Less is more: low expression of MT1-MMP is optimal to promote migration and tumourigenesis of breast cancer cells

MCF-7, MDA-MB 231 and HS578t human breast cancer cell lines were obtained from the American Type Culture Collection (Manassas, VA). Cells were maintained in DMEM/F-12 media (Thermo Fisher) supplemented with 10 % FBS, 100 IU/ml penicillin, 100 μg/ml streptomycin, and incubated at 37 °C and 5 % CO₂.

Human MT1-MMP (sc116990), TIMP-2 (sc118083) and MMP-2 (sc321560) cDNA clones were purchased from Origene and subcloned into the vector pcDNA 3.3 (Thermo Fisher). The generation of the ALA + TIMP-2 cDNA construct in pcDNA 3.3 is described in Walsh et al. [38]. The following reagents were used: Recombinant human TIMP-2 and 4-aminophenylmercuric acetate (APMA) (Sigma-Aldrich), BB-94 (Batimastat), U-0126, and AKT inhibitor IV (Santa Cruz), and Furin inhibitor II (Millipore).

For immunoblot analysis, the following primary antibodies were used: MT1-MMP (1:1000, AB6004, Millipore); MT1-MMP (1:1000, AB51074, Abcam); Phospho-ERK1/2 (1:2000, D13.14.4E), ERK1/2 (1:2000, 137 F5) (Cell Signaling Technology); TIMP-2 (1:1000, 3A4), β-Actin (1:1000, C4), and phospho-histone-3 (PH3) (1:5000, C1513) (Santa Cruz). Goat anti-mouse IgG (H + L) (Bio-Rad) and goat anti-rabbit IgG (H + L) (Thermo Fisher) HRP conjugates were used as secondary antibodies (1:10000). For immunofluorescence analysis we used MT1-MMP antibody AB6004 (1:200), and anti-rabbit-IgG-Alexa488 or Alexa594 (Thermo Fisher) as secondary antibodies (1:400).

MCF-7 and MDA-MB 231 cells were seeded at a density of 5×10⁵ cells/ml and incubated for 24 h. Following incubation, cells were transfected with Lipofectamine 2000 (Thermo Fisher) according to the manufacturer’s instructions. For transient transfection experiments, cells were incubated for 24 h after transfection and then utilized for experiments.

Stable cell lines were generated by transfection of cells with the respective cDNAs in the vector pcDNA 3.3, which contains a neomycin mammalian selection marker. Following transfection, cells were split 1:1000 and incubated in media containing 1 mg/ml G-418 (VWR). Individual colonies were selected after four weeks of incubation in selection media and expanded to assay for the levels of MT1-MMP by qPCR and immunoblotting. Stable cells lines expressing an shRNA sequence targeting MT1-MMP in the vector pRS (TR311445, Origene) were generated in the same manner expect using puromycin (2 μg/ml) as the selection antibiotic.

For zsGreen infection, cells were seeded at ~ 40 % density in a 6-well cell culture dish in 3 ml of media with a final concentration of 8 μg/ml polybrene and infected with 250 μL of virus. For virus production, the pLVX-ZsGreen1-N1 lentiviral plasmid was used. Twenty-four hours post-infection, the media containing virus was removed and replaced with puromycin selection media (2 μg/ml) for three days of incubation to select for infected cells.

Conditioned media (CM) containing high levels of MMP-2, TIMP-2, and ALA + TIMP-2 protein was created by transfecting MCF-7 cells with cDNA constructs coding for the respective proteins. Following a 24-h incubation post-transfection, transfected cells were washed with phosphate buffered saline (PBS) and incubated in DMEM/F12 media without FBS for 24 h. The serum-free CM was then collected, aliquoted and stored for later use. Conditioned media from mock-transfected cells was used as a control.

RNA was collected from cells using the RNeasy Kit (Qiagen) and cDNA was synthesized from 1 μg of RNA using qScript cDNA supermix (Quanta). MT1-MMP mRNA levels were assayed by qPCR using PerfeCta SYBR Green Supermix (Quanta) and a CXF connect real time system with CFX manager software (Bio-Rad). mRNA levels were quantified by the ΔΔCT method and are displayed as fold change relative to parental MCF-7 cells. The level of GAPDH mRNA was used as the internal control. Primers are as follows: MT1-MMP; F: gcagaagttttacggcttgca, R: tcgaacattggccttgatctc, GAPDH; F: acccactcctccacctttga, R: ctgttgctgtagccaaattcgt [33].

Post-incubation cells were washed 3× with PBS (pH = 7.2) and disrupted using lysis buffer (150 mM NaCl, 1 % NP-40, 0.5 % NaDC, 0.1 % SDS, 50 mM Tris pH 8.0) supplemented with protease/phosphate inhibitor (Thermo Scientific). Cell lysates were homogenized by sonication and 15 μg aliquots were analyzed by immunoblotting with MT1-MMP, TIMP-2, pH-3, β-actin, pERK1/2 or ERK1/2 primary antibodies, followed by incubation with the appropriate secondary HRP-conjugated antibody and detection using SuperSignal West Pico chemiluminescent substrate (Thermo Fisher). Protein lysate from human 21 T cell lines were acquired as previously described [39].

Gelatin zymography and reverse zymography was done using samples of serum-free media (15 μl) as described previously [40]. To assess the proMMP-2 activation ability of MCF-7 cells, which endogenously express very low levels of MMP-2 [34], proMMP-2 CM was added to cells at a dilution of 30 μl proMMP-2 CM/mL SF media. For reverse zymography, serum-free media conditioned for 24 h by HS578t cells was used as the source of active MMP-2 within the gel, as these cells naturally express high levels of MMP-2 [33].

