NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Coffin JM, Hughes SH, Varmus HE, editors. Retroviruses. Cold Spring Harbor (NY): Cold Spring Harbor Laboratory Press; 1997.

Cover of Retroviruses


Show details


Retroviruses were first discovered in association with cancer around the turn of this century. An avian erythroblastosis virus (AEV) was isolated from a spontaneous erythroleukemia in a chicken in 1908 (Ellerman and Bang 1908), and soon after, Peyton Rous isolated Rous sarcoma virus from a chicken fibrosarcoma (Rous 1911). Strains descended from Rous' original isolate are still studied today. These experiments were eventually followed by many others in which tumor-inducing retroviruses were isolated from rodents, cats, cows, primates, and other animals (for review, see Gross 1970; Weiss et al. 1982). Viruses capable of inducing sarcomas, a wide range of hematopoietic tumors, and, more rarely, carcinomas were discovered.

Studies of these viruses laid much of the foundation for our current understanding of the molecular mechanisms involved in tumor induction (for review, see Bishop 1991). The discovery of viral oncogenes (v-onc genes) and the realization that they were derived from cellular genes called proto-oncogenes (c-onc genes; see Chapter 4 provided strong clues about the role of c-onc genes in other types of tumors. This link was solidified when it became clear that oncogenic retroviruses lacking v-onc genes integrate near some of the same c-onc genes and “activate” their expression. Increased appreciation of the common molecular basis of viral and nonviral oncogenesis came when it was found that mutation of c-onc genes has a key role in tumors induced by other nonviral agents. These connections are illustrated by the ways in which the c-myc oncogene can be activated (Fig. 3). Several avian retroviruses, including MC29, OK-10, and MH2, have captured c-myc (see Table 2), and the gene is activated via insertional mutagenesis in lymphomas induced by ALV, Mo-MLV, and several other viruses that do not carry v-onc genes (see Table 3). c-myc is also activated by chromosomal translocation and mutation in Burkitt's lymphoma, a nonretrovirus-associated tumor afflicting humans (for review, see Rabbitts and Boehm 1991; Klein 1994).

Figure 3. Proto-oncogenes can be activated in multiple ways.

Figure 3

Proto-oncogenes can be activated in multiple ways. (Top) The c-myc locus and its protein product encoded by exons 2 and 3; (A) activation via oncogene capture as illustrated by the genome and fusion protein product of MC29 virus, one (more...)

Table 2. Retroviruses Containing Oncogenes.

Table 2

Retroviruses Containing Oncogenes.

Table 3. Loci Targeted by Provirus Insertional Mutagenesis.

Table 3

Loci Targeted by Provirus Insertional Mutagenesis.

Types of Oncogenic Retroviruses

Retroviruses that induce malignancies belong to one of five different genera: avian sarcoma/leukosis viruses (ASLV); mammalian C-type viruses; B-type viruses; D-type viruses; and the HTLV-BLV group (for review, see Coffin 1996). Most oncogenic retroviruses are simple C-type viruses that contain the gag, pol, and env genes (see Chapter 2, and many share features with endogenous viruses found in their hosts (Table 1). These viruses are classified according to their host range, a feature that reflects the structure of the Env protein (see Chapter 3. ALVs, which belong to subgroups A through D, are one large group of related oncogenic viruses (for review, see Payne 1992). Some of these viruses were first isolated from stocks of RSV and are called Rous-associated viruses (RAVs) to reflect their origin. Oncogenic ALVs are related to the subgroup-E chicken endogenous viruses. They usually induce B-lymphocyte tumors called bursal lymphomas because the tumors arise in the bursa of Fabricius. Members of the reticuloendotheliosis virus (REV) group can also induce tumors in birds (for review, see Witter 1984; Payne 1992). These viruses are related to mammalian C-type viruses and belong to a single interference group.

Table 1. Representative Oncogenic Retroviruses That Lack Oncogenes.

Table 1

Representative Oncogenic Retroviruses That Lack Oncogenes.

Endogenous and exogenous MLVs and feline leukemia viruses (FeLVs) induce tumors of T lymphocytes and some other hematopoietic cells (Table 1). Oncogenic FeLVs belong to either subgroup A or B; subgroup-A viruses are exogenous and subgroup-B viruses are related to endogenous feline viruses (for review, see Hardy 1993). MLVs are classified according to the species which express the viral receptor recognized by the Env protein (see Chapter 3 for review, see Weiss 1993). Ecotropic viruses infect mouse cells only; polytropic and amphotropic viruses infect both murine and nonmurine cells using receptors distinct from the ecotropic receptor and from each other; xenotropic viruses use another receptor expressed only on nonmurine cells. All known exogenous viruses have either ecotropic or amphotropic host range. Endogenous viruses can belong to any of these groups except amphotropic. Oncogenic MLVs belong to all groups except xenotropic.

Many isolates of oncogenic ALV, MLV, and FeLV contain oncogenes (Table 2). These viruses cause a wide range of tumors but are most commonly associated with either sarcomas or tumors involving hematopoietic cells. Most of these viruses transform cells in vitro and they usually induce tumors after a much shorter latent period than viruses that lack oncogenes. They were generated by recombination between a replication-competent virus and a proto-oncogene sequence, and thus their retroviral “backbone” was originally contributed by a replicating C-type virus (see Chapter 4. Because most viruses that have captured v-onc genes are replication-defective, animals infected with them are also infected with a replication-competent helper virus. The helper virus provides the Gag and Env proteins and thus controls the host range.

HTLV and BLV are exogenous, complex retroviruses that comprise their own genus. These agents are associated with lymphoid malignancies in humans and cattle, respectively (for review, see Kettmann et al. 1994; Cann and Chen 1996). The murine mammary tumor virus (MMTV), a B-type retrovirus, induces mammary tumors and, more rarely, T-cell lymphomas (for review, see Salmons and Gunzburg 1987). Jaagsiektse, a lung disorder of sheep resembling adenocarcinoma, is associated with a D-type retrovirus (for review, see Sharp 1987; DeMartini et al. 1988). Retrovirus-associated tumors are also commonly found in many types of bony fish, but the role of the viruses in many of these tumors has not been established firmly (for review, see Poulet et al. 1994). The unique organization of at least one of the fish retroviruses suggests that these agents may comprise a distinct genus (Holzschu et al. 1995).

Common Molecular Themes in Oncogenesis

The diversity evident among oncogenic retroviruses seems to suggest that many different molecular mechanisms are involved in tumorigenesis. Even though retroviruses that carry v-onc genes are closely related to the replicating viruses from which they were derived, the ways in which these two types of viruses induce tumors may seem quite distinct. In one instance, a viral gene, the v-onc gene, is the dominant force and tumors arise after a short latent period. For viruses lacking v-onc genes, tumors arise after an extended latent period, and proviral integration near a proto-oncogene, a phenomenon termed proviral insertional mutagenesis, plays a central part in tumor induction. Despite these differences, important central principles unify the oncogenic processes and extend to tumor systems that lack a retroviral etiology. A large subset of the genes affected by proviral insertional mutagenesis can be found as oncogenes in one or more retroviruses (see Tables 2 and 3), and many of the proteins encoded by v-onc genes and activated proto-oncogenes belong to the same array of functional categories. Altered expression of proto-oncogenes, reflecting the loss of tissue and developmental and cell-cycle-specific regulators, is a second common thread connecting the oncogenic mechanisms used by all retroviruses. In addition, both v-onc gene capture and proviral insertional mutagenesis often result in production of an oncogene product that has altered or lost sequences which are important for regulating its function.

The common themes that unite the mechanisms involved in retroviral oncogenesis extend to tumors that lack a retroviral etiology (for review, see Bishop 1991). Many genes, including ras, abl, erbB, and myc, all first identified as v-onc genes, are now known to be activated in certain types of spontaneous tumors (for review, see Barbacid 1987; Bos 1989; Rabbitts and Boehm 1991; Sawyers 1992; Van Etten 1993; Carter and Kung 1994; Hynes and Stern 1994; Klein 1994). Indeed, some of the mutations found in c-ras genes in human bladder, lung, and breast carcinomas affect the same codons that are altered in the v-ras genes found in several murine sarcoma viruses (MSVs) (Barbacid 1987; Bos 1989). Changes affecting elements controlling gene expression resulting from gene amplification, translocation, or other mutations play a prominent part in most tumor systems. These unifying features extend the significance of studies conducted with retroviral tumor models to broader issues of carcinogenesis.

Tumor Induction by Simple C-type Retroviruses That Lack Oncogenes

Most tumors induced by retroviruses that lack oncogenes involve hematopoietic cells; a few of these viruses induce carcinomas (see Table 1). These malignancies contribute significantly to the incidence of tumors in cats, chickens, and some mouse strains. Endogenous viruses have important roles in some instances; in other cases, exogenous viruses that naturally infect the host population are involved. The oncogenic agents are replication-competent, and extensive viral replication occurs during the long latent period that characterizes tumor development. This long latent period reflects the inability of these viruses to transform cells directly and the need for multiple cooperative changes in growth control mechanisms to induce the tumor. Consistent with these features, the tumors are clonal.

Early Events in the Oncogenic Process

For most viruses in this group, extensive replication occurs during the latent period or preleukemic phase of the disease. Many of the infected cells proliferate, and changes in the cell composition and morphology of infected tissue are often evident (for review, see Weiss et al. 1982; Fan et al. 1991). For example, follicles of ALV-infected cells are prominent in the bursa long before malignant disease develops (Fig. 4) (Cooper et al. 1968; Neiman et al. 1980; Baba and Humphries 1985; Thompson et al. 1987), and proliferative changes have been documented in the thymus of mice developing lymphoma (O'Donnell et al. 1984, 1985; Davis et al. 1986). These preleukemic changes do not occur exclusively in the tissue where the tumor eventually develops. Proliferative changes and preleukemic cells can be detected in the bone marrow and spleen of mice that go on to develop thymic lymphoma before they are evident in the thymus (Haran-Ghera 1980; Asjo et al. 1985; Storch et al. 1985). Some of the proliferating cells are not infected and may be stimulated via cytokines or other soluble mediators that are released as a consequence of the infection (Brightman et al. 1990). At least some of the changes that characterize the preleukemic phase are critical to the oncogenic process because viruses that grow well but fail to induce these changes are poorly oncogenic (Davis et al. 1985, 1987; Li and Fan 1991b).

Figure 4. Stages in the development of ALV-induced bursal lymphomas.

Figure 4

Stages in the development of ALV-induced bursal lymphomas. (A) Histologic section of an ALV-infected bursa showing a reddish-stained abnormal follicle in the center of the photograph. Adjacent normal follicles have a bluish appearance. ( (more...)

Proviral Insertional Mutagenesis

Pioneering studies of ALV-induced bursal lymphomas revealed that most of the tumors contain a provirus integrated in the vicinity of the c-myc proto-oncogene and express high levels of c-myc (Hayward et al. 1981; Payne et al. 1981). This phenomenon, termed proviral insertional mutagenesis, occurs in many types of retrovirus-induced tumors and can involve a wide variety of genes that modulate growth and differentiation (Table 3) (Peters 1990; van Lohuizen and Berns 1990; Kung et al. 1991; Tsichlis and Lazo 1991). As noted earlier, the c-myc gene had already been identified as the cellular homolog of an oncogene found in several avian transforming retroviruses. Other targets of insertional mutagenesis have been captured by retroviruses or are activated in tumors induced by other means. In addition, a number of genes now known to have key roles in growth control were first discovered as common targets of proviral insertion. Most of these genes have been identified in murine, feline, or avian tumors induced by exogenous viruses because large numbers of these animals can be screened. This feature may bias the types of genes that have been identified.

Like all retroviral integrations, insertions near oncogenes occur by chance as a consequence of the normal viral life cycle (Hughes et al. 1978; Steffen and Weinberg 1978; Withers-Ward et al. 1994; see Chapter 5. However, when a provirus integrates near a gene controlling growth and alters its expression, the host cell may have a selective growth advantage. Expansion of this cell can eventually lead to the formation of a clonal tumor in which each cell contains a provirus integrated at the same site. Thus, even though such an event is very rare on a cellular basis, it is certain to occur at least once in a tissue if enough cells become infected. The time required for this event as well as the requirement for additional genetic changes influences the latent period required for tumor induction by this mechanism.

Identification of Genes Activated by Proviral Insertion

Any system in which retroviral infection induces multiple, independent clonal tumors can be analyzed for the presence of common integration sites. The provirus itself serves as an important tag to identify common sites of viral insertion. DNAs from a panel of independently derived tumors induced by the same virus can be examined by Southern blotting to identify a tumor with only a few newly acquired proviruses. Next, these proviruses and a portion of the cellular sequence flanking them are cloned and probes are prepared from the cellular sequences. Each of the flanking sequence probes should identify two forms of the locus in the tumor from which it was isolated—the copy altered by the proviral insertion and the normal copy. Any probe that detects proviral insertions in other members of the tumor panel identifies a site of common proviral insertion. The frequency with which particular common integration sites are found in different types of tumors varies considerably. Sometimes, a particular locus is targeted about 90% of the time; in other instances, a site may be occupied in less than 5% of tumors (Table 3). Even this low frequency is significant given that repeated integrations in an unselected region would be expected to occur at a frequency on the order of 1 in 106.

Once a common integration site is identified, a search for coding sequences in the vicinity of the locus must be initiated. Approaches commonly used include screening the newly identified locus with probes specific for known oncogenes, screening Northern blots for altered forms of viral RNAs that contain a cellular sequence, and using exon amplification methods designed to recover exons from uncharacterized stretches of cellular sequence (Buckler et al. 1991; Church et al. 1994). Genetic mapping can pinpoint the chromosomal location of the insertion, and a search of the genome database can rapidly rule out most known genes, which will be located in other regions of the genome. Occasionally, this approach reveals a candidate gene that maps to the same region as the proviral insertion. Once a gene is identified, experiments should demonstrate that the product of the gene has a role in the tumorigenic process. At a minimum, changes in expression of the gene are expected to correlate with tumor formation. Other experiments should explore the ability of the gene product to induce transformation in vitro and in vivo when it is overexpressed. Transgenic mice in which overexpression of the gene is targeted to particular tissues can also be studied. Often, if a gene identified corresponds to a known oncogene, a gene encoding a growth factor, a transcription factor, or any protein with a function associated with regulating cell growth, the assumption is made that the product facilitates tumor development.

