Stem-like features of cancer cells on their way to metastasis

Circulating Tumor Cells (CTCs)

Circulating tumor cells (CTCs) represent cancer cells that detach from a solid tumor mass and, using the blood circulatory system, are in the process of colonizing distant organs to give rise to metastasis [1]. CTCs have been detected in the majority of cancers of epithelial origin, including breast, colon, prostate, and pancreas but also in cancers of non-epithelial origin, such as glioblastoma multiforme and melanoma [7–12]. Most importantly, their presence has been correlated with a poor prognosis in several major epithelial cancer types, suggesting that CTCs contribute actively to the metastatic process [7, 13, 14]. However, in patients, CTCs are extraordinarily rare compared to normal blood cells (approximately one CTC among a billion normal blood cells), and their isolation and molecular characterization has been hampered for a long time by technological constraints. Furthermore, variable expression levels of cancer-associated cellular markers represent a major challenge in identifying these rare cells in a viable state, among a large number of blood cells, highlighting the need for specialized technologies for their isolation and characterization.

In the recent years, several platforms have been established that take into account both cell surface expression patterns as well as physical or functional properties of CTCs for their isolation [9, 15–18]. For epithelial cancer types, antigen-dependent isolation devices rely mostly on the expression of cell surface Epithelial Cell Adhesion Molecule (EpCAM) to isolate CTCs, given its high expression in most of the tumor cells of epithelial origin [19]. For example, the only FDA-approved platform for CTC enumeration is the CellSearch system, which uses a two-step procedure to isolate CTCs [17, 18]. In the first step, plasma components are removed by centrifugation, while EpCAM-positive CTCs are captured using immunomagnetic beads. In the second step, CTCs are permeabilized and stained for the expression of additional markers, such as the epithelial cytokeratins (CK) 8, 18, and 19. Meanwhile, contaminating leucocytes are identified with a staining against CD45, a widely expressed leukocyte marker. At the end of the procedure, captured cells are placed on an integrated analyzer for scanning and CTCs are identified as being positive for the expression of CKs and DAPI, while being negative for CD45. Even though CellSearch is currently considered as the gold standard in capturing CTCs in a clinical setting, a disadvantage of this technology is that ultimately isolated CTCs are non-viable due to fixation and additionally, the antigen-dependent approach for CTC-isolation may in fact favor the enrichment of CTCs with higher EpCAM expression, over CTCs with lower EpCAM levels as a consequence of cancer heterogeneity [20], dynamic protein turnover [21], or the down-regulation of epithelial markers in favor of mesenchymal-like traits, in a process known as Epithelial-to-Mesenchymal Transition (EMT) [8, 22–24]. As a consequence, new emerging technologies tend to focus on antigen-independent methods for CTC isolation, yet they are designed based on different principles. For example, isolation of CTCs independently of the expression of tumor-associated antigens has been achieved via size-based depletion of normal blood cells, leveraging on the notion that CTCs are slightly bigger than red or white blood cells (approximately 12-25 μm for a single CTC, vs 8 μm and 7-15 μm, for red and white blood cells respectively) [25, 26]. Filter-based platforms such as Parsortix and ScreenCell capture larger CTCs, while smaller white blood cells (WBCs) and blood components like platelets and red blood cells (RBCs) flow through the microfluidic channels [27, 28]. Alternatively, the spiral biochip is another newly developed ultra fast, size-based microfluidic device that can isolate larger CTCs from patient blood samples within a few minutes. This chip relies on a combination of hydrodynamic forces that are generated when blood flows through spiral microchannels. These forces position larger CTCs towards the inner walls of the spiral channel, while smaller blood components are pushed towards the opposite outer wall [29]. However, further studies need to be performed with this technology to assess whether the higher shear forces, compared to those that are normally present in physiological blood circulation, could influence some of the properties of CTCs. On the other hand, the CTC-iChip combines an initial hydrodynamic cell separation of nucleated cells, including CTCs and WBCs, from other blood components, followed by an immunomagnetic depletion of antibody-tagged WBCs [7]. This approach, similarly to the other size-based approaches mentioned above, results in the antigen-independent enrichment of CTCs from whole blood samples, allowing their molecular characterization.

