Canine mammary tumor-U27 and CMT-U309 CMT cell lines (20, 21) were a kind gift from Prof. Eva Hellmen at Uppsala University. CMT-U27 was isolated from simple carcinoma and CMT-U309 from spindle cell tumor. Cells were cultured in RPMI 1640 with 10% FBS and penicillin–streptomycin in an incubator with 5% CO₂, at +37°C temperature and were plated on dishes 1 day prior to experiments.

Cells were cultured on glass coverslips, washed with PBS and fixed with 4% PFA. Fixed cells were permeabilized with 0.1% Triton X-100 in TBS for 5′ and moved to 0.2% Dulbecco/BSA. Immunofluorescence stainings were performed as described previously (26) and were repeated at least three times. The following primary antibodies were used in stainings: anti-E-cadherin (#14472, Cell Signaling Technology, Inc., Danvers, MA, USA), anti-vimentin (#5741, Cell Signaling Technology, Inc., Danvers, MA, USA), and anti-vinculin antibody (1:50) (hVin-1, Sigma, Saint Louis, MO, USA). The following secondary antibodies were used to detect the primary antibodies: Alexa Fluor α-rabbit 488 and α-mouse 568 (Life Technologies™, Carlsbad, CA, USA). Other reagents: Alexa-488- and -647-Phalloidins were used to visualize actin cytoskeleton in 1:200 dilution (Life Technologies™, Carlsbad, CA, USA), DAPI for DNA (Life Technologies™, Carlsbad, CA, USA), and DABCO/Mowiol was used in mounting. The images were acquired with Leica DM6000 upright fluorescence wide field microscope equipped with Hamamatsu Orca-Flash4.0 V2 sCMOS camera.

Cells were washed with PBS and lysed in 1% Triton X-100/PBS with protease and phosphatase inhibitor cocktail sets (539131 and 539131, Calbiochem, San Diego, CA, USA). Protein concentrations were measured with Qubit^® Protein Assay Kit (ThermoFisher Scientific, Waltham, MA, USA). Sample loading buffer 4× LSB-DTT buffer was added to lysates and samples were boiled for 5′ before loading the samples in SDS-PAGE gels. Semidry transfer and Immobilon-P Membrane, PVDF filter (Millipore, Billerica, MA, USA) was used for blotting. Mixture of 5% milk/BSA was used for blocking. Following antibodies were used for detection of specific proteins: mouse anti-p63 (4A4, Abcam, Cambridge, UK), rabbit anti-slug (#9585, Cell Signaling Technology, Inc., Danvers, MA, USA), mouse anti-E-cadherin (#14472, Cell Signaling Technology, Inc., Danvers, MA, USA), rabbit anti-N-cadherin (13116, Cell Signaling Technology, Inc., Danvers, MA, USA), rabbit anti-claudin-1 (#13255, Cell Signaling Technology, Inc., Danvers, MA, USA), mouse anti-cytokeratin 5 (XM26, Abcam, Cambridge, UK), rabbit anti-cytokeratin 14 (EPR17350, Abcam, Cambridge, UK), anti-Pan-keratin mAb (C11, #4545, Cell Signaling Technology, Inc., Danvers, MA, USA), rabbit anti-vimentin (#5741, Cell Signaling Technology, Inc., Danvers, MA, USA), rabbit anti-P-Thr172-AMPK (#4188, Cell Signaling Technology, Inc., Danvers, MA, USA), anti-calponin 1/2/3 Antibody (FL-297, Santa Cruz, CA, USA), anti-tropomyosin 1, 2, 3 (TM31; T2780, Sigma, Saint Louis, MO, USA), rabbit anti-P-thr18/ser19-myosin light chain (MLC) (3674, Cell Signaling Technology, Inc., Danvers, MA, USA), mouse anti-β-actin (clone AC-15, Sigma, Saint Louis, MO, USA), mouse anti-smooth muscle actin (α-SMA, clone 1A4, A5228, Sigma, Saint Louis, MO, USA), mouse anti-GAPDH (G8795, Sigma Aldrich, Saint Louis, MO, USA). Anti-mouse or -rabbit HRP-linked secondary antibodies (Cell Signaling Technology, Inc., Danvers, MA, USA) and Western HRP substrate (Luminata™Crescendo, WBLUR0100, Millipore, Billerica, MA, USA) were used for chemiluminescence detection of the protein bands. All Western Blot experiments were repeated at least three times.

Cell proliferation assays were performed with Cell IQ live-cell imaging setup as in wound-healing assays. For measuring cell doubling time, CMT-U309 and CMT-U27 cells were sparsely plated on 12-well plate 1 day before the actual experiment. The following day, cell growth/proliferation was monitored for 24 h and images were acquired every 30 min. Cell growth rate (GR) was calculated by using the formula: GR=ln(N(t)/N(0))t, and cell doubling time was calculated: ln(2)GR, where N (t) = number of cells in the end of experiment, N(0) is cell number in the beginning and t = time in hours.

