Tetraspanin CD151 regulates TGF beta signaling: implication in tumour metastasis

Abstract

Tetraspanin CD151 is associated with laminin-binding integrins and controls tumour cell migration and invasion. By analysing responses of breast cancer cells to various growth factors, we demonstrated that depletion of CD151 specifically attenuates TGFβ1-induced scattering and proliferation of breast cancer cells in 3-D Matrigel. CD151-dependent cell scattering requires its association with either α3β1 or α6 integrins but it is independent of the recruitment of CD151 to tetraspanin-enriched microdomains. We also found that CD151 regulates compartmentalisation of TGF-beta type I receptor (TβRI/ALK-5) and specifically controls TGFβ1–induced activation of p38. In contrast, signaling leading to activation of Smad2/3, c-Akt and Erk1/2 proteins was comparable in CD151(+) and CD151(−) cells. Attenuation of TGFβ1-induced responses correlated with reduced retention in the lung vascular bed, inhibition of pneumocytes-induced scattering of breast cancer cells in 3-D Matrigel and decrease in experimental metastasis to the lungs. These results identify CD151 as a positive regulator of TGFβ1–initiated signaling and highlight the important role played by this tetraspanin in TGFβ1–induced breast cancer metastasis.

INTRODUCTION

Tetraspanin protein CD151/Tspan24 is a member of a large family of four transmembrane domain proteins, which directly interacts with laminin-binding integrins (i.e. α3β1, α6β1/β4 and α7β1) (1, 2). It affects their ligand-binding properties, thus regulating adhesion-dependent signaling and post-adhesion events including cell migration (3). In addition, it has been shown that CD151 regulates internalisation of the associated integrins, a process which is integral to cell motility (4). Finally, we have recently reported that CD151 specifically changes glycosylation of α3β1 and CD151-dependent differences in glycosylation of this integrin contribute to the modulatory activity of CD151 towards this integrin (5).

In addition to this CD151 regulates activity of laminin-binding integrins via tetraspanin-enriched microdomains (TERM or TEM), molecular aggregates on the plasma membrane, which include other tetraspanins and cytoplasmic signaling proteins (3). Association with CD151 is critical for the recruitment of laminin-binding integrins to TERM, where they can co-operate with other TERM-associated transmembrane proteins (6).

Cooperation between integrins and receptors to various growth and chemotactic factors is well established and occurs at multiple levels (7, 8). This involves cell type specific association of various integrins with growth factor/chemokine receptors (GF-/ChR), integrin-mediated control over ligand-induced activation of GF-/ChR and adhesion-dependent modulation of signaling pathways downstream of activated GF-/ChR. Integrin-associated transmembrane proteins including CD47, tetraspanins and the L6-Ag family members can all function as mediators in communication between integrins and GF-/ChR (3,9-11). We and others have previously described the association of tetraspanins with transmembrane receptors to growth factors including epidermal growth factor receptor (EGFR) (12, 13), receptor to hepatocyte growth factor (c-Met) (14) and receptor for Steel factor (c-Kit) (15). Furthermore, it has been proposed that tetraspanin microdomains may represent a focal point where signaling through integrins and receptors for growth/chemotactic factors can be co-ordinated (3). Indeed, it has been shown that adhesion-dependent cellular responses triggered by activated c-Met are regulated by tetraspanins CD82 and CD151 (14, 16). Likewise, CD81 and CD9 regulated adhesion-dependent proliferation in response to fibroblast growth factor (17). In addition, tetraspanins may target growth factor receptors independent of their association with integrins (12, 14), and this may involve tetraspanin-dependent changes in surface compartmentalisation of the receptors (13, 18).

Here we found that depletion of CD151 attenuated responses of breast cancer cells to TGFβ1 in 3-dimentional ECM. This correlated with specific changes in the TGFβ1-induced signaling leading to activation of p38. Importantly, we found that attenuated responses to TGFβ1 correlated with decreased retention of CD151-depleted cells in the lungs, inhibition of tumour cells growth and scattering in co-culture with lung epithelial cells and, consequently, diminished metastatic burden of breast cancer cells.

