I Cornil

It is known from a variety of experimental systems that the ability of tumor cells to grow locally and metastasize can be affected by the presence of adjacent normal tissues and cells, particularly mesenchymally derived stromal cells such as fibroblasts. However, the comparative influence of such normal cell-tumor cell interactions on tumor behavior has not been thoroughly investigated from the perspective of different stages of tumor progression. To address this question we assessed the influence of normal dermal fibroblasts on the growth of human melanoma cells obtained from different stages of tumor progression. We found that the in vitro growth of most (4 out of 5) melanoma cell lines derived from early-stage radial growth phase or vertical growth phase metastatically incompetent primary lesions is repressed by coculture with normal dermal fibroblasts, suggesting that negative homeostatic growth controls are still operative on melanoma cells from early stages of disease. On the other hand, 9 out of 11 melanoma cell lines derived from advanced metastatically competent vertical growth phase primary lesions, or from distant metastases, were found to be consistently stimulated to grow in the presence of dermal fibroblasts. Evidence was obtained to show that this discriminatory fibroblastic influence is mediated by soluble inhibitory and stimulatory growth factor(s). Taken together, these results indicate that fibroblast-derived signals can have antithetical growth effects on metastatic versus metastatically incompetent tumor subpopulations. This resultant conversion in responsiveness to host tissue environmental factors may confer upon small numbers of metastatically competent cells a growth advantage, allowing them to escape local growth constraints both in the primary tumor site and at distant ectopic tissue sites.

Recent studies have shown that orthotopic (transurethral) transplantation of human bladder cancer cell lines into nude mice permits tumor growth that accurately reflects their clinical malignant status in the original host. Thus, such a system allows a unique opportunity to analyze the genetic events involved in the conversion of low-grade bladder cancer, the vast majority of which are curable, to the high-grade life-threatening form of the disease. Since 5-10% of transitional cell carcinomas (TCCs) have been shown to contain a mutated HRAS gene, and protein expression levels of all forms of HRAS have been correlated with TCC progression, we chose to study the contribution of the HRAS oncogene in bladder tumor progression. We evaluated the effects of transfection of normal or mutated HRAS genes into a human TCC, called RT-4, that behaves as a superficial noninvasive papillary tumor after transurethral orthotopic inoculation into athymic nude mice. We found that overexpression of either transfected normal or mutated HRAS genes converted RT-4 cells to express an invasive phenotype remarkably similar in nature to the clinical behavior of high-grade bladder carcinomas. These results suggest a role for overexpressed normal or mutated RAS genes in human bladder carcinoma progression, and highlight the importance of using orthotopic inoculation systems for evaluation of the contribution of oncogenes to malignant tumor progression.

Direct inoculation of a cloned bovine leukemia virus (BLV) provirus into sheep has allowed study of the viral infectivity of genetic mutants in vivo. Three BLV variants cloned from BLV-induced tumors and 12 in vitro-modified proviruses were isolated and analyzed for viral expression in cell culture. The proviruses were then inoculated into sheep in order to assess viral infectivity in vivo. Of three variants cloned from BLV-induced tumors (344, 395, and 1345), one (344) was found infectious in vivo. This particular provirus was used to engineer 12 BLV mutants. A hybrid between the 5' region of the complete but noninfectious provirus 395 and the 3' end of mutant 344 was infectious in vivo, suggesting that the tax/rex sequences were altered in virus 395. As expected, several regions of the BLV genome appeared to be essential for viral infection: the protease, pol, and env genes. Even discrete modifications in the fusion peptide located at the NH2 end of the transmembrane gp30 glycoprotein destroyed the infectious potential. In contrast, mutations and deletions in the X3 region present between the env gene and the 3' tax/rex region did not interfere with viral infection in vivo. This region of unknown function could thus be used to introduce foreign sequences. A BLV recombinant carrying a ribozyme directed against the tax/rex sequences was still infectious in vivo. Cotransfection of two noninfectious mutants carrying deletions led to infection in two of four independent injections, the infectious virus being then a recombinant between the two deletants. The experimental approach described here should help to gain insight into essential mechanisms such as in vivo viral replication, cooperation between deletants for viral infectivity, and viral superinfections. The gene products in the X3 and X4 region which are dispensable for in vivo infection could be involved in leukemogenesis, and thus proviruses deleted in these sequences could constitute the basis for a live attenuated vaccine.

