Cancer stem cell marker aldehyde dehydrogenase has independent prognostic value in LMS
Elevated aldehyde dehydrogenase 1 (ALDH1) is considered a universal marker of CSCs [10]. Studies investigating bone sarcoma cell lines indicated that ALDH1high cells had increased capacity to form spheres and colonies and had significantly greater frequency of tumor-initiating cells [11, 12]. We have therefore analyzed by using IHC the expression of the ALDH isoform 1 expression in two independent series of LMS (cohort 1 n = 145, cohort 2 n = 89) as shown in Fig. 1a, b. On multivariate analysis, high ALDH1 expression score was significantly associated with poor metastases-free survival (hazard ratio = 1.5 [95% CI 1.3–1.8], p = 0.04), together with grade 3 (hazard ratio = 1.5 [95% CI 1.3–1.8], p = 0.004), and a tumor size of > 10 cm (hazard ratio = 1.5 [95% CI 1.3–1.8], p = 0.001). We did not find any correlation between the level of ALDH1 expression and the level of p-S6 expression (data not shown).
Generating dual PI3K/mTOR inhibitor resistant LMS cells in vitro and in vivo
In order to confirm that enhanced ALDH1 activity is a hallmark of LMS stem cells and the potential role of these cells in resistance to therapy, we decided to generate resistant LMS cells from three LMS cell lines we have previously described to be sensitive to dual PI3K/mTOR inhibition [6]. We exposed these cell lines were to BEZ235, a dual PI3K/mTOR inhibitor, using doses that were increased in a stepwise manner, and surviving cells were selected until normal cell growth resumed. Thus, we independently established three BEZ235-resistant LMS cell lines (resIB112, resIB134, and resIB136), which exhibited BEZ235 IC50 values that were > 8-fold higher than the parental cell line (Fig. 2a). Cells required approximatively 48 weeks in culture to achieve resistance. As indicated in Fig. 2b, secondary resistant cells were significantly more resistant than parental cells to BEZ235-induced apoptosis. No cross-resistance with other anti-cancer agents approved for the management of sarcoma patients (doxorubicin, gemcitabine) was observed (Additional file 1: Figure S1A and S1B). We further explored the PI3K/mTOR pathway by western blotting using phosphorylation of S6RP as a marker of pathway activation. No change in pS6RPS240/244 accumulation was observed between the three parental and resistant LMS cell lines (Fig. 2c).
These in vitro findings prompted us to examine whether these cells were drug resistant in an in vivo setting. Thus, IB136 and resIB136 xenografts were established and grew to a size of 100 mm3, after which mice were treated orally with either vehicle or BEZ235 5 days a week for 3 weeks. Our results showed that 40 mg/kg of BEZ235 treatment potently inhibited tumor growth of IB136-derived xenografts (55% tumor inhibition, p < 0.05), while tumor growth of resIB136-derived xenografts was strongly similar to the vehicle-treated group (Fig. 3a). Also, BEZ235 treatment highly significantly slowed the rate of parental IB136 tumor growth compared to the control (8.4 vs. 6.5 days for median survival; p < 0.0001) and had no effect on resIB136-derived xenograft growth as shown in Fig. 3b. No apparent toxicity events were observed in the drug-treated animals. There were no significant changes in animal weight (data not shown). The number of tumor cells positive for Ki-67, a cell proliferation marker, and p-S6RPs240/244, a protein indicating activity of the PI3K/mTOR pathway, was substantially lower in IB136 and resIB136 xenografts treated with BEZ235 compared with control tumors (Fig. 3c). Altogether, these results showed BEZ235 could inhibit PI3K/mTOR signaling pathway of both tumors but had no effect on resIB136 tumor growth suggesting, in vivo, the presence of an acquired-resistance to treatment.
