Transcription (original) (raw)

Basic Retroviral Transcription Events

The Template

In the retroviral life cycle, proviral DNA is the template for transcription. Thus, the formation of an integrated DNA template is a vital step in the replication of retroviruses. The integration of the DNA template is a strategy unique to retroviruses. In other viruses (e.g., hepadnaviruses) that replicate with both reverse transcription and transcription steps, unintegrated DNA serves as a template for transcription. It is not clear why integration precedes transcription in retroviruses. It is possible that the structure of the preintegration particle or efficient targeting of the viral DNA to the host chromosome prevents the assembly of transcriptional machinery until integration has occurred.

The Enzyme

Retroviral transcription is mediated by the host-cell RNA polymerase II, which synthesizes cellular mRNAs and some small nuclear RNAs (snRNAs). In a variety of assay systems, retroviral transcription is sensitive to α-amanitin, which inhibits the function of this polymerase. Additional evidence for the role of RNA polymerase II comes from the characterization of the _cis_-acting machinery encoded by the retroviral genome. As discussed in detail below, these elements are characteristic of cellular sequences that direct RNA polymerase II in the production of mRNAs.

The Transcript

The full-length viral transcript, which is packaged as the viral genome, contains a unique copy of all of the information encoded in the proviral DNA, plus a short direct repeat at each end termed R (Fig. 1). The R region is defined by the transcription start site in the 5′ LTR and the location of a 3′-end processing event in the 3′ LTR. Although the size of the R region varies, the genomic RNA of each retrovirus has a characteristic organization because there is precise control of both the start site positioning and the 3′-end processing. Cellular machinery caps the 5′ end of the viral transcript with m7G5′ppp5′Gmp. This common cellular modification is necessary for proper function of mRNAs in protein synthesis, but its role in viral replication is unknown. It is necessary for the retroviral transcript to resemble cellular transcripts. The first nucleotide (Gm) after the cap group is templated and represents the transcription start site within the proviral DNA. The proviral sequences contain no obvious control site for the termination of transcription, and viral transcription into flanking host sequences is well established. If the viral sequences within these extended transcripts are not processed and polyadenylated, translation and packaging into infectious virions can occur. However, most transcripts in infected cells are processed precisely to generate a 3′ end that is subsequently polyadenylated.

RNA Levels

The efficiency with which transcription of integrated DNA is initiated is a key step in determining the steady-state levels of RNA in an infected cell. Each provirus contains a set of _cis_-acting control signals that determine the potential rate of transcriptional initiation. However, the cell type, the differentiation state of the cell, and the physiological state of the cell with respect to extracellular signals result in considerable variation in the availability and activity of the transcriptional machinery. These variables generate a remarkably wide range of viral gene expression levels. For example, tissue culture cells actively producing infectious virions of avian or murine leukemia viruses (ASLV or MLV) contain such high levels of viral RNA that 5–10% of the mRNA in a cell can be of viral origin (Fan and Baltimore 1973; Lee et al. 1979). As described below, in complex retroviruses, the levels of viral transcripts are regulated by both cellular and viral processes. Thus, RNA production in HIV infection is extremely variable. Early in an infectious cycle, relatively low levels of RNA are detected, whereas late in the cycle, Tat mediates the production of much higher levels of RNA (Kim et al. 1989).

Overview of Eukaryotic Transcriptional Control

Basal Machinery

In eukaryotic cells, the initiation of transcription by RNA polymerase II is directed by two types of _cis_-acting control elements: core and regulatory. Core control elements mediate the assembly of a multiprotein complex that consists of basal transcription factors, RNA polymerase, and other proteins found stably associated with polymerase (for review, see Koleske and Young 1995; Orphanides et al. 1996; Pugh 1996). For most promoters, including those of all retroviruses, a TATA box is the core promoter element. In some cases, an initiator, which resides at the transcription start site, also contributes to core promoter activity. The first step in complex assembly onto the promoter is the binding of TFIID to the TATA box. TFIID is itself a multiprotein complex that includes the TATA-binding protein (TBP) as the DNA-binding polypeptide and several TBP-associated factors (TAFs). This first step is followed by the binding of other basal factors and RNA polymerase (Fig. 2). This preinitiation complex is capable of directing basal transcription with the addition of ribonucleotide triphosphates. Initiation of transcription is accompanied by ATP hydrolysis and phosphorylation of the carboxy-terminal tail of the large subunit of RNA polymerase II. As the transcript is elongated, some portion of the initiation complex is retained for additional rounds of RNA synthesis. However, these subsequent steps are less well understood than the initial assembly process. Elongation also appears to be regulated in part by basal transcription factors TFIIE and TFIIF, as well as more specific elongation factors such as elongins.

Figure 2

Eukaryotic transcriptional control machinery for RNA polymerase II. Core transcription factors associate with DNA polymerase within the promoter. Regulatory transcription factors may be positioned within the distal enhancer or within the more proximal (more...)

Regulatory Machinery

Regulatory control elements modulate the rates of transcription executed by the basal machinery (for review, see Johnson and McKnight 1989; Tjian and Maniatis 1994). In some cases, these regulatory elements are located in the immediate vicinity of the basal promoter and are also termed promoter elements (Fig. 2). However, they frequently occur at considerable distance from the basal promoter, and at such distal locations, they are termed enhancers. In retroviral LTRs, the spacing between the transcription start site and the enhancer is limited to less than 1 kb (see Fig. 1); however, in cellular genes, this distance can exceed 10 kb. Insertional activation of cellular genes by retroviruses also can occur over considerable distances. It has been proposed that regulatory control elements can act at a distance because the DNA bends to allow distal sites to be brought into proximity with the basal promoter (Fig. 2).

Cellular and viral gene expression is modulated by extensive arrays of _cis_-acting elements. These regulatory elements are sequence-specific-binding sites (10– 20 bp in length) for proteins that function as transcriptional activators or repressors. Multiple regulatory proteins within an array function in a combinatorial manner (Johnson and McKnight 1989; Goodrich et al. 1996). For example, some factors bind DNA as homodimeric or heterodimeric complexes. Furthermore, coactivators and corepressors that cannot themselves bind DNA mediate transcriptional activity of a particular enhancer or promoter element by direct protein-protein interactions with the DNA-binding proteins. Some transcription factors appear to play solely an architectural part by mediating DNA bending necessary for the assembly of multiprotein enhancer complexes. Evidence for the combinatorial action of factors is also supported by the observation that multiple activators act synergistically to stimulate transcription.

Functional Domains of Transcription Factors

Transcription factors are modular in nature and possess independent domains for DNA binding, subunit associations, ligand binding, and transcriptional activation or repression. The mechanism of activation is an area of extensive investigation; however, few conclusions can be made at present. Distinct classes of activation domains have been identified, including those marked by a preponderance of acidic, proline, or glutamine residues. A number of models for activation have been proposed. Regulatory factors can both recruit and stabilize the binding of the initiation complex and/or the RNA polymerase. Activation also can involve regulation of promoter clearance and polymerase elongation. Accumulating evidence suggests that some activation involves the remodeling of chromatin structure (for review, see Felsenfeld 1992; Paranjape et al. 1994). A dynamic view of the interplay between chromatin structure and the transcriptional machinery has emerged with the discovery that some regulatory transcription factors can bind DNA that is assembled in nucleosomes. Furthermore, large complexes of proteins appear to function in remodeling of chromatin (Peterson and Tamkun 1995; Tsukiyama and Wu 1995). The role of chromatin has been studied with the LTRs of both mouse mammary tumor virus (MMTV) and HIV, as discussed below.

The best understood property of transcription factors is their ability to bind DNA in a sequence-specific manner. DNA-binding domains contain structural motifs that fit into the grooves of DNA for the recognition of functional groups on DNA base pairs. Transcription factors can be grouped into structural families that are defined by common DNA-binding motifs (Steitz 1990; Pabo and Sauer 1992). The existence of these structural families implies that related proteins can bind similar or even identical binding sites. Determining which members of a family function on a particular element is a challenge, and this point will be illustrated in the case of LTR-binding proteins. A consensus recognition sequence for a particular DNA-binding protein can be derived from the analysis of many binding sites. However, such studies have demonstrated that any one site can diverge significantly from the consensus. Therefore, DNA sequence elements for the function of a particular protein must be tested experimentally.

The glucocorticoid receptor, which regulates MMTV gene expression (as discussed below), serves to illustrate the functional domains of a typical transcription factor. It belongs to a superfamily of nuclear hormone receptors that includes receptors for steroid hormones, thyroid hormone, and retinoic acid (Evans 1988; Mangelsdorf et al. 1995). Figure 3 (top) illustrates the functional domains of the glucocorticoid receptor and two closely related hormone receptors. An activation domain has been mapped to the amino-terminal portion of the glucocorticoid receptor; the hormone-binding domain is located in the carboxy-terminal portion of the glucocorticoid receptor. The centrally located DNA-binding domain can be separated from the remainder of the glucocorticoid receptor protein and still function efficiently. Structural studies of several members of the nuclear hormone receptor superfamily indicate that the structural motif for DNA binding is composed of two distinct zinc fingers in each subunit of the dimeric structure (Fig. 3, bottom) (Luisi et al. 1991).

Figure 3

Functional domains of proteins in the hormone receptor superfamily. (Top) Schematic of the functional domains of three members of the superfamily of nuclear hormone receptors, identified by their hormone ligands. Each domain (activation, DNA binding, (more...)

In summary, the potential activity of a promoter is set by the number of regulatory elements and the binding affinity of these elements for their cognate binding proteins. The activity of a promoter is then determined by the availability of active transcription factors. The simplest way to control the availability of a transcription factor is to regulate the concentration of the protein in the nucleus. More refined control pathways involve modulation of transcription factor activity by posttranslational modification, compartmentalization within the cell, or the availability of cofactors.

