Clinical and functional significance of loss of caveolin-1 expression in breast cancer-associated fibroblasts

Abstract

Loss of caveolin-1 (Cav-1) expression in breast cancer-associated fibroblasts (CAFs) is predictive of poor prognosis in breast cancer, but its function has not been established. Our study tested the hypotheses that loss of Cav-1 expression in breast fibroblasts was associated with poor prognosis in breast cancer, through promotion of breast cancer cell invasion. Cav-1 stromal expression was immunohistochemically assessed in 358 breast cancers. Cav-1 expression in primary breast fibroblasts was analysed by western blot. Modified Boyden chamber assays determined fibroblast ability to promote invasion of breast cancer cells. The impact of siRNA silencing of Cav-1 in fibroblasts was evaluated using invasion assays and 3D co-culture assays. Loss of Cav-1 expression in breast stroma was significantly associated with decreased breast cancer-specific and disease-free survival (p = 0.01). Mean survival was 72 months (Cav-1⁺ group) versus 29.5 months (Cav-1⁻ group). This was confirmed in multivariate analysis. Cav-1 expression was significantly decreased in CAFs compared to normal fibroblasts (p = 0.01) and was associated with increased invasion-promoting capacity. Cav-1 siRNA-treated fibroblasts promoted significantly increased invasion of MDA-MB-468 and T47D breast cancer cells from 27% (control) to 67% (p = 0.006) and from 37% to 56%, respectively (p = 0.01). 3D co-cultures of MDA-MB-468 cells with myoepithelial cells led to the formation of organized cohesive structures when cultured with conditioned media from fibroblasts but resulted in a disorganized appearance in the presence of conditioned media from Cav-1 siRNA-treated fibroblasts, accompanied by loss of E-cadherin expression in tumour cells. Our data confirm that loss of stromal Cav-1 in breast cancer predicts poor outcome. At a functional level, Cav-1-deficient CAFs are capable of significantly increasing the invasive capacity of breast cancer cells. Copyright © 2012 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

Introduction

There is growing recognition that the tumour microenvironment can impact upon tumour cell behaviour, with fibroblasts implicated as key modulators of cancer progression 1, 2. It has been shown that a sub-population of these cells, termed carcinoma-associated fibroblasts (CAFs), found in tumour stroma, undergo changes in protein expression which represent an ‘activated’ myofibroblastic phenotype, typically up-regulating markers such as α-smooth muscle actin (α-SMA) 3, 4. By imitation of activated ‘wound repair’ fibroblasts, CAFs are thought to promote carcinoma progression by stimulating epithelial cell growth, migration, and invasion 1, 5, 6.

Caveolin-1 (Cav-1) is the principal scaffolding protein of plasmalemmal caveolae, specialized lipid rafts located within the plasma membrane 7. Caveolae are 50–100 nm uncoated omega-shaped invaginations of the membrane, involved in vesicular transport, endocytosis, and abundant signalling cascades 8. Cav-1 has been identified as a stromal biomarker, which is predictive of poor prognosis in breast cancer by several independent groups 9–13. Loss of expression of Cav-1 in breast CAFs was associated with increased tumour progression, local metastases, and oestrogen receptor negativity 9, 11, all features related with poor outcome. This has recently been reported in melanoma 14, suggesting that this may also apply to other types of cancer. Sloan et al demonstrated that 10-year survival rates for breast cancer patients with tumour stroma expressing Cav-1 was 91%, yet fell to 43% in the group with no Cav-1 expression 10. These studies identify a potential role of Cav-1 in suppressing tumour progression, yet little explanation has been offered regarding the functionality of Cav-1 loss in CAFs. To date, neither the mechanism responsible for the down-regulation of Cav-1 nor the pathological effect of its loss is conclusively understood.