Samples for fluorescent gelatin degradation and 3D culture experiments were fixed and prepared for immunofluorescence according to their respective protocols (see below). MT1-MMP primary antibody (AB6004) was detected using anti-rabbit IgG Alexa594 or Alexa488 for Oregon green-488 gelatin degradation or 3D culture experiments respectively. F-actin was stained with Alexa633 phalloidin (1:100, Thermo Fisher) and nuclei with 4′,6-diamidino-2-phenylindole (1 μg/ml, BioShop Canada). Samples were imaged using a Nikon A1R+ confocal microscope (1.2 au) with a 20× dry or 60× oil-immersion lens and presented using NIS Elements software.

The migratory potential of cells was measured using 24-well 8 μm pour transwell inserts (Corning Costar). Cells (2×10⁴) were seeded on the upper chamber of the transwell in serum-free media and allowed to migrate towards the bottom chamber which was placed in DMEM/F-12 media supplemented with 10 % FBS. Migration assays were done with uncoated transwell inserts, whereas invasion assays were done with inserts coated with 20 % Matrigel. Cells that migrated to the lower chamber of the transwell insert were quantified as described previously [41]. Migrated cells were normalized to MCF-7/MDA-MB 231 parental cells in control conditions and are presented as a mean percentage ± SEM.

Cells were seeded at a density of 1×10⁶ cells in a 35 mm cell culture dish and allowed to form a monolayer for 24 h. Following incubation, media was removed and a scratch was made down the middle of the cell culture dish with a 100 μl pipette tip. Cells were washed 3× with PBS (pH 7.2) to remove cell debris and then incubated with fresh media. After 2 h, 10 images were captured down the length of the scratch that represent the ‘initial’ size of the scratch for that sample. Every 24 h for 3 days, the same area of the scratch was imaged to examine how cells migrated to close the scratch. Scratch closure was quantified using ImageJ 2.0.0 software by measuring the width of the scratch each day and normalizing it to the initial size of the scratch. Scratch closure is presented as a mean percentage of the initial scratch size ± SEM.

Cells were seeded in triplicate (5000 cells/well) onto a 96-well cell culture dish in media supplemented with 10 % FBS or serum-free media. Immediately after seeding, 20 ul of Celltiter96® AQueous One Solution (Promega) was added in triplicate to each sample to obtain an initial reading and ensure that the different cell lines were seeded equally. The OD at 490 nm was measured using a Bio-Rad model 3550-UV microplate reader after incubating the cells with the substrate for 2 h. The substrate was added at the indicated day intervals and measured in the same way as the initial measurement to create a growth curve over that span, or to compare the effect of chemical inhibitors on cell viability during prolonged serum-free incubation.

Coverslips were coated with either Oregon Green-488 gelatin (for IF analysis) or gelatin labeled using an Alexa594 protein labeling kit (for live imaging analysis) (Thermo Fisher) and used to analyze ECM degradation as per Martin et al. [42]. Immunofluorescence conditions are as described above.

Cells (2.5×10⁴) were embedded in Matrigel (Corning Costar) and processed for immunofluorescence as per Cvetkovic et al. [43]. To quantify colony morphologies observed in 3D culture, five random 50 μm Z-stacks (2 μm step size) were acquired using DIC microscopy at 10× magnification every day for 5 days post-embedding. For each individual Z-stack, invasive features (disseminations and protrusions, see text) were blindly counted and normalized to the number of circular colonies per field of view. Immunofluorescence procedure was done as described above using 3 % BSA/PBS as the blocking buffer. Single cells (marked by DAPI), F-actin disseminations, F-actin protrusions and zsGreen protrusions were blindly counted from 20× 3D volumes and normalized to the number of circular colonies per field of view.

Cells were embedded in Matrigel or seeded on Alexa594 gelatin coverslips as described above and placed in a live imaging chamber mounted on the stage of a Leica DM16000 B fluorescent microscope. To analyze the relationship between ECM degradation and migration, cells stably expressing zsGreen were incubated on Alexa594 gelatin coverslips and imaged at the same stage position at 10× magnification every 10 min for 20 h. These images were then compiled into timelapse movies using ImageJ. The zsGreen channel timelapse movies were used to quantify the migration of individual cells in an automated manner using the ADAPT plugin for ImageJ [44]. The ADAPT plugin (v1.146) was used with default conditions and the trajectory visualization output was used to group cells according to the distance migrated. The Alexa594 gelatin channel was used to manually quantify the percentage of cells that had degraded the underlying gelatin at different time points. To visualize the dynamics of cells in 3D culture, cells were embedded in Matrigel and z-stacks (100 μm, 5 μm step size) were acquired at 20× magnification every 30 min for 72 h. Focal panels showing colony features with the greatest clarity were isolated and compiled into timelapse movies using ImageJ.

For implantation experiments, the superficial layer of the CAM of day 9 chicken embryos was removed to expose the underlying capillaries [45]. MDA-MB 231, MCF-7 or MCF-7 MT1-MMP cell lines stably expressing zsGreen were then resuspended in Matrigel and pipetted onto the exposed capillaries (5×10⁵ cells in 10 μl Matrigel/embryo). Eight days post-implantation the resulting tumour was imaged using a fluorescent stereoscope to examine vascularization of the tumour shown by the presence of non-fluorescent CAM vessels within the zsGreen tumour.