Cells were cultured on fibronectin (FN)-coated, shape-determined micropatterns (CYTOOchips™, Grenoble, France) for 6 h and fixed with 4% PFA for 20′. Actin cytoskeleton was stained with Alexa-488-Phalloidin and nuclei with DAPI. Analysis of contractile function is based on the curvature of the formed thick contractile actomyosin bundle (see also Figure 3B). Radius of curvature (R) was calculated by using the formula: R=(L/2)2+w2w, where L = length and w = width. Measurements of cell edges were done with ImageJ program.

To measure cell-exerted traction forces, CMT-U27 and -U309 cells were cultured for 2–4 h on elastic collagen-1-coated polyacrylamide-based gel substrates of known stiffness (Young’s Modulus/elastic modulus = 11 kPa). Substrates were surface-coated with sulfate fluorescent microspheres (Invitrogen, Carlsbad, CA, USA, diameter 200 nm). Single isolated cells together with the underlying microspheres were imaged with 3I Marianas imaging system (3I intelligent Imaging Innovations, Germany), at multiple locations. To perform the imaging, a 63×/1.2 W C-Apochromat Corr WD = 0.28 M27 objective was used. The system was placed in a heated sample chamber (+37°C) and controlled for CO₂. Following live cell imaging, the cells were detached from the substrates with 10 × Trypsin (Lonza Group, Basel, Switzerland) and a second set of microsphere images were obtained in a cell-free configuration. These images served as reference. Cell-exerted traction forces cause displacement of microspheres within the elastic substrate and spatial maps of microsphere displacements were achieved by comparing the reference microsphere images together with the experimental images. With knowledge of the cell-exerted displacement field, substrate stiffness (11 kPa), and a manual trace of the cell boundary, we computed the cell-exerted traction field using Fourier Transform Traction Cytometry (27, 28). From the traction field, we computed the root mean squared magnitude.

Wound-healing assays were performed to study cell migration by using Cell IQ live-cell imaging setup consisting of a microscope (Nikon) equipped with 10×/0.30 Ph1 objective and an incubator at +37°C with CO₂. Images were captured with Qimaging Retiga EXi camera using Cell IQ Imagen software. For wound-healing assay, CMT-U309 and CMT-U27 cells were grown to confluency on 12-well plate (CellStar). The wound (gap) was created by scratching the confluent monolayer with a 200 µl pipette tip, and the debris was removed immediately by rinsing twice with DMEM. Cells were monitored for 24 h and images were acquired every 30 min. Data were analyzed using Cell IQ Analyser. The half-closure time was calculated by using the formula: t1/2gap=Initial gap Area2xslope. Gap area was measured and plotted as a function of time. Cell migration rate (v_migration, micrometer per hour) was calculated from the slope of this plot: υ migration=slope2xl, where v = velocity and l = length of the gap.

ECM gel from Engelbreth-Holm-Swarm murine sarcoma (Sigma E1270) was prepared according to manufacturer’s instructions. CMT-U309 and CMT-U27 cells were trypsinized and ~5,000 cells in DMEM were seeded on top of the matrigel-coated eight-chamber slides. 3D spheroids were observed under microscope daily for 2 weeks.

For the inhibition of AMPK activity, cells were treated with AMPK inhibitor, compound C (Sigma-Aldrich). On 2D cultures, final concentration of compound C was 5 μM and incubation time 3–6 h.

In 3D spheroid cultures, 15 µM compound C was applied on cells for 6 days to completely inhibit AMPK activity. For the activation of AMPK activity, cells were treated with 25 μM AICAR (Sigma Aldrich, Saint Louis, MO, USA) for 16 h.

SDs and statistically significant difference between means of two groups (t-test) were analyzed in Microsoft Excel 2013. Box charts were done with OriginPro 8.6 program (Whisker range 5–95, showing outliers) and column charts were done with Microsoft Excel 2013.

We examined the morphological features and a set of lineage-specific markers of the two established CMT cell lines, simple carcinoma CMT-U27, and spindle cell carcinoma CMT-U309 (20, 21). Some of these markers have previously been shown by other studies (20, 21, 29). CMT-U27 exhibited pleomorphic cells as shown in Figure 1A, left panel. These cells were contact inhibited by increasing cell confluency, while the spindle carcinoma cells, CMT-U309, had an elongated-morphology and tended to overgrow the neighboring cells (Figure 1A, right panel). The examination of epithelial cell adhesion proteins showed that E-cadherin and claudin-1 are not expressed in CMT-U309 cells (Figure 1B). In contrast, CMT-U309 cells were showing high levels of N-cadherin (Figure 1B), a protein, which is often associated with the decrease in E-cadherin expression in so-called cadherin switch during epithelial-to-mesenchymal transition and linked to metastatic breast cancer (30, 31). CMT-U27 cells did not show any expression of N-cadherin but had clear expression of epithelial adhesion proteins E-cadherin and claudin-1 (Figure 1B).

