Transforming growth factor beta receptor type III is a tumor promoter in mesenchymal-stem like triple negative breast cancer

SUM159 cells (Asterand, Detroit, MI, USA) were maintained in (Dulbecco’s) Modified Eagle’s Medium: Nutrient Mixture F12 ((D)MEM-F12, GIBCO, Grand Island, NY, USA) supplemented with 5% fetal bovine serum (FBS) (GIBCO) and 0.5 μg/ml hydrocortisone. MDA-MB-231 and MDA-MB-157 (ATTC, Manassas, VA, USA) were maintained in (D)MEM (GIBCO) supplemented with 10% FBS. Stable TβRIII-KD SUM159 cell lines were generated by lentiviral infection with virus carrying four independent short hairpin RNA (shRNA) clones (sequence-verified shRNA, pLKO.1-puro), (Sigma-Aldrich, St. Louis, MO, USA), Mission shRNA library #SHCLNG-NM_003243: clone#TRCN0000033433 (TβRIII-KD), clone#TRCN0000359000 (TβRIII-KD2), clone#TRCN0000359001 (TβRIII-KD3), and clone# TRCN0000359081 (TβRIII-KD4)) followed by puromycin selection (Invitrogen-Life Technology, Inc, Carlsbad, CA, USA). MDA-MB-231 and MDA-MB-157 were stably infected with clone# TRCN0000033433. Integrin-α2 was stably knocked down in TβRIII-KD MSL cells using lentiviral particles carrying shRNA to integrin-α2 (α2-KD) (Sigma-Aldrich, Mission shRNA validated library, #SHCLNG-NM_002203, clone#TRCN0000308081).

The wells in 48-well plates were coated with 50 μl of growth factor reduced BD Matrigel (BD Biosciences #356231, San Jose, CA, USA) and allowed to polymerize at 37°C for 15 minutes. Then, 5 x 10⁵ cells were resuspended in 200 μl of growth factor reduced BD Matrigel and plated onto the matigel-coated wells. Plates were incubated for 30 minutes after which 1 ml of media was added to the top of the matrigel. Media was replenished every 48 hours. Images were taken at day six. Quantification of the images was performed using Fiji Software.

One milllion cells embeded in collagen were implanted into the number four gland of six- to eight-week-old female athymic nude- Foxn1^nu/nu mice (purchased from Harlan Sprague- Dawley, Inc., Indianapolis, IN, USA). Mice were monitored weekly for tumor growth. Tumor measurements were performed once a week for five weeks after palpable tumors formed. Tumor volume was measured at the indicated times with calipers, and tumor volumes were calculated as width² x length/2. All mouse experiments were approved by the Vanderbilt University Institutional Animal Care and Use Committee (IACUC).

Cells were seeded at a density of 2 X 10⁴ cells/well in 12-well tissue culture plates. The following day, the cells were transiently transfected using Transfectin lipid reagent following the manufacturer’s protocol (Bio-Rad #170-3351, Hercules, CA, USA). Cells were transfected with 1.5 μg 3TP-Lux [22] or CAGA(9)-Luc [23]. pRL-CMV-renilla (Promega #E2261, Madison, WI, USA) was co-transfected and used as an internal control to correct for transfection efficiency. Eighteen hours after transfection, cells were treated with 1 ng/ml TGF-β1 or TGFβ-2 (R&D Systems, #102-B1 and #102-B2, respectively). Twenty-four hours after TGF-β treatment, cells were harvested and assayed for promoter specific luciferase activity using a Dual-Luciferase Reporter Assay System (Promega #E1910) according to the manufacturer’s protocol. Luciferase activity was measured using a BD/Pharmigen Monolight 3010 luminometer.

RNA was isolated and purified using an RNeasy Mini Kit and an RNase-Free DNase Set (Qiagen, Valencia, CA, USA). A total of 750 μg of RNA was used to synthesize cDNA using Superscript III reverse transcriptase as described by the manufacturer (Invitrogen). Bio-Rad iCycler and CFX96 machines were used for qPCR employing Power SYBR Green (Applied Biosystems, Carlsbad, CA, USA) or SsoAdvanced SYBR Green Supermix (Bio-Rad), respectively. C _t values were normalized to GAPDH for statistical analyses. Primer sequences are available in Additional file 1.

Standard protein preparation and electrophoresis procedures were used as described [4]. Western membranes were blocked in 5% milk and incubated with primary antibody overnight. The antibody list with concentrations and the catalog numbers are available in Additional file 1.

