A pre-existing population of ZEB2+ quiescent cells with stemness and mesenchymal features dictate chemoresistance in colorectal cancer

Abstract

Background: Quiescent/slow cycling cells have been identified in several tumors and correlated with therapy resistance. However, the features of chemoresistant populations and the molecular factors linking quiescence to chemoresistance are largely unknown.

Methods: A population of chemoresistant quiescent/slow cycling cells was isolated through PKH26 staining (which allows to separate cells on the basis of their proliferation rate) from colorectal cancer (CRC) xenografts and subjected to global gene expression and pathway activation analyses. Factors expressed by the quiescent/slow cycling population were analyzed through lentiviral overexpression approaches for their ability to induce a dormant chemoresistant state both in vitro and in mouse xenografts. The correlation between quiescence-associated factors, CRC consensus molecular subtype and cancer prognosis was analyzed in large patient datasets.

Results: Untreated colorectal tumors contain a population of quiescent/slow cycling cells with stem cell features (quiescent cancer stem cells, QCSCs) characterized by a predetermined mesenchymal-like chemoresistant phenotype. QCSCs expressed increased levels of ZEB2, a transcription factor involved in stem cell plasticity and epithelial-mesenchymal transition (EMT), and of antiapototic factors pCRAF and pASK1. ZEB2 overexpression upregulated pCRAF/pASK1 levels resulting in increased chemoresistance, enrichment of cells with stemness/EMT traits and proliferative slowdown of tumor xenografts. In parallel, chemotherapy treatment of tumor xenografts induced the prevalence of QCSCs with a stemness/EMT phenotype and activation of the ZEB2/pCRAF/pASK1 axis, resulting in a chemotherapy-unresponsive state. In CRC patients, increased ZEB2 levels correlated with worse relapse-free survival and were strongly associated to the consensus molecular subtype 4 (CMS4) characterized by dismal prognosis, decreased proliferative rates and upregulation of EMT genes.

Conclusions: These results show that chemotherapy-naive tumors contain a cell population characterized by a coordinated program of chemoresistance, quiescence, stemness and EMT. Such population becomes prevalent upon drug treatment and is responsible for chemotherapy resistance, thus representing a key target for more effective therapeutic approaches.

Keywords: Cancer stem cells; Chemotherapy resistance; Colorectal cancer; Dormancy; Epithelial-to-mesenchymal transition; Quiescence.

Fig. 1

Untreated tumors contain chemotherapy-resistant quiescent cells with an EMT/stemness phenotype and increased ZEB2 levels. a SW480 cells were stained with PKH26, treated with 2,5 μM oxaliplatin (OXA) at day 11 and monitored by flow cytometry. FACS plots are shown in Additional file 4. b FACS plots showing PKH26 positivity in CCSCs-derived tumors the day of injection (Day 0) and at 1/3/6 weeks. c Percentage of PKH26⁺ cells (black line) in relation to tumor size (red line). Mean ± SD or SEM (tumors), n = 6 tumors/group. d Spatial representation of principal component (PC) analysis with genes as rows (statistical units) and samples as columns (variables). n = 5 samples of 2 pooled tumors each. e Numerical PC plot identifying genes with the highest absolute score in the discriminant component. A full list of transcripts modulated in PKH26⁺ versus PKH26⁻ cells is reported in Additional file 5. f Categories of transcripts enriched in PKH26⁺versus PKH26⁻ cells. Replicated entries are reported in Materials and Methods. Transcripts categories are detailed in Additional file 6. g qRT-PCR analysis of ZEB2 in EpCAM⁺/PKH26⁺ versus EpCAM⁺/PKH26⁻ cells from CCSCs-derived xenografts. *P < 0.05 (two-tailed t test), mean ± SD, n = 3 pools of 6 tumors each. h qRT-PCR of ZEB2 expression in SW480 cells untreated (day 0) or treated with 2,5 μM oxaliplatin (OXA). Mean ± SD of 3 experiments. i Representative confocal image of CCSCs-derived xenograft sections showing overlapping areas of ZEB2 (red) and PKH26⁺ (yellow) positivity. Scale bar 80 μm. l qRT-PCR of xenograft-derived EpCAM⁺/PKH26⁺ versus EpCAM⁺/PKH26⁻ cells. *P < 0.05, **P < 0.01 and ***P < 0.001 (two-tailed t test). Mean ± SD, n = 3 pools of 6 CCSCs-derived tumors each. m qRT-PCR analysis of ZEB2 expression in TOP-GFP.mcherry negative (grey) and positive (purple) CCSCs sorted from in vitro culture. Mean ± SD of 3 experiments

Fig. 2

Reverse-phase proteomic analysis of quiescent/slow cycling xenograft cells. a Hierarchical clustering of RPPA results obtained on 3 EpCAM⁺/PKH26⁺ and EpCAM⁺/PKH26⁻ cell samples isolated from CCSCs-derived tumor xenografts. Clusters, identified for either antibodies or samples and based on optimal cut of dendrograms, are indicated by coloured bars adjacent to dendrograms. The values represented by the heatmap correspond to normalized intensities of antibodies, standardized over the sample set analyzed (z score). n = 3 pools of 12 tumors each. A list of RPPA antibodies and modulated endpoints are reported respectively in Additional file 1: Table S1 and Additional file 7: Table S6. b Principal component analysis (PCA) of RPPA results showing that PKH26⁺ samples have a common molecular signature. c Volcano plot showing the antilogarithm (base = 10) of the adjusted P value versus base 2 logarithm of the ratio between PKH26⁺ and PKH26⁻ samples. Kruskal Wallis test was performed for each RPPA analyte on the 3 samples stratified by PKH26 positivity. RPPA analytes where Kruskal Wallis test resulted in a statistically significant (*P < 0.05) change between PKH26-stratified samples, underwent a further analysis by means of Wilcoxon signed-rank test. All the resulting p values were adjusted for multiple comparisons using the Benjamini-Hochberg correction. d Schematic representation of pathways that emerged from experiments described in Figs.1 and 2 as present in PKH26-negative fast proliferating cells (left, red area) or in slow proliferating/quiescent cells (right, blue area). Phosphorylation sites are outlined in green when they result in protein activation, in red when they inhibit protein function. Activated pathways are highlighted in colors while inhibited pathways are depicted in light grey