Mechanism of Altered Gene Expression

Insertion of a provirus in the vicinity of a proto-oncogene disrupts the normal organization of the locus and also introduces strong promoter and enhancer regions into the locus. In most instances, these changes are sufficient to alter gene expression; deletions and rearrangements of cellular sequences in the vicinity of the integration site are rare (Robinson and Gagnon 1986; Goodenow and Hayward 1987; Nottenburg et al. 1987; Lazo and Tsichlis 1988). In addition, the relationship of the viral and cellular sequences is usually quite flexible. Insertions can occur over a broad range upstream or downstream from many targeted loci, and the proviruses can be in the same transcriptional orientation as the gene or in the opposite orientation (Table 3).

When a provirus integrates upstream of a proto-oncogene in the same transcriptional orientation, the promoter and enhancer elements in one of the proviral long terminal repeats (LTRs) can direct increased or developmentally inappropriate levels of proto-oncogene expression. This situation is known as promoter insertion (Fig. 5A). More than 80% of ALV-induced bursal lymphomas contain proviruses that alter c-myc expression by this mechanism (Fung et al. 1981, 1982a; Neel et al. 1981; Payne et al. 1981; Westaway et al. 1984; Robinson and Gagnon 1986; Goodenow and Hayward 1987; Swift et al. 1987). In most of these tumors, the promoter in the 3′LTR is used to generate a hybrid ALV-c-myc transcript, and transcription initiating in the 5′LTR is reduced (Hayward et al. 1981; Goodenow and Hayward 1987; Swift et al. 1987). The promoter in the 3′LTR is normally silent, but most of the ALV proviruses integrated near c-myc contain deletions which may favor use of the downstream promoter (Cullen et al. 1984; Herman and Coffin 1986; Boerkoel and Kung 1992). Although promoter insertion is the most obvious mechanism of insertional activation, it has been rarely seen in other types of tumors (Table 3).

Figure 5. Mechanisms of proto-oncogene activation by proviral insertion.

Figure 5

Mechanisms of proto-oncogene activation by proviral insertion. Four exons of a model proto-oncogene are shown at the top. In this and subsequent diagrams, the spliced mRNA produced is shown below the proto-oncogene or provirus–proto-oncogene (more...)

Many proviruses are inserted upstream of target genes in the opposite transcriptional orientation or downstream from target genes in either orientation (Fig. 5B). These proviruses increase gene expression by placing the gene under the influence of strong enhancer sequences within the retroviral U3 region. This type of activation occurs in a wide range of tumor systems (Table 3). Occasionally, the proviruses are separated from the gene by as much as 300 kb (Lazo et al. 1990; Bartholomew and Ihle 1991). In these instances, the viral enhancer is thought to interact with the target gene following looping out of the region between the LTR and target gene as postulated for other enhancers that affect expression over long distances (Dillon and Grosveld 1993; Kennison 1993; see Chapter 6. In some of these instances, there may be additional, as yet unidentified, genes within these regions. More than one gene has been found in the vicinity of a common integration site in several instances. For example, Mlvi2, a locus targeted in Mo-MLV-induced thymomas (Table 3), maps close to the growth hormone receptor gene and the prolactin receptor gene. However, integrations into the Mlvi2 locus do not affect levels of expression of either of these genes (Barker et al. 1992). The Evi2 locus maps to a region containing four genes, but only one of these appears to have a clear role in oncogenesis (Buchberg et al. 1990; Viskochil et al. 1990, 1991; Cawthon et al. 1991; Largaespada et al. 1995). Thus, the presence of a provirus in the vicinity of a gene may not alter its expression and does not always indicate that the gene has a role in tumorigenesis.

Proviral insertional mutagenesis can also lead to expression of altered protein products. A provirus inserted within a gene can generate transcripts that initiate at the normal retroviral promoter in the 5′LTR and read through into cellular sequences. Because the provirus has inserted within the gene, the resulting transcript contains viral leader sequences fused to sequences from those exons that are downstream from the proviral insertion (Fig. 5C). This complex process is commonly observed in ALV-induced erythroblastosis in which a provirus integrates between two exons of c-erbB, a gene that encodes a receptor which binds epidermal growth factor (EGF) and tumor growth factor-α (TGF-α) (Downward et al. 1984; Ullrich et al. 1984; Lax et al. 1988). As a consequence, sequences that encode the ligand-binding domain of the molecule are separated from those that mediate protein tyrosine kinase (PTK) activity and downstream signaling (Fung et al. 1983b; Nilsen et al. 1985; Raines et al. 1985).

Another way in which gene expression can be affected by proviral insertion involves loss of other types of control sequences (Fig. 5D). Proviral insertions downstream from the coding sequence of pim1 that are in the same orientation insert the viral poly(A) site before the normal cellular poly(A) site in the pim1 gene. This configuration causes the deletion of 3′-noncoding sequences that render the normal pim1 mRNA unstable, leading to accumulation of abnormal mRNA levels (Selten et al. 1985, 1986). Proviral insertions can also facilitate the use of cryptic promoters (Dickson et al. 1990) or destroy important regulatory elements such as silencer regions or pause sites (Corcoran et al. 1984; Selten et al. 1984). In rare instances, proviral insertion can inactivate gene expression (see below Common Biological Themes in Oncogenesis). Theoretically, integrations might also influence the methylation of sequences in the vicinity of the insertion, altering gene expression (Jahner and Jaenisch 1985). The possible contribution of this mechanism to tumorigenesis has not been investigated in detail.

When the disrupted c-erbB locus is expressed, several forms of mRNA are produced. One of these contains a small amount of gag sequence spliced directly to c-erbB sequences; in another, a small portion of env sequence encoding the leader is included between the gag and c-erbB sequences (Nilsen et al. 1985; Goodwin et al. 1986). Fusion of the viral Env signal sequence to c-erbB sequences affects the processing and membrane localization of the truncated c-ErbB protein (for review, see Maihle and Kung 1988). The kinase is constitutively active because most of the ligand-binding domain is lost (Lax et al. 1985). Indeed, the amino-terminal sequences missing in the truncated c-ErbB protein are similar to those missing in the v-ErbB proteins produced by AEV isolates. Activation of other proto-oncogenes via the production of similar readthrough transcripts has been documented in some ALV-induced bursal lymphomas that have integrations in the c-myb locus (Pizer and Humphries 1989) and in MLV-induced myeloid tumors (Shen-Ong et al. 1986; Weinstein et al. 1986; Rosson et al. 1987).

Role of Viral Genes in Oncogenesis

Infection of the host and seeding of the target organ are the first steps in tumor induction. Large numbers of virus-infected cells are found in the host and many of these cells proliferate. As discussed earlier, the changes in cell growth and tissue architecture that accompany this proliferative phase have a key role in tumor induction. Although viral replication alone is not sufficient for tumor induction, the viral elements that control replication have a strong influence on these steps. The ability to infect a large number of cells within the target tissue multiple times enhances the probability that a proviral insertion will occur in the vicinity of a gene involved in growth control. Clearly, all of the viral sequences required for replication are essential for this process. However, the env gene and the U3 sequences within the LTR have central roles in determining the incidence and types of tumors that arise. MMTVs have an additional gene, sag, that encodes a superantigen and is critical for establishing infection and disease induction.

Env Proteins and Direct Stimulation of Growth

In addition to their classical role in mediating viral entry, some SU proteins can bind to growth factor receptors on the cell surface and trigger a growth-stimulating signal by mimicking normal ligand-receptor interaction. Binding probably occurs at a site distinct from that used by the natural ligand and is not involved in viral entry. However, this interaction can expand the available target pool and promote viral replication in three ways. First, because simple retroviruses establish infection only in cycling cells, interaction of SU protein with a surface receptor that stimulates growth can render a cell susceptible to infection. Second, stimulation of growth can expand the number of available target cells. Third, expanding the number of infected cells increases the amount of viral replication. This combination of effects can have a large impact on tumor development.

The clearest example of Env-mediated growth factor receptor stimulation in tumor induction occurs in the rapid onset erythroleukemia induced by the polycythemic strain of Friend virus (FV-P). The virus induces massive erythroid proliferation which causes splenomegaly (spleen enlargement) (Fig. 6A). FV-P resembles v-onc gene-containing viral stocks in that it contains two components—a replication-competent virus called Friend MLV (Fr-MLV) and a replication-defective component called spleen focus-forming virus (SFFV-P) (Troxler et al. 1977). This latter virus induces the erythroid proliferation (Wolff and Ruscetti 1985; 1988; Wolff et al. 1986). However, unlike most replication-defective viruses that induce rapid onset disease, SFFV-P does not contain an oncogene; the active gene product is a deleted recombinant Env protein called gp55 (Fig. 6B). This molecule contains amino-terminal sequences related to polytropic MLVs and carboxy-terminal sequences related to ecotropic MLVs (Amanuma et al. 1983; Wolff et al. 1983). A large central portion of the ecotropic sequences, including the portion required for gp70-p15E cleavage, is missing because of an in-frame deletion. In addition, a single-base insertion in the p15E-coding sequence leads to production of a protein that is truncated at the carboxyl terminus. Similar changes are found in many independent isolates of MLV that induce erythroleukemia, indicating that these features are important for the oncogenic potential of the virus (for review, see Kabat 1989).

Figure 6. Induction of erythroleukemia by FV.

Figure 6

Induction of erythroleukemia by FV. (A, Right) A normal spleen (top) and a spleen from an FV-P-infected mouse (bottom); (left) a normal mouse (top) and an infected mouse (bottom). The arrows denote the spleens. (more...)

Two types of erythroid precursor cells, late BFU-E and CFU-E, respond to gp55 (Kost et al. 1979, 1981; Hankins and Troxler 1980; Peschle et al. 1980). Under normal conditions, these cells are stimulated to grow when the hormone erythropoietin (Epo) binds to the Epo receptor on their surface. When the cells are infected with FV-P, the gp55 protein binds to the Epo receptor and mimics a normal receptor-ligand interaction (Fig. 6C) (Hoatlin et al. 1990; Li et al. 1990; Ruscetti et al. 1990). Binding occurs in the extracellular portion of the Epo receptor at a site distinct from the Epo-binding site and requires surface expression of both the receptor and the gp55 molecule (Casadevall et al. 1991; Ferro et al. 1993; Li et al. 1995). The importance of this interaction in mediating the polycythemia that characterizes early FV-induced disease is highlighted by the fact that the closely related anemia-inducing strain of FV (FV-A) does not stimulate Epo-independent growth of erythroid precursors. This property maps to a small region in the 3′end of the env gene that encodes the transmembrane portion of the protein (Ruscetti and Wolff 1985; Wolff et al. 1985; Chung et al. 1987; Amanuma et al. 1989).

Interactions with other growth factor receptors similar to those observed with FV-P may be important for the induction of other types of tumors. Because their growth is tightly regulated by interleukins, hematopoietic cells may be particularly susceptible to this type of stimulation. Indeed, interleukin-2 (IL-2) stimulates the growth of many T lymphocytes, and some MCF viruses have been shown to stimulate the IL-2 receptor in vitro in a fashion analogous to the interaction between gp55 and the Epo receptor (Li et al. 1990; Li and Baltimore 1991). However, not all factor independence reflects this type of interaction. Infection of an IL-2-dependent Mo-MLV-induced rat thymoma line with MCF or xenotropic virus allowed selection for IL-2-independent cells (Tsichlis and Bear 1991; Flubacher et al. 1994). In this case, the IL-2-independent cells have proviral insertions that activate the IL-9 receptor gene, creating an IL-9 autocrine loop (Flubacher et al. 1994).

Recombinant Env Genes and Tumor Induction

The proteins encoded by the env gene have an essential role in viral infection by mediating interactions between the virus and its cellular receptor (see Chapter 3. Once a cell is infected and produces env-encoded proteins, the viral receptor becomes blocked by the SU protein. This phenomenon, known as superinfection resistance, limits the number of proviruses that can integrate in a cell. Superinfection resistance also limits the number of rounds of replication that can occur and decreases the chances that recombinations will occur during mixed infections with closely related viruses. The importance of these effects to the tumorigenic process is highlighted by the changes that occur in env sequences during the development of virus-induced thymic lymphomas in mice and cats.

The first clues concerning the important part played by env-encoded sequences came from analyses of AKR mice. About 90% of these animals spontaneously develop thymic lymphoma during the first 7–12 months of life (for review, see Gross 1970). Genetic studies revealed that inheritance and expression of an ecotropic endogenous provirus derived from the Akv1 locus are essential for tumor development (Lilly et al. 1975). However, the relationship of the Akv1 virus to tumor development was initially difficult to understand because the virus is highly expressed in AKR mice from birth but does not induce tumors in susceptible, low-leukemia strains of mice (Cloyd et al. 1980; for review, see Gross 1970).

The role of Akv1 was clarified when it was discovered that the tumors themselves and the preleukemic thymic tissue expressed new viruses that had a polytropic host range (Hartley et al. 1977). The polytropic viruses are called mink cell focusing-inducing viruses because they induce cytopathic effects in mink lung cell cultures (Hartley et al. 1977). They arise via recombination between Akv and other endogenous proviruses carried by AKR mice (Elder et al. 1977; Chattopadhyay et al. 1982; Khan et al. 1982; Thomas and Coffin 1982; Herr and Gilbert 1983; Stoye et al. 1991; for a discussion of the mechanisms involved in recombination, see Chapters 4 and 8). The acquisition of sequences encoding the amino-terminal portion of the SU protein is responsible for the polytropic host range (Fig. 7) (Battini et al. 1992; Ott and Rein 1992; see Chapter 3. The endogenous polytropic virus that donates these sequences has not been conclusively identified.

Figure 7. Proviruses present in AKR mice and AKR thymomas.

Figure 7

Proviruses present in AKR mice and AKR thymomas. (Top) (Akv) Endogenous ecotropic provirus; (Bxv1) endogenous xenotropic provirus; (Pmv) representative of one or more unidentified endogenous polytropic proviruses that can recombine (more...)

The broadened host range of the recombinants allows them to infect cells that already express ecotropic viruses, enhancing the probability that growth-stimulating proviral insertions will occur. Perhaps more importantly, the presence of mixed viral populations in individual cells increases the chances that additional recombination events will occur. Indeed, recombination events affecting sequences in the 3′portion of the env gene and the LTR are also important for tumor development (see below). MCF viruses can also indirectly stimulate splenic hyperplasia, perhaps by suppressing bone marrow hematopoiesis (Brightman et al. 1990; Li and Fan 1991a,b), and may stimulate T-cell proliferation by interaction with the IL-2 receptor (Li and Baltimore 1991). As a consequence, an increased number of target cells may be present in the spleen and the periphery, facilitating viral replication. The part played by MCF viruses in tumorigenesis is emphasized by the ability of some of them to accelerate lymphoma development when inoculated into newborn AKR mice and to induce lymphomas in low-incidence strains (see Cloyd et al. 1980; O'Donnell et al. 1981; Holland et al. 1985).

