We have found that recurrent tumors are more densely composed of putative cancer stem cell as characterized by ALDH1A1, CD44 and CD133 than their matched primary ovarian cancer specimens, suggesting that their expression is clinically significant and may correlate with residual chemoresistant populations that must be present at the end of primary therapy. Presumably targeting these populations with some other treatment modality would be required to achieve durable cures in ovarian cancer patients. Additionally, we identified several genes from a large panel of 84 genes involved in stem cell biology to be significantly overexpressed in recurrent patient samples, further suggesting that resistant tumors are enriched with genes involved in stem cell pathways. With this methodology, the TGF-β co-receptor endoglin was found to be overexpressed in residual tumors cells, and thus important to the chemoresistant population of tumor cells. This represents a previously unrecognized function of this gene, as a mediator of survival in tumor cells, in addition to its known role in angiogenesis. Moreover, the hedgehog transcription factor Gli2 was also overexpressed and functional in the chemoresistant population, and with correlative in vitro data, was found to play a novel role in platinum resistance.
It is hypothesized that CSCs may be responsible for tumor initiation or recurrent disease. There are many facets of this hypothesis that are still under debate, including what level of stemness such populations may have, how best to identify the true stem cell population, and whether these marker-defined cells are also the ones surviving initial chemotherapy (32
). However, there clearly are subpopulations within a heterogeneous tumor that have more aggressive, chemoresistant features than others in ex vivo, and now de novo models (2
). This is clinically evident in the observation that patients often have outstanding initial responses to chemotherapy, suggesting the majority of primary tumor is actually chemosensitive. It is important to note that although we do see an increase in these populations, recurrent tumors are not completely composed of these cells. This indicates that either additional chemoresistant populations are yet to be identified, or these cells have such differentiating capacity that they rapidly produce marker-negative cells, or both. An additional limitation of our analysis is the specific examination of stem cell pathways. Other pathways almost certainly play important roles in mediating survival of the therapy-resistant population; one example being, altered DNA repair mechanisms. Recent evidence suggests that ovarian cancers can arise from specific defects in DNA repair pathways, and that inhibitors of the proteins involved in these pathways, such as poly (ADP-ribose) polymerase (PARP), could be used to reverse chemoresistance (34
), It is reasonable to postulate that cancer stem cells, like normal stem cells, would have enhanced mechanism of DNA repair, allowing for survival with prolonged exposures to DNA-damaging insults. Analysis of RNA from FFPE samples showed the extract was of quality appropriate for qPCR analysis, but not enough samples had sufficient quality for full microarray analysis, which could be used in future studies to examine the role of DNA repair or other pathways in mediating chemoresistance. Further characterization of the recurrent chemoresistant tumors with evolving high-throughput methods that can be performed on FFPE samples, or identification of a cohort of patients with snap frozen tumors, would be required to fully characterize this aggressive population.
Whether the chemoresistant population is composed of predominantly cancer cells with stem cell biology or not, we propose a model of how such a population may comprise the overall tumor during different clinical settings. Because most patients have an initial positive response to chemotherapy, the presenting tumor must be composed of mostly therapy-sensitive cells (TSCs), with a small component of therapy-resistant cells (TRCs). Treatment selectively kills TSCs, resulting in predominantly TRCs, but in a small enough volume that they are not clinically detectable (“persistent tumor”). Therefore, the patient is observed, but in about 75% of cases, tumors will recur 18–24 months after completion of therapy (with an “untreated recurrent tumor”). Because of the differentiation capacity of the resistant cells, this tumor has become repopulated with CSC marker-negative differentiated cells and is again heterogeneous, with a significant portion of chemosensitive cells. This would seem to be the case given the observed 50% response rate seen in patients receiving second-line chemotherapy. However, either due to genetic changes in genetically unstable tumor cells, or further selective growth of the therapy-resistant population, ultimately the TRCs dominate, patients get no further response with multiple agents, and succumb to tumor burden (“treated recurrent tumor”). The observed increase in CSC marker staining, particularly ALDH1A1 and CD133, in samples collected immediately at the completion of primary therapy, suggests these cells have preferential survival and can go on to give rise to recurrent disease. These cells may represent a population that could be targeted in order to achieve increased response rates and survival in ovarian cancer patients.
