Abstract After demonstrating, with karyotyping, polymerase chain reaction (PCR) and fluorescence in-situ hybridization, the retention of certain human chromosomes and genes following the spontaneous fusion of human tumor and hamster cells in-vivo, it was postulated that cell fusion causes the horizontal transmission of malignancy and donor genes. Here, we analyzed gene expression profiles of 3 different hybrid tumors first generated in the hamster cheek pouch after human tumor grafting, and then propagated in hamsters and in cell cultures for years: two Hodgkin lymphomas (GW-532, GW-584) and a glioblastoma multiforme (GB-749). Based on the criteria of MAS 5.0 detection P-values ≤0.065 and at least a 2-fold greater signal expression value than a hamster melanoma control, we identified 3,759 probe sets (ranging from 1,040 to 1,303 in each transplant) from formalin-fixed, paraffin-embedded sections of the 3 hybrid tumors, which unambiguously mapped to 3,107 unique Entrez Gene IDs, representative of all human chromosomes; however, by karyology, one of the hybrid tumors (GB-749) had a total of 15 human chromosomes in its cells. Among the genes mapped, 39 probe sets, representing 33 unique Entrez Gene IDs, complied with the detection criteria in all hybrid tumor samples. Five of these 33 genes encode transcription factors that are known to regulate cell growth and differentiation; five encode cell adhesion- and transmigration-associated proteins that participate in oncogenesis and/or metastasis and invasion; and additional genes encode proteins involved in signaling pathways, regulation of apoptosis, DNA repair, and multidrug resistance. These findings were corroborated by PCR and reverse transcription PCR, showing the presence of human alphoid (α)-satellite DNA and the F11R transcripts in additional tumor transplant generations. We posit that in-vivo fusion discloses genes implicated in tumor progression, and gene families coding for the organoid phenotype. Thus, cancer cells can transduce adjacent stromal cells, with the resulting progeny having permanently transcribed genes with malignant and other gene functions of the donor DNA. Using heterospecific in-vivo cell fusion, genes encoding oncogenic and organogenic traits may be identified. Introduction Primary human tumor transplants, particularly to immunosuppressed rodents, such as nude and NOD/SCID mice, are used as preclinical models for evaluating tumor biology and drug sensitivity [31][1]–[32][7]. These studies are based on the supposition that such xenografts retain the properties and critical genotypes of their donor tumors, thus being predictive for clinical translation. However, we and others have demonstrated that such transplants can induce tumors in their rodent recipients, such as golden hamsters [33][8]–[34][10], nude/SCID mice [35][11]–[36][24], and immunosuppressed rats [37][25], although infrequently (either because of low incidence or rare testing). One plausible explanation is the horizontal transfer of oncogenic DNA [38][25]–[39][27]. Indeed, lateral oncogenesis between tumor and its stromal cells can be traced back to Ehrlich and Apolant in 1905, who showed that stromal cells of a tumor can become a sarcoma when a carcinoma is grafted in mice, and in fact the authors conjectured that a chemical factor was implicated [40][28]. Seventy-six years later, a human carcinoma transplanted to nude mice also was reported to induce fibrosarcomas that killed the nude mouse recipients and could propagate as malignant tumors in immune competent mice of the same genetic background [41][12]. In addition, a human ovarian cancer transplant to nude mice showed two cancer populations, an epithelial and a sarcomatous, the former showing human and the latter murine properties [42][14], thus suggesting lateral transduction or DNA transfer. Only the murine sarcoma cells, which were postulated to be induced by the human carcinoma cells, were metastatic and lethal in nude mice or immunocompetent mice of the same genetic background [43][14]. This induction of stromal tumors in host animals after xenotransplantation of human epithelial cancers has