Institute of Cancer Research, College of Physicians & Surgeons, Columbia University, 99 Fort Washington Avenue, NewYork,N.Y.10032

I. The Strategy of the Search for RNA Tumor Viruses in Human Malignancies

Our overall purpose has been and remains to explore the possible involvement of RNA tumor viruses as etiologic or cofactor agents in human neoplasias and to exploit any leads that emerge that could be of any conceivable use in the prevention, diagnosis, or therapy of human cancer . The task of identifying the existence and the causative role of the animal RNA tumor viruses was inadvertently made easier by breeding high cancer incidence animal strains. In the process, a homogeneous genetic background was created that was permissive for viral replication. As a consequence, virus particles reached levels that made their detection inevitable. Those who are concerned with human neoplasias are for the most part faced with the same difficulties encountered by the 1 early animal oncologists prior to the availability of inbred strains. It follows that a search for putative human viral agents requires more sensitive devices than those which sufficed to establish their presence in the genetically homogeneous animal systems. In the quest for such tools, we quite naturally turned to molecular hybridization and the other methodologies developed by molecular biologists in the past several decades. Our investigations evolved through a number of stages that are conveniently identified by the questions we posed for experimental resolution : 1) Do human neoplasias contain RNA molecules possessing detectable homology to the RNA of tumor viruses known to cause similar cancers in other mammals? 2) If a positive outcome is obtained, do the RNA molecules identified in tumors possess the size and physical association with reverse transcriptase that characterize the RN A of the animal oncornaviruses ? 3) If such RNA exists in human tumors, is it encapsulated in a particle possessing the density and size of the RNA tumor viruses? 4) Is the RNA of human tumor particles homologous to the RNA of the viruses causing the corresponding disease in animals? 5) The "virogene-oncogene" concept proposes that all animals prone to cancer carry in their germ line a complete copy of the information required to convert a cell from normal to malignant for the production of tumor virus particles. Is this concept valid for randomly bred populations and, in particular, for the human disease ?

II. The Animal Models as a Point of Departure

When we began our investigations, there were relatively few animal oncornaviruses a vailable in amounts adequate for the sort of biochemical experiments required. Table I lists these and records certain relevant features that served as a

Table I: Comparsion of Some Representative Oncornaviruses


*The results of molecular hybridizations between [³H]DNA complementary to the various RNAs and the indicated RNAs. The plus sign indicates the hybridizations were positive and the negative sign indicates none could be detected.

guide in these expermients. There are two avian viruses, myeloblastosis virus (AMV) and the Rous sarcoma virus (RSV), that cause mesenchymal malignancies in chickens. In addition, we have the murine leukemia virus (MuL V) and the murine sarcoma virus (MuSV) that induce similar diseases in mice. Finally, we have the murine mammary tumor virus (MMTV), which is the unique etiologic agent for mammary tumors.
When these viruses are examined for sequence homologies amongst their nucleic acids, a rather informative pattern emerges. I twill be noted that the two chicken agents have sequences in common, but do not show detectable homology with any of the murine agents. Turning to the murine viruses, we find that the nucleic acids of the leukemia, lymphoma, and sarcoma agents are homologous to one another, but not to either of the two avian agents or to the mammary tumor virus. Finally, the mouse mammary tumor virus has a singular sequence homologous on]y to itself.
It is important to understand that a plus sign does not indicate identity, but simply sufficient similarity to be detectable by the relaxed hybridization conditions used in these initial studies. Similarly, a negative sign does not imply the complete absence of sequence homology, but rather that none was observable by the procedures used.
If analogous, or similar, virus particles are associated with the corresponding human diseases, certain predictions may be hazarded on the basis of the specificity patterns exhibited in Table I, and they may be listed as follows:

(a) In view of the lack of homology between the avian and murine agents, it is unlikely, from simple evolutionary considerations, that human agents, should they exist, would show more homology to the avian group than to the murine oncornaviruses.

(b) It follows that the murine tumor viruses would represent the more hopeful source of the molecular probes required to search for similar information in the analogous human cancers.

(c) If particles are found to be associated with human mesenchymal tumors (in leukemias, sarcomas, and lymphomas), their RNAs might show homology to one another and possibly to that of the murine leukemia virus.

(d) If RNA particles are identified in human breast cancer, they should not exhibit homology to the RNA of virus-like particles associated with the human mesenchymal neoplasias or to MuL V RNA, but might exhibit some homology to the RNA of the murine mammary tumor virus.
On the basis of both availability and the specificity considerations outlined above, it is clear why the murine agents were initially chosen for producing the necessary molecular probes to look for corresponding information in the human disease. Furthermore, the desire to monitor the biological consistency of our findings dictated that we examine in parallel the human neoplasias listed. Such a parallel examination would permit us to determine whether our findings in humans mirrored biologically what was known from the animal experimental models. For this purpose, we focused our efforts on the mesenchymal neoplasias and on breast cancer.

III. Molecular Hybridization with Radioactive DNA Probes


The DNA-RNA hybridization procedure we used to answer the question whether human tumors contain viral-related RNAs was one that we had designed (1) some fifteen years ago to answer questions of almost precisely this nature in the case of virus-infected bacteria. The method depends on the ability of any piece of single-stranded DNA to find its complementary RNA and form, under the proper conditions, a double-stranded hybrid structure. The reaction is highly specific and has proved to be of considerable value in molecular biology over the past decade.
The required radioactive DNA was synthesized by supplying detergent-disrupted virus preparations with magnesium and the deoxyrjboside triphosphates, with one of them being labeled with tritium. When the synthesis is completed, the protein and the RNA present are eliminated, and the residual radioactive DNA is purified to completion. Each [³H]DNA preparation is then rigorously examined for specific hybridizability to its appropriate template and for its inability to complex with irrelevant RNAs. After satisfying the specificity criteria, the purified viral-specific tritiated DNA is mixed with cytoplasmic RNA prepared from a variety of tumors and annealed under the conditions described in Figure 1. The hybridizations are always carried out with a vast excess of tumor RNA. Since the viral-specific tritiated DNA is small compared with the RNA, any complexes formed between them will behave physically more like RNA than DNA. Such complexes are readily detected by isopynic separation in equilibrium density gradients of cesium sulfate. At the end of the centrifugation, the distribution of the tritiated DNA is examined across the gradient. Any uncomplexed DNA will remain at a density corresponding to about 1.45. The DNA molecules that have annealed either partially or completely to RNA will band at or near the density of RNA (epsilon = 1.65). The movement of the tritiated DNA from the DNA density region to the RNA density region is then the signal that the probe used has found complementary sequences in the tumor RNA with which it is being challenged.

