Fig 2: Cytoplasmic LINC00238 sponges miR-522. (A) The bar graphs show that LINC00238 was predominantly localized in cytoplasmic Huh7 and HepG2 cells, as calculated via reverse transcription-quantitative PCR. Cytoplasmic control (GAPDH) and nuclear control (U6) were determined in their expected localization. (B) The expression of miR-522 was downregulated in LINC00238-overexpressing Huh7 cells, but upregulated in LINC00238-knockdown HepG2 cells. (C) Schematic model indicating that three variants of LINC00238 contain the target region of miR-522. (D,E) In vivo and in vitro RNA pull-down assays indicated the interaction between LINC00238 and miR-522. (F) Complementary pairing and corresponding mutation between the sequences of miR-522 and LINC00238. (G) Compared with NC, the miR-522 mimic inhibited the relative fluorescence activity of the LINC00238 wild-type PGL3-control plasmids, but not the mutant type. (H) Western blotting detected that alterations in two known miR-522 targets (SFRP2 and DKK1) were upregulated in LINC00238-overexpressing Huh7 cells and downregulated in LINC00238-knockdown HepG2 cells. miR, microRNA; NC, negative control; SFRP2, secreted frizzled related protein 2; DKK1, dickkopf1; shRNA, short hairpin RNA.

Fig 3: The DKK1 RNAscope assay in tumor resections. Representative images from 3 tumor resections with no (top row), moderate (middle row) and high (bottom row) DKK1 signal are shown with the PPIB control for RNA integrity and the dapB background control. DKK1 H-scores (range 0–300) were semi-quantified manually as described in the methods. Black arrows denotes non tumoral cells without DKK1 signal. Blue arrows denote tumor cells with low levels of DKK1 signal (one dot per cell). Scale bar: 50 µm.

Fig 4: Specificity and accuracy of the DKK1 RNAscope assay. (a) Cell lines were identified using RNA-Seq data from the Cancer Cell Line Encyclopedia database that expressed very high levels of the indicated Dickkopf family member and low levels of DKK1. Red boxes denote the expression level of the Dickkopf family member that is highly expressed in the indicated cell line. Reads per kilobase per million mapped reads (RPKM). (b) The specificity FFPE cell pellet array with the indicated cell lines was stained for DKK1, quantified using QuPath morphometric software and an H-score (range 0–300) was calculated as described in the methods. Scale bar: 50 µm. (c) FFPE cell pellet arrays with different cancer cell lines were stained for DKK1, quantified using QuPath morphometric software and an H-score (range 0–300) was calculated as described in the methods. DKK1 RNA-Seq data was obtained from the Cancer Cell Line Encyclopedia database. Reads per kilobase per million mapped reads (RPKM). (d) The indicated cells were stained for DKK1 RNA by RNAscope (top row), DKK1 protein by IHC (middle row) or an isotype control antibody (bottom row). Scale bar: 50 µm.

Fig 5: CSC colony formation in bone marrow is mediated by Wnt signalling. a Treatment of MCF-7 cells with CM in vitro, or intra-femoral growth in vivo increased expression of AXIN2 compared to cells treated with control media in vitro. b Addition of 100 nM recombinant Wnt (rWnt) to control media replicated the effect of CM in increasing mammosphere formation in MCF-7 and MDA-MB-231_BH cells. Adding Wnt to CM did not further increase mammosphere formation (bars labelled as CM + Wnt). c Inhibition of Wnt signalling with 50 ng/ml DKK1 or 50 µg/ml Vantictumab reversed induction of mammosphere formation by CM in early breast cancer samples in vitro (n = 10). d There was no significant difference in tumour formation between control and Vantictumab treated mice following intra-femoral injection of MCF-7 cells. e Cells isolated from femurs of Vantictumab treated mice formed significantly fewer mammsopheres ex vivo than cells isolated from control mice. Representative photomicrographs shows brightfield and tdTomato images of mammospheres. Scale bar 150 µM. f Survival analysis based on DKK1 expression (Low (orange) vs High (purple)) in 2 published datasets (NKI, Cb560) in patients with bone metastases. Low DKK1 expression in primary breast cancers was associated with poor outcomes in patients with bone metastasis. (NKI; significant cut points P < 0.05 92/244 (38%), Cb560; significant cut points P < 0.05 170/420 (40%)). HR; Hazard ratio. BMFS; Bone metastasis free survival. g Knockdown of DKK1 with 25 nM siRNA sensitised MDA-MB-231 cells to CM, resulting in induction of mammosphere formation. h DKK1 gene expression is decreased in MDA-MB-231_BH cells compared to the parental MDA-MB-231 cells. Data presented as fold change in gene expression in MDA-MB-231_BH cells compared to parental cells. All graphs represent mean ± SEM, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001