been confirmed by others [44][15]–[45][25], thus suggesting that cancer xenografts be carefully evaluated for horizontal oncogenesis [46][13], [47][24]. How this transformation or induction occurred was not elucidated, but a viral role has been discussed [48][17]. In some of these experiments involving primary human tumor transplants, transfer of functional human genetic information by in-vivo cell hybridization of the donor tumor and recipient host cells, showing chromosomal, immunological, or genetic features of both partners [49][9], [50][29]–[51][33], was proposed as the mechanism for induction of these tumors that exhibited highly invasive and metastatic behavior in their animal hosts [52][34], [53][35]. For example, we reported that after long-term propagation of human-hamster hybrid tumors derived from a glioblastoma multiforme [54][33] and two Hodgkin lymphomas, human DNA and genes could be confirmed by fluorescence in-situ hybridization (FISH) and polymerase chain reaction (PCR), and their donor organoid features by histology [55][36], [56][37]. Translation of some of these gene products was found by immunohistochemistry (IHC) in the glioblastoma multiforme transplants, even after propagation for over a year [57][36]. These results indicate that human genes can remain functional within human-hamster hybrid tumors propagated in the animal host, emphasizing the horizontal transmission of human DNA implicated with malignancy and the organoid features of the original patient donor tumors. However, the scope of human DNA transduced and transcribed in these interspecies hybrid cells has not been investigated. Accordingly, we examined (i) if such formalin-fixed, paraffin-embedded (FFPE) tumor grafts, which were stored for over 40 years since they were made, could be tested globally for the expression of transcribed human genes, (ii) if human genes are retained during long-term serial passage, and (iii) if there are specific human gene families indigenous to these human-hamster hybrid tumors. By using tumors and hosts of different species, we are able to identify each party's genetic contribution, which is especially problematic when attempting to prove cell-cell fusion in humans, whether involving normal-normal, malignant-normal, or malignant-malignant fusions. We postulate that these results of heterospecific fusions provide a general mechanism of tumor DNA transfer to stromal cells that results in genetic instability, heterogeneity, and aneuploidy, leading to stable genomic changes associated with cancer progression, while also retaining the tumor's original organoid phenotype, as well as other genes derived from the donor human tumor. This merging of tumor and normal genomes into a new population of malignant hybrid cells could be a mechanism whereby a cancer escapes host immunity by reducing the immunological disparity between the tumor and its host [58][34], [59][35]. Various aspects of the role of cell-cell fusion in cancer are now gaining increased attention [60][35], [61][38]–[62][47]. Results Human mRNA transcripts present in each of four different human-hamster hybrid tumor FFPE samples ([63]Table 1) were identified by analysis of total RNA, in comparison to a control hamster melanoma line (CCL-49), using Affymetrix Human U133 X3P arrays. Probe sets with MAS 5.0 detection P-values ≤0.065 in a hybrid sample, a detection P-value>0.065 in the hamster control, and an expression signal value that was at least 2-fold greater in the hybrid sample than in the hamster control, were considered to represent expressed human gene transcripts. Using these criteria, we identified a total of 3759 probe sets (ranging from 1040 to 1303 probe sets in at least one hybrid sample), which unambiguously mapped to 3107 unique Entrez Gene IDs ([64]Table S1), representing genes from all human chromosomes. Among these, 39 probe sets passed all of the expression criteria in all four hybrid specimens ([65]Figure 1, [66]Table 2), with 34 probe sets detecting 33 unique Entrez Gene IDs ([67]Table S2), two probe sets detecting either MUC3A or MUC1B, and the remaining probe sets detecting an uncharacterized