The human neoplasias examined included adenocarcinoma of the breast (2-4), the leukemias (5), the sarcomas (6), and the lymphomas (7). The leukemias encompassed both acute and chronic varieties of the lymphatic and myelogenous types. The human sarcomas studied included fibro, osteogenic, and liposarcomas. The lymphoma series contained Hodgkin's disease, Burkitt's tumors, lymphosarcomas, and reticulum cell sarcomas. Control adult and fetal tissues were always examined in parallel, and these were invariably negative. In the case of breast tissue, the two benign diseases, fibroadenoma and fibrocystic disease, were also included and were found to be negative.
Table II summarizes in diagramatic form the outcome of the survey of human neoplasias with the animal virus probes. The pluses signify that the corresponding tritiated DNA complexed with the indicated tumor RNAs and the minuses, that no such complexes were detected. The positives in these earlier studies ranged from 67 0/0 for breast cancer to 92 0/0 for the leukemias. What is most noteworthy of the pattern exhibited in Table II is its concordance with predictions deducible from the murine system. Thus, human breast cancer contains RNA homologous only to that of the murine mammary tumor virus. The human leukemias, sarcomas, and lymphomas all contain RNA sharing sufficient homology to that of the Rauscher murine leukemia virus (RL V) to make a stable duplex. These mesenchymal neoplasias contain no RNA homologous to the MMTV RNA. Finally none of the human tumors contains RNA detectably related to that of the avian myeloblastosis virus. The homology of leukemic RNA to that of RL V and the homology of RNA from human breast cancer to that of MMTV have been confirmed (8,9).
In summary, the specificity pattern of the unique RNA found in the human neoplasias is in complete agreement with what has been described for the corresponding virus-induced malignancies in the mouse.

Table II: Homologies among Human Neoplastic RNAs and Animal Tumor Viral RNAs


The results of molecular hybridization between [³H]DNA complementary to the various viral RNAs and pRNA preparations from the indicated neoplastic tissues. The plus sign indicates that hybridizations were positive and the negative sign, that none could be detected (5).

IV. TheSimultaneous Detection Test


The existence of RNA in human tumors having sequence homology to virus particles causing homologous diseases in mice does not of course establish a viral etiology for these diseases in man. The next step requires the performance of experiments designed to answer the second and third questions raised in the introductory paragraphs, i. .e., those relating to the size of the RNA being detected and whether it is associated with the reverse transcriptase in a particle possessing other features of complete or incomplete oncornaviruses.
What we sought was a method of detecting the presence of particles similar to the RNA tumor viruses that would be simple, sensitive, and sufficiently discriminating so that a positive outcome could be taken as an acceptable signal of the presence of a viral-like agent. To achieve this goal, we devised a test that depended on the simultaneous detection of two diagnostic features of the animal RN A tumor viruses.
The oncornaviruses exhibit two identifying characteristics. They contain a large (1 x 107 daltons in molecular weight and composed of subunits each of which is 3 x 106 daltons) single-stranded RNA molecule having a sedimentation coefficient of 705, or 355 if the 705 molecule has broken down into its subunits. They also have reverse transcriptase (10, 11), an enzyme that can use the viral RNA as a template to make a complementary DNA copy.
The possibility of a concomitant test for both the enzyme and its template was suggested by our prior experience with RNA transcriptase in which we found (12) that the growing RNA chain could be detected as a complex with its DNA template on removal of the protein from the reaction mixture. Similar observations were made in examinations of the early reaction intermediate (13, 14) of the reverse transcriptase reaction.
I t was on this basis that 5chlom and 5piegelman ( 15) developed the simultaneous detection test that was used to demonstrate ( 16) the presence in human milk of particles containing 705 RNA and the reverse transcriptase. The test was modified (17) to be applicable to tumor tissue using the mouse mammary tumor as the experimental model.
Figure 2 diagrams the procedure used. Tumor cells are first broken by the use of the Dounce homogenizer and nuclei, mitochondria, and large cell membrane fragments removed by low speed centrifugation. The supernatant is subjected to trypsin digestion to inactivate any nucleolytic enzymes and the trypsin is neutralized by trypsin inhibitor. The supernatant is then centrifuged at 150,000 X g to yield a cytoplasmic pellet containing virus particles, if present. The resulting pellet is then banded isopycnicly in a sucrose density gradient and the fraction between 1.16 and 1.19 g/ml is collected by centrifugation. The recovered pellet is then treated with a nonionic detergent (NP40) to disrupt possible viral particles, and the disrupted preparation is used in a brief endogenous reverse transcriptase reaction. The product of the reaction, with its RNA template, is freed of protein and analyzed in a glycerol velocity gradient to determine the sedimentation coefficient of the tritiated DNA. In addition, the product is subjected to equilibrium centrifugation in a Cs2SO4 gradient to determine its density.
The presence of particles encapsulating 705 RNA and a reverse transcriptase will be indicated by the appearance of a peak of newly synthesized DNA traveling at a speed corresponding to either a 705 RNA or a 355 RNA molecule. That the apparently large size of the [³H]DNA is due to its being complexed to an RNA molecule can be readily verified by subjecting the purified nucleic acid to ribonuclease prior to velocity examination. The disappearance of the 705 and 355 [³H]DNA peaks following RNase treatment proves that the [³H]DNA was complexed to large RNA molecules. Similarly, if the reaction is positive, newly synthesized DNA should appear in the RNA and/or hybrid regions of the Cs2S04 gradient, and these peaks should again be eliminated by prior treatment with ribonuclease.
The simultaneous detection test was first applied to human breast cancer (18) in a series including 38 adenocarcinomas and ten non-malignant controls. It was found that 79 0/0 of the malignant samples were positive for the simultaneous detection reaction and all of the control samples from normal and benign tissue were negative. It was further shown that the particles possessing the reverse transcriptase activity and its 70S RNA template localize at a density between 1.16 and 1.19 g/ ml, the densi ty characteristic of the oncogenic viruses.
The data obtained therefore indicate that one can, with a high probability, find in human breast cancers particulate elements of the right density that encapsulate RNA-instructed DNA polymerase and a 70S RNA.

Fig. 2: Simultaneous detection test for 70S RNA and reverse transcriptase in neoplastic tissue (see text for further details).

V. Application of the Simultaneous Detection Test to Mesenchymal Tumors


In our initial study of the leukemias (19), peripheral leukocytes were prepared from the buffy coats of both leukemic and nonleukemic control patients. Cells were disrupted and fractionated as described in Figure 2. Representative experiments examining the effects of ribonuclease treatment of the product and omission of one of the deoxytriphosphates during the reaction are shown in Figure 3. We see the telltale 70S peaks of DNA synthesized by the pellet fractions from the leukocytes of patients with acute lymphoblastic and acute myelogenous leukemias. The elimination of the complex by prior treatment with ribonuclease (Figure 3A) shows that the tritiated DNA is indeed complexed to a 70S RNA molecule. Further, the omission of dATP (Figure 3B) leads to a failure to form the 70S complex, a result expected if the reaction is in fact leading to the synthesis of a proper heteropolymer. In similar experiments, it was shown that omission of either dCTP or dGTP also resulted in the absence of the 70S RNA-[³H]DNA complex, all of which argues against nontemplated end addition reactions.

In some cases leukemic cells were obtained in amounts adequate to permit a more complete characterization of the product. Hybridization of the human product to the appropriate viral RNAs provides the most revealing information since it tests sequence relatedness to known oncogenic agents. We summarize in Table III the results of examining the peripheral leukocytes of 23 patients, all in the active phases of their disease, including both acute and chronic leukemias. Of the 23 leukemic patients examined, 22 showed clear evidence that their peripheral leukocytes contained particles mediating a reaction leading to the appearance of endogenously synthesized DNA in the 70S region of a glycerol gradient. Nine of these were tested for ribonuclease sensitivity and in all cases the complexes were destroyed. In nine others, the DNA was recovered from the complex and annealed to RLV RNA and

Table III: Simultaneous Detection of 705 RNA and Reverse Transcriptase in Leukemic Cells (19)

Table IV: Simultaneous Detection on Mesenchymal Tissues

The simultaneous detection test was carried out as described in Figure 2. Peripheral white blood cells (WBC) were used in the leukemias, acute myelogenous (AML), chronic myelogenous (CML), acute lymphocytic (ALL) and chronic lymphocytic (CLL).

