Could gastrointestinal tumor-initiating cells originate from cell-cell fusion in vivo?

Tumor-initiating cells (TICs) or cancer stem cells are believed to be responsible for gastrointestinal tumor initiation, progression, metastasis, and drug resistance. It is hypothesized that gastrointestinal TICs (giTICs) might originate from cell-cell fusion. Here, we systemically evaluate the evidence that supports or opposes the hypothesis of giTIC generation from cell-cell fusion both in vitro and in vivo. We review giTICs that are capable of initiating tumors in vivo with 5000 or fewer in vivo fused cells. Under this restriction, there is currently little evidence demonstrating that giTICs originate from cell-cell fusion in vivo. However, there are many reports showing that tumor generation in vitro occurs with more than 5000 fused cells. In addition, the mechanisms of giTIC generation via cell-cell fusion are poorly understood, and thus, we propose its potential mechanisms of action. We suggest that future research should focus on giTIC origination from cell-cell fusion in vivo, isolation or enrichment of giTICs that have tumor-initiating capabilities with 5000 or less in vivo fused cells, and further clarification of the underlying mechanisms. Our review of the current advances in our understanding of giTIC origination from cell-cell fusion may have significant implications for the understanding of carcinogenesis and future cancer therapeutic strategies targeting giTICs. ©The Author(s) 2021. Published by Baishideng Publishing Group Inc. All rights reserved.

Core Tip: Currently, there are many controversial hypotheses concerning the generation of gastrointestinal tumor-initiating cells (giTICs). Here, we mainly review the current advances in the understanding of giTIC origination from the cell-cell fusion of cancer cells and bone marrow-derived cells.

INTRODUCTION

Tumors are composed of cells with different levels of differentiation, and tumor-initiating cells (TICs) are the least differentiated cancer cells, which are then capable of giving rise to other cancer cells[1,2]. TICs are the source of gastrointestinal tumor initiation, progression, metastasis, and drug and radiation resistance. Moreover, they are capable of self-renewal, can differentiate into multiple cell lineages (such as cancer cells), and can undergo asymmetric cell division. TICs are the most carcinogenic subpopulation of cells in most cancer types[3-5], including gastrointestinal cancers[6]. The origin of TICs remains unknown; however, many hypotheses[7] have been proposed to explain it, including those involving gene mutations[8], endogenous reprogramm-ing[9,10], and cell-cell fusion[11-14].

Gastrointestinal TICs (giTICs) may originate from gene mutations[15]. Some hypothesized that gastrointestinal stem cells, similar to other types of stem cells, have protective mechanisms that reduce tumorigenesis. These mechanisms include asymmetric cell division via chromosomal segregation and relatively slow cell cycles[16], which can protect cells from DNA damage and cellular stress[17]. To form giTICs, these mechanisms must be circumvented. The development and progression of colorectal cancer (CRC) are associated with a number of identified gene mutations, in genes such as KRAS, adenomatous polyposis coli (APC), and p53, that promote the conversion of normal epithelial mucosal tissue to cancerous tissue[18,19]. The tumor suppressor gene p53 ensures the genomic stability of stem cells, and can therefore act as a barrier to the formation of TICs[20]. Wild-type p53 can be experimentally replaced with a mutant version of p53 via PCR, CRISPR/Cas9, and knock-in techniques. When a related gene mutation occurs, p53 loses its tumor-suppressing ability and acquires additional carcinogenic capabilities. This process is termed as mutant p53 gain of function (GOF). Experimental evidence suggests that mutant p53 GOF can mediate cancerous properties, such as cell death resistance, sustained proliferation, metastasis and invasion, and tumor-promoting inflammation[21-23]. Mutant p53 is highly expressed in colorectal TICs and CRC tissues[8]. Most evidence that supports this hypothesis arises from the observation that common mutations in CRC would affect normal stem cell behavior. For example, deletion or inactivation of the APC gene is often the initiating step in colorectal carcinogenesis[18] and as such, acts as a gatekeeper in CRC. The absence of APC is rare and APC is commonly found in gastrointestinal cells, including normal populations of gastrointestinal stem cells, as it plays a major role in regulating normal stem cell function[24]. There is little direct evidence demonstrating that giTICs originate from gene mutations in stem cells. Regardless, it is generally believed that giTICs originate from mutated stem cells because stem cells are long-lived gastrointestinal cell types. Thus, there is sufficient time for them to accumulate oncogenic mutations[19]. In addition, TICs and normal stem cells have many identical or similar properties, indicating that they have a common source or originate from the same ancestor.

Another hypothesis is that giTICs may originate from endogenous reprogramming. A specific combination of transcription factors can reprogram differentiated cells into pluripotent stem cells[25]. Following the same reasoning, gastrointestinal epithelial cells can be dedifferentiated into progenitor/stem cells via specific matched signal transduction pathways. Notably, bidirectional transformation between TICs and non-TICs was observed in intestinal tumors. Nuclear factor kappa-B (NF-κβ) induces the stabilization of β-catenin and activation of the β-catenin/T-cell factor transcription complex, which, together with the cancer-causing Kras, can induce dedifferentiation of non-stem colon cancer cells into stem-like cancer cells[9,26] or TICs[27,28]. However, the mechanisms underlying their regulation remain unclear[28]. Epithelial-mesenchymal transition (EMT) may also be involved in endogenous reprogramming[29] by inducing overexpression of the transcription factors Snail[30-33] and zinc finger E-box-binding protein 1 (Zeb1)[34-37]. It is worth noting that Zeb1 activation is associated with Slug (Snai2) in TICs[36]. Zeb1, a transcription factor known to be involved in EMT, is necessary for the conversion of non-TICs to TICs. EMT in TICs also induces the expression of CD44, which was shown to be highly expressed in giTICs[36].