MCF-7 and MCF-7 MT1-MMP cell lines stably expressing zsGreen were used for intravenous injection into the CAM vasculature of day 14 chicken embryos and extravasation efficiency was quantified 24 h post-injection as per Kim et al. [37].

Quantitative analysis of immunoblots was done using QuantityOne software (Bio-Rad). Band intensity was obtained for the pERK and total ERK signal of each sample from three independent experiments. ERK activation is presented as a ratio between the pERK and the total ERK band intensity within each sample normalized to MCF-7 cells under control conditions.

Statistical analysis and graphing was performed using GraphPad Prism version 6.0 (GraphPad software, La Jolla, CA, USA). Data is presented as mean ± SEM. One-way ANOVA followed by Tukey’s post-hoc test was used unless otherwise indicated in the figure legend. Significant differences denoted by asterisks are shown versus MCF-7 or MDA-MB 231 parental cells in control conditions unless otherwise shown in figure. Different levels of statistical significance are denoted by a different number of asterisks and are as follows: ****p ≤ 0.0001, ***p ≤ 0.001, **p ≤ 0.01, *p ≤ 0.05. All experiments were repeated in triplicate with comparable results.

We transiently transfected MCF-7 breast cancer cells with wild-type untagged MT1-MMP cDNA and assessed proMMP-2 activation, ERK activation, migration and invasion (Fig. 1). qPCR and immunoblot analysis of MT1-MMP mRNA and protein levels, respectively, of MCF-7 cells transiently transfected with MT1-MMP cDNA showed that 24-h post-transfection these cells have very high levels of MT1-MMP mRNA (~17,000 fold relative to mock transfected cells) and produced predominately the pro- but also active isoform, along with degradation forms, of MT1-MMP protein (Fig. 1a). Gelatin zymography analysis demonstrated that MCF-7 cells transiently transfected with MT1-MMP are capable of activating proMMP-2 after 24 h of incubation with MMP-2 CM, as shown by the transition to the intermediate and active isoforms of MMP-2. Parental MCF-7 cells, which do not secrete endogenous proMMP-2, did not increase MMP-2 production after transfection with MT1-MMP, as determined by gelatin zymography (Fig. 1b, top gel). Addition of proMMP-2 conditioned media to transfected cells demonstrated that MT1-MMP overexpression activated proMMP-2 after 12 h of incubation, shown by the presence of an intermediate band, whose activation was enhanced by the addition of low levels of TIMP-2 (100 ng/ml) that resulted in the presence of more active MMP-2 isoform (Fig. 1b, bottom gel). Differing from published reports [24, 27, 28] we found no difference in ERK activation (Fig. 1c) or migration and invasion (Fig. 1d, ns > 0.05) in MCF-7 cells that produced functional MT1-MMP protein after transient transfection.

To test whether the very high and heterogeneous levels of MT1-MMP expression achieved using transient transfection was a factor behind the discrepancies between our observation and other reports, we created clonal MCF-7 cell lines that stably express significantly different levels of MT1-MMP (Fig. 2). These cell lines, MCF-7 MT1-MMP C1, C2, and C3 cells (hereby referred to as C1, C2 or C3 cells), significantly differ in their stable expression of MT1-MMP transcript from high (C1, ~2,500 fold compared to MCF-7 parental cells), medium (C2, ~1100 fold) to low (C3, ~11 fold) as determined by qPCR analysis (Fig. 2a). Immunoblot analysis of these MCF-7 MT1-MMP cell lines showed that both pro- and active- isoforms of MT1-MMP protein were detected in C1 cells, whereas the active isoform was predominantly detected in C2 cells and no MT1-MMP protein could be detected by immunoblot in C3 cells (Fig. 2b). Cell lines where active MT1-MMP was detected via immunoblot (C1 and C2 cells) were capable of activating proMMP-2 in a time-dependent manner, and as expected this was enhanced by low levels of recombinant TIMP-2 (100 ng/ml), shown by a transition to the intermediate form of MMP-2, which was completely inhibited by addition of BB94 (10 μM) (Fig. 2c). Addition of diluted (1:100) TIMP-2 CM demonstrated a greater enhancement of proMMP-2 activation by C1 and C2 cells as shown by the presence of both intermediate and active forms of MMP-2. Densitometry quantification of a representative zymography with diluted TIMP-2 CM demonstrates that C2 cells have the highest ability to activate proMMP-2, as shown by increased levels of intermediate and active forms, despite not expressing the highest level of MT1-MMP (Fig. 2c, bottom graph). Active MMP-2 was not detected in C3 cells.

