Breast cancer stem cells tolerate chromosomal instability during tumor progression via c-Jun/AXL stress signaling

Abstract

Chromosomal instability (CIN) is critical for tumor evolution, yet its relationship with stemness is unclear. Here, we describe CIN as a key stress induced during tumor initiation that is uniquely tolerated by breast cancer stem cells in an activated signaling state (aCSCs). While we noted elevated CIN specifically in tumors from aCSCs, this was not intrinsic to these cells, as baseline levels were similar to non-stem cell types. This suggests that CIN is induced during tumor initiation, and that aCSCs can better tolerate this stress. Further, this increased CIN may be transient, as it was only in low-burden aCSC tumors, with levels diminishing in more established disease. Phospho-array profiling revealed specific activation of c-Jun stress signaling in aCSCs, which we hypothesized could induce genes responsible for CIN tolerance. Indeed, we identified AXL as a c-Jun dependent gene enriched in aCSCs that enhances resistance to this stress. Thus, CIN tolerance mediated by c-Jun/AXL signaling may be a defining feature of stemness, contributing to breast cancer progression.

Keywords: Breast cancer; Chromosomal missegregation; Stem cell activation; Stress tolerance; Tumor initiation.

Fig. 1

CIN is a bottleneck to tumor initiation uniquely tolerated by aCSCs. (A) Representative H&E images of anaphase cells in tumor xenografts initiated by aCSCs. Examples are displayed of normal anaphase (top) and chromosomal missegregations including anaphase bridges (middle) and lagging strands (bottom). Scale bar, 5 μm. (B) Quantitation of the missegregation frequency in tumors formed from limiting numbers of sorted HCC38 cells. Data from aCSCs (EpCAM^Low/αvβ3⁺) was separated into small (<100 mm3) versus large tumors (>100 mm3). P-values relative to Small aCSC tumors (n = 3): Large aCSC (n = 7), P = 0.0379; EpCAM^Low/αvβ3^- (n = 4), P = 0.0255; EpCAM^High/αvβ3⁺ (n = 6), P = 0.102; EpCAM^High/αvβ3^- (n = 4), P = 0.0325. >50 total anaphase cells assessed per tumor type. *P < 0.05. n. s. = not significant. (C and D) CIN gene signature scores for freshly sorted HCC38 cells from culture (C) or breast cancer cell lines representing different molecular subtypes (D). (D) Center line represents the mean with each dot indicating a different cell line. Cells in each category: Luminal B; n = 7, HER2; n = 7, Basal-like; n = 6, Claudin-low; n = 8. (E and F) Assessment of chromosomal missegregations by DAPI DNA staining. (E) Representative images of LM2-4 cancer cells in normal anaphase (top) or with an anaphase bridge (bottom). Scale bar, 10 μm. (F) Frequency of missegregations in breast cancer cell lines. *P < 0.05 for MCF-7 and T47D versus MDA-MB-468, BT549 and MDA-MB-231 cells. For each cell line, 150 total anaphase cells were analyzed from n = 3 independent experiments. (B-D and F) Data represent the mean ± s. e.m. Statistics by one-way ANOVA and Holm-Sidak multiple comparisons test. See also Supplementary Fig. S1.

Fig. 2

Enhanced c-Jun/JNK stress signaling in aCSCs. (A) Phospho-array profiling of sorted HCC38 cell types showing enrichment for pS63 c-Jun (red ovals) in aCSCs (EpCAM^Low/αvβ3⁺). Membranes contain duplicate dots per target phospho-protein. (B and C) Representative immunoblots confirming enrichment for pS63 c-Jun relative to total c-Jun in sorted aCSCs from HCC38 cells (B) as well as stem-like breast cancer cell lines (C). Hsp90 is a loading control and molecular weight markers are indicated in kilodaltons. (D) HCC38 methylcellulose tumorspheres after treatment with two different JNK kinase inhibitors. Data represent the mean ± s. e.m. Statistics by one-way ANOVA and Tukey's multiple comparisons test. *P < 0.05 relative to vehicle control. (B-D) n = 3 independent experiments. See also Supplementary Fig. S2.

Fig. 3

AXL expression is induced by c-Jun/JNK stress signaling. (A) TCGA analysis comparing established c-Jun transcriptional targets and CIN (fraction of genome altered) in patient breast cancers (invasive carcinoma). Spearman correlation coefficients were determined for each gene using cBioportal and data from the PanCancer Atlas data set (1066 patients). (B) Schematic describing the selection criteria for AXL with the indicated gene sets. (C) QPCR validation of AXL mRNA expression in sorted HCC38 cells types. Samples were run in duplicate with GAPDH as a loading control. Expression is shown relative to the EpCAM^Low/αvβ3^- cells (dashed line). Data represent the mean ± s. e.m. (D and E) Representative immunoblots for AXL protein expression in sorted HCC38 cells types (D) or breast cancer cell lines (E). β-actin or Hsp90 are loading controls and molecular weight markers are indicated in kilodaltons. (C-E) n = 3 independent experiments. See also Supplementary Fig. S3.

Fig. 4

A c-Jun/AXL signaling module is required for CIN-tolerance in aCSCs. (A) QPCR for AXL mRNA after treatment of LM2-4 cells with JNK inhibitors for 24 h. All samples were run in triplicate with β-actin used as a loading control. Data represent the mean ± s. e.m. (B) AXL immunoblots of surrogate stem-like cell lines after treatment with DMSO (Veh) or 5 μM of the JNK inhibitor (JNKi) IN-8 for 24 h Hsp90 is a loading control and molecular weight markers are indicated in kilodaltons. (C and D) Methylcellulose tumorsphere assays with control or AXL KO LM2-4 and BT549 cells treated with Reversine to induce CIN. (C) Representative images showing differential response to CIN-induced stress in control and AXL KO LM2-4 cells. Scale bar, 200 μm. (D) Statistics by two-way ANOVA and Tukey's multiple comparisons test. *P < 0.05. (A and D) Data represent the mean ± s. e.m. n = 3 (A, B and D; BT549) or n = 5 (C and D; LM2-4) independent experiments. See also Supplementary Fig. S4.