Which of the following best describes eukaryotic gene expression via chromatin modifications

2.5.1 Histone variants

Histone genes are invariably present in multiple copies, the level of reiteration varying from two copies of each gene in yeast to several hundredfold in sea urchin. In some organisms, most notably the sea urchin, distinct batteries of different forms of core histone genes are transcribed at precise periods in development [Poccia, 1986]. Such variants are particularly prevalent in gametes, where the key function of the histone is to compact the DNA. Metabolic activities involving DNA are generally inhibited in spermatozoa and it is likely that DNA can be compacted in a wide variety of ways. This is because once it is compacted no significant process involving DNA will occur until fertilization [Section 2.5.5]. The sea urchin has specialized histone H2A variants [five in all] which have different expression profiles for the cleavage, blastula and gastrula stages of embryogenesis, and four different histone H2B variants that are also developmentally regulated [Table 2.2; see Section 3.4]. Sea urchin sperm has an H2B variant which has an ammo-terminal tail with a 21 amino acid extension. This tail interacts with linker DNA and may contribute to the unusual stability of sperm nucleosomes [Bavykin et al., 1990; Hill and Thomas, 1990]. Differences have also been observed in the stability of nucleosomes containing early and late sea urchin histones [Simpson and Bergman, 1980]. Interestingly, a single form of both histone H3 [excluding CENP-A homologues] and histone H4 is present throughout development, reflecting the central role of these histones in nucleosome structure [Section 2.2.4] and chromatin assembly [Sections 3.2]. Variations in the primary structure of histones H2A and H2B are likely to alter the compaction of DNA into both the nucleosome and the chromatin fibre. This could be due either to a direct effect on nucleosome structure or an altered binding of histone H1 to the nucleosome core particle [Section 2.3.1].

Table 2.2. Histone modification in development

An evolutionarily conserved variant of histone H2A that is essential for Drosophila development is H2A.vD [van Daal and Elgin, 1992]. This variant is known as H2A.Z in mammals, H2A.F/Z in chickens, and hvl in Tetrahymena [Harvey et al., 1983; Ernst et al., 1987; van Daal et al., 1988; White et al., 1988; Hatch and Bonner, 1988]. This particular H2A variant associates preferentially with actively transcribed chromatin in Tetrahymena [Stargell et al., 1993]. H2A.Z has an amino-terminal tail that is similar to that of histone H4, and it is post-translationally modified through acetylation to a greater extent than normal H2A. A second conserved variant is H2A.X [Mannironi et al., 1989]. This histone has an extended carboxyl-terminal tail beyond the histone-fold domain. Such extensions are not uncommon in variant H2A molecules. Wheat H2A1 has a 19 amino acid carboxyl-terminal extension which has the potential to protect about 20 bp of DNA immediately adjacent to the nucleosome core [Lindsey et al., 1991]. This is again consistent with the proposed position of this domain within the nucleosome [Section 2.2.4]. Until recently, core histone variants were believed to contain the most extreme changes from a normal histone architecture that might be incorporated into a nucleosome. However, these variants are now recognized as only one example of a family of proteins that share the histone-fold structure and that might be incorporated into a nucleosomal architecture, although these proteins contain wide deviations from normal histone sequences [Baxevanis et al., 1995].

The core histones have evolved from a DNA-binding protein that contained only the three α-helices of the histone-fold domain and lacked any tail domains [Baxevanis et al., 1995]. The archaebacterial protein HMf consists of only the histone-fold domain and wraps DNA around itself within nucleosome-like structures [Sandman et al., 1990]. The eukaryotic core histones retained this property, but added the capacity of the assembled nucleoprotein complex to interact with other proteins outside the nucleosome through the addition of the tail domains. Other regulatory proteins have made use of their histone-fold domains to confer specialized properties on individual nucleosomes by replacing normal histones within chromatin [Sullivan et al., 1994; Stoler et al., 1995; Shelby et al., 1997]. These include the deviant histones: CENP-A, which is found at the centromere, and rat macro H2A [Fig 2.42]. The function of macro H2A is not yet known, although an extended C-terminal tail has been added to the histone-fold domain of H2A [Pehrson and Fried, 1992]. This tail contains the leucine zipper, which is a dimerization motif often found in transcription factors. Macro H2A might interact with a sequence specific DNA-binding protein to tether a specific nucleosomal structure at a defined site.

More extreme variations from normal histone sequence are found in transcriptional regulatory proteins. These proteins maintain the histone-fold domain and make use of it both to direct specific protein–protein interactions and to bind DNA [Kokubo et al., 1993; Sinha et al., 1995]. Regulatory proteins with these properties include the important components of the basal transcription factor TFIID, known as TATA-binding-protein associated factors [TAF]II60 and -40, and the related CCAAT-box-binding proteins, NF-Y [CBF] and HAP2, -3 and -5 [Fig. 2.42]. TAFII40 resembles H3 and TAFII60 resembles H4. Both proteins have extended C-terminal tails that interact with other components of TFIID and transcriptional activators. It has been proposed that TAFn40 and TAFII60 participate in the assembly of nucleosome-like structures, excluding normal histones from the TATA box, yet maintaining DNA in a semi-compacted state competent for transcription [Kokubo et al., 1993; Nakatani et al., 1996; Xie et al., 1996]. Metazoan NF-Y [CBF] and S. cerevisiae HAP2, -3′ and -5 are highly related trimeric proteins. The evolutionarily conserved peptide sequences of two of the subunits [CBF-A and CBF-C, or HAP3 and HAP5] resemble the histone-fold domains of histones H2B and H2A, respectively [Fig. 2.42]. These domains are essential for DNA binding in the presence of the third protein [CBF-B or HAP2] that confers sequence specificity [Sinha et al., 1995]. The NF-Y [CBF] and HAP2, -3 and -5 proteins are transcriptional activators.

It appears advantageous for a large number of eukaryotic DNA-binding proteins to retain their histone-like character [Wolffe and Pruss, 1996a]. This requirement might be similar to the architectural role of the HMG box in the assembly of large regulatory nucleoprotein complexes [Grosschedl et al., 1994; Section 2.5.8]. In certain instances, it is possible to facilitate nucleoprotein-complex assembly through the use of a generic HMG-box protein, such as HMG1, which alters or stabilizes a bend in the double helix. In more specialized cases, a specific transcription factor has evolved that incorporates a structure related to HMG1 that bends DNA, but that also fulfils additional functions by virtue of interaction with other transcription factors. Thus, it might be important to retain DNA in a relatively compact structure within a nucleosome-like architecture, but also useful for the histone-like protein to assume other more specialized functions.

Human centromeric chromatin contains a highly deviant histone called CENP-A. This has significant identity over the histone-fold domain to histone H3, yet has a very different N-terminal tail [Fig 2.42; Section 2.4.2]. CENP-A is found within nucleosomes and heterodimerizes with H4 [Sullivan et al., 1994; Palmer et al., 1987]. Importantly, Sullivan and colleagues discovered that the histone-fold domain of CENP-A targeted the protein to the centromere. This result implies that other histone variants, such as H2A.Z, might be targeted to particular DNA sequences, because the DNA-binding surfaces of the histone-fold domains show a similar number of sequence differences compared to the normal somatic histone H2A as seen between histone H3 and CENP-A [Fig. 2.43].

CENP-A might have arisen through the necessity of having a specialized nucleosomal structure at the centromere where the Nterminal-tail domain makes highly selective contacts with other centromeric components, such as the large hydrophilic DNA-binding proteins CENP-B and CENP-C [Fig. 2.44; Section 2.4.2]. Experimental support for this hypothesis comes from genetic experiments in S. cerevisiae where a CENP-A homologue, CSE4 is essential for correct sister chromatid segregation [Stoler et al., 1995]. This suggests that nucleosomes that include CENP-A will assemble a specialized differentiated chromosomal domain that might serve to facilitate attachment of the kinetochore and the function of the centromere in the segregation of chromosomes. The assembly of a differentiated domain at the centromere establishes an interesting precedent for the targeting of core histone variants to particular sites within chromatin.

The form of linker histone varies during sea urchin development and during that of many other organisms, often in a tissue-specific way [Risley and Eckhardt, 1981; Wolffe, 1991a; Khochbin and Wolffe, 1994; Grossbach, 1995; Section 3.4]. At least six variants of histone H1 are present during sea urchin embryogenesis. Like histone H2A, there are distinct cleavage stage, blastula and gastrula proteins. Interestingly, these latter proteins do not contain short peptide sequences known as ‘SPKK motifs’ [see Section 2.5.3]. These sequences are found multiple times in the tails of linker histones and are the sites of phosphorylation by the major mitotic kinase in the cell [called the cdc 2 or MPF kinase] [Section 2.5.3]. Hence the gastrula form of histone H1 in the sea urchin cannot be phosphorylated by this particular kinase. Once again, an exact role for the different histone H1 variants has not yet been established, but it is clear that the chromatin of the early embryo is less compacted than that of the gastrula [Longo, 1972] and may therefore be more accessible to trans-acting factors and more easily replicated. It is also possible that the synthesis of histone H1 variants that cannot be phosphorylated by the MPF protein kinase during embryogenesis reflects a change in the mechanism by which chromosome structure is regulated during the cell cycle [Section 2.5.3].

Linker histones, such as histone H1, have been shown to direct the exact positioning of nucleosomes with respect to DNA sequence [Meersseman et al., 1991]. This positioning relies on the sequence and structure selective recognition of DNA by the linker histone, and protein–protein contacts made between the winged helix domain of the linker histone and the histone-fold domains of the core histones [Ura et al., 1995, 1996; Section 2.3.1]. The mouse serum albumin enhancer exists in the active state within an array of precisely positioned nucleosome-like particles [McPherson et al., 1993]. Specific enhancer-binding factors, including the winged-helix protein HNF3, are part of the nucleosome-like particles and HNF3 can actively direct their positioning with respect to DNA sequence. These observations indicate that HNF3 replaces linker histones within chromatin containing the serum albumin enhancer [McPherson et al., 1993], thereby establishing a precise regulatory nucleoprotein architecture [Fig. 2.45]. The replacement of histone H1 by HNF3 would be analogous to the replacement of histone H3 by CENP-A at the centromere. In both instances, a sequence-selective histone-like regulatory protein would direct the assembly of a differentiated chromatin domain. This might encompass a single variant nucleosome, as at the serum albumin enhancer, or long arrays of variant nucleosomes, as at the mammalian centromere.

What are the advantages for regulatory proteins of maintaining a histone-like structure within a nucleosomal architecture? One immediate advantage is the stability of the nucleosome during the cell cycle. Histones H3 and H4 do not exchange out of chromatin outside of S-phase, moreover, molecular mechanisms exist to reassemble nucleosomes efficiently on newly replicated DNA [Jackson, 1990; Section 4.3]. Thus, CENP-A could maintain a very stable association with the centromere even through the replication process. Histones H2A and H2B do exchange out of chromatin, but do so predominantly during transcription [Jackson, 1990]. It is very difficult to disrupt core histone interactions within a nucleosome in vivo. Regulatory proteins that assemble in association with the core histones will be resistant to displacement from DNA by competing protein–DNA interactions. Linker histones have a much less stable association with nucleosomal DNA. Relatively weak interactions offer the potential for a role in regulatory events where transcription needs to be reversibly activated. For example HNF3 might replace linker histones on the serum albumin enhancer to prevent their repressive influence on transcription [Cirillo et al., 1998; Shim et al., 1998 see Fig. 2.45].

A second advantage of transcription factors resembling the histones lies in the utilization of nucleosomal architecture. As the physical analysis of chromatin clearly demonstrates, nucleosomal arrays do efficiently self-assemble into higher-order chromatin structures. The absence of a nucleosome or the presence of a large non-nucleosomal nucleoprotein complex will probably interfere with this self-assembly process [Hansen and Ausio, 1992]. Thus, the requirement for the assembly of higher order chromatin structures might impose constraints on the properties of nucleoprotein complexes in particular chromosomal regions. The easiest way to maintain higher order structure, yet have functional specialization, would be to modify the proteins within a nucleosome, while maintaining the basic function of DNA compaction. A distinct feature of histone interactions with nucleosomal DNA is the exposure of DNA on the surface of the nucleosome. One side of DNA is occluded on the histone surface, but the other is exposed and potentially accessible to other regulatory proteins. Thus, histone-like interactions with DNA could facilitate the assembly of multicomponent complexes [Truss et al., 1995].

2.5.2 Post-translational modification of core histones

Core histones undergo two major post-translational modifications: acetylation and phosphorylation. Both have been the subject of intense interest. Core histone acetylation is now believed to have an integral role in the transcription process because some transcriptional activators have histone acetyltransferase activities [Brownell et al., 1996] and some transcriptional repressors are histone deacetylases [Taunton et al., 1996] [see Section 2.5.4]. Acetylation of the four core histones occurs in all animal and plant species examined [Csordas, 1990]. The sites of modification are the lysine residues of the positively charged amino terminal tails [Section 2.1.2], where each acetate group added to a histone reduces its net positive charge by 1. The number of acetylated lysine residues per histone molecule is determined by an equilibrium between histone acetylases and deacetylases. Two populations of acetylated histone appear to exist in a particular cell nucleus. For example, in the embryonic chicken erythrocyte, 30% of core histones are stably acetylated while the acetylation status of about 2% changes rapidly. The pattern of specific lysine residues in the histone tails that are acetylated varies between different species. This non-random usage suggests that considerable sequence specificity exists for the relevant acetylases and deacetylases [Turner, 1991]. Histone H4 that contains three or more acetylated lysines is described as hyperacetylated, whereas the protein containing one or no acetylated lysines is described as hypoacetylated.

Hyperacetylation of the histone tails leads to relatively subtle changes in nucleosome conformation [Bode et al., 1983; Oliva et al., 1990]. However, it appears that the most significant consequences are for protein–protein interactions, either between nucleosomes, with histone H1 or with non-histone proteins. The amino terminal tails of the core histones are accessible to trypsin, suggesting that they are exposed on the outside of the nucleosomal core particle [Section 2.2.4]. Chemical cross-linking experiments have shown that weak interactions can occur between the N-terminal domain of H2B and linker DNA, although these experiments used a special sperm H2B with a particularly long N-terminal tail [Bavykin et al., 1990; Hill and Thomas, 1990]. The amino termini of histones H3 and H4 also interact with core-particle DNA. High resolution protein nuclear magnetic resonance data indicate association of the amino terminal tails of histones H3 and H4 with DNA in the nucleosome core particle at physiological ionic strength [< 0.3 M NaCl]. In contrast, the tails of the normal somatic variants of histones H2A and H2B are relatively mobile at all ionic strengths examined [Cary et al., 1982; but see Lee and Hayes, 1997]. The weak interaction of the histone tails with DNA in the nucleosome is reflected in the lack of structural change in the organization of DNA and in the integrity of the nucleosome following their proteolytic removal [Ausio et al., 1989; Hayes et al., 1991b; but see Lilley and Tatchell, 1977]. Acetylation or removal of the histone tails from nucleosomes that contain recognition sites for trans-acting factors can facilitate factor access to these sites [Lee et al., 1993; Vettese-Dadey et al., 1996]. This result suggests that histone acetylation might have a major regulatory role in the transcription process. In addition, the acetylation of the histone tails or removal of the histone tails from arrays of nucleosomes deficient in linker histones, impedes the capacity of these arrays to compact into higher-order structures [Garcia-Ramirez et al., 1992, 1995]. These observations suggest that a major influence of the histone tails on chromatin structure may be through their influence on the integrity of the chromatin fibre [Section 2.3.1]. Others report that histone hyperacetylation appears to have less effect on the assembly of higher-order structures [McGhee et al., 1983b; Ridsdale et al., 1990]. However, these older studies use heterogeneous preparations of chromatin whereas the more recent work of Garcia-Ramirez et al., utilize well-defined preparations of synthetic chromatin reconstituted in vitro. Subtle changes in chromatin fibre stability might have large effects on transcriptions. The formation of compacted structures in vitro can severely impede both the initiation of transcription by RNA polymerase and elongation by the polymerase [Hansen and Wolffe, 1992, 1994; see.Fig. 2.32].

Induction of histone hyperacetylation with the deacetylase inhibitor sodium butyrate increases the accessibility of DNA in chromatin to DNase I and can improve expression of transfected DNA [Gorman et al., 1983]. High levels of histone acetylation improve chromatin solubility, suggesting a reduced tendency to aggregate [Perry and Chalkley, 1981]. This also implies that higher order chromatin structures containing acetylated nucleosomes are less stable [Perry and Annunziato, 1989]. Maintenance of histone acetylation in nascent chromatin immediately after the replication fork might also contribute to a reduction in the stable sequestration of linker histones [Reeves et al., 1985; Perry and Annunziato, 1989; Ridsdale et al., 1990]. However in a purified system histone hyperacetylation does not directly influence the association of histone H1 with nucleosomes [Ura et al., 1994, 1997]. Consistent with these in vitro observations constitutive hyperacetylation of the core histones does not reduce the association of histone H1 with bulk chromatin [Dimitrov et al., 1993; Almouzni et al., 1994], however it inhibits the complete condensation of interphase chromatin [Annunziato et al., 1988].

Bradbury and colleagues have quantitated the number of superhelical turns introduced into DNA per nucleosome in a purified system, dependent on whether the core histones are acetylated. They found that fewer superhelical turns are introduced by a population of nucleosomes when the core histones are acetylated [Norton et al., 1989, 1990]. Detailed analysis reveals that the helical periodicity of DNA in the nucleosome is unchanged by histone acetylation. This suggests, by elimination, that the path of DNA between nucleosomes or writhe of DNA on the histone core is affected by histone acetylation [Bauer et al., 1994]. These changes could also have important consequences for trans-acting factor access to DNA in the nucleosome. An influence of the histone tails on the path of linker DNA could explain their role in the assembly in vitro of stable higher-order chromatin structures [Garcia-Ramirez et al., 1992, 1995].

The generation of antibodies against acetylated histones has allowed a number of general correlations to be made concerning possible functional roles of histone acetylation. In Tetrahymena, Xenopus and humans there is excellent evidence for the presence of acetylated histones in chromatin immediately following replication [deposition-related] [Lin et al., 1989]. It has long been known that histone H4 is diacetylated in the cytoplasm immediately after synthesis [Ruiz-Carrillo et al., 1975]. Histone acetylation appears to play an important role in facilitating chromatin assembly [see Section 3.2 for discussion]. There is also a strong correlation between histone acetylation and the transcriptional activity of chromatin [Allfrey et al., 1964; Gorovsky et al., 1973; Mathis et al., 1978]. In S. cerevisiae most of the genome is transcriptionally active and contains hyperacetylated core histones [Clark et al., 1993a]. However, there are also transcriptionally inactive domains of chromatin in yeast, such as the silent mating type cassettes and telomeric sequences. These contain histone H4 that is hypoacetylated except at one position lysine 12 [Braunstein et al., 1993, 1996]. In higher eukaryotes, acetylation of histone H4 increases during the reactivation of transcription in the initially inactive chicken erythrocyte nucleus following the fusion of the erythrocyte with a transcriptionally active cultured cell to form a heterokaryon [Turner, 1991; Section 3.1]. Histone acetylation is particularly prevalent over specific genes that are actively transcribed in erythrocytes [Hebbes et al., 1988]. More recent studies have demonstrated convincingly that histone hyperacetylation is actually restricted to the domain of chromatin that contains the potentially active chicken β-globin gene locus [Hebbes et al., 1992, 1994]. This result is indicative of a very specific targeting of histone acetyltransferase. Structural data indicate that the hyperacetylated histone tails retain some contact with DNA in vivo at the globin gene locus [Ebralidse et al., 1993]. Immunolabelling of polytene chromosomes in Chironomus and Drosophila also reveals a non-random distribution of histone H4 acetylation correlating with transcriptional activity [Turner et al., 1992; Bone et al., 1994]. Within female mammals the transcriptionally inactive X chromosome is distinguished by a lack of histone H4 acetylation [Jeppesen and Turner, 1993]. Therefore, several independent experimental approaches have shown that actively transcribed and potentially active chromatin domains are selectively enriched in hyperacetylated histones, whereas transcriptionally inactive chromatin contains hypoacetylated histones.