Cell-cell fusion can be easily induced in vitro by physicochemical or biological molecules but also occurs in vivo, such as the fusion of sperm and egg cells. Cell fusion is an essential physiological process, which plays a role in fertilization, virus entry, muscle differentiation, and placenta development. It was also reported to be closely associated with the occurrence and development of cancer. Fused cells display the genotype and phenotype of the maternal cells, and hybrids produced by the fusion of different cell types have distinct properties. Cell-cell fusion can be identified by cell size and shape, karyotypes, DNA, gene expression, cell-specific markers, and other properties. Both fused cells and TICs display aneuploidy, such as being tetraploid, and chromosomal instability, as well as have the ability to induce metastasis and drug resistance[38], which suggests that cell-cell fusion may produce TICs. In other words, cell-cell fusion may be a better explanation of TIC generation than the aforementioned conventional gene mutation and endogenous reprogramming hypotheses. In addition, cell-cell fusion may play a role in giTIC formation by introducing endogenous reprogramming, as cell fusion hybrids retain transcripts from both parental cells and also express a unique subset of transcripts[39].

Cell-cell fusion in vivo and tumor-initiating capacity in vivo should be the criteria used to determine whether giTICs originate from cell-cell fusion. Non-tumor initiating cancer cells can also proliferate and generate tumors when enough of such cells are used. However, theoretically, only TICs can initiate tumor formation using a limited number of cells. Generally, unsorted cancer cells contain both TICs and non-TICs. Therefore, it is difficult to determine which cells are responsible for tumor initiation. Here, we review giTICs that can initiate tumors in vivo using 5000 or fewer in vivo fused cells, as well as tumor-initiating like cells (TILCs) that can generate tumors using more than 5000 fused cells. However, we do not exclude the possibility that more than 5000 hybrid cells may be needed to initiate tumors when cell-cell fusion occasionally induces rare genetic changes that lead to tumor development.

GiTICs originate from cell-cell fusion: Supporting and opposing evidence

The fusion of human cells in vivo was confirmed by reports describing the fusion of melanoma cells and osteoclasts in 2007 and the fusion of BRAF mutated melanoma and stromal cells in 2016[40,41]. The fusion of macrophages and peripheral blood melanoma cells, which was discovered in 2015, also provided evidence for human cell fusion in vivo. Moreover, studies have shown that fusion cells exhibit high expression levels of cell fusion factors, including the cell fusion molecule chemokine receptor 4 (CXCR4), as well as that fusion cells may cause tumor metastasis. However, in the tumorigenic experiments described, 5 × 106 fusion cells were inoculated into mice to generate primary tumors at a number that was much higher than the 5000 cells criteria used for in vivo tumor generation. Therefore, the fusion cells were not concluded to be TICs when using the aforementioned restrictions[42].

Currently, there are no reports of giTICs originating from cell-cell fusion in vivo[43]. We summarize the reports regarding giTICs originating from cell-cell fusion, including the fusion of gastrointestinal cells with various cell types, the study methods used (in vitro or in vivo), evidence of cell-cell fusion, tumorigenic and tumor-initiating properties of the fused cells, and possible mechanisms of cell-cell fusion (Table 1).

