Surface markers
A fraction of cells with clonogenic and self-renewing ability was identified in early studies based on semi-solid supports such as soft agar and methylcellulose [
74,
75]. However, the first indication of in vivo validated OCSC came from pioneering studies of Bapat and colleagues who identified cell clones with tumorigenic activity from the ascites of a patient with HGSC, and showed that the tumors obtained in xenotransplanted mice recapitulated the histopathological features of the original disease [
48]. Thereafter, several reports described the isolation and characterization of OCSC from patient-derived samples, mouse models of OC, or established OC cell lines. Many of those studies relied on the use of surface markers for the purification of OCSC, as outlined below. Table
1 summarizes some of the information on surface proteins proposed as OCSC markers.
Table 1
Cell-surface markers of OCSC
CD44 | HA receptor. Stimulates EGFR-Ras-ERK. Cell proliferation, differentiation, motility, chemoresistance. | 102 (CD44+/CD117+) | |
CD24 | Transmembrane glycoprotein. Activates STAT3. Stemness, cell adhesion, tumor cell malignancy, metastasis. | 5x103 (CD24+) | |
CD117 | Receptor tyrosine kinase. Regulates PI3K/Akt, Ras/ERK, Src and JAK/STAT pathways. Cell signaling, apoptosis, cell differentiation, proliferation, cell adhesion. | 103 (CD117+) | |
CD133 | Transmembrane glycoprotein. Induces PI3K/Akt pathway. CSC maintenance, tumor formation, chemoresistance. | 102(CD133+) | |
The screening of OC sample cohorts lent further support to the association of CD44 with stemness traits and with the clinical course of the disease. As mentioned above, recurrent OC expressed higher levels of CD44 as compared to matched primary tumors [
59], and CD44 levels correlated with poor outcome in several studies [
73,
86,
87], including a recent meta-analysis on 957 cases [
72]. The latter study, however, reported no association with the response to chemotherapy, which is somehow counterintuitive for a CSC marker. In addition, a number of studies either found no correlation of CD44 with patients’ survival [
88,
89] or even reported CD44 to be associated with better outcome [
90‐
92]. While there might be several reasons that account for such a discrepancy, including different experimental approaches and molecular tools (e.g., CD44 antibodies), different inclusion criteria between the patient cohorts and different degrees of tumor heterogeneity within the individual cohorts, it appears that the clinical value of CD44 as an OCSC biomarker remains controversial. A possible solution would be to focus on alternatively spliced variants of this molecule. Indeed, CD44 exists in various isoforms, depending on ten exons that can be added in different combinations to the standard form of the molecule [
93]. While the functional and clinical implications of CD44 variants have been investigated in various tumor types, only limited information is available on their role in CSC [
94,
95]. In OC, in particular, CD44v6 was recently shown to be up-regulated in peritoneal metastasis and, more important, a fraction of CD44v6
+ tumor cells displayed metastasis-initiating activity [
96], pointing to this variant as a putative marker of OCSC.
While the functional role of CD24 in OCSC (and in CSC in general) remains poorly defined, recent data have provided novel insights. In particular, CD24 has been shown to enhance the activation of signal transducer and activator of transcription 3 (STAT3) in different tumor types [
101‐
103]. STAT3 activity is a well-established player in cancer stemness and, indeed, CD24-induced STAT3 activation in nasopharyngeal carcinoma cells triggers their reprogramming towards a CSC state [
103]. While the causal role of the STAT3 pathway in OCSC has been described [
104], its regulation by CD24 remains to be explored. In agreement with the hypothesis of a CD24/STAT3 interplay in ovarian cancer stemness, CD24
+ OCSC exhibit increased levels of STAT3 phosphorylation and of STAT3-dependent expression of stemness factors such as Nanog and c-Myc [
105].
It should be mentioned, however, that there are studies which found no evidence for the association of CD24 expression in OC cells with stem cell phenotype or chemoresistance [
68]. In fact, other groups including ours (Lupia and Cavallaro, unpublished observation) attributed CSC properties to subsets of OC cells with no or low expression of CD24 [
58,
106].
Furthermore, as mentioned above, the combined expression of CD117 and CD44 defined a subpopulation of OC cells endowed with tumor-initiating capacity and chemoresistance [
76,
77]. CD117 was also efficient in the isolation of OCSC as a single marker: Luo et al., indeed, showed that the <2% fraction of CD117
+ cells isolated from OC xenografts had a tumorigenic potential 100-fold higher than CD117
- cells. Tumors derived from CD117
+ cells, in addition, recapitulated the heterogeneity of the original disease and could be serially transplanted, confirming the differentiation and self-renewal abilities of these cells [
110]. In other experimental systems, however, CD117
+ OC cells failed to show increased tumor initiation with respect to CD117
- cells [
105].
