Single-copy Snail upregulation causes partial epithelial-mesenchymal transition in colon cancer cells

Cells were cultured in Dulbecco’s Modified Eagle Medium (Gibco) containing 4.5 g/L D-Glucose and L-Glutamine and supplemented with 10% tetracycline-free foetal bovine serum (Clontech) and 1% penicillin and streptomycin (Gibco). Cells were passaged when confluent with 0.05% trypsin (Gibco) and incubated in a Heracell 150 incubator, 20% oxygen and 5% carbon dioxide at 37 °C. DLD-1 cells were a kind gift from Stephen Taylor, University of Manchester. HCT-116, SW480, and SW620 cells were from group stocks. All cell lines were verified by short-tandem repeat genotyping undertaken by the in-house Biorepository. Cell cultures were regularly tested for mycoplasma. Antibiotics were added for selection in the following concentrations: Blasticidin S HCl (Invitrogen) 8 μg/ml, hygromycin B (Invitrogen) 400 μg/ml, puromycin (Invitrogen) 3.5 μg/ml, and zeocin (Sigma) 60 μg/ml.

Snail^wt and Snail^lo: DNA fragments were isolated from restriction digest or polymerase chain reaction of plasmids, assembled by ligation using Infusion Cloning (Clontech) and transformed into XL-2 Ultracompetent Blue Cells (Agilent Technologies) by heatshock. Plasmids were extracted from bacteria, sequence verified by Sanger sequencing, then inserted into a DLD-1 Flp-In T-Rex cell line containing a single Frt site (a kind gift from Professor Steve Taylor, University of Manchester) following the protocol in Tomic et al., 2018 [30]. The copy number and location of integration of plasmids into cells were verified by loss of b-galactosidase staining, PCR and quantitative PCR, protocol as in Tomic et al., 2018. Details of primers and plasmid construction can be found in Sup. Table 2 and single copy confirmation of Snail in Sup. Fig. 2. The location of the frt site in DLD-1 Flp-In T-Rex cells was identified using the Universal GenomeWaler 2.0 Kit (Clontech) and following standard protocols. Snail^hi: protocol followed from Siemens et al., 2011.

Gene expression was measured by qPCR. The following Taqman probes were used: E-cadherin (Hs01023894 m1), N-cadherin (Hs00983056 m1), RPL19 (Hs02338565 gH), Snail (Hs00195591 m1) and vimentin (Hs00185584 m1) (all Thermo Fisher Scientific). Technical and biological repeats were both run in triplicate. Plates were run on a QuantStudio 6 Flex Real-Time PCR System (Thermo Fisher Scientific) with the following programme: 95 °C 20 s then 40 cycles of 95 °C 3 s and 60 °C 30 s. Data was analysed using the Ct method normalised to RPL19.

A multiplex qPCR of 84 EMT-associated genes was conducted using the RT² Profiler PCR Array Human Epithelial to Mesenchymal Transition (QIAGEN) according to the manufacturer’s instructions.

Cells that had been incubated with doxycycline for 14 days were plated at 90% confluence in Essen Image-Lock 24-well plates. A 1 mm scratch was drawn across the centre of the well using 10ul pipette tips and a wound maker. Plates were loaded into an IncuCyte ZOOM System inside a 37 °C incubator. Images of the wound were taken every 3 h and measurements calculated by the system’s software.

Cells were cultured in serum levels stated for each experiment for 24 h then 2 × 10⁵ were seeded in triplicate into the top chamber of 6.5 mm Corning Transwells with 8 μM pores. Transwells were placed in 24-well plates filled with stated serum levels in each experiment and incubated at 37 °C for 24 h. Media was aspirated from the plate and the upper membrane wiped with water and cotton to remove cells that had not migrated. Transwells were fixed in methanol (Sigma-Aldrich) for 10 min, incubated in Giemsa Stain (Fluka Analytical) for 10 min, washed several times in water then left to dry at 25 °C for 24 h. The membrane was excised and fixed between a slide and a cover slip. The number of visible nuclei in three fields at a 40 × objective on a Nikon Eclipse 50i microscope was counted and averaged.