Table 1

Tumor initiating cell origination from cell-cell fusion

Cell I	Cell II	Cell fusion method	Cell fusion evidence	Tumor initiating method	Tumor initiating evidence	Mechanism	Ref.
Colorectal cancer cell	Human dendritic cell	In vitro. DCs and cancer cells fused using PEG	Purified cells		Enhanced induction of antigen-specific CTL	Streptococcal preparation OK-432 promotes fusion efficiency	[69]
Human esophageal carcinomas cell	Human dendritic cell	In vitro. ECs and DCs fused using PEG	Co-expression of MHC class II, CD86, and MUC1		Induced specific antitumor response		[55]
Human gastrointestinal cancer cell	Human dendritic cell	In vitro. Fusion via PEG and electroporation			Th1/Th2 and Tc1/Tc2 balance improved		[73]
Human gastric cancer cell	Human dendritic cell		FACS analysis		Induced CTLs, reduced metastases		[56]
Human gastric cancer cell (HGC-27 or SGC-7901)	hucMSC	In vitro. GC-DIO and hucMSCs-DID fusion using PEG	Double positive cells	BALB/C nude mice (2 × 106 cells)	In vivo. Tumors formed from fused cells		[44]
Human breast cancer cell (MDA-MB-231)	Human MSC	In vivo. 2 × 106 MSC300415-GFP and 2 × 106 MDA-MB-231-cherry subcutaneously injected into 5 female NOD/SCID mice	Hybrid cells GFP/cherry fluorescence	1 × 106 MDA-hyb3-GFP/cherry cells were injected subcutaneously into 3 female NOD/SCID mice	In vivo. Tumors formed from fused cells		[80]
Human colon adenocarcinoma cell	Human HeLa cell (D98OR)	In vitro. Fused using PEG, isolation of hybrid cells by selecting isolated colonies	Flow cytometry analysis had more DNA than expected. A range of 71–140 chromosomes		Fusion cell characteristics were consistent with cancer cells		[75]
Human colon cancer cell (SW480)	Human dendritic cell	In vitro. Tumor cells- PKH26-red and DCs- PKH67-green fused using PEG	Dual red and green fluorescence and highly expressed CD80, CD86, and HLA-DR	CD8+ T lymphocytes co-cultured with dendritic cells at a ratio of 10:1	CTLs were activated to proliferate and the number of T cells increased		[55]
Human colon cancer cell (SW620)	Human dendritic cell	In vivo. DCs and tumor cells fused using PEG	Fusion efficacy was evaluated by FM and FC	In vivo. 1 × 107 fusion hybrids injected intraperitoneally	Cellular immune responses, significant inhibition of tumor growth		[55]
Human colon carcinoma line (VACO-411)	Human breast cancer line (MCF-7)	In vitro. VACO-411 (1 × 106 cells) and MCF-7 (1 × 106 cells) fused using PEG	Morphology of VACO-411 × MCF-7 fused cells	In vitro. The fused cells were treated with TGF-β	Fusion cells were inhibited by TGF-β		[76]
Human colon epithelial cancer cell	Human normal colon cell	In vitro. Cancer cells and normal cells (1:10) fused using PEG	Comparison of DNA synthesis (P < 0.01)	Male mice nu/nu injected subcutaneously with1 × 106 fused cells	The fused cells could not grow into tumors		[77]
Human colorectal carcinoma cell	Human dendriticcell	In vitro. DCs-CMFDA-green,colorectal carcinoma-CMTMR-red cells fused using PEG/electrofusion	Double-positivecells		Efficiently activated autologous tumor-specific T cells		[68]
Human esophageal cancer cell (EC109)	Human dendritic cell	In vitro. DCs and ECs (5:1-10:1) fused using PEG	Co-expression of MHC-CiaSSII and CD86 and MUC1 antigens	Cytotoxic T lymphocytes	Antitumor capabilities		[60]
Human esophageal cancer cell (EC9706)	hucMSC	In vitro. ECs-DiO hMSCs- DiD fused using PEG	Double positive hybrids are yellow and multinuclear	In vivo. Xenograft assays in immunodeficient mice	Both ECs and their self-fusion groups developed tumors
Human esophageal carcinoma cell (EC9706)	Human hemopoietic stem cell	In vitro. ECs and HSCs (10:1) fused using PEG	CD34+CD38-Scal+ cells isolated using immunomagnetic beads; HSCs cannot grow in DMEM supplemented with 10% FBS	In vivo. 5 × 105 fused cells injected into 12 NOD/SCID mice	All mice formed tumors; however, the tumor weight of the fused cell group was lower than that of the EC9706 group		[54]
Human esophageal carcinoma cell	Human dendritic cell	In vitro. DCs and ECs (5:1) fused using PEG; incubated with FA-FITC and CD80-PE	Analysis using FATICan	In vivo. Fusion vaccine (0.2 mL; 1 × 106 cells) injected	Anti-tumor effects		[58]
Human esophageal carcinoma cell (EC109)	Human dendritic cell	In vitro. DCs and ECs (5:1) fused using PEG. FA-FITC CD80-PE mAbs-CD80, CD83 and CD86	FC	In vitro. Cytotoxicity assays	Antitumor activity		[59]
Human female pancreatic adenoepithelial neoplasm cell	Human male BMDC	In vivo. 4 female pancreatic cancer patients transplanted with male BMDCs	Peripheral blood cell: EpCAM (yellow)/CD45 (green), Y chromosome, CK+/CD45+, MФ proteinsCD14, CD16, CD11c, CD163 MUC4				[43]
Mouse colon cancer cell (MC38)	Mouse R26R- YFP Cre mice	In vivo. MC38 cells were injected ventrally into r26R-YFP Cre mice	RFP+ YFP+				[51]
Mouse primary melanoma cell	Mouse MФ	Melanoma cells were injected into mice intradermally	RFP and GFP cells	300 and 3000 cells, respectively injected into mice (n = 9, 3)	Tumor initiation	The characteristics of parental cells provided adhesive affinity	[51]
Human gastric cancer cell (MKN-1)	Dendritic cell	In vitro. DCs- PKH-26 and GC cells-PKH-67 fused via PEG/electrofusion	Double positive cell populations	In vitro. Co-cultured DCs (1 × 105 cells) with 1 × 106 T cells	Induced tumor antigen-specific CD8+ T cells		[70]
Human gastric epithelial cell (GES-1)	CM-MSC	In vitro. GES-1- PHK-26 (2 × 106 cells) and CM-MSCs- CFSE (1 × 107) cells fused using PEG	Most cells express PKH26 and CFSE	In vivo. The fused cells (1 × 107 cells) were injected into 8 BALB/c nude mice	Tumors from the fused cells formed in vivo		[47]
Human gastric cancer cell (SGC7901)	Human dendritic cell	In vitro. SGC7901 and DCs fused using PEG	Pure fused cells were obtained by selective culture with HAT/HT culture system	In vivo. Fusion cells (5 × 108) were injected into BALB/c mice	In vivo. The fused cells could not grow into tumors		[61]