VII. On the Problem of Germ-line Transmission of Viral Information


We now come to grips with the fifth question raised in the introductory paragraphs, the virogene-oncogene concept (22), which derives from animal experiments and argues that all animals prone to cancer contain in their germ line at least one complete copy of the information necessary and sufficient to convert a cell from normal to malignant and produce the corresponding tumor virus. This hypothesis presumes that the malignant segment normally remains silent and that its activation by intrinsic or extrinsic factors leads to the appearance of virus and the onset of cancer .
There are various ways of testing the validity of the virogene-oncogene hypothesis, but the pathways differ in the technical complexities entailed. One approach commonly used attempts to answer the question: Does every normal cell contain at least one complete copy of the required viral-related malignant information? The methodologies used included the techniques of genetics, chemical viral induction, and molecular hybridizations. However, for a variety of reasons, none of these gave, or could give, globally conclusive answers. Genetic experiments do not readily distinguish between susceptibility genes and actual viral information. Further, even if genetic data succeeded in identifying some structural viral genes, it would still be necessary to establish that all the viral genes are represented in the genome. Attempts to settle the question by demonstrating that every cell of an animal can be chemically induced to produce viruses have thus far, for obvious reasons, not been tried. The best that has been achieved along these lines is to show that cloned cells do respond positively. However, the proportion of clonable cells is small and clonability may well be a signal for prior infection with a tumor virus.
Finally, the quantitative limitations of molecular hybridization make it almost impossible to provide definitive proof that each cell contains one complete viral copy in its DNA. Although it is not very difficult to show that 90 0/0 of the information is present, it is the last 10 0/0 that constitutes the insurmountable barrier and 10% of 3 x 10 high 6 daltons amounts to a far from trivial 3 x 10 high 5 daltons, the equivalent of about one gene.
A useful way to obviate these technical difficulties is to invert the problem. Instead of asking whether one complete copy exists in normal cells, the question can be phrased in the following terms: Does the DNA of a malignant cell contain viral-related sequences that are not found in the DNA of its normal counterpart? Phrasing the issue in this manner leads to the design of experiments that avoid the uncertainties generated by the demonstrated fact that many indigenous RNA tumor viruses share, completely or partially, some sequences with the normal DNA of their natural hosts (23 ). The crucial point is of course whether all of the viral sequences are to be found in normal DNA. The approach we adopted requires removal of those viral sequences that are contained in non-neoplastic DNA by exhaustive hybridization of the viral probe to normal DNA in vast excess. Any unhybridized residue can then be used to determine whether malignant DNA contains viral-related sequences not detectable in normal tissue.
We first investigated this question in the case of the human leukemias (24) and the strategy, as diagrammed in Fig. 4 (a and b), may be outlined in the following steps:

a) Isolate from leukemic cells the fraction enriched for the particles encapsulating the 70S RNA and RNA-directed DNA polymerase;

b) Use this fraction to generate [³H]DNA endogenously synthesized in the presence of high concentrations of actinomycin D to inhibit host and viral DNAdirected DNA synthesis;

c) Purify the [³H]DNA by hydroxyapatite and Sephadex chromatography with care being exercised to remove by self-annealing and column chromatography all self-complementary material in the tritiated probe;

d) Use the resultant [³H]DNA to detect complementary sequences in normal and leukemic leukocyte DNA;

e) If viral-related sequences are detected in both, remove those found in normal leukocytes by exhaustive hybridization to normal DNA; and

f) Test the residue for specific hybridizability to leukemic DNA.
In carrying out the recycling and test hybridizations, it is imperative that conditions be chosen to account for the possibility that the leukemia-specific sequences are present in only one copy per genome, a possibility which is in fact realized (24). To this purpose, the concentration in moles per liter (Co) of DNA and the time (t in seconds) of annealing is adjusted to Cot values of 10,000, which are adequate to locate unique sequences.

A typical outcome of hybridizing such recycled tritiated DNA to normal and leukemic DNA is shown in Fig. 5. It is evident that no complexes stable at temperatures above 88° are formed with normal DNA. On the other hand, 57 0/0 of the recycled [³H]DNA probe forms well-paired duplexes with leukemic DNA. A series of such experiments was performed with particle-generated [³H]DNA and nuclear DNA obtained from 8 untreated patients with either acute or chronic myelogenous leukemia. In every case (Table V), the [³H]DNA, after being sub-

jected to exhaustive annealing to normal DNA, yielded a residue that forms stable duplexes only with leukemic DNA, in agreement with the experiment of Fig. 5.
In estimating the implication of these results, it must be recalled that the leukemia-specific sequences found (24) in leukemic cells are present as non reiterated copies per genome. This was established by the Cot values ( concentration of nucleotides X time) required to detect them. The sensitivity used to examine normal cells for the leukemia-specific sequences was such that l/50th of an equivalent of that found in leukemic cells would have been readily detected. Consequently, one may conclude that the vast majority of normal cells do not contain this particular stretch of malignant-associated information and it cannot therefore be represented in the germ line of nonleukemic individuals.

VIII. Unique Sequences in Hodgkin's and Burkitt's Lymphomas and their Relatedness


We have already noted that, like the leukemias, Hodgkin's and Burkitt's lymphomas have particles containing reverse transcriptase and a 70S RNA template related in sequence to that of RL v. It was of obvious interest to determine whether the lymphomas also parallel the leukemias in possessing a unique sequence not detectable in normal tissue. If they do, one can in addition ascertain whether the sequences found in Hodgkin's and Burkitt's lymphomas are related to each other. The outcome has evident significance for the possible relevance of the sequence to malignancy.
[³H]DNA probes were synthesized with particles isolated from four Burkitt's tumors and three Hodgkin's disease specimens. Sequences shared with normal DNA (between 35010 and 400/0) were then removed as described for the leukemias (24) to yield the recycled [³H]DNA probes (25).

Figure 6 shows the outcome of challenging recycled Hodgkin's disease [³H]DNA with nuclear DNAs from normal spleen, Hodgkin's disease spleen, and from Burkitt's lymphoma. Few, if any, stable duplexes are formed with normal DNA. Note, however, that although the probe was made with Hodgkin's disease particles, the [³H]DNA hybridized to Burkitt's lymphoma nuclear DNA virtually as well as it complexed to DNA from Hodgkin's spleen. The converse is also true, as may be seen from Table VI, which summarizes the results of our findings in the recycled [³H]DNA challenged with nuclear DNA from normal and malignant tissues (25 ). Again, normal DNA is unable to form significant amounts of stable complexes (elution at 88° and above) with the [³H]DNA probes. In all instances, the lymphoma [³H]DNAs hybridized in stable complexes to the nuclear DNA of the original types from which the particles were obtained and used to generate the labeled DNA. Further, with only one exception, all of the Burkitt's and Hodgkin's disease [³H]DNAs cross hybridize with each other's DNA.
In summary, several features emerged from this study of the lymphomas. The particle-related sequences found in Burkitt's and Hodgkin's lymphomas possess sequences in common, an observation in accord with our earlier findings (7, 20, 21, 26 ), that Hodgkin's and Burkitt's particles both share sequences with the Rauscher murine leukemia agent. Further, in view of the previous association of the Epstein-Barr virus with Burkitt's lymphoma (27, 28) and the non-neoplastic infectious mononucleosis (29, 30), it is revealing to note from Table VI that the leukocyte DNA of patients with infectious mononucleosis was devoid of the Burkitt's sequences detected by the recycled [³H]DNA lymphoma probe, indicating that these latter sequences are specific for neoplastic tissues. The fact that the particle-related sequences in Hodgkin's and Burkitt's tumors are related to each other adds further weight to this conclusion. Finally, the observation that cells carrying multiple copies of the DNA of the Epstein-Barr virus do not complex with recycled [³H]DNA probes from either Hodgkin's or Burkitt's particles proves that these particle sequences have no detectable relation to the DNA of the Epstein-Barr virus.