Experimental Approaches to Mapping Transcriptional Control Elements

The interaction of transcription factors with DNA is exquisitely sensitive to changes in nucleotide sequence. Thus, the key to the identification of transcriptional control elements is their sensitivity to mutation. Retroviral LTRs have been analyzed by mutagenesis in several different contexts. The easiest strategy for extensive analysis is to use plasmids in which the expression of a reporter gene is driven by the transcriptional control elements in the LTR. These plasmids can be introduced into almost any mammalian tissue culture cell line.

The plasmids become localized in the nucleus, where the vast majority are not integrated into host-cell DNA. They express the reporter gene only transiently (for 24–48 hours) and hence the term “transient expression” assays. In vitro transcription assays using purified plasmid DNA and crude or purified nuclear protein preparations have also been used to study LTR transcriptional function. In the latter assay, an LTR can be assembled into chromatin to resemble more closely an integrated provirus. Studies of LTRs in intact, integration-competent retroviruses are also informative. Recently, retroviral vectors have provided an additional source of integration-competent templates with which to study LTR function in the context of chromosomal DNA.

Deletion mutagenesis of the DNA sequences provides initial mapping information for the location of transcriptional control elements. Further analysis is expedited by the identification and characterization of the proteins that recognize these DNA elements. Simple DNA-protein interaction assays including DNase I footprinting and electrophoretic mobility-shift gels have been used extensively to identify binding activities in nuclear extracts. The sequence specificity of DNA binding allows the identification of key sequence components by base substitution mutations. LTRs containing these mutations can then be tested in a transcription assay. Correlation between functional data and DNA-binding data is a prerequisite for concluding that a particular protein factor functions at a particular DNA site.

Providing definitive evidence for the function of specific transcription factors in mammalian systems, including viral control elements, has posed a challenge. After mutated LTR elements have been used to establish a correlation between the binding of a factor and transcriptional function, several directions can be taken. A key advance is to obtain a cDNA clone for the production of protein in an expression system. The gene encoding a transcription factor or the purified protein can then be added to the transcription assay. Another approach is to delete a factor from an in vivo assay: This can be accomplished by using antisense technology or by mutating the chromosomal copy of a cellular gene by mammalian gene targeting technology. These techniques are just beginning to be applied in relevant systems.

In the next sections, selected LTRs are used to illustrate the interplay of host-cell machinery and viral transcriptional control. Highlighted are the responsiveness of viral gene regulation to cellular signaling pathways and the role of transcriptional control in disease. Viral _trans_-activators encoded by the complex retroviruses are presented in detail to demonstrate the added level of control exerted by these viruses.

Moloney Murine Leukemia Virus

Positive-acting Transcriptional Control Elements

The LTR of Moloney murine leukemia virus (Mo-MLV) is a prototype for the transcriptional control machinery of simple retroviruses. The U3 region of the LTR is divided into two parts, designated the promoter and the enhancer, and each has multiple components (Fig. 4) (Laimins et al. 1984b; Graves et al. 1985). The promoter contains the TATA element and a set of positive _cis_-acting regulatory signals that lie immediately upstream (Graves et al. 1986). The more distal element, the enhancer, was first identified by analogy with the 75-bp direct repeats in the simian virus 40 (SV40) viral enhancer, since the Mo-MLV LTR also displays a direct repeat of approximately 75 bp (Levinson et al. 1982). Some sequence similarity exists between the Mo-MLV repeat and a region of the SV40 and polyomavirus enhancers known as the enhancer core (Weiher et al. 1983). The Mo-MLV enhancer can augment the function of heterologous promoters.

Figure 4

Transcriptional control elements of the Mo-MLV LTR: Structure of the LTR of the Mo-MLV with transcriptional control elements marked. Positive regulatory elements (+) are located in the enhancer and promoter. Negative control elements (–) lie both (more...)

A variety of cellular transcription factors bind to the promoter and enhancer regions of the Mo-MLV LTR (Fig. 4). Many of these binding sites are closely packed or even overlapping, allowing for the synergistic or competitive interactions thought to be necessary to explain enhancer function in different cellular environments. The binding proteins that recognize the enhancer repeats are best understood: (1) Multiple members of the Ets family of transcription factors bind to two sites in each repeat unit (Gunther and Graves 1994). The Ets family is represented by more than 20 transcription factors in vertebrates (Wasylyk et al. 1993), and several family members are expressed in cell types permissive for Mo-MLV replication and disease progression. It is currently unclear which Ets protein(s) is involved in Mo-MLV gene expression. (2) The core-binding factor (CBF) is a heterodimeric binding protein that recognizes the enhancer core region (Ogawa et al. 1993a,b; Wang et al. 1993). The α-subunit of CBF interacts directly with DNA, whereas the β-subunit increases the stability of the CBF-DNA complex. There are three CBFa genes (Speck and Stacy 1995), two of which display a hematopoietic expression pattern. CBFb is encoded by a single, ubiquitously expressed gene. Recent genetic and biochemical evidence suggests that Ets and CBF proteins cooperate to activate transcription mediated by the Mo-MLV enhancer (Wotton et al. 1994; Sun et al. 1995). (3) NF1, which binds to two sites in each enhancer repeat, also belongs to a small family of DNA-binding proteins (Reisman 1990). (4) Mammalian type-C retrovirus enhancer factor 1 (MCREF-1) is a partially characterized DNA-binding protein with a binding site that overlaps the Ets- and CBF-binding sites. Other sites overlap the weaker NF1-binding site in the direct repeat and in the region that lies downstream from the direct repeat (Manley et al. 1993; Sun et al. 1993). (5) the glucocorticoid receptor also binds to the enhancer (DeFranco and Yamamoto 1986), and LTR-driven reporter genes can be activated by glucocorticoids (Miksicek et al. 1986). However, the glucocorticoid-receptor-binding site overlaps an E-box element (Murre and Baltimore 1992), which is recognized by several transcription factors from the basic helix-loop-helix (bHLH) structural family (Corneliussen et al. 1991; Nielsen et al. 1992).

As is the case with the Ets family of proteins, it is difficult to conclude from existing data which proteins in the CBF, NF1, and bHLH groups activate the Mo-MLV LTR in vivo. It is tempting to speculate that the members of each family which display lymphoid or hematopoietic expression patterns are the most important. However, it is not unreasonable to propose that a viral enhancer structure is designed for versatility, allowing the virus to employ different family members in distinct contexts.

The identification of LTR-binding proteins has guided more refined mutagenesis studies of the enhancer and the promoter. LTRs bearing wild-type or mutated binding sites have been tested using transient expression assays with reporter plasmid constructs in tissue culture (Graves et al. 1986; Speck et al. 1990a; Gunther and Graves 1994) as well as intact virus (Speck et al. 1990b). Even though the exact identity of each of the cellular proteins that function at particular sites is still in question, these functional analyses indicate that all of the identified binding sites correspond to positive-acting elements within the LTR. Thus, a large portion of the U3 region of the Mo-MLV LTR contains binding sites for cellular proteins that are necessary for high-level LTR transcriptional activity. Indeed, every retroviral LTR that has been analyzed displays an array of binding sites for cellular transcription factors, although the exact arrangement and the identity of the factors vary with the virus (see Figs. 5, 7, 8, and 9). It is striking that in some closely related retroviruses, such as the mammalian type-C group, the transcription-factor-binding sites represent the most highly conserved regions of the LTR (Golemis et al. 1990). In avian sarcoma/leukosis virus (ASLV), however, the transcription-factor-binding sites in U3 are divergent among the viral strains (Majors 1990).

Negative-acting Transcriptional Control Elements

Proviruses depend on host-cell machinery for transcription, and this can lead to extreme variability in the expression of integrated DNA in different cell types. For example, the Mo-MLV provirus is not expressed in preimplantation embryos or in embryonal carcinoma cell lines. Initial investigations of this phenomenon suggested an important role for DNA methylation in the repression of integrated genomes, as the unexpressed viral DNA is hypermethylated (Jähner et al. 1982). Furthermore, subsequent activation during mouse development or upon differentiation of embryonal carcinoma cell lines is associated with demethylation (Speers et al. 1980; Jähner and Jaenisch 1985). Although the correlation between silenced loci and methylation is strong, the causal role of methylation in the initial establishment of the off-state has not been established in these cell types or in other cases of proviral silencing (Gautsch and Wilson 1983; Hoeben et al. 1991; Challita and Kohn 1994). In several cases, it is clear that lack of expression precedes methylation. Nevertheless, it is generally agreed that methylation has a key role in maintaining the silenced state.

If methylation does not provide the primary control, then what does determine whether a particular provirus becomes an active locus? In the case of the early mouse embryo, two possible mechanisms have been investigated: the absence of positive regulatory proteins and the presence of negative regulatory elements. Support for the first model comes from the apparent absence of CBF and NF1-type factors in embryonal carcinoma cells (Speck and Baltimore 1987). Support for the second mechanism comes from the identification of two potential negative regulators of LTR expression in embryonal carcinoma cells. One lies just upstream of the 75-bp repeat and binds to the cellular embryonal LTR-binding protein (ELP), which is a member of the nuclear hormone receptor superfamily (Fig. 4) (Tsukiyama et al. 1992). Experiments in which ELP is transiently overexpressed have shown that this protein has modest repressor activity on the LTR. Moreover, ELP is present in embryonic cells but not in more differentiated cells. Another silencer of Mo-MLV maps to the primer-binding site (Barklis et al. 1986; Loh et al. 1988; Feuer et al. 1989). Again, a binding activity that recognizes this silencer has been identified (Yamauchi et al. 1995). At the 5′ distal region of the Mo-MLV LTR, a distinct negative-acting control element has been identified, and it is well-conserved among the type-C viruses. A cellular factor, upstream conserved region binding protein (UCRBP; also known as YY-1 or NF-E1) binds this region and mediates repression of the LTR in transient expression assays (Flanagan et al. 1992). The role of this negative regulatory element has not been tested in the context of a viral infection or for its relevance to the repression of the Mo-MLV enhancer in embryonal carcinoma cells.

