* This work was generously supported by the Deutsche Forschungsgemeinschaft (SFB 47 "Virologie", 103 "Zellenergetik und Zelldifferenzierung", 118 "Früherkennung des Krebses"), by the Bundesminister fur Forschung und Technologie, by the President of the University of Giessen, 6300 Giessen, FRG, and by the President of the University of Erlangen, 8520 Erlangen, FRG The present paper supplements that of "The Biology of an Oncogene" published in vol. 28 of this series [ I]

A. Introduction

Currently in cancer research much emphasis is being placed on mutation [2-4], rearrangement [5], amplification [6-8], and demethylation [9] of oncogenes that are supposed to represent the primary events leading to oncogene activation when carcinogens trigger neoplasia in animals or humans (the reader is referred to ref. [10] for more extensive details on this topic). The data underlying this supposition were mainly derived from experiments performed with tumor cells in vitro. In vivo studies on this issue are rare. A model which has proven to be suitable for such in vivo studies is Xiphophorus, i.e., a fish genus from Central America known as the platyfish and swordtails [ II ]. All individuals of this genus carry an oncogene, designated Tu, which was analyzed by formal genetics long before the viral and cellular oncogenes were identified by methods of modern molecular biology [ l2 -16]. Our phenogenetic, cytogenetic, and developmental genetic studies on Xiphophorus, however, have failed so far to detect any structural alterations of the Tu oncogene which were expected to occur when the gene switches over from the silent to the tumorigenic state. In contrast, it appears that impairment of the Tu oncogene might reduce rather than promote its tumorigenic potential and, therefore, it is difficult to detect. By means of seven experiments the present study aims to show that the common primary events in the causation of neoplasia in Xiphophorus are changes in the regulatory gene systems controlling the oncogenes rather than changes in the oncogenes themselves.

B. The Oncogene and the Neoplasm Used in This Study

I. The Oncogene

It appears that the genetically identified oncogene Tu may mediate neoplastic transformation in poorly differentiated cells of all tissues. The type of tissue in which transformation occurs (for instance in the pigment-cell system) depends on regulatory genes of Tu. Oncogenes other than Tu which were expected to exist in Xiphophorus were not detected by genetic methods (for discussion of this problem see [16]). From the eleven cellular oncogenes (concs) that were identified molecularly in Xiphophorus (c-erb, c-src, c-myc, c-fgr, c-ab!, cfes, c-myb, c-rasH, c-rasK, c-sis, c-yes) [16] only c-,src will be taken into consideration because most of our knowledge comes from this gene. Its inheritance was identified by molecular hybridization of viral src probes with genomic Xiphophorus DNA, and its activity was monitored by the transcript and by the pp60c-src kinase activity assayed according to Collet and Erikson [ 17]. The results parallel completely the inheritance and phenotypic expression of the Tu oncogene, specified by tumor development [I, 18-20].


II. The Neoplasm

Out of the 25 different types of neurogenic (neuroblastoma, melanoma, etc.), mesenchymal (fibrosarcoma, reticulosarcoma, etc.), and epithelial (adenocarcinoma, hepatoma, etc.) neoplasms that have been induced and identified in Xiphophorus only melanoma is taken into consideration in this study. This tumor has the advantage over the tumors of other histogenesis in that its development can easily be traced from a single transformed pigment cell detectable by its typical heavy black pigmentation, up to the extreme malignant melanoma that is lethal (Fig. I).

C. Results and Discussion

I. First Experiment: Tumors Induced by Dislocation of Regulatory Genes for the Oncogene in a Germ Line Cell

Information about the crucial event in the causation of this type of hereditary melanoma comes from an X-ray-induced structural chromosome change involving a translocation of the Tu oncogene from the X chromosome of Xiphophorus maculatus (platyfish) to an autosome of Xiphophorus helleri (swordtail) and the corresponding Tu deletion that took place in meiosis of an F1-hybrid female [13] (see Fig. 2a-d). In its new position (Fig. 2 c) Tu is no longer controlled by its formerly linked regulatory genes that act only in the cis position [ 13]. A powerful regulatory gene nonlinked to Tu that acts in the trans position persisted in the system. Fertilization of the egg with sperm of the swordtail and breeding the mutant using the swordtail as the recurrent parent resulted in offspring in which all animals containing the Tu translocation segregated into a I: I ratio of animals lacking or containing the nonlinked regulatory gene for Tu. In both segregants, as a consequence of the lack of the linked regulatory genes,

Fig. 2 a-f. a-c Translocation of the Tu oncogene from the X chromosome (X) of x. maculatus to an autosome (A) of X. helleri and d the respective Tu deletion. Note separation of Tu from its linked regulatory genes RMel and Rco (see first experiment of this study); e X chromosome of X. maculatus that is used in the second experiment (see Figs. 5, 9); f The X chromosome used in the third experiment (see Fig. 10). Note that e and f differ only in the RMel locus. F, female-sex-determining region; Pter, pterinophore locus. Rco region containing at least 13 compartment-specific regulatory genes such as Rpp (posterior part-specific) and RDf (dorsal fin-specific); RMel, melanophore-specific regulatory gene; Tu, oncogene

neoplastic transformation takes place in the pigment-cell precursors of the early embryo. In the segregants lacking both the linked and the nonlinked regulatory genes, neoplastic transformation continues in pigment-cell precursors in all areas of the body, thus forming a "whole body melanoma " (Fig. 3 B) which kills the fish before or shortly after birth. If, however, the nonlinked regulatory gene is present in the system, melanoma development is retarded at the time of birth (Fig. 3 C). The fish develop melanoma but may reach sexual maturity, thus providing the possibility to breed by further backcrosses with x. helleri, a stock that continuously produces lethal-melanoma-developing and nonlethal-melanoma-developing embryos. If, furthermore, the regulatory gene is introduced into the system in the homo zygous state, no Tu-mediated melanoma develops (Fig. 3 A). Fish carrying the Tu deletion chromosome (see Fig. 2 d) have a highly diminished potential for the formation of all kinds of neoplasms. The development of the whole body melanoma in the early embryo reflects the genuine oncogenic effect of the com pletely deregulated Tu on the pigment-cell system. These observations suggest to us that in the early embryo Tu normally exerts important functions in cytodifferentiation, which, if Tu is completely deregulated, cannot be stopped (see below). So far we have not tested pp60c-src kinase activity in the embryos that develop melanoma. However, pp60c-src kinase activity was measured in the normal developing embryos (Fig. 4) [21]. The activity is detectable from the very outset of cleavage and strongly increases during early organogenesis. While organogenesis morphologically culminates, kinase activity apparently becomes choked, decreases during the growth phase before and after birth to a lower level, and thereafter (not shown in Fig.4) remains constant at a basic level throughout the whole life of the fish [ 19]. In adults it was found that pp60c-src kinase activity is organ specific, with the brain always showing the highest values [ 18, 19]. The results suggest that c-src activity is related to differentiation and specificity of organs rather than to cell growth. Similar sequences of pp60c-src kinase activity during life were found in frogs and chickens (see Fig. 4), indicating that our results are more general rather than fish specific [21]. These genetic, embryological, and molecular biological observations suggest to us that the melanoma-mediating oncogene, irrespective ofits appearance as Tu or c-src, exerts important normal functions as a developmental gene in the early embryo. Moreover, we assume that in normal embryogenesis these functions become switched off or choked by regulatory genes after cell differen tia ti on progresses to organogenesis. If, however, the regulatory genes (i.e. the entire choke in the lethal Tu translocation) are lacking, the oncogene continues to exert its early embryo-specific functions which, as an extension of the cellular development in the early embryo,