These observations have been dramatically raised in significance following the discovery that transcriptional activators and repressors exist that acetylate or deacetylate the core histones [see Section 2.5.4]. Moreover genetic and biochemical experiments have suggested other interesting functions for histone acetylation.

Replacement of all four acetylatable lysines in the H4 tail in S. cerevisiae, with arginine such that basic charge is maintained leads to extremely slow growth, whereas substitution with glutamine, which mimics an acetylated lysine leads to a delay in G2/M progression [Meegee et al., 1995]. None of these mutations alters the eventual assembly of replicated DNA into nucleosomes, therefore the acetylation and deacetylation of lysines in the H4 tail appear necessary for cell cycle progression itself. The N-terminal tails of histones H3 and H4 have redundant functions in the chromatin assembly process whereas they have quite specific functions in transcription [Wan et al., 1995; Lenfant et al., 1996; Ling et al., 1996]. The essential role of specific histone modification might reflect the requirement for structural changes in chromatin necessary for the transcription of genes that regulate or drive the cell cycle. However, the aberrant cell cycle characteristics could also be related to other check-points that monitor chromosome integrity. In particular the mutations in histone H4 increase reliance on DNA-damage-sensitive cell-cycle check-point controls [Meegee et al., 1995], suggesting that increased DNA damage occurs in the H4-mutant cells. How might histone H4 acetylation be involved in cell-cycle check-point control, DNA damage and chromosome repair?

Newly synthesized histones H3 and H4 are acetylated [Ruiz-Carrillo et al., 1975; Chang et al., 1997] and deacetylated shortly after their incorporation into the nascent chromatin assembled immediately after replication [Jackson et al., 1976]. The histone acetyltransferase and deacetylase involved in these modifications have been characterized [Taunton et al., 1996; Parthun et al., 1996]. These enzymes interact with H3 and H4, and appear to share a common subunit known in mammalian cells as p48/p46. The molecular chaperone involved in the assembly of chromatin on newly replicated DNA is CAF1, which also interacts with H3 and H4 and contains p48/p46 [Verreault et al., 1996; Gaillard et al., 1996]. Dynamic alterations in H3 and H4 acetylation might be necessary to drive the exchange of p48/p46 between the acetyltransferase, deacetylase and CAF1. These dynamic transitions will not occur if the lysine residues in histone H4 are mutated [Meegee et al., 1995]. A failure to mediate these transitions might in turn impact on the role of CAF1 in the repair of damaged chromatin. For instance CAF1 might be irreversibly sequestered on nascent chromatin and not be available to facilitate chromosomal repair on damaged DNA. Alternatively if DNA damage occurs more readily, because of alterations in chromatin compaction following from the inability to appropriately acetylate or deacetylate histones, then inappropriate sequestration of CAF1 and p48/p46 on damaged DNA might interfere with cell cycle progression due to a decrease in the rate of chromatin assembly [Figure 2.46]. It is dangerous for a cell to synthesize naked DNA in the absence of chromatin assembly. This is because of the multiple roles of chromatin both in constraining inappropriate gene activity and in directing the appropriate packaging of DNA into chromosomes. Consequently it might be anticipated that molecular mechanisms will exist to monitor chromosomal integrity.

The association of histone acetylation with transcription [see also Section 2.5.4] and the maintenance of chromosomal integrity point to a central biological role for this regulatable modification within chromatin. A fusion protein generated by a recurrent chromosomal translocation associated with acute myeloid leukemia incorporates two putative acetyltransferase domains [Borrow et al., 1996; Ogryzko et al., 1996]. This suggests that aberrant histone acetylation might contribute to cellular transformation. The two genes fused in this translocation encode the coactivator/histone acetyltransferase CBP [see Section 2.5.4] and MOZ [for monocytic leukemia zinc finger] protein. MOZ contains both the CBP acetyltransferase domain and a region of identity with a yeast protein involved in transcriptional silencing known as SAS2, which shares homology with the Gcn5p protein in the acetyltransferase catalytic domain [Reifsnyder et al., 1996]. In addition, the P/CAF acetyltransferase interacts with the p300/CBP acetyltransferase at the same interface as the adenovirus oncoprotein EIA, such that EIA can modulate the association of these proteins [Yang et al., 1996; Ogryzko et al., 1996] and potentially their function. A connection between histone acetylation and cell differentiation has long been known. Histone deacetylase inhibitors such as sodium butyrate and Trichostatin A both promote cell lines to differentiate [Yoshida et al., 1987] and restrict cell transformation [Sugita et al., 1992]. These drugs also induce defects during early vertebrate embryogenesis [Almouzni et al., 1994]. Clearly inappropriate changes in acetylation patterns might contribute to loss of the differentiated phenotype and cell transformation. How might aberrant acetylation contribute to such events?

Many controls in early vertebrate development depend on the capacity to establish the stable functional differentiation of chromosomal domains: for example the imprinting of chromosomes dependent on parental origin. These epigenetic effects are known to contribute to control of growth and tumorigenesis [Reik and Surani, 1989]. Maintenance of histone acetylation states provides an excellent mechanism for the propagation of stable chromosomal imprints determining gene activity [see also Section 4.3]. This is because [1] the distributive segregation of nucleosomes during DNA replication will ensure that the parental histone acetylation states are present on both daughter chromatids [Perry et al., 1993], and [2] states of chromosomal acetylation are preserved through mitosis [Lavender et al., 1994]. A speculative model for the maintenance of elements of chromatin structure through the cell cycle [Fig. 2.47] would involve a causal role either for histone acetylation states within the nucleosome itself, or for proteins that specifically recognize particular acetylation states and that might segregate in association with the core histones. Strong candidates for such regulatory molecules include the coactive ators/histone acetyltransferases themselves. Once segregated, the histone acetyltransferases would spread the appropriate state of acetylation over a contiguous imprinted domain of chromatin. Disruption of these imprints by expression or localization of a dysfunctional histone acetyltransferase would therefore be expected to contribute to cellular transformation [Wade et al., 1997].

The second type of core histone modification to receive extensive experimental study is phosphorylation. Histone H3 is rapidly phosphorylated on serine residues within its basic amino terminal domain, when extracellular signals such as growth factors or phorbol esters stimulate quiescent cells to proliferate [Mahadevan et al., 1991]. The basic amino terminal domain of histone H3, like that of histone H4, may interact with the ends of DNA in the nucleosomal core particle and therefore perhaps with histone H1 [Glotov et al., 1978]. Several studies have suggested a change in either nucleosomal conformation or higher-order structure within the chromatin of the proto-oncogenes c-fos and c-jun following their rapid induction to high levels of transcriptional activity by phorbol esters [Chen and Allfrey, 1987; Chen et al., 1990]. DNase I sensitivity of chromatin rapidly increases and proteins with exposed sulphydryl groups accumulate on the protooncogene chromatin. The proteins containing exposed sulphydryl groups include both non-histone proteins, such as RNA polymerase, and molecules of histone H3 with exposed cysteine residues. The histone H3 cysteine residues, the only ones in the nucleosome, are normally buried within the particle. Exposure of the sulphydryl groups might imply a major disruption of nucleosome structure, for example the dissociation of an H2A/H2B dimer might allow access from solution to this region of histone H3. Phosphorylation and acetylation of histone H3 might act in concert to cause these changes. There is likely to be an important yet currently unexplored link between cellular signal transduction pathways and chromatin targets.

In vivo phosphorylation of H4 and H2A occurs in the cytoplasm shortly after histone synthesis [Sung and Dixon, 1970; Ruiz-Carillo et al., 1975; Jackson et al., 1976; Dimitrov et al., 1994]. The phosphorylation of these histones, together with the diacetylation of histone H4, may selectively target them to the molecular chaperones involved in nucleosome assembly at the replication fork [Kaufman and Botchan, 1994; Wade et al., 1997]. Histone H2A.X is also phosphorylated and has been proposed to have a role in nucleosome spacing during chromatin assembly on naked DNA [Kleinschmidt and Steinbeisser, 1991]. H2A.X is a specialized variant synthesized outside of the S phase of the cell cycle [Mannironi et al., 1989], which is stored in Xenopus oocytes [Dimitrov et al., 1994]. Phosphorylated H2A.X accumulates in decondensing Xenopus sperm chromatin such that it eventually represents 50% of the total H2A in the paternal pronucleus. Nevertheless, removal of the modification using phosphatases does not influence the spacing of nucleosomes [Dimitrov et al., 1994]. Thus the function of phosphorylation in chromatin assembly remains enigmatic.

Core histones are also methylated on their lysine residues without clearly defined functional consequences. However, it is likely that any inhibition of lysine acetylation would contribute to transcriptional repression. Most methylation in vertebrates occurs on histone H3 at lysines 9 and 27 and histone H4 at lysine 20 [reviewed by Annunziato et al., 1995]. Histone H3 is found with up to three methyl groups on each lysine, while lysine 20 in histone H4 maximally contains two methyl groups. Methylation of lysines begins after nucleosome assembly and reaches peak levels at mitosis. In interphase, histone methylation is preferentially targeted to histones H3 and H4 that are already acetylated. This may however only reflect the relative accessibility of the acetylated N-terminal tails. Some evidence suggests that ADP-ribosylation of core histones may lead to localized unfolding of the chromatin fibre. ADP-ribosylation may play a particularly important role in DNA repair [Section 4.4]. Here the disruption of chromatin structure cannot always rely on the processive enzyme complexes involved in DNA replication or transcription. The synthesis of long negatively charged chains of ADP-ribose may well facilitate a partial disruption of nucleosomes, presumably by exchange of histones to this competitor polyanion. Histone H2B and especially H2A can also be modified by addition of the small protein ubiquitin [West and Bonner, 1980]. Ubiquitin has been found to participate in regulating protein degradation. The protein is covalently attached, via an ATP-dependent reaction, to a protein to be targeted for proteolysis.

Ubiquitin is a 76-amino acid peptide that is attached to the C-terminal tail of histone H2A and H2B. Ubiquitinated H2A is incorporated into nucleosomes, without major changes in the organization of nucleosome cores [Levinger and Varshavsky, 1980; Kleinschmidt and Martison, 1981]. Since the C-terminus of histone H2A contacts nucleosomal DNA at the dyad axis of the nucleosome [Guschin et al., 1991; Usachenko et al., 1994], ubiquitination of this tail domain might be anticipated to disrupt the interaction of linker histones with nucleosomal DNA. The bulky ubiquitin adduct might also be anticipated to disrupt higher-order chromatin structures by impeding internucleosomal interactions.

The cell cycle provides an additional useful context with which to consider all these histone modifications and transcription as a whole [Fig. 2.48]. RNA polymerase I and II mediated transcription are severely inhibited during mitosis [Johnson and Holland, 1965; Morcillo et al., 1976]. This inhibition involves the post-translational modification of components of the basal transcriptional machinery as well as chromatin structural components [Hartl et al., 1993]. During S phase, all the core histones are acetylated but the predominant modifications are mono- and diacetylation of histones H3 and H4 [Waterborg and Matthews, 1982, 1984 see above]. In G2, histones H3 and H4 become hyperacetylated and in mitosis all four core histones are deacetylated. With respect to phosphorylation, histone H2A is phosphorylated throughout the cell cycle [Gurley et al., 1978], histone H3 is phosphorylated during mitosis [Paulson and Taylor, 1982] and histone H1 phosphorylation occurs throughout S phase, increases during G2 and becomes maximal at metaphase with 22–24 phosphates per H1 molecule [Mueller et al., 1985; see Section 2.5.3]. ADP ribosylation and ubiquitination are present through S phase, becoming maximal in G2 [Kidwell and Mage, 1976; Mueller et al., 1985]. These modifications also decline during mitosis. These results indicate the dynamic nature of these modifications, and the necessity of reconfiguring the chromosome continually during the cell cycle.

2.5.3 Linker histone phosphorylation

A widely studied post-translational modification of chromatin is the reversible phosphorylation of histone H1. This modification varies through the mitotic cell cycle [Fig. 2.48]. Studies of both the slime mould Physarum and mammalian cells in culture show that phosphorylation of H1 is highest in rapidly dividing cells and decreases in non-proliferating cells; levels of histone H1 phosphorylation are lowest in G1 and rise during S phase and mitosis. During mitosis, phosphorylation of histone H1 peaks at metaphase when chromosomes are at their most condensed. These results have led to the suggestion that a causal relationship exists between histone H1 phosphorylation and chromosome compaction [Bradbury et al., 1974].

Histone H1 consists of a globular central domain flanked by lysine-rich highly basic amino terminal and carboxyl terminal tails [Section 2.3.1]. The globular domain interacts with DNA in contact with the core histones, whereas the tails bind to linker DNA. Phosphorylation of the histone H1 tails occurs predominantly at conserved [S/T P-X-K/R, serine/threonine, proline, any amino acid, lysine/arginine] motifs of which several exist along the charged tail regions [Churchill and Travers, 1991]. The carboxyl-terminal tail has the capacity to adopt an extended α-helical structure [Clark et al., 1988]. The central structured domain of histone H1 is not phosphorylated [Langan, 1982]. It might be expected that the neutralization of positive charge on the tails would weaken the interaction of histone H1 with the linker DNA. This might be a prerequisite for chromosome condensation, but is also paradoxical since the association of histone H1 with linker DNA has been thought to direct the folding of nucleosomal arrays into the chromatin fibre [Section 2.3.2]. The importance of linker histones in the assembly of chromosomes and nuclei has recently been examined through cell biological approaches in which they have been depleted from Xenopus egg extracts. It was found that it is possible to assemble mitotic chromosomes and functional nuclei in the complete absence of linker histones [Ohsumi et al., 1993; Dasso et al., 1994b]. These experiments unambiguously establish that phosphorylated linker histones are not essential for the chromosome compaction essential for nuclear function.

In order to determine the significance of linker histone phosphorylation for chromosomal function it has been useful to examine systems in which mitosis and chromosome compaction are uncoupled [Roth and Allis, 1992]. Tetrahymena is a ciliated protozoan in which two distinct nuclei exist differing in structure, function and mitotic behaviour. The somatic macronucleus is responsible for maintaining cell growth, is transcriptionally active and divides amitotically without any apparent condensation of chromatin. Surprisingly, however, macronuclear H1 phosphorylation is controlled through a kinase [the cdc 2 or MPF activity] that is similar to that regulating the cell cycle in normal mammalian cells. Nevertheless, the activity of this kinase and chromosome condensation can be uncoupled. The phosphorylation state of macronuclear histone H1 is highly dependent on cell growth conditions. If the cells are starved, growth ceases and histone H1 is moderately dephosphorylated. More significantly, during conjugation the macronucleus becomes completely inert, chromatin condenses and histone H1 is completely dephosphorylated [Lin et al., 1991]. Thus phosphorylation of H1 is inversely related to chromosome condensation. In contrast to the macronucleus, the germ-line micronucleus which is responsible for the sexual cycle is normally transcriptionally silent and undergoes a normal mitotic cycle including the formation of mitotic chromosomes [Gorovsky, 1986].

Several very useful studies correlating nuclear function and histone modification have been carried out comparing these two nuclei within the single Tetrahymena cell – a natural heterokaryon [Section 3.1.2]. Briefly, special variants of histone H2A are present in the transcriptionally active macronucleus but absent from the micronucleus. Histones are also more extensively acetylated in the macronucleus [Section 2.5.2]. The association of the linker histone H1 with chromatin in the macronucleus also decreases with transcriptional activity. This macronuclear histone H1 is highly phosphorylated during vegetative growth. In micronuclei, macronuclear histone H1 is replaced by four specialized linker histone polypeptides [Allis and Gorovsky, 1981; Roth et al., 1988]. One of these linker histone polypeptides becomes heavily phosphorylated on transcriptional activation of the micronucleus during the sexual cycle in Tetrahytnena [Sweet et al., 1996].

A second system in which linker histone phosphorylation can be uncoupled from mitosis concerns the function of the specialized linker histone H5 during development of chicken erythroid cells. During the final stages of chicken erythrocyte development, the nucleus is condensed into inactive heterochromatin due in part to the appearance of histone H5 [Section 2.5.6]. Topoisomerase II also becomes much reduced during this differentiative process [Section 2.4.2]. Newly synthesized histone H5 is highly phosphorylated, but when the erythrocyte chromatin becomes condensed, histone H5 is quantitatively dephosphorylated. Hence once again dephosphorylation of a linker histone correlates with chromatin compaction [Aubert et al., 1991]. That these two events are directly linked receives further support from experiments in which the gene for histone H5 was expressed in fibroblasts. This specialized linker histone would not normally be found in these cells. The accumulation of histone H5 in the fibroblasts inhibited cell growth concomitant with chromatin compaction [see also Sun et al., 1989]. Under these circumstances histone H5 was not phosphorylated. Introduction of the protein into transformed cells led to phosphorylation of histone H5. In this case nuclear condensation did not occur and the cells continued to grow and divide. Phosphorylation of the linker histone clearly prevents chromosome folding as might be expected from biophysical analysis [Section 2.3.1].

The final example of a correlation between linker histone phosphorylation and chromatin compaction concerns sea urchin spermatogenesis. Here, dephosphorylation of a sperm-specific histone H1 correlates with chromatin condensation [Green and Poccia, 1985]. Following fertilization, sperm histone H1 is phosphorylated in parallel with decondensation of the sperm pronucleus. In all of these examples we see that histone H1 dephosphorylation correlates with chromosome compaction. The co-ordinate phosphorylation of non-histone proteins at the same time as histone H1 during the cell cycle seems more likely to regulate the compaction of chromatin during mitosis [see below].

Phosphorylation of histone H1 has been shown directly to weaken interaction of the basic tails of the protein to DNA. Surprisingly, these changes influence the binding of the protein to chromatin even more than to DNA [Hill et al., 1991]. Perhaps phosphorylation of histone H1 is required to weaken the interaction of the linker histone with chromatin and thereby ‘loosen’ the chromatin fibre to allow other trans-acting factors required for gross changes in chromosomal architecture to bind to DNA or the fibre itself. For example, such proteins might include SMC proteins such as XCAP-C/E [see Section 2.4.2].