Antigen-independent isolation of CTCs from cancer patients is key to understanding the biology of cancer metastasis in a variety of cancer types, including those of epithelial but also non-epithelial origin. Recently, using the CTC-iChip and the Herringbone ^HBCTC-chip microfluidic platforms, as well as density gradient centrifugation, CTCs were isolated from the blood of breast, prostate and pancreatic cancer patients [7, 30, 31], as well as patients with small cell lung cancer, glioblastoma multiforme and melanoma [32–34]. Downstream, single cell resolution transcriptional profiling of these cancer cells that are en route to metastasis has revealed a great degree of heterogeneity among them within the same patient, but also among CTCs from different patients. Interestingly, these studies revealed a role for non-canonical WNT signaling in drug resistance and establishment of metastases in pancreatic and prostate cancer patients [30, 31]. In human breast CTCs, a dynamic expression of epithelial versus mesenchymal markers in response to treatment was observed using quantitative RNA-in situ hybridization, demonstrating for the first time a mesenchymal-like phenotype in human metastatic cells [8]. Similarly, in glioblastoma multiforme, mesenchymal markers were enriched in CTCs over neural differentiation markers [33]. In small cell lung cancer, CTCs were shown to be tumorigenic upon transplantation in immunocompromised mice and more importantly, the xenograft tumors matched those morphological and genetic features of the primary tumor in the patient of origin, and were predictive of treatment response [32].

All together, recent technological breakthroughs are allowing us to gain fundamental insights into CTC heterogeneity in different types of cancers and patients. However, it is very important to highlight that in any given tumor type, the number of CTCs present in the bloodstream appears to largely exceed the number of clinically detectable metastatic foci, indicating that most CTCs will not lead to metastasis, and that only very few will have those features that will enable them to seed a metastatic disease.

CTC clusters

The identification and characterization of the subset of metastasis-initiating cells among the CTC population in patients is of paramount clinical importance. The majority of CTCs circulate in the blood of cancer patients as single cells, however they can also be found as clusters of 2-50 cells, with the ratio of single vs clustered CTCs varying significantly among different patients, and along disease progression [7, 30, 31]. While the role of CTC clusters in the metastatic process remained unknown for a long period, recently, their presence in the blood circulation of patients with metastatic breast, lung or prostate cancer was correlated with poor metastasis-free survival and overall survival, suggesting that CTC clusters are key players in the spread of cancer cells to distant metastatic sites [7, 35, 36]. By using the CTC-iChip technology in combination with a micromanipulator, both single CTCs and CTC clusters from patients with metastatic breast cancer were recently isolated and subjected to RNA sequencing profiling [7]. Data analysis revealed that CTC clusters upregulate a set of genes that include the cell-cell junction component plakoglobin. In breast cancer patients, increased expression of plakoglobin in the primary tumor is indicative of a decreased metastasis-free survival, while in mouse xenograft models, knockdown of plakoglobin expression in orthotopic mammary tumors suppresses spontaneous CTC cluster formation and lung metastases [7].

In the same study, using two independent mammary tumor mouse models, it was shown that CTC clusters are oligoclonal in origin and do not arise from the aggregation or proliferation of single CTCs within the circulatory system [7, 8]. Rather, CTC clusters arise when a group of malignant cells detaches from a solid tumor deposit and enters into the blood circulation. By color-coding primary tumor cells to distinguish single versus clustered CTCs, it was also shown that even though single CTCs are more frequent in blood circulation, lung metastases arise preferentially from CTC clusters. In fact, CTC clusters were found to be up to 50 times more likely to form lung metastatic deposits compared to single CTCs in this experimental setup. Using in vivo flow cytometry in mouse models to assess circulation half-life of single and clustered CTCs, it was found that larger CTC clusters are rapidly cleared (usually within 10 min) from the blood circulation, due to their entrapment in smaller capillaries at distant organs [7]. Thus, compared to the 30 min half-life of single CTCs in blood circulation, it is possible that current blood sampling techniques underestimate the number of CTC clusters released in circulation due to their shorter half-life. However, the rapid clearance of CTC clusters at distant organs indicates that clusters reach potential metastatic sites at a higher rate compared to single CTCs. Whether the higher metastatic propensity of CTC clusters correlates with the faster rate at which they reach distant organs in patients, or whether this indicates that metastasis-initiating cells with stem-like properties are preferentially enriched in clusters versus single CTCs is currently unknown.