Human gastric cancer cell (SGC7901)	Human dendritic cell	In vitro. SGC7901 and DCs fused using PEG	Selective culture with the HAT/HT culture system	In vivo. Fused cells (5 × 106/mL) + T cells (5 × 106/mL)	In vivo. The fused cells could not grow into tumors		[62]
Human hepatobiliary stem/progenitor cell	Human hematopoietic precursor-derived myeloid intermediate	In vitro. Cultures of CD34+ LTICs and xenograft cells (the xenografts were produced by CD34+ hybrid cells)	CD34+ LTICs co-expressed liver stem cell and myelomonocytic cell markers			HSPCs were fused with a CD34+ hematopoietic precursor-derived myeloid intermediate to form CD34+ hybrid cells
Human hepatocellular carcinoma cell (HepG2)	Human embryonic stem cell	In vitro. HepG2-red mitochondrion selective probe and hESCs-Oct-GFP cells fused via laser-induced single-cell fusion	Transfer of cytoplasmic GFP from hESCs to HepG2 cells	In vivo. HepG2 cells and the fused cells (5 × 104 and 1 × 105 cells, respectively) were injected into nude mice	Tumors were generated from fused cells
Human hepatocellular carcinoma cell (HepG2)	Mouse MSC	In vitro. MSCs- DiI (5 × 105 cells) and HepG2-eGFP (1 × 105 cells) fused using PEG	Dual fluorescence, two nuclei	In vivo. The fused cells were injected into 7 nude mice/group with 2.4 × 107 cells/group	Tumors were formed from fused cells
Human intestinal cancer cell (HT-29)	Human MSC	In vitro. PM7-eGFP and HT-29-DsRED cells were cocultured	eGFP and DsRED double positive cells		Acquired epithelial characteristics		[51]
Human intestinal epithelial cells	Human hematopoietic cell	In vitro. X- and Y-chromosome determined by FISH. Female recipients of hematopoietic cell transplant from male donors	Stained for X- (green) and Y- (red) chromosomes and Lamin B1 (white)			In mice, hematopoietic fusion with non-hematopoietic cell types occurs endogenously in the absence of disease	[86]
Mouse intestinal epithelial cell	Mouse bone marrow-derived cell	In vivo. CMV-CreGFP+ mice BM were transplanted into iDTR mice	Co-staining for GFP and EpCAM.GFP+ cells in the intestine		Cell fusion is dispensable for tissue homeostasis		[52]
Mouse intestinal stem cell	Human bone marrow-derived cell	In vivo. Donor female mice BMDCs-GFP, male recipient mice	EGFP expression in all principal intestinal epithelial lineages				[63]
Mouse colon adenocarcinoma cell (CT26)	Mouse dendritic cell	In vitro. Tumor cells- PKH67-Green and DCs fused using PEG	Assessedvia the trypan-blue exclusion test	In vivo. BALB/c mice injected with 5 × 105 cells	The fused cells could not generate tumors		[64]
Mouse colon adenocarcinoma cell line (CT26)	Mouse dendritic cell	In vitro. DCs-anti-CD11cmAb and tumor cells- CFSE fused using PEG	Analyzed by FC	In vivo. Injection of 1 × 104, 105, or 106 cells/mouse	The fused cells did not generate tumors. CTL anti-tumor effects		[72]
Mouse colon carcinoma cells (CT26CL25)	Allogeneic and semi allogeneic dendritic cells	In vitro. DCs-PKH26-red andCT26CL25-PKH67-green fused using PEG	Analyzedby FC	In vivo. 1 × 106 fused cells and 5.0 × 105 CT26CL25 cells	Anti-tumor effects in vivo		[65]
Mouse colon epithelial cell	Mouse BMDC	In vivo. Female mice BMDCs-GFP (1 × 107 cells) transplanted into irradiated male mice	Co-expression of GFP and the Y chromosome	In vivo. Parabiosis surgery (GFP and ROSA mice)	Bone-marrow/epithelial cell fusion causes genetic reprogramming	Inflammation and proliferation act together to mediate intestinal cell fusion	[87]
Mouse gastric epithelial cell	Mouse BMDC	In vivo. Male irradiated C57BL/6 mice received female C57BL/6 mice BMDC-GFP	Direct. Positive for the Y chromosome and expressed GFP as determined by FM	In vivo. GCs were induced with a carcinogen	Tumor formed from fused cells in vivo	Chronic inflammation (adenocarcinoma, glandular stomach, not squamous cell carcinoma)	[48]
Mouse hepatocellular carcinoma cell	Mouse dendritic cell	In vitro. HCCs PKH-26-red and DCs-PKH-2-green fused using PEG	The fusion cells were yellow under the confocal microscope	In vitro. CTL assay	In vitro. Activated cytotoxic T lymphocytes		[66]
Mouse hepatoma cell line (Hepa1-6)	Mouse embryonic stem cell	In vitro. Cancer cells-GFP and ES cells-RFP fused using PEG	Double fluorescence-positive	In vivo. 1 × 106 ES-cancer fused cells injected into nude mice	Tumor formed from fused cells in vivo
Mouse intestinal epithelial cancer cell	Mouse WBM (macrophage)	In vivo. WBM-GFP (5 × 106 cells) injected in recipient mice (male WT, ApcMin/+, ROSA26, ROSA26/ApcMin/+). Parabiosis	Co-localization of GFP (green) and β-galactosidase (red)		Nuclear reprogramming	Fusion between circulating blood-derived cells and tumor epithelium origin at the natural course of tumorigenesis	[39]
Mouse intestinal epithelial cells (IEC-6). Human cervical adenocarcinoma cells (HeLa)	Mouse intestinal epithelial cells (IEC-6) Human cervical adenocarcinoma cells (HeLa)	In vitro. IEC-6- CFSE andIEC-6- SNARF-1 (HeLa -Cy3-dUTP-red and HeLa- Cy5-dUTP-green) fused using PEG	The fused cell emits both CFSE and SNARF-1 fluorescence (IEC-6). Eight daughter cells contain both dyes (HeLa)	In vivo. The IEC-6 fused cells (Two million cells) were injected in 18 mice	Tumor formed from the fused cells in vivo (n = 11 generated tumors)		[78]
Mouse intestine stromal cell	Mouse bone marrow-derived macrophage	In vivo. Female mice BMDCs-GFP injected in male recipient mice	Co-localization of GFP and Y-chromosome		Organ fibrosis	Depleting macrophages genetically reduced the number of cells	[53]
Mouse prostate cancer cell (PCa)	Mouse BMDC	In vivo. 2 × 106 cells/mice BMDCs-GFP transplanted into 10 C57BL/6 mice	Co-expression of GFP and CK8	C57BL/6 mice-GFP, induced prostate cancer by MNU	GFP-positive cells in the prostate cancer tissue		[79]
Whole tumor cell	Human dendritic cell	In vitro. The purified DCs and tumor cells fused using PEG					[67]