Characterization of the major mitotic kinase [cdc 2 or MPF] in eukaryotic cells has allowed many of the nuclear events driven by phosphorylation to be defined [Dunphy and Newport, 1988]. In the course of these studies it became clear that MPF was similar, if not identical, to the major histone H1 kinase in eukaryotic cells [Langan et al., 1989]. During mitosis in higher eukaryotes, MPF induces the ultrastructural changes required for cell division, including nuclear envelope disassembly [nuclear membrane and lamina], chromatin condensation and construction of the mitotic spindle. Although histone H1 becomes hyperphosphorylated during mitosis, it is clearly not the only substrate for MPF during the cell cycle. Newport, Gerace and colleagues have shown that disassembly of the nucleus, nuclear membrane vesicularization, lamin disassembly and chromosome condensation are all independent processes [Ohaviano and Gerace, 1985; Newport and Spann, 1987; Newport et al., 1990]. The disassembly of the nuclear lamina appears to be driven by phosphorylation. It is quite possible that phosphorylation of the other proteins found in the nuclear scaffold fraction, including Sc I [topoisomerase II], Sc II [XAP-C/E] and Sc III, may influence chromatin and chromosome folding.

2.5.4 Activators and repressore that make use of chromatin modifications to regulate transcription

The acute regulation of transcription in response to the addition or withdrawal of inductive agents, such as hormones and nutrients, requires the rapid turn-on and turn-off of gene activity. This type of gene regulation is particularly prevalent in the yeast, Saccharomyces cerevisiae where the entire organism is continually responding to changes in the environment. The elegant genetic dissection of transcriptional control circuits in S. cerevisiae has recently been complemented by substantial progress in our understanding of the biochemistry of transcriptional regulation in metazoans. Remarkable similarities have been found to exist between S. cerevisiae and metazoans in the molecular mechanisms used to reversibly regulate gene activity in response to diverse signalling pathways. These mechanisms have been found to employ the modification of chromatin as a central element in gene control.

Central to this grand unification of mechanism is the observation that diverse sequence-specific transcription factors such as steroid and nuclear hormone receptors, Mad/Max, c-Jun/v-Jun, C-Myb/v-Myb, c-Fos, MyoD and CREB utilize a limited number of transcriptional coactivators and/or corepressors to effect their regulatory functions. Thus, the coactivators and corepressors integrate diverse regulatory signals to determine gene activity. Transcriptional coactivators and corepressors have multiple activities that together contribute to their regulatory function. They have the capacity to interact both with the regulatory domains of sequence-specific transcription factors and with the basal transcriptional machinery [Section 4.1]. In addition coactivators and corepressors directly modify the chromatin environment within which the transcriptional machinery functions. In fact it appears that the transcriptional machinery requires a chromatin environment in which to function most effectively. Coactivators and corepressors have been found to utilize chromatin to amplify the range of transcriptional regulation far beyond what might be achieved on naked DNA [Wolffe et al., 1997b].

Early experiments established the existence of large molecular machines with the dedicated function of disrupting chromatin structure and facilitating transcription [reviewed by Peterson and Tamkun, 1995]. Most recently the covalent modification of the core histone proteins that assemble DNA into nucleosomes has been recognized as having a major role in transcriptional regulation. Multiple coactivators have been shown to possess histone acetyl-transferase activity [Brownell et al., 1996; Mizzen et al., 1996; Ogryzko et al., 1996; Yang et al., 1996], and corepressors have been identified that recruit histone deacetylases [Alland et al., 1997; Heinzel et al., 1997; Kadosh and Struhl, 1997; Laherty et al., 1997] and that selectively bind to deacetylated histones [Edmondson et al., 1996].

Genetic selections and screens for yeast mutants that influence the transcription of genes limiting for growth have been successful in defining many structural and regulatory components of the transcriptional machinery. These include proteins with well defined functions such as RNA polymerase II subunits and the Tata binding protein TBP [Arndt et al., 1989; Eisenmann et al., 1989]. Other genes were identified whose functional roles in transcription were less clearly defined. Among this latter group were positive regulators of both the HO endonuclease gene required for mating type switching and of the SUC2 [invertase gene] [Neigeborn and Carlson, 1984; Stern et al., 1984].

SWI [switch] and SNF [sucrose non-fermenting] genes have been found to encode proteins that together assemble a large multisubunit complex required for the regulation of a specific group of inducible genes [Cairns et al., 1994; Peterson et al., 1994]. A major clue to the molecular mechanisms by which the SWI/SNF activator complex functions came from a genetic screen for mutations of genes that would allow transcription from the HO gene in the absence of specific SWI genes [Herskowitz et al., 1992]. These studies identified the SIN genes [SWI INdependent]. SIN 1–4 have been found or inferred to have a direct impact on chromatin structure and function. A simple model would predict that the SWI/SNF activator complex functions by overcoming the repressive effects of the SIN gene products on transcription [Fig. 2.49]. Consistent with this hypothesis: in vivo experiments in S. cerevisiae suggest that the SWI/SNF activator complex activates transcription by altering chromatin structure [Hirschhorn et al., 1992]; and in vitro experiments using purified SWI/SNF complex indicate that stoichiometric amounts of SWI/SNF complex can alter histone–DNA interactions in the nucleosome [Côté et al., 1994]. The subsequent analysis of the SIN gene products and other repressive components of the yeast transcriptional machinery and their effects on chromatin have been particularly informative.

SIN1p is a highly charged nuclear protein, somewhat similar in sequence to the vertebrate HMG1/2 proteins and containing the HMG box domain [Kruger and Herskowitz, 1991, Section 2.5.8]. The carboxy-terminal domain of RNA polymerase II has been proposed to antagonize the repressive effects of SIN1 on transcription [Peterson et al., 1991]. However, mutations in SIN1 decrease expression of some genes, suggesting a more complex structural role for this protein in chromatin. In vertebrates, HMG1/2 have been shown to replace linker histones within the nucleosome and to directly influence the transcription process in a chromatin context [Nightingale et al., 1996; Ura et al., 1996].

SIN2p is either histone H3 or H4 [Kruger et al., 1995]. These histones assemble a tetramer [H3/H4]2 that forms the foundation of nucleosomal architecture [Section 2.2.4]. The SIN mutations cluster in β-bridge motifs within the heterodimer of histones H3 and H4 [Fig. 2.50]. Because of the juxtaposition of two [H3, H4] heterodimers at the dyad axis of the nucleosome, the SIN mutations have the potential to disrupt histone–DNA interactions involving the central turn of DNA at the dyad axis. This has a major impact on the integrity of the nucleosome [Kurumizaka and Wolffe, 1997; Wechser et al., 1997].

SIN3p is a large 175 kDa polypeptide containing four paired amphipathic helices, and is proposed to interact with bona fide yeast DNA binding proteins [Wang et al., 1990; Wang and Stillman, 1990]. Targeting of SIN3p by fusion to a DNA binding domain will direct transcriptional repression [Wang and Stillman, 1993]. Genetic experiments indicate a close functional relation between SIN3p and another transcriptional regulatory protein RPD3p [Vidal and Gaber, 1991; Vidal et al., 1991]. RPD3p and SIN3p participate in the same transcriptional regulatory functions and appear to be components of one pathway. Both genes are required for the regulation of inducible genes responding to external signals [PH05], cell differentiation [SPO11 and SP013] and cell type [HO, TY2 and STE6]. It has recently been shown that transcriptional repression by the sequence specific DNA binding protein Ume6p involves recruitment of a complex containing the SIN3p corepressor and the RPD3p histone deacetylase to target promoters [Kadosh and Struhl, 1997]. In RPD3p-deficient strains both gene activation and repression can be impaired by as much as five fold, thereby leading to a 25-fold reduction in the range of transcriptional regulation for particular genes [Vidal and Gaber, 1991].

RPD3p is very similar in sequence to the mammalian histone deacetylase [Taunton et al., 1996]. Subsequent work by Grunstein and colleagues have confirmed this identity, characterizing a 350 kDa histone deacetylase complex [HDA] from yeast nuclei containing three polypeptides, HDA1, HDA2 and HDA3 [Carmen et al., 1996]. The HDA1 polypeptide is very similar in sequence to RPD3p, and antibodies against both proteins immunoprecipitate histone deacetylase activity [Rundlett et al., 1996]. Mutations in either HDA1p or RPD3p, that lead to a general increase in the acetylation of histones H3 and H4 increase transcriptional repression at the telomeres of yeast chromosomes [Rundlett et al., 1996; De Rubertis et al., 1996]. These results could either reflect a direct role for high histone acetylation levels in gene repression at the telomeres, or more likely a consequence of a general increase in the access of transcription factors to non-productive sites in chromosomes following an increase in acetylation levels. A general titration of the transcriptional machinery might account for the difficulty in transcribing telomeric genes. These complications in interpretation are unfortunately inherent to genetic approaches to transcriptional control in chromatin. To understand molecular mechanism, biochemical data are necessary to complement genetics.

SIN4p is a 111 kDa protein that shares no significant homology with other proteins in the database [Jiang and Stillman, 1992]. Mutations in SIN4p alleviate the repression of some genes such as HO, and activate others such as HIS4 and MATα2 [Jiang and Stillman, 1995; Wahi and Johnson, 1995]. These mutations have phenotypes similar to those observed in strains with histone mutations suggesting a modulatory role for SIN4p in organizing chromatin. SIN4p is a component of the mediator complex within the RNA polymerage II holoenzyme [Li et al., 1995; Section 4.1], however this does not exclude functions relevant to chromatin organization. The loss of SIN4p has no effect on nucleosome positioning, but does lead to a striking increase in the sensitivity of chromatin to digestion by micrococcal nuclease [Macatee et al., 1997]. The molecular basis for this transition in chromatin structure is unknown.

TUP1p is a yeast global transcriptional repressor that is also required for the repression of the SUC2 [invertase] gene [Carlson et al., 1984; Williams and Trumbly, 1990]. Tup1p is a 713 amino acid protein: the C-terminal domain contains eight repeats of a 43 amino acid sequence rich in aspartate and tryptophan [WD-40] repeats [Fong et al., 1986]. These WD-40 repeats facilitate the targeting of TUP1p to particular promoters through interaction with the DNA sequence-specific α2 repressor [MATα2p] [Komachi et al., 1994; Section 3.3.4]. The N-terminal domain of TUP1p also interacts with proteins. Here there are two regions of defined function: the first 72 amino acids interact with SSN6p – a large 107 kDa phosphoprotein that contains 10 tandem copies of a tetratricopeptide repeat [Schultz and Carlson, 1987], this region also facilitates multimerization of Tup1p [Tzamarias and Struhl, 1994]. Both TUP1p and SSN6p contribute to the establishment of a repressive chromatin structure targetted by the DNA bound MATα2p/MCM1p complex [Cooper et al., 1994]. The MATα2p/MCM1p complex had been shown by nuclease mapping studies to direct the specific positioning of nucleosomes [Roth et al., 1990; Shimizu et al., 1991]. This positioning is dependent on interactions with the N-terminus of TUP1p and histones H3 and H4 [Edmondson et al., 1996]. Amino-terminal mutations in histones H3 and H4 that interfere. with this interaction with TUP1p relieve the repression of genes by MATα2p/MCM1p. Changing specific lysines [12 and 16] to glutamines, which resemble the consequences of acetylation in the N-terminal tail of histone H4, interfere with Tup1 binding [Edmondson et al., 1996] suggesting that Tup1 prefers to bind to deacetylated histone H4.

These results provide a firm genetic and biochemical basis for considering a specific role for the active modification and organization of chromatin in transcriptional control. This conclusion is further substantiated by the definition of a distinct activator complexes in S. cerevisiae.

The GCN5p/ADA2p/ADA3p activator complex was identified using a distinct type of genetic screen carried out by Guarente and colleagues to identify mutations in genes that confer resistance to the toxic chimeric transcriptional activator GAL4-VP16 [Berger et al., 1992]. Genes identified by this screen might be involved in facilitating gene activation by the VP16 acidic activation domain. In this way two ‘adaptor’ proteins, ADA2p and ADA3p were identified that were proposed to bridge interactions between activation domains and the basal transcriptional machinery [Guarente, 1995]. A comparable mutation in the gene GCN5 impaired the activation of transcription by the transcription factor GCN4p [Georgakopoulos and Thireos, 1992]. Subsequent genetic and biochemical experiments established that GCN5p/ADA2p/ADA3p exist as a coactivator complex in yeast [Georgakopoulos et al., 1995; Marcus et al., 1994; Horiuchi et al., 1995] and that the ADA2p interacts with both acidic activation domains and TBP [Barlev et al., 1995].

The GCN5p/ADA2p/ADA3p coactivator is a histone acetyltransferase [Brownell et al., 1996] that selectively modifies lysine 16 in the N-terminal tail domain of histone H4 [Kuo et al., 1996]. This property suggested for the first time that coactivators have the capacity to directly modify the chromatin template in order to facilitate transcription. GCN5 is not an essential gene in yeast, however the capacity to induce gene expression by GCN4p is reduced by 60% if GCN5 is non-functional. This suggests that like the histone deacetylases [HDA1p and RPD3p], the individual histone acetyltransferases may not be essential in yeast. This might reflect the presence of numerous genes with overlapping functions, and/or merely that the modification of chromatin structure is only one contributor to transcriptional regulation. The existence of multiple potentially redundant histone acetyltransferases is substantiated by recent observations in metazoans.

There is excellent precedent for pioneering experimental work in S. cerevisiae leading to the recognition of comparable regulatory mechanisms in metazoans. The identification of the SWI/SNF activator complex [Peterson and Herskowitz, 1992] offered insight into potential regulatory roles for related proteins in Drosophila [Tamkun et al., 1992]. It was also shown that metazoan regulatory proteins including the glucocorticoid receptor introduced into yeast could make use of the SWI/SNF activator complex to activate synthetic promoters containing their recognition elements [Laurent and Carlson, 1992; Yoshinaga et al., 1992; Laurent et al., 1993]. Mammalian homologs of components of the SWI/SNF complex were characterized [Khavari et al., 1993; Muchardt and Yaniv, 1993; Muchardt et al., 1995; Chiba et al., 1994]. These proteins hSNF2α and hSNF2β, possess amino terminal proline- and glutamine-rich regions which resemble transcriptional activation domains. Their capacity to interact with other components of the transcriptional machinery including the glucocorticoid and oestrogen receptor is shown by their activation of transcription in transient co-transfection assays which may be largely independent of chromatin-mediated effects [Muchardt and Yaniv, 1993; Chiba et al., 1994].

Evidence for the targeted disruption of chromatin by the mammalian SWI/SNF complex has remained elusive. A 100-fold molar excess of the 2 × 106 Da SWI/SNF complex can disrupt a synthetic nucleosome core [containing 0.1 × 106 Da of histone] in vitro [Imbalzano et al., 1994]. It has also been suggested that the RNA polymerase II holoenzyme contains SWI/SNF and might remodel chromatin [Wilson et al., 1996]. If this was true, the targetting problem would be solved. However SWI/SNF is present at only 200 copies per yeast or mammalian cell [Côté et al., 1994; Gerald Crabtree, Stanford University, personal communication], therefore only a small subset of total holoenzyme would contain SWI/SNF and recent purification schemes have not detected SWI/SNF in holoenzyme preparations [Cairns et al., 1996]. Moreover recent experiments suggest that the yeast RNA polymerase II holoenzyme might under certain circumstances disrupt chromatin independent of the presence of SWI/SNF [Gaudreau et al., 1997]. Crabtree and colleagues have identified multiple complexes containing mammalian SWI/SNF homologs [Wang, W. et al., 1996a,b] and suggested the existence of development tally distinct functions. Certain cell lines lack hSNF2α and β [also known on hbrahma and brahma related gene 1, respectively] entirely, indicating that they are not essential for cell viability [Wang, W. et al., 1996b]. The association of hSNF2α and β with chromosomes is modulated during the cell cycle and following cell transformation [Muchardt et al., 1996], leading to the suggestion that these proteins might be involved in the control of cell growth. The exact functions of the metazoan SWI/SNF complex remain to be determined, however a connection to transcriptional regulation for a subset of this diverse family of complexes seems to be a reasonable speculation [Wang, W. et al., 1996a, b].

Historie acetyltransferases: PCAF, p300, TAFII250. The discovery that S. cerevisiae GCN5p has histone acetyltransferase activity [Brownell et al., 1996] led to the recognition that comparable regulatory mechanisms might exist in metazoans [Yang et al., 1996b]. A human homolog of GCN5p known as p300/CBP associated factor [PCAF] acetylates histones [Yang et al., 1996], as does p300/CBP itself [Ogryzko et al., 1996]. p300/CBP serves as an integrator to mediate regulation by a wide variety of sequence-specific transcription factors [Kamei et al., 1996] including steroid and nuclear hormone receptors, c-Jun/vJun, cMyb/v-Myb, c-Fos and MyoD [Janknecht and Hunter, 1996]. To strengthen the analogy with the GCN5p/ADA2p/ADA3p complex, p300/CBP has a domain highly similar to part of ADA2p and associates with PCAF, the homolog of GCN5p [Yang et al., 1996]. Most recently a component of the DNA-binding basal transcription factor TFIID has also been shown to have histone acetyltransferase activity [Mizzen et al., 1996]. TAFII250 is the architectural core of TFIID interacting with the other TAFs [TBP associated factors] as well as with TBP [Section 4.1]. TAFII250 is required for the activation of particular genes indicative of coactivator function, and associates with components of the basal transcriptional machinery such as TFIIA, TFIIE and TFIIF [Dikstein et al., 1996]. In addition, TAFII250 functions as both a kinase and a histone acetyltransferase [Mizzen et al., 1996; Dikstein et al., 1996]. Although GCN5p and PCAF are related proteins, there is no significant sequence identity or known structural similarity with p300/CBP or TAFII250. Thus diverse proteins in metazoans [and potentially in S. cerevisiae] possess histone acetyltransferase activity. In an interesting link between the mammalian SWI/SNF activator complex, monoclonal antibodies against p300 immunoprecipitate a complex of at least seven cellular proteins [Dallas et al., 1997]. Within this complex is TBP, TAFII250 and hSNF2β [BRG1] suggesting that functions of histone acetyltransferases might be linked to those of other activators that contend with chromatin.

Histone deacetylase and mammalian SIN3. The purification of the mammalian histone deacetylase and the recognition of the similarities to S. cerevisiae RPD3p [Taunton et al., 1996] has provided considerable insight into transcriptional repression in metazoans. The first direct evidence for mammalian homologs of RPD3p being involved in transcriptional repression came from two hybrid screens indicating that thetranscriptional regulatory factor YY1 interacted with mouse and human RPD3p [Yang et al., 1996a]. The fusion of mammalian RPD3p to a targeted DNA binding domain directed transcriptional repression by more than 10-fold. Mutations in a glycine rich domain of YY1 that directs binding to RPD3p could abolish transcriptional repression by YY1 suggesting that YY1 negatively regulates transcription by tethering RPD3. YY1 is a mammalian zinc-finger transcription factor [Shi et al., 1991] that is proposed to regulate cell growth and differentiation [Shrivastava and Calarne, 1994].

