The concept of cell–cell fusion in cancer was already proposed more than a century ago by the German physician Otto Aichel
[1]. He assumed (a) that the fusion of tumor cells and cancer infiltrating leukocytes could be an explanation for aneuploidy and (b) that the combination of different chromosomes and their qualitative differences could lead to a metastatic phenotype (for review, see
[1][2][3][4][5][6][7]). Since then, the fusogenic capacity of cancer cells has been demonstrated in a variety of in vitro and in vivo studies. It was shown that cancer cells could either fuse with other cancer cells or normal cells, such as macrophages, fibroblasts, stromal cells, and stem cells, thereby generating tumor hybrid cells exhibiting an increased metastatogenic capacity, an enhanced resistance to chemo- and radiation therapy, and even properties of cancer stem/initiating cells (CS/ICs) (for review, see
[3][4][5][6][7][8][9][10][11]). Moreover, tumor hybrids have been clearly identified in human cancer patients with a former bone marrow transplantation (BMT) history
[12][13][14][15][16][17]. For instance, tumor hybrid cells with overlapping donor and recipient alleles were found in the primary tumor, lymph node metastases, and brain metastases of melanoma patients
[13][14][15]. Likewise, Y-chromosome positive tumor hybrid cells were found in the primary cancer and the circulation of female pancreatic adenocarcinoma patients, which received a BMT from a male donor
[17]. Moreover, the presence of circulating tumor hybrid cells as compared to normal circulating tumor cells was correlated to a statistically significantly increased risk of death in pancreatic cancer patients
[17], suggesting that the cancer cells’ malignancy was dramatically enhanced by fusion. Briefly, these human cancer data support the theory that (a) cell–cell fusion events really occur in human cancers and that (b) cell–cell fusion could give rise to tumor hybrid cells exhibiting an altered phenotype.
What sounds simple and promising on the one hand, is actually very complex. First, prospective CS/ICs are usually characterized by the expression of specific markers (for review, see
[18][19][20][21][22][23][24]), but no common CS/IC marker has been identified so far. Moreover, a few studies suggest some plasticity of cancer cells, such that non-CS/ICs can become CS/ICs
[39][40][41][42]. Second, the capacity of CS/ICs to induce tumor formation is studied in immunocompromised mice
[43]. This is advantageous since defined mouse strains guarantee comparable and reliable results. In contrast, the murine tumor environment is barely comparable with the human tumor microenvironment. Thus, human CS/ICs will respond differently to murine cytokines, matrix components, or murine cells, which all have an impact on the cells’ tumorigenicity. Moreover, the tumorigenicity of CS/ICs is also influenced by co-implantation of, e.g., matrix components and stromal cells
[19][44]. Third, the originally proposed CS/IC model cannot explain intra-tumoral heterogeneity, which is a hallmark of cancer
[45]. Instead of a clone of more or less identical CS/ICs, which were derived from the initial CS/IC by symmetric division, it is currently assumed that the CS/IC pool within a tumor is more heterogeneous, suggesting that different CS/IC subclones co-exist and drive tumor progression
[19][23].
2. Some Brief Facts about CS/ICs
Even though prospective CS/ICs have been characterized by expression of specific markers, such as CD24, CD44, CD90, CD133, and ALDH1 (for review, see
[18][19][20][21][22][23][24]), no common CS/IC marker has been identified so far and, for some CS/ICs, different markers have been proposed (
Table 1). For instance, CD133 has been suggested as a marker for glioblastoma, pancreatic, and prostate CS/ICs
[46][47][48], while putative breast CS/ICs have been designated as CD44
+/CD24
−/low [27]. Interestingly, mouse mammary Brca1 tumors contained distinct CD44
+/CD24
− and CD133
+ cells with CS/IC characteristics
[49], whereas three distinct triple negative breast CS/IC populations (both CD44
+/CD24
− and CD44
+/CD24
+ in estrogen receptor α-negative breast tumors, and CD44
+/CD49f
hi/CD133/2
hi) were tumorigenic in murine xenograft models
[50]. Hermann and colleagues identified two distinct prospective CS/IC populations in pancreatic cancer. CD133
+ CS/ICs were found in the center of the primary tumor, whereas CD133
+ CXCR4
+ CS/ICs were present in the invasive front of pancreatic cancer and determined the metastatic phenotype
[48]. However, a contrastingly different CD44
+CD24
+ESA
+ pancreatic CS/IC phenotype has been suggested by Li and colleagues
[29]. Epithelial cell-adhesion molecule (EpCAM) and CD44 have been suggested as more robust colon CS/IC markers
[31] than CD133
[51][52]. However, CD133
+CD44
+CD49
high colon cancer cells were highly tumorigenic in a study of Haraguchi and colleagues
[53].