The receptor tyrosine kinase activity of CD117 regulates a wide spectrum of signalling cascades, including PI3K/Akt, Ras/ERK, Src and JAK/STAT [
111], all of which are plausible candidates as functional effectors of this surface protein in OCSC. Unfortunately, despite the numerous studies that capitalized on CD117 as an OCSC marker, very little knowledge is available on its biological role in these cells. Functional experiments conducted in two OC cell lines revealed that both gene silencing and pharmacological inhibition of CD117 kinase activity with imatinib reduced significantly their sphere-forming potential [
112], pointing to CD117 as a possible therapeutic target in the context of OCSC. The same study implicated the Wnt/β-catenin signalling cascade as an effector of CD117 in the regulation of stemness function [
112].
Several multi-kinase inhibitors which have CD117 among their targets have been investigated as therapeutics in ovarian cancer [
113‐
115]. Nevertheless, none of these studies has studied whether such drugs affect the OCSC subpopulation. Future efforts should be devoted to understanding the biological role of CD117 in OCSC and to testing CD117-targeted compounds as OCSC-directed therapies.
The functional contribution of CD133 to the pathophysiology of CSC, including OCSC, is poorly defined. Nevertheless, various lines of evidence suggest that CD133 is not only a surface marker but it also plays a causal role in CSC. For example, CD133 is required for the maintenance of glioblastoma stem cells [
131] and confers stem-like properties and chemoresistance to other cancer cell types [
133‐
135]. Experimental data implicated the PI3K/Akt pathway as a CD133 effector [
133,
134,
136]. Akt pathway inhibitors, are being actively investigated in ovarian cancer therapy [
137]. It would be worthwhile, therefore, to study the involvement of Akt downstream of CD133 in ovarian cancer and, in particular, to assess whether targeting Akt signalling affects the self-renewal and tumorigenic potential of CD133-positive OCSC.
Functional markers
The definition of “functional markers” refers to stemness-associated biological activities or functional states that have been exploited for the prospective isolation of OCSC.
While some of the studies discussed above pointed to the causal role of ALDH1 in conferring CSC traits to OC cells, the underlying molecular mechanism have not been fully elucidated yet. Recent data suggested that ALDH1A1 exerts a regulatory function on the levels of ATP-Binding Cassette (ABC) drug transporters, thus modulating the resistance of OC cells to chemotherapeutics, although the molecular mechanisms remains to be pinpointed [
141]. Interestingly, the role of ALDH1 in OCSC extends beyond its detoxifying activity: Meng et al. showed that the knockdown of ALDH1A1 in the OC cell line A2780 caused a decrease in the cell cycle checkpoints regulators KLF4 and p21 which, in turn, resulted in enhanced cell proliferation [
142]. Actively proliferating cells are more susceptible to cytotoxic drugs and, therefore, forcing the cell cycle entry likely contributes to the sensitization of OC cells to chemotherapy upon loss of ALDH1A1. Furthermore, the ablation of ALDH1A1 triggered DNA damage with a concomitant reduction in various DNA repair pathways [
142], implying that ALDH1A1 exerts a genome-protecting role in OCSC. The anti-proliferative role of ALDH1A1 reported by Meng et al. was contradicted by data obtained in primary OC-derived cells where the genetic or pharmacological inactivation resulted in decreased proliferation [
127]. In this experimental system, the authors also identified an intriguing interplay between ALDH1A1 and the stemness-associated gene
SOX2 and showed that their reciprocal regulation orchestrates sphere formation and OCSC survival and proliferation [
127].
The isolation of label-retaining OCSC entails, by definition, the manipulation of the system (e.g., biochemical or genetic labelling, expansion of tumor cells to allow for label dilution, etc.), which obviously increases the risk of artifacts. In addition, label retention marks quiescent cells which, however, not necessarily coincide with CSC. These limitations should be taken into account when applying label retention-based strategies. Yet, the latter offer the possibility to discover molecular and biological features of CSC, provided that the findings are validated in clinically relevant systems. For example, the transcriptomic profile of PKH-retaining mammary cells has yielded a panel of differentially expressed genes that have then been proven useful for the prospective identification of CSC from breast tumor tissue [
150]. Quiescent CSC can also be tracked genetically, for example through the retention of green fluorescent protein-tagged histone 2B (H2B-GFP). This approach can be combined with animal models of spontaneous tumorigenesis, thus making it possible to identify label-retaining CSC in preclinical tumor models that recapitulate, at least to some extent, the natural history of the disease. Long-term retention of H2B-GFP favoured the identification and characterization of stem cells in the ovarian and oviductal epithelium of transgenic mice [
151,
152]. Applying this strategy to mouse models of spontaneous ovarian carcinogenesis should help defining novel molecular and functional features of OCSC, not only in the context of tumor initiation, but also in dissemination, recurrence and chemoresistance.
It should not be neglected, however, that PKH-retaining cells isolated from OC specimens failed to show higher tumorigenic potential when compared to their PKH-negative counterparts and, in fact, exhibited lower clonogenic activity in vitro and longer tumor latency in xenograft models [
153]. Results in our laboratory with either PKH or H2B-GFP indicated that label retention in primary OC cells does not mark cells with specific OCSC-like traits (Lupia and Cavallaro, unpublished data). These observations might imply that OC is highly heterogeneous with regard to the quiescent nature of its CSC compartment, and that slow cycling is not an absolute and universal feature of OCSC.