BMDC: Bone marrow-derived cell; CAM: Cell adhesion molecules; CM-MSCs: Cord matrix-derived mesenchymal stem cells; CTL: Cytotoxic lymphocytes; DC: Dendritic cell; DMEM: Dulbecco's modified eagle medium; EC: Esophageal carcinoma; FACS: Fluorescence-activated cell sorting; FBS: Fetal bovine serum; FM: Fluorescence microscope; FC: Flow cytometry; GC: Gastric cancer; hucMSCs: Human umbilical cord mesenchymal stem cells; HSPC: Hematopoietic stem and progenitor cell; hESC: Human embryonic stem cell; HSC: Hematopoietic stem cell; HCC: Hepatocellular carcinoma; HLA: Human leukocyte antigen; LTICs: Liver tumor-initiating cells; MSC: Mesenchymal stem cell; MNU: N-methyl-N-nitrosourea; PEG: Polyethylene glycol; WBM: Whole bone marrow; WT: Wild-type; IEC: Intestinal epithelial cell; HGC: Human gastric cancer cell; GFP: Green fluorescent protein; MDA: Malonaldehyde; TGF: Transforming growth factor; FITC: Fluorescein isothiocyanate; PE: Physical examination; CM: Chylomicron.

Cell fusion between gastric cells and stem cells may generate giTICs

Xue et al[44] fused DIO-labeled (green) HGC-27 gastric cancer cells with DID-labeled (red) human umbilical cord mesenchymal stem cells (hucMSCs) using polyethylene glycol (PEG) 1500 in vitro. The fused cells with double nuclei were then stained with Hoechst 33342 (blue) and DIO-GC and DID-hucMSC double labels (yellow) were observed after 7 d by sorting via flow cytometry. Then, 20 male BALB/C nude mice were injected subcutaneously with 2 × 106 gastric cancer cells or fused cells. Mice in the fusion group exhibited tumor nodules at 4 d post-injection, while mice in the gastric cancer group showed no tumor nodules. Moreover, the fusion cells were shown to form more colonies than their parental cells and had higher Cyclin D1 and proliferating cell nuclear antigen (PCNA) expression levels. Cyclin D1 and PCNA expression in tumor tissues is usually positively correlated with cancer cell proliferation. The expression levels of the stem cell transcription factors Sox2, Oct4, Nanog, and Lin28, as well as those of the cancer cell markers CD133 and CD44, were also shown to be increased in the fused cells[44]. In addition, real-time RT-PCR analysis revealed that E-cadherin mRNA expression was decreased in fused cells, whereas that of mesenchymal markers, such as α-SMA, FAP, vimentin, snail, N-cadherin, slug, and twist, was significantly increased, indicating that the fused cells underwent EMT. EMT is associated with the metastatic ability and invasiveness of cancer cells. As such, the obtained fusion cells were shown to have EMT properties, which is similar to TICs[44-46]. Cell fusion in vitro between gastric epithelial cells and MSC also resulted in fusion cells with tumorigenic capabilities that underwent EMT[47]. However, these hybrid cells were formed in vitro and the number of cells used for the tumorigenic experiments was much higher than 5000 cells.

