1. Introduction
Rhabdomyosarcoma is the most common malignant mesenchymal tumor in children, with a peak incidence in patients younger than 5 years old [
1]. The embryonal subtype represents ~70% of all rhabdomyosarcoma cases and is associated with a good prognosis. Yet approximately one-third of embryonal rhabdomyosarcoma patients die, mostly from metastatic disease [
1,
2]. Despite the significant improvement in the survival of rhabdomyosarcoma patients, the prognosis of children with relapsed or metastatic disease has not changed over the last 30 years [
2].
As in other cancers, cancer stem cells (CSCs) have been implicated in the recurrence and progression of rhabdomyosarcoma. Although the first studies suggested prominin-1 (CD133) [
3,
4] or nestin [
3] as rhabdomyosarcoma CSC markers, our recent study showed that, regardless of these proteins, only sarcoma cell lines that express high levels of the transcription factor SOX2 form tumors in immunodeficient NOD/SCID gamma (NSG) mice [
5]. An in vivo tumorigenicity assay remains a gold-standard functional assay of the CSC phenotype because it enables us to test (directly in an animal model) the ability of cancer cells to self-renew and form a tumor that exhibits the cellular heterogeneity of the primary tumor [
6]. Commonly, this assay has been used to determine the tumorigenicity of different cell populations sorted based on prospective CSC surface markers (e.g., CD133, ABCG2, CD24, CD44 [
7,
8]). Ideally, when a xenograft tumor forms, cells are sorted again and transplanted into secondary recipient mice. This serial xenotransplantation should validate the specificity of a studied marker and verify that the sorted population of cells retains its CSC characteristics [
6]. Although it may provide some important evidence, such an approach is to a large extent limited because it assumes sustained levels of marker proteins on the cell surface of CSCs. However, it is becoming evident that the subcellular localization of these putative CSC markers, e.g., CD133 [
5,
9,
10] or CD24 [
10,
11], is dictated by a complex dynamic process, which presumably reflects cellular needs under different conditions and impairs the use of such proteins as surface markers to distinguish between CSC and non-CSC states [
8].
A growing body of evidence shows that enhanced cell plasticity, one of the intrinsic characteristics of CSCs, may provide another explanation for the contradictory results reported regarding the validity of CSC surface markers [
7,
8,
12]. Many of the previously suggested markers have been recently demonstrated as being non-specific, either because both marker-positive and marker-negative cell populations exhibited the same tumorigenic capacities or because marker-positive cells did not form tumors in mice [
8]. Thus, recent CSC studies have focused on proteins that are functionally involved in the regulation of stemness and cellular plasticity, such as SOX2, OCT4, or NANOG [
8,
13]. Indeed, many studies clearly demonstrated a crucial role for these stemness regulators in the induction and maintenance of CSCs in various cancers [
8,
13,
14]. However, because most of these proteins are intracellular, cells must be genetically manipulated to enable their sorting and enrichment according to these markers [
13].
Here, we present another approach to study CSCs utilizing an in vivo tumorigenicity assay. It has been previously reported that serial xenotransplantation of cancer cells in immunodeficient mice (passaging in vivo) selects highly tumorigenic cells that generate more aggressive tumors [
15,
16,
17]. Similarly, the acquisition of stemness has been linked with cancer progression and more advanced disease [
15]. Thus, we employed serial xenotransplantation of unsorted rhabdomyosarcoma cells as a platform for an unbiased screening of molecular targets that are relevant for the acquisition and maintenance of a CSC phenotype. Our results demonstrate that serial xenotransplantation in NSG mice might be a useful tool for CSC enrichment, allowing subsequent analyses to identify prospective CSC markers and potential therapeutic targets. Using serial xenotransplantation of embryonal rhabdomyosarcoma cells, we revealed an upregulation of several genes associated with stemness and early myogenic precursors. More importantly, this approach allowed us to unveil complex molecular changes that may underlie the induction of stemness in rhabdomyosarcoma, such as a hybrid epithelial/mesenchymal signature, which was recently linked with CSCs in other cancers.