To confirm that C3 cells produced MT1-MMP protein (not detectable by immunoblot) we used an imaging approach whereby cells were seeded on Oregon-green 488 gelatin coated coverslips and examined using immunofluorescence for both MT1-MMP protein presence and degradation of the underlying fluorescently labeled ECM [42]. MDA-MB 231 breast cancer cells were also used in this analysis to compare MCF-7, C1, C2 and C3 cells to a cell line demonstrated to express naturally increased levels of MT1-MMP (Fig. 3). Quantification of fields of view acquired at 20× magnification demonstrated no detectable cytoplasmic MT1-MMP immunological signal nor gelatin degradation in parental MCF-7 cells. In contrast, 84 and 60 % of C1 and C2 cells displayed ECM degradation (gelatin -), respectively, with approximately 20 % of these cells also displaying punctate cytoplasmic immunological signal for MT1-MMP protein (Fig. 3, bar graph). A low percentage of MT1-MMP C3 cells (~1.5 %) showed ECM degradation, and a higher percentage (~3 %) showed a cytoplasmic MT1-MMP signal, although these amounts were non-significantly different than parental MCF-7 cells. Interestingly, in all conditions where MT1-MMP was overexpressed, there were cells that had degraded the underlying ECM that, paradoxically, were devoid of MT1-MMP protein signal (Fig. 3, white arrows). MDA-MB 231 breast cancer cells were not capable of widespread cell-associated ECM degradation, and only ~6 % of cells displayed cytoplasmic MT1-MMP signal. Taken together with the data in Fig. 2, these analyses show that ECM degradation and proMMP-2 activation is associated with elevated MT1-MMP expression levels and immunodetection of MT1-MMP protein (C1 and C2), consistent with other reports. However, there is also an inconsistency when comparing these results to C3 and MDA-MB 231 cells, as C3 cells display no detectable levels of MT1-MMP protein via immunoblot, but some C3 cells did degrade the underlying ECM and displayed cytoplasmic MT1-MMP protein via immunofluorescence. Additionally, the pattern of ECM degradation and cytoplasmic MT1-MMP protein in MDA-MB 231 cells is most consistent with those of C3 cells, suggesting that cells that express low levels of MT1-MMP acquire invasive capabilities, despite reports that suggest high levels of MT1-MMP are typically needed for such phenotypes [24, 27, 28].

As our analysis using transient transfection of MT1-MMP demonstrated no change in cell migration, we wanted to further examine this migratory effect as well as the potential role of ERK activation using our stable cell lines that express lower levels of MT1-MMP than transiently transfected cells. As others have previously shown that TIMP-2 can modulate rapid ERK activation by MT1-MMP expressing breast cancer cells [24, 25], we utilized conditioned media (CM) that contained high levels of TIMP-2, or ALA + TIMP-2 (a TIMP-2 mutant that cannot inhibit MMPs enzymatic activity, but which has been shown to still bind cell surface MT1-MMP and upregulate ERK activation [24, 46]) to test this parameter. Analysis of these CMs via immunoblot showed that high but equal levels of TIMP-2 and ALA + TIMP-2 are present in the media, whereas reverse zymography analysis demonstrated that only TIMP-2 was capable of inhibiting MMPs (Additional file 1: Figure S1a). Using these CMs, we tested the ability of MCF-7 MT1-MMP cell lines to sequester TIMP-2 from the media by treating these cells with diluted CM. TIMP-2 levels remained equal after 15 min, but after 12 h of incubation TIMP-2 levels in the media decreased corresponding to the level of MT1-MMP expression (Additional file 1: Figure S1b), further confirming that these cells produced differing levels of functional cell surface MT1-MMP. Quantification of TIMP-2 protein in TIMP-2 CM via immunoblot was performed which demonstrated that undiluted TIMP-2 CM contains approximately 10 μg/ml TIMP-2 protein (Additional file 1: Figure S1c). The proMMP-2 activation ability of MCF-7 MT1-MMP cells was also tested in the presence of increasing levels of TIMP-2 or ALA + TIMP-2 CM by gelatin zymography and reverse zymography (Additional file 1: Figure S1d). Only C1 and C2 cells were capable of activating proMMP-2, and this was enhanced by addition of low levels of TIMP-2 (~100 ng/ml), but inhibited by high levels of TIMP-2 (> ~1 μg/ml) [25, 47, 48], consistent with the observations of others that TIMP-2 at 100 ng/ml is optimal for proMMP-2 activation by MT1-MMP expressing cells [49]. ALA + TIMP-2 did not enhance proMMP-2 activation in these cells consistent with the requirement for both the N- and C-terminal domains of TIMP-2 to act as an adapter for proMMP-2 [20]. Taken together, this analysis confirmed the functionality of our TIMP-2/ALA + TIMP-2 CMs and demonstrated how MCF-7 cells producing high levels of MT1-MMP protein follow the well-defined mechanism described for MT1-MMP/TIMP2 mediated activation of proMMP-2.

To then examine the level of ERK activation by MCF-7 MT1-MMP cell lines and the role of TIMP-2 in mediating this process, cells were incubated in media containing either 10 % FBS or in serum-free (SF) media containing different dilutions of TIMP-2 or ALA + TIMP-2 CM. Phospho-ERK levels were then examined using immunoblotting (Additional file 2: Figure S2) and quantified via densitometry (Fig. 4a). In media containing 10 % FBS, only C2 cells had a significantly higher level of phospho-ERK compared to parental MCF-7 cells. A 12-h incubation with TIMP-2 or ALA + TIMP-2 CM demonstrated a significant decrease in ERK activation in C2 cells. In contrast, a short 15 min incubation period with CMs demonstrated a significant increase in ERK activation of C2 and C3 cells, but not C1 cells, particularly with ALA + TIMP-2 treatment. Additionally, short incubation periods with dilutions of ALA + TIMP-2 CM showed that ERK activation in C2 and C3 cells, but not C1 cells, is significantly upregulated by high levels of ALA + TIMP-2 protein. Overall, analysis of phospho-ERK levels demonstrated that cells expressing medium (C2) and low (C3) levels of MT1-MMP demonstrate increased ERK activation that is upregulated by high levels of TIMP-2 (> ~ 200 ng/ml) and independent of MMP inhibition.