In a report by Yan et al[48], the bone marrow of green fluorescent protein (GFP) transgenic female C57BL/6 mice was transplanted into irradiated male homologous mice (68/68), all of which survived. Then, the transplanted bone marrow-derived cells (BMDCs) became the main bone marrow cells of the chimeric mice. Tumors were induced using the tumor-causing drug 3-methylcholanthrene. Three of the 12 treated mice successfully developed tumors. Hematoxylin & eosin staining showed two diffuse-type carcinomas in the glandular stomach and one squamous cell carcinoma (SCC). Analysis of CK-18 (mostly expressed in epithelial cancer cells and determined via immunohistochemistry staining) and GFP expression (fluorescence microscopy) showed that cells derived from both cancer types were positive for CK-18 and GFP expression, indicating that they are epithelial tumors originating from BMDCs. Moreover, co-expression of the Y chromosome and GFP in the cytoplasm was detected in a large number of adenocarcinoma cells via fluorescence in situ hybridization (FISH) and immunofluorescence microscopy. In SCC tissues, GFP expression was mainly detected in the interstitium and keratin pearl, but FISH did not detect the presence of the Y chromosome. Instead, the Y chromosome and GFP were co-expressed in the epithelial cells surrounding the SCC. Gastric cancer may originate from the BMDCs of transplant donors and develop initially via trans-differentiation and then cell-cell fusion[48-50]. These authors revealed that BMDC-gastric epithelial cell fusion may contribute to the renewal of the gastric mucosa and lead to increased carcinogenesis potential. Additionally, the aforementioned experiments demonstrated that fusion cells exhibit stem cell and cancer cell markers in chemical-induced tumor tissues in vivo but did not prove that fusion cells can initiate tumors. As such, it is not possible to distinguish between drug-induced tumors or fusion cell-induced tumors.

Cell fusion between CRC cells and stem cells

In a previous study[51], researchers directly co-cultured PM7 cells, which are eGFP-labeled bone marrow-derived MSCs, with the DsRED-labeled colon cancer cell line HT-29. After co-culture, some cells showed eGFP and DsRED double-positive labels and these fused cells were shown to be positive for epithelial-specific antigen (ESA) and cytokeratin expression. However, the authors did not investigate the tumorigenic capacity and other stemness properties of the fused cells. Other reports have found similar results[52,53].

Notably, a study revealed that the fusion of intestinal epithelial cancer cells and macrophages from BMDCs in vivo leads to nuclear reprogramming and the authors suggested that the fusion cells may play a role in tumor development and metastasis[39].

Cell fusion between liver cancer cells and stem cells may generate giTICs

In a previous study, human embryonic stem cells (hESCs) were labeled with Oct-GFP and HepG2 hepatocytes and stained with a mitochondrial (red) probe, and the cells were then fused via laser-induced fusion. Later, it was shown that GFP was transferred from hESCs (green) to liver cancer cells (red), confirming the successful generation of fusion cells. Subsequently, different amounts of 5 × 104 - 1 × 106 fused cells were injected subcutaneously into nude mice, and mice injected with normal liver cancer cells were used as controls. The fused cell group exhibited a tumor incidence of 9/12, while the liver cancer cell group had a tumor incidence of only 1/8. Moreover, a lower number of fused cells were necessary for tumor generation when compared to the liver cancer cell group. These results demonstrated that in vitro cell fusion between liver cancer cells and stem cells could generate cells with giTILC properties. The tumorigenicity of the fusion-generated giTILCs was also shown to be significantly higher than that of the maternal cancer cells. However, the number of cells used for tumorigenicity experiments was higher than 5000 and cell-cell fusion was induced in vitro. Similar reports have also shown that in vitro cell-cell fusion produces tumorigenic hybrid cells or giTILCs.

CD34+ hybrid cells extracted from liver cancer cell lines were shown to express high levels of hepatic stem cell and bone marrow mononuclear cell markers. The cells were also shown to be drug-resistant and express some TIC markers. As such, these results suggested that liver TILCs may be formed by the fusion of hepatobiliary stem/progenitor cells and hematopoietic precursor-derived myeloid cells.

Evidence opposing or not supporting the hypothesis that giTICs originate from cell-cell fusion

The fusion of gastrointestinal cells and dendritic cells does not produce giTICs or giTILCs, which is generally used to generate tumor vaccines[54-73]. In addition, it was shown that cell fusion in the pancreas and esophagus did not generate giTICs or giTILCs[54,74]. Moreover, in vitro fusion cells obtained from human colon adenocarci-noma cells and metastatic human cervical cancer HeLa cells were shown to possess cancer cell properties but were not considered to be giTICs or giTILCs[75]. Currently, cell-cell fusion between tumor cells and tumor cells has not been shown to lead to the generation of TICs[76]. Notably, cell-cell fusion between human colon epithelial cancer cells and normal colon cells not only fail to induce TICs or TILCs but also to generate tumorigenic hybrids in some cases[77]. Similarly, cell-cell fusion between intestinal epithelium cells did not generate giTICs or giTILCs[78].