Fig.3A-C. Backcross hybrids (X. maculatusxX. helleri X X. helleri) carrying the Tu translocation according to Fig. 2 c. A A nonlinked regulatory gene is present in the homozygous state. There is no tumor; B The nonlinked regulatory gene is lacking. Lethal "whole body melanoma" develops; COne copy of the nonlinked regulatory gene is present. Melanoma develops but the animals reach sexual maturity. A Adult fish, 5 cm long; B, C neonates, 6 mm long

appear as neoplastic transformation. Support for this idea comes from the high pp60c-src kinase activity that is found in the melanoma of young and adult fish (see second and third experiments). In any case, the crucial event leading to neoplasia in this experiment is the X-ray-induced translocation -condi tioned loss of oncogenespecific regulatory genes.

II. Second Experiment: Tumors Induced by Crossing-Conditioned Elimination of Regulatory Genes for the Oncogene from the Germ line

The basic crossing procedure for the production of melanoma-developing Xiphophorus is shown in Fig. 5: Crosses of a wild platyfish female from Rio Jamapa (Fig. 5 A) with a wild swordtail male from Rio lancetilla (Fig. 5 B) result in F 1 hybrids (Fig. 5 C) that develop uniformly in all individuals, with melanomas consisting mainly of well-differentiated transformed pigment cells that are morphologically similar to those of certain spots of the parental platyfish. They occur only in those compartments of the body where the platyfish parent infrequently exhibits the spots, e.g., in the skin of the dorsal fin and the posterior part of the body. In older F1 animals, the compartment-specific melanomas combine to form a large benign melanoma. Backcrosses of the F 1 hybrids with the swordtail as the recurrent parent (Fig. 5 D) result in offspring (BC1) exhibiting three types of segregants: 25% of the BC1 (Fig. 5 E) develop benign melanoma like that of the F1, 25% (Fig. 5 F) develop malignant melanoma consisting mainly of incompletely differentiated transformed cells which invade other tissues ( except for brain, gonads, intestine) and eventually kill the fish, whereas 50% (Fig. G, H) develop neither spots nor melanomas. Further backcrosses (not shown in Fig. 5) of the fish carrying benign melanoma with the swordtail result in a BC2 that exhibits the same segregation pattern as the BC1. The same applies for further backcrosses of this kind.

Fig.4. pp60c-src Kinase activity during cleavage, organogenesis, and growth. Da ta from [21 ]

Backcrosses of the fish carrying the malignant melanoma with the swordtail show a different result: 50% of the BC segregants develop malignant melanoma, whereas the remaining 50% are melanoma free; benign melanomas do not occur. Whenever melanomas occur in these crossing experiments, they develop in both the compartment of the dorsal fin and the compartment of the posterior part of the body. The results obtained in this crossing experiment (and a large variety of similar experiments) may be explained by the assumption of four prominent genetic components that are involved in melanoma formatioll or protection from melanoma, respectively. They are contributed by x. maculatus (for those contributed by x. helleri but not involved in melanoma formation in this crossing experiment, see below) and are normal homozygous constituents of the natural gene pool of the wild platyfish population from Rio Jamapa. Three of the genetic components (see the following points 1, 2, 3), as was confirmed by about 70 structural changes, are closely linked together at the end of the X chromosome (see Fig. 2). The fourth (point 4) is autosomally located.

1. The Tu oncogene. This gene is apparently responsible for neoplastic transformation. The restriction of neoplastic transformation to pigment cells and to the development of melanoma in the dorsal fin and in the skin of the posterior part of the body comes from other genes.

2. The regulatory gene RMel. Impairment of this gene (R'MeV specifies the tumorigenic effect of Tu to the melanophore system. RMel is a member of a series of tissue-specific regulatory genes which, in the impaired state, determine the histogenesis of the tumors (mesenchymal, epithelial, and neurogenic neoplasms). The mutations of these genes indicate that tissue specificity of neoplasia depends on regulatory genes rather than on oncogenes.

3. The regulatory genes Rpp and RDf. Impairment of these genes (RFp, R'Df) permits the development of melanoma in the skin of the posterior part of the body (Pp) and in the dorsal fin (Df) provided that R Mel is also impaired. Thirteen compartments of the body (anal fin, tail fin, mouth, eye, peritoneum, meninx, etc.) that have been identified correspond to a series of different regulatory genes (Rco in total) which in turn correspond to sites of the body where melanomas may occur (see Fig. 13).

Fig. 5A-H. Crossing. procedure for the production of melanoma-developing hybrids of Xiphophorus. A X. maculatus from Rio Jamapa (Mexico ); some small spots in the skin of the dorsal fin and the side of the body are visible. The spots consist of terminally differentiated, neoplastically transformed pigment cells. B X. helleri from Rio lancetilla (Mexico), always lacking the spots. C Fl , developing benign melanoma instead of spots (100% of the F1). D X. helleri from B used in the backcross as the recurrent paren t. E Backcross hybrid with benign melanoma (25% of the BC generation). F Backcross hybrid with malignant melanoma (25% of the BC generation). G, H Backcross hybrids that do not develop melanoma (50% of the BC generation). Spots and melanomas (in the posterior part of the body and the dorsal fin) depend on the deregulated Tu of the X chromosome shown in Fig. 2 e. Depending on the pterinophore locus ( Pter of this chromosome), spotted and tumorous fish (A, C, E and F) exhibit a reddish coloration. pp60c-src Kinase activity expressed as counts/min per milligram protein; note basic and excessive activity and correlation between c-src expression and Tu expression. Data from [15, 19] were combilled. See Figs. 9 and 10

4. The differentiation gene Diff This gene is indicated by the I: I segregation of benign- and malignant-melanoma-bearing fish. If it is present, the majority of the melanoma cells are well differentiated (benign melanoma). If, however, Diff is lacking, the majority of the melanoma cells are poorly differentiated (malignant melanoma ).