Changes in tumour stroma drive invasion and metastasis—hallmarks of malignancy responsible for treatment failure, recurrence, and death. Wound-healing stroma has been shown to stimulate cancer cell invasion in animal models 15, with myofibroblast cells principally implicated. Cav-1-deficient fibroblasts have been shown to exhibit characteristics of myofibroblasts, with a significant correlation in the transcriptome 16 and overexpression of a number of myofibroblast marker proteins 17. Pro-invasive cross-talk between cancer cells and CAFs may occur through secreted molecules such as growth factors, cytokines, and chemokines. Cav-1 null mammary fibroblasts have been shown to secrete increased levels of several pro-angiogenic and tumourigenic factors including vascular endothelial growth factor (VEGF), platelet-derived growth factor (PDGF), and interleukin-6 (IL-6) 16. Cav-1 and caveolae are associated with a plethora of signal transduction cascades 18, and thus loss of Cav-1 in breast cancer fibroblasts may be linked to an invasion-promoting phenotype through deregulation of these signalling pathways.

This study has analysed the predictive ability of Cav-1 stromal status in an independent cohort of invasive breast cancers. Noting a reduction in Cav-1 expression in primary fibroblasts derived from breast cancer samples, this study also analysed the ability of normal mammary fibroblasts in which Cav-1 had been experimentally silenced to promote tumour cell invasion. The methodology was designed to test the hypotheses that (i) loss of caveolin-1 in breast CAFs was associated with poor prognosis in breast cancer, and (ii) loss of caveolin-1 in breast CAFs promoted invasion of breast cancer cells in vitro.

Materials and methods

Clinical material and immunohistochemistry

Following ethical approval [Leeds (East) REC:06/Q1206/180], a tissue microarray (TMA) was assembled from 358 cases of breast cancer diagnosed at the Leeds Teaching Hospitals NHS Trust between 1987 and 2005 with a mean follow-up of 80.5 months. Patient characteristics are shown in Table 1. Five-micrometre serial sections were mounted onto Superfrost Plus slides (BDH, Poole, Dorset), dewaxed in xylene, and rehydrated through graded alcohols. Endogenous peroxidases were blocked (3% hydrogen peroxide, 10 min) and antigen retrieval was performed by heating slides in 10 mM sodium citrate buffer, pH 6.0, for 2 min using a pressure cooker. Slides were then mounted in Sequenza racks, washed in phosphate-buffered saline (PBS), and blocked with 10% casein in PBS for 30 min at room temperature. Antibodies to caveolin-1 (rabbit anti-human polyclonal, clone N20; Santa Cruz Biotechnologies, Santa Cruz, CA, USA; dilution 1:1000) and α-SMA (mouse anti-human monoclonal, clone 1A4; Dako, Glostrup, Denmark; dilution 1:500) were applied for 1 h at room temperature. After washing with PBS, slides were incubated for 40 min at room temperature with appropriate horseradish peroxidase (HRP)-conjugated secondary antibodies (Santa Cruz Biotechnologies). Bound antibodies were visualized using diaminobenzidine (En-vision, Dako) before counterstaining with haematoxylin, dehydration, and mounting with coverslips. Cav-1 immunohistochemistry was scored semi-quantitatively by SAS and DLH under the supervision of AMH, a specialized consultant breast histopathologist, as 0 (no/little expression; < 25% positive fibroblasts), 1 (intermediate expression; 25–75% positive fibroblasts), or 2 (strong expression; > 75% positive fibroblasts). SMA-stained serial sections were not formally scored but used to highlight the position of fibroblasts within the TMAs, thus ensuring that only fibroblasts were evaluated in Cav-1 stained TMAs.

Table 1. Clinicopathological features of the 358-case breast cancer cohort and their relationship with Cav-1 status

		Cav-1 status
Characteristic	Number (%)	+	−	p value
Age, years
< 50	105 (29)	47	42	0.0001
> 50	253 (71)	169	55
Tumour type
Invasive ductal	281 (78)	171	85	0.07
Lobular	39 (11)	30	5
Other	38 (11)	11	4
Grade
I	78 (22)	55	13	< 0.0001
II	156 (43)	105	32
III	124 (35)	54	53
Lymph node status
N0	162 (45)	101	45	0.62
N1–3	179 (50)	95	55
No nodes taken	18 (5)
ER status
Positive	281 (78)	174	72	0.05
Negative	77 (22)	28	22
Tumour size, mm
≤ 10	44 (12)	20	2	0.02
> 10	284 (79)	188	88
Multifocal	20 (6)
Unable to estimate	10 (3)
Treatment
Endocrine
Yes	281 (79)	174	72	0.05
No	77 (21)	28	22
Chemotherapy
Yes	86 (24)	50	21	1.0
No	272 (76)	101	43
Radiotherapy
Yes	124 (35)	65	75	0.05
No	126 (35)	48	31