Generation of recombinants with extended host range is not unique to AKR mice. Similar events are involved in leukemogenesis in other mouse strains with a high incidence of lymphoma, including HRS, C58, and CWD (Green et al. 1980; Famulari 1983; Thomas et al. 1984, 1986). MCF strains are also generated following infection of mice with exogenous leukemogenic MLVs such as the Fr-MLV, Rauscher (Ra-MLV), and Mo-MLV strains (Fischinger et al. 1975; Troxler et al. 1978; Vogt 1979; Hoffman et al. 1981; Ishimoto et al. 1981; Ruscetti et al. 1981). In the latter instance, generation of MCFs has been linked to efficient induction of lymphoma (Brightman et al. 1991). Recombinants between exogenous and endogenous MMTV env genes can be detected in mammary tumors (Golovkina et al. 1994). Despite these strong correlations, rats do not contain endogenous viruses that could generate MCFs, yet they develop thymic lymphomas at a high frequency after infection with Mo-MLV (Tsichlis et al. 1983; Lemay and Jolicoeur 1984; Steffen 1984). Thus, recombinants are not required for leukemogenesis in all settings.

Recombinations affecting env sequences also occur in FeLV-induced thymic tumors. Most of these tumors arise from naturally occurring infections with uncloned mixtures of virus. The tumors often contain FeLV-A, an exogenous virus, and FeLV-B, an env gene recombinant generated between FeLV-A and endogenous sequences (Stewart et al. 1986; Overbaugh et al. 1988a; Neil et al. 1991; Sheets et al. 1992). Indeed, the tumors contain mixtures of many different viruses displaying a variety of changes in the env gene, including point mutations, insertions, deletions, and recombinations (Boomer et al. 1994; Rohn et al. 1994). These events are very reminiscent of those just described in mice. However, cats are not susceptible to FeLV-B infection in the absence of FeLV-A (Jarrett et al. 1978; Jarrett and Russell 1978), probably because SU protein encoded by endogenous proviruses blocks FeLV-B receptors (McDougall et al. 1994). This phenomenon has made it difficult to test the oncogenic potential of the recombinants in the absence of FeLV-A.

Control Elements in the LTR

The LTR controls expression of simple retroviruses and contains both the promoter and a complex set of enhancer binding motifs in the U3 region that mediate expression of sequences placed under their control (see Chapter 6. These elements influence the replication cycle of the virus. As noted above, infection of a large number of target cells and the accumulation of large numbers of proviruses in each cell enhance the chances that a provirus will integrate in the vicinity of a gene modulating growth. High levels of replication increase the chances that recombinants with enhanced oncogenic potential will be generated. Thus, the effects of sequences within the LTR on viral replication strongly influence the tumorigenic process. In addition, once appropriate proviral insertions have occurred, the targeted gene must be expressed at levels sufficient to affect cell growth. Sequences within the U3 region of the LTR, via interactions with different combinations of transcription factors, influence this process and control the types of tumor cells that arise.

The importance of the LTR in modulating oncogenesis was first suggested by experiments in which the LTR sequences from closely related viruses that differ in oncogenic potential were compared. Analyses of chimeric viruses revealed that these sequences are one of the major determinants that distinguish oncogenic and nononcogenic ALVs and MLVs (Robinson et al. 1982; DesGroseillers et al. 1983, 1984b; Lenz and Haseltine 1983; DesGroseillers and Jolicoeur 1984a; Lenz et al. 1984; Brown et al. 1988). Finer mapping revealed that the enhancer within the U3 region of the LTR, particularly the core elements, has a dominant role (Celander and Haseltine 1984; DesGroseillers and Jolicoeur 1984b; Thiesen et al. 1988; Boral et al. 1989; Yuen and Szurek 1989; Speck et al. 1990; Hallenberg et al. 1991; Tupper et al. 1992; see Chapter 6. LTR sequences also influence the types of tumors that develop (Chatis et al. 1983, 1984; Rosen et al. 1985; Ishimoto et al. 1987; Li et al. 1987; Golemis et al. 1989; see section below, Common Biological Themes in Oncogenesis).

Consistent with the importance of U3 sequences, changes in this region distinguish leukemogenic MCF viruses from closely related nonleukemogenic isolates (Fig. 7) (Thomas and Coffin 1982; M. Kelly et al. 1983; Lung et al. 1983; Quint et al. 1984; Holland et al. 1985, 1989; Stoye et al. 1991). Leukemogenic MCF viruses contain a portion of the env gene including that encoding TM residues and a large part of the U3 region of the LTR derived from the endogenous xenotropic provirus Bvx1 (Stoye et al. 1991). In addition, sequence information within the enhancer region of U3 is often amplified (Li et al. 1984; Chattopadhyay et al. 1989; Stoye et al. 1991; Morrison et al. 1995). Similar events occur in other high-lymphoma mouse strains including HRS, C58, and CWD (Thomas et al. 1984, 1990; Massey et al. 1990) and in FeLV-induced tumors (Miura et al. 1989; Fulton et al. 1990; Athas et al. 1995; Rohn and Overbaugh 1995). In all of these cases, LTR changes together with the changes affecting env gene structure discussed earlier contribute to the oncogenic properties of the viruses.

Sag Genes and MMTV Infection

Unlike other simple retroviruses, MMTVs contain an additional gene that has a novel and important role in mediating host infection and tumorigenesis. This gene, called sag for superantigen, was discovered when it became clear that a series of superantigen genes in mice first called Mls (minor lymphocyte-stimulating antigen) (Abe and Hodes 1989; Janeway et al. 1989; Janeway 1990) were really endogenous mtv loci (Frankel et al. 1991; Marrack et al. 1991; Woodland et al. 1991a,b; Coffin 1992; see Chapter 8. Like other superantigens (for review, see Chatila and Geha 1993; Marrack et al. 1993; Acha-Orbea and MacDonald 1995), Sag proteins interact with major histocompatibility complex (MHC) class II molecules and the variable region of the β-chain of the T-cell receptor (TCR). However, neither of these interactions involves sites that are engaged during presentation of conventional antigens (Fig. 8) (Janeway et al. 1989; Cazenave et al. 1990; Pullen et al. 1990, 1991; Torres et al. 1993). The sequences on the TCR that interact with superantigen, unlike those that form the binding pocket for conventional antigen, are shared by a large number of Vβ chains derived from one or a small subset of about two dozen germ-line genes. Thus, in contrast to a conventional antigen response in which only 0.01–0.001% of the T cells are stimulated, Sag protein can stimulate as many as 10% of all T cells. Like the conventional antigen response, stimulation initially induces proliferation of T cells; prolonged stimulation induces apoptosis and leads to the loss of the responding cell (Acha-Orbea and MacDonald 1995). This last feature of Sag expression by endogenous MMTV controls the Vβ gene families that are expressed by T cells in different strains of mice (see Chapter 8.

Figure 8. T-cell stimulation by conventional antigen and superantigen.

Figure 8

T-cell stimulation by conventional antigen and superantigen. (A) Processing and presentation of conventional antigen to a T cell; (B) presentation of a superantigen such as the MMTV-encoded Sag protein to a T cell. (Concept (more...)

The sag gene is located in the U3 region of the MMTV LTR and encodes a low-abundance, glycosylated, type II membrane protein (Choi et al. 1991, 1992; Beutner et al. 1992; Korman et al. 1992; Pullen et al. 1992). The low levels of expression have made the molecule difficult to study. However, it is synthesized as a precursor which is cleaved and expressed on the surface of cells (Brandt-Carlson and Butel 1991; Acha-Orbea et al. 1992; Choi et al. 1992; Knight et al. 1992; Korman et al. 1992; Winslow et al. 1992). The cleaved form of the molecule is the active moiety that interacts with MHC class II molecules (Winslow et al. 1994; Park et al. 1995). Sequences at the carboxyl terminus of Sag proteins encoded by different Mtv loci are variable, and this portion of the molecule determines the families of TCR β variable regions with which the Sag protein will interact (Choi et al. 1992; Knight et al. 1992; Yazdanbakhsh et al. 1993).

Expression of a functional Sag protein, presentation of the molecule by MHC class II proteins, and stimulation of reactive T cells are all required to establish MMTV infection (Golovkina et al. 1992, 1995; Beutner et al. 1994). In the natural setting, the virus is transmitted in infected milk to newborn mice before the stomach is acidified, and spreads to the Peyer's patch where it infects B cells (Held et al. 1993a,b, 1994; Tucek et al. 1993; Karapetian et al. 1994). The initial infection is followed by a vigorous Sag-mediated response in which T cells expressing TCRs containing Vβ proteins of a reactive type are stimulated (Golovkina et al. 1992; Held et al. 1993a, b). Stimulation of the T cells induces a proliferative B-cell response, amplifying the pool of infected and susceptible B cells. T cells can also be infected, but this appears to occur at later times postinfection (Waanders et al. 1993; Beutner et al. 1994).

The infected B cells transmit the virus to the mammary gland, and the virus infects the mammary epithelium. The virus replicates in these cells and stimulates proliferative changes which eventually lead to tumor formation. Like many other tumors induced by retroviruses that lack oncogenes, proviral insertional mutagenesis events have a key role in the development of a mammary carcinoma (Table 3). Very little is known about many of the features that are important in virus–host cell interaction in the mammary gland. The MMTV receptor has not been identified and the virus is very difficult to grow in vitro, complicating many genetic approaches that are routinely used with other retroviruses. Indeed, most studies of MMTV pathogenesis are still conducted in vivo, and transgenic animals expressing portions of the MMTV genome remain the approach of choice.

Tumor Induction by Viruses Containing v- onc Genes

Retroviruses that carry v-onc genes (sometimes called transforming viruses) induce a wide range of malignancies including sarcomas and hematopoietic cell tumors (see Table 2). Most of these agents were probably isolated from the original tumors in which they arose, making their discovery highly dependent on the detection and analysis of the tumor. This feature may contribute to the high frequency of isolates derived from chickens, which can be screened for disease in very large numbers in slaughterhouses, and cats, which are treated by veterinarians. Both of these species of animals are also frequently infected with replicating simple retroviruses that can recombine with cellular sequences. Viruses containing v-onc genes do not contribute significantly to the incidence of naturally occurring retrovirus-induced tumors because oncogene capture is rare. In addition, the low frequency of horizontal spread and the rapid and lethal consequences of infection limit the number of animals that become infected.

Although viruses containing v-onc genes are not a significant cause of tumors in the wild, these viruses have been very useful in experimental settings. Their ability to induce tumors rapidly and to transform cells in vitro has made them powerful tools for studies addressing the ways in which oncogenes alter cell growth. Most of these viruses are replication-defective, and the v-onc-encoded protein is often the only viral product. Although a replication-competent helper virus is present in most natural infections, in vitro transformation does not require replication. In some cases, viral replication is not required for tumor induction (Fung et al. 1983a; Green et al. 1987a; Poirier and Jolicoeur 1989). These features and the simple genetic organization of most v-onc-containing viruses facilitate studies addressing oncogenic mechanisms.

Three features of v-onc genes are important for their dominant effects on cell growth. First, they were selected for this property in the original tumor in which the gene was captured. The relatively high mutation rate inherent in retroviral replication and the opportunity for multiple replication cycles after oncogene capture foster selection of a recombined gene with high oncogenic potential (Eychene et al. 1990; Felder et al. 1991; Vennstrom et al. 1994). This process distinguishes v-onc genes from proto-oncogenes activated by proviral insertion; the latter are no more likely to undergo further mutation than other cellular sequences. Second, because v-onc genes are transmitted by viruses, they can be introduced into cells that may not express the proto-oncogene parents and are not equipped to regulate the activity of the v-onc gene products correctly. Third, the strong promoter and enhancer elements in the viral LTR ensure that high levels of the v-onc gene products are expressed.

Structure of v- onc Genes

The recombination and selection events that give rise to viruses containing v-onc genes allow for a wide range of final structures (Fig. 9). As noted earlier in Chapter 2 v-onc genes are often fused in-frame to parts of genes derived from the retrovirus parent. The viral sequences can target the v-onc gene product to particular subcellular compartments where it then interacts with components of growth-signaling pathways and induces transformation. Fusions involving gag sequences are particularly common (see Table 2 and Fig. 9) and sometimes direct proteins to the inner surface of the plasma membrane. In a similar fashion, fusion with env sequences can contribute a signal peptide that mediates intracellular transport and localization (Hannink and Donoghue 1984). In some cases, the gene fusion causes the v-onc gene product to localize in a subcellular compartment where the product of the c-onc gene is not normally found. For example, the v-abl oncogene is expressed as a Gag-Abl fusion and is localized to the membrane (Reynolds et al. 1978; Witte et al. 1978; Boss et al. 1981; Rohrschneider and Najita 1984); the normal c-abl gene product is a nuclear protein (Van Etten et al. 1989).

Figure 9. Representative v-onc-containing retroviral genomes and protein products.

Figure 9

Representative v-onc-containing retroviral genomes and protein products. The genome structures and protein products of several different v-onc-containing retroviruses are shown. Note that the structure of the different viruses (more...)

The flexibility inherent in v-onc gene structure is emphasized by comparing the structure of v-onc genes that have been captured multiple times (see Table 2). Between 30% and 50% of chickens that develop ALV-induced erythroblastosis contain viruses that have captured the erbB gene (for review, see Maihle and Kung 1988). The 5′end of the c-erbB-coding sequence in these viruses is the same, and many express a signal peptide derived from env-coding sequences. These features probably reflect the way in which proviruses inserted into the c-erbB locus direct expression of the gene (see above Tumor Induction by Simple C-type Retroviruses That Lack Oncogenes). Despite these similarities, the precise recombination points in all of the viruses are distinct (Gamett et al. 1986; Pelley et al. 1988; Raines et al. 1988a,b; Robinson et al. 1992a; Vennstrom et al. 1994). In addition, each isolate has a unique constellation of point mutations and deletions in the newly captured v-erbB gene that affect sequences in the carboxy-terminal portion of the protein. Although the transforming ability of these isolates varies (Tracy et al. 1985; Gamett et al. 1986; Pelley et al. 1988; Raines et al. 1988b; Vennstrom et al. 1994), no particular activating mutation is required and the impact of any one change depends in part on the other erbB mutations that are present.