IX. Evidence from Studies of Identical Twins


Although the comparison of leukemic patients with normal suggests that healthy individuals do not contain the leukemia-specific sequences, the data do not rule out the possibility that those who do come down with the disease do so because they in fact inherit the required information in their germ line. One way to resolve this issue is to study the situation in identical twins. Since identical twins are monozygous, i. e., derive their genomes from the same fertilized egg, any chromosomally transmitted information must be present in both. It had already been shown by Goh and his colleagues (31, 32) in the case of chronic myelogenous leukemia that only the leukemic member of each of two identical twin pairs contained the marker Philadelphia chromosome. It was of obvious interest to examine this situation for the leukemia-specific sequences. If the leukemic member of the pair contains the particle-related DNA sequences, and does so because he inherited them through his germ line, then these same sequences must be found in

X. Implications of the Unique DNA Sequences in Leukemias and Lymphomas


The data we have summarized on the existence of virallateral-related sequences unique to the DNA of human malignant cells imply that they are inserted in somatic DNA, a process known to occur with RNA tumor viruses in animal cells in tissue cultures (34) and in whole animals (35). The fact that these viral sequences can be incorporated into somatic DNA suggests that this could also occur in early embryogenesis and thus involve a cell destined to differentiate into the germ line. An event of this nature would be selected for in any attempts at developing inbred strains characterized by high frequency of cancer .
Indeed this seems to have occurred in the course of producing the AKR mouse, a strain in which spontaneous leukemia occurs with virual certainty. It has been shown (36) that the DNA of the AKR mouse contains murine leukemia virus sequences that are not present in the DNA of the NIH Swiss mouse. These sequences were localized by genetic and molecular hybridization, and it was found that they are either identical to or closely linked to the Akv-1locus.

Fig. 7: Hydroxyapatite elution profile of a hybridization reaction of the recycled leukemic twin [³H]DNA probe to nuclear DNA from normal leukocytes, normal twin leukocytes, and leukocytes from the same leukemic twin. The annealing reaction mixtures contained 20 A260 units of cellular DNA, 0.004 pmol of [³H] DNA, and 15 µmol NaH2PO4 (pH 7.2) in a final vol of 0.01 ml. The reaction was brought to 98 ° for 60 sec and 0.04 mmol of NaCI was added. The reaction mixture was then incubated at 60° X 50 hr. The reaction was stopped by the addition of 1 ml of 0.05 M NaHP04 (pH 6.8). The sample was then passed over a column of hydroxyapatite of 20-ml-bed v01 at 60°. The column was washed with 40 ml of 0.15 M NaHP04 (pH 6.8) at 60°,80°,88°, and 95°. Fractions of 4 ml were collected, the A260 of each fraction was read, and the DNA was precipitated with 2 µg/ml of carrier yeast RNA and 10 per cent trichloroacetic acid. The precipitate was collected on Millipore filters, which were dried and counted. In all cases, greater than 80 per cent of the nuclear DNA reannealed. A background count of 8 cpm was subracted in all instances (33).

The case of the AKR mouse was very likely an inadvertent result of its selection. An even more remarkable instance is the deliberate insertion of viral sequences into the germ line (37). This was accomplished by infection of preimplantation mouse embryos at the 4-8 cell stage with the murine leukemia virus (MuL V) followed by reimplan.tation in the uteri of surrogate mothers. Of 15 such animals born, one developed lymphatic leukemia at 8 weeks of age. Molecular hybridizations revealed leukemia-specific viral sequences in the DNA of all eight different organs examined, whether they were of mesenchymal origin or not. In agreement with our earlier findings (35), these sequences were not found in normal mesenchymal tissue nor were they detected in the DNA of non-target tissues in animals made leukemic by injection of virus after birth.
In summary, except for strains deliberately inbred for high spontaneous occurrence of disease, the mesenchymal neoplasias of mice and men would appear tO have a similar underlying mechanism. In both instances new viral-related sequences are found in the DNA of the malignant cells and these are not found in the DNA of uninvolved tissues. They are therefore not germinal unless one wishes to invoke a rather unlikely specific elimination in the course of the differentiation of every cell but the malignant one. Despite its implausibility, this possibility should be exploited by testing for the leukemic-specific sequences in the germ line DNA (sperm) of leukemic individuals.
It must be emphasized that conclusions as to the validity of the virogene-oncogene hypothesis are only relevant to the particular instances examined and cannot be generalized to any other viral-related cancers even in the same animal, let alone to other species. In any event, it is evident that our findings with respect to the human mesenchymal tumors suggest more optimistic pathways for the control of these diseases than would be available if the total information were already in the genome. The data imply that we may not be forced to master the control of
our own genes in, order to cope with these neoplasias.
The fact that the human particles possess sequences homologous to those found in viral agents known to cause the corresponding neoplasias in mice encourages the hope that they are relevant to human disease. However, despite the considerable progress that can be recorded, no definitive proof exists at the present writing that the virus-like particles found in the human neoplasias are either viruses or etiologic agents of the cancers in which they are found. Proof will ultimately come when it proves possible to produce the relevant malignancy in a susceptible animal by injection of the particles purified from human tumors. It should, however, be noted that we have known of the mouse mammary tumor virus for more than 35 years and no one has yet succeeded in producing mammary tumors with this agent in any animal other than the mouse.
Under the circumstances, it would seem prudent not to wait for the definitive experiment with the human particles, but rather to proceed with attempts at further exploration of their significance and possible clinical usefulness. We should like to list a few areas of possible exploitation and then turn our attention to a brief description of what has been accomplished along these lines.

1. The existence of the leukemia-specific sequences can provide the clinician with a hitherto unsuspected parameter that could potentially be a useful adjunct in monitoring therapy.

2. One could attempt to grow the human particles in tissue culture in order to provide a more accessible source of these particles for further biochemical characterization with the ultimate hope of generating useful reagents for diagnostic, therapeutic or monitoring purposes.

3. Another, less ambitious approach is to purify one of the protein subcomponents from the human particles for further characterization. This could then be used for the production of a monospecific antiserum that might be clinically useful.

