- © 1998 American Society of Plant Physiologists
Suppose that you have several kilometers of very fine thread and must find a way to fit it inside a sphere the size of a golf ball. The thread cannot be packed too tightly because all points along its length must be accessible. Your packaging plan must also accommodate the need to unravel and repack the thread repeatedly without it becoming hopelessly tangled. This hypothetical engineering problem should elicit an appreciation for the packaging of chromosomal DNA into a eukaryotic nucleus, a task which requires that DNA be compacted in length as much as 10- to 50,000-fold while maintaining a form that accommodates transcription and replication.
How DNA is packaged into chromatin and chromosomes is not entirely clear, especially at the highest levels of organization. However, certain aspects are fairly well understood. The first step is the assembly of DNA into nucleosomal arrays. When viewed under the electron microscope, these arrays have the appearance of beads 10 nm in diameter distributed along a 2-nm DNA string (Kornberg, 1974; Olins and Olins, 1974). Each bead is a nucleosome core particle that includes ~146 bp of DNA wrapped almost twice around a core histone octamer. Nucleosome assembly compacts the DNA approximately six-fold in its linear dimension. Histone H1 or a related “linker” histone binds the 40 to 70 bp of linker DNA that separates adjacent core particles and helps to organize the beads-on-a-string into fibers ~30 nm in diameter (Finch and Klug, 1976; Thoma et al., 1979). Viewed under the electron microscope, these 30-nm fibers appear to be helical structures with approximately six nucleosomes per turn, an arrangement that compacts the DNA ~40-fold in its linear dimension.
Our understanding of levels of organization beyond the 30-nm filaments is more sketchy, but current models derive from the stunning electron micrographs of Laemmli and colleagues, which were published in the late 1970s (Marsden and Laemmli, 1979; Paulson and Laemmli, 1977). These initial images of HeLa cell metaphase chromosomes stripped of histones show DNA spooling out in 30- to 90-kb loops from a dark proteinaceous “scaffold” that still retains the X shape of the paired sister chromatids (Paulson and Laemmli, 1977). The loops appear to emanate from and return to the same point, suggesting that the DNA is tethered to the scaffold at the base of the loops. Preparations in which histones are not removed from the DNA reveal loops of chromatin made up of 180 to 300 nucleosomes coiled in 30-nm fibers (Marsden and Laemmli, 1979). Organized in this way, each loop would account for ~700-fold packing of the DNA relative to the long axis of the chromosome. In cross-section, the loops appear to radiate from the scaffold as if tracing the outline of the petals on a daisy flower.
These observations led to the formal “radial loop model” for chromosome organization in which adjacent loop attachment sites are arranged in a helical spiral along the long axis of the metaphase scaffold (Marsden and Laemmli, 1979). Organizing 15 to 18 such loops per turn along the chromatid would account for ~1.2 million bp of DNA (Nelson et al., 1986). This arrangement predicts the stacking of loops into a cylinder of chromatin ~800 to 1000 nm in thickness, which is in good agreement with the diameter of the metaphase chromosome (Marsden and Laemmli, 1979; Nelson et al., 1986). This model also accounts for the dimensions of metaphase chromosomes, which are ~10,000-fold shorter and 400- to 500-fold thicker than the double stranded DNA helices contained within them. Twisting the cylinder into a superhelix would further compress it in the linear dimension and account for the corkscrew appearance of metaphase chromosomes viewed at high magnification.
We know less about the organization of interphase chromosomes. This is primarily because interphase chromatin tends to be dispersed throughout the nucleus and is therefore more difficult to see. Exceptions include the giant “lampbrush” chromosomes of amphibian oocytes. These chromosomes consist of highly elongated sister chromatids, which remain attached to each other by multiple chiasmata in a prolonged meiotic interphase that can last for months (Callan, 1982). Loops emanate from the central axis of each chromatid and are visible under the light microscope due to the association of large amounts of newly transcribed RNA, protein, and ribonucleoprotein processing complexes.