A second well-defined protein complex that influences cell growth and differentiation in mammalian cells is the Mad-Max heterodimer [Lahoz et al., 1994; Chen et al., 1995; Hurlin et al., 1995]. Max is a widely expressed sequence-specific transcriptional regulator of the basic region-helix-loop-helix-leucine zipper family [bHLH-ZIP]. Max heterodimerizes with the Myc family of bHLH-ZIP proteins including Myc, Mad and Mxi-1 [Ayer et al., 1993; Zervos et al., 1993]. While the Myc-Max complex activates transcription and transformation, the Mad-Max complex represses these events. Eisenman and colleagues identified two mammalian proteins mSin3A and mSin3B that interact with Mad and that have striking homology to S. cerevisiae Sin3p including the four paired amphipathic helix [PAH] domains. The association between Mad-Max and mSin3A and B requires the second PAH domain. Mutations in this domain eliminate the interaction with mSin3A and prevent the Mad-Max complex from repressing transcription [Ayer et al., 1995]. The next step was to establish that the mSIN3 proteins interact with the mammalian histone deacetylases. Mad, mSIN3 and the mammalian histone deacetylases coimmunoprecipitate [Laherty et al., 1997; Alland et al., 1997]. The third PAH domain of mSIN3 interacts with the mammalian RPD3p homologs and can confer transcriptional repression when attached to a DNA binding domain. More subtle mutational analysis suggests that the cell transformation and transcriptional repression suppressed by the Mad-Max complex depend on distinct domains of the mSIN3 proteins [Alland et al., 1997]. However, an active role for histone deacetylation in transcriptional control is demonstrated by the use of deacetylase inhibitors such as Trichostatin A [Yoshida et al., 1990], that abolish Mad’s ability to repress transcription. The existence of a conserved transcriptional repression mechanism that utilizes SIN3p and histone deacetylase emphasizes the significance of the chromatin environment for transcriptional control. Histone deacetylation directs the assembly of a stable repressive chromatin structure [Fig. 2.51].

A role for chromatin had already been established in the control of transcription by the thyroid hormone receptor [Wong et al., 1995, 1997a]. These studies provide a useful example of how the histones can contribute to gene regulation. The assembly of minichromosomes has been utilized within the Xenopus oocyte nucleus to examine the role of chromatin in both transcriptional silencing and activation of the Xenopus TRβA promoter. Transcription from this promoter is under the control of thyroid hormone and the thyroid hormone receptor [Ranjan et al., 1994], which exists as a heterodimer of TR and RXR. Microinjection of either single-stranded or double-stranded DNA templates into the Xenopus oocyte nucleus offers the opportunity for examination of the influence on gene regulation of chromatin assembly pathways that are either coupled or uncoupled to DNA synthesis [Almouzni and Wolffe, 1993a]. The staged injection of mRNA encoding transcriptional regulatory proteins and of template DNA offers the potential for examining the mechanisms of transcription factor-mediated transcriptional activation of promoters within a chromatin environment. In particular, it is possible to discriminate between pre-emptive mechanisms in which transcription factors bind during chromatin assembly to activate transcription, and post-replicative mechanisms in which transcription factors gain access to their recognition elements after they have been assembled into mature chromatin structures. TR/RXR heterodimers bind constitutively within the minichromosome, independently of whether the receptor is synthesized before or after chromatin assembly. Rotational positioning of the TRE on the surface of the histone octamer allows the specific association of the TR/RXR heterodimer in vitro [Wong et al., 1997b]. The coupling of chromatin assembly to the replication process augments transcriptional repression by unliganded TR/RXR without influencing the final level of transcriptional activity in the presence of thyroid hormone.

The molecular mechanisms by which the unliganded thyroid hormone receptor makes use of chromatin in order to augment transcriptional repression also involve mSin3 and histone deacetylase [Alland et al., 1997; Heinzel et al., 1997]. The unliganded thyroid hormone receptor and retinoic acid receptor bind a corepressor NCoR [Horlein et al., 1995]. NCoR interacts with Sin3 and recruits the histone deacetylase [Alland et al., 1997; Heinzel et al., 1997, see Fig. 2.51]. AU the transcriptional repression conferred by the unliganded thyroid hormone receptor in Xenopus oocytes [Wong et al., 1995, 1997a] can be alleviated by the inhibition of histone deacetylase using Trichostatin A [Fig. 2.52] indicative of an essential role for deacetylation in establishing transcriptional repression in a chromatin environment.

The addition of thyroid hormone to the chromatin bound receptor leads to the disruption of chromatin structure [Wong et al., 1995, 1997a]. Chromatin disruption is not restricted to the receptor binding site and involves the reorganization of chromatin structure in which targeted histone acetylation by the PCAF and p300/CBP activators may have a contributory role [Yang et al., 1996b; Ogryzko et al., 1996, see Fig. 2.51]. Recently, yet another targeted coactivator with histone acetyltransferase activity has been discovered [Chen et al., 1997]. The exact coactivators that function in Xenopus are being defined. It is possible to separate chromatin disruption from productive recruitment of the basal transcription machinery in vivo by deletion of regulatory elements essential for transcription initiation at the start site and by the use of transcriptional inhibitors [Wong et al., 1995, 1997a]. Therefore chromatin disruption is an independent hormone-regulated function targeted by DNA-bound thyroid hormone receptor. It is remarkable just how effectively the various functions of the thyroid hormone receptor are mediated through the recruitment of enzyme complexes that modify chromatin. These results provide compelling evidence for the productive utilization of structural transitions in chromatin as a regulatory principle in gene control [Wolffe, 1997].

The genetic, biochemical and cell biological evidence that we have outlined provides a substantial rationale for considering chromatin structural proteins as integral components of the transcriptional machinery. It is important to recognize that chromatin structure is not necessarily static and obstructive to transcription but provides a means of display for the DNA template that determines function. Variation in the quality of histone–DNA interactions and in the three-dimensional path of the double helix can directly influence transcription [Ura et al., 1997; Schild et al., 1993 see Sections 4.1 and 4.2]. Conformation is a well-known determinant of enzymatic activity, alterations in chromatin conformation may well determine transcriptional activity.

The facts that: [1] core histone acetylation greatly facilitates the access of transcription factors to DNA in a nucleosome; [2] transcriptional repressors recruit histone deacetylases; and [3] transcriptional coactivators are histone acetyltransferases leads to a model for transcriptional regulation in which the recruitment of repressors could direct the local stabilization of repressive histone–DNA interactions and where the recruitment of activators could destabilize these interactions [Fig. 2.53]. Repressive nucleosomes might prevent either the association or function of the basal transcriptional machinery on a particular promoter. However, it is important to note that certain transcriptional regulators such as the thyroid hormone receptor can bind to their recognition elements in a nucleosome even when the histones are deacetylated [Wong et al., 1995, 1997a] and nucleosome assembly is not always repressive [Schild et al., 1993].

The recruitment of histone deacetylase by chromatin bound repressors will potentially eliminate basal levels of histone acetylation and impede the recruitment or function of the basal transcriptional machinery. Targeted acetylation provides a means of allowing the basal machinery to displace nucleosomes, assemble a functional transcription complex and never have to deal with chromatin again. For example we can propose three steps in the regulation of transcription by thyroid hormone receptor: [1] thyroid hormone receptor binds to chromatin on the surface of a positioned nucleosome and facilitates the assembly of a repressive chromatin structure; [2] in response to hormone, the receptor recruits molecular machines or enzymes that disrupt local chromatin structure; [3] the hormone-bound receptor and associated activators facilitate the recruitment and activity of the basal transcriptional machinery to further activate transcription. Additional interesting possibilities include the regulated association and activity of histone acetyltransferases and deacetylases within a common complex. In this way transcriptional activity could be continually modulated through variation in chromatin conformation [Fig. 2.53].

2.5.5 Remodelling of chromatin during spermatogenesis

Histones represent only one way of packaging DNA such that it can fit into the volume of a nucleus. There are many possible ways of rendering DNA compact in a reversible fashion. It is a measure of the major role of histone structure in many other nuclear processes [Sections 2.5.4 and 4.1] that histones have been so conserved through evolution. Perhaps the best example of the reversible compaction of DNA by multiple pathways concerns the condensation of DNA into sperm nuclei during spermatogenesis. This provides an excellent example of roles for histone variants, post-translational modification of histones and non-histone proteins in regulating chromosome structure and function.

The easy availability of pure populations of spermatozoa led to the early realization that the types of proteins in the sperm nucleus could vary greatly in different organisms. These proteins have been divided into certain classes. One class, the protamines, comes in two types: one type is rich in polyarginine tracts [4–6 residues], punctuated with proline, and potentially phosphorylatable serine and threonine residues; the other type is rich in cysteine. Both types of protamine are small [3000–5000 Da] and highly basic. Protamine–DNA complexes often represent the final state of chromatin in fish and mammalian sperm nuclei. However, during the process of spermatogenesis other proteins can transiently replace the histones [Poccia, 1986].

The transition proteins replace histones during the initial stages of condensation of chromatin in spermiogenesis and are later replaced by protamines, which are the only basic nuclear structural proteins in the sperm of most mammals. The transition proteins presumably facilitate the replacement of histones by protamines, although little is known about their specific function. The amino acid sequence of a mouse transition protein [TP 2] suggests two domains of protein structure: an ammo-terminus which, like the protamines, is rich in cysteine, proline and phosphorylatable serine and threonine; and a highly basic carboxyl-terminus rich in arginine [Kleene and Flynn, 1987; Luerssen et al., 1989]. The transition proteins are of special interest because they are the only proteins that are proven to dissociate histones from DNA in a physiological context by competitive mechanisms without the use of molecular machines such as coactivators and SWI/SNF or the progression of RNA or DNA polymerase. Several mechanisms have been proposed. First, the electrostatic binding of the transition proteins to DNA should be intermediate between histones and protamines because the concentration of basic amino acids near the carboxyl-terminus is greater than that of histones and less than in protamines. Second, the putative DNA-binding domain contains aromatic amino acids: phenylalanine and tyrosine. The tyrosine residues in the transition proteins have been shown to intercalate between the bases of DNA lowering its thermal stability [Singh and Rao, 1987]. Intercalation of aromatic amino acids can induce bends or kinks in native DNA, and these bends might alter the path of DNA around the histone core and potentially destabilize the nucleosome.

Special histone variants exist that are specific for the testis. In sea urchin sperm, specific variants exist of histones H2B and histone H1. Both molecules are considerably larger and richer in arginine than their relatives in somatic cells. Most of the extra size is due to amino-terminal extensions. These additional amino acids increase the interaction of both histones with linker DNA [Bavykin et al., 1990; Hill and Thomas, 1990]. It is possible that they might also further facilitate the stabilization and condensation of chromatin by creating inter strand linkages between chromatin fibres. Phosphorylation of the histone H1 tails regulates their interaction with DNA [Hill et al., 1991]. As chromatin is compacted, the tails are dephosphorylated [Section 2.5.3]. Testis-specific histone variants also exist in mammals. Variants of histone H1, histones H2A and H2B accumulate during meiotic prophase. As all of these transitions in chromatin structure occur after replication, the movement of a processive enzyme complex along the DNA duplex is not a prerequisite for the remodelling of chromosomal structure [Sections 3.1.1 and 3.1.2]. In the rat, the core histone variants are heavily acetylated prior to their dissociation from DNA, which is driven by the accumulation of the transition proteins. Finally, accumulation of the protamines, whose binding is also regulated by their phosphorylation state, leads to the progressive displacement of the transition proteins and the nucleus is completely condensed. The problem of decondensing the sperm nucleus following its arrival in the cytoplasm of the fertilized egg is considered later [Section 3.1].

Why is sperm chromatin condensed through such a special mechanism? Suggestions include the protection and stability of the genetic material DNA in such structures. DNA in the nucleus of a spermatozoon is much less accessible to nucleases than in a somatic cell. It is also much more resistant to physical and chemical perturbants. A concomitant effect of chromatin compaction, as normal histones are replaced in chromosomes, is the suppression of gene activity. It is possible that the compaction of DNA into the sperm nucleus not only renders the DNA inaccessible to RNA polymerases but also erases the developmental history of a chromosome. Specifically the trans-acting factors responsible for directing specific events in the nucleus could be displaced. Evidence to support this concept comes from an analysis of DNase I hypersensitive sites in chromatin [Section 4.2] which are lost in sperm for all genes. However, those genes that are constitutively expressed are marked during spermatogenesis by hypomethylation at sites of future hypersensitivity [Groudine and Conkin, 1985].

2.5.6 Heterochromatin, position effect, locus control regions and insulators

Early studies by cytologists led to the realization that some chromosomal regions have properties distinct from the rest of the chromosome [Henikoff, 1990, 1994; Pardue and Hennig, 1990]. Large segments of chromatin were found to be highly condensed and to replicate late in S-phase. Geneticists determined that these chromosomal regions, which they called heterochromatin, did not participate in meiotic recombination. However, heterochromatin does have significant genetic effects. The most common observed consequence of heterochromatin formation is the repression of transcription either in heterochromatin itself or in regions of chromatin that lie adjacent to the heterochromatin domain. The variability in gene expression at the border of the heterochromatin is described as ‘position effect variegation’. Three explanations have been offered for this phenomenon. The first is that special proteins, such as HP1 [see below], exist that cause heterochromatin to adopt its distinct structure, and these proteins can ‘spill over’ into regions of normal chromatin. The second is that heterochromatin represents the sequestration of chromosomal domains in specialized nuclear compartments from which the transcriptional machinery is excluded. The third applies only to Drosophila and other insects with polytene chromosomes in that, following placement adjacent to heterochromatin, a gene will undergo fewer rounds of replication than would normally occur [Spradling and Karpen, 1990]. Fewer copies of the gene would cause a concomitant reduction in transcription. Most investigators accept that heterochromatin-specific proteins can diffuse on to normal chromatin and thereby influence gene expression of juxtaposed genes. Although initially defined from work on insects, there is an increasing body of evidence that heterochromatin domains exist in all eukaryotic chromosomes including those of yeast. Moreover, it is clear that metazoans have made use of this type of chromosomal organization for the developmental regulation of gene expression [Singh, 1994].

An approach to the molecular basis of heterochromatin formation has been to look for mutations in Drosophila or more recently in S. cerevisiae, that enhance or suppress position effects on gene expression [modifiers of position effect variegation]. Modifiers of position effect variegation would include mutations in the genes encoding the chromosomal proteins involved in forming heterochromatin [Renter and Spierer, 1992]. Support for this approach comes from the observation that chromosomal deletions that reduce the number of copies of histone genes in Drosophila reduce the influence of position effects on gene expression [Moore et al., 1983]. A role for histones is consistent with the observation that chemicals that maintain histones in a hyperacetylated state suppress position effect [Mottus et al., 1980]. Using this approach a gene encoding a non-histone protein [HP1, heterochromatin protein 1] has been identified [James and Elgin, 1986]. Mutation of the HP1 gene reduces position effects on gene expression [Elgin, 1990]. HP1 is preferentially associated with the heterochromatin regions of polytene chromosomes. Other proteins homologous to HP1 include Polycomb, which is also chromatin associated and is known from genetic experiments to influence the expression of many genes in normal chromatin [Zink and Paro, 1989]. Neither HP1 nor Polycomb interact with DNA directly but presumably recognize some aspect of nucleosome or chromatin fibre structure. Any such direct interaction is however yet to be defined. Polycomb and HP1 share a common amino acid sequence known as the chromodomain [chromatin modification]. This domain is highly conserved through evolution and can be found in the S. pombe SWI6 gene. The SWI6 protein is involved in the assembly of the chromosomal domain containing the transcriptionally silent mating type cassettes [Klar and Bonaduce, 1991; Singh, 1994; Lorentz et al., 1992, 1994]. The chromodomain does seem to have a role in the subnuclear compartmentalization of chromodomain proteins such as HP1 and Polycomb. Mammalian homologs of HP1 are found in subnuclear structures or foci termed ‘polymorphic interphase karyosomal associations’ [Saunders et al., 1993]. Polycomb has a similar distribution in Drosophila nuclei. Recent studies have established that the chromodomain family of proteins comprising more than 40 members [Aasland and Stewart, 1995; Koonin et al., 1995] can be subdivided into two major groups. Proteins such as HP1 contain both an N-terminal chromodomain and a C-terminal ‘shadow’ chromodomain [Epstein et al., 1992; Aasland and Stewart, 1995]. The N-terminal chromodomain appears to direct binding to heterochromatin whereas the C-terminal ‘shadow’ chromodomain determines nuclear localization and assists in binding to chromatin [Messmer et al., 1992; Platero et al., 1995; Powers and Eissenberg, 1993]. The second group of proteins including Polycomb and SWI6 rely on interactions with other proteins to target association with particular chromatin domains [Lorentz et al., 1994; Ekwall et al., 1995].

The structure of the chromodomain was recently solved using nuclear magnetic resonance [NMR] [Ball et al., 1997]. The chromo-domain has strong homology to two archaebacterial DNA binding proteins [Baumann et al., 1994; Edmondson et al., 1995]. However, the eukaryotic chromodomain does not interact with DNA [Ball et al., 1997] and appears to be involved in protein–protein interactions [Le Douarin et al., 1996]. Each chromodomain consists of an N-terminal three-stranded anti-parallel β sheet which folds against a C-terminal α-helix. The presence of both a chromodomain and a shadow chromodomain are thought to allow proteins such as HP1 to function as adaptors in the assembly of large multicomponent proteins.

The Polycomb-group [PcG] of proteins are found to be non-randomly associated with over 100 chromosomal domains within polytene chromosomes [Zink and Paro, 1989; De Camillis et al., 1992; Rastelli et al., 1993]. Of particular interest are the homeotic genes that are regulated by PcG proteins. How PcG proteins are directed to particular regions such as the homeotic loci of the chromosome remains substantially unknown. Mutations in the genes encoding the PcG proteins generally lead to the activation of target genes. Changes in the expression of homeotic genes that control the segmental identity of the insect body following mutation of the PcG proteins have been studied in detail [Moehrle and Paro, 1994; Pirotta and Rastella, 1994].

The general properties of ‘modifier’ genes that enhance or suppress position effect variegation provide some insight into how PcG proteins might work to change chromosomal structure. The phenotypes of modifier mutations show a strong dependence on the number of gene copies within the cell [Locke et al., 1988; Reuter et al., 1990]. This has led to the proposal that the modifier proteins that assemble heterochromatin act through simple mass action. The more modifier protein there is in the nucleus the more repressive heterochromatin will be assembled. It is probable that certain modifier proteins co-operate to assemble multimeric complexes that alter the structure of entire chromosomal domains [Fig. 2.54] [Moehrle and Paro, 1994]. The distribution of Polycomb along the chromosomal domain including the bithorax complex which encodes three homoeotic genes supports this hypothesis. Orlando and Paro [1993] found that the Polycomb protein is associated with transcriptionally inactive chromatin over more than 200 kb of DNA, while it is absent from regions of gene activity. The boundaries of Polycomb-associated chromatin and Polycomb-free chromatin are very distinct, which implies that regulatory elements exist that initiate and terminate Polycomb binding [Fig. 2.54].