It is an interesting finding that CD44+ cells were less dense in recurrent tumors than CD133 and ALDH1, despite multiple studies showing that CD44+ cells have CSC properties. Many of these studies have used CD44 in combination with other markers, such as c-kit (4
), My88 (5
), CD133 (6
), and CD24 (35
). It is for this reason that we examined CD44 by itself as potentially important, but at the same time may have introduced a limitation by not being able to evaluate dual-positive populations. It is yet to be determined the degree of crossover between individual markers. Likely the combination of markers will identify a more aggressive population than either alone, as previously shown with CD133 and ALDH1 (11
), but it is unknown whether such combinations then exclude other aggressive populations. This disparity, however, highlights the limitations in defining the key population by marker status alone, instead relying on clinical behaviors such as resistance to chemotherapy.
Recent studies have shown that developmental pathways (such as Notch, Wnt, Hedgehog and TGF-β) play an important role in the self-renewal and maintenance of CSCs and that inhibiting these pathways may provide useful therapeutic strategies both alone and in combination with traditional chemotherapies (36
). In our study, genes identified as being significantly overexpressed in persistent tumors included endoglin (a member of the TGF-β superfamily) and the primary mediators of hedgehog transcription, GLI1
, among others (). The most significant and consistent increase in expression from primary to persistent tumor occurred in endoglin (CD105), a TGF-β co-receptor. This molecule interacts with TGF-β receptor II (TGFBR2, which was also significantly increased in persistent tumors (2.76-fold, p=0.0190)), both dependently and independently of the TGF-β ligand (38
). This interaction subsequently promotes gene transcription mediated by the Smad family of transcription factors (Smad2 and 4). In contrast, a proteolytically cleaved, secreted form of endoglin, known as soluble endoglin (Sol-Eng) appears to inhibit TGF-β signaling by scavenging circulating TGF-β ligands (39
). Endoglin is a well-described marker of angiogenesis, whose expression is turned on in growing/sprouting endothelial cells (such as those supplying vasculature to tumors). This characteristic of endoglin has made it a desirable target for anti-angiogenic cancer therapy, with monoclonal antibodies being developed for future clinical use (29
). Previous studies have shown that endoglin expression in the stroma of ovarian tumors is associated with poor survival (40
), but the role of this receptor in cancer cell biology remains largely unexplored. Based on our data, it appears that endoglin plays a role in ovarian cancer chemoresistance and recurrence. Moreover, endoglin appears to be important for continued ovarian cancer cell survival as evidenced by our in vitro
data. In a study performed by Li et al., it was demonstrated that endoglin prevents apoptosis in endothelial cells undergoing hypoxic stress, either in the presence or absence of TGF-β ligand (42
). It could be speculated that endoglin serves a similar anti-apoptotic function in tumor epithelial cells and thereby promotes ovarian cancer cell survival. Whether this is due to the promotion of TGF-β signaling or through a TGF-β-independent mechanism remains to be determined. Taken together, these data suggest that inhibiting endoglin could be used to target both the tumor and its developing vasculature, thereby having a potentially greater therapeutic benefit. Additional studies will determine the viability of endoglin as a therapeutic target, as antibodies have been developed that disrupt the interaction of endoglin and TGF-β receptor II (43
Previous studies have implicated hedgehog signaling in multi-drug resistance (45
); however, the role of this pathway in resistance to platinum-based compounds remains largely unexplored. While both Gli1 and Gli2 appeared to mediate ovarian cancer cell survival in vitro
, only downregulation of Gli2 sensitized cells to cisplatin in a synergistic fashion, with a 5-fold reduction in IC50 concentrations in two different cell lines. It is suggested that the mechanism underlying this sensitization involves apoptosis. Inhibition of apoptosis is known to mediate cisplatin resistance (47
) and Gli2 has previously been shown to serve an anti-apoptotic function through transcriptional regulation of apoptotic inhibitor molecules (48
). In our study, we found that downregulation of Gli2 alone induced apoptosis, and this may have contributed to the increased sensitivity of ovarian cancer cells to cisplatin in vitro
. Interestingly, downregulation of Gli1 had no effect on cisplatin toxicity. Future studies of the link between Gli2, apoptosis and cisplatin resistance are warranted.
Collectively, the data presented in this study demonstrate that cells with stem cell properties enrich recurrent ovarian tumors, especially in their more chemoresistant forms. The varied density of these subpopulations in different clinical scenarios provides insight into the dynamic heterogeneity during the typical natural history of ovarian cancer progression. Additional stem cell pathways contribute to the continued survival and chemoresistance of ovarian cancer, and targeting these pathways may be necessary in order to achieve durable clinical response in for this disease. In addition, the TGF-β co-receptor endoglin (CD105) and the Hedgehog mediator Gli2 were found to be overexpressed in recurrent ovarian tumors and are promising targets in overcoming chemoresistance.