TIC generation from cell-cell fusion in other tissue types

Gast et al[43] intradermally injected mouse primary melanoma cells (RFP+, actin–green fluorescent protein, 5 × 104 cells) into mice with GFP+ macrophages (actin–green fluorescent protein, n = 12). This resulted in tumor formation and mouse macrophages (MФ, GFP) and melanocyte fusion cells (RFP+GFP+) were detected in the tumors. The researchers then implanted 300 RFP+GFP+ cells, which were isolated by fluorescence activated cell sorting (FACS), into 19 recipient mice (300 cells per mouse) and found that the fusion cells led to tumor growth. Then, 3000 fusion cells per mouse were implanted into three mice for time-dependent analysis of tumor growth characteristics. It was found that the fusion hybrids obtained in vivo had different tumor growth rates, which indicated that the obtained hybrid cells had heterogeneous growth abilities. Therefore, the authors demonstrated that melanoma TICs originate from cell-cell fusion in vivo[43]. Notably, MФ-tumor fusion cells were found in the peripheral blood and were shown to have a stronger ability to metastasize and proliferate. Moreover, the authors showed that the presence of hybrid cells in the peripheral blood of female pancreatic cancer patients with bone marrow transplants from male donors was correlated with disease stage and patient survival.

Xie et al[79] reported that glioma stem cells reorganized the inflammatory microenvironment at the implanting site in mice. Cell-cell fusion between glioma cells and immunoinflammatory cells was also demonstrated in vitro and the fusion cells were shown to be tumorigenic in nude mice and have TILC characteristics.

The formation of cancer cell/MSC hybrids was observed in breast and prostate cancers. Researchers transplanted stem cells into experimental mice and identified in situ CK8+ prostate tumors derived from GFP-labeled transplanted stem cells. This demonstrates that 1 × 106 fusion cells formed from stem cells and breast cancer cells can generate tumors. However, due to the excessive number of cells used for the tumorigenesis experiment, these fusion cells may not be TICs[79,80].

GiTICs originate from cell-cell fusion: Possible mechanisms

Cell-cell fusion is a process involving cell chemotactic trafficking, membrane fusion, intramembrane structure fusion (including nucleus), and formation of functional fusion cells[81,82]. Moreover, it requires two or more cells to undergo cell membrane merging. However, nuclear fusion is not necessary for the formation of functional fused cells. After the first mitotic division, the binuclear hybrid may undergo nuclear fusion to produce mononuclear cells[43,48,83]. Membrane fusion involves the physical merging of membranes from different cells into a single bilayer, allowing for the exchange of cellular contents[84,85]. Generally, cells undergo cell fusion as an adaptation to unfavorable environments or factors and for the acquisition of favorable phenotypes.

Here, we summarize and hypothesize the mechanisms of giTIC generation from cell-cell fusion (Table 2 and Figure 1). The generation of TICs via cell-cell fusion may involve several fusion partners: (1) BMDCs fusing with local differentiated cells; (2) BMDCs fusing with local stem cells; (3) Local differentiated cells fusing with local stem cells; and (4) Migratory cells from different tissues fusing with local stem cells. In all cases, fusion cells may inherit the self-renewal ability of stem cells[7,11].

Table 2

Molecules potentially involved in gastrointestinal tumor-initiating cell generation from cell-cell fusion

Tumor type	Cell surface molecules involved in cell fusion	Intracellular molecules involved in cell fusion	Signaling pathways involved in cell fusion	giTIC molecules
Gastric	CD44, CD133, EpCAM, CXCR4, Lgr5, CD54[3], ALDH1[3]	OCT4, SOX2[120], Twist, Slug[47], Nanog, Lin28[44]	CXCL12/CXCR4, Lgr5[3]	CD44[3], CD133[111]
Colorectal	CD44, CD133, EpCAM, CXCR4, CD166[3]. CD81, CD9, GTP-binding protein α13, radixin[85], Syncytin 1, CD47	APC, p53, Kras, NF-κB, OCT4, SOX2[3]. ADAM10, myosin regulatory light chain, RhoA[85]	CXCL12/CXCR4, Wnt/β-catenin[3], c-Jun	CD133, CD44, ALDH1, EpCAM, CD44, CD166, CD24, LGR5, CD26[3]
Liver	CD44, CD133, CD13, EpCAM, CD24, CD90[3], E-cadherin, matrix metalloproteinase	p53[3], OCT4, SOX2[119]. Vimentin, Twist, Snail[113]	CXCL12/CXCR4[77], Wnt, TGF-β, Notch, Hedgehog[3]	CD133, CD49f, CD90, CD13[3]
Esophageal	CD44, CD133, EpCAM[115], CXCR4[3]	OCT4, SOX2	Lgr5[3], CXCL12/CXCR4	CD44, ALDH1[3]
Pancreatic	CD44, CD133, EpCAM, CXCR4, CD24[3], ALDH1[3]	KRAS, TP53, SMAD4, OCT4, SOX2[116]	CXCL12/CXCR4[3], Lgr5[118]	CD133, CD44, CD24, ESA, CXCR4[3]

giTIC: Gastrointestinal tumor-initiating cell; CD44 and CD133: Cell surface glycoproteins; CXCL12: Chemokine; CXCR4: Chemokine receptors 4; EpCAM: Epithelial cell adhesion molecule, transmembrane glycoprotein; ALDH: Acetaldehyde dehydrogenase; NF-κB: Nuclear factor kappa-B; APC: Adenomatous polyposis coli; GTP: Guanosine triphosphate; ESA: Epithelial-specific antigen; TGF-β: Transforming growth factor-β.