Although the frequent capture characteristic of the ALV-erythroblastosis system does not occur with other oncogenes, analyses of other viruses reveal similar patterns. There are multiple isolates of RSV (for review, see Payne 1992), and additional strains have been recovered from chickens infected with transformation-defective mutants of RSV (Hanafusa et al. 1977; Halpern et al. 1979; Enrietto et al. 1983). Common features are shared by all of these v-src genes, but none are identical. The v-fps gene has been found in five different avian sarcoma viruses, in each case as a gag-fps fusion. However, each of these encode slightly different Fps proteins (Wang et al. 1981; Neel et al. 1982a; Shibuya and Hanafusa 1982; Wong et al. 1982; Hammond et al. 1985). The v-sis oncogene is fused to env sequences in simian sarcoma virus (SSV) and to gag sequences in Parodi-Irgens feline sarcoma virus (PI-FeSV) (Devare et al. 1982; Besmer et al. 1983a). Similar differences are evident when the retroviruses that have captured the fos gene are compared (Curran and Verma 1984; Van Beveren et al. 1984; Nishizawa et al. 1987).

v- onc Gene Nomenclature and Families

Most v-onc genes were named as they were discovered using designations based on the name of the virus or the disease associated with the agent (e.g., src for Rous sarcoma virus, abl for Abelson murine leukemia virus; Coffin et al. 1981). In many cases, the names provide very few clues concerning any of these features. Indeed, some names were given before it was realized that two independently isolated viruses had captured the same gene (see Table 2). For example, the v-fes and v-fps genes are the feline and avian homologs of the same gene (Hampe et al. 1982), and v-mil and v-raf were derived from the same gene of chickens and mice, respectively (Jansen et al. 1984; Kan et al. 1984). Sometimes, as in the case of fes and fps, the viruses induce the same types of tumors; in other instances, such as mil and raf, they do not (Table 2). This feature probably reflects both host differences and tissue tropisms of the viruses that carry the oncogenes.

v-onc genes are usually classified in families based on the function of the protein encoded by their protooncogene parent. Most classes of proteins involved in growth stimulation are represented among these families which include growth factors, growth factor receptors, nonreceptor protein tyrosine kinases, adapter signaling proteins, G proteins, serine/threonine kinases, and transcription factors (Fig. 10). v-onc genes encode only a subset of molecules that belong in each class. The central role of proto-oncogenes in cancer and in regulating normal cell growth and differentiation has expanded study of these molecules beyond the bounds of retrovirology. Thus, only a brief overview of the function of the different classes of v-onc genes is provided here. For more detailed information on this area, see reviews that focus on the function of the different types of oncogenes and mechanisms of growth control (Blackwood et al. 1992a; Marcu et al. 1992; Hunter 1994, 1995; Marshall 1994, 1995; Pawson 1994, 1995; van der Geer et al. 1994; Williams and Roberts 1994; Cohen et al. 1995; Heldin 1995; Hill and Treisman 1995; Ihle et al. 1995; Karin and Hunter 1995; Miyamoto and Verma 1995; Spaargaren et al. 1995).

Figure 10. Signal transduction pathways involved in growth regulation through receptor tyrosine kinases.

Figure 10

Signal transduction pathways involved in growth regulation through receptor tyrosine kinases. Some of the pathways that control cellular growth in response to external signals are illustrated. For clarity, only a subset of possible (more...)

Protein Tyrosine Kinases

Protein tyrosine kinases (PTKs) are one of the largest groups of oncoproteins. This class of proteins was first discovered when the v-Src protein encoded by RSV was shown to phosphorylate proteins on tyrosine (Collett et al. 1980; Hunter and Sefton 1980). Soon it became clear that tyrosine phosphorylation was an important way in which cells regulate their growth and that the Src protein was one of a family of PTKs. This family includes proteins that localize to the inner surface of the plasma membrane such as Src and others such as ErbB that are expressed as transmembrane proteins and function as growth factor receptors (see Table 2). The kinase activity of the receptor PTKs is normally activated following interaction with ligand; the nonreceptor PTKs are activated in response to extracellular signals transmitted from receptors, including those transmitted by receptor PTKs (for review, see Hunter 1994, 1995; Schlessinger and Bar-Sagi 1994; van der Geer et al. 1994; Heldin 1995). Nonreceptor PTKs contain two domains called SH2 and SH3 (src homology domains 2 and 3) that are not found in receptor PTKs but are present in other types of signaling molecules (for review, see Pawson 1994; Cohen et al. 1995). SH2 domains recognize and bind tyrosine phosphorylated residues, whereas SH3 domains interact with proteins via proline-rich motifs. Both of these types of protein-protein interactions facilitate signal transmission.

Genes encoding both receptor and nonreceptor PTKs have been captured by retroviruses (Table 2). Uncontrolled tyrosine phosphorylation of multiple cellular proteins by these oncoproteins induces transformation. Although some of the key substrates have yet to be identified, the oncogenic PTKs transmit growth signals downstream via multiple pathways, including those feeding through Ras proteins, phospholipase-Cγ, phosphatidylinositol-3-kinase, and other molecules (Fig. 10) (for review, see Kapeller and Cantley 1994; van der Geer et al. 1994; Ihle et al. 1995; Karin and Hunter 1995; Liscovitch and Cantley 1995; Pawson 1995). Eventually, these signals reach the nucleus, where they affect transcription of growth regulatory genes and alter components that regulate cell cycle progression and cell survival.

Several types of mutations can be found in the v-onc genes that encode PTKs. All of these changes affect sequences that normally regulate the kinase activity. For most of the nonreceptor PTKs, loss of a carboxy-terminal tyrosine residue has a key role (Fig. 11A) (Cartwright et al. 1987; Kmiecik and Shalloway 1987; Piwnica-Worms et al. 1987; Reynolds et al. 1987). Phosphorylation of this residue normally allows the PTKs to assume an inactive conformation through interactions between the carboxyl terminus and the SH2 region; when the tyrosine is missing, the regulatory circuit is disrupted (for review, see Liu and Pawson 1994). Deletion of either sequences encoding the SH3 domain or a portion of the carboxy-terminal region of the c-Abl protein leads to constitutive activation of the v-Abl PTK by interfering with the binding of a cellular factor that appears to control kinase activity (Bergold et al. 1987; Franz et al. 1989; Jackson and Baltimore 1989; Dai and Pendergast 1995; Shi et al. 1995).

Figure 11. Mechanisms by which protein tyrosine kinases can be activated.

Figure 11

Mechanisms by which protein tyrosine kinases can be activated. (Top, A) Relationship between the c-Src and v-Src proteins. (Top, B) Relationship between c-ErbB and a representative v-ErbB protein. In addition to the structural (more...)

Most commonly, v-onc genes that encode receptor PTKs have lost the sequences that encode some or all of the ligand-binding region of the proteins, rendering the kinases constitutively active (Fig. 11B). The v-kit, v-erbB, v-sea, v-ros, and v-eyk oncogenes have all undergone changes of this sort (Downward et al. 1984; Matsushime et al. 1986; Qiu et al. 1988; Huff et al. 1992; Jia et al. 1992; Jia and Hanafusa 1994). Additional mutations, affecting sequences involved in intracellular signaling, are also found in many of these oncogenes. The v-fms oncogene of SM-FeSV, which encodes an activated CSF-1 receptor, retains a mutated version of the sequences that encode the ligand-binding domain. One of the mutations causes a single-amino-acid substitution in the extracellular portion of the molecule. As a consequence, the conformation of this domain is altered, mimicking the activation that occurs during normal ligand binding (Roussel et al. 1988; Woolford et al. 1988).

Growth Factors

Although growth factor receptors are common targets of oncogenic activation, v-sis is the only v-onc gene that encodes a growth factor—the β-chain of the platelet-derived growth factor (PDGF) (Doolittle et al. 1983; Waterfield et al. 1983). This molecule, normally expressed by platelets, endothelial cells, and cells in the placenta, stimulates the growth of mesenchymal and endothelial cells by binding to the PDGF receptor, a member of the receptor PTK family (Eck et al. 1982; Nishimura et al. 1982; Yarden et al. 1986). Under normal circumstances, PDGF binding activates the kinase and triggers growth stimulatory signals. The response is down-modulated by receptor turnover and downregulation of the receptor molecule.

Like PDGF-β, the v-Sis protein binds the PDGF receptor. This interaction establishes a continuous autocrine loop that is required for transformation (Leal et al. 1985). Although binding can occur intracellularly (Keating and Williams 1988; Bejcek et al. 1989), cell surface localization of the v-Sis receptor complex is probably required for full transforming potential (Hannink and Donoghue 1988; Fleming et al. 1989; Hart et al. 1994). Membrane expression may facilitate interactions between the activated receptor and downstream signaling components. Although the interaction of v-Sis with its receptor supports a simple autocrine model for transformation, continuous stimulation with high levels of PDGF does not transform cells (Bejcek et al. 1989). This difference probably reflects the ability of normal cells to down-modulate the signaling circuit by controlling receptor-ligand interaction, a mechanism that is inoperative or overwhelmed in cells expressing v-Sis.

Adapter Proteins

One way in which PTKs transmit signals to downstream effectors is via molecules called adapter proteins (for review, see Pawson 1995). These proteins lack enzymatic activity but can interact with tyrosine and serine/threonine kinases and other proteins via SH2 and SH3 domains. When they are tyrosine-phosphorylated, adapter proteins bring components of signal transduction pathways together and facilitate downstream signal transmission. Mutations affecting adapter protein function or overexpression of some of these proteins can transform cells (Katzav et al. 1989; Chou et al. 1992; Pellici et al. 1992; Egan et al. 1993). However, only one gene in this group, the c-crk gene, has been captured by retroviruses (Mayer et al. 1988; Tsuchie et al. 1989). This gene and the related crkl gene (ten Hoeve et al. 1993) encode widely expressed proteins that play a still poorly understood part in signal transduction.

The v-Crk protein is expressed as a Gag-Crk fusion. As a consequence of the capture events, amino-terminal c-Crk sequences that may regulate c-Crk function are missing in v-Crk; the protein also contains several amino acid substitutions (Mayer et al. 1988; Reichman et al. 1992; Fajardo et al. 1993). Cells transformed by v-Crk contain elevated levels of phosphotyrosine, and the v-Crk protein binds tyrosine-phosphorylated proteins, including some PTKs, in a stable fashion (Matsuda et al. 1990; Mayer and Hanafusa 1990a). These interactions correlate with transformation (Mayer and Hanafusa 1990b) and appear to alter the normal regulatory circuits that control the PTK activity of several cellular kinases. The exact mechanisms involved are not known. However, v-Crk protein interacts with cytoskeletal elements (Birge et al. 1993; Sakai et al. 1994; Polte and Hanks 1995) and the c-Abl protein (Feller et al. 1994; Ren et al. 1994). The protein can also stimulate the Ras pathway by interacting with GTP exchange factors (Knudson et al. 1994; Tanaka et al. 1994).

Ras Proteins

Ras proteins are members of the G-protein superfamily, molecules that participate in signaling by cycling between GTP- and GDP-bound forms (for review, see Boguski and McCormick 1993; McCormick 1994). Normally, the activity of Ras is regulated by upstream signals from receptors that have been triggered by growth factors or other cellular stimuli. These signals are transmitted via adapter proteins such as c-Crk, Grb-2, and Shc which interact with receptors or molecules associated with the cytoskeleton and bring guanine nucleotide exchange factors like the Sos and C3G proteins to the membrane (Lowenstein et al. 1992; Pelicci et al. 1992; Rozakis-Adcock et al. 1992, 1993; Buday and Downward 1993; Egan et al. 1993; Gale et al. 1993; Li et al. 1993; Knudson et al. 1994; Schlaepfer et al. 1994; Tanaka et al. 1994). Membrane localization of the exchange factors stimulates the formation of Ras-GTP (Aronheim et al. 1994) which in turn can transmit the signal downstream via several pathways. One of these involves the c-Raf serine/threonine kinase (see below; Moodie et al. 1993; Vojtek et al. 1993; Warne et al. 1993; Zhang et al. 1993), a protein that activates the MAP kinase pathway (for review, see Marshall 1994; McCormick 1994); another involves the MEKK serine/ threonine kinase which stimulates the JNK kinase pathway (Derijard et al. 1994; Coso et al. 1995; Minden et al. 1995). Down-modulation of Ras-GTP is mediated by GAP proteins, molecules that stimulate the GTPase activity of Ras (Trahey and McCormick 1987; Vogel et al. 1988; Martin et al. 1990; Xu et al. 1990b; Maekawa et al. 1994).

Three different ras genes have been captured by retroviruses (see Table 2). Despite their diverse origins and the presence of several sequence differences, all of the v-ras genes contain point mutations affecting codon 12 (Dhar et al. 1982; Tsuchida et al. 1982; Rasheed et al. 1983; Reddy et al. 1985; Fredrickson et al. 1987). This change is particularly important; substitution of almost any amino acid at this position activates the transforming potential of c-Ras protein by altering the way in which the protein interacts with GTP and lowering intrinsic GTPase activity (Gibbs et al. 1984; McGrath et al. 1984; Seeburg et al. 1984; Sweet et al. 1984; Tong et al. 1989; Krengel et al. 1990; Milburn et al. 1990; Chung et al. 1993). As a consequence, the high levels of GTP-bound v-Ras protein present in infected cells produce a sustained activation signal. This response is heightened because v-Ras proteins are resistant to the effects of GAP proteins (Trahey and McCormick 1987; Vogel et al. 1988; Martin et al. 1990; Xu et al. 1990a). Mutations affecting codon 12 are often found in c-ras genes in tumors that lack a retroviral etiology (for review, see Barbacid 1987; Bos 1989), highlighting the broad effects that mutation of ras genes can have on cellular growth.

Serine/Threonine Kinases

Serine/threonine phosphorylation events have a very important role in stimulating cell growth, and not surprisingly, two genes encoding these kinases have been captured by retroviruses. One of these, v-raf/mil, encodes an altered form of the c-Raf protein. As noted above, c-Raf interacts with Ras and other signaling components and activates the MAP kinase pathway (for review, see Williams and Roberts 1994; Magnuson et al. 1995). Interaction with Ras brings Raf to the membrane, a site that is essential for Raf-mediated signaling (Leevers et al. 1994; Stokoe et al. 1994). The factors that control c-Raf kinase activity are still unknown. However, 14-3-3 proteins, serine/threonine ceramide-activated kinase, and several other proteins have been implicated as possible regulators (Freed et al. 1994; Irie et al. 1994; Dent et al. 1995; Yao et al. 1995; for review, see Magnuson et al. 1995).