Stem-like properties of cancer cells

By definition, stem cells are cells that are resident in virtually all normal tissues and have the ability, upon cellular division, to both self-renew and give rise to cells that will differentiate into more specialized cell types [37, 38]. As a consequence to this definition, in cancer, the so called “cancer stem cells” are defined as those cells with stem-like properties, which can self-renew and differentiate, resulting into a population of cells that has the long-term ability to clonally reproduce the tumor itself and its inherent heterogeneity [39, 40]. Functional repopulation assays are currently the gold standard for measuring stem-like properties of cells. For cancer biology, serial transplantations in mice or in situ tracking are used to ensure that cancer cells with stem-like properties have the long-term ability to clonally maintain the tumor and regenerate the heterogeneous tumor mass. These assays are designed in order to simultaneously assess the self-renewal and differentiation potential of putative stem-like cancer cells. The principle for the serial transplantation assay is that cancer stem-like cells must have the ability to re-establish the primary tumor, while maintaining both heterogeneity and self-renewing ability upon sequential transplantations into animal models [41]. During in situ clonal assays, a population of cells is initially labeled, most commonly by the addition of a fluorescent marker, to allow long-term tracing of the progeny in the recipient mouse, in order to assay again for self-renewal and differentiation potential [42].

However, one of the major challenges in these studies is to identify those expression markers that faithfully define a cancer stem cell phenotype. Undeniably, there is a great degree of heterogeneity between stem-like cancer cells themselves and among different tumor types [37, 38], arguing that the identification of faithful stem-like markers in cancer cells is an extraordinarily difficult task. The identification of a small population of cells enriched with a tumor-initiating potential was first reported in Acute Myeloid Leukemia (AML) and it was based on the expression pattern of cell adhesion markers CD34 and CD38 [43]. Using flow-sorting techniques it was shown that a fraction of CD34⁺/CD38^- but not CD34⁺/CD38⁺ cells were able to initiate AML upon serial transplantations in immunodeficient mice. These studies provided proof that not all tumor cells are equal in their tumor-initiating potential and paved the way for the identification, based on expression of cell surface markers, of similar populations with stem-like properties in other solid tumors. For instance, in brain and lung cancer, the expression of CD133, a transmembrane protein with poorly understood signaling function, correlates with tumor initiating potential upon serial transplantations in immunodeficient mice [44, 45]. Similarly, in colon cancer as well as in ovarian tumors, cancer-initiating cells are highly enriched in the CD133 expressing fraction [46–48]. In pancreatic cancer, combination of two distinct expression patterns, such as of CD44/CD24 together with the epithelial-specific antigen (ESA) or CD133/CXCR4, can identify the cancer stem-like population [49, 50]. However, CD133 expression combined with that of CXCR4, a mediator of cell migration, was also shown to enrich for a population that in addition to stem-like features, also displayed a metastatic potential in pancreatic cancer [50]. In breast and prostate cancer in particular, the subpopulation of cancer cells with stem-like properties is identified as being enriched in the CD44⁺/CD24^- fraction [51, 52]. In breast cancer, combination of CD44 expression with that of alcohol dehydrogenase (ALDH), which is also expressed in hematopoietic progenitor cells, could further refine the tumorigenic population [53, 54]. ALDH alone was also used as an expression marker in both breast and prostate cancer to identify a distinct population of stem-like cells in addition to the CD44⁺/CD24^- fraction [55, 56].