3. Discussion
Bottom-up approaches that would enrich for CSCs based on their functional characteristics and might allow the objective screening of CSC-specific markers and therapeutic targets are currently of broad interest to many cancer researchers. Here, we report a proof-of-principle platform to study rhabdomyosarcoma CSCs that successfully combines serial xenotransplantation in vivo with short-term culture of xenograft-derived cells in vitro.
A fundamental observation made while testing our approach is that rhabdomyosarcoma CSCs are maintained during short-term in vitro culture of xenograft-derived cell lines, which was confirmed by the high tumorigenicity of primary and secondary xenograft-derived cells. In fact, as serial xenotransplantation selected for CSCs, their numbers in xenograft-derived cell lines increased, which resulted in a statistically significant difference between the parental cell line NSTS-11 and the tertiary xenograft-derived cell line LTB24. This increase in CSC number was documented by the enhanced ALDH activity as well as high colony and sphere formation capacity of LTB24 cells. As these functional assays for CSC phenotype were performed using biological replicates from different passages in vitro, the obtained data also underpinned the stability of CSC numbers during short-term in vitro culture. Such stability was further demonstrated by similar expression patterns detected by immunofluorescence and qRT-PCR among different in vitro passages of each of the examined cell lines. Thus, the major advantage of our approach resides in the continuous supply of xenograft-derived cell lines that (i) may serve as a cost-effective model for the selection of CSCs for comparative studies, (ii) may be stored for future studies, and (iii) provide enough material for subsequent analyses and assays.
One of the intriguing changes detected in our model was the increase in ALDH activity despite the decrease in ALDH1 level detected over the passages in vivo. It has only recently come to light that ALDH activity but not necessarily ALDH1 expression marks CSCs [
33]. Indeed, it has been demonstrated in hematopoietic cells that ALDH1A1 deficiency does not affect Aldefluor staining [
34]. Similarly, although Aldefluor positivity was associated with breast CSCs, ALDH1A1 did not correlate with Aldefluor positivity and performed poorly as a predictor of breast carcinoma progression [
35]. In agreement with our results, the latter study showed that increased expression of other ALDH isoforms, e.g., ALDH6A1, correlated with high ALDH activity and was associated with metastatic disease in breast carcinoma [
35]. Recently, proteomic analysis of primary and metastatic prostate cancer has also demonstrated that ALDH6A1 is highly specific to progressive metastatic disease [
36]. Using our model of gradual selection of rhabdomyosarcoma CSCs, we revealed an apparent upregulation of ALDH6A1 at mRNA and protein level, which was associated with Aldefluor positivity and increased stemness, while
ALDH1A1 gene expression was markedly downregulated. Importantly, the survival analysis performed using TCGA-SARC database supported the observed decrease in
ALDH1A1 expression. We show here that it is in fact the downregulation (not upregulation) of
ALDH1A1 gene expression that is associated with aggressive tumors and strongly predicts (
p < 0.0001) poor prognosis among various sarcomas. Notably, these results contradict the previous report by Martinez-Cruzado et al., whom used artificially transformed mesenchymal stem cells as an experimental model of sarcomagenesis and proposed that ALDH1 expression is enhanced in xenograft-derived cells and associates with sarcoma CSCs [
37]. In contrast, our model utilizing primary tumor-derived cells suggests ALDH6A1 as a candidate molecule that plays a role in rhabdomyosarcoma CSCs. Indeed, the expression of
ALDH6A1 has already been identified among the genes that correlate with poor outcome in rhabdomyosarcoma patients [
31]. This is in agreement with our survival analysis on soft-tissue sarcomas demonstrating that upregulated expression of
ALDH6A1 but not
ALDH1A1 or
ALDH3A1 significantly correlates with poor survival when combined with other genes identified in our model, i.e.,
CDH15,
MYOD1,
SOX4,
ARMCX1, and
RYK.