Further insight into the phenomenon of embryonic restriction has come from analysis of proviral integrations that circumvent repression. Addition of a novel positive-acting element or removal of the negative-acting element can lead to an increase in the level of expression. For example, insertion near a strong cellular promoter can overcome repression (Barklis et al. 1986; Peckham et al. 1989). Similarly, a natural variant that acquired new positive-acting elements in its LTR shows enhanced expression (Prince and Rigby 1991). Insertion of an enhancer of a mutant polyomavirus that is capable of growing in embryonal carcinoma cells also rescues expression of Mo-MLV (Linney et al. 1984). Other natural variants also have been obtained in which the primer-binding site repressor element (Barklis et al. 1986; Weiher et al. 1987) or the U3 region repressor (Hilberg et al. 1987) has been altered. In summary, a combination of negative-acting and positive-acting elements can establish different levels of expression. Subtle changes in binding sites or the availability of factors can dramatically alter the transcriptional activity of an LTR.

Transcription Factors Mediate Responsiveness to Cellular Signaling Pathways

As illustrated above, the combination of positive- and negative-acting factors found in a specific cell type or at a particular developmental stage affects the relative activity of transcriptional control elements in the LTR. It is important to note that cell signaling can alter the activity of transcription factors such that the function of a promoter or enhancer is changed as the physiological state of the cell is changed. Two examples of the activation of retroviral gene expression by cell signaling are discussed below.

Induction of MMTV by Steroid Hormones

Milk-borne transmission of MMTV from mother to offspring requires efficient expression of viral proteins in the lactating mammary epithelium. As hormone levels rise, lactation is stimulated, viral production in the mammary gland increases, and new viral integrations occur. This increased tissue-specific viral load leads to the insertional activation of proto-oncogenes and the genesis of mammary tumors. Thus, physiological changes in a lactating female mouse lead to viral spread and disease induction.

Expression of viral proteins is influenced by steroid-hormone-mediated stimulation of the LTR. The steroid hormone induction of MMTV gene expression has been a useful system for dissecting the steroid hormone-signaling cascade (for review, see Gronemeyer 1991; Lucas and Granner 1992; Truss et al. 1992). Steroid hormones enter cells passively and bind to receptors that are located in the cytoplasm or the nucleus. Glucocorticoids bind to the cytoplasmic form of the glucocorticoid receptor, which is associated with the heat shock protein, Hsp90 (Pratt 1993). Hormone-bound glucocorticoid receptor dissociates from Hsp90 and is transported to the nucleus, where it binds with high affinity to DNA sites. Glucocorticoid receptor and other steroid hormone receptors are competent to activate transcription of a basal promoter in the absence of other transcription factors; however, there are nearby binding sites for other factors that influence hormone responsiveness of a particular cellular gene or provirus.

The responsiveness of the MMTV LTR to steroid hormones has been studied in whole animals as well as in tissue culture cells bearing proviruses and gene reporter systems (for review, see Günzburg and Salmons 1992; Truss et al. 1992). In cells treated with glucocorticoids, RNA levels rise up to 100 times above basal levels. The hormone-responsive element (HRE) of the LTR contains four binding sites for activated glucocorticoid and progestin receptors (Fig. 5). Binding sites also exist for additional proteins within the MMTV proximal promoter, indicating that this is a complex hormone-responsive element. Most important is the requirement for the NF1-binding site. The current model for the synergism that exists between transcription factors NF1 and the glucocorticoid receptor provides a paradigm for the interface of chromatin structure and eukaryotic transcription factors. The glucocorticoid receptor can bind an MMTV LTR that is assembled into chromatin, whereas NF1 cannot. It has been proposed that the binding of glucocorticoid receptor disrupts the chromatin structure of the LTR to allow NF1 binding and that NF1 binding to the LTR is necessary for transcriptional activation (Archer et al. 1992).

Figure 5

Transcriptional control elements of the MMTV LTR: Structure of the LTR of the MMTV with transcriptional control elements marked. Key components are the hormone-responsive element (HRE), the negative regulatory element (NRE), and the mammary-restricted (more...)

Induction of HIV during T-cell Activation: Role of NF-κB

In HIV-infected T cells and monocytes, transcription of proviral DNA is induced by a class of cellular transcription factors known as NF-κB proteins (Nabel and Verma 1993), encoded by the NF-κB/Rel gene family and forming a variety of homodimers and heterodimers (Fig. 6) (for review, see Siebenlist et al. 1994; Miyamoto and Verma 1995). Proteins in this family contain a conserved region of approximately 300 amino acids termed the Rel homology domain (RHD), which functions in DNA binding, nuclear localization, and dimerization. The carboxy-terminal segments of RelA, c-Rel, and RelB contain strong transcriptional activation domains, whereas those of NF-κB-1 and NF-κB-2 contain a series of ankyrin repeats. The latter motif is found in a variety of cellular proteins and appears to mediate protein-protein interactions that inhibit NF-κB function. NF-κB-1 and NF-κB-2 cannot participate in transcriptional activation until the ankyrin repeats are removed by proteolytic processing.

Figure 6

The NF-κB/Rel transcription factors. (Left) Structures of the members of the mammalian NF-κB/Rel family of transcription factors and IκB inhibitors. The positions of the conserved Rel homology domain (RHD), containing amino acids (more...)

The transcription function of NF-κB proteins is negatively regulated by IκB proteins (for review, see Beg and Baldwin 1993; Miyamoto and Verma 1995), which interact with NF-κB dimers in the cytoplasm and inhibit entry into the nucleus (Baeuerle and Baltimore 1988). The IκB proteins contain ankyrin repeats that mediate interactions with NF-κB dimers. IκBα, IκBβ, and Bcl-3 (a B-cell oncogene with an IκB-like structure) are encoded by unique genes, whereas IκBγ is produced by the NF-κB-1 gene through an internal transcription initiation. The ankyrin repeats of the unprocessed forms of NF-κB-1 and NF-κB-2 also perform IκB-like inhibitory functions by retaining the NF-κB dimers in the cytoplasm.

T-cell and monocyte activation, as well as cytokines such as tumor necrosis factor-α (TNF-α) and interleukin-1 (IL-1), induce phosphorylation and rapid proteolytic degradation of IκBα (Fig. 6, right) (Beg and Baldwin 1993; Miyamoto and Verma 1995), allowing NF-κB dimers to translocate to the nucleus. Degradation of IκBβ appears to be associated with a slower but more sustained activation of NF-κB (Thompson et al. 1995). Recent studies suggest that the rapid degradation of IκBα and the limited proteolytic cleavage of the NF-κB-1 and NF-κB-2 precursors are mediated through the ubiquitin-proteosome pathway of proteolysis (Palombella et al. 1994). A novel kinase has recently been purified that mediates the phosphorylation of IκBα (Chen et al. 1996). This kinase itself seems to be activated by ubiquitination, suggesting an active role for this protein modification in the NF-κB signal transduction pathway. The second-messenger pathways linking cellular stimulation with activation of the IκB kinase remain unclear. Proteins have been identified that interact with the TNF-α receptors (Rothe et al. 1995; Cheng and Baltimore 1996) and apparently participate in a signal transduction cascade to activate NF-κB; similarly, a kinase associated with the IL-1 receptor also appears to have a role in NF-κB activation (Cao et al. 1996). Furthermore, the oxidation/reduction state of the cell may have a role, as inhibitors of oxygen free radicals, such as dithiocarbamates (Staal et al. 1990; Kalebic et al. 1991; Schreck et al. 1992), block activation of NF-κB by all known inducers.

Several lines of investigation have implicated NF-κB proteins as important for HIV transcription. The two NF-κB-binding sites in the HIV LTR (Nabel and Baltimore 1987) are conserved in all HIV-1 isolates and NF-κB sites are also found in the related HIV-2 (Fig. 7) (Arya 1990; Tong-Starksen et al. 1990) and simian immunodeficiency viruses (SIVs). A variety of NF-κB homodimers and heterodimers have been shown to bind to the HIV NF-κB sites in vitro. Transcription assays demonstrate that NF-κB proteins can function at these sites; however, individual tests of different family members have led to the conclusion that the various NF-κB species exert differential effects on HIV gene expression. The NF-κB-1(p50)/RelA heterodimer is the major inducible NF-κB dimer in T cells and strongly activates HIV expression in in vitro transcription assays (Fujita et al. 1992; Kretzschmar et al. 1992). Homo-dimers of NF-κB-1 may actually inhibit the effects of more potent activators (Franzoso et al. 1992). The NF-κB-2 gene product may also have an important role in activating HIV transcription since NF-κB-2(p52)/RelA heterodimers synergize with HIV Tat in the _trans_-activation of LTR transcription (Liu et al. 1992).

Figure 7

Transcriptional control elements of the HIV-1 and HIV-2 LTRs. The structures of the HIV-1 and HIV-2 LTRs are shown with the locations of the binding sites for cellular transcription factors, including NF-κB and Sp1, marked. Enhancer sequences (more...)

Strong support for a role of NF-κB in HIV transcription comes from mutational analysis of the NF-κB-binding sites. In transient expression assays with reporter genes, a large number of stimuli that induce nuclear NF-κB have been shown to activate HIV LTR-mediated transcription (Nabel and Baltimore 1987; Siekevitz et al. 1987b). These include a variety of cytokines, T-cell activation signals, infection of cells by heterologous viruses, and exposure to stress and DNA-damaging agents (for review, see Antoni et al. 1994a). Deletion of the NF-κB sites results in loss of part or all of the LTR activation by these stimuli. Activation of the expression of latent HIV proviruses in both T-cell and monocyte cell lines by cytokines such as TNF-α and IL-1 is correlated with induction of nuclear NF-κB (Duh et al. 1989; Osborn et al. 1989). T-cell lines that contain an HIV provirus lacking the LTR NF-κB sites fail to undergo activation following cytokine treatment (Antoni et al. 1994b).