XI. Particulate Reverse Transcriptase in the Leukocytes of Leukemic Patients in Remission


We have already noted (Table V) that positive simultaneous detection tests, indicating the presence of particles containing reverse transcriptase and the 70S RNAtemplate, were obtained in more than 99°% of the leukemic patients examined. It was of obvious interest to see whether these particles could be detected in the leukocytes of leukemic patients who are in good clinical remission.
Peripheral blood leukocytes were obtained from patients at the Baltimore Cancer Research Center and from the M. D. Anderson Hospital. The leukocytes from some of the leukemic patients were obtained by leukophoresis and immediately stored at -7° until used. The clinical statuses of the patients at the time of leukophoresis are summarized in Table VII. A complete remission was defined as the absence of symptoms related to the disease, normal results on physical examination, a hemoglobin of greater than 10 g/100 ml, leukocyte count greater than 3000/mm³, platelet count greater than 100,000/mm³, no blasts in the peripheral blood smear, and less than 5 010 blasts in the bone marrow.
Table VIII summarizes the results of simultaneous detection assays for high molecular weight RNA and reverse transcriptase in the leukocytes from the patients examined. Outcomes are designated as positive only when the peaks of tritiated DNA found in the 705 and 355 regions were eliminated by prior treatment with ribonuclease, a feature establishing that the [³H]DNA is complexed to a large RNA molecule. If the peaks are not removed subsequent to RNase digestion, the reaction is scored as a negative outcome. In the present study two untreated leukemic patients were available for testing prior to remission induction and both were positive at that time. Three of the nine patients in complete remission demonstrated a 70S or 35S peak of acid-precipitable radioactivity that was abolished by RNase treatment. The "negatives" were subjected to a simultaneous detection assay via a cesium sulphate gradient, a procedure that obviates the problem generated by fragmentations of the RNA template during manipulation.
It will be noted from Table IX that samples from the untreated patients were all positive. The [³H]DNA-RNA hybrids were detected in nine out of eleven patients in complete remission. In this group, six of seven AML patients and two of three acute lymphocytic leukemia (ALL) patients demonstrated a positive reaction. In five of the patients, the simultaneous detection tests were negative by velocity sedimentation analysis but were positive when analyzed in the cesium

Table VII:



+AML = acutc myelogenous leukemia; ALL = acute lymphocytic leukemia
*UN = untreated; CR = complete remission; REL = re]apse
Clinical status of leukemic patients when leukophoresis was performed for enzyme studies.



Table VIII:



Test for 70S and 35S RNA-[³H]DNA in leukocytes from leukemic patients. CPM in 70S or 35S represents acid-precipitable radioactivity that was removed from the 70S and 35S by prior treatment with ribonuclease A and T 1.

sulphate gradients. Hybrids of small molecular size would not be identified as 70S with 35S complexes, but can be detected as complexes in the hybrid region of the cesium sulphate gradient. Only one of the patients (#9) was negative by both glycerol and cesium sulphate gradient analyses. This patient did not appear to differ clinically at the time of the examination from the other remission patients exhibiting positive reactions. In the course of these studies we also examined by cesium sulphate analysis two pooled, normal white blood cell samples and five non-neoplastic spleens for the presence of particles capable of yielding RNAtritiated DNA hybrids in an endogenous reaction and all were negative, as had been true in our previous studies (Table V).
It is obvious that finding the leukemia characteristic particles in the white blood cells of patients in remission is disappointing and does not accord with the generally accepted assumption that there is a normal and a leukemic population of leukocytes in acute leukemia (38). The goal of contemporary chemotherapy and immunotherapy is to reduce the size of the leukemic component (to zero if possible) to allow the bone marrow and the peripheral blood to repopulate with non-neoplastic cells. A number of clinical observations suggest that remission leukocytes are in fact normal cells. First, the prolongation of life is directly proportional to the duration of the remission. Secondly, a small but increasing number of patients with acute lymphoblastic leukemia in long-term remission appear to go on to cure, indicating a permanent extinction of the leukemic cell population (39).
The morphology and functional properties of remission leukocytes have been studied by a number of techniques. These include karyotype analysis (40-45), ability to form colonies in agar ( 46-49), and the detection of leukemia-related antigens (50, 51). In general, these studies have supported the concept that remission leukocytes represent the return of a population of normal cells. Also, relapses are usually heralded by the detection of the abnormality associated with leukemic cells, and a number of these techniques have been suggested as ancillary tools in following the response of patients to chemotherapy. However, there have been instances in the above reports of patients in well-consolidated remissions whose peripheral leukocytes or bone marrow cells demonstrated persistence of aneuploidy, leukemia-related antigens, or abnormal colony formations in agar; these abnormalities appear unrelated to the effects of maintenance of chemotherapy. In this connection, mention should be made of Killmann's deductions (52) based on the demonstrated capacity of leukemic cells to differentiate; on these grounds he questions whether the normal-looking cells observed in the bone marrow of AML patients in remission are in fact derived from non-leukemic ancestor cells.
The data presented indicate that with regard to certain biochemical markers, which may be virus-related, remission leukocytes may more closely resemble the leukemic cells than normal cells. Quite surprisingly, two of the three ALL patients in long-term remission still had evidence of particles in their peri pheralleukocytes. Further, the enzyme was detected in seven of eight patients with AML in remission. We are unable to determine if all or a fraction of the peripheral white cells studied possessed the leukemic characteristics. Thus we cannot directly answer the question whether one or two white cell populations are present.
Mak and his colleagues (53, 54) have described particulate activity in the supernatants of short-term cultures derived from bone marrow of leukemic patients in remission. The activity of the cultures from remission patients equaled, and in some instances exceeded, that detected in the cultures derived from patients in relapse.
There are a number of plausible explanations for the persistence of the particulate enzyme and its associated template in the remission leukocytes. The normal cell found in remission could be infected with a non-oncogenic C-type virus or conversely the remission leukocyte could have acquired resistance to transformation whereas susceptibility to infection was unaltered. Second, as a result of chemotherapy, a portion of the leukemic clone could have evolved into a non-neoplastic clone still capable of expressing some viral function. There are a number of in vitro models for the latter phenomenon. Thus, it has been shown that cells transformed with a murine sarcoma virus can spontaneously, or after exposure to antimetabolites, revert to a normal morphology. Certain clones of these morphological revertants behave in a non-malignant manner, yet some viral functions are expressed or can be induced (55).
A more direct method for examining such questions is to use the molecular hybridization to answer the following questions: 1) Do remission cells have leukemia-specific DNA nucleotide sequences? 2) If present, are some leukemia-specific DNA sequences not expressed or are critical DNA sequences deleted? These areas are presently under investigation.

In these experiments, cDNA probe synthesized endogenously with HL23V particles, was annealed to RNAs from SSV-1 or BV-M7 or both. Figure 10 shows that the HL23V-cDNA hybridized 37 % and 44 % to the RNAs of BV-M7 and SSV-1, respectively. These individual hybridizations were additive as demonstrated by the complete complexing of HL23V-cDNA to a mixture of BV-M7 and SSV-1 RNAs. These data indicate that the genetic information of HL23V virions is completely accounted for, within the limits of the sensitivity of the molecular hybridization technique used, by the complete genomes of both SSV-1 and BV-M7.
To supplement and confirm these findings by an independent method, competition molecular hybridizations were performed using cDNA synthesized from SSV. Viral RNA from SSV was labeled with 125I. SSV-cDNA can protect this 125I
SSV-RNA more than 79 % from ribonuclease digestion at molar ratios of 5:1 (cDNA:RNA). Table X shows that when unlabeled viral RNAs were added in vast excess to compete this homologous reaction, both SSV and HL23V RNA

