Metastatic Growth from Dormant Cells Induced by a Col-I Enriched Fibrotic Environment

Cell lines and culture

Mouse mammary cancer D2.0R and D2A1 cells (21) (kindly provided by Ann Chambers, London Cancer Center, Ontario) were cultured as previously described in 2-D or 3-D Cultrex® (BME) (Trevigen Inc, Gaithersburg, MD) (20) or in BME mixed with neutralized rat tail collagen I (final concentration 2mg/ml). Cell morphology was imaged by confocal microscopy (LSM-META 510 Zeiss MicroImaging, Inc, Thornwood, NY). ML-7 (Biomol International L.P Plymouth Meeting, PA), W13 and U0126 (Calbiochem, San Diego, CA), and PP1 (Biomol International L.P), were used to inhibit MLCK, calmodulin, ERK and Src activities, respectively. Anti-Intβ1 antibody clone 9EG7 (azide-free) was used to antagonize Intβ1 function and non-specific IgG was used as a control (BD Biosciences, San Jose, CA). Sublines of D2.0R and D2A1 cells were generated to express GFP using the pSICO lentivirus (kindly provided by Tyler Jacks, MIT)

shRNA silencing experiments

Cells were transfected with PLKO-1 plasmids expressing either short hairpin RNAs targeting Src (shSrc) (Clone ID: TRCN0000023596) or scrambled shRNA (Open Biosystems, Huntsville, AL) and cultured in the 3D system 48h later as previously described (20). Additionally, total protein extracts were isolated from the transfected cells in 2D culture at 72 h for western blot analysis. For stable knockdown of Intβ1, cells were transduced with Mission shRNA lentiviral particles targeting either mouse Intβ1 (sh-Intβ1) or with non-target sh-RNA (sh-NT) control (Sigma, St. Louis, MO) and selected with puromycin as per the manufacturer’s protocol.

Western blot

Western blot was carried out as previously described (20). The following primary antibodies were used: goat anti-mouse Intβ1 (0.2μg/ml) (R&D system, Minneapolis, MN); monoclonal antibody against Src (1:1000) (Upstate cell signaling solutions, Temecula, CA); and monoclonal antibody against β-actin (1:40,000) (Sigma, St. Louis, MO).

Animal studies

All mice were treated in accordance with the guidelines of the Animal Care and Use of Laboratory Animals (NIH publication No. 86-23, 1985) under an approved animal protocol and in accordance with the guidelines of the Canadian Council of Animal Care.

Immunofluorescence Staining

Cells were cultured in 8 well chamber glass slides fixed and blocked as previously described (20). The following primary antibodies were used: Armenian hamster monoclonal antibody to Intβ1–(FITC) (Abcam, Cambridge, MA), Rabbit polyclonal to FAK ^Y397 (Abcam, Cambridge, MA), antibody against di-phosphorylated MLC (25) (kindly provided by Dr. Sabina Jankowska, Eisai London Research Laboratories, London UK), phospho-ERK1/ERK2 (Cell Signaling Technologies, Danvers, MA) and Alexa-Fluor ® 488 Phalloidin (Molecular Probes, Eugene, Oregon). Co-staining for Src ^Y416 and FAK^Y397: Cells were blocked and stained with M.O.M kit FMK-2201 (Vector laboratories Inc, Burlingame, CA) and incubated with Src^Y416 monoclonal antibody (9A6) (Assay Designs, Ann Arbor, MI). For cell viability staining: Calcein AM staining (Trevigen Inc., Gaithersburg, MD) was utilized according to the manufacturer’s protocol

For Col-I staining, frozen lung sections were prepared as previously described (20) and incubated with goat anti-type I Collagen (1:100) (SouthernBiothec, Birmingham, Alabama).