MATERIALS AND METHODS

Cells, antibodies and reagents

MDA-MB-231/CD151(+), MDA-MB-231/CD151(−), MDA-MB-231/CD81(−), MDA-MB-231/α3β1^low, MDA-MB-231/α6β4^low, DCIS.com/CD151(+) and DCIS.com/CD151(−) cell lines were generated using appropriate shRNA lentiviral constructs with cells expressing low levels of the target proteins being selected after cell sorting (5, 19, 20). MDA-MB-231/CD151rec, MDA-MB-231/CD151palm and MDA-MB-231/CD151-QRD cell lines were established after transfection of MDA-MB-231/CD151(−) cells with the wild-type and appropriate mutants of CD151 and subsequent selection CD151-expressing cells by cell sorting (5). Cell lines expressing low levels of α6 integrins (MDA-MB-231/α6^lowβ1/β4 and MDA-MB-231/α3β1^lowα6^lowβ1/β4 were established by infecting of MDA-MB-231/α3β1^low cells with pLVTHM-based lentivirus encoding shRNA that targets α6 integrin subunit (target sequence – 5′-GGUCGUGACAUGUGCUCAC-3′). To generate GFP-expressing cells MDA-MB-231/CD151(+) and MDA-MB-231/CD151(−) were transduced with pLVTHM lentivirus (encoding GFP) which was produced in 293T cells using a standard protocol (20). All MDA-MB-231 and 293T cells were grown in DMEM (Sigma) supplemented with 10% foetal bovine serum (FBS). HPL1D cells (21) were cultured in Ham F12 media with 5 ug/ml insulin, 5 ug/ml transferrin, 10⁻⁷ Mhydrocortisone, 2×10⁻⁷ M thyronine and 1% of FBS. The mouse anti-CD151 mAb: 5C11 (22), 11B1G4 (provided by Dr.L.Ashman, University of Newcastle, Australia), NCL-CD151 (Novocastra). Rabbit polyclonal antibodies against TGFβR1 were from SantaCruz; mouse mAb against β-actin was from Sigma, mouse mAb against β4 integrin subunit (3E1) was from Chemicon. Rabbit polyclonal antibodies to α3 and α6 integrin were generously provided by Dr F. Watt (Cambridge, UK) and Dr A. Cress (Tuscon, AZ). Mouse monoclonal to α3- (A3-IVA5), α6- (A6-ELE) and β1- (TS2/16) integrin subunit were described previously (23-25). The rest of the antibodies used in this study were purchased from Cell Signaling Technology. All growth factors were purchased from PeproTech. SB431542 was from Calbiochem.

Culturing cells in 3-D Matrigel

Culturing of cells in 3-D Matrigel was previously described (20). In some experiments cells were cultured in Endothelial Basal Medium (Promocell) which was supplemented with 2% FBS and various growth factors/chemokines (at 10ng/ml); growth media was replaced every third day. To evaluate cell proliferation in 3-D Matrigel AlamarBlue (Invitrogen) was added at different time points to the growth media and cell were further incubated for 4h at 37°C. Fluorescence of the samples was measured using Fluoroskan plate reader. In cell co-culture experiments, HPL1D cells were grown in 48-well plates until they reached 80% confluency: MDA-MB-231 cells suspended in Matrigel (8×10³/200μL) were subsequently overlaid and cultured for 7 to 10 days with growth media (DMEM/10%FCS) being replaced every third day. In some experiments GFP-expressing MDA-MB-231 cells were premixed with HPL1D cells at the ratio 1:100 and incubated for 30-60 min on slowly rocking platform to induce formation of cell aggregates. Aggregates were carefully suspended in Matrigel and cultured for 7-10 days. For all 3-D culture experiments representative pictures were taken every 2-3 days using Nikon Eclipse TS100 microscope.

Analysis of TGFβ1-induced signaling

Cells spread on Matrigel-coated (10μg/ml) culture plates were serum starved overnight and subsequently stimulated with 10ng/ml TGFβ1 for different length of time. Cells were lysed in the ice-cold Laemmli buffer supplemented with 2mM phenylmethylsulfonyl fluoride, 10μg/ml aprotinin, 10μg/ml leupeptin, 100mM N_a3VO₄, 10mM NaF, 10mM Na₃P₄O₇. Proteins separated by SDS-PAGE were transferred onto nitrocellulose membrane and probed with various antibodies using a standard protocol.

Fractionation in sucrose gradient

Fractionation of cellular lysates in the gradient of sucrose was carried out as described previously (13, 18). In brief, cells were homogenised in 100mM Na₂CO₃ pH 11.0 by passing through a 25G hypodermic needle and the lysates were subsequently sonicated (4×20 s bursts, 1 min interval, power 20). The lysates were mixed with two volumes of 2 M sucrose solution and overlaid with two sucrose solutions (35% and 5%). Samples were centrifuged at 100,000 g for 16-18 hours at 4°C in a Beckman SW60 rotor, and 10 equal volume fractions were collected from the top of gradient. The pellet was suspended in 200 μl of the lysis buffer. Equal amounts of each fraction were mixed with 4x Laemmli loading buffer and the proteins were resolved by SDS-PAGE. Fractional distribution of proteins was subsequently analysed by western blotting.