Table x: Analysis of HL23V for SSV Genomic Content by Competition Hybridization

Hybridization reactions (5.5 µl) were performed as descibed in Fig. 9 and contained 0.11 ng 1251-55V RNA (1.2 X 108 cpm/µg), 0.68 ng [³H]-55V cDNA (2 X 10³ cpm/µg) and 0.5-0.25 µg of the indicated competing RNA. Following incubation at 68° for 48 h, the reactions were diluted with 0.01 M Tris-HCI, pH 8.0,0.4 M NaCI, 0.01 M EDTA and divided into four equal aliquots. Ribonuclease A (25 units/mI) and ribonuclease T 1 (5 units/ml) were added to two aliquots and the samples incubated at 37° for 1 h. Nuclease resistance was the ratio of acid-precipitable 125I in the samples with and without ribonuclease. Recovery of input acid-precipitable 1251 was greater than 90 %. SSV RNA was isolated from virions by disruption with SDS, Pronase treatment, rate sedimentation in a sucroseSDS gradient and equilibrium density gradient centrifugation in potassium iodide. Viral RNAs from GALV-l and MuLV-R were isolated by similar procedures excepting the KI gradient. HL23V RNA was total RNA from purified virus and the mouse 18S and 20S RNAs were extracted from purified ribosomal subunits from N1H/3T3 tissue culture cells. [³H]-SSV cDNA was synthesized from SSV RNA and oligo dT with AMV DNA polymerase in the presence of 0.1 mg/µl actinomyc in D and 0.5 mg/ml distamycin A. The reaction contained TTP, dATP and dGTP at 1 mM each and [³H]-dCTP (25 Ci/mMole) at 0.05 mM. This SSV cDNA protected 1251-SSV RNA 33 %, 75 % and 87 % from ribonuclease digestion at molar input ratios of 0.4,3 and 15, respectively (cDNA:RNA).

Fig. 10: Reconstruction hybridizations. HL23V [³H]-DNA probe was
hybridized to SSV and BV-M7 RNAs individually and in combination.
0 = SSV RNA, white triangle = BV-M7 RNA,
black circle = SSV and BV-M7 RNAs.

XIII. Purification of Reverse Transcriptase from Human leukemic Spleens


We have already mentioned the biologic and logistic difficulties that have attended attempts at obtaining whole human viral particles for characterization. At the present writing there exists no tissue culture source of authentic human RNA tumor viruses for use in biochemical and immunological investigations. One way to obviate these problems is to forego temporarily the more ambitious goal of characterizing the whole particle and focus rather on individual protein components of the human particles. Of these, one of the most accessible is the reverse transcriptase since its activity can be followed during fractionation.
The human leukemic reverse transcriptase has been partially purified from fresh peripheral blood cells obtained from patients with acute myelogenous leukemia (65, 66). However, the very large amounts of white blood cells required precluded definitive isolation of the enzyme in amounts that would establish its purity by gel analysis and thus permit an unambiguous characterization of its biophysical and biochemical properties. One way out of this dilemma was to explore the use of spleens as a source of leukemic enzyme. In addition to providing a possible solution of the logistic problem, such experiments would provide useful information on the relation between diseased tissue and the presence of the virus-like particles. Leukemias often involve the spleen, and splenectomies are occasionally performed therapeutically in instances of massive spleen enlargement or persistent platelet destruction.
We first worked out the technology of enzyme isolation from leukemic spleens using the murine Rauscher leukemia model. I t was found that isopycnic separation of virus particles and their conversion to cores by non-ionic detergents (67, 68) provided material suitably enriched for enzyme. Usually in such experiments 70 grams of chronic lymphocytic spleens were used. After thawing, mincing and resuspension, and homogenization, the extracts were clarified by low speed centrifugation to remove nuclei and mitochondria. The remaining particulate elements were then recovered by high speed centrifugation at 80,000 X g for 90 min. Particles were concentrated by isopycnic centrifugation in sucrose gradients and converted to cores as described previously (67, 68). After recovery, the cores were disrupted with 1 % NP-40 and 0.7 M KC1 and then subjected to column fractionations on DEAE cellulose, phosphocellulose, and agarose gels with results as described in Fig. 12. One main peak of activity is observed on the DEAE column (Fig. 12A) containing 5 % of the input and greater than 90 % of the enzyme, yielding a 19-fold enrichment. When the active fractions of the DEAE cellulose columns are pooled and then chromatographed on phosphocellulose (Fig. 12B), about 25 % of the protein contains the enzyme activity, with a recovery of 71 % and an increase in specific activity of 2.8-fold.

Fig. 12:
(A) DEAE-cellulose chromatography of leukemic spleen enzyme. Core-Iike particles isolated from a leukemic spleen were treated with 1 % Nonidet P-40 and 0.7 M KCI, and the solubilized enzyme activity was chromatographed on a 16.5 cm X 2.5 cm column. Elution was with 0.4 potassium phosphate. 10 µ of each fraction (3.2 ml) were assayed using on oligo dT-poly rA template.

(B) Phosphocellulose chromatography of leukemic spleen enzyme. The fractions from the pooled DEAE-cellulose peak activity were diluted and chromatographed on a 17 cm X 1.5 cm column. Elution was with 160 ml of an 0.01 M-0.05 M potassium phosphate gradient; 1.6.m1 fractions were collected and 10 µI aliquots assayed for oligo dT-poly rA-templated activity.

(C) Agarose gel filtration of leukemic spleen enzyme. The peak of activity eluted from phosphocellulose was subjected to gel filtration on a 50 cm X 0.9 cm agarose column. The elution rate was 4 ml/hr and 0.4 ml fractions were collected. Aliquots of 4 µI were assayed for enzyme activity with an oligo dT-poly rA template.



The concentrated phosphocellulose enzyme is then passed through a 0.5 M agarose column on which one routinely observes two peaks of activity (fig. 12C). If the first peak of activity is rechromatographed on the same type of column, a shift of most of the activity to the position of the second peak occurs. Thus, the first peak would appear to be an aggregate (possibly a dimer) of the enzyme. A summary of the column fractionations in terms of recoveries and specific activities is recorded in Table XI.

Table XI: Purification of RNA-dependent DNA Polymerase from a Viral Core Fraction of Human Leukemic Spleen

Fig. 13: Sodium dodecyl sulfate-polyacrylamide gel agarose enzymes. Fractions 48(A) and 60(B) of Fig. 12C were subjected to electrophoresis on sodium dodecyl sulfate polyacrylamide gels. After staining with Coomassie blue, the gels were scanned in a Gilford Model 2400 spectrophotometer. The stained gel of each fraction is also shown.

Table XII: Deoxynucleoside Triphosphate and RNA Requirements of Leukemic RNA-dependent DNA Polymerase

The critical diagnostic criterion of a putative reverse transcriptase is the ability to transcribe heteropolymeric RNA into a DNA complement. To test this, a reaction was run with the purified enzyme employing isolated AMV-RNA as the template. The product synthesized was purified, alkali-treated and hybridized either to AMV or RL V-RNA. An analysis on cesium sulfate gradient showed clearly that the tritiated product was complementary only to AMV-RNA. Thus, the enzyme purified from the human leukemic particles satisfied this operational definition of a reverse transcriptase.
The results (69) we have just described represent the first instance of a human reverse transcriptase isolated to the purity required for complete characterization. In addition, the procedure provides enough protein to generate monospecific antisera for ultimate use as detecting devices.