Recent genetic approaches have delineated regulatory elements within genes such as engrailed that confer silencing in response to Polycomb [Polycomb response elements, PREs]. Each PRE is several hundred base pairs in length and can independently direct position effect variegation in the presence of Polycomb [Chan et al., 1994; Rastelli et al., 1993]. The sequences of individual PREs show little homology to each other suggesting that features of DNA structure or diverse DNA binding proteins could target Polycomb mediated silencing. For the PRE in the Ubx gene a binding site for the GAGA factor has been defined suggesting that this regulatory factor has a role in facilitating Polycomb function [Pirotta, 1997]. The GAGA factor has been implicated in the remodeling of chromatin [Tsukiyama et al., 1994; Section 4.2].

An important aspect of the functional role of the PcG proteins is that they are not involved in establishing the expression state of a particular gene but in the maintenance of the repressed state through replication and chromosomal duplication. Since they act through large multicomponent complexes it is possible that they might subdivide through a replicative event, thereby maintaining a repressive chromatin structure [Fig. 2.55]. Specialized mechanisms appear to exist within the chromosome to enable the removal of Polycomb from chromatin and the resetting of chromatin to a transcriptionally competent state [Tamkun et al., 1992]. The molecular machines involved in this process include the Brahma protein which is structurally related to a component of the yeast general activator complex SNF2/SWI2 [Section 2.5.4]. Evidence for competition between transcriptional activators and repressors such as Polycomb comes from experiments in which PRE/Polycomb-mediated transcriptional silencing can be overcome by the expression of high levels of Gal4p [Zink and Paro, 1995]. Immunostaining of polytene chromosomes revealed that the Polycomb complex was removed from the vicinity of the Gal4p binding site without the need for DNA replication and cell division.

Gottschling and colleagues have successfully established that position effect variegation also occurs in the chromosomes of S. cerevisiae. When a gene is located near a telomere, transcriptional activity is reduced [Gottschling et al., 1990]. Transcriptional repression at yeast telomeres appears to be due to the assembly of a distinct chromatin structure that initiates at the telomere. The sequence of the DNA is important since internal tracts of telomeric DNA have the capacity to also act as silencers of transcription in S. cerevisiae [Stavenhagen and Zakian, 1994]. Mutations in the amino-terminal tails of histones H3 and H4 relieve transcriptional silencing [Aparicio et al., 1991; Thompson et al., 1994a]. Histone H4 at the telomeres is hypoacetylated and mutations in the amino terminal acetyltransferases also relieve silencing [Aparicio et al., 1991; Braunstein et al., 1993]. DNA methyltransferases expressed in yeast have restricted access to telomeric chromatin compared to most of the chromosome [Gottschling, 1992]. This implies that telomeric chromatin is either more compacted or is sequestered away from freely diffusible trans-acting factors in the nucleoplasm. There are also several similarities to properties of yeast and Drosophila heterochromatin. Genes within Drosophila heterochromatin are normally maintained in a stably repressed state, but will occasionally escape repression [Henikoff, 1990], and similar phenomena occur in yeast [Gottschling et al., 1990].

The efficiency of transcriptional repression decreases when a gene is placed further from the S. cerevisiae telomere [over a 10–20 kb range; Renauld et al., 1993]. This result supports the idea that a repressive chromatin structure originates at the telomere and spreads along the chromosome. The expression of a protein SIR3 influences the extent of silencing at the telomeres, suggesting that it is essential for the assembly of repressive chromatin [Renauld et al., 1993]. SIR3 and the histone tails are also involved in the sequestration of yeast chromosomal telomeres at the periphery of the nucleus [Palladino et al., 1993; Thompson et al., 1994a]. Thus telomeric silencing could be related to the sequestration of this portion of the chromosome within a transcriptionally incompetent compartment of the nucleus adjacent to the nuclear envelope [Franke, 1974].

Many of the modifiers of position effect in yeast chromosomes are shared between the telomeres of the chromosomes and the silent mating type cassettes in yeast. Mutations in four silent information regulator genes, SIRI, 2, 3 and 4, cause derepression of the silent mating type cassettes [Laurensen and Rine, 1992]. Mutations in three of these, SIR2, 3 and 4, influence telomeric repression [Aparicio et al., 1991]. The assembly of specialized chromosomal domains at the silent mating type cassettes is reflected in the limited access of restriction enzymes compared to bulk chromatin [Loo and Rine, 1994a]. The silencers at the mating type cassettes are required after each round of replication to re-establish the transcriptionally repressed state [Holmes and Broach, 1996]. This indicates that repressive chromatin structures do not self-template the reassembly of a repressive structure. It is possible that both the telomeres and the mating type cassettes are sequestered in specialized nuclear structures or compartments in which the transcriptional machinery does not function efficiently.

Recent results also implicate heterochromatin-mediated silencing mechanisms in the nucleolus, the site of ribosomal RNA transcription. SIR2, SIR4, H2A and H2B have been found to influence the transcription of genes requiring RNA polymerase II that are integrated into ribosomal DNA [Bryk et al., 1997; Smith and Boeke, 1997]. Remarkably, silencing also appears to control aging in yeast. Interference with the targeting of silencing complexes by expression of mutant forms of SIR4 allows yeast to live longer [Kennedy et al., 1997]. Nucleolar organization is implicated in this phenomenon because mutant SIR4 proteins localize to the nucleolus. In the fission yeast Schizosaccharomyces pombe, the determinants of gene silencing and heterochromatin assembly at the silent mating type cassettes are shared with the centromeres [Thon and Klar, 1992; Klar, 1992; Allshire et al., 1994, 1995].

The existence of these diverse sites of heterochromatin-mediated silencing has led to the idea that there is a competition for limiting components and that the telomere might act as a molecular sink to form a reservoir of silencing factors [Maillet et al., 1996; Gotta et al., 1996]. Alternatively individual targeting factors such as SIRI at the mating type loci might modulate the stability of heterochromatin assembly.

The heterochromatin structures that control transcription are likely to be dynamic, and ubiquitination may have a role in regulating silencing. A de-ubiquitinating enzyme Ubp3 binds to SIR4p and mutations in the Ubp3 gene increase silencing at telomeres and mating type loci [Moazed and Johnson, 1996]. Mutations in the yeast SAS acetyltransferases also influence silencing at the telomeres and mating type loci [Reifsnyder et al., 1996]. Histone modification is likely to have a significant role in the silencing process.

In an apparently related phenomenon, components of the yeast origin recognition complex [ORC] that normally regulate the site and frequency of chromosomal initiation of replication, direct transcriptional silencing within the same chromosomal domain [Bell et al., 1993a, b]. The molecular mechanism exerting this effect is unknown. It is possible that a gene adjacent to the ORC is directed to reside in a replication competent but transcriptionally incompetent environment. Alternatively, perhaps replication and ORC function are necessary to remodel chromatin from an active to a repressed state [Almouzni and Wolffe, 1993a; Almouzni, 1994], as replication is required to repress the mating type cassettes [Miller and Nasmyth, 1984].

Position effects in mammalian chromosomes have been a recurrent problem for transgenic research since highly variable levels of transcriptional activity follow from the random introduction of reporter genes into the genome [Forrester et al., 1987; Grosveld et al., 1987; Stief et al., 1989]. These effects can be relieved by the introduction of a Zocus control region [LCR] that exerts a dominant transcriptional activation function over a chromatin domain [10–100 kb]. The mechanism of this activation function remains to be determined; however, communication between LCRs, enhancers and promoters either directly or through modifications of chromatin structural components are favoured hypotheses [Felsenfeld, 1992; Wijgerde et al., 1995; Section 4.2]. Recent evidence suggests that both the locus control regions and enhancers act in eis to actively suppress position-effect variegation [Walters et al., 1996; Festenstein et al., 1996]. In this regard locus control regions basically function as operationally defined ‘powerful’ enhancers. The coexistence of heterochromatin domains that can transmit repressive effects and the definition of the extensive long-range activation function of LCRs emphasize the necessary compartmentalization of the chromosome into discrete functional units. These discrete functional units are prevented from influencing each other in a natural chromosomal context [Fiering et al., 1995]. This is due in part to the existence of special chromosomal regions that prevent the transmission of chromatin structural features associated with the boundaries of repressive or active domains. These specialized chromosomal regions are known as insulators [Chung et al., 1993; Wolffe, 1994a].

The original evidence for an insulator function within the chromosome came from genetic experiments in Drosophila. Each boundary of the 87A7 heat-shock locus is defined by a pair of nuclease hypersensitive sites bordering a 250–300 bp segment of DNA. These specialized chromatin structures [scs] are located at the junctions between the decondensed chromatin of the transcriptionally active 87A7 heat-shock locus and adjacent condensed chromatin. The scs were found to have three functional properties: [1] they establish a domain of independent gene activity at many distinct chromosomal positions; [2] scs elements are necessary at each edge of the domain; and [3] the elements are independently neither inhibitory nor stimulatory to transcriptional activity within the domain [Kellum and Schedi, 1991, 1992]. Subsequent work found that the introduction of an scs element between an enhancer and a promoter blocked communication between the two elements. Thus the scs elements prevent both the transmission of repressive effects on transcription from proximity to heterochromatin and the transmission of stimulatory effects on transcription from an enhancer. How this insulation is achieved is unknown; moreover, the nature of the nucleoprotein complex assembled on the scs elements has yet to be defined. Fortunately, similar phenomena have been described associated with a well-defined nucleoprotein complex between the Drosophila suppressor of hairy-wing protein [su[Hw]] and the gypsy retrotransposon [Corces and Geyer, 1991].

Insertion of a gypsy element as far as 10–30 kb from a promoter can cause a mutant phenotype [Jack et al., 1991]. The mutant phenotype requires the su[Hw] protein to interact with the inserted gypsy element at a 350 bp region containing 12 copies of a 10-bp sequence separated by AT-rich sequences. The su[Hw] protein has a molecular weight of 100 kDa, and its sequence includes several motifs characteristic of eukaryotic transcription factors, including 12 zinc fingers, a leucine zipper and two acidic domains [Parkhurst et al., 1988; Fig. 2.56]. The complex of the su[Hw] protein with a gypsy element has many of the properties of an insulator element: the complex blocks enhancer activity when placed between an enhancer and a promoter [Holdridge and Dorsett, 1991; Geyer and Corces, 1992], and when the complex is placed at the boundaries of a gene-containing fragment, the gene is protected from the repressive effects of heterochromatin on transcription [Roseman et al., 1993].

Although in certain circumstances the su[Hw] protein does not independently stimulate or repress transcription of a reporter gene, the su[Hw] protein can occasionally function as a transcriptional activator [Corces and Geyer, 1991]. This suggests that the function of an insulator may be conferred on sequences by DNA-binding proteins that might under other circumstances have more conventional roles in the transcription process. Although it is clear that the su[Hw] protein does not bind to scs elements [Roseman et al., 1993] it seems likely that these elements will form large nucleoprotein complexes with a similar composition.

How might insulators function? Any models must explain how characteristics of both repressive or active chromatin are restricted to particular chromosomal domains. Several models have been suggested to explain the activities of LCRs and enhancers [Fig. 2.57] [Section 4.1]. These elements might function as entry points for transcription factors, RNA polymerase or other components of the transcription machinery, which might then track along the DNA until reaching the promoter. Alternatively the LCR or enhancer complex might associate with the promoter complex by stable looping of the intervening DNA or chromatin, forming a complex that increases the efficiency with which RNA polymerase is recruited and used. Another possibility is that the LCR or enhancer complex might cause the gene to assemble into a chromatin structure capable of being transcribed through its association with nuclear compartments [or organelles] that act as transcription factories [Section 2.5.9], or that associate with proteins that modify repressive chromatin structure by disrupting nucleosomes [Section 2.5.4]. Similar models can explain the repressive effects of heterochromatin. Repressive chromatin proteins, such as the Drosophila HP1 protein or histone deacetylases, may undergo local diffusion enlarging the heterochromatin domain. Alternatively, heterochromatin may be sequestered in a transcriptionally incompetent region of the nucleus [see above].

In considering these models, it is important to recognize that the eukaryotic nucleus is a highly organized structure in which DNA is compacted by its association with histone proteins into nucleosomes and the chromatin fibre [Section 2.5.9]. Although it is possible that insulators prevent protein tracking or diffusion between active and repressive domains, it is difficult to envisage how this might occur in the nucleus – where DNA segments separated linearly by many kilobases can be juxtaposed by folding of the DNA helical axis in three dimensions – without invoking some specific attachment of inactive chromatin domains, insulators and active chromatin domains to a nuclear framework. Similar attachments might be required to prevent the juxtaposition, as a result of DNA looping, of LCRs, enhancers and promoters. Perhaps the most economical suggestion is that insulators are nucleoprotein complexes that associate neither with regions or structures in the nucleus where ‘transcription factories’ load on to DNA [Jackson et al., 1993], nor with regions or structures from which the transcriptional machinery is excluded [Palladino et al., 1993]. Instead, the insulators might associate with distinct ‘neutral’ nuclear structures. The ‘neutral’ nuclear structures would tether promoter elements where the transmissible activating effects of enhancers or silencing effects of heterochromatin could not occur. This absence of transmissible effects could be accounted for by the exclusion from the ‘neutral’ nuclear structures of particular transcriptional coactivators normally associated with communication between promoters and enhancers, or of chromatin modification proteins or enzymes normally associated with heterochromatin.

Examples of mammalian chromosomal regions that contain heterochromatin are the centromere and the telomere [Section 2.4.2]. The telomeres of mammalian chromosomes have an unusual chromatin structure in which nucleosomes are closely packed with a repeat length of 157 bp [Makarov et al., 1993]. Mammalian telomeres consist of the sequence [TTAGGG]n repeated for 10–100 kbp [Section 2.4.2]. Heterochromatin at the centromere contains tandemly repeated simple sequence ‘satellite’ DNA, for example the α-satellite DNA at the human centromere. This α-satellite heterochromatin appears to play a structural role by mediating attachment of the kinetochore [Section 2.4.2]. The inactive X chromosome of female mammals is also heterochromatic. The properties of the inactive X chromosome have been the focus of much interesting research.

Female mammalian embryos begin development with two active X chromosomes; however, very early in embryogenesis almost all of the genes on one of the two X chromosomes become inactivated. This transcriptional inactivation is concomitant with the chromosome both taking on the appearance of heterochromatin and also becoming late replicating during S-phase. Although in eutherian [placental] mammals the initial choice between inactivation of the maternal or paternal X chromosome is random, once established in a repressed state the same X chromosome will be inactivated after every cell division. This is an excellent example of the establishment and maintenance of a chromosomal state of determination [Section 4.3]. Abnormalities in X-chromosome inactivation have allowed the definition of the X inactivation center that is required to present for inactivation to occur [Willard et al., 1993]. Within the X inactivation center is the Xist gene which is the key regulator of inactivation [Brown et al., 1991; Brockdorff et al., 1991]. Remarkably the gene does not encode mRNA, but instead a long, imtranslated RNA that remains associated with the inactive X chromosome [Brockdorff et al., 1992; Brown et al., 1992]. This result means that the inactive X chromosome actually has a transcriptionally active Xist gene at the X-inactivation center. The Xist RNA through some unknown mechanism has a causal role in directing the heterochromatinization of the inactive X-chromosome. The process of heterochromatinization leads to several differences between active and inactive X-chromosomes including DNA methylation and histone acetylation status. DNA methylation is a characteristic of inactive promoters in eukaryotic chromosomes [Section 2.5.7]. Many genes in the inactive X chromosome are heavily methylated in contrast to the active X chromosome [Grant and Chapman, 1988]. However, the kinetics with which particular sites within the X-linked genes become methylated during the differentiation of embryonic female somatic cells do not always correlate with the timing of transcriptional inactivation [Lock et al., 1987]. Although it remains to be determined if a subset of key sites around regulatory elements is always methylated before transcription is repressed it seems probable that other mechanisms must supplement any influence of DNA methylation on transcription.

Heterochromatin normally replicates late during S-phase. Replication timing has been proposed as a determinant of transcriptional activity [Section 4.3]. Genes that replicate late during S-phase might do so under conditions of limiting transcription factors, or might be assembled into a repressive chromatin structure using components translated only late in S-phase. The active X chromosome normally replicates early in S-phase whereas the inactive X chromosome replicates late [Takagi, 1974]. However, female lymphoma cell lines have been isolated in which the opposite occurs [Yoshida et al., 1993]. Thus the inactive X chromosome does not have to replicate late in S-phase in order to be transcriptionally quiescent.

The formation of local repressive chromatin structures in which key genetic regulatory elements are rendered inaccessible to transcription factors by inclusion within positioned nucleosomes is an important mechanism for transcriptional repression [Section 4.2]. Promoters in the inactive X chromosome appear to be incorporated into positioned nucleosomes whereas promoters in the active X chromosome are free of such structures and have transcription factors bound to them [Riggs and Pfeifer, 1992]. Thus specific local chromatin structures clearly appear to have a role in regulating differential gene activity between the two X chromosomes. Nevertheless, a causal relationship between chromatin structure and transcription is yet to be established. Nucleosomes on the active X chromosome contain predominantly acetylated histones whereas those on the inactive X chromosome are not acetylated [Jeppesen and Turner, 1993] [Section 2.4.3]. Thus the establishment and maintenance of specific chromatin structures containing modified histones is an excellent candidate mechanism for establishing and maintaining differential expression of genes between the two X chromosomes. Differential methylation and replication timing may serve to stabilize these different states of gene activity.

Occasionally, an entire nucleus will become heterochromatinized, one example being the inactivation of the erythrocyte nucleus in chicken. Here the special linker histone variant H5 accumulates, which represses transcription and compacts nucleosomal arrays very effectively [Sections 2.5.3 and 3.1.2]. Histone H5 is more arginine rich than the normal linker histone H1 found in somatic cells. This increase in arginine content probably strengthens the interaction of H5 with DNA and stabilizes chromatin structure.

2.5.7 DNA methylation and chromatin

The covalent modification of DNA provides a direct and powerful mechanism to regulate gene expression [Kass et al., 1997a]. Considerable experimental evidence supports the existence of such a mechanism in the majority of plants and animals [Bird, 1986, 1995; Szyf, 1996; Yoder et al., 1997]. The genome of an adult vertebrate cell has 60-90% of the cytosines in CpG dinucleotides methylated by DNA methyltransferase [Riggs and Porter, 1996]. This modification can alter the recognition of the double helix by the transcriptional machinery and the structural proteins that assemble chromatin [Nan et al., 1997; Kass et al., 1997b]. How these events might together contribute to gene control is the major theme of this section.

DNA methylation could control gene activity either at a local level through effects at a single promoter and enhancer, or through global mechanisms that influence many genes within an entire chromosome or genome [Tatte and Bird, 1993]. An attractive suggestion is that DNA methylation evolved as a host-defence mechanism in metazoans to protect the genome against genomic parasites such as transposable elements [Yoder et al., 1997]. An increase in methyl-CpG correlates with transcriptional silencing for whole chromosomes, transgenes, particular developmentally regulated genes and human disease genes [Li et al., 1993; Szyf, 1996]. All of these systems exhibit epigenetic effects on transcriptional regulation in which identical DNA sequences are differentially utilized within the same cell nucleus. These patterns of differential gene activity are clonally inherited through cell division. Because specific methyl-CpG dinucleotides are maintained through DNA replication, DNA methylation states also provide an attractive mechanism [epigenetic mark] to maintain a particular state of gene activity through cell division and, thus, to contribute to the maintenance of the differentiated state [Holliday, 1987]. We discuss how the molecular mechanisms that accomplish this important goal might also involve the assembly of specialized chromatin structures on methylated DNA.