Cell lines

MDA-MB-468 and T47D breast cancer cell lines were cultured from stocks originally purchased from the American Type Culture Collection (Manassas, VA, USA). The normal breast tissue fibroblast cell line HMFU-19 and myoepithelial cell line MYO1089 were gifts from Professor Mike O'Hare (Ludwig Institute, London, UK). Both lines were generated from isolated primary cell populations by transduction with SV40 large T antigen 19. Mycoplasma checks were performed every 2–3 months and were consistently negative.

Isolation of primary fibroblasts

Breast tissue was obtained from patients undergoing surgery for breast carcinoma (n = 3) or reduction mammoplasty (n = 6) after informed consent was provided (06/Q1206/180). Tissue surplus to histopathological diagnosis was selected and fibroblasts were isolated as described previously 20. Tissue was digested for 12 h in Dulbecco's modified Eagle's medium (DMEM), 10% fetal calf serum (FCS), 2 mM L-glutamine, 100 IU of penicillin and streptomycin, 2.5 µg/ml fungizone, and 400 IU of collagenase IA (all Sigma-Aldrich, Poole, UK). Following sedimentation at 1 g for 30 min, the supernatant was removed and plated in DMEM with 10% FCS, penicillin and streptomycin (100 U/ml), and fungizone (2.5 µg/ml). Cells were cultured for 24 h before the medium was aspirated and the cells washed with DMEM (to remove any non-viable cells) before re-feeding with DMEM supplemented with 10% FCS. All primary fibroblasts were used at passages 3–5. Cells were characterized using indirect immunofluorescence as previously described 21 (Supporting information, Supplementary Figure 1).

Figure 1. Use of α-SMA to identify CAFs and aid Cav-1 analysis. Cav-1 expression in CAFs (a), with adjacent α-SMA-stained serial section used to identify activated CAFs (b). Cav-1 is also expressed in tumour epithelium (c, arrows), whereas α-SMA is not (d, asterisks). Some cases were completely negative for Cav-1 (e), despite the presence of CAFs in α-SMA-stained serial sections (f). Original magnification: × 40

Caveolin-1 (Cav-1) knockdown

Small-interfering (si)-RNAs targeted to Cav-1 (Hs_CAV_7; AACTAAACACCTCAACGATGA and Hs_CAV_10; AAGCAAGTGTACGACGCGCAC) plus AllStars Neg siRNA AF 488 scrambled control were purchased from Qiagen (West Sussex, UK). Transfection was performed according to the manufacturer's reverse-transfection protocol (HiPerFect^® transfection reagent, Qiagen). Briefly, fibroblasts were seeded into 12-well plates at 1 × 10⁴ cells per well 24 h prior to transfection. Cultures were incubated for 6 h with 10 µl of 10 nM siRNA, diluted in 5% FCS. Control fibroblasts were transfected with scrambled (non-coding) siRNA. After incubation, plates were washed and cells were allowed to recover in normal growth conditions (10% DMEM) for 24 h post-transfection, after which invasion assays were performed as described below. Cav-1 knockdown in HMFU-19 fibroblasts was confirmed by western blot.