Statistical analyses

Student’s t-test was used for proliferation assays. Statistical analyses of metastatic lesion measurements by SCOM: The pixel counts of GFP-positive metastatic lung lesions were quantified for the entire surface area of each lung. Lesions with >10 and <1000 pixels represented individual tumor cells; lesions with >1000 pixels represented multi-cellular metastatic lesions. Total tumor burden/lung was represented by the total number of pixels detected on the entire surface of each lung. Average tumor burden/lung was represented as the median size of the metastatic lesions. Single-cell lesions (<1000 pixels) and multi-cellular clusters (>1000 pixels) were quantified and the distribution of the sizes of metastatic lesions (represented by the number of pixels per lesion) per lung was calculated. Statistical analyses of differences in the distribution of tumor burden and metastatic lesion size between groups utilized unpaired Student’s t test for two-group comparisons and one-way analysis of variance (ANOVA) planned comparisons for multiple groups. The contrast comparisons tested the differences between fibrotic lungs vs. normal lungs for each cell line, and between fibrotic lungs in the different cell lines, after subtracting the control baseline value of the normal lungs for each cell line. Both unadjusted and Bonferroni corrected P-values were determined and considered significant if <0.05.

Fibrosis triggers metastatic outgrowth of dormant D2.0R cells

Mice instilled with Ad-TGFβ1 developed fibrosis by 3 weeks with extensive deposition of matrix including Col-I, whereas mice treated with Ad-empty or left untreated did not develop fibrosis (Fig. 1A and Supp-Fig. 1 and 2A). 21 days post-viral transduction, after virus clearance and absence of exogenous viral TGFβ1 expression (24, 26) D2.0R-GFP cells were injected via tail vein into all three mice groups. Lungs were removed 4 weeks later (Supp-Fig. 1) and tumor cells in the lungs were imaged by SCOM (Fig. 1B–D). Since initial experiments demonstrated that growth of D2.OR-GFP cells in the Ad-empty-instilled lungs and untreated lungs was not significantly different (Supp-Fig. 2B–D), subsequent experiments were conducted using either Ad-empty or Ad-TGFβ1-treated mice.

Mice with fibrotic lungs (Ad-TGFβ1) exhibited a significant increase in proliferative metastatic lesions with a marked increase in collagen deposition (Fig. 1A) compared to Ad-empty control mice (Fig. 1B–D). Total tumor burden per lung was 18-fold higher in mice with fibrotic lungs (6.3×10⁶ pixels) compared to mice that received null vector (3.5×10⁵ pixels) (Fig. 1B p ≤ 0.001). The average size of the metastatic lesions was 18-fold higher in mice with fibrotic lungs (2.3×10⁵ pixels) compared to mice receiving Ad-empty (1.3 ×10⁴ pixels) (Fig. 1C; p ≤ 0.01). Multicellular metastatic lesions and individual metastatic cells expressing GFP were quantified on the entire lung surface. Fluorescence density >1000 pixels represented multicellular, proliferative metastatic lesions, whereas foci <1000 pixels indicated individual, dormant metastatic cells (Fig. 1D insert). 79 ± 3.3% of the foci in fibrotic lungs were multicellular lesions, with only 21 ± 3.3% of the foci persisting as single, quiescent cells. In contrast, only 40 ± 10.8% of the lung lesions were multi-cellular in mice without fibrosis (p ≤ 0.01), while 60% ± 10.8% of the lung lesions persisted as single quiescent cells (Fig. 1D).

Although Ad-TGFβ1 vector no longer produces TGFβ1 at the time of cell injections into mice, we wished to determine whether TGFβ1 might be capable of initiating the dormant-to-proliferative switch. D2.0R cells cultured in 3-D with BME remained quiescent and did not proliferate in response to TGFβ1. Interestingly, TGFβ1 slightly increased the proliferation of D2.0R cells, which were induced to proliferate with Col-I. (Supp-Fig. 2E, F). These results strongly suggest that TGFβ1 is not an initiating factor to induce the proliferation of dormant D2.0R cells in vivo, but may augment their proliferative response after induction of the dormant-to-proliferative switch by Col-I.