We then examined how TIMP-2-mediated ERK activation related to the migratory potential of MCF-7 MT1-MMP cell lines. A scratch closure assay over a span of 3 days demonstrated that only C3 cells closed the scratch significantly faster than parental MCF-7 cells, indicating increased migratory ability of C3 cells (Fig. 4b). Migration was then examined using transwell assays where the cells were incubated in SF media containing TIMP-2 and ALA + TIMP-2 CMs in the upper compartment and allowed to migrate towards the lower compartment containing 10 % FBS. TIMP-2 CM caused a significant increase in the number of migrated C2 and C3 cells, while ALA + TIMP-2 CM also caused a significant increase in the number migrated C3 cells (Fig. 4c, top). Higher dilutions of ALA + TIMP-2 CM showed a similar trend whereby only C2 and C3 cells demonstrated a significant increase in the number of migrated cells (Fig. 4c, bottom). C1 cells, which express the highest level of MT1-MMP, demonstrated an inability to increase migration potential when incubated with TIMP-2 CMs, and this was consistent with the lack of ERK activation upon TIMP-2 treatment in these cells.

To determine if the enhanced migratory ability of C3 cells was a result of the relatively low (11 fold) increase in MT1-MMP expression, we transfected these cells with a construct expressing an shRNA targeting MT1-MMP, and selected two cells lines: C3 SH 1, a partial knockdown which displayed a small but significant increase in MT1-MMP mRNA level (1.8 fold change p ≤ 0.01) as compared to parental MCF-7 cells, and C3 SH 2, a complete knockdown with 1.1 fold change (ns p = 0.62). A cell line was also derived from C3 cells stably expressing a scrambled control shRNA (C3 SH SCR) (Fig. 4d, top). Migration of these C3 cell line variants were assayed during incubation in SF media containing TIMP-2 or ALA + TIMP-2 CM, or ALA + TIMP-2 CM along with the ERK inhibitor U0126 (10 μM). Consistent with the previous transwell assays, C3 and control C3 SH SCR cells were significantly more migratory than parental MCF-7 cells when exposed to high levels of TIMP-2 (> ~1 μg/ml), and this correlated with ERK activation as ALA + TIMP-2/U0126 incubation attenuated this migration (Fig. 4d, bottom). This analysis also showed that complete knockdown of MT1-MMP expression (C3 SH 2) inhibited the TIMP-2-mediated increase in migration, indicating that MT1-MMP is needed for enhancement of migratory potential via TIMP-2. Surprisingly, C3 SH 1 cells, which express a small but significant (1.8 fold) increase in MT1-MMP expression, demonstrated a similar TIMP-2-mediated increase in migration via ERK activation. Additionally, we attempted to knockdown MT1-MMP expression in C1 and C2 cells to rescue migratory potential using the same approach, however, this was unsuccessful using transfection of shRNA (not shown) presumably because of the excessively high expression of MT1-MMP in these cells. Taken together, this data demonstrated that TIMP-2 enhancement of migratory potential is indeed dependent on MT1-MMP and ERK activation, but low levels (1.8 -11 fold) of MT1-MMP expression appear to be optimal to increase migratory ability, whereas high (> ~1000 fold) MT1-MMP expression does not result in augmented cell migration.

To analyze how MT1-MMP overexpression mediated survivability and how this might explain the different migratory potential of MCF-7 MT1-MMP cell lines, we measured cell viability using CellTiter 96® AQueous One Solution each day for 7 days during incubation in media containing 10 % FBS or SF media, which represents the two compartments in the transwell assay. In 10 % FBS, C1 cells (high MT1-MMP level) showed a significantly lower viability throughout the 7 days. Conversely C3 cells (low MT1-MMP level) were significantly more viable from seeding until day 3 (Fig. 5a, top). Similarly, in SF media C1 cells were less viable than parental MCF-7 cells from seeding until day 5, whereas C3 cells were significantly more viable during this time (Fig. 5a, bottom). After 6 days of incubation in SF media, all MT1-MMP expressing cells were significantly more viable than MCF-7 cells indicating that MT1-MMP has an essential role in mediating survivability during prolonged absence of growth factors. To confirm that MT1-MMP was responsible for this increased viability during SF incubation, we repeated the analysis using the C3 knockdown variants. Consistent with the previous analysis, only C3 cells had significantly higher viability at day 3, but by day 6 and 9 all MT1-MMP expressing cell lines were significantly more viable than MCF-7 cells (Fig. 5b). Viability measurements of the C3 knockdown variants demonstrated that MT1-MMP is responsible for enhanced viability during SF incubation as C3 SH 2 cells, which are MT1-MMP deficient similar to parental MCF-7 cells, are less viable than other cell lines at day 6, and show the same viability as MCF-7 cells at day 9. To corroborate these results, we also analyzed cell lysate from all MCF-7 cell lines via immunoblot (Fig. 5b, bottom) for the cell proliferative marker phospho-histone 3 (PH3), and quantified this using densitometry (Fig. 5b, bar graph). All MCF-7 cell lines that expressed significantly higher levels of MT1-MMP than MCF-7 cells showed higher levels of PH3 after incubation in SF media, particularly C2 cells, which accumulated high levels of active MT1-MMP protein, and showed significantly higher levels of PH3 than MCF-7 cells.