Flow cytometry analysis

Cells were detached with enzyme free cell dissociation buffer (Gibco) and kept for 1h on ice with saturating concentrations of appropriate mouse monoclonal antibodies. Staining was visualised using Phycoerythrin-conjugated goat anti-mouse IgG antibodies and samples were analysed with COULTER Epics XL.

Metastasis experiments

All experiments were performed in accordance with institutional and national animal research guidelines. Cells were injected intravenously into the lateral tail vein of 4-week-old to 5-week-old athymic mice to evaluate lung colonization. Mice were injected with 5×10⁵ cells/150μl PBS/mouse. Six to ten animals per group were used in the experiments. The experiments were terminated 8 weeks post injection (or earlier if mice adverse symptoms were observed). At termination, the lungs were removed and fixed in 10% buffered formalin and processed for histological and immunohistochemical analyses.

Histology and morphometry

Paraffin sections of formalin-fixed lungs were processed routinely for histology. The sections stained with Haematoxylin and Eosin were analyzed and images captured using the Nikon E400 DS-U2/L2 software (Japan). For quantitative analysis, 4 sections at the distance of 200 μm on either side of the hilum were cut and immunostained for vimentin (Dako, UK) using a standard avidin-biotin-peroxidase complex (ABC) method. Metastatic foci were counted in 2 low power fields (4x) at each level and their area quantified using the AxioVision LE software (Zeiss, Germany). The numbers represent the average from 8 counts/measurements. The mean ± SEM were calculated and statistical analyses were carried out by the unpaired t test using a GraphPad Prism 3.2 program (GraphPad Software, San Diego, CA, USA).

Analysis of tumour cell trafficking in vivo using intravital microscopy

Male C57BL/6 mice were anaesthetised with ketamine hydrochloride (100 mg/kg Vetalar; Amersham Biosciences and Upjohn Ltd., UK) and xylazine hydrochloride (10 mg/kg; Millpledge Pharmaceuticals, UK). Animals subsequently underwent carotid artery cannulation to allow administration of cells fluorescently labelled with BCECF-AM (PPL 40/2749). Following laparotomy, the left hepatic lobe was exteriorised and visualised intravitally using an inverted fluorescent microscope (Olympus, Middlesex, UK). 1 × 10⁶ MDA-MB-231/CD151(+) or MDA-MB-231/CD151(−) cells were administered in a 100ul bolus via the carotid artery. Seven fields of view containing a post-sinusoidal venule and surrounding sinusoidal capillaries were analysed 1 hr post injection of cells. Lung tissue was isolated from sacrificed mice at the end of the intravital experiments. Tissue was washed in 0.9% saline and also analysed on the inverted fluorescent microscope. Ten fields of view were selected in a set pattern and adherent cells counted. The average from these ten fields was obtained from each of the five separate experiments.