Figure 1

Origins of gastrointestinal tumor-initiating cells. Gastrointestinal tumor-initiating cells may originate from gene mutations, endogenous reprogramming, or cell-cell fusion. GICC: Gastrointestinal cancer cell; CMG: Cell membrane glycoprotein; CAMs: Cell adhesion molecules; TIC: Tumor-initiating cells; APC: Adenomatous polyposis coli; NF-κβ: Nuclear factor kappa-β; CXCR4: Chemokine receptor 4; CXCL12: Chemokine.

The mechanisms of giTIC generation via cell-cell fusion in vivo are very rarely elaborated. Cell-cell fusion between gastrointestinal cells and stem cells may be spontaneous[39,86], unexplained[44], or induced by carcinogens (carcinogenic chemicals) or carcinogenic factors[48,87], such as chronic inflammation and body damage. It is hypothesized that stem cells may initiate changes in the local microenvironment, which then recruits differentiated cells and leads to the fusion of local stem cells with differentiated cells, thereby generating TICs via cell-cell fusion. Similarly, Helicobacter pylori can cause chronic inflammation leading to gastric epithelial mucosal damage, which may recruit BMDCs. These BMDCs can differentiate through cell fusion with local gastric epithelial cells, leading to giTIC formation via cell-cell fusion and adenocarcinoma development[49].

Fusion proteins, also called fusogens, play an important role in mediating membrane fusion[84,85]. Cell adhesion molecules (CAMs) and cell membrane glycoproteins can mediate cell fusion. Most CAMs are involved in the process of membrane fusion and some in cell transfer[88,89]. CAMs, such as CD44, EpCAM[89,90], and cell membrane glycoprotein CD133, are highly expressed in gastrointestinal tumors, especially in giTICs or giTILCs[3,91].

CXCR4, which is a receptor of the chemokine CXCL12, is preferentially expressed in gastrointestinal tumors and promotes invasion and metastasis of gastrointestinal cancer cells[92-103]. The binding of CXCL12 to CXCR4 promotes the directed migration and homing of BMDCs[104,105]. Moreover, CXCL12 was shown to attract organ-specific metastases of CXCR4-expressing tumor cells[106,107] and CXCR4-positive MSCs were shown to migrate to the destination area, such as the stem-cell initiated tumor microenvironment[108]. Moreover, CXCL12 was shown to attract organ-specific metastases of CXCR4-expressing tumor cells and CXCR4-positive MSCs were shown to migrate to the destination area, such as the stem-cell initiated tumor microenvironment, thereby clarifying the mechanism of the induction/activation-cell migration-adhesion-cell fusion process[82,109]. Fujita et al[110] found that diffuse-type gastric cancer-derived CXCR4-positive stem-like cells penetrate into the gastric wall and migrate to the CXCL12-expressing peritoneum, resulting in the formation of peritoneal tumor lymph nodes and malignant ascites in an immunodeficient mouse model[110], which were also found to contain tumorigenic hybrid cells. Many factors, such as inflammatory factors, exosome secretion, cancer-related signal transduction pathways, and chemokines (such as CXCR4/CXCL12), can promote or cause cell chemotaxis; however, no report has shown that these factors are actually involved in the membrane fusion process[111].

In 2019, the cell-cell fusion of two colon cancer cell lines (HCT116 and LoVo) using cobalt chloride showed that syncytin 1, CD9, CD47, and c-Jun were overexpressed in the polyploid giant cancer cells (fusion cells), while PKARIα and JNK1 expression was decreased. Molecules that mediate cell fusion are usually highly expressed in fusion partner cells and hybrid cells. These highly expressed molecules or molecular pathways may be further studied as candidate cell fusion molecules that mediate cell-cell fusion.

The molecules or molecular pathways summarized in Tables 1 and 2 are likely involved in cell-cell fusion processes and the properties of TICs. As such, they may have potential as cell-cell fusion and TIC markers[112-118].

CONCLUSION

Understanding giTIC generation from cell-cell fusion may have significant implications for the understanding of carcinogenesis and the development of future cancer therapeutic strategies targeting giTICs. Under the aforementioned restrictions for giTICs and TILCs, to date, there is little evidence demonstrating that giTICs originate from cell-cell fusion in vivo, although there are reports showing that giTILCs and mouse TICs can form in vivo via the cell-cell fusion of melanoma cells and macrophages[4,5,43]. Human cell-cell fusion in vivo has also been reported, namely, the fusion of stem cells with microglia and mature neurons after the transplantation of bone marrow-derived stem cells[119]. In addition, the mechanisms of giTIC generation via cell-cell fusion are poorly understood. As such, we propose potential mechanisms involving a multi-step cell fusion process of different cell fusion partners, which is mediated by chemokine and fusogen molecules. Studies on in vitro cell-cell fusion may promote our understanding of the possible mechanisms of giTICs generation via cell-cell fusion in vivo. We suggest that future research should focus on giTIC generation via cell-cell fusion in vivo, isolation of giTICs that have tumor-initiating capabilities when using 5000 or less in vivo fused cells, and the understanding of their underlying mechanisms.