Single cell suspensions (4 × 10⁴) were seeded onto 8-well μ-slides (ibidi) and fixed with 4% paraformaldehyde at room temperature for 15 min. Antibodies used were as follows: E-cadherin (BD Transduction Laboratories, 1:100) with AlexaFluor 647 monkey anti-mouse IgG (Invitrogen 1:500); Snail (Santa Cruz SC-10433, 1:50) with AlexaFluor 555 donkey anti-goat 1gG (Invitrogen, 1:500); Vimentin (Sigma C9050, 1:400), and DAPI (Invitrogen, final concentration 300 nM).

Cells were lysed with Laemmli buffer for protein collection then 33 μg total protein was loaded into each well of a 4–12% Bis–Tris gel (Thermo Fisher Scientific). Antibodies used were as follows: E-cadherin (Abcam ab1416, 1:1000) with IRDye 800CW goat anti-mouse (Li-Cor Biosciences, 1:5000); GAPDH (Sigma G8795, 1:1000) with IRDye 800CW goat anti-mouse (Li-Cor Biosciences, 1:5000); N-cadherin (Cell Signaling D4R1H, 1:1000) with IRDye 680LT goat anti-rabbit (Li-Cor Biosciences, 1:20,000); Snail (Cell Signaling C15D3, 1:1000) with IRDye 680LT goat anti-rabbit (Li-Cor Biosciences, 1:20,000); and Vimentin (Cell Signaling D21H3, 1:1000) with IRDye 680LT goat anti-rabbit (Li-Cor Biosciences, 1:20,000).

Xenografts were incubated in 4% paraformaldehyde for 24 h at 4 °C then remained in 70% ethanol until processed for paraffin wax embedding by the in-house histology core facility. Antibodies used were as follows: β-catenin (BD Biosciences 610,154, 1:100); cleaved caspase-3 (Cell Signaling Technology 9664, 1:200); E-cadherin (Daki M3612, 1:25); human Ki67 (Dako M7240, 1:400); Vimentin (Novocastra NCL-L-VIM-572, 1:400); and secondary antibodies from the Bond Polymer Refine Kit.

Cells were maintained in DOX or dimethyl sulfoxide (DMSO) for 14 days then 3 × 10⁴ cells per well were seeded onto 24-well plates and loaded into an IncuCyte ZOOM System inside a 37 °C incubator, and maintained in DOX or DMSO. After 24 h, the media was removed and replaced with oxaliplatin or 5-fluoruracil, still in DOX- or DMSO-rich media. Images of the wells were taken every 3 h and the confluence measured at each time point as calculated by the system’s software.

Phase contrast images of live cells were taken on a Zeiss Axiovert Microscope. Images of tumour histology were taken on a Nikon Eclipse 50i microscope. Immunofluorescent images of cells were taken on a Leica Tandem Confocal microscope using LAS-AF software.

All mouse work was conducted in line with UK Home Office regulations. Mice were kept on a 12-h light–dark cycle in individually ventilated cages with free access to food and drink. Standard diet (5R58, LabDiet) or doxycycline supplemented diet (200 ppm, 5058, LabDiet) was provided, as per requirements of the experiment. Mice were checked once per day by staff for signs of disease progression affecting eating, breathing or locomotion. All mice used were non-obese diabetic severe combined immune deficiency interleukin 2 receptor-null (NSG) bred by Alexandra Bruna (in-house, CRUK-CI). No animals were excluded from the final analysis.

Cells were counted on a Vi-Cell counter (Beckmann Coulter) and resuspended in 1:1 growth factor reduced matrigel (Corning) and cell culture media at a concentration of 10⁵/ml. Under anaesthetic, mice were injected with 100 μl cell suspension in each flank. Mice were alternately allocated to each group, ensuring equal numbers of mice in each group in each cage. Twice weekly measurements were taken of total tumour mass with electronic scales and tumour size with digital callipers by a single individual blinded to group allocation.