Cells were detached using Accutase (Life Technologies), pelleted, washed and counted. One million cells were incubated with TβRIII antibody (Cell Signaling, #5544, Danvers, MA, USA) for 30 minutes, washed, and then incubated at 4°C with Alexa Fluor 488 conjugated secondary antibody (Life Techologies, #A11034) for 30 minutes. One million cells were labeled with fluorescence-conjugated integrin-α2 antibody (BioLegend, #314308, San Diego, CA, USA) for 30 minutes at 4°C. Cells were washed three times then analyzed on a FACSCalibur flow cytometer (Becton Dickinson, San Jose, CA, USA) using CellQuest Pro software. Data were analyzed with FlowJo software (Tree Star).

All data were analyzed using the unpaired two-tailed Student’s t test (GraphPad Prism 5). Error bars show mean ± SEM. A two-sided P value less than 0.05 was considered significantly different.

Using a gene expression data set generated from 587 TNBC tumors, we examined the relative mRNA levels of TGF-β receptors and ligands across subtypes of TNBC. We observed elevated expression of TGFBR3 in basal-like1 (BL1), mesenchymal (M) and MSL tumors (Figure 1A). The highest relative level of TGFBR3 expression was in the MSL subtype (Figure 1B). Average probe intensities for the TGF-β receptors I and II as well as TGF β ligands 1 and 3 were also elevated in the MSL subtype in comparison to the rest of the TNBC subtypes (Additional file 2: Figure S1). Similarly, analysis of TGFBR3 gene expression across a panel of TNBC cell lines, representative of the various subtypes, demonstrates that the M and MSL subtypes have relatively higher levels of TGFBR3 mRNA (Figure 1C-D). These findings were validated by qPCR (Figure 1E) and immunoblot analyses for TβRIII protein levels (Figure 1F). Although the TNBC M subtype cell lines also showed increased levels of TβRIII expression, we focused our studies of this receptor on the MSL subtype as their expression is more consistent with human datasets (Figure 1A-B).

In order to determine the significance of the TβRIII expression in MSL TNBC cell behavior, we knocked down TβRIII in MSL cells and performed orthotopic xenograft tumor studies. We used a panel of four shRNA expression vectors to optimize TβRIII knockdown, as validated by immunoblot and flow cytometry analyses (Figure 2A-C). We utilized immunocompromised nude mice to establish orthotopic xenograft tumors from cell lines representing the MSL subtype of TNBC with and without TβRIII knockdown. Initially we tested SUM159 cells with two shRNA expression vectors (TβRIII-KD and TβRIII-KD4) to eliminate off target effects of the shRNA (Additional file 2: Figure S2). After establishing that both expression vectors resulted in a similar phenotype, we used a single shRNA (TβRIII-KD) in all subsequent experiments across three MSL cell lines. Knockdown of TβRIII in the SUM159 and MDA-MB-231 MSL cell lines significantly decreased xenograft tumor growth (Figure 2D-E). MDA-MB-157 showed inconsistent results (Additional file 2: Figure S3A) and after further investigation we discovered that the TβRIII-KD tumors expressed TβRIII (Additional file 2: Figure S3B). Thus, either there was a selection against the knockdown in vivo and, therefore, the tumor cells expressed TβRIII, or there was a small subpopulation of MDA-MB-157 cells at the start of the experiment that retained expression and seeded the tumor growth. Regardless, both explanations provide further evidence for the tumor-promoting effect of TβRIII.

Since TβRIII-KD markedly decreased the tumorigenic potential of mesenchymal TNBC cells, we further explored whether this was due to a proliferation defect. TβRIII can bind to all TGF-β ligands but with highest affinity for TGF-β2 [29, 30]; therefore, cells were treated with TGF-β2 in addition to TGF-β1. Both controls and TβRIII-KD MSL cell lines responded similarly to the ligands (Figure 3A-B). TβRIII-KD did not alter the proliferation rates of MSL cell lines (SUM159, MDA-MB-231 and MDA-MB-157) by live cell counts (Figure 3A) or ³H-thymidine incorporation assay (Figure 3B). Consistent with an intact TGF-β signaling pathway [22, 23] we have observed an increase in phospho-SMAD2 following ligand treatment (Additional file 2: Figure S4). In order to examine cell viability and determine whether knockdown of TβRIII influenced apoptosis, we analyzed cleaved-caspase 3 and cleaved-PARP and we did not detect any difference between control and TβRIII-KD MSL cells (Figure 3C).