XIV. Present Status and Future Prospects

The application of nucleic acid hybridization and the other techniques of molecular biology have permitted us to illuminate some of the issues of human cancer noted in the questions listed in the introductory paragraphs. The answers obtained may be summarized in the following statements:
(1) Human neoplasias do contain RNA molecules possessing detectable homologies to the RN A of tumor viruses known to cause similar cancers in animal systems .
(2) The RNA molecules indentified in the human tumors possess the size and physical association with reverse transcriptase that characterizes the RNA of the animal oncornaviruses .
(3) Further, the tumor-specific RNA is encapsulated in a particle possessing the size and density of the RNA tumor viruses .
(4) The RNA of the human tumor particles possess homology to the RNA of the viruses causing the corresponding diseases in animals .
(5) Viral-related sequences are unique to the DNA of tumor cells, are inserted postzygoticall y, and are therefore not resident in the germ line. The availability of the mesenchymal and mammary tumor animal model systems provided us with the viral agents that generated the radioactive DNA probes used to search for and find the homologous sequences in the particulate fractions of the corresponding human neoplasias. Our invention and perfection of the "simultaneous detection test" enabled us to characterize the size and particulate nature of the RNA detected in the human tumors. Beyond this, it evolved into an important technical advance since it permitted us to bypass the restriction of being forced to start with a known animal viral agent. Thus, we were able to extend our investigations beyond breast cancer and the mesenchymal tumors to other clinically important human neoplasias. The use of the simultaneous detection test established that a variety of human carcinomas contained particles having the diagnostic criteria associated with the oncornavirus-like particles. The neoplasias examined included a variety of cancers of brain (70), the gastrointestinal tract and lung (71 ), and skin (72).

Having found these particles in the human tumors, where does one go from there? An obvious need is to characterize them more fully biochemically and immunologically so that they can be compared with the similar agents found in the animal systems. However, here we are faced with a logistical problem that has thus far interposed serious obstacles. The amount of tumor material available and its particle content are such that it is difficult to isolate particles in sufficient quantity and purity to perform the desired biochemical examinations. Adequate , characterizations will not be readily feasible until these particles are obtainable in suitable yield from established cell lines. To date this desideratum has not been achieved. Nevertheless, it may still prove possible to exploit some of the implications derivable from certain of their features. In particular, at least a portion of I the RNA sequences found in the particles associated with each primary tumor site ,1 appears to be unique. As we have already noted, the particles found in human " breast cancer share no sequences in common with those found in the mesenchymal tumors. These differences appear to extend to the particles found in other sorts or primary tumors. Thus, we have been able to distinguish by cross hybridization the sequences found in particles from stomach cancers from those found in the colon. Similarly, the sequences found in lung cancer particles were easily differentiated from those found in other tumor sites tested. It would appear from our survey that the sequences found in these particles are histogenic ally specific.
These results suggest the possibility of developing a novel pathway for specific tumor detection. From the outset, it was evident that the nucleic acid hybridization technology we had developed and used to provide fundamental information of the molecular basis of the cancer cell was not likely to be useful in a clinical setting. In the first place, it has thus far been successfully applied only to tumor cells, and there is little likelihood that there would be enough circulating tumor-specific nucleic acids to be useful for diagnostic purposes. Further, the hybridization procedure is too sophisticated, too laborious, and too expensive to be introduced into the clinical pathology laboratory.
It seems clear that one must find a way of translating the sequence differences of the tumor particles into a parameter that would be more amenable to detection by the devices more commonly used in the clinical laboratory. One approach would depend upon the plausible expectation that the sequence differences observed in the different particles would be reflected in proteins that might be distinguishable antigenically. Were this realized, one could immediately hope to use the very sensitive and less restrictive methods of immunology. These are not only very sensitive, but they are in routine use in clinical laboratories.

To exploit this approach and obviate the logistic and other difficulties attending attempts to study the whole particles, one might focus rather on isolating and characterizing individual protein components. Of these, one of the most amenable is the reverse transcriptase since it can be followed during fractionation by means of its enzyme activity. As we have shown here, this approach has led to the successful isolation and purification of the reverse transcriptase of the particles found in leukemic spleens. These same methodologies can be and have been applied to other human neoplastic tissue. With these specific proteins available, monospecific antisera can be generated that could hopefully be used for detection in the body fluids of tumor-bearing individuals. The potential value of providing a use able specific assay for the presence of tumor cells has been even further enhanced by the recent advances in adjuvant chemotherapy. Efforts along these lines could convert what has thus far been an exercise in molecular biology into a potentially powerful clinical tool.

Acknowledgments

Supported by grant CA-02332 from the U.S. Public Health Service and by
Virus Cancer Program contract NO1-6-1010 with the National Cancer Institute.

References:


1. Hall, B. D., and Spiegelman, S. (1961) Proc. Nat. Acad. Sci. USA 47, 137-146.

2. Axel, R., Schlom, J. and Spiegelman, S. (1972) Nature 235,32-36.

3. Axel, R., Schlom, J. and Spiegelman, S. (1972) Proc. Nat. Acad. Sci. USA 69, 535-538.

4. Schlom, J., Spiegelman, S. and Moore, D. H. (1971) Nature 231,97-100.

5. Hehlmann, R., Kufe, D. and Spiegelman, S. (1972) Proc. Nat. Acad. Sci.USA 69, 435-439.

6. Kufe, D., Hehlmann, R., and Spiegelman, S. (1972) Science 175, 182-185.

7. Hehlmann, R., Kufe, D. and Spiegelman, S. (1972) Proc. Nat. Acad. Sci. USA 69, 1727-1731.

8. Gallo, R. C., Miller, N. R., Saxinger, W. C. and Gillespie, D. (1973) Proc. Nat.Acad. Sci. USA 70, 3219-3224.

9. Vaidya, A. B., Black, M. M., Dion, A. S. and Moore, D. H. (1974) Nature 249, 565-567.

10. Baltimore, D. (1970) Nature 226, 1209-1211.

11. Temin, H. M., and Mizutani, S. (1970) Nature 226, 1211-1213.

12. Spiegelman, S., Hall, B. D. and Storck, R. (1961) Proc. Nat. Acad. Sci. USA 47, 1135-1141.

13. Rokutanda, M., Rokutanda, H., Green, M., Fujinaga, K., Ray, R. K. and Gurgo, C. (1970) Nature 227, 1026-1029.

14. Spiegelman, S., Burny, A., Das, M. R., Keydar, J., Schlom, J., Travnicek, M.and Watson, K. (1970) Nature 227, 563-567.

15. Schlom, J., and Spiegelman, S. (1971) Science 174, 840-843.

16. Schlom, J., Spiegelman, S. and Moore, D. H. (1972) Science 175, 542-544.

17. Gulati, S. C., Axel, R. and Spiegelman, S. (1972) Proc. Nat. Acad. Sci. USA 69, 2020-2024.

18. Axel, R., Gulati, S. C. and Spiegelman, S. (1972) Proc. Nat. Acad. Sci. USA 69, 3133-3137.

19. Baxt, W., Hehlmann, R. and Spiegelman, S. (1972) Nature New BioI. 240, 72-75.

20. Spiegelman, S., Kufe, D., Hehlmann, R. and Peters, W. P. (1973) Cancer Research 33, 1515-1526.

21. Kufe, D., Margrath, I. T., Ziegler, J. L. and Spiegelman, S. (1973) Proc. Nat. Acad. Sci. USA 70, 737-741.

22. Todaro, G. J. and Huebner, R. J. (1972) Proc. Nat. Acad. Sci. USA 69, 1009-1015.

23. Ruprecht, R. M., Goodman, N. C. and Spiegelman, S. (1973) Proc. Nat. Acad. Sci. USA 70, 1437-1441.

24, Baxt, W. G. and Spiegelman, S. (1972) Proc. Nat. Acad. Sci. USA 69, 3737-3741.