Saccharomyces cerevisiae and Drosophila melanogaster live without any detectable methyl-CpG in their genomes. DNA-methylation-dependent gene regulation is not necessarily essential for cell division or metazoan development, because other gene regulatory mechanisms can compensate for the lack of DNA methylation in these organisms. In the mouse, primordial germ cells, embryonal stem cells and the cells of the blastocyst also progress through the cell cycle and divide without detectable DNA methylation [Jaenisch, 1997]. Nevertheless, once embryonic stem cells begin to differentiate, normal DNA methylation levels are essential for individual cell viability [Panning and Jaenisch, 1996]. The de novo methylation of CpG dinucleotides is a regulated process [Szyf, 1996]. In the embryo, normal DNA methylation levels are essential for post-gastrulation development [Li et al., 1992]. Jaenisch has proposed that DNA methylation has no role in cell viability in mammalian embryonic lineages including the germ line, but that it has an important role in the differentiation of somatic cells. In complex organisms, such as vertebrates, that contain a large number of tissue-specific genes, DNA methylation provides a mechanism to turn-off permanently the transcription of those genes whose activity is not required in a particular cell type [Bird, 1995]. This stable silencing of a large fraction of the genome would allow the transcriptional machinery to focus on those genes that are essential for the expression and maintenance of the differentiated phenotype. Consistent with this hypothesis, the inhibition of DNA methyltransferase activity with 5-azacytidine leads to the activation of several repressed endogenous genes [Jones, 1985]. How might CpG methylation contribute to this global control of gene activity?

The most direct mechanism by which DNA methylation could interfere with transcription would be to prevent the binding of the basal transcriptional machinery and of ubiquitous transcription factors to promoters. This is not a generally applicable mechanism because some promoters are transcribed effectively as naked DNA templates independent of DNA methylation [Busslinger et al., 1983; Iguchi-Ariga et al., 1989; Kass et al., 1997b]. Certain transcription factors [e.g. the cyclic AMP dependent activator CREB] bind less well to methylated recognition elements, however the reduction in affinity is often insufficient to account for the inactivity of promoters in vivo [Hoeller et al., 1988; Weih et al., 1991]. It seems unlikely that DNA methylation would function to repress transcription globally by modifying the majority of CpGs in a chromosome, if the only sites of action are to be a limited set of recognition elements for individual transcription factors.

The second possibility is that specific transcriptional repressors exist that recognize methyl-CpG and, either independently or together with other components of chromatin, turn off transcription. This mechanism would have the advantage of being substantially independent of DNA sequence itself, thereby offering a simple means of global transcriptional control. It would be especially attractive if the methylation-dependent repressors work in a chromatin context because then DNA could maintain the nucleosomal and chromatin fibre architecture necessary to compact DNA [Jost and Hofsteenge, 1992; McArthur and Thomas, 1996]. Moreover, because chromatin assembly also represses transcription, methylation-dependent repression mechanisms would add to those already in place.

Bird and colleagues have identified two repressors MeCPl and MeCP2 that bind to methyl-CpG without apparent sequence specificity [Meehan et al., 1989, 1992; Lewis et al., 1992]. Like DNA methylation itself, MeCP2 is dispensable for the viability of embryonic stem cells, however it is essential for normal embryonic development. Consistent with the capacity of methylation-dependent repressors to operate in chromatin, recent studies indicate that MeCP2 is a chromosomal protein with the capacity to displace histone H1 from the nucleosome [Nan et al., 1996]. Moreover, MeCP2 contains a methyl-CpG DNA-binding domain, which might alter chromatin structure directly, and a repressor domain, which might function indirectly to confer long-range repression in vivo [Nan et al., 1993, 1997]. The capacity for MeCP2 to function in chromatin explains several phenomena connected with unique aspects of chromatin assembled on methylated DNA.

A role for specialized chromatin structures in mediating transcriptional silencing by methylated DNA has been suggested by several investigators. High levels of methyl-CpG correlate with transcriptional inactivity and nuclease resistance in endogenous chromosomes [Antequera et al., 1989, 1990]. Methylated DNA transfected into mammalian cells is also assembled into a nuclease-resistant structure containing unusual nucleosomal particles [Keshet et al., 1986]. These unusual nucleosomes migrate as large nucleoprotein complexes on agarose gels. These complexes are held together by higher-order protein–DNA interactions despite the presence of abundant micrococcal nuclease cleavage points within the DNA. Individual nucleosomes assembled on methylated DNA appear to interact together more stably than on unmethylated templates [Keshet et al., 1986]. Each nucleosome normally contains an octamer of core histones [H2A, H2B, H3 and H4] around which is wrapped approximately 160 bp of DNA, and a single molecule of histone H1, which constrains the linker DNA between adjacent nucleosomes [Section 2.1]. The replacement of histone H1 with MeCP2 is a possible explanation for the assembly of a distinct chromatin structure on methylated DNA [Nan et al., 1997].

The accessibility of chromatin to nucleases could also be affected directly by the stability with which the histones interact with DNA within the nucleosome. DNA methylation does not influence the association of core histones with the vast majority of DNA sequences in the genome [Felsenfeld et al., 1983; Englander et al., 1993]. However, for certain specific sequences, such as those found in the Fragile X mental retardation gene 1 promoter, methylation of CpG dinucleotides can alter the positioning of histone–DNA contacts and the affinity with which these histones bind to DNA [Godde et al., 1996; Section 2.2.5]. The exact chromatin structure found in vivo can also be a consequence of gene activity. Linker histones, such as H1, are relatively deficient on the transcribed region of genes [Kamakaka and Thomas, 1990]. So it is not surprising that transcriptionally inactive chromatin containing methyl-CpG should show an increase in the abundance of histone H1, whereas DNA sequences lacking methyl-CpG are deficient in H1 [Ball et al., 1983; Tazi and Bird, 1990]. In vitro studies indicate that histone H1 can interact preferentially with methylated DNA under certain conditions, although there is no measurable preference for the assembly of H1 into a nucleosomal architecture containing methylated DNA [Levine et al., 1993; Campoy et al., 1995; Nightingale and Wolffe, 1995]. Recent in vivo studies indicate that rather than functioning as a general transcriptional repressor, histone H1 is highly specific with respect to the genes whose activity it regulates [see Section 2.5.1]. It seems probable that the major differences between chromatin assembled on methylated versus unmethylated DNA will be determined by the inclusion of methyl · ation-specific DNA-binding proteins, such as MeCP2.

There are features of transcriptional repression dependent on methylated DNA that can be explained by methylation-specific repressors operating more effectively within a chromatin environment. Transcriptional repression is strongly related to the density of DNA methylation [Boyes and Bird, 1992; Hsieh, 1994]. There is a nonlinear relationship between the lack of repression observed at low densities of methyl CpG and repression at higher densities. These results led to the demonstration that local domains of high methyl-CpG density could confer transcriptional repression on unmethylated promoters in cis [Kass et al., 1993]. This observation is consistent with MeCP2 containing DNA binding and transcriptional repression domains [Nan et al., 1997]. Thus MeCP2 does not necessarily have to function by occluding regulatory elements from the transcriptional machinery [by binding to promoter sequences itself], but might bind to a methyl-CpG sequence at one place on a DNA molecule and then use the repression domain to silence transcription at a distance. Chromatin assembly itself might promote this ‘action at a distance’ by juxtaposing MeCP2 and the regulatory elements under control through the compaction of intervening DNA [Fig. 2.58].

Early experiments using the microinjection of templates into the nuclei of mammalian cells suggested that the prior assembly of methylated, but not unmethylated, DNA into chromatin represses transcription [Buschhausen et al., 1987]. The importance of a nucleosomal infrastructure for transcriptional repression dependent on DNA methylation was reinforced by the observation that immediately after injection into Xenopus oocyte nuclei, methylated and unmethylated templates both have equivalent activity [Kass et al., 1997b]. However, as chromatin is assembled, the methylated DNA is repressed with the loss of DNase I hypersensitivity and the loss of engaged RNA polymerase. The requirement for nucleosomes to exert efficient repression can be explained in several ways. The repression domain of MeCP2 might recruit a co-repressor complex that directs the modification of the chromatin template into a more stable and transcriptionally inert state [Fig. 2.58]. One potential candidate corepressor for MeCP2 is the SIN3-histone deacetylase complex, because inhibition of histone deacetylation can reverse some of the transcriptional repression conferred by DNA methylation [Jones et al., 1998]. MeCP2 also copurifies with SIN3 and histone deacetylase in Xenopus oocytes [Jones et al., 1998] and MeCP2 and SIN3 interact in in vitro binding studies [Nan et al., 1998]. Alternatively, like histone H1, MeCP2 might bind more efficiently to nucleosomal rather than to naked DNA. Any co-operative interactions between molecules could propagate the association of MeCP2 along the nucleosomal array even into unmethylated DNA segments [Fig. 2.58]. This latter mechanism is analogous to the nucleation of heterochromatin assembly at the yeast telomeres by the DNA-binding protein RAPI, which then recruits the repressors SIR3p and SIR4p that organize chromatin into a repressive structure [Grunstein et al., 1995; Hecht et al., 1996]. All of these potential mechanisms could individually or together contribute to the assembly of a repressive chromatin domain. Although these molecular mechanisms are speculative, they illustrate the advantages of a nucleosomal infrastructure. It should be noted that MeCP2 can repress transcription in an in vitro extract, although this might be by direct occlusion of transcription factor binding over the methylated promoter.

If methylated DNA directs the assembly of a specialized repressive chromatin structure, it might be anticipated that the transcriptional machinery will have less access to such a structure than the orthodox chromatin assembled on unmethylated promoters and genes. Activators such as Gal4-VP16 can normally penetrate a preassembled chromatin template to activate transcription, even in the presence of histone H1 [Laybourn and Kadonaga, 1992]. However once chromatin has been assembled on methylated DNA, Gal4-VP16 can no longer gain access to its binding sites and activate transcription. This suggests that the specialized features of chromatin assembly on methylated DNA provide a molecular lock to silence the transcription process permanently [Siegfried and Cedar, 1997]. This capacity of DNA methylation to strengthen transcriptional silencing in a chromatin context could be an important contributor to the separation of the genome into active and inactive compartments in a differentiated cell.

DNA methyltransferase maintains the methyl CpG content of both daughter DNA duplexes following replication [Holliday, 1987]. Methyltransferase localizes to the chromosomal replication complex and maintenance methylation takes place less than one minute after replication [Leonhardt et al., 1992; Gruenbaum et al., 1983]. By contrast, chromatin assembly takes 10–20 minutes in a mammalian tissue culture cell [Cusick et al., 1983]. Histone deposition occurs in stages, and it is not until a complete histone octamer is assembled with DNA that histone H1 is stably sequestered [Worcel et al., 1978]. Comparable limitations might restrict the stable association of methylation-specific repressors. This would account for the lag time before methylated DNA is repressed following injection as a naked template into the nuclei of mammalian tissue culture cells or Xenopus oocytes [Kass et al., 1997b; Buschhausen et al., 1987].

A significant feature of transcriptional repression on methylated DNA is that it is not only time dependent but also potentially dominant [Kass et al., 1997b]. Thus, at early times when chromatin assembly is incomplete, the transcriptional machinery has the potential to associate with methylated regulatory DNA. As chromatin structure matures, the basal transcriptional machinery is potentially erased from the template. This provides a general mechanism for the global silencing of transcription dependent only on DNA methylation state [Fig. 2.59], although Gal4-VP16 cannot function if chromatin is assembled on methylated DNA before exposure to the activator. If a very strong activator, such as Gal4-VP16, is present during chromatin assembly then transcriptional activity can resist methylation-dependent transcriptional silencing [Fig. 2.60]. Therefore, under certain circumstances, regulatory nucleoprotein complexes might be assembled that resist this powerful silencing mechanism. Such a mechanism has been suggested to be dependent on SP1 sites in the promoter of a housekeeping gene in the mouse [adenine phosphoribosyl-transferase] that is maintained in a methylation-free state [Macleod et al., 1994]. For example, if components of regulatory complexes could bind to DNA immediately after replication with reasonable efficiency and before DNA methyltransferase can begin to modify the template, then they might prevent DNA methylation around their binding sites. These sequences might then become progressively demethylated and eventually resist transcriptional repression [Fig. 2.60]. This would provide a mechanism for the demethylation of regulatory DNA in particular differentiated cell lines.

Other mechanisms might contribute to the maintenance of transcriptional repression through DNA synthesis. The assembly of a specialized chromatin structure on methylated DNA might result in the presence of additional proteins [e.g. MeCP2] and histone modifications [e.g. histone deacetylation] that could be maintained in daughter chromatids. Nucleosomes segregate dispersively in small groups to daughter DNA molecules at the replication fork [Sogo et al., 1986]. Particular modified histones and repressors such as MeCP2 would be anticipated to segregate within the nucleosomal context [Perry et al., 1993]. These proteins could therefore provide at least 50% of the chromatin proteins necessary to restrict transcription. Their continued presence on DNA could help to re-establish transcriptional repression on both daughter chromatids through any of the mechanisms illustrated in Fig. 2.58. Therefore, demethylation alone might be insufficient to relieve transcriptional repression until successive cell divisions eventually unravel the repressive chromatin structure.

Although I focus on molecular mechanisms that might influence DNA methylation and gene expression in dividing cells, DNA demethylation is also important in non-dividing terminally differentiated cells. Under these circumstances demethylation at particular promoters must occur in the absence of replication [Sullivan and Grainger, 1986; Saluz et al., 1986]. Presumably mechanisms must also exist to destabilize any repressive chromatin structure associated with methylated DNA in order to allow the demethylation machinery access to the template.

The importance of DNA methylation and methylation-specific DNA-binding proteins for the viability of a differentiated mammalian somatic cell is well documented. An attractive explanation for the importance of DNA methylation is that it helps to turn off transcription from the large number of genes not required in a particular differentiated cell through global mechanisms. The major problem is in determining how this global repression might first be achieved and then maintained through successive cell generations? We have described evidence for methylation-specific and chromatin-dependent transcriptional repression mechanisms operating in vivo. Recent studies discussed here suggest that these two mechanisms operate together to regulate gene expression more tightly. There is now excellent precedent for transcriptional activators and repressors operating most effectively in a nucleosomal environment. Clearly, future experiments should explore how MeCP2 is incorporated into a nucleosomal array, together with the physical and functional consequences of this inclusion. The potential interaction of the MeCP2 repression domain with co-repressor complexes that might modify chromatin is also an area of active interest. Finally, a mutual dependence on DNA methylation and chromatin assembly for transcriptional silencing provides a potential mechanism not only for the stable propagation of the repressed state through cell division, but also for the targeted demethylation of promoter DNA. The availability of replication systems capable of propagating methylation states as well as directing chromatin assembly will allow this model to be tested directly [Harland, 1982].

2.5.8 The HMGs and related proteins

Chromatin structure can be modified by the selective association of abundant non-histone proteins that interact with DNA histone complexes. Primary among these are the high mobility group proteins. Early methodologies for the fractionation of the linker histone H1 employed perchloric acid extraction of chromatin. Several other proteins were also found to be solubilized during this process. Later it was noticed that extraction of chromatin at moderate ionic strengths [0.35 M NaCl] released similar proteins. The addition of trichloroacetic acid [2%] to these salt-extracted proteins separated them into an insoluble fraction of large proteins [low-mobility group, LMG, when molecular size was assayed by gel electrophoresis] and a soluble fraction of small proteins [a group of high-mobility proteins during electrophoresis, HMG] [Johns, 1982]. Four major proteins are found in the HMG group. These fall into two classes: HMG1 and 2 are one pair of homologous proteins [~29 000 Da in size]; HMG14 and 17 are the other [10 000–12 000 Da in size]. The content of HMG14 and 17 in chromatin may range up to 10% of DNA weight, similar to that of histone H1. In addition there are also several minor HMG proteins, for example HMG-I/Y which binds to the α-satellite sequences in the centromere [Bustin and Reeves, 1996; Section 2.4.2].

The genes encoding all four of the major HMGs have been cloned. The HMG14 and 17 proteins are highly conserved from human to chicken, certain basic stretches of amino acids being completely identical. These amino terminal basic regions are believed to interact with nucleosomal DNA. HMG14 and 17 also have an acidic carboxyl-terminal tail [Srikantha et al., 1988]. Both proteins bind selectively to nucleosomal DNA in preference to naked DNA of a comparable length. It appears that two HMG molecules can bind per core particle [Mardian et al., 1980; Sandeen et al., 1980; Crippa et al., 1992]. Surprisingly both HMG-14 and HMG-17 bind as homodimers to nucleosomes but do not interact together directly suggesting that they induce specific allosteric transitions in the nucleosome core to promote this selective association [Postnikov et al., 1995]. One model based on chemical cross-linking suggests that the HMG14 and 17 proteins can interact with DNA where it exits and enters the nucleosome [Shick et al., 1985; Alfonso et al., 1994]. Incorporation of HMG14 and 17 alters the stability of protein–DNA interactions at the nucleosome boundaries. This leads to a change in micrococcal nuclease digestion and potentially in the spacing of nucleosome cores [Crippa et al., 1993; Tremethick and Drew, 1993]. In this respect HMG14 and 17 function in the nucleosome very much like some models for linker histone function [Pruss et al., 1995]. Like acetylation or phosphorylation of the core histones, interaction of HMG14 and 17 at the nucleosomal boundaries is likely to modify histone H1 interaction and hence higher-order chromatin structure. Although definitive proof is lacking, considerable circumstantial evidence suggests that HMG14 and 17 are involved in potentiating the transcription of genes in vivo [Einck and Bustin, 1985; Section 4.3].

Recent experiments have examined the consequences for basal transcription of incorporating HMG14 and 17 into chromatin during the assembly process [Crippa et al., 1993; Tremethick, 1994; Trieschmann et al., 1995a; Paranjape et al., 1995]. Deletion mutagenesis of HMG-14 and HMG-17 suggests that the negatively charged C-terminal region of the proteins is required for transcriptional enhancement [Trieschmann et al., 1996]. In general, there is a modest 10-fold increase in transcription that appears related to a more accessible chromatin structure. The observation that RNA polymerase II elongation is stimulated by inclusion of HMG14 into chromatin is also consistent with a less compact, more accessible chromatin environment in which transcription can occur more efficiently [Ding et al., 1994]. A role for HMG14 and 17 in transcription would explain their recruitment to chromatin domains associated with elongating RNA polymerase within poly tene chromosomes [Section 2.4.3].