We then examined how inhibition of select signaling pathways affected the viability of MCF-7 MT1-MMP cell lines during prolonged SF incubation to gain insight into the mechanism behind MT1-MMP-mediated survivability. The chemical inhibitors U0126, BB94, furin inhibitor II, and AKT inhibitor IV were used to examine their effect on the viability of MCF-7 MT1-MMP cell lines after 6 days of SF incubation. ERK inhibition and AKT inhibition significantly decreased the viability of all MT1-MMP expressing cells consistent with the observations of others [21, 24]. Furin inhibitor (which would inhibit intracellular activation of MT1-MMP) showed a non-significant dose dependent decrease in the viability of cells expressing MT1-MMP, a trend not seen in MCF-7 parental cells, indicating that high levels of pro-MT1-MMP likely negatively mediated viability. Unlike all the other inhibitors that generally decreased viability, low concentrations of BB94 trended to increase the viability of all MT1-MMP expressing cells, especially C1 cells which displayed a significant increase in viability after 6 days of SF incubation with 2.5 μM BB94 (Fig. 5c). Taken together, this data demonstrated that MT1-MMP mediated survivability to serum-free stress, but that excessive MT1-MMP proteolytic activity may be counterproductive to increased viability. Furthermore, this analysis also demonstrated that cell viability has to be considered as a component to the pattern of migration seen in transwell assays, as C3 cells which were the most migratory are also the most viable, whereas in contrast C1 cells were the least viable and migratory. Therefore, it is difficult to delineate the relative contribution of changes in cell proliferation or changes in migratory capability to the difference in the number of cells quantified in the bottom compartment of a transwell at the endpoint of the assay.

Here we present evidence that high MT1-MMP overexpression (transient transfectants, C1 and C2 cells) does not correlate with the migration of MCF-7 breast cancer cells, as seen using the transwell migration assay, which we showed also included a substantial cell viability component. Increased MT1-MMP-mediated viability may have yielded a higher number of cells in the transwell compartments, which may have resulted in incorrect conclusions about migratory potential if viability were not considered. So to confirm that high MT1-MMP overexpression (>1000 fold compared to MCF-7 cells) does not result in increased migration, we used a live imaging approach with automated migration quantification of MCF-7 MT1-MMP cells that were stably expressing fluorescent zsGreen protein and incubated on Alexa594 gelatin coated coverslips in control conditions (0.1 % DMSO) (Fig. 6a, Additional files 3, 4, 5 and 6) and with BB94 (10 μM) (data not shown). Automated quantification using the ADAPT plugin for ImageJ software [44] allowed us to track the migration of all individual cells in three independent experiments, which were then grouped according to their distance migrated (Fig. 6b, Additional file 7). This analysis showed that high MT1-MMP overexpression inhibited migration of MCF-7 cells as a higher proportion of C1 and C2 cells migrated a shorter distance from the initial point of tracking compared to parental MCF-7 cells, as 53 % and 79 % of C1 and C2 cells, respectively, migrated more than 25 μm, compared to 86 % of parental MCF-7 cells that migrated the same distance. In contrast, C3 cells did have increased migration ability compared to MCF-7 cells, as 92 % of C3 cells migrated more than 25 μm. Importantly, this 6 % increase in migratory ability between MCF-7 and C3 cells does not reflect the increase seen using the transwell assays (20-100 % increase), confirming that there is a cell viability component underlying the number of cells that transgressed to the bottom transwell compartment. BB94 incubation did not affect the migration of MCF-7 cells or C2 cells, but did alter the migratory patterns of C1 and C3 cells. Inhibition of MMP activity by BB94 decreased the migratory ability of C3 cells as 92 % of cells migrated more than 25 μm in control conditions compared to 66 % during BB94 incubation. C1 cells, which express the highest level of MT1-MMP and are the least migratory under control conditions, markedly improved their migratory potential during BB94 incubation as 73 % of cells migrated more than 25 μm (compared to 53 % in control conditions). This migration data is consistent with the cell viability assay during BB94 incubation, as C1 cells saw an augmentation rather than decrease in viability and migration when MMPs were inhibited, suggesting that excessive MMP activity is counterproductive to those cellular processes. Analysis of the percentage of cells that degraded the underlying ECM demonstrated a time-dependent increase in ECM degradation correlated to high MT1-MMP expression in C1 and C2 cells (Fig. 6c), confirming functional MT1-MMP protein production in these cells during the course of this assay. Incubation with BB94 (10 μM) did not allow for degradation of the fluorescent gelatin by MT1-MMP expressing cells (not shown). Taken together, this analysis demonstrated that low levels of MT1-MMP expression (11 fold compared to parental cells) are optimal for increased migratory ability of MCF-7 breast cancer cells, whereas high MT1-MMP overexpression (>1000 fold) does not increase migration but allows for widespread ECM degradation.

To further investigate the correlation between MT1-MMP expression levels and the migratory potential of cells, we extended our observations to MDA-MB 231 and HS578T breast cancer cells by first assessing endogenous MT1-MMP levels in these cell lines. Both MDA-MB 231 and HS578t cells expressed significantly higher levels of MT1-MMP mRNA than MCF-7 cells (~172 and ~100 fold, respectively), and are significantly different from each other (Fig. 7a). Consistent with their high levels of expression, MDA-MB 231 and HS578t cells produce active MT1-MMP protein, and also have higher levels of phospho-ERK than MCF-7 cells. Reverse zymography analysis showed that MDA-MB 231 and HS578t cells also produce higher levels of TIMP-2 protein compared to MCF-7 cells (Fig. 7b). Gelatin zymography showed that MDA-MB 231 cells predominately secreted proMMP-9, whereas HS578t cells secreted high levels of proMMP-2 (Fig. 7c). Transwell migration assays demonstrated that MDA-MB 231 and HS578t were significantly more migratory than MCF-7 cells (Fig. 7d). Importantly, HS578t cells were significantly more migratory than MDA-MB 231 cells even though they express lower levels of MT1-MMP.