The average cell doubling time for CMT-U27 and CMT-U309 was found to be 28.5 and 21.1 h, respectively (Figure 1C). Faster proliferating spindle carcinoma CMT-U309 cells showed clear reactivity for the basal cell marker p63, while CMT-U27 cells expressed very low levels of this marker (Figure 1D). Transcription factor p63 has been shown to be highly expressed in both human and canine spindle cell mammary cancer tumors (32, 33). CMT-U309 cells also expressed snail family transcription repressor slug that has been implicated in maintaining the basal phenotype [reviewed in Ref. (34)], while CMT-U27 cell line had almost non-detectable levels of this protein (Figure 1D). Furthermore, CMT-U27 cells did not show any reactivity for the mesenchymal marker vimentin, while it was highly expressed in CMT-U309 spindle cells (Figures 1E,F).

Cytokeratins (CKs) play a significant role in the progression of mammary tumors (35). Therefore, we examined the expression of CKs 4, 5, 6, 8, 10, 13, 14, and 18 by utilizing a Pan-Keratin antibody as well as by using a specific antibody against CK5 and CK14. Both CMT-U27 and CMT-U309 showed similar expression of CK5, as detected by western blotting (Figure 1G). Basal CK14 was distinctly expressed in the spindle carcinoma cells (Figure 1G). The western blot analyses with Pan-Keratin antibody showed similar pattern of protein bands between these cell lines. However, the intensity of bands corresponding to the size of luminal epithelial markers CK8 and CK18 (53 and 48 kDa, respectively), showed clearly lower levels in CMT-U309 cell line (Figure 1G). Thus, the expression patterns, verified by several lineage-specific markers, were suggestive for distinct luminal- and basal-like features within the studied cell lines.

Actin cytoskeleton undergoes dramatic reorganization during the process of malignant transformation (36). During this transformation, several actin-binding proteins are known to be deregulated and contribute to the development of abnormal adhesive and migratory properties of the cancer cells. We examined the organization of actin cytoskeleton as well as expression of cytoskeletal proteins in CMT-U27 and CMT-U309 cancer cell lines (Figure 2A). CMT-U27 cells exhibited variable organization of their actin cytoskeleton, including both thick contractile bundles as well as thin actin-based structures both parallel and perpendicular to the leading edge of the cells (Figure 2A). These thin structures resembled structures that act as precursors for the mature thick contractile actomyosin bundles (26, 37, 38). In contrast, CMT-U309 cells had very prominent, thick contractile actin-based bundles, bound to focal adhesion sites from their both ends (Figure 2A).

Both CMT-U27 simple carcinoma and CMT-U309 spindle carcinoma cells appeared to have similar expression level of β-actin (Figure 2B). Additionally, spindle carcinoma cells expressed high levels of smooth muscle actin (α-SMA) and calponin-1 (Figure 2B), typical for these type of carcinoma cells. In respect of non-muscle myosin and high molecular weight tropomyosins (Tm1, 2, and 3), these cell lines showed also clear differences. CMT-U27 possessed high levels of phosphorylated non-muscle MLC (P-thr18/ser19-P-MLC), indicative of active myosin that is required for the contraction. In contrast, in CMT-U309, its level was almost non-detectable (Figure 2B). These cell lines had also distinct expression patterns of tropomyosins as CMT-U309 showed predominantly expression of Tm1 and Tm2, while CMT-U27 had high expression of Tm3 isoform (Figure 2B).

As the maturation of contractile actomyosin bundles from their precursor was recently found to be dependent on the activity of 5′ AMPK (26), we examined the levels of active AMPK, P-Thr-172 AMPK, from the lysates of both CMT cell lines (Figure 2C). Lysates from CMT-U27 cells contained low levels of active AMPK. In contrast, CMT-U309 cells were found to have high levels of P-thr172-AMPK, correlating well with the presence of thick, mature actin-based bundles in these cells. The results suggest that the high AMPK kinase activity could be responsible for the great number of mature actomyosin bundles in CMT-U309 cells.