Additional regulatory genes contributed by the platyfish to the hybrid genome have been identified but are considered only in general in this experiment. Xiphophorus helleri also contains the Tu oncogene and presumably the linked regulatory genes, which, however, are not mutated and thus are fully active. Tu and its linked regulatory genes, therefore, are not detectable with the methods used in this experiment. Furthermore, since the linked regulatory genes act only in the cis position, this "oncogene-regulatory-gene complex," contributed by the swordtail, does not significantly influence the expression of the platyfish-derived melanoma-mediating Tu copy. No Diffand nor any other nonlinked regulatory gene was found in the swordtail that might correspond to those of the platyfish. The outcome of this crossing experiment suggests the following interpretation: Following crossings and backcrossings according to Fig. 5, the regulatory-gene-carrying chromosomes of the platyfish are replaced stepwise by the homologous chromosomes of the swordtail, resulting in the gradual disintegration of the regulatory gene system for the platyfish-derived Tu. Thus the Tu hybrids spontaneously develop benign melanoma if some regulatory genes such as Diff are still present in the genome and malignant melanoma if the regulatory genes are lacking. If the platyfish-derived Tu is lacking, no melanomas occur . Based on the assumption that the activity of the pp60c-src-associated phosphokinase monitors the activity of the c-src oncogene, we carried out comparisons between kinase activity, i.e., c-src expression, and tumor development, i.e., Tu expression. The preparatory work to this study showed that pp60c--src kinase activity in normal tissues of non tumorous and tumorous fish is tissue specific, with the brain always showing the highest values [ 18]. It furthermore showed that this activity is elevated in the brain of benign-melanoma-bearing fish, and even higher in the brain of malignant-melanoma-bearing fish. In any case, pp60c--s'rc kinase activity varies in both melanoma and the brain in the same direction. Hence, we were able to determine pp60c-src-associated protein kinase activity mainly in brain extracts and relate the activity observed to the expression of Tu ascertained by the development of melanoma. The possibility that the differences in kinase activity measured in the fish of different Tu genotypes are due to secondary processes in the melanoma appears unlikely. Therefore, the results reflect the actual genetic activity of the c-src oncogene in the nontumorous brain tissue of the tumorous and nontumorous fish. For critical evaluation of the methods see refs. [18, 19]. As indicated in Fig. 5, the purebred platyfish female (Fig. 5 A) carrying the population-specific "Tu-regulatory-gene complex" in both of its X chromosomes, as well as the purebred swordtail and the BC hybrids lacking this special complex, display the same activity of c-src kinase. We interpret this activity to be the basic expression of c-src. In contrast, the melanomabearing hybrids which contain the derepressed Tu show an increase in c-src activity, with the malignant-melanoma-bearing BC hybrids displaying the highest activities. To accomplish the crossing experiment (not shown in Fig. 5) we crossed two melanoma-bearing BC hybrids (FX F) that exceptionally reached sexual maturity. The resulting Tu/Tu hybrids (25% of the offspring) are of special in terest. They show an oncogene dosage effect in both melanoma forma tion and pp60c-.src kinase activity as compared with the parental BC hybrids [1, 20]. In summary, all experiments of this kind showed a clear-cut correlation between c-src expression (measured by pp60c--s'rc kinase activity) and Tu expression (measured by melanoma formation). Many features that are assumed to be .involved in the causation of neoplasia have been found in melanoma of Xiphophorus. About 70% of the mel'anomatous cells in cell cultures derived from embryos exhibit at least two pairs of double minute chromosomes (Fig. 6) [22) that might be interpreted

Fig.8. RNA-dependent DNA polymerase activity in Xiphophorus.
Sucrose density gradient centrifugation of the microsomal pellet

as carriers of amplified oncogenes [7, 8]. Virus particles have been found in the melanoma following treatment with bromodesoxyuridine [23]. Following sucrose density centrifugation, in the supernatant of nontreated melanomatous cells of cultures mentioned above, RNA-dependent DNA polymerase associated with particles similar to fish iridovirus [24] (Fig. 7) was found in the density range of 1.15-1.17 g/ml. Three to five particles have been counted per grid, which is equivalent to 105 particles/ml. RNA-dependent DNA polymerase activity has also been detected in the melanoma tissue and, although three to five times lower, in the muscle (Fig. 8). Activity of many other enzymes, such as tyrosinase [25], lactate dehydrogenase [26, 27], malate dehydrogenase, and pyruvate kinase [27], is changed in the melanoma.

Table 1. SCE frequency in intestinal cells of Xiphophorus [29]

Fig.9A-H. Crossing procedure f'or the suppression of melanoma in Xiphophorus. A MalIgnant melanoma-bearing backcross hybrid according to Fig. 5 F. B X. maculatus accor.ding to Fi.g: 5 A as the recurrent parent. C Quasi Fl-exhlblting benign melanoma. D X. macul.atus as the recurrent parent. E-H Backcross hybrids (quasI X. maculatus) exhibiting spots only. Melanomas and spots depend on Tu of the X chromosome shown In Fig. 2 e. See Fig. 5

Fig.l0A-H. Crossing procedure for the production of Xiphophorus hybrids.that are sensItive to carcinogens. The procedure IS basically the same as that outlined in Fig. 5 (A X. maculatus: B X. helleri; C Fl ; D X. helleri). A, C, E, and F exhibit a reddish coloration. c-src, pp60c-src kinase activity. See text

(Fig. 9 D) result in spotted and nonspotted fish that are similar to the purebred platyfish (Fig. 9 E-H). Genetically they segregate into those that inherit the capability to develop melanoma after crossings .with swordtails and those that do not Inherit the capability to develop melanoma (not shown in Fig. 9). The outcome of this final series of backcrosses indicates the step wise reintroduction of the platyfish chromosomes carrying the regulatory genes for the oncogene into the descendants, and the reconstruction of the original regulatory gene system that suppresses the activity of the oncogene in the platyfish genome.

Fig.11A,B. Backcross hybrids according to Fig.1O E and F after treatment with MNU. A Diff-carrying segregant (identified by an esterase) developing benign melanoma; B Diff-lacking segregant (identified by lack of the esterase) developing malignant melanoma

Presumably the oncogene itself remained unchanged during the crossing procedure (Figs. 5, 9) that first induced and thereafter suppressed melanoma formation. Carcinogens or tumor-suppressing agents were not required in this experiment. The outcome of the experiment, however, indicates that the experimenter must interfere in the regulatory gene system of the oncogene to be able to induce or suppress melanoma ( or any other neoplasm). To examine this idea, we carried out the following experiment.