It is worth noting that HIV proviruses containing deletions or mutations in the NF-κB-binding sites still replicate both in primary T cells and in T-cell lines in culture (Leonard et al. 1989), although the NF-κB site in the SIV LTR is important for SIV replication in monocytes (Bellas et al. 1993). The multiple remaining transcription-factor-binding sites (such as the Sp1-binding sites; Fig. 7) apparently compensate for the loss of NF-κB-binding sites in supporting HIV replication (Jones et al. 1986). Mutation of the HIV NF-κB sites and adjacent Sp1-binding sites severely reduces viral replication. Interactions between RelA and Sp1 augment both binding to the HIV LTR and transcription itself (Perkins et al. 1993); thus, the highly conserved arrangement of NF-κB- and Sp1-binding sites may enhance the efficiency of these factors in activating HIV transcription. Interactions between RelA and Sp1 appear to be particularly effective in activating transcription of HIV LTR DNA assembled into chromatin (Pazin et al. 1996); binding of either factor in vitro to chromatin-assembled HIV DNA produces alterations in nucleosome structure similar to those observed for integrated HIV proviruses (Verdin et al. 1993).

In summary, HIV transcription is influenced by the activation state of infected immune cells. NF-κB proteins are critical for the activation of silent integrated HIV proviruses and apparently also play a part in replication in monocytes. Thus, an important evolutionary advantage conferred upon HIV by the presence of NF-κB-binding sites is likely to be the potential for activation of integrated provirus following immune or cytokine stimulation of infected cells. These features of HIV transcriptional regulation may play a part in the pathogenesis of HIV-induced disease (see Chapter 11.

Transcription Factors Mediate Disease Induction

The importance of viral transcriptional control pathways is further illustrated by replication-competent, nonacutely transforming retroviruses, which can cause disease by the insertional activation of oncogenes (see Chapter 10. High levels of viral transcription in target tissues can induce disease in multiple ways. The level of viral RNA influences the frequency of recombination with endogenous retroviral transcripts, a well-characterized proximal agent for leukemogenesis in the murine system (see Chapter 8. In addition, high viral titer increases the chance of insertion near a proto-oncogene. After insertion, the expression of the transforming gene is dependent on the strength of the viral transcriptional control elements. The enhancer and promoter activity of acutely transforming viruses also affects their ability to express their oncogenes in target tissues.

Thymocyte Transformation Induced by Mo-MLV Depends on Enhancer Elements

A variety of studies have suggested that the pathogenic potential of the type-C retroviruses is strongly affected by tissue-dependent activity of the transcriptional control elements of the LTR. Investigation of the enhancer of Mo-MLV and closely related viruses has been the most extensive. Changes in the enhancer elements affect the disease phenotype and the length of the latent period. Specifically, viruses constructed with only one copy of the 75-bp enhancer repeat element display an increased period of latency prior to tumor development (Li et al. 1987), and reduplication of this region is a common mutation that occurs during oncogenesis by endogenous MLVs (Stoye et al. 1991). Similarly, disease latency is increased by alterations in the strongest NF1-binding site, at least two glucocorticoid-receptor-binding sites, the strongest Ets-binding site, or the CBF-binding site (see Fig. 4) (Speck et al. 1990b). Thus, remarkably small changes in the DNA sequence of the viral U3 region have significant consequences for induction of disease.

Mo-MLV causes T-cell lymphomas. Infection of newborn mice with Mo-MLV, which mimics the horizontal route of transmission, leads to viral proliferation in the spleen. After 2 weeks, high viral titers are observed in the thymus, the site of transformation. Mutation of either the CBF-binding or the Ets-binding site causes a switch from the usual thymic disease toward the development of erythroleukemia (Speck et al. 1990b). These results are consistent with the observed relative importance of the core site for enhancer activity in transient expression assays in lymphoid cell lines versus fibroblast lines (Speck et al. 1990a). Additional support for the importance of the core site for T-cell transformation is found in natural variations in the core site in other MLVs. In the T-lymphomagenic SL3-3 virus, the CBF-binding site is crucial for high levels of gene expression mediated by the LTR in T cells (Thornell et al. 1988; Boral et al. 1989; Zaiman et al. 1995). A specific change in the core site in SL3-3 to resemble the sequence of the weakly leukemogenic Akv reduces the disease-induction capacity of the virus (Hallberg et al. 1991). Thymic tumors arising after infection of animals with this defective virus contain proviruses with additional alterations in the U3 region, including changes that restore the original CBF-binding site sequence (Morrison et al. 1995).

Similar observations have been made with chimeric viruses bearing portions of the Friend murine leukemia virus (Fr-MLV) and Mo-MLV. Although the genomes of the two viruses are extremely similar, their disease phenotypes are distinct. Chimeric viruses bearing reciprocal exchanges in the U3 region show reversed disease specificity (Chatis et al. 1983). Recombinant virus bearing the Fr-MLV direct repeats and additional downstream sequences in place of the Mo-MLV direct repeats cause only erythroleukemia. A reciprocal switch in which the Mo-MLV repeat replaces the Fr-MLV repeat shifts the virus-induced disease to T-cell lymphoma in 85% of infected animals (Li et al. 1987). These analyses map the viral determinants of tumor induction to the enhancer element within the LTR. Attempts to further localize the determinants of erythroid and lymphoid disease within the Mo-MLV and Fr-MLV enhancer revealed that important sites are scattered throughout these enhancers. Reciprocal exchanges of smaller regions of the enhancer only partially switched disease specificity (Golemis et al. 1989). Because the sequences bound by Ets and CBF proteins (Manley et al. 1989) are identical in these two enhancers, determinants of erythroid and lymphoid disease specificity must include sequences that flank this conserved region.

Avian Type-C Retroviral LTRs Vary in Disease Induction and Enhancer Strength

Studies of ASLVs link the efficiency of retroviral transcription with pathogenesis. Sequence variations in the LTRs can be directly correlated with the strength of promoter activity and with the ability of viruses bearing these LTRs to induce tumors. The R and U5 regions are highly conserved among the ASLV LTRs; however, important differences are found in U3 (Fig. 8) (for review, see Majors 1990; Ruddell 1995). The LTRs of Rous sarcoma virus (RSV) and exogenous leukemogenic ASLVs, such as Rous-associated virus (RAV-1), are approximately 320–350 bp in length and contain both enhancer and promoter elements.

Figure 8

Transcriptional control elements of ASLV LTRs. The structures of exogenous (RSV) and endogenous (RAV-0) ASLV LTRs are shown. The positions of important regulatory sequences and binding sites for cellular transcription factors are indicated. The EFII site (more...)

The RSV LTR, as the prototypic ASLV LTR, has been studied extensively. It is a strong transcriptional promoter (Gorman et al. 1982) that functions well in both avian and mammalian cells. The RSV LTR contains an enhancer (Luciw et al. 1983; Laimins et al. 1984a; Cullen et al. 1985; Norton and Coffin 1987), within which are a number of distinct segments to which transcription factors bind, including enhancer II (EFII) composed of CCAAT/enhancer-binding protein (C/EBP)-related motifs and EFIII or CArG sequences similar to serum response elements (Fig. 8). Additional elements demonstrated to have a role in RSV LTR transcriptional regulation include a second EFIII sequence and two enhancer I (EFI) sequences related to the Y-box DNA motif (Sealy and Chalkley 1987; Goodwin 1988; Ruddell et al. 1988, 1989; Ruddell 1995). As in the case of the Mo-MLV LTR, most of the host cellular factors important for RSV LTR expression are present in many different cell types. Several have been cloned and shown to be related to previously identified families of mammalian transcription factors (Ryden and Beemon 1989; Ozer et al. 1990; Kandala and Guntaka 1994; Sears and Sealy 1994; Smith et al. 1994). Furthermore, as in other promoters, these factors may cooperatively interact with other cellular transcription factors; for example, C/EBP forms a functionally active complex with the RelA NF-κB protein on the RSV LTR (Bowers et al. 1996).

In contrast, the LTRs of the major endogenous avian ALVs, such as RAV-0, a nonleukemogenic, sub-group-E endogenous virus, exhibit much weaker promoter activity than the exogenous LTRs and are inactive in assays for enhancer activity when linked to heterologous promoters (Cullen et al. 1983, 1985; Weber and Schaffner 1985; Norton and Coffin 1987). The endogenous LTRs are considerably shorter (∼280 bp in length) and exhibit considerable sequence divergence from the RSV enhancer elements (Fig. 8). Although the endogenous LTRs retain one CArG box, they are missing important transcription-factor-binding sites present in 5′ U3 sequences, including a second CArG sequence, as well as the EFI-binding site (Zachow and Conklin 1992). The lack of these sites could explain the low levels of expression of these integrated proviruses.

The nucleotide sequence and functional variations between the LTRs of different ASLVs have important biological consequences. For example, the LTR of the highly oncogenic RAV-1 subgroup-A ALV exhibits strong transcriptional activity, whereas the U3 region of the nononcogenic RAV-0 functions as a weaker enhancer (Cullen et al. 1985; Weber and Schaffner 1985; Greenhouse et al. 1988). The correlation between pathogenesis and LTR sequence leads to the hypothesis that the ability of different avian retroviruses to induce avian lymphomas is dependent on the transcriptional activity of the LTR (Tsichlis and Coffin 1980). Experimental tests of this proposal have come from the study of chimeric viruses. Substitution of the RAV-1 LTR into RAV-0 results in a chimeric virus with increased in vivo replication and enhanced oncogenicity (Robinson et al. 1982, 1985; Brown et al. 1988). Both LTR and env sequences are found to affect the level of ASLV replication in the bursa, presumably resulting in increased opportunities for lymphomagenesis through promoter insertion (see below and Chapter 10. The precise sequences in the avian LTRs responsible for these effects on pathogenicity have not been identified; unlike the analysis of MLV-induced leukemogenesis, a detailed analysis of ASLV leukemogenicity involving LTR point mutations has not yet been performed. Furthermore, it is clear that other _cis_-acting, non-LTR sequences may also affect the efficiency of ASLV gene expression. An enhancer has been identified in the gag gene of several ASLV strains (Arrigo et al. 1987; Carlberg et al. 1988) (see below), and pol region sequences have also been shown to influence ASLV gene expression (Petropoulos and Hughes 1991), although a defined intragenic enhancer in the pol gene has not been reported.