We then tested how overexpression of MT1-MMP in MDA-MB 231 cells affects the migration of cells that natively express MT1-MMP and have endogenous migratory ability. MDA-MB 231 MT1-MMP stable cell lines were created and three were selected which produce the isoforms of MT1-MMP shown via immunoblot (Fig. 8a). MDA-MB 231 C1 cells produce pro- and active- MT1-MMP, whereas MDA-MB 231 C2 and C3 produce predominantly active MT1-MMP at higher levels than parental MDA-MB 231 cells. Transwell migration analysis of these MDA-MB 231 MT1-MMP cell lines showed that overexpression of MT1-MMP significantly decreased their migratory potential compared to parental MDA-MB 231 cells (Fig. 8b). Viability of these cell lines was analyzed in 10 % FBS (top) and SF media (bottom) in the same manner as MCF-7 cell lines to show that MT1-MMP overexpression significantly inhibited viability, particularly of MDA-MB 231 C1 and C3 cells (Fig. 8c). Further, MDA-MB 231 C2 cells, which accumulate high levels of active MT1-MMP protein, exhibited increased viability in SF media and during prolonged incubation in 10 % FBS, which was consistent with both our PH3 analysis of MCF-7 C2 cells, and with their higher migration levels relative to the other MDA-MB 231 MT1-MMP cell lines. Taken together these analyses show that naturally increased MT1-MMP expression in transformed cancer cells (MDAMB-231, HS578t cells) does result in enhanced migration, but this is not the case for experimental MT1-MMP overexpression (MT1-MMP cell lines), as increasing overexpression of MT1-MMP decreases rather than enhances the migratory potential of breast cancer cells.

As our work using in vitro 2D culture demonstrated that increased, but relatively low levels of MT1-MMP expression were optimal for enhanced survivability and migration, we subsequently examined how these observations translated to ex vivo 3D culture. In this culture system breast cancer cells are embedded in Matrigel, whose composition resembles the components of a natural BM, and as such mimics a more physiologically relevant 3D environment [43]. MCF-7 MT1-MMP cell lines were embedded in Matrigel and cultured for 5 days in media containing 10 % FBS. During this time, cell morphology was monitored using DIC microscopy to examine how MT1-MMP expression affected cell behavior. Non-invasive MCF-7 cells have been shown to form and remain in circular colonies during 3D culture [50] (Fig. 9a, white arrow), which resemble the acini-like structures that normal breast cells form that are representative of the terminal ductal lobular unit (TDLU) in the human breast [51]. In contrast, invasive MDA-MB 231 cells lose this circular morphology and develop into an interconnected network within the Matrigel (Fig. 10a, red arrow).

During 3D culture, C1, C2 and C3 cells do not form an interconnect network as seen with MDA-MB 231 cells. Instead, their colonies remained circular but display two distinct phenotypes; disseminations - distinct particles being released from circular colonies (Fig. 9a, green arrow), and short dynamic protrusions emerging from colonies (Fig. 9a, red arrows), which were confirmed using time-lapse microscopy (Additional file 8: Figure S3a, Additional files 9, 10, 11 and 12). Comparison of these morphological features between cell lines were blindly quantified per 10× magnification fields of view each day for 5 days and normalized to the number of circular colonies per field of view (Fig. 9a, line graphs). This quantification demonstrated that MCF-7 expressing high levels of MT1-MMP (C1 and C2) had significantly higher number of disseminations per colony than MCF-7 and C3 cells, particularly C1 cells which showed a time-dependent release of particles (Additional file 10). In contrast, the number of protrusions per colony at day 5 was inversely proportional to MT1-MMP expression, as C2, and especially C3 cells had significantly more protrusions per colony, whereas C1 cells had a significantly lower number of protrusions than MCF-7 cells. Immunofluorescence analysis of these MCF-7 MT1-MMP cell lines (Fig. 9b), confirmed that while MCF-7 colonies retained circularity during 3D culture (white arrows), C1 and C2 cells had clear disorganization of colony structure with released cell fragments (disseminations) containing MT1-MMP protein (green arrows). In contrast, C3 cells demonstrated long F-actin protrusions emerging from circular colonies (red arrows), consistent with the protrusive phenotype.

To further characterize this novel dissemination phenotype mediated by MT1-MMP, the cellular composition of released fragments was examined by repeating the immunofluorescence analysis with the zsGreen MCF-7 tagged cells. The fluorescently tagged variants were used to provide information about the origin of morphological structures as zsGreen accumulated in the cytoplasm and could be used to better visualize the origins of disseminations or protrusions. The number of single cells (marked by DAPI), F-actin disseminations, F-actin protrusions, and zsGreen protrusions were quantified per colony from 20× magnification 3D volume views (Additional file 8: Figure S3b). This quantification demonstrates that the disseminations are likely enucleated cells, as C1 cells show significantly higher levels of single cells (0.18/colony) and F-actin disseminations (0.67/colony) (Fig. 9c), which added together represented a density of ~0.85/colony, similar to the ~1 dissemination/colony seen by day 5 in our DIC microscopy analysis (Fig. 9a). In contrast, C3 cells demonstrate higher density of F-actin (0.65/colony) and zsGreen protrusions (0.15/colony), which when added together represent a density of ~0.8/colony similar to the ~1 protrusion/colony seen by day 5 in our DIC analysis, confirming that these cells demonstrate a protrusive morphology in 3D culture. Volume views at 60× magnification also showed C3 colonies (Fig. 9c) that demonstrated many zsGreen protrusions (blue arrow) that emerged from existing F-actin protrusions (red arrow). These protrusions are indicative of invadopodia, which are characterized by extension of the cell body due to F-actin polymerization, and subsequent F-actin retraction to result in a protrusion devoid of F-actin [35].