Western blot analysis

Intracellular proteins from whole cells were obtained by the direct application of lysis buffer [50 mM Tris–HCl (pH 7.4), 150 mM NaCl, 1% Triton X-100, and protease inhibitor cocktail; Roche Diagnostics, West Sussex, UK] onto cell monolayers grown to 80% confluence. After rocking at 4 °C for 4 min, the efficiency of cell lysis was examined under an inverted microscope. Cell lysates were then collected and centrifuged for 10 min at 4 °C to remove insoluble material. Proteins were quantified by the Bradford assay (Bio-Rad, Hertfordshire). Ten micrograms of protein per sample was separated by SDS-PAGE electrophoresis. Separated proteins were then transferred to a Hybond^™ ECL nitrocellulose membrane (Amersham Biosciences, Buckinghamshire, UK) using the X Cell II-Blot module (Invitrogen, Carlsbad, CA, USA) and blocked for 1 h in 5% milk–Tris buffered saline–Tween [TBST; 10 mM Tris–HCl (pH 8.5), 150 mM NaCl, 0.1% Tween-20]. Membranes were probed with anti-caveolin-1 (1:1000; N20; Santa Cruz Biotechnology) diluted in 5% milk–TBST for 1 h at room temperature. Membranes were then washed in TBST (3 × 5 min) and incubated for 45 min with HRP-conjugated polyclonal goat anti-rabbit antibody (1:5000; SC-2004; Santa Cruz Biotechnology). Immunoreactive protein was detected using an enhanced chemiluminescence detection system (Thermo Scientific, Loughborough, UK). To verify equal protein loading and transfer, membranes were also probed for β-actin antibody (clone AC-15; Sigma) at a 1:5000 dilution in milk–TBST.

Preparation of conditioned media (CM)

HMFU-19 fibroblasts were cultured and transfected with Cav-1 siRNA as described. Cells were then allowed to recover in normal growth conditions (10% DMEM) for 24 h. The cultures were rinsed twice in PBS and incubated for 48 h with serum-free media. Media were removed and centrifuged to remove cell debris and then stored at − 80 °C until use.

Matrigel invasion assays

Invasion assays were performed over 48 h as previously reported 20. Briefly, the underside of an 8.0 µm-pore polyethelene terephthalate track-etched membrane was coated with 10 µg/ml fibronectin (Sigma-Aldrich Ltd). The upper surface was coated with Englebreth-Holm-Swarm basement membrane (Matrigel; Becton, Dickinson and Company, Franklin Lakes, USA) at a concentration of 5 µg per filter. Breast cancer cells (MDA-MB-468 cells or T47D cells) were seeded at a concentration of 5 × 10⁴ into the upper chamber, with 1 × 10⁴ siRNA-treated HMFU-19 fibroblasts in the lower chamber. Assays were performed in quadruplicate. Cell counts were performed on haematoxylin and eosin (H&E)-stained membranes and an invasion index was calculated as a percentage of the number of cells on the lower membrane (invaded cells) compared with the total number of cells (on the upper and lower membranes).

3D culture

MDA-MB-468 breast cancer cells (2 × 10⁵) were cultured with MYO1089 myoepithelial cells (2 × 10⁵) in 200–400 µl of collagen gel at a concentration of 2 mg/ml, as previously described 22. 3D cultures were incubated with CM from Cav-1 siRNA-treated HMFU-19 fibroblasts or scrambled siRNA controls for 7 days prior to fixing in formalin for 2 h. 3D collagen gels were subsequently embedded in paraffin wax, sectioned, and stained with H&E.

Statistical analysis

Statistical analysis was performed using GraphPad Prism 5. Comparison between normal and tumour donors was carried out using an independent samples t-test. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated as an index of the donor group. For comparison of siRNA Cav-1 knockdown versus control experiments, a repeated-measures analysis-of-variance (ANOVA) analysis was performed. For multivariate analysis, SPSS was used. All statistical tests were two-sided. p < 0.05 was considered significant.

Results

Cav-1 expression in CAFs is associated with increased breast cancer-specific survival and disease-free survival

In order to identify CAFs, serial sections were stained with α-SMA. Figure 1a shows strong expression of Cav-1 in CAFs, with the adjacent serial section used to identify regions of activated CAFs stained with α-SMA (Figure 1b). This was a necessary control to ensure that only CAFs were scored; as shown in further serial sections, Cav-1 is sometimes expressed in tumour epithelium (Figure 1c, arrows), whereas α-SMA is not (Figure 1d, asterisks). Some cases were completely negative for Cav-1 (Figure 1e), despite the presence of CAFs in α-SMA-stained serial sections (Figure 1f). Thus, α-SMA expression can be used to highlight the position of fibroblasts within breast tissue, ensuring that only fibroblasts were assessed for Cav-1 immunoreactivity in subsequent experiments.