Col-I triggers the transition from dormancy to proliferation through Intβ1

Given that Col-I is enriched in the fibrotic environment, we explored whether this ECM protein might play a direct role in the dormant-to-proliferative switch. D2.0R and D2A1 cells were either cultured on BME or on BME+Col-I. While D2.0R and D2A1 cells cultured in BME alone displayed a rounded morphology at 24h, both cell lines cultured in BME+Col-I acquired a flattened, spindle morphology within 24h (Supp-Fig. 4A). D2.0R cells in BME remained quiescent during the nine-day assay period, while D2.0R cells cultured in BME + Col-I proliferated within 24h (Fig. 2A; p ≤ 0.0001). Similarly, supplementation of BME with Col-I accelerated the transition from quiescence to proliferation of D2A1 cells (Fig 2A; p ≤ 0.0001).

Since Intβ1 serves as a receptor for Col-I, we examined whether the Col-I induced dormant-to proliferative switch was dependent on Intβ1 expression. sh-Intβ1 or sh-NT control was stably expressed in pooled and selected clones of D2.0R cells (dormant phenotype) and D2A1 (metastatic) cells (Fig. 2B). Both D2.0R and D2A1 cells with reduced Intβ1 expression were significantly less responsive to Col-I induced proliferation (Fig. 2C&D; p ≤ 0.0001) than parental cells or cells with the non-target shRNA vector. As demonstrated for quiescent D2.0R cells cultured on BME, quiescent D2.0R and D2A1 cells with reduced Intβ1 expression remained viable as demonstrated by calcein AM staining while maintaining a rounded morphology when cultured on BME+Col-I (Supp-Fig. 4B).

Intβ1 mediates the switch of solitary dormant D2.0R tumor cells to metastatic growth in vivo

Mice were tail vein-injected with D2.0R-GFP cells expressing shRNA for Intβ1 (clone #47-6,D2.0R-sh-Intβ1), or D2.0R-GFP expressing non-target shRNA (D2.0R-sh-NT). Lungs were removed 4 weeks post-injection and analyzed by SCOM (Fig. 3A–C). Metastatic burden in the lungs was analyzed by ANOVA in relation to both presence or absence of fibrosis and the Intβ1 expression (Fig. 3A). A significant 5-fold difference in the total tumor burden/lung in mice with fibrotic lungs receiving D2.0R sh-non target cells (sh-NT) compared to mice with non-fibrotic lungs was observed (n=8; p=0.018 unadjusted, p=0.05 with Bonferroni correction). Importantly, no difference was observed in total burden between normal and fibrotic lungs receiving D2.0R-sh-Intβ1 cells (sh-Intβ1) (n=10; p=0.89 unadjusted, p=1 with Bonferroni correction), indicating that fibrosis is associated with significant induction of metastatic lesions that could be prevented by the loss of Intβ1 expression in D2.0R cells. Viable single dormant cells were observed in these lungs, suggesting that the reduction in metastatic burden was not due to the lack of cell viability as a result of the reduction in Intβ1 expression, in keeping with the in vitro results above (Supplementary Figure 4B). Comparing the fibrosis effect only, revealed that in fibrotic lungs the total tumor burden per lung of D2.0R-sh-NT cells was 5-fold higher then the total tumor burden/lung of D2.0R cells with low expression of Intβ1 (sh-Intβ1) (Figure 3A). However, this did not reach statistical significance (p=0.086 unadjusted, p=0.26 with Bonferroni correction) due to the higher error variance when the previous two class comparisons were directly compared.

The average size of metastatic lesions in the fibrotic lungs was significantly higher for D2.0R-sh-NT lesions (3.1×10⁵ pixels) compared to D2.0R-sh-Intβ1 lesions (812 pixels) (p ≤ 0.05; Fig. 3B). Similarly, the average size of the metastatic lesions of D2.0R-sh-NT cells in the fibrotic lungs was significantly higher compared to D2.0R-sh-NT cell lesions in non-fibrotic lungs (p ≤ 0.001; Fig. 3B). Importantly, as shown in Fig. 3C, the majority of D2.0R-sh-Intβ1 cells persisted as single cells in lungs with or without induced fibrosis (82%± 12.5% and 97% ± 3.3%, respectively, p = n.s.). In contrast, mice with pre-existing lung fibrosis and D2.0R-sh-non target cells had only 7± 4.2% single cell foci and 93± 4.2% proliferative, multicellular foci compared to mice with fibrotic lungs injected with D2.0R-sh-Intβ1 cells exhibiting only 18% ± 12.5% proliferative foci (p ≤ 0.01; Fig. 3C).