Cell-type-specific differences in transcription factors interacting with ASLV LTRs may also have a role in leukemogenesis. Proteins binding to both enhancer regions I and II have been identified in B cells and cloned (Smith et al. 1994); these labile proteins are clearly distinct from those found in other cell types which also bind to these regions (Ruddell et al. 1989). These B-cell-specific binding activities may be important for transcription of the ASLV LTR in bursal cells, leading to subsequent development of B-cell lymphomas.

Tissue-specific Expression of MMTV Is Key to Disease Induction

Although control of MMTV expression by steroid hormones is necessary for optimal viral production and disease, additional positive- and negative-acting factors modulate MMTV gene expression and influence pathogenesis. MMTV-induced mammary tumorigenesis requires the activation of proto-oncogenes in the mammary gland by insertional activation. Recent findings indicate that this insertional activation is not hormone-dependent. Instead, positive-acting elements in the most distal region of the MMTV LTR function in mammary-specific gene expression and are necessary for continued expression of MMTV and proto-oncogenes (see Fig. 5) (Stewart et al. 1988; Mink et al. 1990; Lefebvre et al. 1991; Mok et al. 1992; Mellentin-Michelotti et al. 1994). The region can function in the absence of the hormone-responsive elements and does not mediate hormone inducibility of the LTR. Dissection of this region has identified multiple elements that are bound by cellular proteins (Michalides et al. 1982; Mink et al. 1992). These binding sites are also present in the enhancer regions of acidic whey protein, consistent with their putative role in mammary-gland-specific gene activation. A better understanding of this element requires identification of the relevant cellular proteins and an analysis of their tissue distribution.

MMTV also can cause T-cell lymphomas in mice, and the disease determinants map to the LTR. Analysis of the new MMTV LTRs found in T-cell lymphomas indicates that a central portion of the U3 region is deleted (Michalides et al. 1982; Dudley and Risser 1984; Hsu et al. 1988). Moreover, an infectious MMTV variant (dimethylbenz [α]-anthracene leukemia virus [DMBA-LV] or thymotropic type-B leukemia virus [TBLV]) that causes T-cell tumors contains a similar LTR deletion (Ball et al. 1988). The deletion includes a region that has been mapped as a negative regulatory element (NRE) in transient assays of MMTV LTR function in nonmammary cell lines (Morley et al. 1987; Hsu et al. 1988; Lefebvre et al. 1991) and in transgenic mice (Ross et al. 1990). A binding activity, NRE-binding protein (NBP), that recognizes sequences in this region has the features of a transcriptional repressor (see Fig. 5) (Dudley and Risser 1984; Bramblett et al. 1995). These findings suggest that MMTV gene expression is repressed in many tissues, restricting the replication of the virus to a few cell types, including those from the mammary gland.

In conclusion, the ability of retroviruses to induce transformation is most clearly associated with the induction or transduction of cellular proto-oncogenes. The studies of MMTV, MLV, and ALV demonstrate that this activation is dependent on virus-specific elements that likely have been optimized for viral transmission.

Retroviral Transcriptional **Trans**-activators

The complex retroviruses (the lentiviruses, spumaviruses, and viruses of the HTLV/bovine leukemia virus [BLV] family) have evolved regulatory mechanisms that employ virally encoded transcriptional activators (Table 1) as well as cellular transcription factors. These _trans_-acting viral proteins serve as activators of RNA transcription directed by the LTRs. This situation is analogous to that observed in a variety of DNA tumor viruses, such as papovaviruses, herpesviruses, and adenoviruses, which encode transcriptional _trans_-activators as early gene products (for review, see Jones et al. 1988; Nevins 1991). Although the specific modes of action of the retroviral _trans_-activators vary, each acts in conjunction with cellular proteins to induce the transcription of proviral DNA. This establishes a strong positive feedback loop that enhances gene expression. Following integration of viral DNA, relatively small amounts of the viral _trans_-activator protein may be synthesized (in response to cellular transcription factors as described above). Even low levels of the _trans_-activators will strongly activate transcription of all of the viral RNAs, resulting in high levels of viral production. It has been suggested that the expression of retroviral _trans_-activators may serve to delineate two distinct phases of retroviral infection (Cullen and Greene 1989; Kim et al. 1989; Feinberg et al. 1991): an early phase with gene expression restricted to regulatory genes and a late phase with high-level expression of structural genes, enzymes, and genomic RNA.

Table 1

Retroviral Transcriptional Trans-activators.

Retroviral _trans_-activators may also explain the ability of complex viruses to induce latent or productive infection. Latent infection might occur when an absence of important cellular factors results in the transcription of only low levels of viral mRNAs that encode the _trans_-activator proteins. In the presence of cellular transcription factors, synthesis of viral _trans_-activator proteins is induced, resulting in high-level viral gene expression (Antoni et al. 1994a). Viral gene expression is therefore either “on” or “off.” Such a clear discrimination between the different forms of infection would be particularly advantageous for a cytopathic virus such as HIV-1. Continuous low-level expression of potentially toxic HIV proteins might result in cell death prior to the production of large amounts of progeny virus. Therefore, in the absence of conditions favoring a fully productive infection, it would be advantageous for viral gene expression to be shut off. Furthermore, the ability to establish a latent infection could be advantageous in the face of an effective host immune response. Productive viral infection and concomitant cytotoxicity would then only occur in the context of high-level viral expression mediated by the viral _trans_-activator.

The Tax Protein

HTLV-1 (Felber et al. 1985; Fujisawa et al. 1985; Sodroski et al. 1985a), the related HTLV-2 (Chen et al. 1985), and BLV (Rosen et al. 1985c; Derse 1987) encode viral _trans_-activators known as Tax proteins (for review, see Cullen 1992). The properties of the tax gene product of HTLV-1 described here are quite similar to those of the HTLV-2 and BLV proteins. HTLV-1 Tax is a 40-kD phosphoprotein encoded by a multiply spliced RNA (for review, see Smith and Greene 1991). Tax is found in the nucleus of infected cells. At its amino-terminal end, the protein contains zinc-finger-like cysteine-rich and histidine-rich motifs that are important for _trans_-activation.

Tax activates HTLV-1 transcription through DNA sequences in the U3 region of the HTLV-1 LTR (Fig. 9) (Rosen et al. 1985a; Fujisawa et al. 1986; Paskalis et al. 1986; Shimotohno et al. 1986) known as the Tax-responsive elements (TREs). TREs consist of three copies of an imperfectly repeated 21-bp element that are sufficient to confer Tax responsiveness to a heterologous promoter.

Figure 9

Transcriptional control elements of the HTLV-1 LTR. The locations of the three 21-bp repeat Tax-response elements (TREs) are shown, as well as the locations of binding sites for cellular transcription factors.

Although the mechanisms by which Tax activates HTLV transcription are still being elucidated, certain principles are clear. Tax does not directly bind to TRE DNA. Instead, the TRE is recognized by cellular transcription factors (Jeang et al. 1988; Nyborg et al. 1988; Yoshimura et al. 1990) whose functions are enhanced by Tax. The CREB/ATF (cAMP response element/ activating transcription factor) family of transcription factors, which normally mediate cAMP activation of cellular gene expression (Lalli and Sassone-Corsi 1994), bind to the TREs. CREB/ATF proteins belong to the class of transcription factors known as bZIP proteins, which utilize a basic region-leucine zipper (bZIP) motif for DNA binding. DNA binding requires the formation of different homo- and heterodimers via the leucine zipper structure. Tax has been shown to interact physically with CREB/ATF proteins (Zhao and Giam 1991; Suzuki et al. 1993) and can be identified in ternary complexes with both CREB/ATF proteins and DNA. One model of Tax activation is that it functions to increase association of cellular transcription factors with TRE DNA (Fig. 10) (Matthews et al. 1992; Zhao and Giam 1992) by increasing the dimerization of CREB/ATF proteins (Wagner and Green 1993). Tax interacts with amino acid residues in the basic region of these bZIP proteins and acts both to enhance dimerization and DNA binding and to alter the specificity of specific CREB/ATF factors for particular DNA sequences (Perini et al. 1995; Baranger et al. 1995). CREB/ATF factors also interact with transcriptional coactivators, such as CBP (Creb-binding protein; for review, see Janknecht and Hunter 1996); thus, another function of Tax may be to help bring CBF and/or the related p300 proteins to the promoter. These coactivators have recently been suggested to possess histone acetyltransferase activity which may contribute to transcriptional activation through altering the structure of local chromatin (Ogryzko et al. 1996). A direct effect of Tax on the basal transcriptional apparatus has also been suggested (Caron et al. 1993; Clemens et al. 1996). Full Tax-mediated activation may also involve other factors that bind to the HTLV LTR (Fig. 10) (Marriott et al. 1990), such as the Ets proteins (Gitlin et al. 1993b), which do not contain leucine zippers.

Figure 10

Possible mechanisms of HTLV-1 Tax _trans_-activation. The HTLV-1 Tax _trans_-activator activates HTLV-1 transcription through the Tax response elements (21-bp repeats, TRE) in the U3 region of the HTLV-1 LTR. The TRE binds dimers of CREB/ATF-related proteins, (more...)