III. Third Experiment: Tumors Induced by Impairment of Cell- Type-Specific Regulatory Genes for the Oncogene in Somatic Cells

This experiment is based on the two assumptions that (a) the total number of pigment-cell precursors competent for neoplastic transformation is 10 high 6 (this is the average number in the pigment-cell system of a young fish) and (b) the mutation rate for a given gene in these cells is 10 high-6. Then tumor incidence is one ( on average 100% of the treated animals develop one tumor) provided the oncogene is under control of only one regulatory gene. If, however, the oncogene is regulated by two genes, the rate of simultaneous mutations of both regulatory genes in one cell is 10high-12 and the tumor incidence is 10high-6. This calculation shows that if the oncogene is controlled by more than one regulatory gene, it is difficult to succeed in inducing somatic mutation-conditioned neoplasms at a frequency suitable for mutagenesis experiments. To establish a Xiphophorus strain suitable for mutagenesis studies on a single regulatory gene of the Tu oncogene, we replaced the R`Pp RDf R'Mel platyfish X chromosome used so tar (Fig. 2 e) by the Rpp RDj R Mel Tu X chromosome (Fig. 2 f), which, instead of the impaired RMel, contains the normal RMel. Crossings were performed according to Fig. 10, which are essentially the same as those described in the first part of the previous experiment (Fig. 5). Due to the active RMel that controls the oncogene in the pigment cell system, no melanomas develop in the hybrids. The BC hybrids E and F that correspond to the melanoma-developing BC hybrids in the former experiment (they can be recognized by a reddish coloration encoded in the pteridine marker gene PIer also linked to the Tu oncogene; see Fig. 2) are, however , sensitive to carcinogens and may develop

Table 2. Tumor initiation (somatic mutation) and promotion (promotion of cell differentiation) distinguished
by tester strains of Xiphophorus constructed according to the third and sixth experiments

melanoma after carcinogen treatment. Those BC hybrids carrying the Diff gene (autosomal; identified by a Diff-linked esterase marker gene [30,31]) may develop benign melanomas and those lacking Diff may develop malignant ones (see Figs. I, II ). Crosses between two BC hybrids of the latter genotype were the basis for the establishment of a strain homozygous for the R`pp R'Df R Met Tu chromosome [32, 33]. Because each of the two copies of the Tu oncogene is repressed by its own linked RMet' which acts in the cis position only, the incidence of animals developing melanoma following treatment with carcinogens potentially doubles. These fish are highly suitable as test animals for mutagenic carcinogens in the water (Table 2, see column "Initiation"). In contrast to the crossing-conditioned hereditary melanomas that, due to their multicellular origin [ 15], cover large areas in the compartments of the body from their very beginning, the carcinogen-induced melanomas appear first as small foci of cells in these compartments (Fig. 1), indicating that they originate from the expected single mutational event in a particular regulatory gene of the oncogene in a particular somatic cell in a particular compartment of the body. All purebred and hybrid animals of this experiment display uniform pp60C-srC kinase activity in the brain [19]. This activity seems to represent the basic c-src expression like that of the nontumorous fish in the previous experiment ( cf. Figs. 5, 10). Following the induction of melanomas and of any other kind of neoplasia, pp60c-src kinase activity in the brain remains unchanged. Kinase activity in the melanomas, however, is heterogeneous, with the malig-

Table 3. Elevated pp60c-src kinase activity in Xiphophorus bearing tumors
as compared with nontumor ous controls

nant melanoma showing values that are at least as high as those of the corresponding melanomas of crossing-conditioned hereditary origin. In an induced malignant amelanotic melanoma pp60c-src kinase activity was even six times higher than that of the brain. As in melanomas, elevated levels of kinase activity were also found in other carcinogen-induced neoplasms such as fibrosarcoma, retinoblastoma, and epithelioma (see Table 3). However, changes in the pp60c-src kinase activity in the brain that are typical in the hereditary tumor-bearing animals were not observed. More studies are necessary to find out whether the elevation of pp60c-csrc activity found in the neoplasms is due to the same impaired regulatory gene that permits the Tu oncogene to mediate carcinogenesis.

IV. Fourth Experiment: Transfer of the Regulatory Gene-Dependent Tumorigenic Potential by Genomic DNA

To perform this transfection experiment we used different laboratory stocks of x. maculatus as donors. These stocks were uniform with respect to Tu and the impairment of the compartment-specific regulatory genes RAp and Rpp (see fifth experiment). They were, however, different in the condition of the melanophore-specific regulatory gene RMeI, which was normal, impaired, strongly impaired, or deleted (Fig. 12). Total genomic DNA of the donors was injected into the neural crest region of early embryos (where the pigment-cell precursors originate) of x. helleri. This species has excellent potential for expressing a deregulated oncogene (see previous experiments). If a donor containing the normal RMeI-Tu complex (Fig. 12 A) was used, no recipient exhibited transformed pigment cells. If a donor strain was used that carried a Tu slightly derepressed in the pigment-cell system by a "weak" mutation of its linked RMeI, i.e., R'Mel Tu (Fig. 12 B), then 0.4% of the recipients developed colonies of neoplastically transformed cells. If the DNA originated from a strain that carries the Tu derepressed to a greater degree, due to a stronger mutation of RMel. i.e. R``MeI (Fig. 12 C), the incidence of recipients exhibiting transformed pigment cells increased to 2.6%. If finally, a donor was used in which the Tu lacks the RMeI (due to a chromosomal translocation) ( Fig. 12 D), this incidence increased to 6.3%. DNA from animals carrying additional but rigidly repressed accessory Tu copies (not shown in Fig. 12) did not influence the incidence of transformants [34]. Besides the fact that the information for neoplastic transformation, presumably the

Fig.12A-D. Transfection activity of donor DNA extracted from male gonads of fish differing in Tu control by the pigment-cell-specific regulatory gene R Mel. Data from [34 ]. See text

oncogene itself, was transferred via total genomic DNA, it is important to note that the transforming donor DNA did not originate from tumor cells but from the nonneoplastic testes, indicating that oncogenes must not necessarily be changed or amplified in order to acquire the transforming potential. The many oncogene transfection experiments accomplished during the past years by several authors with other systems in which DNA extracts from tumors were used [2, 3, 35-37], with the expectation that tumor DNA differs from DNA from normal tissues, should be, in our opinion, reconsidered from the viewpoint of repression and derepression of oncogenes exerted by intact and defective regulatory genes. The main factors responsible for neoplasia are, in view of our results on Xiphophorus, not the onc genes, but their regulatory genes. This view is supported by the fact that the incidence of transformants, i.e., the incidence of transformation events mediated by Tu, was independent of the number of Tu copies in the donor DNA (we tested DNA containing up to eight copies), but was exclusively dependent on the degree of impairment of RMel. In this light it also appears reasonable to assume that R Mel, if present, is so closely linked to Tu that both Tu and RMel are always cotransferred [34]. Although the donor DNA originated from fish exhibiting different degrees of Tu expression, the transformed cells of the recipients all looked alike, and the cell colonies were all about the same size. This indicates that growth of the tumor is influenced neither by Tu nor by the intact or impaired RMel.