Taken together, these studies indicate that combinations of different cell surface expression markers have been used to identify distinct tumor-initiating populations in solid tumors, but that use of these expression markers may not yet be sufficiently refined to identify unequivocal stem-like populations in different human cancers, especially considering the high degree of mutational and cellular heterogeneity of human tumors in the metastatic setting, and upon resistance to therapeutic agents [57]. Further, it has been shown that the percentage of putative stem-like cancer cells in each model may vary dramatically, depending on experimental conditions. Factors such as the use of different recipient mouse strains, as well as assaying tumors at different stage of progression and treatment may explain some of the discrepancies between those studies aimed at assessing the stem-like population in solid tumors (Reviewed in [57]). Univocally though, the studies summarized above point to the fact that, in order to dissect the heterogeneity within the putative stem-like cancer populations, those functional assays that are currently used to assess stemness such as the serial transplantation assay, will need to be redesigned in order to assay for both self-renewal and differentiation potential in clonal populations arising from single cells. Tracing both self-renewal, differentiation and tumorigenic potential to a single cell level will also be crucial for identifying true tumor-initiating cells endowed with stem-like potential in cancer, and discriminating them from subpopulations of cells that are only enriched in proliferation or survival properties but lack true stem-like potential. To this regard, there is also a need to reassess culture methods in vitro, and most importantly ensure that clonal populations can arise from single cells in vitro without the need to exogenously supplement saturating amounts of growth factors. Altogether, these studies point towards the need to refine both the use of expression markers and related functional assays, especially considering the possibility that is now available, to interrogate cancer cells at higher resolution and at the single cell level [58, 59].

Stem-like properties of cancer cells in circulation

Expression of cancer stem-like markers in various cancer types has been shown to correlate with the occurrence of metastasis and reduced survival in patients, supporting the hypothesis that cancer cells enriched in stem-like features are the precursors of metastasis (reviewed in [57, 60]). Recent single cell-resolution RNA sequencing data obtained from CTCs of patients with prostate and breast cancer may help gain insights into the expression of stem-like markers, specifically in cancer cells as they transit through the bloodstream, as CTCs isolated from these cancer patients were shown to express CD44. Moreover, 60 % of the CTCs profiled from prostate cancer patients were also enriched for ALDH7A1 and other putative stem-like markers such as KLF4 [7, 31]. However, the expression dynamics and the heterogeneity regarding the expression of these markers have not yet been quantitatively assessed in CTCs. Most importantly, it will be necessary to link the expression patterns of these putative stem-cell markers in CTCs with enhanced tumor-initiating and metastatic potential. To date, even though CTCs isolated from patients are able to form tumors in mice, they have not been assessed for their ability to long-term self-renew and regenerate a highly heterogeneous cancer upon serial transplantations [7, 30]. Arguably, additional functional assays should be implemented, that can assess the stem-like properties of cancer cells at the single cell level, and will enable the correlation of stem-like marker expression with the metastatic potential of putative stem-like CTCs.

On the other hand, it is increasingly evident that stem-like properties may represent an extraordinary advantage for CTCs as they encounter a number of challenges during the metastatic process. For example, a first major challenge that CTCs face, is the need to suppress anoikis, a form of programmed cell death that is triggered in epithelial cells upon loss of adhesion to surrounding cells. Once in blood circulation, CTCs are striving to survive in a hostile, immune cells-rich environment while they are subjected to high shear forces that challenge their structural integrity. Additionally, if successful, CTCs that reach distant organs will need to adapt to the foreign environment in order to seed a metastatic lesion. In this context, it is not surprising that the majority of molecular pathways that regulate self-renewal, such as PI3K/AKT, PTEN, JAK/STAT, seem to be also essential for cancer onset and metastasis, while increasing survival signals that provide resistance to cell death and chemotherapeutic agents [61–64]. Whether stem-like properties are a feature of some CTCs and whether they are enriched in single versus clustered CTCs remains an open question. The fact that CTC clusters are more prone to establishing metastases compared to single CTCs in mouse xenograft models could indicate that stem-like properties are enriched in CTC clusters over single CTCs. On the other hand, it is also possible that stem-like properties are equally enriched in both single CTCs and CTC clusters, but the higher metastatic propensity observed with CTC clusters is reflective of their physical advantages, such as the greater likelihood to be trapped in small capillaries at distant sites. To this regard, the presence of cell-cell junctions in CTC clusters could help overcome anoikis in the bloodstream and provide survival signals along the metastatic process, including during the early dissemination to distant organs. The actual contribution of specific mutations and deregulated signaling pathways, mesenchymal-like features and CTC-clustering for the generation of metastasis-initiating cells remains to be addressed (Fig. 1).