A TMA of 358 breast cancers was scored for stromal Cav-1 expression. At least 296 cases (82%) had sufficient stromal tissue and unambiguous staining to be included in the analysis. For patients treated with chemotherapy or radiotherapy, a total of 215 and 219 patients were available, respectively. As illustrated in Figure 2, the extent of Cav-1 staining was scored as 0 (no expression, < 25% of fibroblasts positive; 33/295), 1 (intermediate expression, 25–75% fibroblasts positive; 76/295) or 2 (strong expression, > 75% of fibroblasts positive; 186/295). In accordance with previous studies 9, 10, cases with strongly positive and intermediate Cav-1 staining were grouped for comparison with Cav-1-negative cases. Stromal Cav-1 expression was significantly associated with age (> 50 years), higher grade, size (> 10 mm), and ER positivity (Table 1). In addition, stromal Cav-1 was significantly associated with both increased breast cancer-specific overall survival (OS) and disease-free survival (DFS) (log rank test p = 0.02 and p = 0.01, respectively). Kaplan–Meier plots are shown in Figures 2d and 2e, respectively. Mean breast cancer-specific overall survival was 72 months in the Cav-1-positive group, compared with 29.5 months in the Cav-1-negative group. Uni- and multi-variate Cox regression analysis is shown in Table 2, with the latter showing that stromal Cav-1 was a significant predictor of OS and DFS independent of ER, grade, size, and node status.

Figure 2. Immunohistochemical detection of Cav-1 in CAFs and its effect on survival. Cav-1 staining was scored as 0 (no expression; a), 1 (intermediate expression; b) or 2 (strong expression; c). Cav-1 staining of vascular endothelial cells (arrows) acted as internal positive controls. Original magnification: × 20. Kaplan–Meier curves showed that expression of stromal Cav-1 was significantly associated with both increased disease-free (p = 0.01; d) and breast cancer-specific survival (p = 0.02; e)

Table 2. Cox uni- and multi-variate analysis of disease-free (DFS) and overall survival (OS)

	Univariate				Multivariate
	DFS		OS		DFS		OS
	HR	p	HR	p	HR	p	HR	p
HR = hazard ratio.
Grade	3.7	0.0001	3.7	0.0001	0.13	0.05	1.9	0.04
	(2.3–5.9)		(2.3–6.0)		(0.02–1.0)		(1.25–2.9)
LN	3.1	0.001	3.1	0.0001	0.52	0.08	3.59	0.07
	(1.6–6.0)		(1.6–6.0)		(0.25–1.1)		(2.2–5.6)
Size	1.04	0.001	1.0	0.0001	1.03	0.001	2.9	< 0.001
	(1.0–1.04)		(1.0–1.05)		(1.0–1.05)		(1.5–5.7)
ER	0.36	0.001	0.36	0.001	1.96	0.07	0.62	0.001
	(0.2–0.63)		(0.2–0.64)		(0.93–4.1)		(0.4–1.0)
Cav-1	0.26	0.001	0.28	0.001	2.04	0.04	2.0	0.05
	(0.1–0.53)		(0.15–0.5)		(1.0–3.9)		(1.0–3.8)

Cav-1 expression is reduced in CAFs and defines the invasive capacity of fibroblasts

Loss of Cav-1 has been described as a marker of oncogenic transformation in fibroblasts 22, 23. Intracellular protein lysates from normal fibroblasts (NFs) (n = 7, including the normal human breast fibroblast cell line HMFU-19) and CAFs (n = 3) obtained from primary samples and cell lines were subjected to western blot analysis. The results are displayed in Figure 3a. Invasion assays were performed using a modified Boyden chamber assay to determine the invasive capacity of NFs, CAFs, and CAFs stratified by their Cav-1 status. CAFs were significantly more invasive then NFs (p = 0.01). NFs induced a mean invasion of 35% (range 13–49%, n = 10). In CAFs, this was 62% (range 17–88%, n = 8). Cav-1-negative CAFs had increased invasive capacity over those which were Cav-1-positive, although this did not reach statistical significance (p = 0.09; n = 3 in each group). Data are shown in Figure 3b.