gene (LOC286068), GUSBP2 or mutlple GUSB pseudogenes, and FAM91A2 or multiple uncharacterized genes. Thus, at least 33 unique human genes were transcribed in these FFPE tissues from 3 different human tumor xenografts representing different transplant generations, including two for GW-532, propagated serially for months to years as highly metastatic tumors. Table 1. Characteristics of test articles used in the microarray study. RNA sample[68]^a Transplant Generation Primary tumor IMM001 GW-532 Gen-2[69]^b Hodgkin lymphoma IMM002 GW-532 Gen-34 Hodgkin lymphoma IMM003 GW-584 Gen-28[70]^c Hodgkin lymphoma IMM004 GB-749 Gen-4[71]^d Glioma IMM006 NA[72]^e NA[73]^e Hamster melanoma [74]Open in a new tab ^a Prepared from FFPE specimens as indicated, except IMM006, which was prepared from CCL-49, a Syrian golden hamster melanoma cell line acquired from ATCC. ^b Human genes of CD74, CXCR4, CD19, CD79b, and VIM were detected by PCR (Ref. 37). ^c Human genes of CD74, CXCR4, CD20, and CD79b were detected by PCR (Ref. 37). ^d The expression of CD74, CXCR4 and PLAGL2 were detected by IHC staining (Ref. 36). ^e Not applicable. Figure 1. Clustered heat map of the 39 human probe sets detected in all four hybrid tumor samples. [75]Figure 1 [76]Open in a new tab The heat map depicts expression signal values for 39 Affymetrix Human U133_X3P probe sets detected in FFPE sections from all four hybrids tested (IMM001-004) and a hamster control (IMM006). Prior to unsupervised hierarchical clustering, the MAS 5.0 signal values were log2-transformed and row mean centered. Samples were clustered by Complete Linkage based on Pearson correlation; probe sets were clustered by Complete Linkage based on Euclidean distance. Criteria for detectable human gene expression included MAS 5.0 Detection p-values ≤0.065 in the hybrid sample and >0.065 in the hamster control, and ≥2-fold increased signal in the hybrid sample vs. the hamster control. Table 2. The 39 probe sets determined to be positive in all hybrid FFPE specimens. Probe Set ID Primary Gene Symbol Chromosomal Location Hs.183274.0.A1_3p_at HOXB8 17q21.3 g2429159_3p_a_at CFLAR 2q33-q34 Hs2.120250.2.S1_3p_a_at PARP15 3q21.1 35666_3p_at SEMA3F 3p21.3 g13376118_3p_at NAA40 11q13.1 Hs.79741.1.S1_3p_at MREG 2q35 4871689C_3p_s_at SEMA3F 3p21.3 Hs.210778.1.A1_3p_at QRSL1 6q21 Hs2.132171.1.S1_3p_x_at SLC9A5 16q22.1 g4502722_3p_at CDH3 16q22.1 g5454081_3p_at RBM17 10p15.1 Hs.241205.0.S1_3p_a_at PXMP4 20q11.22 Hs.147381.0.A1_3p_at POU2F2 19q13.2 g8923482_3p_s_at SSH3 11q13.2 Hs.128691.0.S1_3p_at ZFHX2 14q11.2 g12652612_3p_at PPARA 22q13.31 Hs.126067.0.A1_3p_at TMEM184A 7p22.3 1555620_3p_a_at PTGIR 19q13.3 g4758231_3p_x_at ECEL1 2q37.1 Hs.103978.0.S1_3p_x_at TSSK2 22q11.21 g6912587_3p_at GTPBP6 Xp22.33; Yp11.32 g4506520_3p_a_at RGS9 17q24 g4503430_3p_at DYSF 2p13.3 Hs.146084.0.A1_3p_at GPAT2 2q11.1 g12382772_3p_at CARD11 7p22 Hs.274260.2.S1_3p_at ABCC6 16p13.1 g12653688_3p_a_at DARS 2q21.3 g7705880_3p_a_at ZNF580 19q13.42 Hs.163546.0.A1_3p_x_at UBE2E1 3p24.2 g12751054_3p_s_at RPS6 9p21 Hs.325905.0.A1_3p_x_at FUT7 9q34.3 Hs.101150.0.A1_3p_at PPP1R18 6p21.3 241669_3p_x_at PRKD2 19q13.3 g11065890_3p_a_at F11R 1q21.2-q21.3 1568609_3p_s_at FAM91A2 1q21.1 Hs.129782.1.S1_3p_a_at MUC3A 7q22 Hs2.376165.1.S1_3p_at LOC286068 8q11.21 Hs.129782.0.S1_3p_a_at MUC3A 7q22 g5803174_3p_x_at GUSBP2 5q13///13.2///6p21 [77]Open in a new tab As listed in [78]Table S3, transcripts of the genes expressed in all four hybrid samples include five encoding transcription factors that are known to regulate cell growth and differentiation (HOXB8, PPARA, POU2F2, ZFHX2, and ZNF580), and five encoding cell adhesion and transmigration-associated proteins that participate in tumorigenesis and/or invasion/metastasis (CDH3, FUT7, F11R, MUC3A, and SEMA3F). In addition, genes whose products are associated with signaling pathways, regulation of apoptosis, DNA repair, and multidrug resistance, also were identified (namely, PRKD2, ECEL1, CARD11, CFLAR, PARP15, and MRP6). Recognizing that the degraded nature of the FFPE RNA and the high background of hamster RNA in the FFPE hybrid samples could interfere with the sensitivity