Table 1. Prospective CS/IC marker.
| Cancer Type |
CS/IC Marker |
Reference |
| Breast cancer |
CD44+/CD24−/low |
[27][49][50] |
| CD44+/CD24+ |
[50] |
| CD44+/CD49fhi/CD133/2hi |
[50] |
| CD133+ |
[49] |
| ALDH1 |
[28] |
| SP cells |
[54] |
| EMT |
[55] |
| Colon Cancer |
CD44+EpCam+ |
[31] |
| CD133+ |
[51][52] |
| CD133+CD44+CD49high |
[53] |
| Glioblastoma |
CD133+ |
[46] |
| CD133+/CD133− |
[40][41] |
| SP cells |
[56] |
| Lung Cancer |
SP cells |
[57] |
| Malignant Melanoma |
CD271+ |
[58] |
| CD271+/CD271− |
[39] |
| SP cells |
[59] |
| Osteosarcoma |
SP cells |
[60] |
| Pancreatic Cancer |
CD44+CD24+ESA+ |
[29] |
| CD133 |
[48] |
| SP cells |
[61] |
In addition to the inconsistency of prospective CS/ICs markers, various studies revealed that non-CS/ICs could be as tumorigenic as CS/ICs and that CS/IC-related markers were reversibly expressed in non-CS/ICs and CS/ICs
[39][40][41][42]. For instance, CD271 has been suggested as a marker for melanoma CS/ICs
[58], but both CD271
+ and CD271
− melanoma cells were highly tumorigenic and metastatic in nonobese diabetic/severe combined immunodeficient (NOD/SCID) interleukin-2 receptor-gamma mice (NSG mice)
[39]. CD133
+ and CD133
− glioblastoma CS/ICs showed differential growth characteristics and molecular profiles, but both subtypes were similarly tumorigenic in nude mice
[41]. Interestingly, Wang et al. showed that CD133
− glioma cells were tumorigenic in nude rats and could give rise to CD133
+ cells
[40], which have been suggested as glioblastoma CS/ICs
[46]. While these findings point to a possible plasticity of cancer cells, these data again raise the reliability of certain markers in CS/IC research.
Side population cells (SP cells) represent another population of tumor initiating cancer cells
[57][62][63][64][65]. They are characterized by expression of ATP binding cassette (ABC) membrane transporters and the efflux of fluorescent dyes, such as Hoechst blue and Hoechst red, which are usually used for detection and isolation
[57][62][63][64][65]. SP cells have been identified in several tumor cell lines derived from, e.g., breast cancer
[54], lung cancer
[57], glioblastoma
[56], and pancreatic adenocarcinoma
[61], and human cancers, such as breast
[54], melanoma
[59], and osteosarcoma
[60] (
Table 1). While all these studies showed that SP cells were highly tumorigenic, it remains less clear whether SP cells are identical to the above-mentioned population of CS/ICs or represent a unique population of tumorigenic cancer cells. Studies on pancreatic cancer SP cells revealed that, at less than 1000 cells, tumor formation was initiated in nude mice, but that both non-SP cells and SP cells contained CD44
+CD24
+ and CD133
+ cells
[61]. As indicated above, CD44, CD24, and CD133 have been proposed as markers for prospective pancreatic CS/ICs
[29][48]. Likewise, no correlation between breast cancer SP cells and the prospective breast CS/IC phenotype CD44
+CD24
−/low was observed
[54]. About 3.4% SP cells were present in the population of MCF-7/HER2 breast cancer cells and only 100 of these MCF-7/HER2 SP cells sufficiently initiated tumor formation in NOD/SCID mice. However, solely 7.4% of MCF-7/HER2 breast cancer cells were CD44
+/CD24
−/low [54].