Figure 3. Invasive capacity of fibroblasts is related to Cav-1 status. By western blot, reduced Cav-1 expression was observed in CAFs compared with NFs (a), including the normal human breast fibroblast cell line HMFU-19 (asterisk). CAFs had greater invasive capacity than NFs and when stratified for Cav-1 status, there was a trend to greater invasion in CAFs that were Cav-1-negative (b). Error bars represent the mean ± SE of three biological replicates

Cav-1 siRNA knockdown significantly increases breast cancer cell invasion

siRNA-mediated knockdown of Cav-1 protein expression was performed on HMFU-19 fibroblasts. These were chosen in preference to primary fibroblasts as they are an immortalized cell line and not subject to the donor variability often observed with primary cultures. Western blot analysis confirmed that Cav-1 protein knockdown was sustained up to 72 h (Figure 4a). Invasion assays were then performed with T47D and MDA-MB-468 cell lines. HMFU-19 fibroblasts in which Cav-1 had been silenced significantly promoted increased invasion of T47D and MDA-MB-468 breast cancer cells (Figure 4b; p = 0.01 and p = 0.006, respectively).

Figure 4. Cav-1 silencing in HMFU-19 fibroblasts promotes increased invasion of breast cancer cells. siRNA silencing of Cav-1 in HMFU-19 fibroblasts resulted in knockdown of the protein at 72 h, compared with scrambled (Scr) controls in two separate repeat experiments (a). HMFU-19 cells in which Cav-1 had been silenced significantly promoted increased invasion of T47D and MDA-MB-468 breast cancer cells (b). Error bars represent the mean ± SE of three biological replicates

Effect of siRNA silencing of Cav-1 on 3D cell culture

MDA-MB-468 and T47D breast cancer cells were cultured with MYO1089 myoepithelial cells in collagen gels and exposed to CM collected from Cav-1 siRNA-treated or scrambled control HMFU19 fibroblasts for 7 days. In the control MDA-MB-468 cultures, myoepithelial cells organized themselves around the breast cancer cells, forming dual-cell co-units of multiple cells (Figure 5a). These co-units appeared round and tightly cohesive, with approximately 5–10 cells per unit. This phenotype has been observed previously in co-cultures with normal fibroblasts 22. In comparison, cultures treated with CM from HMFU19 fibroblasts in which Cav-1 had been silenced appeared phenotypically as single cells with occasional small groups of widely spaced cells (Figure 5b). These small groups did not form typical round colonies; rather, they appeared to be loosely cohesive strings of 2–3 cells. A significant decrease in the number of co-units of more than five cells was observed in cultures treated with CM from Cav-1 siRNA-treated HMFU19 fibroblasts compared with scrambled controls (p = 0.03; Figure 5c). High magnification co-units treated with CM from scrambled or Cav-1 siRNA-treated fibroblasts are shown in Figures 5d and 5g, respectively. In addition, E-cadherin expression was down-regulated in Cav-1 siRNA-treated cultures compared with control, as demonstrated by immunohistochemistry (Figures 5e and 5h) and immunofluorescence (Figures 5f and 5i). Similar phenotypes were observed with T47D cells (data not shown).

Figure 5. Effect of Cav-1 siRNA silencing on 3D cell culture of MDA-MB-468. In the low (× 200; a) and high (× 400; d) power images of the scrambled (Scr) controls, tightly cohesive dual-cell co-units were seen (d; arrow). Co-unit formation was lost in 3D cultures treated with CM from HMFU-19 fibroblasts in which Cav-1 had been silenced and appeared phenotypically as single cells with occasional loosely cohesive small groups of cells (b, g; × 200 and × 400 images, respectively). Quantification showed significantly more co-units of more than five cells in the Scr controls (c). E-cadherin expression was down-regulated in cells exposed to CM from HMFU-19 fibroblasts in which Cav-1 had been silenced compared with control, both by IHC (e, h) and by IF (f, i)

Discussion

There is increasing recognition of the role of the tumour microenvironment in influencing tumour cell behaviour and progression. It is probable that CAFs play a multifaceted role in this process, aiding tumour cell invasion, proliferation, and survival. As yet, however, the complex interactions between the tumour epithelium and stromal cells are not clearly understood and there is a need to further dissect these important pathways which aid cancer progression.