V. Fifth Experiment: Tumors Induced by Impairment of Compartment-Specific Regulatory Genes
for the Oncogene in Germ Line Cells and/or in Somatic Cells

The compartment-specific regulatory genes for the oncogene, designated Rco in total (see Fig.2), have been studied mainly by means of X-ray-induced germ line mutations that affect one or several sites of the crossing-conditioned melanomas (Fig. 13). These melanomas develop in BC hybrids, for instance (a) in the dorsal fin; (b) in the tail fin; ( c) in both the dorsal fin and tail fin; (d) in the anal fin; (e) in the tail fin, dorsal fin, anal fin, mouth tip, and posterior part of the side of the body (mutations of five compartment-specific regulatory genes are involved); (f) in the anterior and posterior parts of the side of the body; (g) in all compartments except for the mouth, belly, eye, dorsal fin, and tail fin; or (h) even in all compartments of the body. The phenotypes of additional combinations of impaired Rco genes have been described previously [ 13, 15, 16]. The compartment-specific distribution of these melanomas is inherited according to the segregation of the parental Tu-carrying chromosome, indicating that the respective Rco genes are linked to Tu, and structural changes of the chromosome have verified that this linkage is very close. At least 14 genes corresponding to 14 different compartments have been identified. They represent regulatory genes that were designated RAp (anterior part), Rpp (posterior part), RDf(dorsal fin), RTf(tail fin), etc. in reference to each specific body compartment. Intact Rco genes repress Tu, and impaired Rco genes permit Tu activity. They act in the cis position only, indicating that the compartment-specific regulation of Tu exerted by the Rco genes acts at the DNA level. In the active state the Rco genes appear to delay the differentiation of pigment cells in the stem-cell stage (S-melanoblasts; see Rco in Fig. 14). Additional mechanisms that are not understood provide the fish with differentiating pigment cells that mostly escape neoplastic transformation, but very exceptionally may be transformed. If, however, one or several R Co genes are impaired by mutation, compartment-specific melanomas develop in the hybrids and are inherited according to Mendelian prediction. To study whether melanoma can also be induced by impairment of a compartmentspecific regulatory gene in a somatic cell we carried out the following experiment: BC hybrids, which, due to the impairment of the pigment-cell-specific regulatory gene (R`Mel) and of the dorsal-fin-specific regulatory gene (R`Df), develop hereditary melanoma in the dorsal fin only, were irradiated with X-rays. These animals frequently developed additional melanomas in other compartments of the body (Fig. 15) which, in contrast to the hereditary melanoma of the dorsal fin, are nonhereditary and develop from small foci of transformed cells in the skin. This indicates a mutational event in a particular compartment-specific regulatory gene in a particular competent pigment-cell precursor in a particular compartment of the body. The oncogene itself that mediates neoplastic transformation probably remained unchanged.

VI. Sixth Experiment: Tumors Induced by Promotion ofCel1 Differentiation

The neutral crest cells of Xiphophorus which are, like those of all vertebrates, the precursors of different cell types, start migrating and differentiating at the outset of organogenesis in the embryo (Fig. 14) [38, 39]. Those entering their final locations in the skin and extracutaneous tissues become determined to differentiate to chromatoblasts. These chromatoblasts are the common precursors of all types of pigment cells (chromatophores), including pterinophores, purinophores, and melanophores. Those chromatoblasts committed to differentiate to melanophores give rise to stern melanoblasts (S-melanoblasts). These may reproduce throughout the life of the fish,

Fig. 14. Schematic presentation of the differentiation of normal and neoplastically transformed pigment cells. S -, I -, and A -melanoblasts are stem, intermediate, and advanced melanoblasts, respectively. The Tr cells represent the transformed cells. Only J-melanoblasts are competent for neoplastic transformation. Tu, tumor gene (oncogene); RMel, regulatory gene for control of Tu in the melanophore system; Rco, compartment-specific regulatory genes; g, "golden" gene that blocks pigment-cell differentiation; Diff, differentiation gene; I -genes, intensifier genes, which support proliferation of poorly differentiated transformed pigment cells. Macrophages attack melanophores and Tr-melanophores. Modified from [ 16]

but may also differentiate to intermediatestage melanoblasts (l-melanoblasts) that continue differentiation to the advancedstage melanoblasts (A-melanoblasts) that can be distinguished from their precursors by their reaction to dopa. These cells differentiate to melanocytes, and these, finally, to melanophores. Genetic, cytological, and ultrastructural studies of the differentiation of the pigment cells have shown that the l-melanoblasts are the only cells of the system that are competent to undergo neoplastic transformation, i.e., corn petent to undergo the transforming activity of the Tu oncogene [38, 39]. The I-melanoblasts, after being transformed to Trl-melanoblasts (all transformed cells are called Tr cells) differentiate to the easily recognizable, proliferating TrA melanoblasts. These Tr cells differentiate to the heavily pigmented Tr melanocytes, which proceed to the terminal stage of differentiation of the transformed pigment cells, represented by the Tr melanophores [15]. We have analyzed several genes that are involved in pigment-cell differentiation. One of these genes is the "golden " gene (g+), which, if present as a mutation in the homozygous state (g/ g), creates an almost complete stop of melanophore differentiation at the stage of S-melanoblasts (see 9 in Fig. 14). The always present drosopterines become more visible in the skin, thus giving the fish the "golden" coloration. This 9 mutation was introduced into the hereditary melanoma-bearing hybrids by

Fig.16. Littermates that segregate into "golden" (g/ g) animals (left) and tumor (g+ / g) animals (right). The 9 / 9 animals are protected from melanoma by a block of pigment-cell differentiation in the stage of the stem melanoblasts (see Fig. 14). Following treatment with promoters of cell differentiation (see Table 2) they develop melanoma similar to their heterozygous littermates

introgressive breeding, i.e., by the replacement of the g+-bearing chromosome by the g-bearing homolog. In animals heterozygous for 9 (g/ g+) the derepressed Tu oncogene still mediates melanoma formation (Fig. 16, right). In the littermates that are homozygous for 9 (g/ g), however, no melanomas develop (Fig. 16, left): The block of differentiation at the stage of the stem melanoblasts exerted by the homozygous g-mutation protects the fish from melanoma formation or from the activity of its own derepressed oncogene. To break this protection mechanism we tested a large variety of mutagenic and nonmutagenic agents. Most of these agents (see Table 2, column "Promotion") promote almost simultaneously the differentiation of large amounts of the noncompetent cells to the competent stage, which subsequently become neoplastically transformed and give rise to melanomas.