Comparing the gene expression profiles of parental NSTS-11 cells and three subsequent xenograft-derived cell lines, we were able to identify gradual changes in the expression of several genes that confer stemness and/or are linked with undifferentiated myogenic precursors. For instance, our data showed an apparent increase in the expression of the
CDH15 gene, which encodes M-cadherin. In a model of RAS-driven rhabdomyosarcoma tumorigenesis, expression of
KRAS under the
CDH15 promoter resulted in less differentiated and more aggressive tumors [
38]. Cells isolated from such rhabdomyosarcoma tumor were enriched for tumor-initiating activity and expressed markers of early myoblasts and adult muscle stem cells, i.e., satellite cells [
38]. Indeed, M-cadherin expression has been clearly demonstrated to be a marker of satellite cells [
39], which share many similarities with the gene expression signature of embryonal rhabdomyosarcoma [
40].
Although embryonal rhabdomyosarcoma most likely arises from myoblasts, both markers of quiescent satellite cells [
39,
41], e.g., PAX3, PAX7, or HEYL, and markers of activated satellite cells, including MYOD1 [
40], are frequently expressed. Notably,
MYOD1 was one of the most upregulated genes (>8-fold in tertiary xenograft-derived LTB24 cells) in our study. MYOD1 is a transcriptional activator expressed in early muscle progenitors and is required for and regulates muscle progenitor specification [
38]. In rhabdomyosarcoma, MYOD1 is expressed in small, primitive tumor cells, whereas cells that exhibit morphological evidence of skeletal muscle differentiation generally lack MYOD1 expression [
42]. Recently, it has been suggested that MYOD1 regulates common gene programs that lock cells in an arrested myogenic fate and is required for the self-renewal of rhabdomyosarcoma cells and sustained tumor growth [
43]. Corroborating the increase in the expression of
CDH15 along with
MYOD1 in our model of rhabdomyosarcoma CSCs,
MYOD1 knockdown has been demonstrated to significantly downregulate
CDH15 expression [
43].
MYOD1 may also maintain pre-myogenic mesoderm expression by upregulating PAX3, PAX7, and EYA2 [
44]. PAX3 has been shown to prevent the differentiation of myoblasts and satellite cells [
22]. PAX3 also acts as an upstream regulator maintaining stemness in neural crest cells [
22], which suggests its general impact on cell fate. Thus, upregulated expression of
PAX3 might explain the increased stemness of LTB24 rhabdomyosarcoma cells in our study. The gradual increase in rhabdomyosarcoma stemness could be further substantiated by the detected downregulation of 10 out of 18 genes that are associated with mesenchymal stem cell senescence [
25] and by the upregulation of genes involved in early mesenchymal development and maintenance of myogenic precursors, i.e.,
SOX4 [
45,
46] and PITX2 [
23].
SOX4 has been previously demonstrated to contribute to tumor progression by promoting cancer stemness [
46]. Importantly, SOX4 is highly expressed in embryonal rhabdomyosarcoma compared with normal muscle and knockdown of SOX4 leads to a significant decrease in MYOD1 levels and impaired rhabdomyosarcoma cell survival [
47]. Conversely, we have identified downregulation of
MIR29A and
MIR145 genes encoding microRNAs that are crucial for induction of myogenic differentiation [
48] and repression of pluripotency [
8,
25], respectively. Although downregulation of microRNA 145 (miR-145) has been suggested in tumorigenesis of Ewing’s sarcoma [
8,
49], our study provides the first evidence that miR-145 might be involved in regulation of CSC phenotype in rhabdomyosarcoma. Together, the gene expression profiling data indicate that serial xenotransplantation in NSG mice resulted in the progressive dedifferentiation of NSTS-11 rhabdomyosarcoma cells and their reprogramming towards the expression signature of non-senescent myogenic precursors, which was associated with enhanced rhabdomyosarcoma stemness both in vivo and in vitro.