Two retroviruses have captured raf/mil genes (Table 2). One of these, 3611-MSV, contains only the v-raf gene (Rapp et al. 1983a,b); the other, MH2, has also captured the myc oncogene (Coll et al. 1983; Jansen et al. 1983; Kan et al. 1983). Although both viruses are highly transforming, the v-myc sequences in MH2 are required for full transforming potential (Graf et al. 1986; Palmieri and Vogel 1987; Bechade et al. 1988). This property probably reflects differences in the response of chicken and mouse cells to v-raf/mil genes. The products of v-raf/mil are expressed as Gag fusions that have constitutive serine/threonine kinase activity (Hu et al. 1978; Rapp et al. 1983a; Moelling et al. 1984). The proteins lack amino-terminal sequences that appear to regulate c-Raf/Mil activity; deletion of the sequences encoding these residues is sufficient to activate the transforming potential of c-raf (Stanton et al. 1989; Heidecker et al. 1990).

The second retroviral v-onc gene in this group is v-mos, a gene found in several murine sarcoma viruses (MSVs) (Table 2). Normally, the c-Mos protein regulates oocyte maturation by controlling meiosis in both a positive and negative fashion at different points during the process (Sagata et al. 1988, 1989a,b; Yew et al. 1992; for review, see Yew et al. 1993). Although high levels of c-Mos protein are found in skeletal muscle (Leibovitch et al. 1991), most somatic tissues do not express the gene. Indeed, when c-mos is expressed from retroviral constructs in fibroblast cells, the cells become transformed (Blair et al. 1981; van der Hoorn et al. 1982). Thus, ecotopic expression of v-mos has a central role in transformation by MSVs carrying the oncogene.

The serine/threonine kinase activity of the v-Mos protein is required for transformation (Hannink and Donoghue 1985; Maxwell and Arlinghaus 1985) as is v-Mos-mediated activation of the MAP kinase pathway (Fabian et al. 1993; Nebreda and Hunt 1993; Posada et al. 1993; Okazaki and Sagata 1995). Mos protein also interacts with cytoskeletal proteins (Zhou et al. 1991; Bai et al. 1993) and with the components of the mitotic spindle (Wang et al. 1994). This interaction may interfere with spindle formation and mitosis, causing an increase in chromosomal instability and allowing some cells to enter the M phase of the cell cycle inappropriately. These events could facilitate generation of tetraploid or aneuploid cells and retention of uncorrected mutations generated during S phase and contribute to transformation (for review, see Yew et al. 1993).

Transcription Factors

The largest and most diverse group of v-onc genes encodes transcription factors. Each of the oncogene groups discussed so far causes transformation by altering gene expression indirectly. The need for these signals can be bypassed by altering the genes whose products interact with the transcription complex and control gene expression directly. Thus, it is not surprising that a number of retroviruses have captured oncogenes that encode transcription factors (Table 2). A large amount of information has been obtained concerning the types of changes that activate the oncogenic potential of the v-onc genes in this group. For some oncogenes, activation correlates with stimulation of gene expression; for others, repression of gene expression contributes to transformation. However, in most cases, the exact mechanisms that lead to altered gene expression are only partly understood. In addition, very little is known about the target genes that have dominant roles in mediating transformation. Identifying these genes and understanding the way in which altered expression causes transformation remain major challenges for the future.

As with other oncogene groups, mutations that activate the transforming potential of transcription factor proto-oncogenes often involve alterations that affect regulatory domains of their products. For the many of these oncogenes, mutations remove sequences encoding negative regulatory elements. Thus, the proteins encoded by the oncogenes are able to alter transcription of target genes in the absence of appropriate stimulatory signals. For example, the v-myb oncogene, found in AMV, encodes a transcription factor that has lost negative regulatory domains present at both the amino and carboxyl termini of the c-Myb protein; similar truncations are found in the v-myb portion of the v-myb-ets oncogene found in E-26 (Klempnauer et al. 1982; Nunn et al. 1984; Gerondakis and Bishop 1986; Biedenkapp et al. 1988; Weston and Bishop 1989; Ibanez and Lispsick 1990; Lane et al. 1990; Luscher et al. 1990). Interestingly, c-myb genes that have been targeted by proviral insertional mutagenesis encode proteins that are missing these sequences (Lavu and Reddy 1986; Shen-Ong et al. 1986; Gonda et al. 1987; Shen-Ong and Wolff 1987; Weinstein et al. 1987; Kanter et al. 1988; Pizer and Humphries 1989; see Table 3). Point mutations present in v-myb and fusion with v-ets also contribute to transforming potential and influence the types of cells that can be transformed (Introna et al. 1990; Dini et al. 1995; see section below, Common Biological Themes in Oncogenesis).

An internal deletion in the v-jun oncogene of ASV-17 causes the loss of a negative regulatory region called δ (Maki et al. 1987; Bohmann and Tjian 1989; Baichwal and Tjian 1990). Loss of this region is critical for transformation and alters the turnover and phosphorylation of Jun protein (Bos et al. 1990; Adler et al. 1992; Treier et al. 1994). However, whether either of these mechanisms is directly relevant to transformation requires further investigation. As in the case of v-myb, several point mutations in v-jun enhance transformation (Bos et al. 1990; Morgan et al. 1993).

Some of the oncogenes in this group acquire transforming function because expression in a retrovirus allows the accumulation of large amounts of the oncoprotein. This mechanism is important for transformation by the v-myc oncogene (Land et al. 1983; for review, see Weinberg 1985; Bishop 1987). Normally, c-myc is tightly regulated and preferentially expressed in growing cells at particular phases of the cell cycle (K. Kelly et al. 1983; Hann et al. 1985; Rabbitts et al. 1985). Protein-protein interactions between Myc and a second protein called Max provide a second level of regulation (Blackwood and Eisenman 1991; Prendergast et al. 1991; Berberich and Cole 1992; Blackwood et al. 1992b; Kretzner et al. 1992). Although Myc can bind DNA (Blackwell et al. 1990), Myc-Max heterodimers activate transcription to high levels. Max is a relatively abundant protein whose activity is modulated by dimerization with one of several other nuclear proteins (Ayer et al. 1993; Zervos et al. 1993). Elevated levels of Myc protein, expressed by the strong retroviral promoter and enhancer elements in the LTR, swamp these normal regulatory circuits and induce transformation (Amati et al. 1993). Like many other oncogenes, point mutations found in v-myc genes enhance the ability of the viral oncogene to transform cells (Frykberg et al. 1987; Symonds et al. 1989).

The v-fos oncogene uses a similar mechanism. Tightly regulated transcription and rapid turnover of c-fos mRNA normally regulate levels of c-Fos protein (Greenberg and Ziff 1984; Kruijer et al. 1984; Muller et al. 1984; Treisman 1985). High levels of expression from the retroviral promoter elements, coupled with the loss of sequences that normally mediate mRNA turnover, allow oncogene activation. These events are extremely important for oncogenic activation of v-fos. Indeed, even though the v-fos genes found in FBR-MSV and FBJ-MSV and in NK24-ASV contain mutations that affect the coding sequence of the v-Fos protein (Van Beveren et al. 1983, 1984; Curran and Verma 1984; Nishizawa et al. 1987), overexpression of c-fos in a retroviral vector is sufficient to induce transformation (Miller et al. 1984).

Not all transcription factor oncogenes stimulate gene expression. The v-erbA oncogene found in some AEV isolates (Table 2) encodes a mutated form of the thyroid hormone α receptor (Sap et al. 1986; Weinberger et al. 1986). The c-ErbA protein normally activates gene expression when bound by thyroid hormone. The oncogenic form of the molecule fails to bind hormone but still binds DNA, resulting in constitutive repression of target DNA elements (Damm et al. 1989; Sap et al. 1989; Zenke et al. 1990). This effect extends to DNA elements that normally respond to several members of the steroid/retinoid receptor family (Sharif and Privalsky 1991). AEV transforms erythroid cells, and at least three of the genes affected by ErbA are involved in erythroid differentiation (Zenke et al. 1988, 1990; Pain et al. 1990; Schroeder et al. 1990). Repression of these genes suppresses cell differentiation and reduces the requirements of AEV-transformed cells for specific growth signals (see below Common Biological Themes in Oncogenesis). Both of these effects facilitate transformation and induction of erythroblastosis by AEV. Repression of gene expression may also have an important role in transformation by the v-rel oncogene (Ballard et al. 1990, 1992; Inoue et al. 1991; Richardson and Gilmore 1991; McDonnell et al. 1992). However, the v-Rel protein can stimulate gene expression in some settings (Gelinas and Temin 1988; Hannink and Temin 1989; Boehmelt et al. 1992; Walker et al. 1992; Sarkar and Gilmore 1993).

Other Virus Determinants That Modulate Oncogenicity

Although the v-onc gene product is the dominant viral product controlling tumor induction, other elements can modulate the ability of viruses containing v-onc genes to induce tumors and transform cells. Synthesis of the v-onc-encoded protein is controlled by elements in the LTR; these sequences must direct synthesis of sufficient quantities of the oncoprotein to stimulate cell growth. Because viruses containing v-onc genes were originally isolated from tumors, the LTR sequences selected during oncogene capture can clearly accomplish this task. However, unlike retroviruses that lack oncogenes, the particular LTR sequences found in v-onc-containing retroviruses do not have a prominent role in modulating the frequency or type of tumor that is induced (Green et al. 1988; Introna et al. 1990; Jolicoeur et al. 1991a; Feuer et al. 1993a; Engelke and Lipsick 1994). The myeloproliferative sarcoma virus (MPSV) and the malignant histiocytosis virus (MHV) are exceptions to this generalization; the ability of these viruses to induce diseases affecting myeloid cells is controlled by sequences in the LTR (Stocking et al. 1985, 1986; Friel et al. 1990).

As noted earlier, helper viruses are not always required for tumor induction by v-onc-containing viruses. However, they can affect tumorigenesis and transformation under some circumstances. Abelson MLV (Ab-MLV) stocks containing some helper viruses fail to induce pre-B-cell tumors in certain strains of mice, a feature that maps to the LTR of the helper virus (Rosenberg and Baltimore 1978; Scher 1978; Savard et al. 1987). The requirement for helper virus integrations at the ahi1 locus and other, as yet unknown, loci appears to control tumorigenesis in this situation (Poirier et al. 1988). This requirement does not extend to thymic tumors induced by the same virus (Poirier and Jolicoeur 1989), highlighting the ways in which cell type can influence the outcome of virus-host interaction.

Common Biological Themes in Oncogenesis

Despite the diversity of oncogenic retroviruses, tumor induction by these agents involves several common biological themes. First, a multistep process is often involved in tumor induction. Despite the dramatic differences in latent period that distinguish tumors induced by viruses carrying v-onc genes and those lacking them, the dominant growth signals provided by v-onc gene products are usually not sufficient to convert a normal cell to a full-blown tumor cell. A second related feature is the requirement for cooperativity between different oncogenes. Cooperating genes may be present in a single v-onc-gene-containing retrovirus, they may be targeted by multiple proviral insertions in an individual tumor, or they may be activated as a consequence of spontaneous mutation. In some retroviral systems, cooperativity involves stimulatory signals from oncogenes and the loss of negative regulatory signals from tumor suppressor genes. Finally, the specificity involved in virus-cell interaction, reflected in the tendency of a particular virus to induce a specific type of tumor, is a property shared by all oncogenic retroviruses.

Many of the biological features characteristic of retrovirus-induced tumors parallel those evident in tumors that lack a retroviral etiology. The multistep nature of the development of many spontaneous tumors has long been appreciated, and cooperativity between multiple genes is a frequent occurrence. Much of our understanding concerning the interaction between upregulation of positive growth signals from oncogenes and loss of negative growth signals from tumor suppressor genes comes from these studies (for review, see Vogelstein and Kinzler 1993; Weinberg 1995a). In addition, the association of particular genetic alterations with specific types of cancers, first established by the presence of tumor-specific chromosomal markers (for review, see Rabbitts 1994), is a common feature of most tumor systems. These unifying features are another way in which the significance of studies conducted with retroviral tumor models extends to broader issues of carcinogenesis.

Cooperativity between v- onc Genes

Although the majority of retroviruses that carry oncogenes have captured a single proto-oncogene sequence, some contain sequences derived from two different c-onc genes. In many of these instances, both oncogenes cooperate to induce disease. For example, the avian erythroblastosis virus AEV-ES4 expresses both the v-erbA and v-erbB oncogenes and induces a more rapidly fatal erythroleukemia than that of the related AEV-H isolate of AEV which expresses only v-erbB (for review, see Beug and Graf 1989; Hayman and Beug 1992; Metz 1994). Similarly, AMV-BAI, an avian myeloblastosis virus, expresses the v-myb oncogene and induces a leukemia of myeloid precursors. AMV-E26, which expresses both v-myb and v-ets as a Gag-Myb-Ets fusion protein, induces both myeloblastosis and erythroblastosis because the presence of both oncogenes as a gene fusion facilitates transformation of a broader range of target cells, including multipotential progenitor cells (Metz and Graf 1991a,b; Graf et al. 1992).

Despite these examples of cooperativity, all of the cellular sequences present in oncogene-containing viruses do not seem to have an important role in tumor induction. GR-FeSV contains a portion of actin sequences between gag and fgr sequences (Naharro et al. 1984) and FBR-MSV contains a sequence unrelated to fos that has been termed fox (Van Beveren et al. 1984). However, neither of these sequences are absolutely required for transformation (Miller et al. 1984; Kappes et al. 1989; Miyoshi et al. 1989; Sugita et al. 1989). Harvey and Kirsten MSVs (Ha-MSV and Ki-MSV) carry v-ras oncogenes and noncoding portions of an endogenous rat virus element called VL30 (see Chapter 8 (Shih et al. 1978; Chien et al. 1979; Ellis et al. 1980; Young et al. 1980). The VL30 sequences share homology with c-Ha-ras in the vicinity of the recombination breakpoint, a feature that probably facilitates capture of the ras sequences (Firulli et al. 1993; Makris et al. 1993b). These sequences may also facilitate expression of the oncoprotein and contribute to the transforming ability of Ha-MSV (Velu et al. 1989).