Fig.17. Decrease of m5dC methylation of DNA in Xiphophorus hybrids after treatment with 2.5 fig/liter per day 5-azacytidine. Fish were treated for 3 months, beginning at birth. HPLC analysis (filling, Supelco 4C-18-DB). Untreated, six animals; treated, five animals [63]

These tumors are morphologically similar to the spontaneously developing melanomas of the g/ g+ littermates, develop in the same compartments of the body, and are, similar to all hereditary melanomas, of multicellular origin. X-rays, N-methylN-nitrosourea (MNU), and N-ethyl-Nnitrosourea (ENU), which are powerful mutagens or tumor initiators, respectively, may also trigger melanoma by promotion of cell differentiation. It appears, however, that in this case neither of these agents acts as a mutagen, but as a differentiation-promoting agent such as testosterone, TP A, and cyclamate, which are certainly not mutagens. The assumption that any somatic mutation, including that of a possible back mutation or suppressor mutation for the 9 mutation, is not involved in the promoting effect is compatible with the multicellular origin of the melanoma and with the high frequency of the respondents, which, in the case of 17 -methyltestosterone, even reaches 100%. Optimization of the treatment in future experiments will show whether the promoting effect of a certain agent is an allor-llothing effect, as is conceivable on the basis of the concept of this experiment. Since somatic mutations are probably not involved in the trigger for melanoma development in this experiment, one could speculate that the 9 mutation undergoes a general modification that brings the function ofg close to the normal function of the original g+. The promoting effect of 5-azacytidine could be interpreted in this direction. This agent creates undermethylatioll in the DNA and, as a consequence, presumably gene activation [9, 40, 41]. As shown in Fig. 17, 5-azacytidine, which promotes differentiation, also demethylates DNA of Xiphophorus. More data are required to correlate the removal of the tumor-protecting effect of the g-mutation by 5-azacytidine with DNA methylation. The crucial event leading to neoplasia in this experiment is the promotion ofcell differentiation and possibly a more general modification of a gene that is involved in cell differentiation. The oncogene itself probably remained unchanged. VII. Seventh Experiment: Tumors Induced by Elimination ofa Differentiation Gene from the Germ Line and Impairment of That Gene in Somatic Cells To study the effect of the differentiation gene (Diff), thousands of melalloma-bearing BC segregants were produced according to the crossing procedure shown in Fig. 5, and all showed a clear-cut I: 1 segregation into animals developing spontaneously benign (Fig. 5 E) or malignant melanomas (Fig. 5 F). BC hybrids that require the carcinogenic trigger for the development of melanoma also show a clear-cut Diffeffect (Fig. 11, see third experiment). Morphological, histological, cytological, fine structural, biochemical, and molecular studies showed that the majority of the cells of the benign melanoma are well differentiated whereas those of the malignant melanoma are poorly differentiated, and that differentiation of the transformed cells is controlled by Diff [42, 43]. The most convincing data supporting Diff-dependent control of pigment-cell differentiation in melanoma come from transplantation experiments which have shown that pigmentcell precursors present in the transplants taken from fish carrying the deregulated Tu and lacking Diff (material of still tumorfree early embryos of the malignant-mela-

Fig. 18. General cloverleaf structure of the tRNA and positions of modified nucleosides.
Bottom left. structure of queuosine (Q). Modified from [49]

noma-developing genotype according to the fish shown in Fig. 5 F) become transformed and remain incompletely differentiated if transplanted into embryos lacking Tu and Diff. The resulting animals develop malignant melanoma. If, however, the pigment-cell precursors of the same genotype are transplanted into Tu-lacking embryos that contain the Diff gene, the cells of the developing melanoma become terminally differentiated and regain their distance regulation: These resulting animals develop extreme benign melanomas which regress and eventually may become removed by macrophages (see [ I]). Further studies on the Diff effect on melanoma in Xiphophorus were stimulated by the experimental results of other laboratories that have focused on the in volvement of nucleotide modifications of a certain tRNA family in cell differentiation in eilbacteria, slime molds, and cell cultures from different vertebrates [43-48]. These tRNAs include tRNAAsn, tRNAAsp, tRNA His, and tRNA Tyr, which usually con tain qileilosine (Q) instead of guanosine (G) in the first position of the anticodon (position 34; see arrow in Fig. 18). The Q nucleoside (7-(((4,5-cis-dihydroxy2-cyclopenten-l- yl) -amino) -methyl- 7 -deazagilanosine) is unique in that its purine skeleton is modified to a 7 -deazastructure. Eilbacteria synthesize the base qileuine de novo whereas vertebrates are supplied with qileiline by nutrition or the intestinal flora. Qileiline itself is inserted into the nucleotide chain of tRNA by an exchange with guanine. This process is catalyzed by

Fig.19A-C. Incorporation of [³H]guanine in position 34 oftRNA for Asp, Asn, His, and Tyr of Xiphophorus catalyzed by tRNA-guanine-transglycosylase (insertase) of E. coli. The graphs show the kinetics of the exchange of G34 of tRNA by [³H] guanine, a reaction used to evaluate the amount of (Q-)tRNA [50]. A-C according to the fish shown below the curves. These fish correspond to those shown in Fig, 5. High incorporation of ³H] guanine in Diff-lacking animals corresponds to a low content of Q, whereas low incorporation of r H] guanine in Diff-containing animals corresponds to a high content of Q. A Skin of purebred Xiphophorus: white ring, X, helleri; black ring, X, maculatus. B Melanoma of BC segregants: white square , malignant; black square, benign. C Skin of nonmelanomatous BC segregants: white triangle, lacking Diff. Black triangle, containing Diff: Note that comparable Diff-containing animals always have a lower G content and a higher Q content than Diff-Iacking ones. Data from [43,44,51]