Similar results were obtained using metastatic D2A1 cells with reduced Intβ1 expression (Supp-Fig. 5 A–C). Furthermore, the metastatic lesions that developed from D2A1 sh-non target-GFP cells in normal lungs exhibited significant expression of Col-I, whereas regions in the same lungs without D2A1 metastatic lesions did not exhibit significant Col-I expression (Supp-Fig. 5D) further highlighting the association between proliferative growth and Col-I expression for D2A1 cells.

Intβ1 activation by Col-I induces the phosphorylation of Src and FAK

Interactions between focal adhesion kinase (FAK) and Src are known to mediate integrin signaling (27). Within 24h, D2.0R cells on BME+Col-I proliferated with an associated increase in FAK^Y397 phosphorylation, which co-localized with Intβ1. These changes were absent in cells cultured only on BME (Supp-Figs 6A and 7A). Inhibition of Intβ1 in D2.0R cells cultured on BME+Col-I with either αIntβ1 antibody (αIntβ1Ab) or shRNA led to a marked reduction in both in pFAK^Y397 and co-localization with Intβ1 compared to controls (Supp-Figs 6B–C & 7A–B). BME+Col-I significantly enhanced phosphorylation and co-localization of pSrc^Y416 and pFAK^Y397 in D2.0R cells (Fig. 4A).

Inhibition of Src activity in D2.0R cells cultured on BME+Col-I using the inhibitor PP1 or Src shRNA (Supp-Fig. 8) prevented phosphorylation of FAK^Y397, co-localization with Intβ1 (Fig. 4B) and proliferation (p ≤ 0.0001) (Fig 4D). Inhibition of Intβ1 with αInt β1AB or shRNA significantly reduced p FAK^Y397 and p Src^Y416 and their co-localization compared to controls (Fig. 4C and Supp-Fig 7A–B).

FAK and Src activate the extracellular signal regulated kinase (ERK) leading to proliferation (28). Quiescent D2.0R cells cultured on BME had minimal activated ERK (pERK) (Figure 5A and Supp-Fig. 7A). The increased pERK, actin stress fiber formation and proliferation in D2.0R cells cultured on BME+Col-I (Fig. 5A) were inhibited with αIntβ1AB (Fig. 5B), Intβ1 shRNA (Fig. 5C and Supp-Fig 7A–B), or through inhibition of Src activity with PP1 or Src shRNA (Fig. 5D). Scrambled shRNA had no effect (data not shown).

Col-I induced Intβ1 signaling activates ERK leading to phosphorylation of MLC by MLCK and actin stress fiber formation

The dormant-to-proliferative switch of D2.0R cells requires actin stress fiber formation through phosphorylation of MLC (pMLC) by MLCK (20) which can be regulated by ERK (29). D2.0R cells in BME+Col-I demonstrated a marked increase in pMLC and actin stress fibers (Fig. 6A and Supp-Fig. 7A) compared to BME alone. Inhibition of MLCK directly by ML-7 (30) or through inhibition of upstream signaling molecules that regulate MLCK activity (ERK using U-0126 or Ca2+ calmodulin using W13) reduced pMLC, inhibited actin stress fiber formation and proliferation (p ≤ 0.0001) (Fig. 6A–B). Inhibiting Intβ1 function also reduced pMLC and actin stress fiber formation compared to controls (Fig. 6C and Supp-Fig 7A). Inhibiting Src activity with PP1 or shRNA also reduced pMLC, actin stress fiber formation (Fig. 6D) and proliferation.