Although signals from v-onc genes have the dominant role in transformation, changes in cellular genes are often required for transformation to occur. The differences observed in transformation of immortalized cell cultures and the primary cell cultures from which they were derived highlight this contribution. Some v-onc genes fail to transform certain kinds of primary cell cultures but will transform established cell lines derived from them (Scher and Siegler 1975; MacAuley and Pawson 1988); in other cases, infection with more than one transforming unit of virus is required for transformation of primary cells (Hartley and Rowe 1966). The nature of the changes that distinguish immortalized cells from primary cell cultures is largely unknown and may involve activation of proto-oncogenes and inactivation of tumor suppressor genes (see below). Consistent with the requirement for multiple changes, primary cell cultures can often be transformed by combinations of oncogenes. For example, ras and myc transform primary rat embryo fibroblasts, cells that are not transformed by either oncogene alone (Land et al. 1983, 1986).

Cooperativity and Insertional Mutagenesis

Oncogene cooperativity is important in tumors induced by proviral insertional mutagenesis. More than a dozen different loci are targeted in Mo-MLV-induced thymomas and at least four different common integration sites have been identified in FeLV-induced thymomas (Table 3). Often, an individual tumor contains insertions into more than one of these loci. For example, many Mo-MLV-induced thymomas contain proviruses integrated at both the c-myc and pim1 loci (Selten et al. 1984, 1986; Cuypers et al. 1986; Fan et al. 1988), and integrations into both wnt1 and int2 are often found in a single MMTV-induced mammary tumor (Peters et al. 1986). Additional examples have been observed in other murine and feline tumors (Tsatsanis et al. 1994; Liao et al. 1995).

Sometimes, particular insertions occur at specific stages of tumor development. Integrations into the c-myc locus, a common feature of avian bursal lymphomas (see above Tumor Induction by Simple C-type Retroviruses That Lack Oncogenes), appear to occur early in the tumorigenic process and promote follicle development (Thompson et al. 1987). A similar phenomenon has been observed in thymic tumors in AKR mice (O'Donnell et al. 1985). Insertions into the c-myb locus can be detected soon after mice are infected with Mo-MLV (Nason-Burchenal and Wolff 1993; Belli et al. 1995) and in chicks infected with RAV-1 as embryos (Pizer et al. 1992). However, expression of myb is not sufficient for tumor induction under these circumstances, and in the case of Mo-MLV-induced thymic tumors, the tumors that finally arise do not contain myb integrations (Belli et al. 1995). This pattern suggests that integrations into c-myb induce a stimulatory signal early in the oncogenic process that is no longer required later in tumor development.

Other loci appear to be targeted at late stages in tumorigenesis. These integrations occur as a consequence of superinfection of cells, demonstrating that the continued presence of replicating virus and reinfection of cells have a key role in development of malignant disease. Two loci, Tpl1 (Bear et al. 1989) and Tpl2 (Patriotis et al. 1993), are targeted at late stages in Mo-MLV-induced thymic tumors. The Tpl1 locus encodes an Ets-related protein (Bear et al. 1989), whereas Tpl2 encodes a serine/threonine kinase that interacts with Ras and Raf and stimulates MAP kinase (Patriotis et al. 1993, 1994). Expression of these genes appears to stimulate factor-independent growth of the tumor cells and may contribute to their ability to metastasize. Integrations into fit1, a locus of unknown function, appear to be a late event in FeLV-induced tumorigenesis and may have a similar role in this system (Tsatsanis et al. 1994).

Transgenic Mice and Oncogene Cooperativity

The development of transgenic mice carrying oncogenes has provided a directed approach to analysis of cooperativity in tumor induction. The transgenic animals are usually susceptible to spontaneous tumors involving the tissue(s) in which the transgenic on cogene is expressed. However, in most cases, only a fraction of the animals develop tumors and a long latent period is required, indicating that expression of the transgenic oncogene is not sufficient for tumor development (Compere et al. 1988; Adams and Cory 1991a,b). An enhanced frequency of tumor development and a shortened latent period have been observed in the progeny of crosses between transgenic animals carrying different oncogenes. Oncogenes that cooperate to induce tumors can be identified in this way. However, such experiments obviously require large numbers of different transgenic animals.

A more directed approach to detect cooperating oncogenes involves infecting animals carrying a single oncogenic transgene with a retrovirus and analyzing the tumors that arise for common proviral insertions (see Table 4). These analyses have shown that multiple pim family members and several members of the myc gene family can cooperate to induce thymic lymphoma (van Lohuizen et al. 1989b, 1991) and that wnt and fgf genes can cooperate to induce mammary carcinomas (Shackleford et al. 1993; Lee et al. 1995; MacArthur et al. 1995). Because the expression of the transgene replaces the need for one of the signals required for tumor induction, the latent period required for disease development is shortened significantly (Fig. 12). To reveal additional genes, transgenic mice carrying an oncogene but lacking a functional copy of a second “cooperating” oncogene have been analyzed. Such a strategy has identified a new member of the pim gene family (van der Lugt et al. 1995). Although many of the cooperative relationships uncovered in these experiments were already supported by analyses of virus-induced tumors in nontransgenic mice, several novel genes have also been found. Similar approaches using different combinations of “engineered” mice and oncogenic retroviruses may reveal additional genes. Even though the oncogenic potential of some of these genes may reflect special conditions that are present in transgenic mice, their identification and study advance our understanding of the complex ways in which cells control their growth.

Table 4. Loci Targeted by Proviral Insertional Mutagenesis in Transgenic Mice.

Table 4

Loci Targeted by Proviral Insertional Mutagenesis in Transgenic Mice.

Figure 12. Induction of Mo-MLV thymomas is accelerated in transgenic mice expressing c-myc.

Figure 12

Induction of Mo-MLV thymomas is accelerated in transgenic mice expressing c-myc. Kaplan-Meier plots illustrate the induction of tumors in wild-type mice, transgenic mice expressing c-myc under control of an immunoglobulin enhancer (more...)

Tumor Suppressor Genes and Retrovirus-induced Tumors

Although proto-oncogenes can cooperate to enhance tumor induction by providing more than one positive growth signal, a positive growth signal can also be complemented by the loss of a growth suppressive signal from a tumor suppressor gene. These genes encode proteins that normally regulate growth, often by mediating cell cycle regulation, and are frequent targets of mutation in human tumors (for review, see Vogelstein and Kinzler 1993; Hinds and Weinberg 1994; Canman and Kastan 1995; Weinberg 1995a). For example, the p53 gene encodes a protein that mediates transit of the G1 checkpoint and is mutated in a high frequency of breast and colon carcinomas and many other tumors (for review, see Levine 1993; Lee and Bernstein 1995; Selivanova and Wiman 1995). Germ-line point mutations in the human p53 gene are responsible for Li-Fraumeni syndrome, a genetic disorder associated with very high frequencies of multiple tumor types (for review, see Malkin 1994a,b). Other tumor suppressor genes include the retinoblastoma gene RB, the neurofibromatosis gene NF1, and the DCC and APC genes which are frequently mutated in colon carcinoma (for review, see Bernards 1995; Cho and Fearon 1995; McCormick 1995; Polakis 1995; Weinberg 1995b; Whyte 1995). Unlike proto-oncogenes, ablating the function of a tumor suppressor gene requires that both copies of the gene be damaged or lost or that the mutation affecting one allele results in a protein which can inactivate the function of the molecule encoded by the second allele.

Studies of retrovirus-induced tumors have not led to the identification of a large number of tumor suppressor genes. No classically defined tumor suppressor genes have been recovered as v-onc genes. Such genes would have to act in a dominant-negative fashion and encode products that interfered with the function of the normal tumor suppressor gene products, a feature that places severe restrictions on the changes that are compatible with oncogenic function. Similarly, there is a low probability that both copies of a tumor suppressor gene will be inactivated by proviral insertional mutagenesis in any individual cell. However, in most tumor systems, loss of tumor suppressor gene function occurs in a stepwise fashion and can involve different types of mutagenic events. Thus, a cell carrying a proviral insertion that inactivated one copy of a tumor suppressor gene could develop into a tumor cell if any inactivating mutation affected the second copy of the gene. The low frequency of such events may reflect the relatively low probability that a proviral insertion will lead to inactivation of a gene, an event that may require more precise targeting than gene activation. Indeed, attempts to inactivate expression of defined haploid target genes in vitro have revealed that such events occur much less frequently than expected and require strong selection (Varmus et al. 1981; Frankel et al. 1985; King et al. 1985).

Although tumor suppressor genes are not common targets of viral capture or insertional mutagenesis, mutation of the p53 gene occurs in several retrovirus-induced tumors. The first clues concerning the importance of p53 in these systems came from studies of the late-stage erythroleukemias induced by Friend virus (FV) (for review, see Ben-David and Bernstein 1991; Johnson and Benchimol 1992; Howard et al. 1993). This disease develops in FV-infected mice that survive the early erythroproliferative phase of the disease mediated by SFFV (see above Tumor Induction by Simple C-type Retroviruses That Lack Oncogenes) and in neonatal mice infected with Fr-MLV (Troxler and Scolnick 1978; MacDonald et al. 1980). The tumors contain proviral integrations in the vicinity of the spi-1 or fli-1 genes (Moreau-Gachelin et al. 1988; Ben-David et al. 1990a), genes that encode transcription factors belonging to the ets family (Klemsz et al. 1990; Ben-David et al. 1991), and mutations affecting p53 (Mowat et al. 1985; Chow et al. 1987; Rovinski et al. 1987). In many cases, large deletions alter one copy of the p53-coding sequence and the second allele is lost (Ben-David et al. 1988; Munroe et al. 1988). In rare instances, a provirus has integrated into the p53 gene (Table 3) (Ben-David et al. 1988; Hicks and Mowat 1988). Point mutations can result in production of a p53 protein that interferes with the function of wild-type p53 and stimulates growth (Munroe et al. 1990). Consistent with the important role of p53, erythroleukemia development is accelerated in transgenic mice carrying a mutant p53 gene (Lavigueur and Bernstein 1991).

p53 mutations play a part in several other retrovirus-based tumor induction systems. The gene is altered in some Cas-Br-E MLV-induced lymphomas (Bergeron et al. 1993), and studies using mice lacking a functional p53 gene have shown that loss of p53 function shortens the latent period required for Mo-MLV to induce thymic tumors (Baxter et al. 1996). In a similar vein, transgenic mice lacking p53 but expressing a wnt-1 transgene develop mammary tumors after a shorter latent period than animals carrying the transgene and a functional p53 gene (Donehower et al. 1995). However, transgenic mice lacking a functional p53 and expressing a myc gene under the control of an MMTV LTR develop accelerated thymic lymphomas but do not develop mammary tumors more rapidly than control mice (Elson et al. 1995). These data, coupled with the observation that the tumors in all of these mice are clonal, indicate that the combination of proviral insertion and p53 mutation is not sufficient for tumor development in many instances.

The NF1 gene is the only other known tumor suppressor gene that has been inactivated by proviral insertion. This gene, which encodes neurofibromin, a member of the GAP family (see above Tumor Induction by Viruses Containing the v-onc Gene), is targeted in some myeloid leukemias arising in BXH-2 mice (Table 3) (Buchberg et al. 1990; Viskochil et al. 1990; Cawthon et al. 1991). Integration events within a large NF1 intron lead to the expression of a truncated mRNA that does not appear to encode a full-length protein (Cho et al. 1995; Largaespada et al. 1995). In some instances, the second copy of NF1 has also been targeted; in others, the absence of neurofibromin suggests that mutation has altered this copy. Although the tumor cells lack neurofibromin, the levels of Ras-GTP in the tumor cells are similar to those in other myeloid tumor cells that do not contain NF1 insertions and express neurofibromin (Largaespada et al. 1995). Thus, other mechanisms must regulate this aspect of Ras metabolism in the tumor cells, and the potential oncogenic effects of the insertion seem to be mediated in another fashion. Interestingly, loss of NF1 occurs in myeloid leukemias that develop in aged mice which carry a single copy of the gene (Jacks et al. 1994), suggesting that alterations of this gene may occur in other myeloid tumors induced by retroviruses.

The possibility that tumor suppressor gene mutations occur in tumors induced by retroviruses containing v-onc genes has received limited attention. When hosts are infected with many of these viruses, the animals succumb rapidly. For example, RSV-induced fibrosarcomas become very large in a short time (7 to 10 days) because many infected cells proliferate and contribute to the tumor mass (for review, see Payne 1992). Similarly, massive proliferation of erythroid precursors leads to decreased erythrocyte levels and function in AEV-infected birds. Infection with AMV stimulates large amounts of myeloid cell proliferation and leads to similar problems in a very short time, with the blood from infected birds containing as much as 75% myeloblasts within 2 weeks of infection (Fig. 13) (for review, see Payne 1992). In all of these instances, the rapid course of the diseases reflects the inability of the host to cope with the proliferative effects induced by the virus. Thus, the cells involved need not be truly malignant and have the capacity to metastasize in order to kill the host.

Figure 13. Extensive myeloid proliferation occurs in AMV-infected chickens.

Figure 13

Extensive myeloid proliferation occurs in AMV-infected chickens. (A) Peripheral blood from a normal chicken. Note the predominance of nucleated red blood cells. (B) Peripheral blood from a chicken with AMV-induced myeloblastosis, (more...)

Despite the examples just cited, many experiments demonstrate that the tumor induction by some v-onc-containing retroviruses is a multistep process which may involve tumor suppressor gene mutation. For example, despite a short 3–4-week latent period, the pre-B-cell lymphomas induced by Ab-MLV are clonal or oligoclonal and arise from only a small subset of the pre-B cells that are infected (Green et al. 1987b, 1989). In addition, only some clones of hematopoietic cells transformed in vitro with Ab-MLV, AEV, AMV, and MC29 viruses develop into immortal cell lines (Whitlock and Witte 1981; Beug et al. 1982a,b; Whitlock et al. 1983). A loss of p53 expression has been correlated with immortalization of the avian cells (Ulrich et al. 1992). However, very few studies have examined the integrity of other tumor suppressor genes in tumors or cells transformed in vitro. Given their central role in other types of tumors, it seems likely that spontaneous mutation of tumor suppressor genes may be important in these and other retrovirus-induced tumors. Future experimentation addressing this possibility could shed new light on the mechanisms by which retrovirus-induced tumors progress and strengthen the theoretical ties between these malignancies and those induced by other means.

Target Cell Specificity

Most oncogenic retroviruses are capable of infecting a variety of cell types in vivo, yet the majority preferentially induce a particular type of tumor (see Tables 1 and 2). This phenomenon is often referred to as target cell specificity and is most striking among viruses that induce hematopoietic cell tumors. Usually, a single hematopoietic lineage is affected and the majority of tumor cells appear to be arrested at a particular stage of differentiation. This feature characterizes tumors induced by retroviruses containing v-onc genes, tumors induced by insertional mutagenesis, and many spontaneous tumors. In all of these instances, altered expression of proto-oncogenes causes the tightly regulated growth and differentiation pattern of the cells to be disrupted. Unraveling the mechanisms by which this occurs has enhanced our understanding of the mechanisms involved in retroviral tumorigenesis. In addition, these investigations continue to reveal insights into the ways in which oncogenes affect growth and differentiation and shed light on the mechanisms by which cells normally regulate these processes.