Another important hint connecting stem-like properties with CTCs is the protocol that has been implemented for CTC cultures [65]. After testing a number of different culture conditions it was shown that CTCs could be propagated in vitro only if they were cultured in hypoxic conditions, as tumor spheres in serum-free media supplemented with epidermal growth factor (EGF), B27 and basic fibroblast growth factor (FGF2). Interestingly, given the fact that CTCs have a limited half-life in blood circulation, the requirement for hypoxic conditions for their propagation could reflect the oxygen levels within the tumor tissue of origin, rather than the oxygen concentration within the arterial or venous circulation. Similarly to CTCs, culture of human embryonic stem cells (hESCs) under hypoxic conditions is also important for regulating pluripotency and proliferation of these cells in vitro [66]. Additionally, for hESC and human induced pluripotent cell (hIPSC) culture in vitro, FGF2 is required for promoting self-renewal and survival in several different ways, such as via the activation of the PI3K/AKT and MAPK signaling pathways [67], but also through promotion of cell adhesion, highlighting the importance of cell-cell contact in maintenance of self-renewal and survival in hESCs [68]. To this regard, hESCs can be propagated in an undifferentiated state in vitro essentially indefinitely by forming rather tightly packed colonies of individual cells that are connected through various complexes, such as adherence junctions, tight junctions, desmosomes and gap junctions [69]. Undoubtedly, not all stem cells require FGF2 signaling for maintaining pluripotency in vitro. In fact, mouse ESCs that are functionally very similar to hESCs, are destabilized by FGF2 and depend on the activation of the LIF/Stat3 signaling pathway [70]. However, as opposed to conventional hESCs that display very poor clonal efficiency and survival when dissociated to single cells, mESCs not only tolerate single cell dissociation but also manage to maintain pluripotency under these conditions [71]. Therefore, we speculate that in systems such as hESCs that depend on FGF2, the link between pluripotency and cell-cell adhesion could support a model in which CTC clusters are enriched in stem-like properties. Further, not only cell-cell adhesion but additionally other signaling molecules could be promoting stemness in CTCs [30, 31, 72].

Recent molecular studies on CTC clusters indicate that there is a great degree of cellular and molecular heterogeneity among different CTC clusters, even among those profiled from the same patient [7, 30, 31]. This heterogeneity unquestionably also reflects differences in expression profiles of individual CTCs that form the same cluster and potentially also reflects the expression of stem-like features. Therefore, it is possible that only a few cells within a CTC cluster are in fact retaining stem-like properties and are able to form metastases, while other surrounding cells display reduced or no stem-like characteristics and do not contribute to the metastatic process. Additionally, other, non-cancerous cell types, such as macrophages, platelets and even fibroblasts and epithelial cells, were shown to bind CTC clusters [7, 73–75]. These interactions can affect the molecular profile of the cells they are in direct contact with, further promoting heterogeneity and enhancing the overall fitness of CTCs. For example, platelets that bind to CTCs may release transforming growth factor β (TGF-β), which in turn can provide additional survival signals [75]. Moreover, platelets have also been shown to induce EMT, a transcriptional program which is orchestrated by a number of embryonic transcription factors, that was also suggested to promote the expression of stem-like markers, conferring malignant traits such as invasiveness and resistance to apoptosis [22, 76]. Alternatively, CTC-associated platelets can release adenine nucleotides and facilitate extravasation by inducing an opening at the endothelial barrier [77], while macrophages that bind to CTCs can promote tissue remodelling and help extravasation through capillary walls in the lung [78]. Moreover, tumor-derived fibroblasts or endothelial cells that bind to CTC clusters could act as passenger stromal cells that form part of the tumor microenvironment at the distant site, therefore facilitating the initial growth of CTCs and ultimately promoting metastasis [74, 75].

Therefore, it is evident that in order to dissect the heterogeneity that is present in CTCs, additional studies that aim at profiling these cells at single cell resolution are necessary. In respect to identifying stem-like features in CTCs, it will be important to first identify robust expression markers that faithfully report the stem-like potential of these cells, as well as appropriate functional assays that assess their self-renewal and differentiation potential while also evaluating their metastatic propensity.