Finally, the induction of stemness through serial xenotransplantation in NSG mice was also associated with the modulated expression of many genes involved in the EMT program. Intriguingly, functional analysis of these genes suggested an acquisition of an equilibrium between EMT and its reverse process, mesenchymal-epithelial transition (MET). Moreover, some of the typical markers of an epithelial phenotype (
CDH1) or a mesenchymal phenotype (
CDH2) were downregulated, which suggests a loss of differentiation and a shift towards an undifferentiated metastable phenotype that combines partial mesenchymal and epithelial traits. In carcinomas, recent accumulating evidence has shaped a widely accepted view that EMT generates CSCs, which appear to reside at an intermediate state along the epithelial-mesenchymal spectrum of phenotypes [
49,
50]. Carcinoma CSCs thus most likely undergo EMT only partially, attaining a hybrid epithelial/mesenchymal phenotype, which confers plasticity, invasiveness and tumor-initiating capacity [
49,
50]. In fact, hybrid epithelial/mesenchymal cells can integrate various epithelial and mesenchymal traits that facilitate collective cell migration and promote metastasis [
51]. Corroborating these observations, excessive activation of the EMT program leading to a highly mesenchymal phenotype has been shown to be detrimental to the tumorigenic activity of CSCs [
52,
53]. In parallel to the results obtained in carcinomas, partial MET has been discussed as a process contributing to the metastasis and stemness of sarcomas, although this mechanism remains elusive [
54,
55,
56].
Here, we report novel evidence for the prospective role of MET in tumors of mesenchymal origin. Notably, some of the identified rhabdomyosarcoma CSC-specific targets have already been linked with MET. For instance, it has been demonstrated that overexpression of PAX3 induces MET in mesenchymal cells [
57]. Importantly, MET has also been established as a key cellular mechanism in the process of reprogramming mesenchymal somatic cells, fibroblasts, towards pluripotency induced by transcription factors, including SOX2 [
58,
59]. These results suggest that enhanced expression of these stemness-related factors may, in the context of embryonal rhabdomyosarcoma, result in a partial loss of mesenchymal phenotype, as indicated here by the progressive downregulation of a prominent EMT inducer,
SNAI1, and
CDH2, which encodes N-cadherin.
Among sarcomas, osteosarcoma tumor tissues have been shown to exhibit extremely low expression levels of N-cadherin [
60]. Furthermore, several cadherins, including P-cadherin, E-cadherin, and N-cadherin, are markedly downregulated in osteosarcoma cell lines in vitro [
61], which is in line with the gene expression signature associated with the enhanced rhabdomyosarcoma stemness in our study. Mechanistically, N-cadherin overexpression has been demonstrated to significantly impair osteosarcoma migration in vitro and metastasis in vivo [
61]. Experimental re-expression of N-cadherin also restores cell-cell contacts and inhibits cell migration in glioma, another highly aggressive non-epithelial tumor, which is frequently characterized by low levels of N-cadherin [
62]. Similarly, metastatic dissemination of neuroblastoma is strongly correlated with low N-cadherin expression [
63]. Notably, N-cadherin downregulation is crucial for the migration of neural crest cells [
64] and smooth muscle cells [
65], which offers an explanation for how partial MET, i.e., a loss of the typical mesenchymal marker, N-cadherin, may promote the progression of sarcomas towards a more aggressive phenotype.
Although epithelial differentiation has been reported in various sarcomas, it has never been associated with a full transition of sarcoma cells to the epithelial state and rather presents a hallmark of phenotypic plasticity resulting from the active MET program [
54]. In fact, the combined presence of epithelial and mesenchymal features has been proposed to contribute to aggressive sarcomas [
55]. From this perspective, EMT-related signaling that would counteract complete MET is crucial for the maintenance of mesenchymal traits. Indeed, such activation of the EMT program has been identified in the present study. In addition to the MET-related gene expression signature, the increase in rhabdomyosarcoma CSC number during serial xenotransplantation correlated with the upregulation of a typical EMT-inducing gene,
ZEB1, and a stemness-associated gene,
SOX4, which has recently been identified as a master inducer of EMT, controlling several EMT-relevant genes [
46,
66]. Based on the results reported here, we hypothesize that it is the activation of both MET and EMT signaling that sets the rhabdomyosarcoma cells in a “ready-to-act” stem-like state in which they can easily exploit various microenvironmental cues to promote their survival, proliferation, and/or migration. Remarkably, another evidence supporting this hypothesis has been recently reported in a mouse model demonstrating that rhabdomyosarcoma CSCs arise from genomically instable satellite cells, which undergo MET-like process [
67]. Further mechanistic studies to examine the prospective link between the hybrid epithelial/mesenchymal phenotype and rhabdomyosarcoma progression are therefore needed to advance our understanding of this rare but aggressive disease.