Fig. 3

Coordinated expression and modulation of ZEB2, pCRAF and pASK1. a Left: immunoblot analysis of ZEB2, CRAF pS338, and ASK1 pS83 on whole lysates of SW480 cells treated for 4 days with 5 μM 5-fluorouracil (5-FU) or 2,5 μM oxaliplatin (OXA). Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used as a loading control. Right: quantification of immunoblot shown on the left. b qRT-PCR analysis of ZEB2 levels in CCSCs (left panel) and SW480 (right panel) 24 h after siRNA-mediated silencing of ZEB2. ***P < 0.001 from two-tailed t test. Data of qRT-PCR are the mean ± SD, n = 3. c Immunoblot analysis of ZEB2, CRAF pS338, and ASK1 pS83 on whole cell lysates 24 h upon siRNA-mediated silencing of ZEB2 in CCSCs (left panel) and SW480 (right panel). The respective quantifications are shown on the right. d qRT-PCR analysis of ZEB2 levels in CCSCs (left panel) and SW480 (right panel) transduced with empty pLenti-GFP (Vector) or with pLenti-GFP-ZEB2 (ZEB2) and sorted on the basis of GFP expression. ***P < 0.001 from two-tailed t test. Data are the mean ± SD, n = 3. e Immunoblot analysis of ZEB2, CRAF pS338, and ASK1 pS83 on whole lysates of CCSCs (left panels) and SW480 cells (right panels) transduced with pLenti-GFP (Vector) or with pLenti-GFP-ZEB2 (ZEB2) and sorted as above. The respective quantifications are shown on the right. f Viability of CCSCs (left) and SW480 (right) transduced with pLenti-GFP or pLenti-GFP-ZEB2, sorted on the basis of GFP expression and immediately treated for 48 h with 10 μM oxaliplatin (OXA) and 10 μM 5-fluorouracil (5-FU). *P < 0.05 from two-tailed t test, n = 3. Data are the mean ± SD of three independent experiments

Fig. 4

ZEB2 induces a transition towards a quiescent/slow cycling and mesechymal-like state in CRC xenografts. a Left: Xenograft volume of SW480 cells transduced with pLenti-GFP (Vector, black line/triangles) or with pLenti-GFP-ZEB2 (ZEB2, red line/squares). Mean ± SEM, 6 tumors/group. **P < 0.01 (two-tailed t test). Middle: representative confocal images of Vector- and GFP-ZEB2-transduced SW480 xenografts stained with anti-ZEB2 (red) and anti-Ki67 (white) antibodies. Scale bar 60 μm. Right: quantification of Ki67-, ZEB2- and GFP-positive cells performed on 3 sets composed of 5 fields/group. *P < 0.05 and **P < 0.01. Mean ± SD (two-tailed t test, n.s. = not significant). AU, arbitrary units. b Cell cycle analysis of GFP⁺ cells from Vector- and ZEB2-transduced tumors. c qRT-PCR analysis of GFP⁺ cells from Vector- and ZEB2-transduced tumors, n = 3 pools of 2 tumor each. *P < 0.05 and **P < 0.01 (two-tailed t test). Mean ± SD. d mRNA levels of cell cycle genes in GFP⁺ cells from Vector- and GFP-ZEB2-transduced tumors. Mean ± SD, n = 3. *P < 0.05 (two-tailed t test). e Volume of xenografts expressing pLenti-GFP (Vector, black line) or GFP-ZEB2 (ZEB2, red line), untreated or treated (Vector, gray line/triangles and ZEB2 yellow line/diamonds) with 5-fluorouracil plus oxaliplatin (5FU + OXA). Mean ± SEM, 6 tumors per group. *P < 0.05 and **P < 0.01 from one-way ANOVA and Bonferroni post-tests. f Upper panels: immunoblot analysis of ZEB2, CRAF pS338 and ASK1 pS83 on whole tumor lysates derived from SW480 xenografts. Lower panels: densitometry analysis of western blots, n = 3, *P < 0.05, **P < 0.01 and ***P < 0.001 (two-tailed t test). g Left: immunoblot analysis of EMT-related proteins on whole xenograft lysates. Every sample is a pool of 2 tumors. Right: densitometry analysis, n = 3, *P < 0.05, **P < 0.01 and ***P < 0.001 (two-tailed t test)

Fig. 5

Higher ZEB2 expression is linked to CMS4 and poor prognosis in colorectal tumors. a Kaplan Meier curve showing the relapse-free survival of 802 CRC patients separated on the basis of ZEB2 expression (red, low expression and blue, high expression). ***P < 0.001 based on log-rank test. b ZEB2 levels in CMS4 tumors as compared with CMSs 1–3. ***P < 0.001 based on one-way ANOVA, n = 2822. Outliers are depicted as crosses. c MKI67 levels in CMS4 as compared with CMSs 1–3. ***P < 0.001 based on one-way ANOVA, n = 2822. Outliers are depicted as crosses. Both the analysis of variance and the single post-hoc pairwise comparison in b and c are highly significant