Cancer stem cell (CSC) markers were upregulated in LMS tumors with acquired resistance to dual PI3K/mTOR inhibition
To identify the mechanisms of secondary resistance to dual PI3K/mTOR inhibition, we performed whole-transcriptome sequencing (RNAseq) of three independent parental and resistant tumors. We identified 985 genes that were differentially expressed between IB136-derived parental and resistant tumor xenografts, among which 583 genes were upregulated, and 402 genes were downregulated (with two times or more fold-change and an adjusted p value < 0.01). Features of differentially expressed genes in resIB136 tumors were summarized (upregulated genes in Additional file 1: Table S2 and downregulated genes in Additional file 1: Table S3). Afterwards, the limma and GSEA were performed to evaluate the different gene expression and pathways between these two groups. The heatmap showed that there was a distinct gene expression pattern between the IB136-derived parental and resistant tumor xenografts (Fig. 4a). The results showed that these differentially expressed genes were highly enriched in proliferative, growth, and embryonic development networks (Additional file 1: Table S4). Transcription levels of most molecules in stem cell pathway are either continuously upregulated, downregulated, or unaffected (Fig. 4b). When analyzing the differentially expressed genes of this pathway, we found in the resistant group, NCAM1, as the most strongly upregulated gene with a fold-change of 375.15 (Additional file 1: Table S5). Moreover, upregulation of ALDH1A2 (fold-change of 10.40), KLF5 (fold-change of 6.73), and SOX2 (fold-change of 2.5) suggested that secondary resistance to dual PI3K/mTOR inhibition could be associated with the emergence of a CSC-like subpopulation. Indeed, these markers have been reported as being involved in the self-renewal and self-protection of cancer stem cells [13, 14]. To confirm our hypothesis at the protein level, we examined the immunohistochemical expression of the stem cell marker SOX2 and ALDH1 in the parental and secondary resistant xenografts and found significantly higher expression in the resistant tumors (Fig. 4c).
LMS cells with ALDH1 activity have tumorigenic and resistant CSC properties
Then, to examine whether secondary resistant LMS cell lines contain a CSC-like subpopulation compared to parental cell lines, we performed the Aldefluor assay in parental and resistant cell lines. Several lines of evidence showed that enhanced ALDH1 activity, measurable by the Aldefluor assay, is a hallmark of cancer stem cells [15]. Flow cytometry analyses revealed that ALDH1 activity was significantly higher in resistant cell lines compared to parental lines (Fig. 5a). These results indicated that the three BEZ235-resistant cell lines contain a subset of cells with high ALDH1 enzymatic activity.
CSCs possess an increased capacity for tumorigenicity and resistance to anti-cancer drugs. We hypothesized that ALDH1high cells are CSCs. To further address the characteristics of these high-ALDH1 LMS cells and their role in resistance to dual PI3K/mTOR inhibition, we isolated ALDH1low and ALDH1high subpopulations from the resistant LMS cell lines by using fluorescence activated cell sorting (FACS). We then compared their respective viability after 72 h of treatment with 5 μM BEZ235 (Fig. 5b). The ALDH1high subpopulation was significantly more refractory to BEZ235 than was the ALDH1low subpopulation in all the three cell lines. To evaluate the stem cell properties of the LMS cells according to the level of ALDH1 expression, we submitted ALDH1low and ALDH1high subpopulations to a tumorsphere assay (Fig. 5c). In the three resistant leiomyosarcoma cell lines, ALDH1high cells formed significantly more tumorspheres after 15 days of in vitro culture than their respective negative counterparts (Fig. 5d). A self-renewal assay with cells taken from the tumorspheres confirmed that the ALDH1high subpopulation was comprised of tumorigenic CSCs (Fig. 5e). To confirm these results in vivo, xenografts were performed in mice with resIB112-, resIB134-, and resIB136-FACS-sorted cells based on ALDH1 activity (Additional file 1: Table S6). In all cases, tumors developed at a significant higher frequency in ALDH1high cells than in their respective ALDH1low cells (cancer-initiating cell frequencies range from 1/144 to 1/3104 for ALDH1high cells vs. 1/1514 to 1/21,465 for ALDH1low cells, p < 0.05). Altogether, these results indicate that ALDH1 is useful for the detection and isolation of CSCs with tumorigenic and resistant properties in leiomyosarcoma.
Pretreatment of resistant LMS cells with an EHZ2 inhibitor significantly re-sensitizes cells to BEZ235 in vitro and in vivo
Since the polycomb repressive complex 2 (PRC2), including EZH2, has been shown to play a crucial role in stem cell maintenance [16, 17], we decided to investigate whether secondary resistance to PI3K/mTOR pathway inhibition observed in vitro and in vivo was correlated with PRC2 activity. When PRC2 is activated, it catalyzed the trimethylation of lysine 27 of histone H3 (H3K27Me3). Indeed, the use of a potent orally available EZH2 inhibitor, EPZ011989, showed a marked decrease in the number of H3K27Me3-positive cells compared with the control (Fig. 6a). Therefore, we wondered whether blocking PRC2 activity with an EZH2 inhibitor could restore sensitivity to dual PI3K/mTOR inhibition. To this end, we pretreated BEZ235-resistant cells with EPZ011989 for 1 week followed by treatment with BEZ235 for 72 h. Pretreatment of resistant cells with EPZ011989 restored significatively the sensitivity to BEZ235 (p < 0.05), generating IC50 values that were similar to the parental cell lines (Fig. 6b). Moreover, flow cytometric results revealed that ALDH1 activity was decreased in resistant cells pretreated with EPZ011989 compared to untreated resistant cells suggesting that EZH2 inhibition activity can reduce the presence of the CSC-like subpopulation in the tumor (Fig. 6c). This effect was confirmed on the formation of tumorsphere by both resistant cell lines, in which EPZ011989 treatment, alone or in combination with BEZ235 treatment, reduced significantly the number of tumorigenic CSCs (Additional file 1: Figure S2A and S2B) while BEZ235 treatment at IC50 only slightly affected the capacity of ALDH1high cells to form spheres.