RESULTS

Depletion of CD151 affects responses of breast cancer cells to TGFβ1

We have analysed whether the absence of CD151 can affect responses of breast cancer cells to various growth factors and chemokines. Initially, we used MDA-MB-231/CD151(+) and MDA-MB-231/CD151(−) cells and assessed growth of these cells in 3-D Matrigel. In agreement with the previous studies, we found that depletion of CD151 did not affect surface expression of predominant integrins in MDA-MB-231 cells (Supplementary Fig.1). Ten growth/chemotactic factors were analysed in these experiments including PDGF-BB, HB-EGF, EGF, bFGF, SDF-1, TGFα, HGF, TGFβ1, IGF-1 and IL-1. Under the control conditions (i.e. in the absence of additional growth factors) both MDA-MB-231/CD151(+) and MDA-MB-231(−) cells formed compact round aggregates (Fig. 1A, upper panels). The morphological appearance of colonies was not affected when cells were cultured in the presence of HGF, IL-1, EGF, IGF-1, PDGF-BB, SDF1, HB-EGF, TGFα (Supplementary Fig.2 and results are not shown). Tight intercellular interactions within aggregates were looser in the presence of bFGF but we observed no apparent difference between the CD151(+) and CD151(−) cells (Supplementary Fig.2). By contrast, the CD151-dependent differential response was seen when cells were grown in the presence of TGFβ1. Whilst less than 20% of CD151-positive aggregates retained compact morphology, the majority of colonies appeared as aggregates of loosely associated cells, some of which developed characteristic thin protrusions (thereafter we referred to this type of colonies as having a “scattered” phenotype) (Fig.1A, left lower panel). Short protrusions were first observed 4 days after cells were seeded in Matrigel and they gradually became more prominent after 8 days in 3-D culture (Fig.1B). In contrast to CD151-positive cells, ~80% of MDA-MB-231/CD151(−) colonies remained compact and had smooth contours during the time-course of the experiments (Fig.1A, right lower panel and Fig.1C). Although there was no obvious difference between CD151-positive and CD151-negative cells in the number of colonies formed in the presence of TGFβ1, the proliferative rate of MDA-MB-231/CD151(−) cells appeared to be consistently lower (Fig.1D). Similarly, TGFβ1–induced proliferation of another breast cancer cell line (DCIS.com) was also attenuated after depletion of CD151 (Supplementary Fig.3). Importantly, the effect of CD151 depletion on TGFβ1-induced signaling was specific as knockdown of tetraspanin CD81 did not change responses of MDA-MB-231 cells (Supplementary Fig.4). These data showed that CD151 specifically regulates TGFβ1–induced responses of breast cancer cells in 3-D Matrigel.

Figure 1.

CD151 regulates responses of MDA-MB-231 breast cancer cells to TGFβ1. A, E. Cells were embedded into growth factor reduced Matrigel as described in Materials and Methods and grown for 8 days without (top panel) or in the presence of TGFβ1 (10ng/ml). Presented photographs of representative fields. Scale bar represents 50μm. B. The time-course of cellular responses to TGFβ1 in Matrigel. Growth of cells in 3-D Matrigel was analysed as described in (A). Scale bar represents 50μm. C. Quantification of TGFβ1-induced changes in morphology of the colonies in 3-D Matrigel. More than 50 colonies were analysed for each of the cell lines in each of three separate experiments. Colonies were defined as having a “compact phenotype” when no individual cell borders could be discerned within the colony. Ratio of average number of “compact” colonies for CD151-positive and CD151-depleted MDA-MB-231 was calculated for each experiment. Data are presented as mean of ratios ± standard deviation. D. Measurements of cell proliferation in the presence of TGFβ1. Cell proliferation was measured using Alamar Blue by a spectrofluorometric method. Data represent mean ±SD of three experiments (each in triplicate). AFU, Arbitrary Fluorescence Units.

The role of CD151-associated integrins and tetraspanin microdomains in regulating cellular responses to TGFβ1

We and others have previously reported that in MDA-MB-231 cells the function of CD151 is closely linked to laminin-binding integrins (i.e. α3β1, α6β1 and α6β4) (5, 20, 26). To examine the involvement of these integrins in CD151-dependent scattering induced by TGFβ1 we initially examined the responses of MDA-MB-231/CD151-QRD cells (5). These cells were derivative of MDA-MB-231/CD151(−) line and they expressed the mutant of CD151 deficient in its ability to form complexes with integrins (5). As illustrated in Figure 2, the response of MDA-MB-231/CD151-QRD cells was comparable to that of the CD151-negative cells with ~90% of all colonies remained compact in the presence of TGFβ1. By contrast, the response to TGFβ1 was restored when MDA-MB-231/CD151(−) cells were reconstituted with the wild-type CD151 (Fig.2A and B). Flow cytometry experiments have shown that CD151-QRD and CD151wt proteins expressed at the comparable levels on the surface of reconstituted MDA-MB-231 cells (Supplementary Table 1). We next investigated which of the laminin-binding integrins contribute to CD151-dependent responses of MDA-MB-231 cells to TGFβ1. To this end we have generated two new cell lines: MDA-MB-231/α6β1/β4^low – deficient in expression of both α6 integrins (i.e. α6β1 and α6β4) and MDA-MB-231/α3β1^lowα6β1/β4 ^low – deficient in expression of all laminin-binding integrins (Supplementary Fig.1A). In these experiments we also used α3β1-deficient cells (MDA-MB-231/α3β1^low) (20). Surprisingly, both MDA-MB-231/α3β1^low and MDA-MB-231/α6β1/β4^low cells were able to form colonies with characteristic “scattered” appearance and only deficiency in all CD151-associated integrins completely abolished responses to TGFβ1 (Fig. 2C, Supplementary Fig.5). These results confirmed that the modulatory activity of CD151 requires its association with integrins and suggested functional redundancy of various CD151-integrin complexes in TGFβ1-induced scattering. Palmitoylation-dependent recruitment of the proteins to tetraspanin-enriched microdomains is critical for many of the tetraspanin activities (27). To investigate whether palmitoylation of CD151 is necessary for TGFβ1-induced responses, we analysed behaviour of MDA-MB-231/CD151palm(−) cells. As illustrated in Figure 2D expression of palmitoylation-deficient CD151 in MDA-MB-231/CD151(−) fully restored the ability of the cells to form scattered colonies in the presence of TGFβ1. These results indicated that the association with other tetraspanins (and recruitment to tetraspanin microdomains) was not essential for CD151-dependent cellular responses to this cytokine.