Rather more functional information is available concerning the other pair of HMG proteins 1 and 2. These proteins have attracted a great deal of attention since conserved amino acid sequence motifs within these proteins are also found in transcription factors [Grosschedl et al., 1994]. HMG1 and 2 have a basic amino terminus and an acidic carboxyl terminus. The carboxyl terminus influences DNA binding selectivity [Wisniewski and Schulze, 1994]. The basic region contains two HMG boxes. Each HMG box is a protein domain consisting of an L-shaped arrangement of three α-helices containing two independent DNA-binding surfaces [Read et al., 1993; Weir et al., 1993]. A single HMG domain may cover 20 bp at a specific binding site and potentially distort the DNA molecule through as much as 130°. HMG1 and 2 also recognize unusual DNA structures such as cruciform DNA, which may simply reflect the presence of multiple DNA-binding sites on the same protein [Bianchi et al., 1989; Lilley, 1992]. Linker histones have similar properties [Varga-Weisz et al., 1994]. It has been suggested that this reflects the affinity of linker histones and potentially HMG1 and 2 for DNA where it enters and exits wrapping around the histone octamer. At this site DNA might cross-over itself [Section 2.3.1]. Recent experiments demonstrate that HMG1 can replace linker histones in chromatin [Nightingale et al., 1996; Ura et al., 1996]. Specific HMG domain proteins have now been defined that control lymphoid transcription, mating-type switching [SIN1, see Section 2.5.4] and mammalian sex determination [Pentiggia et al., 1994; Werner et al., 1995]. Proteins with a single HMG domain associate with DNA sites relatively weakly, probably because of the energy required to direct the distortion of inflexible DNA. However, other proteins often contain several HMG domains, which form more stable complexes with DNA.

Most notable among the sequence-specific HMG domain transcription factors is the protein upstream binding factor or UBF, involved in the transcriptional regulation of mammalian ribosomal RNA genes [Jantzen et al., 1990]. The UBF protein contains five HMG domains flanked by an amino-terminal dimerization motif and an acidic carboxyl-terminal tail. Any adjacent pair of HMG domains will bind to DNA; however, the selectivity of binding is conferred by the three domains closest to the amino terminus [Le Blanc et al., 1993; Hu et al., 1994]. Each UBF dimer contains 10 HMG domains, a binding site that potentially includes up to 200 bp of DNA. This extended region contains a site of DNA distortion every two turns of the double helix as a consequence of the binding of an HMG domain. Because these sites occur on the same face of the helix, DNA is bent into a superhelical turn around the contiguous HMG domains [Bazett-Jones et al., 1994]. Deoxyribonuclease I digestion of UBF-DNA complexes reveals a 10- to 11-bp periodicity of cleavage that is reminiscent of the access that this enzyme has to DNA wrapped around the histones within the nucleosome [Dunaway, 1989; Section 2.2.3]. Looping appears to be facilitated not only by DNA deformation directed by the HMG domains but also by protein–protein interactions between the acidic carboxyl-terminal tail and the HMG domains toward the amino terminus of the UBF molecule.

How might the wrapping of DNA by UBF facilitate the transcription process? Like other eukaryotic genes, transcription by RNA polymerase I requires TBP, which in this system is a component of a sequence-specific transcription factor SL1. UBF and SL1 appear to bind co-operatively to the ribosomal promoter to form a stable complex that recruits RNA polymerase [Bell et al., 1988]. Two binding sites for SL1 are separated by 120 bp within the DNA wound around UBF. These sites function co-operatively, are separated by an integral number of helical turns of DNA, and remain exposed to the solution within the UBF–ribosomal promoter complex. UBF provides the correct scaffolding for productive interaction between individual SL1 molecules bound at the two recognition sites within each complex [Fig. 2.61]. In this way UBF increases the probability and stability of transcription complex formation [McStay et al., 1997].

The architectural role proposed for the UBF transcription factor has been seen with other HMG domain proteins [Giese et al., 1992; Ferrari et al., 1992]. An HMG domain architectural transcription factor LEF-1 [lymphoid enhancer-binding factor, Grosschedl et al., 1994] interacts with a cytoplasmic protein, β-catenin which links the cadherin cell adhesion molecule to the cytoskeleton [Behrens et al., 1996]. LEF-1 and β-catenin bind together to DNA and induce a specific bend in the double-helix. If LEF-1 and β-catenin are coexpressed in Xenopus embryos the axis of the embryo is duplicated reflecting aberrant cell signalling. Thus HMG domain proteins can be components of signal transduction pathways from cell adhesion components to the cell nucleus. In several similar cases, the association of the HMG domain with DNA directs the assembly of clusters of transcription factors bound to DNA into precise higher-order nucleoprotein complexes [Sections 4.1].

The role of the relatively abundant HMG1 and 2 proteins themselves in chromatin remains enigmatic. It has been proposed that these proteins might promote nucleosome assembly [Bonne-Andrea et al., 1984] or prevent nucleosome assembly [Waga et al., 1989]. They have been reported to facilitate transcription [Tremethick and Molloy, 1986] and to inhibit transcription [Ge and Roeder, 1994a; Stelzer et al., 1994]. HMG1 will repress transcription selectively from nucleosomal templates by positioning nucleosomes and restricting octamer mobility [Ura et al., 1996]. The proteins may function as stable components of nucleoprotein structures [Pauli et al., 1993] or transiently as assembly factors required to bend DNA and then dissociate [Travers et al., 1994]. HMG1 appears to function in an analogous manner to the sequence-specific HMG domain proteins in facilitating the binding of the progesterone receptor transcription factor to its recognition element [Onate et al., 1994]. Normally HMG1 and 2 are associated with a relatively minor fraction of chromatin [< 5%; Goodwin et al., 1977; Isackson et al., 1980]. It has been proposed that HMG1 might be capable of functionally replacing linker histones within chromatin [Jackson et al., 1979; Nightingale et al., 1996]. This idea receives support from the observation that early embryonic chromatin in Drosophila and Xenopus is highly enriched in HMG1- and 2-like proteins [Dimitrov et al., 1994; Ner and Travers, 1994]. This strongly suggests that HMG1 and 2 primarily fulfil a structural role in chromatin.

A negative regulator of inducible transcription in yeast known as SIN1 [Kruger and Herskowitz, 1991] [Section 2.5.4] has a very similar structure to HMG1. It is possible that SIN1 might help to fulfil the regulatory role of linker histones [Wolffe, 1994b].

Sequences rich in glutamic and aspartic acid residues [Section 2.4.2], are found in HMG1 and 2, the centromeric protein CENP-B, chromatin assembly proteins N1/N2, nucleoplasmin [Section 3.2], and topoisomerase I. These anionic regions have been postulated to have the capacity to interact directly with histones, since N1/N2 interacts specifically with histones H3/H4 and nucleoplasmin with H2A/H2B. This property may account for the role of HMG1 and 2 in nucleosome assembly. The physiological significance of this assembly activity is unknown [Section 3.2]. It has also been postulated that the acidic regions found within transcription factors might cause a local destabilization of nucleosome structure, perhaps by competing with DNA for interaction with histones, especially histones H2A/H2B.

2.5.9 Functional compartmentalization of the nucleus

Functional compartmentalization in the eukaryotic cell is readily accepted from observation of membrane-bounded organelles that can be fractionated and their properties determined in isolation. The existence of discrete compartments within a given organelle is less immediately apparent, but none the less real. Within the eukaryotic nucleus several independent approaches point to the compartmentalization of particular activities such as transcription, RNA processing and replication. Chromosomes are revealed to occupy defined territories and to represent highly differentiated structures. The numerous activities that use DNA and RNA as a template occur with a defined spatial and temporal relationship. A remarkable fusion of methodologies including cell biology, molecular genetics and biochemistry has contributed to the recognition of nuclear architecture as defined by function [Strouboulis and Wolffe, 1996].

The most dramatic example of the compartmentalization of nuclear function is seen with replication. Analysis of DNA synthetic sites with bromodeoxyuridine or biotinylated dUTP reveals only 150 foci of incorporation within each nucleus during S phase [Nakamura et al., 1986]. The foci are clearly defined with a clear and relatively uniform separation from each other. When replication initiates, these foci are small and appear as ‘dots’, as time progresses they become more diffuse [Manders et al., 1992; O’Keefe et al., 1992]. These foci contain accumulations of the proteins necessary for replication: DNA polymerase a, PCNA, and RP-A as well as regulatory molecules such as cyclin A, cdk2, and RPA70 [Adachi and Laemmli, 1992; Hozak et al., 1993; Cardoso et al., 1993; Sobczak-Thepot et al., 1993]. Immunolabelling synthetic sites with gold particles suggests that nascent DNA is extruded from the replication foci [Hozak et al., 1993]. This implies that DNA moves through a fixed architecture containing the molecular machines directing replication. The advantages of the compartmentalization of DNA replication include a concentration of the necessary regulatory, structural and enzymatic components required to duplicate both DNA and chromosomal structure. The staged assembly of a functional replication elongation complex occurs within a defined macromolecular complex, this allows many check points and controls to be built into the initiation of replication [Almouzni and Wolffe, 1993b].

The essential role of nuclear architecture in determining the functional properties of DNA is perhaps most apparent in connection with chromosomal replication in Xenopus laevis eggs. Injection of prokaryotic DNA into an egg or incubation of the DNA in an egg extract leads to the assembly of a pseudonucleus competent to replicate DNA [Forbes et al., 1983; Blow and Laskey, 1986; Section 3.1]. Importantly, replication is regulated spatially in that it occurs at discrete sites containing clusters of replication forks [Cox and Laskey, 1991]. There is a remarkable similarity between the number and distribution of replication ‘foci’ in the pseudonuclei and those observed in replicating eukaryotic nuclei in tissue culture cells [Nakamura et al., 1986; Mills et al., 1989]. The assembly of functional replication origins is not necessarily dependent on defined DNA sequences in the chromosomes, but on features of nuclear architecture that can be assembled even on prokaryotic DNA. The implication is that general features of nuclear architecture can impose a particular function, in this case that of replication. It should be noted that the early Xenopus embryo is a special case in which normal somatic controls might have been relaxed. Chromatin loop attachments to the chromosomal axis and the number of chromosomal origins of replication are much more frequent in the chromosomes of early embryonic nuclei in Xenopus compared to somatic cell nuclei [Laskey et al., 1983; Micheli et al., 1993]. Only when the cell cycle lengthens at the mid blastula transition are normal controls established [see Hyrien et al., 1995].

In normal somatic nuclei the replication foci do not all engage in replication simultaneously, some are utilized early in S-phase and others late in S-phase [Nakamura et al., 1986; Ariel et al., 1993]. This reflects differential replication timing, which is an important regulatory step in maintaining local chromatin organization and gene activity [Wolffe, 1991c]. The molecular mechanisms controlling the differential utilization of origins are presently unknown [Gilbert, 1986; Guinta and Korn, 1986; Wolffe, 1993]. However, comparable phenomena occur in yeast [Newlon et al., 1993], where origin utilization is found to be dependent on chromosomal position. This is an important area for future study.

These observations on DNA replication, the dependence on nuclear architecture and the movement of the DNA strand through a fixed site have led to speculation that comparable phenomena also govern transcription [Cook, 1994; Hughes et al., 1995]. The concentration of RNA polymerase within defined nuclear compartments together with other components of the transcriptional machinery lends some support to these ideas [see later]. RNA polymerase may be much less mobile and chromatin more mobile than generally considered. An important additional point is that the assembly of a particular nucleoprotein architecture that favors one biological process, e.g. replication, might exert an exclusionary or repressive influence on another, e.g. transcription [see Wansink et al., 1994]. In fact in the S. cerevisiae chromosome, components of the Origin Recognition Complex required for replication exert a silencing effect on transcription [Fox et al., 1993; Section 2.5.6].

Ribosomal gene transcription, rRNA processing and preribosomal particle assembly occur in the nucleolus [Scheer and Benavente, 1990]. All of these events involve the assembly of macromolecular machines that localize within this specialized nuclear compartment. The molecular mechanisms that direct particular proteins and enzyme complexes to this compartment and retain them there are largely unknown [see Hatanaka, 1990]. One simple hypothesis is that the majority of the nucleolar architecture is generated from the activities of the transcriptional machinery itself which assembles reiterated regulatory nucleoprotein complexes on rDNA. Ribosomal RNA genes are tandemly arrayed with approximately 250 copies of a 44 Kb repeat in humans [Scheer and Benavente, 1990]. Thus, more than 10 × 106 bp of rDNA and associated proteins could provide the framework for the nucleolus. Once transcription itself is in progress, additional features of nucleolar architecture would potentially follow from the accumulation and activities of the molecular machines that process pre-rRNA and that assemble ribosomes.

Morphologically the nucleolus has three major organizational areas: [1] the nucleolar fibrillar centers, which are surrounded by [2] a dense fibrillar region, and [3] the granular region. Numerous localization studies using specific antibodies and hybridization probes indicate that the nuclear fibrillar centers are the sites where the ribosomal RNA genes, RNA polymerase I, the class I gene transcription factor UBF and topoisomerase I are localized [Scheer and Rose, 1984; Raska et al., 1989; Rendon et al., 1992; Thiry, 1992a, b]. Accumulation of these particular macromolecules leads to the inference that the nucleolar fibrillar centres are the assembly sites for the regulatory nucleoprotein complexes that direct transcription [Fig. 2.62]. The dense fibrillar component that surrounds the nucleolar fibrillar center consists of nascent ribosomal RNA and associated proteins. It is in the dense fibrillar component that RNA precursors such as [3H]uridine or biotinylated ribonucleotides are initially found on pulse labelling [Thiry and Goessens, 1992]. Specfic hybridization probes localize unprocessed nascent transcripts and associated processing machinery to the dense fibrillar component [Ochs et al., 1985; Kass et al., 1990; Puvion-Dutilleul et al., 1991]. Mature 28S and 18S rRNA, partially processed transcripts and intermediates in ribosome assembly are found in the granular region. These assembly intermediates are visualized as particles 15–20 nm in diameter. Movement of preribosomal subunits from the nucleolar granular region to the cytoplasm might be facilitated by proteins that move within specific pathways or tracks from the nucleolus to the nuclear envelope [Meier and Blobel, 1992].

This hierarchical organization of the nucleolus with particular morphologically distinct compartments reflecting accumulations of specialized molecular machines and their substrates provides an extremely useful model with which to consider the functional compartmentalization of mRNA synthesis, processing and export.

The synthesis of mRNA within the nucleus and the subsequent delivery of the mature transcript to the translational machinery within the cytoplasm also involves the concerted and co-ordinated activities of multiple molecular machines. Transcription requires the assembly of a regulatory nucleoprotein complex, containing the promoter region, associated coactivators and the RNA polymerase holoenzyme. The polymerase must initiate RNA synthesis and traverse the gene. The pre-mRNA must be processed through the addition of a m7G[5′]pp cap, removal of introns [splicing] and polyadenylation. These various biochemical events are interdependent since transcription by RNA polymerase II is a prerequisite for both efficient splicing and polyadenylation [Sisodia et al., 1987], and the 52032 cap and associated proteins also facilitate splicing and mRNA export [Izaurralde et al., 1994; Lewis et al., 1995].

The transcriptional machinery that synthesizes pre-mRNA localizes with the perichromatin fibrils found at the boundaries of condensed chromatin domains. Perichromatin fibrils are nuclear ribonucleoprotein complexes with a diameter varying from 3 nm to 20 nm. They are enriched in nascent pre-mRNA radiolabelled with [3H]uridine or bromouridine, and fibril density correlates with transcriptional activity [Bachellerie et al., 1975; Fakan, 1994; van Driel et al., 1995]. Components of the splicing machinery are found with the perichromatin fibrils [Fakan et al., 1984] consistent with the assembly of the splicing machinery initiating at the site of transcription. This is visualized through immunofluorescent probe detection of the splicing machinery as diffuse nucleoplasmic staining [Spector, 1993]. Considerable morphological and molecular biological evidence indicates that splicing occurs concomitant with transcription [Beyer and Osheim, 1988; Le Maire and Thummel, 1990; Wuarin and Schibler, 1994].

The localization of specific transcripts such as fibronectin premRNA using hybridization probes reveals elongated ‘tracks’ or more compact ‘dots’ at one or two discrete sites corresponding to the chromosomal copies of the gene [Xing et al., 1993, 1995; Huang and Spector, 1991]. Simultaneous RNA and DNA hybridization demonstrates that transcribing genes directly localize with the RNA tracks or dots, with the gene at one end of the track [Xing et al., 1993, 1995; Xing and Lawrence, 1993]. Moreover probes for introns only detect the track near the gene, suggesting that splicing occurs along the track. These results suggest a model of cotranscriptional assembly of the splicing machinery onto pre-mRNA at the perichromatin fibrils, with splicing continuing as the pre-mRNA is released from the gene.

RNA tracks can have a very close association with discrete structures known as interchromatin granules or ‘speckles’ [Xing et al., 1993]. These speckles are sites at which the splicing machinery accumulates together with intron containing pre-mRNA and polyadenylated mRNA [Spector, 1990; Fu and Maniatis, 1990; Carter et al., 1991, 1993; Visa et al., 1993a; Wang et al., 1991]. Thus speckles potentially represent sites of pre-mRNA processing and of mature mRNA accumulation in the nucleus.

Lawrence and colleagues have suggested that actively transcribing genes have a non-random association with the speckles [Xing and Lawrence, 1993; Lawrence et al., 1993; Xing et al., 1995], the implication of this association being that in certain instances speckle structures represent the sites of transcription itself. An immediate limitation to this latter hypothesis is that there are only 20–50 speckles scattered in a punctate distribution throughout the nucleus. Clearly not every active gene can be associated with these structures. However, out of ten transcribing genes investigated in the Lawrence laboratory, seven are associated with speckles [Xing et al., 1995]. It has also been hypothesized that these transcription domains might contain several actively transcribing genes at any one time [Xing et al., 1993; Xing and Lawrence, 1993]. In contrast to this view, the bulk of nascent premRNA labelled with bromouridine accumulates in a distinct pattern which does not correspond to that of speckles containing snRNPs [Jackson et al., 1993; Wansink et al., 1993]. Adenoviral and actin transcripts can be visualized as discrete dots in the nucleus, with no apparent association with snRNP speckles [Zhang et al., 1994]. However, a more detailed study indicates a clear association between actin transcripts and speckles [Xing et al., 1995]. An additional complication is that not all RNA detected in these assays is premRNA, but an ill defined proportion might correspond to a large pool of stable nuclear polyadenylated RNA involved in structural functions within the nucleus [Huang et al., 1994; Mattaj, 1994].

Much of the evidence presented in support of the association between sites of transcription and speckles is based on the transcription of very active genes [e.g. collagen, which comprises 4% of total mRNA in fibroblasts; and the induced expression of the fos gene after serum starvation]. It could be envisaged that the speckles, rich in splicing components, act as processing factories closely associated only with the most actively transcribing genes.