Moreover, we tested whether PRC2 blockade could re-sensitize cells to dual PI3K/mTOR inhibition in vivo. resIB136 xenografts were established and allowed to grow to a size of 100 mm3. Mice were randomly pretreated with either vehicle or 125 mg/kg of EPZ011989 BID for 2 weeks. Then, in the EPZ011989-pretreated group, mice were randomly treated for 3 weeks with 40 mg/kg of BEZ235 or 125 mg/kg of EPZ011989 BID. The EPZ011989 pretreatment plus BEZ235 group showed an anti-tumor effect with a significant (p < 0.05) reduction of tumor growth (average tumor volume at endpoint, 216.5 ± 50.3 mm3) compared with either drug alone (321.8 ± 51.75 mm3 for BEZ235 and 308.9 ± 55.02 mm3 for EPZ011989) or vehicle (355.9 ± 73.83 mm3) as shown in Fig. 7a. Additionally, the survival data of mice (from the first day of the experiment until the day when the tumor size doubled) showed that the combination treatment significantly (p < 0.05) slowed the rate of tumor growth (24.65 days for median survival) compared to the vehicle (18.76 days) and the individual drug (15.12 days for BEZ235 and 16.89 days for EPZ011989) treatment groups (Fig. 7b). No apparent toxicity events were observed in the drug-treated animals. There were no significant changes in animal weight (data not shown). Further immunohistochemical analyses showed that the number of tumor cells positive for p-S6RPs240/244 was substantially lower in tumors treated with BEZ235. None of the tumors pretreated with EPZ011989 displayed SOX2- or H3K27Me3-positive cells (Fig. 7c). Altogether, these results showed that pretreatment of EPZ011989 could affect the differentiation of CSC-like subpopulation to re-sensitize the dual PI3K/mTOR inhibitor.
Clinical validation of the presence of a CSC-like subpopulation as a potential resistance mechanism to a PI3K/mTOR pathway inhibitor
To confirm our in vitro and in vivo observations, we decided to analyze tumor samples from three patients treated with an investigational PI3Kα inhibitor and for whom tumor material was collected prior to treatment onset and at occurrence of secondary resistance. We were able to perform whole-transcriptome sequencing (RNAseq) of paired primary/secondary resistant tumor samples from one of these three patients. Then, 1110 genes were differentially expressed in the resistant tumor, among which 991 genes were upregulated and 119 genes were downregulated (with four times or greater fold-change and an adjusted p value < 0.01). Features of differentially expressed genes in this clinical sample were summarized (upregulated genes in Additional file 1: Table S7 and downregulated genes in Additional file 1: S8). By applying the same analysis methods as described previously, limma and GSEA were performed to show the different gene expression pattern and enriched pathways between the primary and secondary resistant tumor (Fig. 8a). Strikingly, the results once again showed a significant enrichment of stem cell pathway (p < 10− 4) in the resistant tumor sample (Additional file 1: Table S9). Transcription levels of most molecules in the stem cell pathway are either continuously upregulated, downregulated, or unaffected (Fig. 8b). Of these markers, POU5F1 and NANOG were upregulated with a fold-change of 38.75 and 2.1, respectively (Additional file 1: Table S10). POU5F1, also known as Oct-3/4, and NANOG, are one of the most important proteins associated with the pluripotent properties of stem cells and is an essential factor in controlling the early stages of mammalian embryogenesis [18]. We decided to investigate whether secondary resistance to PI3Kα inhibition in these patients was correlated with PRC2 activity. Interestingly, by using IHC, we found a significantly higher level of H3K27Me3 in resistant tumors compared with primitive tumors (Fig. 8c).