Tax also modulates the transcription of a number of cellular genes whose expression is directly relevant to the pathogenesis and clinical characteristics of HTLV-1 diseases (Table 2) (for review, see Gitlin et al. 1993a; Yoshida 1994). The expression of several of these cellular genes is clearly mediated through CREB/ATF and other leucine zipper transcription factors and depends on the effect of Tax on affinity and binding-site specificity of CREB/ATF proteins. Tax also activates a number of cellular genes that are regulated by the Rel/NF-κB transcription factors (see Fig. 6B). Tax induces nuclear translocation of NF-κB dimers and activation of promoters containing κB motifs (Ballard et al. 1988; Leung and Nabel 1988; Ruben et al. 1988). The mechanism by which Tax exerts this effect is unclear, but it is distinct from that by which Tax induces CREB/ATF binding; different regions of the Tax protein mediate effects through NF-κB or through CRB/ ATF proteins (Smith and Greene 1990; Semmes and Jeang 1992). Tax has been shown to induce IκB degradation (Sun et al. 1994; Good and Sun 1996; McKinsey et al. 1996), to disrupt inhibition of NF-κB by p100 NF-κB-2 precursor protein (Kanno et al. 1994), to bind to proteosome components enhancing the proteolytic processing of p105 NF-κB-1 (Rousset et al. 1996), and to bind directly to different NF-κB/Rel proteins possibly enhancing their interactions with target DNA (Yoshida 1994).

Table 2

HTLV-1 Tax: Effects on Cellular Genes.

Whatever the molecular mechanisms, Tax activation of cellular genes through both CREB/ATF proteins and NF-κB has profound effects on T lymphocytes and likely has an important role in HTLV-1-induced lymphomagenesis. Tax induction of NF-κB activates transcription of the IL-2, IL-2 receptor (IL-2R) α-subunit, and IL-6 promoters (Inoue et al. 1986; Cross et al. 1987; Maruyama et al. 1987; Siekevitz et al. 1987a; Ruben et al. 1988; Mori et al. 1994). Expression of genes encoding cytokines and cytokine receptors has been proposed to play an important part in the early events of HTLV-induced tumorigenesis (Yoshida and Seiki 1987; Kitajima et al. 1992). In one model, increased levels of IL-2 and IL-2R as well as of other cytokines and receptors would enhance T-cell proliferation, which in turn would accelerate the occurrence of additional mutagenic events, culminating in the development of a lymphoid malignancy (see Chapter 10. It is not known whether Tax activation of NF-κB has any evolutionary value for HTLV-1 infection or whether, in the context of viral replication, it represents an epiphenomenon that may inadvertently lead to lymphoma development. However, recent studies suggest that association with NF-κB precursors may retain Tax in the cytoplasm, possibly promoting latency of HTLV-1 infection (Béraud et al. 1994).

Other activities of Tax may also contribute to its role in oncogenesis. Tax has been demonstrated to inhibit the transcription of at least one cellular gene, that encoding human β DNA polymerase (Jeang et al. 1990) through effects on bHLH transcription factors (Vittenbogaard et al. 1994). Repression of this gene has been hypothesized to lead to the loss of DNA repair activities in an HTLV-1-infected cell. This might contribute to the accumulation of additional genetic changes responsible for oncogenesis. Furthermore, Tax may affect other signal transduction pathways in transformed cells; for example, activation of the Jak-Stat pathway has been reported in HTLV-1- and Tax-transformed cell lines and may contribute to alterations in cellular proliferation (Migone et al. 1995; Xu et al. 1995).

Spumavirus **Trans-**activators

The primate spumaviruses, the human spumaretrovirus (human foamy virus, HFV), and the simian foamy virus (SFV) also encode _trans_-activating proteins referred to, respectively, as Bel-1 for HFV and Taf for SFV, recently renamed Tas proteins (_trans_-activators of spumaviruses; Zou and Luciw 1996). As in the case of the HTLV/BLV _trans_-activators, those of the spumaviruses activate viral transcription through the U3 regions of the LTR. The transcriptional activity of the HFV LTR is higher in cells infected with HFV than in uninfected cells (Rethwilm et al. 1990). A deletion analysis of an infectious molecular clone identified the bel-1 gene as responsible for _trans_-activation (Rethwilm et al. 1991). Bel-1 is a 300-amino-acid nuclear phosphoprotein (Keller et al. 1991). The activation potential of the Bel-1 protein can be demonstrated by _trans_-activation of the HFV LTR in transient expression assays (Rethwilm et al. 1991). Expression of the bel-1 gene is regulated by an internal, intragenic promoter in the HFV env DNA sequence (Löchelt et al. 1993), which is itself _trans_-activated by Bel-1, creating a positive feedback loop. Bel-1 is required for HFV infectivity (Baunach et al. 1993). Several LTR regions, as well as two regions adjacent to the internal viral transcription intiation site, serve as targets for _trans_-activation by the human and simian Tas proteins (Fig. 11) (Venkatesh et al. 1991; Keller et al. 1992; He et al. 1996; Zou and Luciw 1996). Bel-1 and Taf exhibit direct sequence-specific binding to some of these target DNA sequences (which include a homologous seven-nucleotide sequence present in the LTR and internal promoters) but not to other target regions required for _trans_-activation (He et al. 1996; Zou and Luciw 1996). Thus, the two primate Tas proteins appear to have evolved their own unique mechanism of _trans_-activation that involves both sequence-specific binding to DNA and a requirement for interactions with other cellular transcription factors.

Figure 11

Transcriptional control elements of the HFV LTR. The locations of the Bel-response elements (BREs) are shown, as well as the locations of binding sites for the AP1 cellular transcription factor.

Tat, the **Trans-**activator of HIV

The transcriptional _trans_-activators of the lentiviruses are encoded by the viral tat genes (Table 1). Surprisingly, they are of two types with very different mechanisms of action. The Tat protein of visna virus likely acts through cellular transcription factors binding to U3 sequences in a manner similar to that of Tax and Bel-1. AP1- and AP4-binding sites in the U3 region of the visna virus LTR appear to be important for visna virus _trans_-activation (Gabuzda et al. 1989). In contrast, Tat proteins of primate lentiviruses (including HIV-1, HIV-2, and SIV) and the equine infectious anemia virus (EIAV) influence transcription by binding directly to the 5′ ends of nascent viral RNAs, where they exert effects on transcriptional elongation (for review, see Sharp and Marciniak 1989; Cullen 1993; Jones and Peterlin 1994; Gaynor 1995). This unusual mode of regulating transcription activation has been the subject of extensive study during the past few years.

The Tat protein of HIV-1 is a small (86–102-amino-acid) protein encoded by a spliced mRNA derived from two exons within the central region and env gene of the HIV genome (Arya et al. 1985; Sodroski et al.] 1985b). Tat is a very potent activator of HIV gene expression, enhancing LTR-directed transcription by hundreds to thousands of fold. Introduction of mutations of the tat gene into infectious molecular clones of HIV eliminates HIV production (Dayton et al. 1986; Fisher et al. 1986). Thus, Tat is essential for HIV replication.

Efforts to map the Tat-responsive region of the HIV LTR led to the discovery that Tat binds to a stable stem-loop structure, referred to as TAR, present at the 5′ end of all HIV-1 RNAs (nucleotides +1 to +59) (Figs. 7 and 12) (Rosen et al. 1985b; Muesing et al. 1987; Hauber and Cullen 1988). This RNA secondary structure is highly conserved and required for Tat function. Mutations that disrupt the stem inhibit Tat activation; however, restoration of the stem by introduction of compensatory mutations restores _trans_-activation (Berkhout et al. 1989; Selby et al. 1989; Roy et al. 1990b). The TAR loop is also important because point mutations in this sequence dramatically decrease Tat _trans_-activation (Feng and Holland 1988).

Figure 12

Structures of HIV-1 and HIV-2 TAR RNAs. The positions involved in binding to Tat (arrows) and the bases involved in tertiary structure alterations following Tat binding (color) are shown for HIV-1 RNA.

HIV-1 Tat binds to TAR RNA at a small bulge at nucleotides +22 to +24 (Figs. 12 and 13) (Dingwall et al. 1989; Cordingley et al. 1990; Roy et al. 1990a; Gait and Karn 1993). A subdomain of Tat is sufficient for binding; arginine residues in the basic domain interact with a uridine residue (U23) at the base of the bulge (Frankel 1992). The TAR bulge also serves to distort the TAR RNA, allowing Tat to interact with residues in a widened major groove (Weeks and Crothers 1991; Wang and Rana 1996). Structural studies of Tat and Tat-TAR interactions are beginning to be reported and are shedding further light on this important problem (Calnan et al. 1991; Puglisi et al. 1992; Aboulela et al. 1995). Critical TAR elements for Tat-TAR interactions are positioned in the major groove in the vicinity of the bulge and include functional groups on U23, G26, and A27 as well as phosphate groups P22, P23, and P40. Nuclear magnetic resonance (NMR) analysis of TAR RNA bound to either arginamide, an arginine derivative (Puglisi et al. 1992; Aboulela et al. 1995), or to a 37-amino-acid peptide containing the minimal RNA recognition domain of Tat composed of both basic and core residues (Aboulela et al. 1995), demonstrated that binding to the TAR bulge results in conformational changes in TAR. The conformational changes seen after binding to Tat peptide allow increased accessibility of critical phosphate residues within the major groove, leading to multiple points of contact between TAR and the basic and core regions of Tat (Fig. 13A) and specific high-affinity binding. NMR studies of the free EIAV (Willbold et al. 1994) and HIV Tat proteins (Fig. 13, right) (Bayer et al. 1995) have also been conducted. HIV Tat possesses a central conserved hydrophobic core 16 amino acids in length and a rigid glutamine-rich 17-amino-acid domain. These serve as scaffolds from which project other Tat functional domains, such as the basic domain that binds TAR and a cysteine-rich domain that binds zinc (Derse et al. 1991; Frankel 1992). It is important to note that data derived from amino acid or peptide complexes with TAR or from purified Tat protein may not fully represent the structure of the native Tat-TAR complex; further structural studies are needed.

Figure 13

Structural studies of Tat and TAR. (Left) NMR-based model of TAR RNA in the absence (left) and presence (right) of ligand, as determined by analysis of a 37-residue polypeptide that carries the minimal RNA recognition region of the Tat protein and closely (more...)