Once synthesized and assembled with the splicing machinery, the pre-mRNA has to reach the nuclear envelope and enter the cytoplasm. The pre-mRNA is packaged not only with the splicing apparatus but also with heterogeneous nuclear ribonucleoproteins [hnRNPs]. These proteins provide the ‘workbench’ on which mRNA is processed. Like the packaging of DNA with histones, the resulting architectures are important for the maturation of mRNA [Dreyfuss et al., 1993]. The preparation of nuclear matrix, a process that removes the vast majority of chromatin from the nucleus, retains pre-mRNA, hnRNPs and some elements of the splicing machinery [Huang and Spector, 1991; Mattern et al., 1996]. This is indicative of both the abundance and major structural role for hnRNPs in the nucleus. In general, hnRNPs are diffusely distributed throughout the nucleoplasm, however a subset overlap snRNP speckles, and some even shuttle with mRNA into the cytoplasm before returning to the nucleus [Visa et al., 1996; Pinol-Roma and Dreyfuss, 1992]. The movement of a specific pre-mRNA from gene to cytoplasm has been reconstructed based on the export pathway of the Balbiani ring [BR] pre-mRNP particles in the dipteran Chironomus tentane [Mehlin and Daneholt, 1993]. BR genes are easily visualized as two giant puffs in the polytene chromosomes of the salivary glands. The large nascent transcripts from these puffs are assembled with hnRNPs and the splicing machinery during transcription to form a thin fibre, which, with elongation, becomes thicker and bends into a ring-like structure. The mature pre-mRNP granule, now thought to contain spliced RNA is released into the nucleoplasm making its way to the nuclear envelope, where it positions itself against a nuclear pore and becomes elongated into a rod-shaped structure which goes through the pore in a 5’-head-first manner. As the pre-mRNP granule emerges on the cytoplasmic side it immediately becomes associated with ribosomes [Mehlin and Daneholt, 1993]. The important point for this discussion is that the entire process occurs within precise nucleoprotein architectures.

How does the pre-mRNA reach the nuclear membrane from the sites where transcription takes place? Estimates of the rate of movement of pre-mRNA have been made using a highly expressed hybrid gene in Drosophila salivary glands, which gives a strong signal at the site of transcription and a more diffuse channel-like network pattern throughout the nucleoplasm [Zachar et al., 1993]. These results have been interpreted to suggest that simple diffusion alone could account for the dispersal of mRNA. However, there is also evidence to suggest that pre-mRNA movement through the nucleoplasm occurs in a directed fashion. For example, the need for particular structural features in the pre-mRNAs for their movement into the cytoplasm [Elliot et al., 1994] is inconsistent with a simple diffusion model. Furthermore, the tight association of transcripts, hnRNPs, and functional processing components [e.g. the splicing machinery] with the nuclear skeleton and nuclear matrix argues against the pre-mRNA being freely diffusible in the nucleoplasm. Huang and Spector [1991] were able to visualize ‘tracks’ corresponding to fos gene transcripts frequently extending to the nuclear envelope and exiting over a limited area. This would appear to give credence to one aspect of the gene gating model proposed by Blobel [1985]. This postulates that due to overall three-dimensional architectural constraints in the nucleus, genes will associate with a specific region of the nuclear envelope, hence their transcripts are ‘gated’ to exit at a defined set of nuclear pores. Though the organization of genes relative to the nuclear envelope remains largely unproven, the findings of Huang and Spector [1991] are consistent with the ‘gating’ hypothesis. In contrast, Lawrence and colleagues saw no significant evidence for their transcript ‘tracks’ making contact with the nuclear envelope [Xing and Lawrence, 1993; Xing et al., 1995]. Collagen mRNA was visualized as ‘studding’ or ‘encircling’ the nuclear envelope [Xing et al., 1995], indicative of an exit at many nuclear pores. It has been suggested that the lack of tracks visibly extending to the nuclear envelope may be due to the fact that somewhere along the transport pathway, pre-mRNA rapidly disperses in many directions [Xing and Lawrence, 1993; Xing et al., 1995]. Once again evidence of specificity in a nuclear process is indicative of a high degree of structural organization. Not all mRNAs might require such specificity in their export pathway.

The process of mRNA synthesis has many parallels with that of rRNA: sites of synthesis can be visualized, the pre-mRNA is packaged with processing machinery cotranscriptionally, and then for certain mRNAs processing within a defined structure takes place before release for export from the nucleus. The various structures visualized reflect the molecular machines active at those sites [Fig. 2.62]. Moreover the structures are dynamic with a constant vectorial flow from the sites of synthesis to the next step on the way to the cytoplasm. Components can also recycle between the different functional compartments [e.g. Pinol-Roma and Dreyfuss, 1992]. Everything happens as a nucleoprotein complex that is visually identified as a functional and morphologically discrete compartment. Does it matter that transcription and splicing/processing occur in particular domains? The advantages of compartmentalization are similar to those discussed earlier for replication. There is a concentration of the necessary regulatory, structural and enzymatic components required to transcribe or splice mRNA. The organization of the components within an architectural framework provides many more opportunities for regulation compared to a freely diffusible state. Clearly transcription and splicing can occur in dilute solutions [1 μg/ml] within an in vitro reaction tube, however the efficiency with which these events occur is much less than that achieved in vivo. Organization and channelling of macromolecules from one site of enzymatic activity to another within a specific architecture is clearly advantageous within a nucleus containing nucleoprotein at > 50 mg/ml.

The functional organization of the chromosome into discrete domains has been increasingly recognized through experiments in yeast and Drosophila that have made use of the phenomenon of position effect variegation [Schaffer et al., 1993; Section 2.5.6]. Early cytological experiments demonstrated the positioning of telomeres at the nuclear envelope in salivary gland cells of salamanders [Rabl, 1885]. The telomeres of Drosophila polytene chromosomes and those of Schizosaccharomyces pombe chromosomes in G2 phase of the cell cycle also show comparable localization of the telomeres at the nuclear periphery [Hochstrasser et al., 1986; Funabiki et al., 1993]. Advances in confocal immunofluorescence microscopy and molecular genetics have allowed the demonstration that two proteins, silent information regulators [SIR]3 and 4 are required for the perinuclear localization of Saccharomyces cerevisiae telomeres [Palladino et al., 1993]. These proteins are also required for the heritable inactivation of genes within specific chromosomal domains located at the silent mating type loci and telomeres of S. cerevisiae. Thus a connection is made between the location of a particular chromosomal territory in the nucleus and transcriptional repression per se [Section 2.5.6].

Experiments designed to examine the localization of active genes in the nucleus clearly demonstrate that these are predominantly found within the nuclear interior [Spector, 1993]. Early suggestions that active genes were preferentially located at the nuclear periphery are probably based on experimental artefact [Hutchison and Weintraub, 1985]. With respect to specific active genes, Lawrence and colleagues have suggested that some genes occupy non random positions [Lawrence et al., 1993]. For example: three active genes with very different localizations are the whole EBV genome and the neu oncogene [transcriptionally active] which are positioned within the inner 50% of the nuclear volume, whereas the dystrophin gene is at the extreme nuclear periphery. However, three inactive genes [albumin, cardiac myosin heavy chain and neurotensin] all localize in constitutive heterochromatin at the nuclear periphery or near the nucleolus [Xing et al., 1995]. UV microirradiation and in situ hybridization experiments extend the experiments examining telomeres or specific genes to suggest that individual chromosomes occupy broad, but discrete territories within the nucleus [Cremer et al., 1993; Heslop-Harrison and Bennett, 1990; van Driel et al., 1995].

The spatial relationship between chromosome territories and other subnuclear compartments has been investigated by Cremer and colleagues [Zirbel et al., 1993]. It was shown that the splicing machinery subcompartments were associated with the periphery of chromosome territories and were excluded from their interior [Zirbel et al., 1993]. Similarly, a specific gene transcript visualized as an RNA track was shown to be preferentially localized on the surface of the chromosome territory and a very limited survey of the localization of individual genes again placed them to the exterior of chromosomal territories [Zirbel et al., 1993; Cremer et al., 1993]. On the basis of the above evidence, Cremer and colleagues have postulated that the interchromosome space excluded by the chromosomal territories defines an interconnected functional compartment for transcription, splicing, maturation and transport [Fig. 2.63]. This compartment is intimately associated with actively transcribing genes localized on the surface of the territory, presumably on extended loops [Zirbel et al., 1993; Cremer et al., 1993]. This is an intuitively appealing model which potentially encompasses the observations of Lawrence and colleagues regarding the non-random distribution of genes. However, the nature of the functional interface between an active gene in a chromosome territory and the interchromatin compartment remains unclear, primarily due to our poor understanding of the higher order organization and compaction of DNA into chromosomes.

The compartmentalization of transcription, RNA processing and replication components within the nucleus lends credence to the existence of specialized functional roles for any morphologically distinct structure in which a protein of interest accumulates. Nuclear bodies, originally described at an ultrastructural level, represent such structures in search of a function.

Coiled bodies Morphologically defined as a tangle of coiled threads [Monneron and Bernhard, 1969], coiled bodies are often associated with the periphery of nucleoli. A number of nucleolar proteins and RNAs important for rRNA modification and processing are found in the coiled body including fibrillami and the U3 small nucleolar RNA [Jimenez-Garcia et al., 1994]. snRNPs including U7 snRNA also accumulate in this structure, as do specialized proteins such as p80-coilin [Frey and Matera, 1995; Bohmann et al., 1995]. Accumulation of these proteins suggests a role in RNA processing, however rRNA and mRNA have not been detected in these structures [Huang et al., 1994; Jimenez-Garcia et al., 1994]. Coiled bodies are dynamic, they disassemble at mitosis and reassemble in G1 during the cell cycle, they also increase in abundance in response to growth stimuli [Carmo-Fonseca et al., 1993; Lamond and Carmo-Fonseca, 1993].

The nuclei of amphibian oocytes contain structures known as Csnurposomes with many similarities to mammalian coiled bodies [Bauer et al., 1994; Gall et al., 1995]. These nuclear organelles are found attached to the histone gene loci of lampbrush chromosomes, and since U7 snRNA is involved in the 3′-end modification of histone pre-mRNAs, it has been suggested that this is one function of the C-snurposome and of coiled bodies [Gall et al., 1995; Frey and Matera, 1995]. Alternatively the C-snurposome or coiled bodies might represent assembly sites for the molecular machines that process various RNAs [Bohmann et al., 1995].

PML [promyelocytic leukemia] nuclear bodies Acute PML is a haemopoietic malignancy that is most often associated with a t[15; 17] chromosome translocation which results in an inframe fusion of the PML gene to that of the retinoic acid receptor α [RARα] [Warrell et al., 1993]. The PML protein itself contains a zinc binding RING finger, two cysteine rich domains and a C-terminal coiled-coil domain [Lovering et al., 1993; Reddy et al., 1992; Perez et al., 1993]. In cells from normal individuals, PML predominantly accumulates in a novel nuclear body consisting of a dense fibrillar ring surrounding a central core [Koken et al., 1994; Weis et al., 1994; Dyck et al., 1994]. PML is also found with U1 snRNA and p80-coilin in a distinct compartment or zone surrounding interchromatin granules or speckles [Visa et al., 1993b; Puvion-Dutilleul et al., 1995]. PML nuclear bodies are dynamic with respect to the cell cycle and there appears to be a correlation between their prominence and proliferative states [Koken et al., 1995; Terris et al., 1995]. Viral infections can disrupt PML bodies [for example Puvion-Dutilleul et al., 1995; Kelly et al., 1995]. For adenovirus, this disruption may be a crucial step for replication of the viral genome [Doucas et al., 1996]. Treatment of cells with interferon induces PML expression and represses viral replication, thus offering an additional contribution to antiviral activity [Guldner et al., 1992; Stadler et al., 1995].

When PML is fused to RARα in patients with acute promyelocytic leukemia the normal distribution of PML in defined nuclear bodies is disrupted and a ‘micropunctate’ pattern is observed [Weis et al., 1994; Koken et al., 1994; Dyck et al., 1994]. Any wild-type PML protein is also sequestered into this micropunctate pattern. Treatment of cells with retinoic acid facilitates the restoration of the normal nuclear body distribution of PML through a mechanism that is not understood [Dyck et al., 1994; Weis et al., 1994; Koken et al., 1994]. This correlates with the fact that patients with acute promyelocytic leukemia go into remission following treatment with retinoic acid. It has been suggested that wild-type PML functions to suppress growth, slowing down the growth rate of transformed cell lines and suppressing their tumorigenicity [Koken et al., 1995]. The molecular mechanisms by which this would be achieved remains unknown.

WT1 nuclear domains Wilm’s tumor is a childhood kidney malignancy frequently associated with congenital urogenital abnormalities, thus indicating underlying developmental deficiencies. It is a complex genetic disease, with at least three genetic loci contributing to it. So far only one gene has been isolated, the WT1 tumor suppressor, gene [Hastie, 1994]. Mice homozygous for the WT1 knock-out die. before day 15 of gestation with a clear failure to develop kidneys, gonads and a normal mesothelium [Kreidberg et al., 1993].

The product of the WT1 gene was originally thought to be a transcription factor since it contains an N-terminal proline/glutaminerich domain, frequently associated with transcriptional activators, and four C-terminal zinc-fingers, very closely related to those found in transcription factors such as Sp1, EGR1 and EGR2. WT1 has been shown to bind to a GC-rich motif in vitro and to repress transcription in transient transfection assays in promoters that contain this motif. As yet, no physiological gene target for WT1 activity has been identified.

The subnuclear localization of WT1 in a mouse mesonephric cell line, as well as in fetal kidney and testis, showed a distinct punctate pattern as well as a diffuse nucleoplasmic staining [Larsson et al., 1995; Englert et al., 1995]. Double staining clearly showed that WT1 did not occupy the nuclear transcription factor domains that were stained by Spi. Instead, by using an anti-Sm antibody that detects snRNPs, WT1 was shown to colocalize with the snRNP ‘speckles’ [Larsson et al., 1995]. WT1 was also co-immunoprecipitated with an anti-p80-coilin antibody, suggesting that WT1 is also present in coiled bodies. One recent study in transfected osteosarcoma cells showed that WT1 colocalized with only a subset of the snRNP speckles which did not stain with a monoclonal antibody against the essential non-snRNP splicing protein SC-35 [Englert et al., 1995]. This raises the possibility that the WTl-rich speckles constitute a novel nuclear subcompartment that also contains snRNPs. The distribution of WT1 in the nucleus has been shown to be a dynamic one which paralleled, to a large extent, that of snRNPs. For example, microinjection in the nucleus of antisnRNA oligonucleotides [Larsson et al., 1995] or heat shock [Charlieu et al., 1995] led to an identical rearrangement for both WT1 and snRNPs in both cases. In contrast, treatment with actinomycin D, a transcriptional inhibitor, led to a re-distribution of WT1 that overlapped that observed for p80-coilin, a coiled body hallmark. In both cases, the proteins were seen to be re-distributed around the nucleolus remnants, as opposed to the majority of snRNPs which concentrated in large foci [Larsson et al., 1995].

Alternative splicing gives rise to four WT1 isoforms dependent on the inclusion or omission of two motifs: 17 amino acids encoded by exon 5 included N-terminally to the zinc-fingers, and/or the KTS motif [for lysine-threonine-serine] included between zinc-fingers 3 and 4. The presence or absence of the KTS motif significantly affects the DNA binding properties of WT1, with the + KTS WT1 isoforms having a significantly lower affinity for binding to DNA. In addition, the + KTS isoforms were seen to distribute mostly in a speckled pattern, whereas the –KTS isoforms were distributed more diffusely [Charlieu et al., 1995]. DNA binding-defective –KTS isoforms accumulate in the speckles, thus suggesting a correlation between DNA-binding affinity and subnuclear localization [Larsson et al., 1995; Charlieu et al., 1995]. Moreover, DNA-binding –KTS isoforms are sequestered into a speckled distribution by a WT1 mutant lacking the zinc-finger domain [Englert et al., 1995]. These findings are significant, since naturally occurring dominant negative mutations that give rise to developmental abnormalities, have been mapped within the zinc-finger domain [Hastie, 1994]. These are likely to affect not only DNA binding by the mutant WT1, but also the subnuclear distribution of any wild-type –KTS protein.

This work therefore indicates the existence of a clear subcompartmentalization of WT1 isoforms relating to WT1 function. The dynamic association of WT1 with the splicing machinery, suggests a previously unappreciated role for WT1 in post-transcriptional gene regulation, as well as in transcription per se [Charlieu et al., 1995].

The experimental data discussed here illustrate the diversity of nuclear events in their structural context. There are two major landmarks in the nucleus: [1] the nuclear envelope, associated lamina and nuclear pores [Dingwall and Laskey, 1986; Gerace and Burke, 1988] which marks the outer boundary [Section 2.4.2], and [2] the chromosomes within the interior which represent the reason for the nucleus to exist through transcription and replication of DNA. These sites of chromosomal activity are non-randomly distributed with respect to the nuclear envelope. Unknown features of nuclear architecture direct the spatial arrangement of replication foci [Cox and Laskey, 1991]. A major unresolved issue at this time is whether there is an underlying structure within the nucleus that directs the spatial distribution of functional compartments [Section 2.4.2]. Chromatin is potentially mobile, moving through the replication foci during S-phase [Hughes et al., 1995]. The chromosome provides the template for replication and transcription, but is clearly not a static structure that continually maintains one particular architecture. DNA is packaged with chromatin structural proteins in a way that allows transcription and replication to occur within the functionally differentiated structures.

The molecular machines that transcribe and replicate DNA, as well as those that regulate these events are so extensive that it appears probable that concentrations of these machines together with the associated RNA and DNA, account for many of the structures that can be morphologically distinguished in the nucleus. This is particularly apparent for the nucleolus, where distinct domains are visualized representing: [1] regulatory nucleoprotein complexes controlling transcription, [2] the active transcriptional machinery itself and associated transcripts and [3] transcripts in the process of being assembled into functional ribonucleoprotein complexes. On a more local scale the same domains are visualized for RNA polymerase II transcripts [Fig. 2.62]. Ribonucleoprotein complexes clearly account for many morphologically distinct structures in the nucleus, especially in the interchromosomal domain [Cremer et al., 1993]. This is true both for a normal nucleus within a somatic cell and for the enormous nucleus of the amphibian oocyte. It is, therefore, not surprising that many attempts to characterize a nuclear matrix at a biochemical level reveal ribonucleoprotein as a major structural component [Mattern et al., 1996]. Nuclear structures such as the coiled body may in fact be sites of assembly of the ribonucleoprotein needed to process other ribonucleoprotein complexes.

One role for the chromosome in the overall organization of the nucleus that emerges from these studies is the segregation and dispersal of the DNA template within a particular territory or nuclear domain. There is a directionality in this organization, since telomeres are orientated towards the nuclear periphery and most active genes appear to have an interior location. There are suggestions that this organization might have consequences for the release of mRNA to the cytoplasm and subsequent utilization of the transcripts. This is an important topic for future investigation.

Molecular genetics defines disease genes involved in acute promyelocytic leukemia and Wilm’s tumor. Expression of these genes leads to the accumulation of proteins in nuclear bodies or domains of unknown function: the PML nuclear body and the WT1 domain. Comparable compartmentalization is visualized for other regulatory molecules such as the transcription factors: the glucocorticoid and mineralocorticoid receptor [van Steensel et al., 1996]. These concentrations of specific proteins are dynamic and might represent sites to which particular signal transduction pathways are channelled within the nucleus. Alternatively this compartmentalization might reflect roles for these proteins that are yet to be defined. For example, a potential role for WT1 in post-transcriptional gene regulation emerges from the colocalization of this protein with speckles. The clear significance of these gene products for human disease should further stimulate research on novel functions for nuclear organelles.