Interactions of Viruses Containing v- onc Genes with Their Target Cells

Target cell specificity has been studied most effectively using hematopoietic cells and viruses that contain v-onc genes (for review, see Graf and Beug 1978; Pierce 1989). Most of these viruses transform a restricted subset of cells from a single hematopoietic lineage in vitro, and the transformants are usually phenotypically similar to the tumor cells that arise in infected hosts. For example, infection of chicken bone marrow with AEV induces transformation of BFU-E, a specific type of erythroid precursor, and the immature erythroid transformants that arise resemble the erythroblastic tumor cells observed in vivo. Under similar conditions, AMV and MC29 transform cells of the myeloid lineage, giving rise to immature myeloid cells and macrophages similar to those found in tumors induced by these agents (Beug et al. 1979; Gazzolo et al. 1979, 1980; Graf et al. 1980, 1981). Many other viruses, including S13-ASV, E-26-AMV, Ab-MLV, Ki-MSV, Ha-MSV, and C58-MSV display similar specificity (for review, see Graf and Beug 1978; Pierce 1989).

The precise structure of the v-onc-encoded protein can have important effects on target cell specificity. E26-AMV-infected chickens develop erythroblastosis and myeloblastosis (Sotirov 1981; Radke et al. 1982; Moscovici et al. 1983), and the virus gives rise to erythroid and myeloid transformants in vitro (Radke et al. 1982; Moscovici et al. 1983). Dissection of this system reveals that the “erythroid” cells are really multipotential progenitors that can give rise to several types of myeloid and erythroid cells and thrombocytes (Graf et al. 1992; Kraut et al. 1994; Frampton et al. 1995). The v-ets portion of the molecule has a key role in maintaining the undifferentiated phenotype of the progenitors (Kraut et al. 1994). Although transformation of progenitors requires expression of both the v-ets and v-myb oncogenes as a fusion protein, domains of each oncoprotein affect the ability of the cells to differentiate along particular pathways (Metz and Graf 1991a,b; Domenget et al. 1992; Frampton et al. 1993, 1995).

The v-onc-encoded protein is the major viral determinant of target cell specificity. Intimately linked to this specificity are the similarities between the phenotype of hematopoietic transformants and the immature target cells from which they originated. This relationship suggests that v-onc gene expression interferes with the normal differentiation of the cells. Studies examining the interaction of temperature-sensitive transformation mutants with hematopoietic cells have confirmed this idea (Graf et al. 1978; Moscovici and Moscovici 1983; Beug et al. 1984; Palmieri 1986; von Weizsacker et al. 1986; Frykberg et al. 1988; Golay et al. 1988; Knight et al. 1988; Chen et al. 1994). These temperature-sensitive viruses express an active v-onc-encoded protein and transform cells at the permissive temperature; at the nonpermissive temperature, this protein is inactive. Transformed hematopoietic cells isolated at the permissive temperature stop growing and acquire the properties of more differentiated cells when incubated at the nonpermissive temperature (Fig. 14).

Figure 14. Transformation arrests the differentiation of many hematopoietic cell types.

Figure 14

Transformation arrests the differentiation of many hematopoietic cell types. (A) Morphology of erythroid cells transformed by a temperature-sensitive mutant of AEV at the permissive (left) and nonpermissive (right) temperatures. (more...)

The temperature-sensitive viruses and in vitro hematopoietic transformation systems have been particularly useful for probing the mechanisms by which v-onc genes alter differentiation programs. Myeloid cells transformed with E-26 virus provided the starting point for subtractive cDNA cloning experiments which identified a v-Myb-regulated gene called mim-1 (Ness et al. 1989). This gene is expressed in transformed myeloid cells at the permissive temperature when the v-Myb protein is active but not at the nonpermissive temperature when v-Myb is inactive (Ness et al. 1989). Although the function of the Mim-1 protein is not known, v-Myb and c-Myb stimulate expression of mim1 only in the presence of a second, myeloid-cell-specific DNA-binding factor called NF-M that is related to the C/EBP family (Ness et al. 1993). This observation raises the intriguing possibility that the myeloid target cell specificity of viruses expressing v-myb reflects the restricted expression of NF-M. In such a model, expression of NF-M, in concert with the loss of negative regulatory sequences in the v-Myb protein, would lead to altered expression of some of the genes important for normal myeloid cell growth and differentiation. The observation that a point mutation which distinguishes the v-myb gene found in AMV from that found in E-26 affects both the ability of v-Myb to activate mim1 and the differentiation phenotype of the transformed myeloid cells is consistent with this hypothesis (Introna et al. 1990; Ness et al. 1993).

The complexities involved in understanding the mechanisms by which expression of a v-onc gene can alter both growth and differentiation are well illustrated by studies examining the genes affected by expression of the v-ErbA protein encoded by AEV. Using temperature-sensitive viruses, three genes that regulate the differentiation of erythroid precursors and are suppressed by an active v-ErbA protein have been identified (Zenke et al. 1988, 1990; Pain et al. 1990; Schroeder et al. 1990). Two of these genes, those encoding carbonic anhydrase II (CAII) and the erythrocyte anion transport protein (band-3 protein), influence the growth conditions required to sustain the transformed erythroid cells and modify the transformation response (Fuerstenberg et al. 1992). Indeed, expression of these genes in erythroid cells transformed by AEV-H, which expresses only v-erbB, is probably responsible for the stringent growth requirements of these cells (Beug et al. 1982b; Samarut and Gazzolo 1982). Neither CAII nor band-3 protein has significant effects on v-ErbA-mediated suppression of differentiation. This feature may be modulated by δ-aminolevulinate synthetase, the third known protein regulated by v-ErbA (Zenke et al. 1988), or other as yet unidentified molecules that are affected by v-ErbA.

Other mechanisms may also contribute to the effects of v-ErbA on transformation. The protein influences expression of the AP-1 transcription factor complex by relieving repression normally induced by the retinoic acid receptor and c-ErbA (Desbois et al. 1991). Such a change could stimulate cell growth because AP-1 is involved in the immediate response of cells to many growth stimuli (for review, see Karin 1995; Karin and Hunter 1995). A second mechanism focuses on the possible role of the normal estrogen receptor. This molecule, another member of the receptor family to which v-ErbA belongs, can synergize with c-ErbB protein to stimulate growth of erythroid precursors (Hayman et al. 1993; Schroeder et al. 1993). The heavily mutated v-ErbA protein may perform a similar stimulatory function in erythroid precursors and promote the growth of infected erythroid cells, even though v-erbA alone is not sufficient to transform these cells (Frykberg et al. 1983; Beug et al. 1994).

Experiments with temperature-sensitive viruses have all emphasized the role of v-onc gene products in arresting differentiation. However, v-onc gene expression can alter differentiation programs in other ways. AMV can transform both immature myeloid cells and mature macrophages (Gazzolo et al. 1979; Durban and Boettiger 1981; Graf et al. 1981; Moscovici and Gazzolo 1982). In both cases, the transformants resemble immature myeloid cells because the v-Myb protein causes the mature cells to lose their differentiated phenotype (Beug et al. 1984, 1987). Ab-MLV, a virus that normally arrests differentiation of pre-B cells (Chen et al. 1994), can stimulate erythropoietin-independent differentiation of erythroid cells (Waneck and Rosenberg 1981). Presumably, this reflects the ability of the v-Abl protein to mimic signals normally transmitted via the Epo receptor in a situation where the oncoprotein is not able to stimulate continued growth.

Target Cell Selection and Proviral Insertional Mutagenesis

Most retroviruses that lack oncogenes infect a wide range of different cell types. However, only cells of a particular type ever develop into a tumor. The viral determinants most solidly linked to this phenomenon map to the U3 region of the LTR. The role of these sequences is well illustrated by experiments in which the LTR of Fr-MLV, an erythroleukemia-inducing virus, was swapped with that of Mo-MLV, a closely related virus that induces thymic lymphomas. Chimeras containing the LTR of Fr-MLV induced erythroleukemia and those containing the LTR of Mo-MLV induced thymic tumors (Chatis et al. 1983, 1984; Golemis et al. 1989). Similar experiments have revealed that determinants within the LTRs of amphotropic MLV and Cas-Br-E MLV influence targeting of these viruses to their respective cell types (Jolicoeur and DesGroseillers 1985). These results are thought to reflect the ability of the different LTRs to direct gene expression in particular tissues (Li et al. 1987; Boral et al. 1989) and could affect both viral replication and expression of genes targeted by proviral insertional mutagenesis. A second set of sequences, mapping to the ψ-gag and PR-encoding region of pol in combination with sequences in env, influence myeloid tumor induction by Mo-MLV and Fr-MLV in pristane-primed mice, perhaps by influencing splicing in myeloid cells (Mukhopadhyaya et al. 1994). Whether this mechanism has a role in other tumor systems deserves further study.

The structural relationship between inserted proviruses and the cellular sequences they activate may influence the types of tumors that develop. Even though proviral insertional mutagenesis is not precise, particular patterns are characteristic of different host-virus combinations, suggesting that only some provirus insertion site combinations foster tumor development. For example, most ALV insertions at c-myc are upstream of the coding sequences in the same transcriptional orientation (Fung et al. 1981, 1982a; Neel et al. 1981; Payne et al. 1981; Westaway et al. 1984; Robinson and Gagnon 1986; Goodenow and Hayward 1987; Swift et al. 1987). Insertions near c-myc in Mo-MLV- and FeLV-induced thymomas also tend to cluster upstream of the gene, but in the opposite orientation (Corcoran et al. 1984; Cuypers et al. 1984; Li et al. 1984; Selten et al. 1984; Reicin et al. 1986; Forrest et al. 1987; Miura et al. 1989). These differences may reflect unique properties of the viruses. Alternatively, they may reflect mechanisms controlling c-myc expression in the various tissues and hosts. Integration sites identified by selection for expression of enhancerless or promoterless vectors tend to be located upstream of transcriptionally active genes (von Melchner et al. 1990; Chang et al. 1993; Sablitzky et al. 1993) as do some Mo-MLV integration sites identified in fibroblast cells (Mooslehner et al. 1990; Scherdin et al. 1990). Other evidence suggests that DNase-I-hypersensitive sites influence integration site selection (Vijaya et al. 1986; Rohdewohld et al. 1987).

The age and genetic background of the host as well as the route of inoculation influence the type of tumor that develops. ALV infection induces bursal lymphomas that have integrations into the c-myc locus in most strains of chicken. However, infection of line 151 chickens induces erythroblastosis and is characterized by ALV integrations into the c-erbB locus (for review, see Maihle and Kung 1988; see above Tumor Induction by Simple C-type Retroviruses That Lack Oncogenes). This difference is controlled by an unknown dominant host gene that segregates independently of the c-erbB locus (Robinson et al. 1985). Chickens inoculated with ALV at hatch develop bursal lymphomas containing ALV proviruses integrated into the c-myc locus; chickens infected prior to hatching also develop bursal lymphomas. However, in this case, integrations in the c-myb locus are found (Pizer and Humphries 1989). Different loci are preferentially targeted by MMTV in different strains of mice and in multiparous and virgin mice (Escot et al. 1986; Pathak et al. 1987; Marchetti et al. 1991; Schwartz et al. 1992; Clausse et al. 1993a). These influences of genetic background and target tissue physiology probably reflect the frequency and susceptibility of target cells to infection rather than features that affect the actual integration process.


Human T-cell leukemia virus types 1 and 2 (HTLV-1 and HTLV-2) and bovine leukemia virus (BLV) are complex retroviruses that comprise a distinct genus (Coffin 1996). These agents lack oncogenes and contain gag, pol, and env genes; the 3′region of the genome contains two regulatory genes, tax and rex, that control expression of the virus. Both genes are required for viral replication (I.S.Y. Chen et al. 1985). Tax mediates trans-activation of the viral LTR and Rex influences RNA processing (for review, see Green and Chen 1994; Kettmann et al. 1994; Cann and Chen 1996; see Chapter 6. HTLV and BLV are exogenous viruses associated with naturally occurring infections in humans and cows, respectively. However, unlike other oncogenic retroviruses, only a fraction of individuals infected with these viruses develop tumors.

Malignancies Induced by HTLV and BLV

HTLV-1 was identified in 1980 in a cell line derived from a patient diagnosed with cutaneous T-cell lymphoma (Poiesz et al. 1980). Other reports soon linked HTLV-1 infection to a newly recognized T-cell lymphoma called adult T-cell leukemia (ATL) (Hinuma et al. 1981, 1982; Blattner et al. 1982; Catovsky et al. 1982; Robert-Guroff et al. 1982; Yoshida et al. 1982) and to a central nervous system (CNS) disorder called tropical spastic paraparesis or HTLV-1-associated myelopathy (TSP/HAM) (Gessain et al. 1985; Osame et al. 1986b; Rodgers-Johnson et al. 1987; Vernant et al. 1987; Bhagavati et al. 1988; Jacobson et al. 1988a,b; see below Neurological Diseases, Inflammatory CNS Disease Induced by HTLV-1). HTLV-2 was first identified in a T-cell line derived from a patient with hairy cell leukemia (Kalyanaraman et al. 1982). Subsequent studies have provided some evidence for a link between HTLV-2 infection and a rare form of this leukemia involving T cells (Rosenblatt et al. 1986, 1988; Sohn et al. 1986). However, a clear association between HTLV-2 infection and any disease has yet to be established.

ATL arises after a long latent period and occurs in about 1% of HTLV-1-infected individuals (Kondo et al. 1987; Murphy et al. 1989). Often, 20–30 years pass between HTLV infection and the onset of disease (Kawano et al. 1985). Several different patterns of symptoms characterize the disorder. Some patients develop a pre-ATL syndrome characterized by leukocytosis (elevated number of circulating white blood cells) prior to the onset of symptoms. These individuals have abnormal lymphocytes that arise from one or a few clones of HTLV-infected cells (Gallart et al. 1983; Kinoshita et al. 1985). About 50% of these patients undergo regression. Other patients develop a chronic or smoldering ATL characterized by low levels of circulating lymphoid cells and skin lesions caused by infiltration of leukemic cells (Fig. 15) (Shimoyama et al. 1983; Yamaguchi et al. 1983; Kawano et al. 1985). Some of these patients and some with pre-ATL progress to acute ATL.