Using a validated method (please see methods section for details) for measurement of cell migration [21], we found that TβRIII-KD significantly decreased the migration of SUM159, MDA-MB-231 and MDA-MB-157 cells (Figure 4A-C). Treatment with TGF-β ligands had no effect on migration. In order to determine the invasive properties of MSL lines we analyzed their ability to migrate through a barrier using an invasion transwell assay. TβRIII-KD impaired the ability of the MSL cell lines to invade through matrigel pre-coated transwells and the addition of TGF-β ligands had little effect on invasion in either controls or knockdowns (Figure 4D-F). Next, we examined the effect of TβRIII-KD on the ability of MSL cells to form colonies in three-dimensional matrigel culture. After five days in culture, SUM159 cells with TβRIII-KD had smooth edges around cell spheres while control cells had multiple protrusions invading into the surrounding matrix (Figure 4G). These results were quantified by calculations of the perimeter, which show a significant difference between controls and TβRIII-KD (Figure 4H). Overall, these data indicate that TβRIII modulates migration and invasion, independent of TGF-β stimulation, in MSL cells. To further investigate TGF-β pathway signaling [31] in the MSL lines we used standard CAGA-luc (Additional file 2: Figure S5A) and 3TP-lux (Additional file 2: Figure S5C) reporter assays for measurement of TGF-β activity [22, 23]. Assays were performed either in the presence of TGF-β1 or TGF-β2 ligands and compared to untreated controls [30, 32]. In addition, we performed qPCR analysis for SMAD7 (Additional file 2: Figure S5B) and PAI-1 (Additional file 2: Figure S5D) gene expression as readout for downstream targets for canonical and non-canonical TGF-β activity, respectively [33, 34]. The results of both assays indicate that knockdown of TβRIII does not modulate either arm of the TGF-β signaling pathway. Thus, MSL lines with TβRIII knockdown have resulting phenotypic changes without concomitant changes in the TGF-β signaling pathways measured. Considering these results and knowing that TβRIII can also bind to bone morphogenetic proteins (BMPs) [35], we treated the engineered MSL cell lines with BMP4. We did not observe significant changes in Smad1/5/8 phosphorylation in TβRIII-KD versus control MSL cells (data not shown). The results suggest that TβRIII modulates the tumorigenic potential of MSL TNBC cells through other signaling pathways.

To determine which genes and/or signaling pathways are significantly altered in MSL cells after TβRIII knockdown, we performed gene expression microarray analyses on SUM159 cells grown in three-dimensional cultures. The integrin signaling pathway, along with other cell adhesion pathways, were among the most significant pathways differentially expressed in TβRIII-KD MSL cells relative to control cultures (Additional file 3: Table S1). Analysis of individual genes of the integrin pathway revealed that ITGA2 was a top gene that was significantly increased upon TβRIII knockdown (Additional file 3: Table S2). In vitro qRT-PCR analysis indicates a statistically significant (above two-fold) upregulation of integrin-α2 in the TβRIII-KD MSL cells (Figure 5A-C). The upregulation of integrin-α2 was further validated by flow analysis across all MSL (Figure 5D-F).

Using a clinically relevant, spontaneous mouse model of breast cancer progression and metastasis, Ramirez et al. demonstrated that integrin-α2β1 acts as a tumor suppressor and α2-null cells were more motile and invasive [36]. The in vivo and in vitro findings were further correlated with analysis of microarray gene expression datasets of human breast and prostate cancers, which showed a correlation between decreased expression of ITGA2 and poor prognosis. Considering this role of integrin-α2 in breast cancer, we hypothesized that the decrease in migration and invasion upon TβRIII-KD in MSL cells could be rescued by concomitant knockdown of integrin-α2. To test our hypothesis, we stably knocked down integrin-α2 (α2-KD) in the MSL TβRIII-KD cells and performed migration and invasion assays (Figure 6A-B and Additional file 2: Figure S6A-B). Knockdown of integrin-α2 was sufficient to reverse the migration (Figure 6C and Additional file 2: Figure S6C) and invasion (Figure 6E and Additional file 2: Figure S6D) phenotype of MSL cells with TβRIII-KD to those of control cells. In addition, using an integrin-α2 neutralizing antibody we rescued the migratory phenotype (Figure 6D) in a manner similar to that observed after α2-KD in TβRIII-KD cells. Knelson and colleagues showed that knockdown of TβRIII leads to diminished fibroblast growth factor 2 (FGF2)-mediated ERK phosphorylation [37]. Consistent with this previous study, after knockdown of TβRIII in the MSL cells, the phospho-ERK levels decreased and were restored in the cells after simultaneous integrin-α2 and TβRIII knockdown (Figure 6F and Additional file 2: Figure S6E).

To further investigate the association between TβRIII (TGFBR3) and integrin-α2 (ITGA2) in TNBC, we used the TNBC patient dataset described in Figure 1A [2] to analyze the relationship between TGFBR3 and ITGA2 gene expression. Our results indicate an inverse correlation between ITGA2 and TGFBR3 expression across TNBC subtypes. In particular, we see the strongest inverse correlation in TNBC subtypes with either high TGFBR3 expression (MSL; P = 5.274e-06); or low TGFBR3 expression (basal-like 2; with P = 5.16e-07 and Luminal AR (LAR); with P = 1.759e-07) (Figure 7A-B). The clinical association between ITGA2 and TGFBR3 expression is relevant as it further links the impact of the interplay between the TGF-β and integrins pathways in TNBC.