A series of cellular proteins also binds to the stem, bulge, and loop in TAR RNA (Fig. 14) (Marciniak et al. 1990a; Sheline et al. 1991; Wu et al. 1991; Rounseville and Kumar 1992). Some of these proteins are required for Tat _trans_-activation. For example, disruption of cellular protein binding to the loop region eliminates Tat _trans_-activation, although Tat binding to TAR RNA still occurs (Marciniak et al. 1990a). Furthermore, the binding of TRP-185, a cellular loop-binding protein that can activate LTR transcription, is modulated by Tat (Wu et al. 1991). One function of these cellular factors may be to stabilize the RNA secondary structure, facilitating Tat binding and function (Alonso et al. 1992). NMR studies of Tat-TAR structures have demonstrated a close proximity of the loop to the TAR bulge region (Aboulela et al. 1995), suggesting that loop-binding proteins may either affect the RNA structure of this critical region or directly interact with Tat. Tat itself also binds to cellular proteins, some of which enhance LTR _trans_-activation (Desai et al. 1991; Shibuya et al. 1992; Ohana et al. 1993; Yu et al. 1995). Further evidence for the role of host proteins in mediating Tat _trans_-activation comes from the observation that Tat will _trans_-activate the HIV LTR in rodent cells only if human chromosome 12 is present (Hart et al. 1989; Newstein et al. 1990), suggesting that human cellular proteins are required for Tat function. Thus, Tat seems to function in the context of a protein-RNA complex present at the 5′ end of nascent HIV transcripts.

Figure 14

Possible mechanisms of HIV-1 Tat function. The structure of nascent HIV-1 RNA containing the 5′ TAR element, with bound Tat and cellular proteins is shown. Different potential mechanisms of Tat activation are illustrated: (A) As outlined in the (more...)

One function of Tat binding to TAR RNA is apparently to position Tat in proximity to the RNA start site. In this model, TAR would serve as a “scaffold” placing Tat in an optimal position to _trans_-activate transcription (Fig. 14) (Berkhout et al. 1990). In fact, Tat can activate HIV transcription even if it is tethered to nascent HIV RNA through heterologous RNA-protein interactions (Selby and Peterlin 1990; Southgate et al. 1990). For example, a fusion of Tat with the MS2 bacteriophage coat protein gene was able to _trans_-activate an HIV LTR into which MS2 coat protein RNA-binding sequences were inserted in place of TAR. Binding of Tat to the 5′ end of the nascent RNAs apparently allows transcriptional activation domains of Tat to interact with targets that are a part of the cellular transcriptional apparatus.

How does Tat activate HIV gene expression? Although initial studies suggested that Tat could enhance both the transcription and translation of HIV RNAs (Cullen 1986), it is clear that the predominant effect of Tat is to increase transcription of HIV RNA (for review, see Sharp and Marciniak 1989; Cullen 1993; Jones and Peterlin 1994). The precise mechanisms responsible for this effect remain controversial. As described above, transcription of eukaryotic RNA by RNA polymerase II is dependent on the assembly of a complex array of proteins, including RNA polymerase II and the basal transcription factors, at the RNA start site, resulting in initiation of RNA transcription. The amount of RNA transcription is determined both by the frequency of transcriptional initiation and by the efficiency with which an initiated RNA polymerase II transcription complex continues to elongate the nascent RNA transcript. There is strong evidence that Tat may enhance the efficiency of elongation as well as possibly affect initiation. This provides a fascinating new paradigm of transcriptional regulation. Different possible models of Tat transcriptional activation are shown diagrammatically in Figure 14.

Numerous experiments demonstrate that Tat enhances the efficiency of elongation of TAR containing RNAs. The fundamental observation is that HIV RNA transcripts that are initiated in the absence of Tat tend to be truncated, suggestive of prematurely terminated, poorly processive elongation. Transcripts that are synthesized in the presence of Tat are more likely to be full length. This phenomenon has been observed in in vitro transcription assays (Toohey and Jones 1989; Marciniak et al. 1990b; Marciniak and Sharp 1991; Kato et al. 1992), in nuclear run-on experiments that used transfected Tat and LTR-driven reporter gene plasmids (Kao et al. 1987; Laspia et al. 1989), and in nuclear run-on assays that used integrated HIV proviruses mutated in Tat (Feinberg et al. 1991). In particular, recent in vitro assays have demonstrated that Tat cooperates with a cellular factor to stimulate elongation in a manner clearly distinct from the effects on transcriptional initiation observed with cellular activators such as Sp1 (Zhou and Sharp 1995), and other studies have suggested that Tat remains associated with the elongating RNA polymerase II transcription complexes on the HIV LTR (Keen et al. 1996). The effects of Tat on HIV transcriptional elongation are very reminiscent of the function of the N protein of the bacteriophage λ (Nodwell and Greenblatt 1991; Cullen 1993), which acts as a transcriptional antiterminator of λ transcription. Like Tat, the N protein, in conjunction with host Escherichia coli proteins, binds to an RNA stem-loop in the nascent RNA transcript. This complex acts to modify E. coli RNA polymerase such that it resists premature transcriptional termination, likely in a manner analogous to the effect of Tat on the elongation of the RNA polymerase II complex.

There has also been evidence from both in vitro and in vivo data that Tat increases initiation of HIV RNA transcription. Nuclear run-off experiments have suggested increased transcriptional initiation in the presence of Tat (Laspia et al. 1989), and Tat _trans_-activation is dependent on 5′ promoter elements. Tat activation is dependent on the specific sequence of the HIV TATA box (Berkhout and Jeang 1992); TATA sequences from other genes will not substitute for it. Functional and structural interactions between Tat and Sp1 have also been demonstrated (Jeang et al. 1993). Furthermore, Tat fused to a heterologous DNA-binding domain will activate HIV transcription when bound to DNA sequences upstream of the transcription start site (Kamine et al. 1991; Southgate and Green 1991). In vitro transcription studies from some laboratories (Laspia et al. 1989; Bohan et al. 1992) have also demonstrated clear enhancement of LTR transcriptional initiation in the presence of Tat. Tat is suggested to affect the stability of the basal transcription complex (Bohan et al. 1992) and more recently was shown to bind directly to the TATA-binding protein component of the TFIID basal transcription factor (Kashanchi et al. 1994; Veschambre et al. 1995). Cellular Tat-binding proteins may also interact with the basal transcriptional apparatus to induce transcription (Desai et al. 1991; Shibuya et al. 1992; Ohana et al. 1993; Yu et al. 1995).

Do the observed effects of Tat activation on both HIV transcriptional initiation and elongation represent a single mode of action? Several recent models have been proposed that may accommodate both sets of observations (Fig. 14) (Cullen 1993; Jones and Peterlin 1994; Herrmann and Rice 1995). In one model, it is proposed that two types of transcription complexes may be formed on the HIV LTR at overlapping promoter elements (Cullen 1993; Lu et al. 1993). One promoter mediates the formation of transcription complexes that elongate inefficiently. In fact, an initiator (promoter) of short (nonprocessive) transcripts has been defined in the HIV LTR (Sheldon et al. 1993). A second LTR promoter would be responsible for the formation of initiation complexes which are capable of efficient elongation. In this model, Tat would induce transcription from this second promoter leading to the generation of processive elongating transcription complexes. An alternative model suggests that the same LTR promoter is used in the presence or absence of Tat. Tat would alter the protein constituents of the initiation complex, such that initiation and/or RNA elongation occurs more efficiently. The fact that several general transcription factors have roles in both initiation and elongation (Drapkin et al. 1993) is consistent with this model, and Tat has been shown to have effects on in vitro transcription similar to those of TFIIF (Kato et al. 1992) and may require TFIIF for its activity. One interesting variant of this model derives from the recent observation that a Tat-associated protein kinase (TAK) may enhance the phosphorylation of the carboxy-terminal domain of RNA polymerase II (Herrmann and Rice 1995). Tat-induced phosphorylation of RNA polymerase II or of other basal transcription factors could result in increased initiation and/or processivity of transcription. TAK associates with Tat in vivo, and an intact carboxy-terminal domain of RNA polymerase II is required for Tat function (Yang et al. 1996). Related studies have suggested that Tat interacts with components of the TFIIH basal transcription factor and stimulates phosphorylation of the RNA polymerase II carboxy-terminal domain (CTD) through the TFIIH kinase activity that normally is involved in stimulation of transcriptional elongation and promoter clearance (Parada and Roeder 1996). Furthermore, Tat can be coprecipitated with the RNA polymerase II holoenzyme, even in the absence of TAR (Cujec et al. 1997). Thus, a model is emerging whereby Tat recruits to the promoter and/or enhances the activity of a cellular kinase whose function is to phosphorylate the CTD of RNA polymerase II (Fig. 14C) resulting in enhanced transcriptional elongation. An effect of Tat on initiation could actually reflect enhanced transcript elongation required for promoter clearance, to allow formation of the next initiation complex. This model is somewhat similar to that described for a Drosophila heat-shock-responsive promoter in which transcriptional initiation occurs; however, the transcription complex pauses after synthesis of a short fragment (17–37 nucleotides) and does not continue to elongate until the heat shock factor acts upon it (Lis and Wu 1993). In this regard, it has recently been reported that Tat destabilizes the binding of RNA polymerase II to the stable TAR RNA stem-loop. It is possible that by inhibiting RNA polymerase binding to TAR, Tat could allow elongation of a paused transcription complex, resulting in both increased elongation and increased efficiency of generation of the next transcriptional complex (Mavankal et al. 1996).

Why has HIV developed a _trans_-activation system based on the interaction of its tat gene product with newly synthesized RNA, instead of with an upstream promoter element in the LTR as seen in the other retroviral _trans_-activators? Perhaps interaction of Tat with viral RNA offers an even tighter “on/off” control mechanism than would be provided by interactions with proviral DNA. Both Tat protein and TAR RNA would have to be present before Tat _trans_-activation could proceed. Given the extraordinary potency of Tat, such tight control of the expression of potentially toxic HIV structural proteins would be advantageous. Once HIV gene expression is initiated, the potent and rapid feedback afforded by Tat/TAR interactions would allow rapid production of adequate levels of the Rev protein to allow progression to productive infection (see below). This mechanism could lead to a rapid increase in HIV transcription following the end of the eclipse phase (Cullen and Greene 1989; Feinberg et al. 1991). Interestingly, the effects of Tat are apparently also rapidly shut off following activation by a process that requires new cellular protein synthesis (Drysdale and Pavlakis 1991). The significance of this shut-off is not clear, but it could be hypothesized to lead to the development of latent infection in some cells. The Tat/ TAR system provides an elegant example of the intricate molecular mechanisms employed by HIV to regulate its interactions with its target cells in the human immune system.