We examined Cav-1 expression in CAFs and demonstrated that loss of Cav-1 in these cells is a marker of reduced breast cancer-specific survival and progression-free survival. Moreover, this was independent of ER, size, grade, and node status in a multivariate analysis. Our results are in agreement with previous studies 9–13; however, these studies correlated Cav-1 expression with overall survival rather than breast cancer-specific survival. Our results are the first to conclusively show a significant relationship between breast cancer-specific survival and loss of stromal Cav-1 expression. This strengthens the notion that loss of Cav-1 in breast cancer stroma aids disease progression and increases the risk of breast cancer-specific death, rather than existing as a confounding factor associated with other concurrent disease or increased age.

The significance of stromal markers in breast cancer has been identified in several other studies. It has been recently shown that increased expression of platelet-derived growth factor (PDGF) β-receptor in breast cancer stroma correlates with significant reduction in recurrence-free and breast cancer-specific survival 23. Additionally, stromal CD10 expression is associated with poor prognosis and oestrogen receptor negativity 24. As yet, no stromal marker is routinely used in the conventional prognostication of breast cancer; however, there is mounting evidence that stromal Cav-1 is a strong prognostic indicator and potential target for therapy.

Previously published studies describing Cav-1 stromal expression have remarked on the relative difficulty in accurately scoring stromal Cav-1 expression, with Sloan et al noting that only 60% of patients had unambiguous staining 10. In order to contend with this issue, under the supervision of a specialist breast histopathologist, all Cav-1 immunohistochemical analysis undertaken in this study was carried out with comparative α-SMA staining on consecutive serial sections to aid in the identification of CAFs, allowing 82% of samples to be analysed with confidence. Since TMA cores are designed to capture principally tumour epithelial cells, it could be argued that these may not be the most appropriate for a study of this nature. However, before embarking on this study, we evaluated 140 full tissue sections against TMA cores. This showed similar survival trends between TMAs and full sections where Cav-1 staining was present (unpublished observations); hence TMAs can be regarded as representative.

To determine the possible functional role of loss of Cav-1 in fibroblasts, we used a co-culture Matrigel invasion assay to determine the effect on breast cancer cell behaviour. To eliminate the recognized biological variability inherent with working with primary cultures, we chose to manipulate Cav-1 expression in the normal mammary fibroblast cell line HMFU-19 to allow experimental consistency and reproducibility. Down-regulation of Cav-1 in CAFs was sufficient to significantly increase the invasiveness of breast cancer cells to more than double that of controls (ANOVA test, p = 0.003). This increased invasion was observed in two breast cancer cell lines of differing aggressiveness. If this effect is a true representation of the effect of Cav-1 loss observed in vivo, it could be responsible for the increased local and distant metastasis and thus reduced survival rates seen in patients with Cav-1-deficient stroma. The invasion-promoting function of Cav-1-deficient fibroblasts demonstrated herein is in accordance with a previous study in which Cav-1^−/− null mammary stromal fibroblasts were shown to share characteristics with breast CAFs 16.

It is widely accepted that fibroblasts affect tumour progression by paracrine release of signalling molecules 25. This is supported by the results of this study; since there was no direct contact between the Cav-1 siRNA-treated fibroblasts and the breast cancer cells, it is probable that secreted factor(s) are responsible for the stimulation of tumour cell invasion. Cav-1 and caveolae have been implicated in a plethora of signal transduction pathways 26, and a recent study identified tumours with loss of stromal Cav-1 expression to be associated with 238 up-regulated and 232 down-regulated gene transcripts, compared with those exhibiting high Cav-1 expression 27. Identification of the molecule (s) responsible was beyond the scope of this study but potential candidates could include vascular endothelial growth factor, platelet-derived growth factor or IL-6, as significant increases in all of these have been detected in CM derived from primary murine fibroblasts isolated from Cav-1 null mice 16.