In 2008, two different studies suggested a link between epithelial-to-mesenchymal transition (EMT) and generation of normal stem cells and CS/ICs
[55][66] (
Table 1). The transition of sessile epithelial cells into motile mesenchymal cells is usually associated with cancer metastasis
[67][68][69][70], but, of course, is also mandatory for developmental processes during embryogenesis and wound healing
[71]. EMT is a reversible process and, hence, cells that have undergone EMT could revert to an epithelial state via mesenchymal-to-epithelial transition (MET)
[70]. Several EMT-inducing triggers have been identified, such as transforming growth factor-β (TGF-β), WNT proteins, cytokines, growth factors, and hypoxia, which leads to the expression and functional activation of various master EMT regulators, specifically EMT-inducing transcription factors (EMT-TFs) and microRNAs
[68][72]. In this context, two core EMT circuits/feedback loops have been identified, consisting of SNAIL-miR-34 and ZEB1-miR-200
[73], which are accompanied by additional EMT-TFs, including SLUG, TWIST, and YAP/TAZ, as well as post-translational modifications and splicing
[71][72]. Interestingly, mathematical modeling of the two core EMT circuits revealed that cells could attain either an E, or an M or an E/M phenotype depending on their SNAIL expression levels
[73]. It is this E/M phenotype, which has also been named hybrid E/M state
[73][74][75] or quasi-mesenchymal
[68], that likely possesses CS/ICs properties. For instance, single CD24
+/CD44
+ HMLER cells (human mammary epithelial cells immortalized and transformed with hTERT, SV40LT, and RAS oncogenes
[76]) exhibited a hybrid E/M phenotype and possessed increased stem-like properties, such as an enhanced population of ALDH1+ cells and mammosphere formation capacity
[74]. Likewise, tumorigenic cancer cells expressing both E- and M-specific markers have been found in xenografts isolated from breast
[77] and ovarian tumors
[78]. Data of Kroger and colleagues further indicated that the hybrid E/M phenotype in HMLER cells is likely controlled by SNAIL, whereas the M phenotype is driven by ZEB1
[79]. HMLER cells that were trapped in a hybrid E/M state (ZEB1 knock-out, SNAIL expression) produced much bigger tumors and displayed a 37-fold higher tumor-initiating frequency as compared to appropriate controls
[79]. In this context, integrin-β4 (CD104) has been suggested as a marker for the hybrid E/M state in triple negative breast cancer. Indeed, human breast cancer cells with an intermediate level of integrin-β4 expression exhibited a hybrid E/M state and were more tumorigenic than breast cancer cells in the E-state or M-state, respectively
[80].
Tumor initiation and self-renewal of prospective CS/ICs is usually studied in the so-called serial transplantation assay, which, however, is still imperfect, since human cancer cells are growing in a nonhuman environment, namely immunocompromised mice
[43]. Hence, it remains unclear whether tumor formation was truly related to CS/ICs or to cancer cells, which adopted best to the foreign murine environment. Moreover, the tumor initiation capacity of prospective CS/ICs is strongly related to the used mouse model (detection of tumorigenic cells is several orders of magnitude higher in NSG mice than in NOD/SCID mice
[81]) and co-implantation of matrix components and additional tumor propagating cells
[82]. It is well known that the right composition of the tumor microenvironment and the CS/IC niche, including matrix components, stromal cells, immunocompetent cells, hypoxia, and cytokines, potentially contribute to stemness and an enhanced tumor initiation capacity
[19][44]. In any case, CS/ICs were clearly identified in genetically engineered mouse models of brain, skin, and intestinal cancers
[37][83][84], which are, to date, the strongest evidence that CS/ICs exist and initiate tumor growth.
In summary, the biology of CS/ICs is complex. Although various strategies and protocols have been developed over the past two decades to identify and characterize potential CS/ICs in human cancers, it is still not clear which method and which characterization is the best and most reliable. Moreover, the “classical CS/IC” model fails in intratumoral heterogeneity, which is a hallmark of cancer
[45]. Based on the classical CS/IC model, there should be a population of CS/IC in the tumor that arose by self-renewal and should, therefore, be phenotypically similar. However, this is not the case. Instead, different subclones are likely found in the tumor, each of them most likely originating from one subclone-specific CS/IC
[19][23].
However, how does this pool of subclone-specific CS/IC evolve? One possible mechanism could be the accumulation of genetic mutations in one CS/IC that could lead to the evolution of subclone-specific CS/ICs
[23]. Likewise, it cannot be ruled out that subclone specific CS/ICs evolve independently of the original tumor-initiating CS/IC. As mentioned above, some cancer cells appear to exhibit some plasticity, so that non-CS/ICs can give rise to CS/ICs
[39][40][41][42], which could be the result of a microenvironmental cue or a (epi-)genetic change
[19]. The hybrid E/M state model assumes that cancer cells could acquire a CS/IC phenotype through EMT
[68][73][74][75]. Since EMT is induced in several cancer cells it can be further assumed that a number of hybrid E/M state cancer cells will be generated exhibiting prospective CS/IC properties. Thus, the hybrid E/M state model suits better to intratumoral heterogeneity and the necessity of individual CS/IC subclones.
Cell–cell fusion has been suggested as another possible mechanism by which prospective CS/ICs may arise
[10][36][85][86][87][88][89][90]. Thereby, CS/ICs might originate from the fusion between stem cells and somatic cells
[90] or tumor cells and normal cells, such as macrophages and stem cells
[10][36][85][86][87][88][89][90]. A summary of the different ways that CS/ICs could arise is given in
Figure 1.
Figure 1. CS/ICs can arise through different mechanisms. (A) CS/ICs could originate from normal stem cells (SCs), progenitor cells (PCs), or through cell–cell fusion. The red lightning should indicate mutational events, which drive malignant transformation. (B) Mutational events could lead to a new CS/IC subclone. (C) Differentiated cancer cells could convert into CS/ICs due to an inherent plasticity. (D). CS/ICs could fuse with normal somatic cells (NC), such as stem cells (SC) and macrophages (Mϕ). (E) Induction of EMT in cancer cells might also give to a CS/IC phenotype.