Figure 2.

The association with integrins is critical for CD151-dependent cellular responses to TGFβ1. Growth of cells in 3-D Matrigel (A) and quantification of cellular responses (B) were performed as described in the legend to Figure 1. C, D. MDA-MB-231, MDA-MB-231/CD151rec, MDA-MB-231/CD151-QRD, MDA-MB-231/CD151palm, MDA-MB-231/α3β1^low, MDA-MB-231/α6^lowβ1/β4 and MDA-MB-231/α3β1^lowα6^lowβ1/β4 cells were plated in 3-D Matrigel and grown in presence of TGFβ1 as described in Figure 1. Scale bar represents 50μm.

The effect of CD151 depletion on TGFβ1–induced signaling

Pleiotropic effects of TGFβ1 on cell behaviour is mediated by Smad proteins, which function as transcriptional modulators, and through activation of a number of Smad-independent signaling pathways involving PI3-, Erk1/2- and p38 kinases (28). To analysed how depletion of CD151 affects TGFβ1–induced signaling we assessed phosphorylation of various TGFβ1 targets in cells plated on Matrigel under serum-free conditions and subsequently stimulated with the cytokine for various time intervals. As expected, incubation with TGFβ1 resulted in phosphorylation of Smad2 and Smad3: the phosphorylation levels of the proteins peaked at 1 hour after the stimulation and then decreased to undetectable levels by 12 hours (Fig.3A and 3B). We observed no difference between CD151-positive and CD151-negative cells in the kinetics and degree of phosphorylation of Smad2/3 proteins in these experiments (Fig.3A and 3B). Thus, it appeared that CD151 does not influence Smad-dependent signaling pathway. Similarly, phosphorylation levels of Erk1/2 and c-Akt were comparable at all analysed time points (Fig.3A and 3B and not shown). By contrast, there was a consistent difference in TGFβ1–induced phosphorylation of p38. In MDA-MB-231/CD151(+) cells, after the initial sharp rise (~ 4-fold increase), the level of phospho-p38 has returned to its basal level (i.e. zero time point) by 48 hours. On the other hand, the degree of TGFβ1-induced phosphorylation of p38 in CD151-negative MDA-MB-231 cells was significantly lower when compared to non-stimulated cells, with maximal ~1.5 fold increase by 1 hour (Fig.3A and 3C). Furthermore, p38 phosphorylation was sustained at almost the same level for up to 48 hours. A specific CD151-dependent effect on TGFβ1–induced phosphorylation of p38 was also observed in DCIS.com cells (Supplementary Fig.6). These results showed that CD151 deficiency resulted in specific alterations in TGFβ1-induced signaling. It has been recently reported that TGFβ1–induced activation of p38 was influenced by cholesterol-dependent redistribution of TGF-beta type I receptor (TβRI/ALK-5) on the plasma membrane (29). Fractionation in sucrose gradient was used to examine whether the effect of CD151 on p38 activation also involves changes in membrane compartmentalisation of TβRI/ALK-5. Although the major pool of TβRI/ALK-5 molecules was found in the heavy fractions of sucrose gradient, a significant proportion of the receptor was floating in the light fractions (Fig, 3D). Importantly, we have consistently observed (3 independent experiments) that depletion of CD151 in MDA-MB-231 cells resulted in the decrease in the proportion of TβRI/ALK-5 in the light fraction of the gradient. Control experiments have demonstrated that the total cellular levels of TβRI/ALK-5 were comparable in CD151(+) and CD151(−) cells (Supplementary Fig.7). By contrast, fractional distribution of CD151 and the associated integrins was different with most of the signal detected in the fractions 3-6 (Fig.3D). Furthermore, depletion of CD151 resulted in redistribution of α3β1 in sucrose gradient fractions. In additional experiments we found no evidence for the interaction between CD151 and TβRI/ALK-5 (results are not shown).