tRNA-guanine-transglycosylases. The more the cells are differentiated, the more replacement of G by Q is observed in position 34. The method for estimating the G: Q ratio in a given population of the tRNA family consisted of following the replacement of guanine in position 34 by a ³H-labeled guanine exerted by a guanine-transglycosylase (insertase) of Escherichia coli [50,51]. The results obtained in Xiphophorus by measurement of [³H] guanine incorporation into the tRNAs of the Q family, differing in the ratio of G : Q in position 34, are summarized in Fig. 19. The graphs show the kinetics of the exchange of G 34 of the tRNA family by [³H] guanine, which is the reaction used to evaluate the amount of (Q-)-tRNA. The fish genotypes and phenotypes are identical to those shown in Fig. 5. In accordance with the findings of many investigators working with other differentiation systems [52], [³H] guanine incorporation is high in tRNAs from malignant melanomas that consist predominantly of poorly differentiated cells. In contrast, the incorporation is lower if the tRNAs are derived from benign melanomas that consist predominantly of well-differentiated cells. Therefore, tRNAs of malignant melanomas have a higher amount of Gin place of Q than those of the benign melanomas (Fig. 19 B). To decide whether the distinct difference in G: Q ratios between benign and malignant melanoma is Diff' dependent or represents an epiphenomenon of benignancy and malignancy, the skin of nontumorous littermates that segregate into animals carrying Diff and lacking Diff like the tumorous fish in a 1: 1 ratio was used for analysis (Fig. 19C). The Diff-lacking segregants always had higher amounts of Q-lacking tRNA than the Diff-carrying animals. The skin of the parent animals used

Fig.20A,B. Development of induced somatic mutation-conditioned malignant melanomas on a crossing-conditioned superficial hereditary melanoma. A Untreated fish; B Fish treated with X-rays. It is assumed that the induced melanomas are due to an impairment of the differentiation gene Diff in a pigment-cell precursor each

for the initial crosses showed the same differences (Fig. 19 A) : X. helleri, which lacks the Diff gene, has a high ³H] guanine incorporation (i.e., is G-rich) whereas X. maculatus, which contains the Diff gene, has a lower [³H] guanine incorporation. From these results we suggest that the difference of G: Q ratios between benign and malignant melanoma are not epiphenomena of benignancy and malignancy, but are closely related to the primary effect of the Diff gene. The differences in the functional properties of Q-containing and Q-lacking tRNAs require further elucidation. The (Q)tRNAs have been suggested to prefer codons NAU to NAC, whereas the Q-lacking tRNAs read NAC and NAU equally well [46]. This may be an important mechanism in the regulation of translation. For eukaryotic tRNA Tyr it has been shown that the Q-lacking species reads a terminator codon, probably UAG. Therefore the Q-lacking and Q-containing tRNAs of vertebrates might select mRNAs for translation by a regulatory mechanism similar to that of termination transcription control (see discussions in ref. [51]). If benignancy depends on the presence of a single copy of Diff and malignancy on the lack of this copy then it should be possible to induce easily somatic mutation conditioned malignant melanoma in the area of germinal-conditioned benign melanoma. For this purpose we treated superficial benign melanoma with MNU or X-rays and observed the development of focal malignant melanoma in the area of the benign melanoma (Fig. 20). This is not to say that the benign melanoma changes to the malignant state. In contrast, the somatic mutation-conditioned malignant melanoma develops independently from the already present germ-line-conditioned benign melanoma. The induced malignant melanoma, however, competes with the hereditary benign melanoma for pigmentcell precursors. As a consequence of this competition the focal malignant melanomas are surrounded by a halo-like zone that is sparsely polulated by the cells of the benign melanoma. This experiment shows once more that the genes that carcinogens act upon if they trigger neoplasia are not necessarily the oncogenes themselves, but their regulatory genes in the broadest sense.

Fig.21. Distribution of the c-src oncogene in living organisms. All organisms listed have been tested.
c-src was found in all multicellular animals

D. Conclusions

In our Xiphophorus model we have not found any genetic change of the Mendelian inherited oncogene Tu that might lead to neoplasia, although one would expect this to be possible based on the molecular findings of other laboratories that mutation, rearrangement, amplification, and demethylation can convert a silent oncogene to the transforming state. In contrast, our carcinogenesis studies show that the most important process involved in neoplasia in these animals is loss, impairment, or any other dysfunction of the regulatory gene system of the Tu oncogene. About 20 regulatory genes controlling the oncogene Tu at the gene level have been identified genetically and phenotypically. These genes comprise tissue-specific and compartment-specific (eye, mouth, etc.) regulatory genes (first to fifth experiments). Once the system of these regulatory genes controlling Tu is impaired, a chain of events is begun that can lead straight to neoplasia, but can also be interrupted by a genetic block of cell differentiation which protects the animal from the transforming activity of the oncogene. This protection mechanism can, however, easily be broken by promotion of cell differentiation (sixth experiment). If, finally, the cells are transformed, tumor growth can be stopped by terminal differentiation of the tumor cells exerted by the differentiation gene. Loss or impairment of the differentiation gene, then, leads definitely to neoplasia (seventh experiment). Our carcinogenesis experiments indicate that it is the regulatory genes (in the broadest sense ) and not the oncogene itself that the carcinogens commonly act on, when they trigger neoplasia in Xiphophorus. The inheritance and phenotypic expression of the genetically defined Tu oncogene parallels completely the expression of the molecularly defined c-src oncogene. Regardless of any future findings that might bring Tu substantially in relation to c-src and/ or any other c-onc, both Tu and c-src act, are regulated, and are inherited as if they were the same chromosomal gene of the natural gene pool of the Xiphophorus fish. c-src has been found functioning in all taxonomic groups of multicellular animals ranging from mammals down to the sponges (Fig.21) [16, 53], and intensive efforts are being made in many laboratories to determine whether a cellular oncogene, such as c-src, is capable of mediating neoplastic transformation like its viral counterpart [3, 54-59]. If this should be proven we suggest that all individuals of all metazoa are endowed with the capacity to develop neoplasia. Support for this idea comes from the fact that neoplasia is distributed -although sporadically -in all groups of multicellular animals [60-62]. Ubiquity of the oncogene in metazoa including humans on the one hand and the infrequent occurrence ofneoplasia in all these organisms on the other hand raises the question of the mechanisms that protect the majority of the individuals of all metazoans from the action of their own oncogenes. The Xiphophorus model provides an opportunity to contribute to the study of this problem.

Acknowledgment.

Thanks are due to Kristine Krüger for valuable help in the preparation of this paper.

