br these pathways by immunoblot analysis after BCF treatment
these pathways by immunoblot analysis after BCF treatment, and found that phospho-p38 was indeed significantly increased upon BCF treatment, suggesting that calcium influx activates stress response in BCF-treated cells (Fig. 3K). However, no significant change was ob-served in JNK expression (Fig. 3K). We also found that BCF, but not RCF, activated CAMKII, a known upstream kinase of p38 that is activated by calcium influx  (Fig. 3K and supplementary fig. 2D). When cells were treated with BCF in the presence of CAMKII inhibitor, KN93, p38 activation was rescued, suggesting that CAMKII is indeed the upstream kinase for p38 activation by BCF (Fig. 3L). Furthermore, BCF induced cell cycle arrest and elevated phosphorylation of the alpha subunit of trans-lation initiation factor EIF2 (Fig. 3M and Supplementary Fig. 2E). We then examined the functional consequence of p38 MAPK activation in TAK-242 metastasis by Gene Set Enrichment Analysis (GSEA). p38 MAPK signature was found to be highly enriched in patients without metasta-sis when compared to patients with brain metastasis (Fig. 3N). To fur-ther validate this clinical finding, we expressed wild type and constitutively active p38 (D176A) in SKBrM3 cells (Supplementary Fig. 2F). As shown in Fig. 3O, stable expression of both wild type and ac-tive p38 strongly suppressed the cell growth of SKBrM3 cells. To exam-ine whether p38MAPK pathway is activated by calcium influx, we treated cells with Bay K8644, an agonist of L-type VGCC, for 3 h/day for 7 days. Bay K8644 treated cells strongly activated p38 MAPK and re-duced cell proliferation of SKBrM3 cells, which recapitulates the effect of BCF (Supplementary Fig. 2G and H). Collectively, these results suggest that BCF augments calcium influx through Cav3.2 channel and activates the p38 pathway to suppress cell growth.
3.3. BCF suppress cancer stem cell (CSC) through HMGA2
To further gain insight into the mechanistic action of BCF, we per-formed expression profiling analysis to identify differentially expressed genes following BCF treatment. As shown in Fig. 4A, nine genes were differentially up- or down-regulated commonly in both cell lines by the BCF treatment. Moreover, these genes downregulated by BCF were highly enriched in patients with brain-metastasis compared to the co-hort of patient with no brain metastasis when examined by GSEA (Sup-plementary Fig. 3A). We then performed secondary screening of these 9 genes by analysing a cohort of 710 patients and examined their associ-ation with relapse-free survival (Fig. 4B and Supplementary Fig. 3B). Based on this criterion, the HMGA2 gene was selected for further inves-tigation. Correlation of higher HMGA2 expression to worse outcome in breast cancer patients was further evident in METABRIC dataset (Sup-plementary Fig. 3C). Suppression of HMGA2 expression by BCF in vitro and in vivo was validated by qRT-PCR, western blot analysis and immu-nohistochemistry (Fig. 4C-E). Interestingly, HMGA2 level was found to be higher in patients with brain metastasis compared to patients with no metastatic disease or lung metastasis (Fig. 4F). Similarly, brain met-astatic variants of breast cancer cells expressed higher level of HMGA2 than their parental cell lines, suggesting that the functional role of HMGA2 is linked to brain metastasis (Fig. 4G). HMGA2 is a transcription
Fig. 3. The inhibitory effect of BCF is mediated through Cav3.2 T-type channel and CAMKII/p38 MAPK pathway. (A) 231BrM and SKBrM3 cells were treated with Sham or random frequencies (RCF) or BCF for 7 days followed by thymidine incorporation cell proliferatin assay (n = 5/group). (B) SKBrM3 cells were treated with 1 h, 3 h and 6 h-per day for 7 days and cell proliferation was examined at day 7 by thymidine incorporation assay. (C) Various cell lines were seeded on 96-well plates at day 7 after Sham or BCF treatment, and cell proliferation was examined at day 1, 3 and 5 by MTS assay (n = 8/group). (D) 231BrM and SKBrM3 cells were exposed to Sham or BCF daily for 7 days in the presence of ethosuximide or vehicle, and cell proliferation was quantified by thymidine incorporation at day 7 (n = 5/group). Results are normalised to Sham group. (E-F) The T-type voltage-gated calcium channel subunit genes, Cav3.1, Cav3.2 or Cav3.3, were knocked down in 231BrM or SKBrM3 cells by shRNAs, and they were treated with BCF for 7 days followed by quantifying cell proliferation at day 7. (G) SKBrM3shCav3.2 cells were intracardially injected to NOD/SCID mice. At Day 30, ex vivo tumour signal in brain was quantified by bioluminescence. Right panel shows representative brain images. (H) Cells treated with Sham or BCF were stained with Fluo-4 calcium dye and the level of intracellular calcium level was examined using flow cytometry. The representative histogram is shown. (I) SKBrM3 cells treated with Sham or BCF in the presence of vehicle or ethosuximide were stained with Fluo-4-am dye and cytoplasmic calcium level was quantified by flow cytometry (n = 6/group). (J) SKBrM3 cells cultured in media with or without calcium and examined for cytoplasmic calcium as in I. (K) The expression of total and activated p38 and phosphorylated JNK was examined in 231BrM cells treated with Sham or BCF for 7 days by western blot. α-tubulin was used as a loading control. (L) 231-BrM cells were treated with Sham or BCF or BCF in the presence of KN93 (5 μM) and protein levels of phospho-CAMKII, total CAMKII, phopho-p38 and total p38 were examined by western blot. (M) SKBrM3 cells treated with Sham or BCF for 7 days were subjected to cell cycle analysis using FACS. % of population in G1, S and G2 phase are shown (n = 5/group). (N) Enrichment of biocarta P38MAPK signature in breast cancer patients with or without brain metastasis incidence was analysed by GSEA. (O) p38WT or p38 D176A (active mutation) gene or vector control was ectopically expressed in SKBrM3 cells using retrovirus and expression, and cells were seeded on 96-well plate (n = 500/well) and subjected to cell proliferation assay using MTS reagent at Day 5 (n = 8/group). *, P-valueb.05, ** P-valueb.01 and ***, P-valueb.0001.