25. Kufe, D. W., Peters, W. P. and Spiegelman, S. (1973) Proc. Nat. Acad. Sci. USA 70, 3810-3814.

26. Kufe, D., Hehlmann, R. and Spiegelman, S. (1973) Proc. Nat. Acad. Sci. USA 70, 5-9.

27. Henle, G., Henle, W., Clifford, P., Diehl, V., Kafuko, G., Kirya, B., Klein, G.Morrow, R., Munube, G., Pike, P., Tukel, P. and Ziegler, J. (1969) J. Nat. Cancer Inst. 43, 1147-1157.

28. zur Hausen, H., Schulte-Holthausen, H., Klein, G., Henle, W., Henle, G., Clifford, P. and Santesson, L. (1970) Nature 228, 1056-1058.

29. Henle, G., Henle, W. and Diehl, v. (1967) Proc. Nat. Acad. Sci. USA 59) 94-101.

30. Niederman, J. C., Evans, A. S., Subrahmanyan, L. and McCollum, R. W. (1970) New Eng. J. Med. 282, 361-365.

31. Geh, K. and Swisher, S. N. (1965) Arch. Int. Med. 115, 475-478.

32. Geh, K., Swisher, S. N. and Herman, E. C., Jr. (1967) Arch. Int. Med. 120) 214-219.

33. Baxt, W., Yates, J. W., Wallace, H. J., Jr., Holland, J. F. and Spiegelman, s. ( 1973) Proc. N at. Acad. Sci. USA 70, 2629-2632.

34. Goodman, N. C., Ruprecht, R. M., Sweet, R. W., Massey, R., Deinhardt, F. and Spiegelman, S. (1973) Int. Jour. Cancer 12, 752-760.

35. Sweet, R. W., Goodman, N. C., Cho, J.-R., Ruprecht, R. M., Redfield, R., R. and Spiegelman, S. (1974) Proc. Nat. Acad. Sci. USA 71, 1705-1709.

36. Chattopadhyay, S. K., Rowe, W. P., Teich, N. M. and Lowy, D. R. (1975) Proc. Nat. Acad. Sci. USA 72, 906-910.

37. Jaenisch, R., Fan, H. and Croker, B. (1975) Proc. Nat. Acad. Sci. USA 72, 4008-4012.

38. Frei, E. and Freireich, E. J. (1965) Adv. in Chemotherapy 2, 269-298.

39. Simone, J., Aur, R. J." Hustu, H. 0. et al. (1972) Cancer 30, 1488-1494.

40. Reisman, L. E., Zuelzer, W. W. and Thompson, R. I. (1964) Cancer Res. 1448-1460.

41. Sandberg, A. A., Takaaki, J., Kikuchi, Y. et al. (1964) Ann. N. Y. Acad. Sci. 113, 663-716.

42. Whang-Peng, J., Freireich, E. J., Oppenheim, J. J. et al. (1969) J. Nat. Cancer Inst. 42, 881-897.

43. Trujillo, J. M., Cork, A., Hart, J. S. et al. (1974) Cancer 33, 824-833.

44. Duttera, M. J., Whang-Peng, J., Bull, J. M. C. et al. (1972) Lancet 1,715-717.

45. Craddock, C. G. and Crandall, B. F. (1973) Blood 42,1013.

46. Greenberg, P. L., Nichols, W. C. and Schrier, S. L. (1971) New Eng. J. Med. 284, 1225-1232.

47. Harris, J. and Freireich, E. J. (1970) Blood 35,61-63.

48. Moore, M. A. S., Williams, N. and Metcalf, D. (1973) J. Nat. Cancer Inst. 50) 603-623.

49. Bull, J. M., Duttera, M. J., Stashick, E. D. et al. (1973) Blood 42,679-676.

50. Gutterman, J. U., Mavligit, G., Burgess, M. A. et al. (1974) J. Nat. Cancer Inst. 53,389-392.

51. Halterman, R. H., Leventhal, B. and Mann, D. L. (1972) New Eng. J. Med. 287, 1272-1274.

52. Killmann, S. A. (1968) Ser Haemat 1,103-128.

53. Mak, T. W., Aye, M. T., Messner, H. et al. (1974) Brit. J. Cancer 29,433-437.

54. Mak, T. W., Manaster, J., Howotson, A. F. et al. (1974) Proc. Nat. Acad. Sci. USA 71, 4336-4340.

55. Fischinger, P. J., Nomura, S., Peebles, P. T. et al. (1972) Science 176, 1033-1035.

56. McGrath, C. M., Grant, P. M., Soule, H. D., Glancy, T. and Rich, M. A. (1974) Nature 252, 247-250.

57. Kotler, M., Weinberg, E., Haspel, 0., Olshevshy, U. and Becker, Y. (1973) Nature New BioI. 244, 197.

58. Gallagher, R. E. and Gallo, R. C. (1975) Science 187, 350-353.

59. Teich, N. M., Weiss, R. A., Salahuddin, S. Z., Gallagher, R. E., Gillespie, D. H. and Gallo, R. C. (1975) Nature 256, 551-555.

60. Gallo, R. C., Gallagher, R. E., Miller, N. R., Mondal, H., Saxinger, W. C., Mayer, R. J., Smith, R. G. and Gillespie, D. H. (1974) Cold Spring Harbor Symposium on Quantitative Biology XXXIX, 933-961.

61. Benveniste, R. E., Lieber, M. M., Livingston, D. M., Sherr, C. J. and Todaro, G. J. (1974) Nature 248, 17-20.

62. Theilen, G. H., Gould, D., Fowler, M. and Dungworth, D. L. (1971) J. Nat. Cancer Inst. 47, 881-889.

63. Chan, E., Peters, W. P., Sweet, R. W., Ohno, T., Kufe, D. W., Spiegelman, S., Gallo, R. C. and Gallagher, R. E. (1976) Nature 260, 266-268.

64. Okabe, H., Gilden, R. V., Hatanaka, M., Stephenson, J. R., Gallagher, R. E., Gallo, R. C., Tronick, S. R. and Aaronson, S. A. (1976) Nature 260, 264-266.

65. Sarngadharan, M. G., Sarin, P. S., Reitz, M. S. and Gallo, R. C. (1972) Nature New BioI. 240, 67-72.

66. Mondal, H., Gallagher, R. E. and Gallo, R. C. (1975) Proc. Nat. Acad. Sci. USA 72, 1194-1198.

67. Feldman, S. P., Schlom, J. and Spiegelman, S. (1973) Proc. Nat. Acad. Sci. USA 70, 1976-1980.

68. Michalides, R., Spiegelman, S. and Schlom, J. (1975) Cancer Res. 35, 1003-1008.

69. Witkin, S., Ohno, T. and Spiegelman, S. (1975) Proc. Nat. Acad. Sci. USA 72, 4133-4136.

70. Cuatico, W., Cho, J.-R. and Spiegelman, S. (1973) Proc. Nat. Acad. Sci. USA 70, 2789-2793.

71. Cuatico, W., Cho, J.-R. and Spiegelman, S. (1974) Proc. Nat. Acad. Sci. USA 71, 3304-3308.

72. Balda, B.-R., Hehlmann, R., Cho, J.-R. and Spiegelman, S. (1975) Proc. Nat. Acad. Sci. USA 72,3697-3700.