References

1. Anders F (1983) In: Neth R, Gallo RC, Greaves MF, Moore MAS, Winkler K (eds) Modern trends in human leukemia V. Springer, Berlin New York, pp 186-206 (Haematology and Blood Transfusion, vo128)

2. Reddy EP, Reynolds RK, Santos E, Barbacid M (1982) Nature 300: 149-152

3. Tabin CJ, Bradley SM, Bargmann CI, Weinberg RA, Papageorge AG, Scolnick EM, Dhar R, Lowy DR, Chang EH (1982) Nature 300:143-149

4. Taparowsky E, Suard Y, Fasano 0, Shimizu K, Goldfarb M, Wigler M (1982) Nature 300: 762-764

5. Payne GS, Bishop JM, Varmus HE (1982) Nature 295:209-213

6. Dalla Favera R, Wong-Staal F, Gallo R (1982) Nature 299:61-63

7. Schwab M, Alitalo K, Varmus HE, Bishop JM, George D (1983) Nature 303:497-501

8. Schimke RT (1984) Cancer Res 44: 1735-1742

9. Riggs AD, Jones PA (1983) Adv Cancer Res 40: 1-30

10. Pimentel E (1985) Cancer Genet Cytogenet 14:347-368

II. Kallman KD (1975) Handb Genet 4:81-132

12. Anders F (1967) Experientia 23: 1-10 13. Anders A, Anders F, Klinke K (1973) In: Schröder JH (ed) Genetics and mutagenesis offish. Springer, Berlin New York, pp 33-63

14. Anders F, Anders A, Vielkind U (1974) Int Cancer Congr I1th Florence 3:305 (Abstr) 15. Anders A, Anders F (1978) Biochim Biophys Acta 516:61-95 16. Anders F, Schartl M, Barnekow A, Anders A (1984) Adv Cancer Res 42: 191-275

17. Collet MS, Erikson RL (1978) Proc Natl Acad Sci USA 75: 2021- 2024

18. Barnekow A, Schartl M, Anders F, Bauer H (1982) Cancer Res 42: 2429-2433

19. Schartl M, Barnekow A, Bauer H, Anders F (1982) Cancer Res 42:4222-4227

20. Anders F, Schartl M, Bamekow A (1984) Natl Cancer Inst Monogr 65:97-109

21. Schartl M, Bamekow A (1984) Developmen tal Biology 105:415-422

22. Heil M ( 1984) Thesis, University of Giessen

23. Kollinger G, Schwab M, Anders F (1979) J Cancer Res Clin Oncol 95: 239-246

24. Berthiaume L (1984) Virology 135: 10-19

25. Vielkind U, Schlage W, Anders F (1977) Z Krebsforsch 90: 285-299

26. Ahuja MR, Schwab M, Anders F (1975) Ex perientia 31: 296-297

27. Mäueler W (1984) Thesis, University of Giessen

28. EI-Zawahri MM, Hamdoon NT, Anders F (to be published) Cancer Genet Cytogenet

29. Hamdoon NT (1984) Thesis, University of Giessen

30. Ahuja MR, Schwab M, Anders F (1980) J Hered 71: 403-407

31. Morizot DC, Sicliano MJ (1982) Genetics 102:539-566

32. Anders F, Schwab M, Scholl E (1981) In: Stich HF, San RHC (eds) Short-term tests for chemical carcinogens. Springer, Berlin New York, pp 399-407

33. Anders F, Schmidt CR, Herbert A, Anders A (1983) In: Prüfung von Chemikalien auf Kanzerogenität, Mutagenität und Teratogenität. Ges. f. Strahlen- und Umweltforschung, München, pp 253-274

34. Vielkind J, Haas-Andela H, Vielkind U, An ders F (1982) Mol Gen Genet 185: 379-389

35. Krontiris TG, Cooper GM (1981) Proc Natl Acad Sci USA 78: 1181-1184

36. Perucho M, Goldfarb M, Shimizu K, Lama C, Fogh J, Wigler M (1981) Cel127:467-476

37. Der CJ, Krontiris TG, Cooper GM (1982) Proc Natl Acad Sci USA 79:3637-3640

38. Anders F, Diehl H, Schwab M, Anders A (1979) In: Klaus SN (ed) Pigment cell; vol 4, Karger, Basel, pp 142-149

39. Anders F, Diehl H, Scholl E (1980) Linnean Soc Symp Ser 9:211-224

40. Constantinides PG, Taylor SM, Jones PA ( 1978) Dev Biology 66: 57- 71

41. Jones PA, Taylor SM (1981) Nucleic Acid Res 3:2933-2947

42. Vielkind U (1976) J Exp Zoo1196: 197-204

43. Anders A, Dess G, Nishimura S, Kersten H (to be published) In: Bagnara JT, Klaus S, Paul E, Schartl M (eds) Pigment Ce111985

44. Kersten H (1982) In: Jaenicke L (ed) Bio chemistry of differentiation and morphogen esis. Springer, Berlin New York, pp 116-119

45. Kersten H (1983) in: Nass G (ed) Recent re sults in cancer research: modified nuc leosides and cancer. Springer, Berlin New York, pp 255-263

46. Nishimura S (1983) Prog Nucleic Acid Res Mol BioI 28:49-74

47. Katze JR, Beck WT, Cheng CS, McC1oskey JA (1983) In: Nass G (ed) Recent results in cancer research: modified nucleosides and cancer. Springer, Berlin New York, pp 146 -159

48. Katze JR, Gündüz U, Smith DL, Cheng CS, McCloskey JA (1984) Biochemistry 23: 1171 -1176

49. Sprinzl M, Gauss DH ( 1982) Nucleic Acids Res 10:rl-r55

50. Okada N, Nishimura S (1979) J BioI Chem 254:3062-3066

51. Kersten H, Schachner E, Dess G, Anders A, Nishimura S, Shindo-Okada N ( 1983) In: Curtis HC, Pfleiderer W, Wachter H (eds) Biochemical and pteridines. Walter pp 367-382

52. Nass G (1983) Recent results in cancer re search. Modified nucleosides and cancer. Springer, Berlin New York

53. Barnekow A, Schartl M (1984) Mol Cell Biol 4: 1179-1181

54. Takeya T, Hanafusa H (1982) J Virol 44: 12-18

55. de Klein A, van Kessel AG, Grosveld G. Bartrarn CR. Hagemeijer A, Bootsma D, Spurr NK, Heisterkamp N. Groffen J, Stephenson JR ( 1982) Nature 300: 765- 767

56. Duesberg P (1983) Nature 304:219-225

57. Bishop JM (1982) Sci Am 246:68-78

58. Weinberg RA (1983) Sci Am 249: 102-116

59. Hunter T (1984) Sci Am 251 :60-69

60. Kraybill HF, Dawe C.J, Harshbarger .JC. Tar diffRG (1977) Ann NY Acad Sci 298

61. Kaiser HE (1981) Neoplasms -comparative pathology of growth in animals, plants and man. Williams and Wilkins, Baltimore

62. Dawe Cl, Harshbarger .JC. Kondo S, Sugimura T, Takayama S (1981) Phyletic approaches to cancer. Japan Scientific Societies Press, Tokyo

63. Schmiedel A (1984) Thesis. University of Giessen