Figure 3.

The effect of CD151 depletion on TGFβ1-induced signaling. A, B Serum-starved MDA-MB-231/CD151(+) and MDA-MB-231/CD151(−) cells were incubated with TGFβ1 (10ng/ml) for indicated time intervals and then scraped into Laemmli buffer. Proteins were resolved in 10% SDS-PAGE under reduced conditions. The proteins were transferred to a nitrocellulose membrane and probed with indicated Abs. C. Densitometric analysis of activation p38 (i.e. measurements of phospho-p38) for the results shown in Fig.3A. Measurements were done for three independent experiments. A.D.U. – arbitrary densitometry units. D. The role of CD151 in the membrane compartmentalisation of TβRI. Cell lysates from MDA-MB-231/CD151(+) and MDA-MB-231/CD151(−) were fractionated in gradient of sucrose as described in the Materials and Methods. Equal volumes of each fraction were resolved in 10% SDS-PAGE. Distribution of proteins in the gradient fractions was assessed by Western blotting using specific antibody Abs. The α2–, α3– and α6–panels represent western blots detecting corresponding integrin subunits.

Depletion of CD151 attenuates pulmonary metastasis of breast cancer cells

It has been reported that TGFβ1 plays a critical role in pulmonary metastasis of MDA-MB-231 (30). Thus, we used a well-established experimental lung metastases mouse model (i.e. intravenous injection of tumour cells into the tail vein) to examine whether attenuated responses to TGFβ1 in MDA-MB-231/CD151(−) would correlate with decreased metastatic potential of these cells. We found that although both CD151–positive and CD151–negative MDA-MB-231 cells gave rise to multiple metastatic lesions in the lungs (Supplementary Fig.8), the area occupied by the lesions and their number were decreased in the absence of CD151 (Fig.4A and 4B). In animals implanted with CD151-positive cells an average relative area of all lesions in all examined lung fields was ~1.65%. In comparison, this parameter was 0.46% for mice injected with MDA-MB-231/CD151(−) cells. These results indicate that CD151 is involved in establishing and growth of metastatic lesions in the lungs.

Figure 4.

Depletion of CD151 attenuated pulmonary metastasis. A. The areas occupied by MDA-MB-231/CD151(+) and MDA-MB-231/CD151(−) metastatic lesions were measured as described in Material and Methods and plotted relative to the total lung area (%). Error bars, SEM. B. Numbers of lesions counted in lung sections used for the morphometric assessment the lesions shown in (A). Error bars, SEM.

Depletion of CD151 attenuates the immediate trafficking of breast cancer cells to the lungs in vivo

Whilst the lung metastases model allows to assess the metastatic process as a whole, it doesn’t discriminate between different steps which might be affected by CD151. As a starting point towards uncovering the involvement of CD151 in metastasis we analysed whether CD151 regulates the immediate recruitment of breast cancer cells to the vasculature of the lungs. MDA-MB-231/CD151(+) and MDA-MB-231/CD151(−) cells were injected into the mouse vascular system via the carotid artery and their immediate recruitment to the lung and liver microcirculation was monitored by intravital microscopy. The liver was examined in addition to the lungs to determine whether CD151 is involved in site specific recruitment of breast cancer cells. A significant reduction in lung recruitment of CD151-negative MDA-MB-231 cells was observed (Fig. 5A and 5B). On the other hand, hepatic recruitment of CD151-positive and CD151-negative MDA-MB-231 did not differ during the 1 hour duration of the experiment (Fig. 5B).

Figure 5.

Depletion of CD151 diminishes acute retention of MDA-MB-231 in the lung vascular bed. BCECF – labelled CD151-positive and CD151-negative MDA-MB-231 were introduced via the carotid artery into anaesthetised mice and monitored intravitally. Numbers adherent within the hepatic sinusoidal and pulmonary capillaries were quantified in 7 and 10 random static fields of view, respectively. A. Shown photographs of representative fields. Note, a significant reduction in recruitment of labelled CD151-negative MDA-MB-231 cells in the lungs (right panel) as compared to the control, MDA-MB-231/CD151(+) cells (left panel). B. Quantification of the results from 5 independent experiments. Note, depletion of CD151 decreased retention of the cells in the lungs (top histogram) but not within the liver (bottom histogram). Error bars, SEM.