The effects of Cav-1 knockdown in fibroblasts observed in simple 2D culture assays were replicated using a more physiologically relevant multi-cellular 3D model of breast cancer. MDA-MB-468 breast cancer cells were cultured with myoepithelial cells in collagen gel and exposed to CM from Cav-1 siRNA-treated or scrambled control for 7 days. A previous study has shown excellent reproducibility of CM and fibroblast co-culture with the MDA-MB-468 breast cancer cell line 21. The control 3D cultures organized themselves as dual-cell co-units of multiple cells, appearing highly cohesive and expressing E-cadherin. This phenotype is typical of cells cultured with normal fibroblasts which maintain cell–cell contacts and express E-cadherin 22. In comparison, the 3D cultures treated with CM from Cav-1 siRNA-treated fibroblasts appeared phenotypically as single cells with occasional loosely cohesive small groups of cells and with reduced expression of E-cadherin. This phenotype is akin to 3D models lacking dual cell co-units, a noted feature of 3D models cultured with CAFs 22, and loss of E-cadherin expression is suggestive of reduced cell–cell contacts. A previous study using cells of murine origin also demonstrated the ability of CM from Cav-1 null mammary fibroblasts to promote epithelial–mesenchymal transition (EMT) in wild-type epithelial cells 16. This model further strengthens the evidence supporting loss of Cav-1 in fibroblasts as sufficient to produce a CAF phenotype. This study also confirms that Cav-1 down-regulation at a protein level is a common phenotype of breast CAFs. This observation is consistent with the findings of Mercier et al, who demonstrated a significant reduction in eight out of 11 matched samples of normal and CAFs 28.

Our clinical and experimental data demonstrate that Cav-1 loss in CAFs increases the aggressiveness of breast cancer cells. Absence of stromal Cav-1 has also been shown to be predictive of poor outcome in relation to specific breast cancer types. In triple-negative breast cancer, stromal Cav-1 status was shown to be the most important prognostic factor for overall survival 29. In this cohort, more than 75% of patients with high levels of stromal Cav-1 remained alive at 5 years, compared with only 9.4% of those patients in the Cav-1-negative group 29. Similarly, absence of stromal Cav-1 has been correlated with early disease progression in patients with pre-invasive ductal carcinoma in situ (DCIS) 30. Of 56 patients with DCIS, almost 90% (7/8) with reduced stromal Cav-1 developed invasive breast cancer. In sharp contrast, 35 of 36 patients (97%) with Cav-1-positive stroma did not progress to invasive disease in the 12 years of follow-up. This strengthens the conclusion of our study that Cav-1-deficient CAFs are capable of significantly increasing the invasion of breast cancer cells. Taken together, these data could suggest that Cav-1 may be an initiating factor in the transition of DCIS to invasive cancer and offers the potential of using stromal Cav-1 status as a biomarker for guiding the treatment of DCIS 30, although future studies are required to prove this. Cav-1 status could also be used to distinguish patients at high and low risk of recurrence and breast cancer-related death. This information could potentially be used to inform treatment schedules; for example, chemotherapy for those with Cav-1-deficient stroma to prevent the increased risk of recurrence.

Acknowledgements

This study was supported by the Wolfson Foundation (SAS), the Wellcome Trust (DLH, SAS, VS), and the Breast Cancer Campaign (AMH, DLH, VS).

Author contribution statement

VS and DLH conceived the study. SAS and DLH carried out experiments. All authors analysed data and were involved in writing the paper and had final approval of the submitted and published versions.

SUPPORTING INFORMATION ON THE INTERNET

The following supporting information may be found in the online version of this article.

Figure S1. Immunofluorescent characterization of primary fibroblasts showed that over 95% of cells exhibited staining for vimentin (A) but were uniformly negative for cytokeratin 18 (B), cytokeratin14 (C), and the endothelial-associated CD31 (D). Approximately 30% of cells were positive for α-smooth muscle actin (E). Cav-1 immunostaining in normal fibroblasts is shown in F and is membranous.