Fig 2: The effect of addition or inhibition of TGF-β and Hsp90 on adhesion in the paired SW480 and SW620 colon cancer cell lines. SW480 a and SW620 b cells were treated with 2 ng/ml TGF-β1, 20 ng/ml Hsp90β, 100 nM SB431542, 100 μM novobiocin and 10 μg/ml αvβ6 integrin blocking antibody (anti- αvβ6) singly or in combinations as indicated on the x-axis. Absorbance at 590 nm of crystal violet stained adherent cells was normalized to the untreated control, given as 100%, and depicted by the solid horizontal line in the graph, showing the changes in cell adhesion for each treatment. All graphs are representative of the average of three independent experiments performed in triplicate and error bars indicate the standard error in the mean. Statistical analysis was performed using GraphPad Prism 4.03 software. A two-way analysis of variance (ANOVA) with Bonferroni post-test was performed and significance between untreated cells and those after each treatment (indicated by asterisks) and between particular treatments (indicated by hashes) are shown (*p < 0.05, **p < 0.01, ***p < 0.001, # p < 0.05, ### p < 0.001, ns – not significant)

Fig 3: The effect of addition or inhibition of TGF-β and Hsp90 on migration of the paired SW480 and SW620 colon cancer cell lines. SW480 a and SW620 b cells were treated with 2 ng/ml TGF-β1, 20 ng/ml Hsp90β, 100 nM SB431542, 100 μM novobiocin and 10 μg/ml αvβ6 integrin blocking antibody (anti- αvβ6) singly or in combinations as indicated on the x-axis and the effect of such treatment on migration assessed. All graphs are representative of the average of three independent experiments performed in triplicate and error bars indicate the standard error in the mean. Statistical analysis was performed using GraphPad Prism 4.03 software. A two-way analysis of variance (ANOVA) with Bonferroni post-test was performed and significance between untreated cells and those after each treatment (indicated by asterisks) and between particular treatments (indicated by hashes) are shown (*p < 0.05, **p < 0.01, ***p < 0.001, ## p < 0.01, ### p < 0.001, ns – not significant)

Fig 4: Investigating the status of the canonical TGF-β pathway and levels of Hsp90 in the genetically paired SW480 primary and SW620 secondary tumour-derived colon cancer cell lines. a The secretion of TGF-β1 by SW480 and SW620 cells was determined by means of an ELISA to quantify the levels of extracellular TGF-β1 in spent culture medium for a given cell number after 12 h incubation. Data represent the average of three independent experiments performed in triplicate and error bars indicate the standard error in the mean. b The levels of TGF-βRII receptor on the cell surface was determined by flow cytometry using a fluorescein isothiocyanate-conjugated antibody. Data was analysed using FlowJo software and gating carried out according to the IgG1 isotype control. Histograms show a shift in fluorescence for each sample (black line, unshaded) compared to the isotype control (grey shading) and the percentage of positive events is indicated for each sample. Data are representative of three independent experiments showing consistent results. c Intracellular levels of TGF-β1 in SW480 and SW620 cells were determined by (i) western blot analysis and compared to a histone loading control. The images are cropped from the full length western blot provided as Additional file 1. (ii) The normalized levels of pre-pro TGFβ1 from the western blot in (i) were determined by densitometry using ImageJ and calculated according the histone loading control. AU: arbitrary units. The positive control represents commercially obtained recombinant human TGFβ1 in its active form. d Activation of the TGF-β canonical pathway was determined by the phosphorylation and nuclear localization of Smad2/3. SW480 and SW620 cells were stained for pSMAD2/3 and nuclei were stained with Hoescht-33342. Immunofluorescence was detected using the Zeiss LSM 510 Meta laser scanning confocal microscope and the images analysed using Axiovision LE 4.7.1 software (Zeiss). (i) The first column shows the level of pSmad2/3 staining in the cells, pseudocoloured to white, while the second column shows a merged image of the nucleus pseudo-coloured to red and the pSMAD2/3 staining pseudocoloured to green. The third column shows frequency scattergrams obtained using colocalisation analysis on Image J,. Scale bars represent 20 μm. (ii) Graphs generated from profiles of the nucleus and pSMAD2/3 within individual cells analysed using Zen Lite Software showing the proportion of total pSMAD2/3 that is found in the nucleus. In all cases, analyses were performed on triplicate images containing at least 10 cells per image e The secretion of Hsp90β by SW480 and SW620 cells was quantified by means of an ELISA on spent culture medium after 12 h incubation. The data are representative of two individual experiments performed in triplicate and showing consistent results. f Whole cell lysates from SW480 and SW620 cells were analysed for Hsp90α and Hsp90β expression using western blot analysis. Where relevant, data was analysed using GraphPad Prism 4.03 software with errors bars indicating the standard error in the mean and a students t-test was performed to assess statistical significance. (*p < 0.05, **p < 0.01)

Fig 5: Effect of pre-treatment of paired colon cancer cell lines by the addition or inhibition of TGF-β or Hsp90 under anchorage independent conditions on subsequent sensitivity to colon chemotherapeutics. SW480 (i) and SW620 (ii) cells were treated with either 2 ng/ml TGF-β1, 20 ng/ml Hsp90β, 100 nM SB431542 or 100 μM novobiocin upon seeding in a tumoursphere assay. After 7 days, tumourspheres were reseeded into regular adherent growth conditions and treated with 75 μM 5-fluorouracil a or 550 μM oxaliplatin b for 72 h. Cell viability was assessed compared to an untreated control for each pre-treatment using a MTT Cell Proliferation kit. Data are representative of two individual experiments carried out in triplicate and showing consistent results. Statistical significance was assessed by means of a two-way analysis of variance (ANOVA) with Bonferroni post-tests relative to the non-pre-treated control using GraphPad prism where n = 3 (*p < 0.05, **p < 0.01)