To corroborate our 3D observations with the MCF-7, C1, C2 and C3 cell lines, we performed a similar analysis with MDA-MB 231 MT1-MMP cell lines. MDA-MB 231 cells progressively lost circular morphology and developed into an interconnect network of cells within the Matrigel (Fig. 10a, red arrow), represented by a significantly higher number of protrusions per colony (~25) than MCF-7 cells. In contrast, all MDA-MB 231 cells expressing MT1-MMP showed a significant inhibition to form networks in 3D culture and instead retained a large proportion of circular colonies (Fig. 10a, white arrows). As well, MDA-MB 231 MT1-MMP cells demonstrated a nonsignificant increase in the number of disseminations (~0.4 dissemination/colony), although it was comparatively lower than in MCF-7 MT1-MMP cells. Immunofluorescence analysis of F-actin and nuclear structures in 3D culture confirmed these observations and showed that parental MDA-MB 231 cells formed an extensive cellular network, whereas MDA-MB 231 MT1-MMP cells retained circular colonies marked by MT1-MMP protein (Fig. 10b, green arrows). Taken together, this ex vivo analysis demonstrated that low MT1-MMP expression mediated a protrusive phenotype in 3D culture, shown by C3 (Fig. 9a) and MDA-MB 231 cells (Fig. 10a), whereas high levels of MT1-MMP expression (C1, C2, and MDA-MB 231 MT1-MMP cells) inhibited the ability to form protrusive networks in 3D culture and instead resulted in abnormal release of cell fragments.

We then tested the in vivo implications of our in vitro observations by using the ex ovo chicken embryo [52] and zsGreen expressing MCF-7 C1, C2, and C3 cell lines. These cells, along with zsGreen MDA-MB 231 cells, were suspended in Matrigel and implanted into the ChorioAllantoic Membrane (CAM) of day 9 ex ovo chicken embryos [45]. Eight days post implantation the resulting xenograft was visualized using a fluorescent stereoscope to assess vascularization of the formed tumour. No MCF-7 or C1 tumours were scored as vascularized, whereas a minority (2/14) of C2 tumours demonstrated vascularization by vessels of the chicken embryo (Fig. 11). In contrast, the majority of C3 (14/15) and MDA-MB 231 (15/16) tumours were vascularized, shown by the presence of non-fluorescent vessels present within the zsGreen tumour (Additional file 13: Figure S4, white arrow).

The metastatic potential of these cell lines was subsequently examined by direct intravenous injection of zsGreen MCF-7 C1, C2, and C3 cell lines into the vasculature of day 14 chicken embryos, followed by assessment of extravasation efficiency 24 h post-injection [37]. Extravasation out of the vasculature is a necessary step that precedes metastatic colony formation and as such has been shown to directly correlate with metastatic potential [4]. MCF-7 and C1 cells, which express the highest level of MT1-MMP, demonstrate poor extravasation efficiency, as < 5 % of cells were able to extravasate after 24 h (Fig. 12a). C2 cells showed a higher but non-significant increase in percentage of extravasated cells (~8 %), whereas C3 cells, which express low levels of MT1-MMP, displayed a significant increase in extravasation efficiency (~15 %) compared to all other MCF-7 cell lines. Orthogonal sections of extravasated C1 and C3 cells acquired using confocal microscopy at 60× magnification showed that C1 cells are capable of extravasating out of the CAM vasculature (Fig. 12b) but display membrane blebbing (white arrow) and cell fragment release (green arrow), reminiscent of our observations in 3D culture (Fig. 9). In contrast, C3 cells exhibited a uniform morphology as they extravasated into the stromal space (blue arrows) and contained cell protrusions trailing from the CAM capillary bed into the stroma (Additional file 14, red arrow), suggesting that these cells formed invadopodia in vivo.

To extend and corroborate our observations that low levels of MT1-MMP are optimal to promote metastatic features in 3D culture and in vivo, we assayed via immunoblot the level of MT1-MMP protein in the 21 T series cell lines. These cells were isolated from a single patient and represent a mammary tumor progression series that mimic specific stages of breast cancer progression [39, 53] (atypical ductal hyperplasia -21PT-ADH, ductal carcinoma in situ -21NT – DCIS and invasive mammary carcinoma- 21MT-1- IMC) (Fig. 13). Assaying MT1-MMP protein levels in these cell lines demonstrated that ADH and DCIS variants produced active MT1-MMP, with the non-invasive DCIS cells producing higher levels of MT1-MMP protein. Direct comparison to the MCF-7 MT1-MMP cell lines showed that C1 and C2 cells produced more active MT1-MMP than ADH or DCIS 21 T cells. In contrast, 21MT-1 cells, which represent an invasive mammary carcinoma, produced undetectable levels of MT1-MMP protein as determined by immunoblot, similar to C3 cells. The 21MT-1 IMC cells have been show to possess metastatic qualities in 3D culture and in vivo when compared to the non-invasive variants [39], consistent with our observations that low levels of MT1-MMP are representative of metastatic breast cancer. Taken together, the observations in our study demonstrate that low levels of MT1-MMP expression are optimal for tumorigenicity and metastatic potential in vivo, and importantly, that abnormally high MT1-MMP overexpression correlated with a decrease, rather than enhancement of tumorigenic features (see schematic representation Fig. 14).