of MAS 5.0 detection P-values, we relaxed the detection P-value criterion by requiring a detection P-value ≤0.065 in only one of the four hybrid samples, instead of all four, and produced a larger list of human genes that potentially were commonly expressed in all of the hybrid samples. This second list contained 1120 probe sets, representing 982 unique Entrez Gene IDs ([79]Table S4). These results indicate the presence of genes for CD20 (MS4A1), CD22, and CD44 (signaling component of the macrophage migration inhibitor factor (MIF)-CD74-CD44 receptor complex), thus corroborating the previous PCR results for the presence of CD20 and, also, CD74 genes in the GW-532 and GW-584 lymphoma hybrid tumors [80][36], [81][37]. A number of other human genes, such as those encoding CD24, CD27, CD47, CD52, CD84, CD151, and tenascin XB (TNXB), were found to be transcribed in these hybrid cell lines when the detection P-value criterion was relaxed ([82]Table S4). Pathway enrichment analysis of the larger, relaxed, common gene set and the individual gene sets from each of the four hybrid samples was performed with Webgestalt [83][48], [84][49], using the KEGG [85][50]–[86][52], and Pathway Commons databases [87][53], to identify similar pathways that are commonly represented in all four samples of the three hybrid tumors ([88]Table S5). Pathways that were enriched in all five gene sets (the large common gene set and the four individual hybrid sample gene sets) fall into two general categories related to cell-cell communication/focal adhesion/cell junctions/ECM (extracellular matrix) interactions, and cytokine or growth factor signal transduction (including various ErbB signaling pathways). Pathways in two other general categories related to nuclear hormone receptors and MHC antigen processing/presentation were enriched in four of the five gene sets. Enrichment analysis using the DAVID Bioinformatics database [89][54], [90][55] identified six functional annotation clusters that were represented in all five gene sets: embryonic morphogenesis, cyclic AMP/adenylate cyclase activity, mitosis/ubiquitin-mediated proteolysis, nuclear hormone receptors, lymphocyte proliferation/activation, and apoptosis ([91]Table S6). These results, from both the pathway and functional enrichment analyses, indicate that the various sets of human genes expressed in each hybrid tumor sample affect related cellular processes, and thereby likely produce similar effects on cellular function and growth. To further corroborate the microarray findings, PCR was performed on six additional FFPE tissue samples: three from GW-532 (generations 11, 52, and 82), one from GW-584 (generation 3), and two from GB-749 (both of generation 2), to assess the presence of human DNA in these tissue blocks, using a pair of primers directed to the 171-bp monomer of human alpha satellite DNA [92][56]. As shown in [93]Figure 2, four of the 6 samples (GW-532 generations 52 and 82, GW-584 generation 3, and GB-749 generation 2) were positive for the expected PCR product of human αlpha satellite DNA (the 171-bp), which was detected also in the DNA of human lymphoma Raji cells (positive control), but not in the DNA of CCL-49 hamster melanoma cells (negative control). Moreover, we were able to confirm the expression of the F11R gene detected by the cDNA microarray studies in two of the six samples by one-step reverse transcription-PCR, using human hepatic cancer HepG2 cells as the positive control [94][57]. As shown in [95]Figure 3, the presence of a 141-bp band was prominent in both GW-532 generation 11 and GW-584 generation 3, as well as in human HepG2 cells (positive control), but not in the tissue of a hamster spleen (negative control). These results were confirmed in a repeat experiment ([96]Figure S1), using CCL-49 cells as the negative control. Figure 2. PCR of human alpha satellite DNA. [97]Figure 2 [98]Open in a new tab The presence of human DNA was demonstrated by the detection of the 171-bp product in GW-532 generation 52 (lane 2), GW-532 generation 82 (lane 3), GB-749 generation 2 (lane 5), and GW-584 generation 3 (lane 6), but not in the negative