Significance values for Student’s t-test, analysis of variance (ANOVA) and Kruskal–Wallis test were calculated on statistical package GraphPad Prism. Comparison of the means between multiple samples was then conducted by Tukey’s multiple comparisons test for ANOVA and Dunn’s test for Kruskal- Wallis tests. Significance of non-linear fit, comparing the differences in gap closure in a scratch wound assay, was calculated on GraphPad Prism. Statistical analysis of the multiplex PCR was calculated in the PCR Array Data Analysis Software (QIAGEN). Significance of the rates of tumour growth were based on the linear regression of interaction between conditions calculated using the programme R. This compares the likelihood that two data sets follow the same line of best fit.

To produce a model of EMT that is mediated by levels of Snail overexpression comparable to those that arise spontaneously in cancer cell lines, a single copy expression cassette of human SNAI gene (Snail) that was doxycycline-inducible was engineered into a defined, open, genomic region in DLD-1 cells using frt recombination (Sup. Fig. 1) [31]. This region was identified by PCR-based genome walking as the final intron of ARID5B on chromosome 10. Targeted single-copy insertion of Snail and vector-only cassettes was confirmed by PCR, qPCR copy number assay, and ablation of lacZ staining (Sup. Fig. 2). For comparison, Snail was also introduced into DLD-1 cells using a pRTR episome that has been previously published, though here the human coding sequence was substituted for the murine one (Sup. Fig. 1) [26].

The level of Snail RNA expression was consistent in the single copy Snail line, with an all-or-nothing response to escalating levels of doxycycline (DOX) induction contrasting to dose-dependent responses in episomal-induced Snail cells (Figs. 1a, b). Snail RNA remained near absent expression in the vector-only line (Snail^wt). When single or episomal Snail cells were cultured with 100 ng/mL DOX, the level of Snail RNA expression in the single-copy Snail line (Snail^lo) was comparable to other Snail-expressing colon cancer cell lines, whereas the episomal–Snail line (Snail^hi) expression was 75-fold higher (Fig. 1c).

Snail expression and changes in the transcriptional abundance of responsive genes (E-cadherin and N-cadherin) in Snail^lo cells remained similar from 5 to 14 days of DOX induction indicating a stable maintenance of phenotype over this time (Sup. Fig. 3). Subsequently, Snail^lo and Snail^wt lines were cultured for 14 days in DOX for use in functional assays, using dimethyl sulfoxide (DMSO) as a solvent. Of note Snail^lo and Snail^wt lines could be cultured in DOX for weeks without an effect on survival (data not shown). In contrast, Snail^hi lines could only be cultured for 5 days before mass cell death commenced. (Previous published work with Snail^hi used 100 ng/mL DOX and data obtained after 4 to 5 days [14]). To enable comparison of models all cells were incubated in 100 ng/mL DOX but analysed at 4 days for Snail^hi and 14 days for Snail^lo and Snail^wt.

DLD-1 cells grow with a distinctive cobblestone appearance typical of epithelial cells. Induction of Snail expression resulted in cells being less rounded and more mesenchymal in appearance and with membrane protrusions. These morphological changes towards a mesenchymal phenotype were more pronounced in Snail^hi compared to Snail^lo expressing cultures (Fig. 2a).

Snail^hi lines were confirmed to express more Snail protein compared to Snail^lo, as shown by immunofluorescence and Western Blot (Figs. 2a, b). Both lines showed a decrease in epithelial E-cadherin protein. On a population basis this decrease was greater in Snail^lo than Snail^hi cells as shown by Western blotting but E-cadherin was completely absent only in a proportion of Snail^hi cells as shown by immunofluorescence. This is likely explained by the different incubation times: 4 days for Snail^hi and 14 for Snail^lo. There were no changes at the protein level of mesenchymal markers N-cadherin or fibronectin in either line. Snail^hi cells showed increased vimentin expression by immunofluorescence.