Actin-based structures provide force for many cellular processes [reviewed in Ref. (39)]. These contractile actomyosin forces can be exerted through the adhesion sites to the surrounding extracellular matrix as traction forces. We analyzed the contractile properties of CMT-U27 and CMT-U309 cell lines by utilizing FN-coated shape-determined micropatterns on 2D cultures (CYTOOchips™) (Figures 3A,B). Cells cultured on these patterns organize their mature contractile bundles at the edges of these patterns and precursor structures are typically found behind the formed leading edge of the cell, at the arc of the micropattern (Figure 3A). The curvature of the formed contractile bundle is related to its contractility. Contractile function of the cells, cultured on these patterns, can be calculated as radius of curvature, R, as described in Figure 3B. Based on this analysis method, CMT-U27 cells were significantly less contractile than CMT-U309 cells (Figure 3C).

Furthermore, we measured the cell-exerted forces directly by utilizing TFM. In this method, cells are cultured on an elastic substrate, embedded with fluorescent microspheres, and the cell-exerted forces cause displacement of these beads. With the knowledge of the microsphere displacement field and stiffness of the elastic substrate, cell-exerted traction forces can be computed. Consistent with the contractility measurements by using micropatterns, described above, CMT-U27 exerted significantly smaller traction forces than CMT-U309 cells (Figures 3D,E).

Recent studies have shown that cell-substrate forces directly correlate with the invasion potential of cancer cells in 3D matrices and that several invasive human cancer cell lines apply higher forces than their non-invasive counterparts (7, 8). We compared the migratory properties of CMT-U27 and CMT-U309 in both 2D cultures and in 3D matrigel cultures. The migration speed of CMT-U27 and CMT-U309 cells was 9.5 and 14.5 μm/h, respectively (see Figures 4A–C). Confocal images showed that, in 3D collagen cultures, the morphology of these two cell types were totally different (Figure 5A): CMT-U27 cells were rounded with cortical actin staining and bleb formation, while CMT-U309 cells exhibited long actin-based protrusion and elongated form. In 3D matrigel cultures, both cell lines initially formed round symmetrical spheroids (Figure 5B). After 3 days in culture, the spindle carcinoma cells started to exhibit invasive features and escape from the primary structures that had already undergone through massive expansion (Figures 5B,C). CMT-U27 cells were highly proliferative but did not invade to the surrounding matrix, while CMT-U309 cultures showed single invading cells that started to form new spheroids outside the primary structures (Figure 5B).

Next, we examined whether the invasion potential of CMT-U309 cell line was dependent on its contractile properties. Since the maturation of force-producing actomyosin bundles was recently shown to be controlled by AMPK (26) and CMT-U309 cells also exhibited high levels of active AMPK (Figure 2C), we tested how inhibition of this kinase would affect contractility and actin-based invasion in this particular cell line. AMPK activity was inhibited by compound C in 3D matrigel cultures (Figures 6A,B). Importantly, in the inhibitor-treated 3D cultures, CMT-U309 cells still proliferated and formed predominantly symmetrical spheroid-like structures. However, their ability to invade was completely prevented (Figure 6A). To verify that AMPK inhibition actually affects the formation and morphology of contractile actomyosin bundles, we performed 2D cultures of these cells and analyzed the organization of actin cytoskeleton in both ctrl and inhibitor-treated cells. Inhibition of AMPK activity for 6 h was found to significantly decrease the amount of fully matured contractile actomyosin bundles and in contrast increase the amount of thinner precursor structures (Figures 6C,D). In line with this, opposite stress fiber phenotype was detected in CMT-U27 cells that were exposed to a commonly used AMPK activator, AICAR (Figure S1 in Supplementary Material), suggesting that the maturation of contractile actomyosin bundles is dependent on AMPK activity also in these particular CMT cell lines. Furthermore, we analyzed whether the effect of AMPK inhibitor on the organization of contractile actomyosin bundles in CMT-U309 cells could be dependent on the expression levels of α-SMA, calponin-1, and tropomyosins. However, we could not detect any alterations in their levels in Western Blots within the same time frame as the reorganization of actin-based structures was detected (data not shown).

Moreover, we wanted to analyze how these changes in actin-based structures after AMPK inhibition would be reflected to cell traction forces. TFM was performed in ctrl (non-treated) and compound C-treated CMT-U309 cells (Figure 6E). After the inhibition, the cell spreading area was about 80% of the size of untreated cells (Figure 6F), while the average Root Mean Square tractions in AMPK-inhibitor-treated cells were approximately five times smaller in comparison with ctrl cells (Figure 6G). The decrease in cell-mediated forces correlated well with the detected decrease in the amount of thick actomyosin bundles after AMPK inhibition (Figures 6C,D). Our results suggest that AMPK activity controls the maturation of force-producing actomyosin bundles in this cell line and that high traction forces are connected to the invasion potential of these canine spindle carcinoma cells.