Transplantation of human tumors into the organ or tissue of their origin (orthotopic transplantation) in nude mice can result in significant enhancement of tumor growth and metastases, compared with se (ectopic) transplantation. Because melanocytes are normally found in the epidermal-dermal junction, intradermal inoculation of melanoma cells might be expected to improve their potential for malignant growth as xenografts. The purpose of our study was to examine this possibility. We found that because mouse epidermis and dermis are so thin, it was not possible to inject a bolus of tumor cells intradermally; instead the cells were actually deposited in the most superficial layer of the subcutis (i.e., subdermally). We evaluated the behavior of cells from a human melanoma cell line after sc or subdennal inoculation into National Institutes of Health Swiss athymic nude mice. The cells used were from (2) the predominantly amelanotic human malignant melanoma cell line MeWo, originally established from a melanotic lymph node metastasis, and (2) two Me Wo variants resistant to wheat germ agglu-tinin (WGAr), which were selected for altered malignant capacities. Whereas 5×105 MeWo cells were required to achieve 100% tumor take with sc injection, only 2×104 cells were required with subdermal inoculation. Subder-mal injection of the MeWo cell resulted in the development of highly melanotic and nonencapsulated primary tumors, which grew quickly into the dermis and epidermis and metas-tasized at high frequency to draining lymph nodes. In contrast, the tumors that developed after sc injection were found in the deepest layer of the subcutis and were predominantly amelanotic and encapsulated; they rarely metastasized to lymph nodes. Neither of the WGAr variants, called 70-W and 3S5, displayed such drastic changes in growth behavior after subdermal inoculation. These results suggest that the subdermis, although not a perfect orthotopic site for a melanoma implant, may be more favorable and appropriate than the deep subcutis. In some cases, the subdermis may facilitate or induce the expression of phenotypic traits characteristic of human malignant melanomas in their natural host, and it may enhance the success rates of transplantation of such tumors in nude mice. [J Natl Cancer Inst 81:938-944, 1989]

Cell surface carbohydrate structures acting as ligands for tissue specific mammalian lectins have been implicated in cell-cell interactions during embryogenesis, lymphocyte homing, and tumor cell metastasis. In this report, we provide evidence that beta 1-4 linked galactose (Gal) residues in N-linked oligosaccharides on the surface of blood born tumor cells serve as a ligand for binding to microvascular endothelial cells. D36W25, a class 1 glycosylation mutant of the MDAY-D2 lymphoreticular tumor cell line, lacks sialic acid and Gal in cellular glycans due to a defect in the Golgi UDP-Gal transporter. Using UDP-Gal and bovine galactosyltransferase in vitro, beta 1-4 Gal was restored to the surface of the cells and 70% of the galactosylated glycans persisted for 8 h in vitro at 37 degrees C. Compared to mock-treated D36W25 cells, galactosylated D36W25 cells showed an 80% increase in binding to microvascular endothelial cell monolayers in vitro. The enhanced binding of galactosylated D36W25 cells to endothelial cell was inhibited by the addition of lactosamine-conjugated albumin to the assay. Consistent with these observations, swainsonine and castinospermine, two inhibitors of N-linked processing that result in loss of lactosamine antennae inhibited the binding of wild-type MDAY-D2 cells to endothelial cells in vitro. Injection of radiolabeled tumor cells into the circulation of syngeneic mice, showed that galactosylation of D36W25 cells resulted in 2-3 more tumor cells retained in the lungs and livers. In addition, galactosylation of D36W25 cells increased by 30-fold the number of visible liver metastases on inspection 4 wk after tumor cell injection. These results suggest that beta 1-4Gal-binding lectins on microvascular endothelial cells can contribute to retention and secondary tumor formation of blood born tumor cells. With the increasing availability of purified glycosyltransferases, reconstruction of a variety of carbohydrate sequences on the surface of class 1 mutants provides a controlled means of studying carbohydrate-lectin interactions on viable cells.