Figure 15. Skin lesions are a common feature of HTLV-1-induced ATL.

Figure 15

Skin lesions are a common feature of HTLV-1-induced ATL. A patient with skin lesions typically associated with HTLV-1-induced ATL is shown. (Reprinted, with permission, from Cann and Chen 1996.)

Acute ATL is characterized by greatly elevated numbers of circulating malignant T cells that usually express CD4 (Hattori et al. 1981; Hanaoka et al. 1982; Yamada 1983; Sugamura et al. 1984; Morimoto et al. 1985). These malignant cells arise from one or a few clones that contain small numbers of integrated HTLV proviruses (Hoshino et al. 1983b; Seiki et al. 1984; Yoshida et al. 1984a,b; Jarrett et al. 1986; Matsuoka et al. 1988). Patients display skin lesions and have enlarged lymph nodes, liver, and spleen (Bunn et al. 1983; Shimoyama et al. 1983; Kawano et al. 1985; Shimoyama 1991). Lysis of bone tissue is another characteristic feature and leads to hypercalcemia (increased levels of calcium in the circulation) (Catovsky et al. 1982; Hanaoka et al. 1982; Blayney et al. 1983). This aspect of the disease seems to reflect the production of a parathyroid-hormone-related protein by the leukemic cells (Ejima et al. 1993; Ikeda et al. 1993a,b; Watanabe et al. 1993). Other patients develop clonal T-cell lymphomas that contain integrated HTLV (Yamaguchi et al. 1983, 1986; Yoshida et al. 1984b; Cappell and Chow 1987; Shimoyama 1991). Patients with acute ATL have an average life expectancy of about 6 months; those with HTLV-1-associated lymphomas have a life expectancy of approximately 10 months (Jaffe et al. 1984; Kawano et al. 1985; Shimoyama 1991).

Bovine leukosis was described in the previous century (for review, see Kettmann et al. 1994). However, BLV was not isolated and characterized until much later (Miller et al. 1969; Dutta et al. 1970; Olson et al. 1970; Mammerickx 1972; Kettmann et al. 1976; Sagata et al. 1985). Although BLV can cause a fatal B-cell malignancy, the majority of BLV-infected cows remain asymptomatic. About one third of the infected cows develop a protracted chronic proliferative disorder called persistent lymphocytosis (PL) in which the numbers of circulating B lymphocytes increase dramatically (Fossum et al. 1988; Matheise et al. 1992; Meirom et al. 1993). This phase of the disease is associated with decreased milk and fat production in some cows (Da et al. 1993). Less than 5% of infected cows develop clonal B-cell lymphomas which contain integrated BLV (Kettmann et al. 1980, 1982). Although sheep are rarely infected under natural circumstances, they can be infected experimentally (Olson et al. 1972; Van Der Maaten and Miller 1976). Most of these animals develop persistent lymphocytosis and progress to lymphoma within 1–4 years after infection, making this a more tractable model for studies addressing pathogenic mechanisms (Djilali et al. 1987; Mammerickx et al. 1988; Lagarias and Radke 1989; Murakami et al. 1994).

Mechanisms of Tumor Induction

Unlike simple C- and B-type retroviruses which induce tumors by expressing viral products or by proviral insertional mutagenesis, HTLV and BLV induce tumors using other, poorly understood, mechanisms. A viral gene product does not seem to be required for maintenance of the transformed state because freshly isolated tumor cells usually lack evidence of viral expression (Kettmann et al. 1980, 1982; Franchini et al. 1982; Seiki et al. 1982; Yoshida et al. 1982; Reitz et al. 1983; Burny et al. 1988; Van Den Broeke et al. 1988; Kinoshita et al. 1989; Lagarias and Radke 1989; Jensen et al. 1991; Korber et al. 1991). Indeed, highly deleted proviruses that have lost the ability to produce most viral proteins are often found in the tumor cells (Kettmann et al. 1982, 1985; Yoshida et al. 1984a; Van Den Broeke et al. 1988; Tanaka et al. 1990; Korber et al. 1991). In addition, there is very little evidence of common integration sites in the tumors (Kettmann et al. 1982, 1983; Gregoire et al. 1984; Seiki et al. 1984). One ATL cell line expresses an RNA containing sequences from the IL-15 gene fused to sequences from the HTLV R region, a situation that facilitates increased expression of the cytokine (Bamford et al. 1996). However, similar events do not appear to be common in vivo, and often, the cell lines that are derived from tumors may not be related to the dominant malignant clone present in vivo (Hoshino et al. 1983b).

Small animal models and in vitro transformation systems, approaches that have provided important insights into the pathogenesis of many oncogenic retroviruses, have been less useful for studies of HTLV and BLV. Rabbits, newborn rats, and severe combined immunodeficient disease (SCID) mice can be used to study HTLV pathogenesis, but the disease induced in any of these settings does not mimic all of the events involved in ATL (Uemura et al. 1987; Cockerell et al. 1991; Suga et al. 1991; Oka et al. 1992; Feuer et al. 1993b). Although there is no in vitro transformation system for BLV, HTLV can immortalize T cells in vitro (Chosa et al. 1982; Yamamoto et al. 1982; Chen et al. 1983; Miyoshi et al. 1983; Popovic et al. 1983). The immortalized cells grow continuously in the absence of IL-2 and, like tumor cells, they usually express CD4. However, transformation occurs at a low frequency and involves a long polyclonal proliferative phase, complicating analyses of the mechanisms involved. In addition, HTLV is largely cell-associated and infection usually involves mixing irradiated, HTLV-producing cells with peripheral blood cells. Thus, studies of different mutants, particularly those in which viral replication is compromised, are difficult. In addition, the in-vitro-transformed cells share only some features with tumor cells. For example, the in vitro transformants produce HTLV. In addition, both HTLV-1 and HTLV-2 induce transformation, even though HTLV-2 is not usually associated with malignant disease in vivo. Thus, direct parallels are not easily drawn between the oncogenic mechanisms involved in vitro and in vivo.

Considerable attention has focused on the possible role of the Tax trans-activator in transformation. Although Tax does not contact DNA directly, the protein affects a variety of cellular transcription factors, including NF-κB (Leung and Nabel 1988; Nyborg et al. 1988, 1990; Ruben et al. 1988). These interactions modulate viral expression and replication by activating the LTR (Lee et al. 1984; Cann et al. 1985; Felber et al. 1985; Fujisawa et al. 1985; Sodroski et al. 1985; Seiki et al. 1986). In addition, they activate expression of a wide range of cellular genes, including many encoding products that stimulate T-cell growth (Inoue et al. 1986; Cross et al. 1987; Maruyama et al. 1987; Siekevitz et al. 1987; Fujii et al. 1988, 1991, 1992; Leung and Nabel 1988; Ruben et al. 1988; Wano et al. 1988; Nimer et al. 1989; Ratner 1989; Kelly et al. 1992). Among these are the genes encoding IL-2 and the IL-2 receptor, proteins that have a key role in regulating T-cell growth. Even though IL-2 is not expressed by all tumor cells (Arya et al. 1984; Farcet et al. 1991; Kimata and Ratner 1991), expression of this cytokine and its receptor could establish an autocrine loop that might facilitate transformation in some instances.

Additional evidence supporting a role for Tax in transformation comes from studies showing that Tax induces transformation of immortalized rodent cells such as NIH-3T3 and Rat-1 cells (Tanaka et al. 1990; Yamaoka et al. 1992). Expression of Tax in a Herpesvirus saimiri vector stimulates growth of T cells that resemble cells isolated from ATL patients in phenotype, but these cells remain dependent on IL-2 for growth (Grassmann et al. 1989, 1992). The vector used in these experiments retains a large amount of H. saimiri sequence, and some of the effects could reflect interactions of Tax with vector sequences. Mice expressing Tax from the HTLV LTR develop mesenchymal tumors 3–4 months after birth (Hinrichs et al. 1987; Nerenberg et al. 1987). Other transgenic mice in which Tax is expressed from the granzyme B promoter develop a T-cell lymphoma characterized as a large granular lymphocytic leukemia after 6 or more months of age (Grossman et al. 1995). The cells involved share properties with natural killer cells, a non-MHC-restricted cytotoxic T cell. Although this disorder is distinct from ATL, the use of the granzyme B promoter to target expression to mature T cells does provide a way to test the effect of various tax mutations on a malignant process involving lymphoid cells.

Although Tax is likely to be important for transformation by HTLV and, by analogy, BLV, all of the data from in vitro studies and from transgenic mice show that expression of Tax alone is not sufficient for transformation. The tax gene is retained by most tumors, but very little Tax is expressed and tax transcripts are difficult to detect in fresh tumor cells even when PCR is used (Franchini et al. 1982; Gotoh et al. 1982; Reitz et al. 1983; Seiki et al. 1984; Yoshida et al. 1985; Jensen et al. 1991; Korber et al. 1991). These data suggest that Tax does not have a key role in maintaining the transformed phenotype of the tumor cells. Instead, Tax may have an important role early in the disease process by stimulating abnormal replication of infected T cells through its effects on the IL-2 gene and other genes involved in growth regulation. Other abnormalities, including chromosomal changes (Fukuhara et al. 1983; Miyamoto et al. 1983; Sanada et al. 1985; Whang-Peng et al. 1985; Itoyama et al. 1990), may be responsible for continued cell growth. Accumulation of these mutations may be facilitated by HTLV-stimulated growth of cells during early phases of the disease.

Other Retroviruses Associated with Malignant Disease

Ovine Pulmonary Carcinoma

Jaagsiekte is an ovine infectious pulmonary adenomatosis that resembles bronchioaveolar adenocarcinoma (for review, see Sharp 1987; DeMartini et al. 1988). A virus related to B- and D-type retroviruses called jaagsiekte retrovirus (JSRV) is found in the epithelial tumor cells (York et al. 1991, 1992; Hecht et al. 1994; Palmarini et al. 1995, 1996). Unfortunately, the virus cannot be grown in vitro, so the pathogenic potential of JSRV has not been tested. Many infected sheep and goats also contain ovine lentiviruses, including VMV, an agent that causes an inflammatory disease affecting the lungs and central nervous system (Payne et al. 1986; Querat et al. 1987; Rosadio et al. 1988; Dawson et al. 1990; see below Other Retrovirus-induced Diseases, Lentiviral Infections of Sheep and Goats). However, none of these viruses appear to induce the tumor (Lairmore et al. 1986; Payne et al. 1986; DeMartini et al. 1988). In addition, some sheep develop jaagsiekte in the apparent absence of lentiviral infection (York et al. 1991). Probably, coinfection of animals with both JSRV and the lentiviruses is very common (Dawson et al. 1990; Gonzalez et al. 1993) just as coinfection with FeLV and FIV occurs frequently in nature (Grindem et al. 1989; Ishida et al. 1989; Pedersen et al. 1990; see below Retrovirus-induced Immunodeficiencies, FIV-induced Immunodeficiency).

Very little is known about the pathogenesis of jaagsiekte. Even though JSRV is specifically associated with tumor tissue (Palmarini et al. 1995, 1996), infected sheep do not appear to contain new proviruses that react with a JSRV probe (Hecht et al. 1994). This feature and the rapid course of the disease (Verwoerd et al. 1980; Sharp et al. 1992) may indicate that the tumors are polyclonal. Indeed, JSRV stocks could even contain an oncogenic replication-defective retrovirus that uses JSRV as a helper. This possibility is difficult to exclude without an in vitro culture system. Further studies are important because this disorder may provide a model for human bronchioalveolar cell carcinoma, a lung tumor that is only weakly associated with cigarette smoking and is occurring with increased incidence (Perk and Hod 1982; Greco et al. 1986; Gazdar and Linnoila 1988; El-Torky et al. 1990; Barsky et al. 1994).

Oncogenic Retroviruses of Bony Fish

Tumors, ranging in type from sarcomas to papillomas and hematologic malignancies, have been found in more than ten species of fish (for review, see Poulet et al. 1994). In several cases, outbreaks of these disorders in aquaculture facilities have supported an infectious etiology. Evidence that retroviruses are involved in these tumors comes most commonly from electron microscopic studies which reveal the presence of C-type particles. Disease induction by transmission of cell-free filtrates has been reported for only a subset of these, including walleye dermal sarcoma (Martineau et al. 1990), chinook salmon plasmacytoid leukemia (Kent and Dawe 1993), and angelfish lip fibroma (Francis-Floyd et al. 1993). In some cases, the tumors appear and regress according to the season (Bowser et al. 1988; Bowser and Wooster 1991; Sonstegard 1976). The factors that control this response probably include seasonal hormonal changes, and changes in water temperature that may affect the immune response and viral replication (Fig. 16). Analysis of a molecular clone recovered from a walleye dermal sarcoma reveals that the agent is a complex retrovirus and shares features with spumaviruses, C-type murine retroviruses, and lentiviruses (Martineau et al. 1992; Holzschu et al. 1995) (see Chapter 2. This picture suggests that the virus probably defines a new retroviral genus. The lack of an in vitro system has made it difficult to test the oncogenic potential of this and other piscine retroviruses.

Figure 16. Dermal sarcomas in walleye show a seasonal pattern.

Figure 16

Dermal sarcomas in walleye show a seasonal pattern. (Top) Walleye caught in spring have nodular sarcomas that can be easily dislodged; (bottom) walleye in the fall have firm lesions that cannot be easily dislodged. (Photographs (more...)

Other Retrovirus-associated Tumors

Retroviruses have been associated with tumors in several other animals. Morphological analyses have demonstrated the presence of retroviral particles in a number of ophidian tumors (Zeigel and Clark 1969; Lunger et al. 1974; Jacobson et al. 1980), and two different viruses, Russell's viper virus and corn snake retrovirus, have been isolated from snakes with tumors (Ziegel and Clark 1971; Clark et al. 1979). The viper virus is immunologically related to primate type-D retroviruses (Andersen et al. 1979). Neither of these agents has been shown to induce malignant disease, and their presence in the tumor cells may reflect chance expression of the viruses in the tumor or the selective replicative advantage provided by cycling tumor cells. A variety of tumors affect different types of molluscs including clams, oysters, and mussels (for review, see Brown et al. 1978). Type-B retroviral particles have been isolated from tumors in clams, and cell-free extracts of these tumors are oncogenic in clams (Oprandy et al. 1981). However, very little is known about the way in which these viruses alter cell growth.

Copyright © 1997, Cold Spring Harbor Laboratory Press.
Bookshelf ID: NBK19409