Depletion of CD151 attenuates responses of breast cancer cells to lung epithelial cells

To explore further the involvement of CD151 in pulmonary metastasis we compared responses of CD151-positive and CD151-negative MDA-MB-231 to lung epithelial cells. HPL1D is a human immortalised small airway epithelial cell line with cells retaining some of the characteristics of type II pneumocytes (21). In the initial experiments we analysed growth of MDA-MB-231 cells in 3-D Matrigel placed above the monolayer of HPL1D pneumocytes. Under these conditions most of the MDA-MB-231/CD151(+) cells developed characteristic scattered aggregates (Fig.6A, upper panels). These colonies morphologically resemble those induced by TGFβ1 (shown in Fig.1). In contrast, most of the CD151-negative cells were non-responsive to the presence of pneumocytes and formed compact aggregates with smooth contours (Fig.6A, lower panels). The CD151-dependent differential responses to pneumocytes were also observed when we cultured pre-aggregated clusters of HPL1D and MDA-MB-231 cells in 3-D Matrigel (Supplementary Fig.9). To examine whether TGFβ1 is involved in pneumocyte-induced scattering, we performed the co-culturing experiment in the presence of SB431542, a specific inhibitor for TβRI/ALK5. As shown in Figure 6B the presence of SB431542 completely abrogated the modulatory effect of pneumocytes on MDA-MB-231 cells. Thus, we concluded that in the absence of CD151 the responses of breast cancer cells to pneumocyte-derived TGFβ1 are suppressed and this is likely to be an important contributory factor for the attenuated metastatic properties of CD151-deficient cells.

Figure 6.

Contribution of pneumocytes and TGFβ1 to growth of MDA-MB-231 cells in 3-D Matrigel. Cells embedded in 3-D Matrigel were grown above the monolayer of HPL1D cells in the absence (A) or presence of 10μM SB431542 (B) as described in the legend to Figure 1. Presented photographs of representative fields. Scale bar represents 50μm.

DISCUSSION

Cytokines of TGFβ superfamily are implicated in a variety of normal and pathological phenomena (31). Here, we identified tetraspanin CD151 as a novel regulator of TGFβ–induced signaling and showed that attenuated cellular responses to TGFβ correlated with reduced metastastic potential of breast cancer cells.

Our results indicate that the modulatory activity of CD151 towards TGFβ1–induced cellular responses is dependent on its association with integrins and correlates with redistribution of TβRI/ALK5 on the plasma membrane. Notably, this does not require recruitment of CD151 to tetraspanin microdomains, critical structural and functional units on the cell surface (3). The fact that TβRI/ALK5 and CD151 do not interact with one another suggests that the mode of regulation involves more general CD151-dependent changes on the plasma membrane which may, as a result, affect distribution of TGFβ receptors. In this regard, not only does CD151 directly bind cholesterol (32), but tetraspanins may also control the expression level and compartmentalisation of gangliosides (13), glycosphingolipids that represent the major structural block of various types of membrane microdomains (33). It is also feasible that in addition to its regulatory role in surface distribution of TβRI/ALK5, CD151 (in complex with laminin-binding integrins) can alter signaling down-stream of the activated receptor. Although this selective mode of action was not previously associated with tetraspanins, it has been extensively documented for various integrin-growth factor receptor pairs (7, 9).

We found that TGFβ – induced scattering (MDA-MB-231 cells) and proliferation (MDA-MB-231 and DCIS.com) of CD151-deficient cells in 3-D Matrigel were dramatically reduced when compared to the control, CD151-positive cells. Attenuated responses to TGFβ1 in CD151-depleted cells correlated with differences in TGFβ–induced activation of p38. On the other hand, signaling pathways leading to activation of Erk1/2, c-Akt and phosphorylation of Smad2/3 were not affected in CD151(−) cells. Whilst CD151-dependent variations in activation of p38 may not be the only factor that underlies differential responses to TGFβ1 they can certainly contribute to CD151-dependent phenotypes in 3-D Matrigel. Indeed, a pre-proliferative function of p38 in various types of cancer cells is well established (34). Although there appears to be no study which specifically addressed the involvement of p38 in scattering of cells in 3-D ECM, activation of this enzyme can certainly contribute to this phenomenon. In this regard, it has been recently reported that TGFβ–induced transcription of matrix metalloproteinase 2, an enzyme which plays an important role in pericellular proteolysis of ECM by mammary epithelial cells, is controlled by ATF2, a well-established target for p38 (35). Furthermore, activation of p38 was shown to be critical for TGFβ1– induced activation MT1-MMP/MMP-14 (36), a key enzyme for invasive migration within 3-D ECM (37). p38 is also known to regulate expression levels and turnover of various cell-cell adhesion proteins and therefore may further contribute to the development of “scattering phenotype” in response to TGFβ1 treatment (38-40). In spite of this, it appears that by itself modulation of p38 activity in response to TGFβ1 is not sufficient to induce cell scattering in 3-D ECM. Indeed, whilst TGFβ1 could potently induce activation of p38 in DCIS.com/CD151(+), the colonies formed by the cells in 3-D Matrigel in the presence of this growth factor did not exhibit typical scattering phenotype (i.e. loosening cell-cell contacts, developing characteristic invasive protrusions). These data strongly suggest that TGFβ1-induced activation of p38 and its contribution to scattering in 3-D ECM relies on other signalling pathways which are either absent or not fully functional in DCIS.com cells.