control of hamster melanoma, CCL-49 (lane 8). The 171-bp and its higher oligomers were detected in the positive control of human Raji lymphoma cells (lane 7). The experimental conditions and the nominal amount of DNA used for each sample are indicated. Figure 3. One-step reverse transcription PCR. [99]Figure 3 [100]Open in a new tab The mRNA transcripts of the F11R gene were detectable in GW-532 generation 11 (lane 1), GW-584 generation 3 (lane 2), and the positive control of human HepG2 cells (lane 6), but not in the negative control hamster spleen cells (lane 5). The experimental conditions and the nominal amount of RNA used for each sample are indicated. Discussion In this study, we utilized human gene expression microarrays to provide further evidence that human genes can remain functional within metastatic human-hamster hybrid tumors propagated in the animal host, and corroborated such findings with additional samples showing the presence of human alphoid (α) satellite DNA and the F11R transcripts by PCR and reverse transcription-PCR, respectively. Our results demonstrate that human tumors transplanted to rodents can merge their DNA with the genome of the animal host, as an example of the larger program of tumor-stromal crosstalk. Cancer cells depend and are influenced by their “soil” or stromal microenvironment [101][58]–[102][60], but it is also known that there can be genetic interchange [103][61], [104][62]. The reciprocal horizontal transfer of genetic material between stromal and tumor cells could explain the heterogeneity and genetic diversity and evolution of cancer cell populations, not only between different patient tumors of the same cancer type, but even different tumors of the same patient, as observed in genetic analyses of human tumor specimens [105][63]–[106][65]. Cell-cell fusion enables the transfer of chromosomes and genetic material from one cell to another, and has been shown to result in viable hybrid progeny capable of replication for different periods, but usually not long-term or as permanent cell lines [107][66]. By using heterospecific cell-cell fusion in-vivo, genes controlling oncogenesis and organoid traits in the donor cancer cells may be elucidated in the fused progeny. The fusion of tumor and myeloid cells was proposed at the beginning of the 20^th century by various German pathologists, such as Aichel, Dor, Hallion, and Kronthal, as cited with the first experimental results and discussion of spontaneous fusion in-vivo in 1968 [108][34]. This was based on the development of highly aggressive and metastatic tumors after grafting four different human cancers, with one of ovarian cancer origin (GW-127) showing hamster chromosomes, but also retention of human antigens [109][8], [110][9], [111][29], [112][30]. A series of subsequent studies described the transplantation of diverse human cancers to the cheek pouch of unconditioned (non-immunosuppressed) golden hamsters, and also showed metastases in their hamster hosts as early as 3–4 weeks after grafting, and the presence of both human and hamster markers within the cancer cells. The transplants displayed mostly hamster properties while retaining features of their human origin, including human chromosomes, isoenzyme patterns, antigens, and stathmokinetic properties in response to colchicine that was more compatible with human than hamster cells [113][30]–[114][33]. Over the course of about 15 years, while grafting more than 1200 primary human cancers to hamsters (cheek pouch site) or nude mice (subcutaneous site), 15 (1.25%) highly aggressive and metastatic tumors resulted from the hamster transplants [115][35]. These were derived from diverse solid and hematopoietic human tumors, and could be propagated in-vitro or in-vivo for years as permanent cell lines, showing rapid growth and metastatic features typical of a hamster tumor [116][10], [117][33], [118][35]. Since gene probes were not available then, it was only recently that FFPE tissues from these earlier transplants were subjected to FISH, PCR, and IHC methods to demonstrate the presence of both species' genetic markers and translation of human genes