A commercially available mRNA array profiler was used to assess the expression of 84 genes associated with EMT (see Methods). There were widespread decreases of epithelial markers and increases in mesenchymal markers in both cell lines on DOX induction of Snail (Figs. 2c, d, full results Sup. Table 1). Of 84 genes known to be associated with changes in morphology, motility and stemness, the expression of 35 were significantly different between Snail-expressing lines compared to Snail^wt controls. In both Snail^lo and Snail^hi lines there tended to be perturbation of expression in the same genes and with the same trend for up and downregulation. This suggests that the same EMT programme is being exploited. Thus Snail-expression led to a downregulation of genes that encode proteins in cellular adhesion junctions (OCLN, DSC2, CDH1), the epithelial cytoskeleton and its interactors (KRT7, KRT19, PLEK2), cell proliferation (ERBB3, WNT5B) and immunity (IL1RN). There was an upregulation of EMT-TFs (ZEB1, SNAI1) and genes that encode proteins with effects on invasion (ITGAV), mesenchymal cytoskeleton (VIM), cell polarity (WNT5A) and rearrangement of the surrounding matrix (COL3A1, MMP3, SERPINE1). However, a striking difference between the two lines was that decreases in some epithelial markers (OCLN, KRT19, PLEK2) were stronger in Snail^lo and increases in some mesenchymal markers (ITGA5, COL5A2, TIMEFF1, VIM, SERPINE1, SNAI1) were greater in Snail^hi with the latter being particularly pronounced.

Overall EMT-related gene expression indicate that Snail^lo is at least equally efficacious as Snail^hi in downregulating epithelial differentiation but that driving a full subsequent mesenchymal differentiation requires the higher levels of Snail expression associated with Snail^hi.

The functional effect of Snail upregulation on Snail^lo cells was measured by migration and chemoresistance. Snail^hi cells died by day 5 of DOX incubation. As DOX treatment itself is well tolerated by Snail^wt and Snail^lo cells this can be ascribed to high levels of Snail expression being lethal.

Time-lapse footage in a scratch-wound assay showed Snail^wt cells moved collectively, as a sheet, with a straight moving edge, whereas Snail^lo cells displayed individual cell migration (Fig. 3a). The moving edge was less uniform and more jagged, with cells displaying mesenchymal-like membrane protrusions. Difference in migratory capacity was confirmed in a Transwell assay. Significantly more Snail^lo cells moved across an 8 µm-pore-lined membrane in the presence of DOX than without (Fig. 3b, c). This difference was consistent, regardless of serum gradient across the membrane, discounting a chemotactic effect arising with the experimental design (Sup. Fig. 4).

Snail^lo cells showed increased chemoresistance, with higher levels of growth in the presence of oxaliplatin and of 5-fluorouracil, chemotherapeutic agents usually used to treat colorectal cancer (Fig. 3d). This difference in chemoresistance was only evident in moderate drug levels; very high or very low doses showed no difference in growth when compared to controls.

To assess the effect of single-copy Snail upregulation in vivo, Snail^lo and Snail^wt cells were subcutaneously injected into the posterior flanks of NSG mice. All cells were injected in the epithelial, uninduced state, then all mice were started on DOX-chow 7 days later. This was to ensure differences in growth rate were not due to seeding or initial cell survival. All injections produced a tumour (96/96).

Snail^lo tumours displayed a partial EMT in vivo, with reduced membranous E-cadherin but no increased vimentin on histology (Fig. 4a). This effect seemed independent of ß-catenin, which remained membrane-bound.

Assessment of tumour size, measured using callipers, showed that Snail^lo tumours were significantly larger and grew at a faster rate, when compared to Snail^wt tumours (Fig. 4b). This difference in tumour size could not be attributed to proliferation, as both sets of tumours had equivalent levels of Ki-67 staining (Fig. 4a). However, Snail^lo tumours showed reduced cleaved caspase-3 staining, a marker for apoptosis, and significantly increased expression of the anti-apoptotic protein Bcl-2 (Fig. 4c). This suggests Snail^lo tumours grow faster and larger due to anti-apoptotic mechanisms, rather than increased cell proliferation.