell surface,^24,26,46 but it is not known whether CD147 is directly responsible for trafficking of these proteins to the cell surface. It is likely, however, that the increased levels of these proteins at the surface of CD147^high cells are at least in part responsible for their more malignant properties.

CD147 has been associated with tumor progression and invasiveness ever since its discovery as a matrix metalloproteinase-inducing factor.^{15–17,33,34,48–53} In accord with these previous studies, we show here that subpopulations of cells expressing high constitutive levels of CD147 are much more invasive than those with low levels of CD147 expression. Recently, we have also shown that experimental overexpression of CD147 to a relatively moderate degree is sufficient to induce invasiveness and formation of active invadopodia in the phenotypically normal breast epithelial cell line MCF-10A, and that down-regulation or up-regulation of CD147 in the metastatic breast carcinoma cell line MDA MB231 correspondingly alters invasiveness and active invadopodia formation.²⁵ In that study, we showed that CD147 induces the expression of MT1-MMP and is present in close association with MT1-MMP in lipid raft-like domains.²⁵ In a more recent study, we demonstrated that CD147 is also associated with EGFR and CD44 in similar lipid domains and that CD147 induces invasiveness via up-regulation of EGFR-Ras-ERK and hyaluronan-CD44 signaling as well as MT1-MMP expression (Grass et al, unpublished data). We have also shown that increased CD147 induces formation of ErbB2-containing complexes within lipid microdomains, accompanied by increased ErbB2 activation.⁵⁴ These results indicate that CD147 plays a fundamental role in the signaling pathways leading to tumor cell malignancy. Two major questions remain unanswered. First, how does CD147 induce formation of these complexes? Second, does CD147 influence the fundamental organization of plasma membrane proteins (eg, into lipid microdomains) or promote protein trafficking and complex formation (thereby in turn influencing membrane organization)?

Recently, we observed enrichment of CD147-containing complexes, similar to those described above, in primary cancer stem-like cells isolated from ascites fluid from ovarian carcinoma patients,⁴⁶ suggesting that the properties of these highly aggressive, therapy-resistant subpopulations of cancer cells may be linked to formation of such complexes. Our results here, showing that CD147^high cells are much more tumorigenic than CD147^low cells, also suggest a relationship between CD147-containing complexes and cancer stem-like cells. Likewise, we have demonstrated that CD147^high cells are much more drug-resistant than CD147^low cells and more readily form spheroids, properties that are also characteristic of cancer stem-like cells.^2,4

Another striking observation is the relationship between the activities of CD147 and CD44, a major marker for cancer stem-like cells.⁵⁵ Several studies have demonstrated close association of CD147 and CD44.^26,31,46 We have shown that up-regulation of CD147 induces synthesis of hyaluronan,³² the major ligand for CD44, as well as several downstream effects of hyaluronan-CD44 signaling, including anchorage-independent growth,³² lactate transport,²⁶ invasiveness (Grass et al, unpublished data), and ErbB2 signaling.⁵⁴ In previous studies, we have demonstrated the involvement of cell-surface complexes containing CD44 and RTKs or transporters. Moreover, we have shown that antagonists of these interactions with CD44 inhibit growth of tumors initiated by cancer stem-like cells.^46,56 The data reported here indicate that CD147, as well as CD44,^45,55 is an important target for treatment of aggressive, therapy-resistant cancers, and in particular their cancer stem-like subpopulations.