Although as a whole cellular interactions with LN-111 in Matrigel matrix are essential, the function of each of the laminin-binding integrins in developing of TGFβ1– induced “scattering” phenotype by MDA-MB-231 cells seems to be redundant. Indeed, only in α3/α6–deficient MDA-MB-231 cells did we observe complete abrogation of the “scattering” response to TGFβ1. The fact that MDA-MB-231/α6β1/β4^low cells are able to respond to TGFβ1 is particularly intriguing and raises the question of how α3β1, the only remaining laminin-binding integrin in these cells, supports scattering behaviour. It is now well established that whereas α6β1 and α6β4 integrins represent principal receptors for LN-111 and can deliver biomechanical signals necessary for scattering, α3β1 does not bind this substrate (41) and, therefore, should utilise an alternative mode of action. One possibility is that whilst other adhesion receptors (and ECM components of Matrigel) in α6–deficient MDA-MB-231 cells provide a “mechanical” support for scattering, the CD151-α3β1 complex regulates development of cell protrusions and movement in 3-D ECM in adhesion-independent manner. Alternatively, in the absence of α6 integrins and upon treatment with TGFβ1, α3β1 controls scattering by mediating adhesion to non-laminin components of Matrigel (42). Further experiments will be necessary to address this question.

Deficiency in responses to TGFβ1 correlated with impaired metastatic potential of CD151-depleted MDA-MB-231 cells. Our observation is complementary to a recent report demonstrating that signaling induced by TGFβ via the Smad-dependent pathway(s) is(are) critical for recruitment to and retention of breast cancer cells in the lungs (30). Interestingly though, we found that in the absence of CD151 TGFβ – induced phosphorylation of Smad proteins was not affected. Therefore, it is unlikely that the mechanisms underlying pro-metastatic activity of CD151 involve the induction of angiopoietin-like 4 (ANGPTL4) as shown by Padua and colleagues (30). In fact, our data suggest that by potentiating cellular responses to TGFβ1 CD151 promotes the metastatic process at two levels: it enhances cell proliferation and facilitates local invasiveness (hence, increased size and number of metastatic lung lesions formed by CD151-positive cells).

Our results indicate that breast cancer cells, which express CD151, are more readily retained in the lung microvascular bed. These results revealed yet an additional metastasis promoting function for CD151. Although relatively small CD151-dependent increase in the numbers of tumour cells in the lung microvascular bed was highly reproducible and specific: indeed, we found that recruitment of cells in the liver vasculature was independent of CD151 expression. These data further emphasise differences in the molecular mechanisms underlying organ-specific recruitment during the metastatic process. A recent report by Zijlstra and colleagues has shown that CD151 may contribute to metastasis by regulating intravasation of the cells into the vascular system at the primary tumour site (43). In their experiments, metastasis-suppressing anti-CD151 mAb did not affect tumour cell arrest in the lungs or alter growth of secondary colonies. Whilst these “discrepancies” can be simply explained by differences in the cell models used, it is also likely that the antibody-induced inhibitory pathways may be distinct from the intrinsic cellular mechanisms which control metastatic potential of tumour cells.

In conclusion, we have identified tetraspanin CD151 as a new modulator of TGFβ – induced signaling. Although in this study we have specifically focussed on TGFβ – induced responses of breast cancer cells, it is likely that a newly established functional link between CD151 and TGFβ1 may be in operation during the metastatic progression in many other cancers where elevated expression of the tetraspanin will sensitise cancer cells to TGFβ cytokines. Thus, future experiments towards unravelling the molecular details of the CD151-TGFβ1 communication will be necessary to understand TGFβ – dependent cancerous progression.