Promoter Occlusion

The LTRs present at the ends of the provirus each contain the requisite _cis_-acting signals for transcriptional initiation. Thus, two questions arise: Does transcription initiate in the 3′ LTR and what, if anything, represses its use as a promoter? Transcripts initiated at the 5′ LTR generate retroviral RNAs; those initiated at the 3′ LTR would run on into an adjacent cellular sequence. Early evidence for the activity of the 3′ LTR came from the structural analysis of insertionally activated proto-oncogenes. The 3′ LTR appeared to be active in transcriptional initiation only if the insertion was accompanied by a deletion of the 5′ LTR (Fung et al. 1981; Neel et al. 1981; Payne et al. 1981). Under most circumstances, initiation only occurs from the 5′ LTR. This dominance is referred to as promoter occlusion (Adhya and Gottesman 1982). On a molecular level, the elongating transcription complexes that originate in the 5′ LTR could disrupt or prevent assembly of requisite transcription factor complexes in the 3′ LTR. Transcriptional interference has been demonstrated experimentally (Cullen et al. 1984) by placing the preproinsulin gene downstream from the 3′ LTR of an ALV provirus. The relative levels of ALV and preproinsulin RNA synthesis were compared. Efficient transcription of the ALV sequences initiated by the 5′ LTR inhibits insulin RNA synthesis. SV40 sequences introduced within the ALV genome could cause termination of transcripts initiated from the 5′ LTR and a concomitant increase in transcription originating in the 3′ LTR.

Alternative Promoters and Intragenic Enhancers

Most retroviruses contain only a single transcription unit, the one initiated at the border of U3 and R in the 5′ LTR, which generates full-length genomic RNA. Alternative promoters (Fig. 15) have now been identified in a few retroviruses. In the case of MMTV, two different additional transcription start sites have been reported. These direct synthesis of RNAs that encode the endogenous viral superantigen (sag gene). One is present in the U3 region upstream of the U3/R border (Günzberg et al. 1993). Transcripts initiated at this site in the 5′ LTR are spliced to the acceptor site for the endogenous superantigen-coding region. These transcripts are produced in B cells, suggesting a role in tissue-specific expression of superantigen. A second transcription start site has been identified in the env gene (Miller et al. 1992). Transcription from this promoter has been observed in activated T cells, apparently as a result of T-cell transcription factors that interact with adjacent putative transcription-factor-binding sites. A similar region has also been shown to be required for transcription in some B cells (Reuss and Coffin 1995).

Figure 15

Retroviral alternative promoters. The locations of alternative promoters in MMTV and HFV are shown, as are the positions of associated regulatory elements (TATA box and binding sites for cellular transcription factors).

The primate spumaretroviruses (HFV and SFV) also contain alternative promoters in the env region that direct the synthesis of bel gene RNAs (Löchelt et al. 1993). Efficient use of this intragenic transcription start site relies on the _trans_-activator function of the Bel-1 protein (Löchelt et al. 1993). Recent reports have also indicated that an internal promoter within the HTLV-1 pX gene can direct Tax expression (Nosaka et al. 1993).

Internal intragenic transcription enhancers that could modulate efficiency of transcription from the classical LTR start site, but which lie outside of the LTR, have also been identified in some retroviruses. The presence of internal enhancer signals adjacent to the 5′ LTR but not the 3′ LTR could provide another mechanism of selecting 5′ LTR sequences. Enhancers have been identified within coding sequences and introns of a number of cellular genes, and thus, it is not surprising that enhancers have been identified within non-LTR sequences. The best characterized of the intragenic retroviral enhancers has been the gag enhancer in some ASLVs (Arrigo et al. 1987). This sequence has been shown to contain three binding sites for the C/EBP transcription factor (Carlberg et al. 1988). Point mutations in the RSV gag enhancer C/EBP sites result in 50–60% reductions in viral RNA synthesis and infectivity (Ryden et al. 1993). Internal enhancer elements also have been identified within the HIV genome (Verdin et al. 1990). One of these, located within the pol gene, is composed of binding sites for AP1 that will activate transcription from heterologous promoters (Van Lint et al. 1991). Other segments of the HIV genome, including a region of the tat gene, also appear to possess weaker enhancing activity. These internal HIV enhancers, in addition to other transcription-factor-binding sites within the HIV LTR, may help to explain the relative independence of HIV replication from the LTR NF-κB enhancer elements. The fact that internal enhancers have been identified in only two genera of retroviruses may simply reflect a lack of effort to detect them in others.

The construction of retroviral vectors has dramatically demonstrated the ability of the retroviral genomes to carry internal start sites of transcription (see Chapter 9. Frequently, the expression of foreign genes is mediated by strong promoter and enhancer combinations placed into the non-LTR region of the vector. Although the orientation of these internal transcription units can influence the efficiency of their use, there is no question that these constructs can provide expression of heterologous genes.

The Host Genome Affects the Expression of Integrated Proviruses

Much of this chapter has focused on the effects of the cellular and viral _trans_-acting factors on retroviral gene expression. After integration of the provirus, gene expression also appears to be influenced in cis by the host genome. In the case of MMTV, several studies have shown that proviruses integrated at different genomic sites in otherwise isogenic cells do not have equivalent expression potential despite the presence of identical transcriptional control elements (Feinstein et al. 1982). Positional effects also are observed in mouse strains in which copies of the Mo-MLV genome are introduced into the germ line. Although these proviruses are silent during embryogenesis, expression is detected in different tissues and developmental stages that appears to be dependent on the site of integration (Jaenisch et al. 1981; Jähner and Jaenisch 1985). Positional effects may also account for HIV latency in some human T-cell lines (Winslow et al. 1993; Chen et al. 1994). Retroviral vectors that carry the β-galactosidase reporter gene also show a dependence on chromosomal location (Hoeben et al. 1991). Since the _trans_-acting factors should be identical in these cases, these differences in expression are proposed to be due to positional effects, some integration sites being more favorable for expression of the proviral genome than other sites.

Position-dependent activation of endogenous proviruses illustrates the susceptibility of integrated proviruses to chromosomal influences. In the case of endogenous MMTV loci, a chromosome rearrangement induces expression. The Mtv-8 endogenous retrovirus, which is located on mouse chromosome 6 near the immunoglobulin light-chain locus, is usually silent. However, a rearrangement of the Igκ L locus that brings the Mtv-8 promoter within 7.5 kb of the κ enhancer leads to activation of viral gene expression (Yang and Dudley 1992). These findings illustrate the long-distance effects of cellular enhancers as well as the sensitivity of proviruses to chromosomal position. A more complex situation occurs with avian endogenous viruses. Findings indicate that silencing occurs as a function of preintegration site activity. However, certain developmental conditions, including 5-azacytidine-induced demethylation, lead to activation of some otherwise silent loci (Conklin and Groudine 1986).

The molecular basis for differential silencing of proviruses is not well understood. As proposed above, the balance of positive- and negative-acting host transcription factors determines the level of transcriptional activity of a provirus. This mechanism can explain general silencing or repression of retroviral transcription as observed during introduction of Mo-MLV into preimplantation embryos or latent state of HIV proviruses in resting T lymphocytes. However, this hypothesis cannot easily explain position effects. In studies of retroviral vectors, retrieval of the DNA from the silenced locus has been shown to generate both active and inactive loci in a second round of integration (Hoeben et al. 1991). These results strongly suggest that position effects are due to epigenetic mechanisms. In many cases, methylation of silent loci occurs (Niwa et al. 1983; Berwin and Barklis 1993); however, a causal role for methylation in silencing has not been established (as discussed above). Chromosome organization or chromatin structure could also epigenetically affect the activity of integrated retroviruses. Interestingly, activation of latent HIV is associated with disruption of a nucleosome in the vicinity of the RNA start site (Verdin et al. 1993). It is also possible that a stochastic process determines the on or off state of a provirus. For example, upon infection, the frequency of assembly of transcription initiation complexes on proviral DNA may be inefficient, leaving some proviruses inactive. It is important to note, as discussed above, that proviruses inactive due to any mechanism (e.g., lack of positive-acting factors, presence of inhibitory factors, chromosome positional effects, or stochastic events) appear to be silenced more permanently by DNA methylation.

Position-dependent silencing of retroviruses has a remarkable similarity to the positional effects observed in transgenic animals bearing heterologous DNA stably integrated into the chromosome. The effect of position on the expression of the introduced genes can be dramatic. In well-studied cases in which relatively small transcriptional control regions are introduced, there is no relationship between the copy number of the transgene and its expression level (Palmiter et al. 1993). In other cases, regions of flanking sequences, extending 20–30 kb, have been shown to mediate position independence of expression (for review, see Evans et al. 1990; Wilson et al. 1990). The entire 20–30-kb region appears to contain relatively less-condensed chromatin as assayed by nuclease sensitivity. The positional independence is mediated by so-called locus control regions (LCRs) that in part consist of binding sites for transcription factors. Alternative models for LCR function include attachment sites to the nuclear matrix, topoiso-merase-binding sites, and origins of early replication. All of these roles can also be classified as referring to the poorly understood phenomena of chromosomal domains and chromatin structure. The failure of retroviral LTRs to maintain position-independent expression could be interpreted as the lack of an LCR. Alternatively, the enhancer elements of the retroviral LTR could be only partially active in mediating LCR function. In either case, LCR functions might be provided efficiently by the host at only a fraction of the viral integration sites.