in some of these permanent transplants, even after years in the foreign, animal host [119][36], [120][37]. For example, the glioblastoma multiforme (GW-749) was reported in 1974 to be a human-hamster hybrid tumor based on retention of up to 15 human and many hamster chromosomes in the same malignant cells, as classified by Giemsa staining, even with definite identification of chromosomes karyotyped from the patient's lymphocytes, thus being a heterosynkaryon [121][33]. More recently, the GW-749 xenograft tumor was shown to have retained 7 transcribed human genes (CD74, CXCR4, PLAGL2, GFAP, VIM, TP53, EGFR), of which CD74, CXCR4, and PLAGL2, continued to be translated to their respective proteins that were visualized by IHC, as well as hamster X chromosome and human pancentromeric DNA in the same nuclei by FISH [122][36]. Surprisingly, these genes are known to have an association with malignancy and, in particular glial tumors, as well as VIM associated with mesenchymal cells. The transplants continued to express features of the original glioma tumor grafted, even after propagation in hamsters for ∼1 year [123][36]. Similar analyses were reported recently for two lymphomas grafted to hamsters [124][37], one of which was described in 1970 and shown to resemble its donor human tumor although gaining highly metastatic properties in the hamster [125][10]. FISH and PCR analyses showed that these two Hodgkin lymphoma-derived hybrid tumors displayed both hamster and human DNA in the same nuclei by FISH, while also retaining the human genes, CD74, CXCR4, CD19, CD20, CD71, CD79b, and VIM. It is noteworthy that the GB-749 glioblastoma hybrid tumor showed retention of glioma-related genes (PLAGL2, GFAP), whereas the lymphoma-derived hybrid tumor retained several B-cell antigen receptor (BCR)-related genes (CD19, CD20, CD71, CD79b). Three human genes, CD74, CXCR4, and VIM, were common to both the glioblastoma and lymphoma transplants. Both vimentin and CXCR4 are mesenchymal markers associated with epithelial-mesenchymal transition (EMT) whose genes were transcribed in all 3 hybrid tumors examined. It was also suggested that the heterosynkaryons of Hodgkin lymphoma with their Hodgkin Reed-Sternberg (HRS) cells retained their B-cell origin [126][37], confirming other evidence for a B-cell origin of this neoplasm [127][67], and again corroborated herein by gene probe analysis disclosing B-cell genes (CD20, CD22) in these specimens. As described, these tumors were observed within 2 weeks of their first transplantation, and showed evidence of metastasis in the hamster within 3–4 weeks [128][10], [129][37], suggesting that the hamster host's early response to the foreign tissue graft may have contributed to this process. Indeed, inflammation and wound healing are known to facilitate cell fusion [130][68]–[131][70]. In the current studies, we were interested in surveying the extent by which human DNA could be transferred and continuously transcribed in the hybrid tumors. Gene expression microarray analysis was performed using total RNA isolated from FFPE sections of these hybrid tumors, including two different transplant generations of GW-532. Unexpectedly, we detected a combined total of >3000 human genes amongst all of the samples, representing genes from all 23 pairs of human chromosomes, and found that 33 human genes were ubiquitously expressed in each of the 4 samples from the 3 tumors. Five of these genes encode transcription factors that are known to regulate cell growth and differentiation (HOXB8, PPARA, POU2F2, ZFH2, ZNF580), while another five encode cell adhesion and transmigration-associated proteins that are known to participate in tumorigenesis and/or metastatic invasion (CDH3, FUT7, F11R, MUC3A, and SEMA3F). Additional genes whose products can promote metastatic growth were also identified, including two signaling pathway enzymes (PRKD2 and ECEL1), two apoptosis regulators (CARD11 and CFLAR), the DNA repair and apoptosis regulator (PARP15), and the multidrug resistance gene (ABCC6). A representative publication for each of these 16 genes is provided in [132]References S1. It is particularly