The similarity between the loop-and-scaffold structures of lampbrush and metaphase chromosomes is striking, suggesting the generality of this organization for both transcriptionally active and inactive chromosomes. However, in interphase cells of species that do not have lampbrush chromosomes, evidence for a loop-and-scaffold-like arrangement is less direct. In such cells, removal of histones with high salt or ionic detergents yields a proteinaceous “nuclear matrix,” which includes part of the nuclear envelope, a weblike nucleoskeleton, and a remnant of the nucleolus (Nelson et al., 1986). DNA is attached to the nuclear envelope and to the nucleoskeleton at intervals that approximate the loop sizes of metaphase chromosomes, suggesting that the metaphase scaffold may be a rearranged and compacted form of the interphase nuclear matrix (Razin and Gromova, 1995).
Whether or not there is a direct relationship between the structural organization of chromosomal DNA into loops and the functional subdivision of the genome into regulatory domains is an important and largely open question. Considerable effort has been focused on mapping scaffold or matrix attachment regions (SARs or MARs) relative to known genes (e.g., Mirkovitch et al., 1984; Nelson et al., 1986; Roberge and Gasser, 1992; Razin and Gromova, 1995). One method involves stripping the histones from lysed nuclei before cutting the matrix-associated DNA with restriction endonucleases whose cleavage sites are known for the gene in question. Some DNA fragments are found in the supernatant when the insoluble matrix fraction is pelleted, whereas other fragments are retained in the pellet. The latter are interpreted to be those that include sequences bound to the nuclear matrix. By judiciously choosing the enzymes used for such studies, SAR/MAR attachment sites can be mapped.
Another method for mapping SARs/MARs is to generate matrix preparations that are free of all contaminating nucleic acid, add exogenous DNA fragments from the cloned gene of interest, and identify those sequences which, because they bind to the matrix, can be recovered by centrifugation. In some cases, these different assays identify the same DNA fragments, but in other cases they do not. Different results are also obtained for matrices prepared using high-salt extraction (e.g., 2M NaCl), ionic detergents (e.g., lithium diiodosalicylate) coupled with heat treatment, or treatment of agarose-encapsulated cells with non-ionic detergents under isotonic conditions (for a review, see Jackson et al., 1992).
Despite their differences, these studies all indicate that attachment sites tend to be AT rich and that they are preferred binding sites for topoisomerase II (Topo II), which is one of the major scaffold proteins (Swedlow et al., 1993). However, it remains controversial whether SARs/MARs are sites at which the DNA itself is fixed directly to the matrix or whether the DNA is instead associated with transcription or replication complexes that are bound to the matrix. Cook and colleagues have argued for the latter hypothesis based on their findings that ~70 to 80% of all newly transcribed RNA or newly replicated DNA is associated with the matrix (Jackson et al., 1992).
Other evidence to suggest that genes are functionally compartmentalized into regulatory domains demarcated by “boundary” or “insulator elements” has accumulated (Eissenberg and Elgin, 1991; Schedl and Grosveld, 1995; Felsenfeld, 1996), and the need for such elements is clear if one considers how it is that active and inactive genes can coexist side by side. For example, transcriptional enhancers can function in either orientation and from either side of a gene promoter, sometimes acting over tens of kb (Blackwood and Kadonaga, 1998). Given their position and orientation independence, why do enhancers activate only specific genes rather than all of the genes in their vicinity? Similarly, how are chromatin modifications that help keep inactive genes silent restricted from spreading to nearby active genes?
Initial evidence that MARs/SARs might fulfill the requirements of boundary elements came from studies using matrix-associated elements that mapped to the borders of a 20-kb DNase hypersensitive region surrounding and including the chicken lysozyme gene (Stief et al., 1989). When placed on either side of a reporter gene integrated into the genome of cultured chicken cells, these boundary sequences protect the transgene from position effects and make expression proportional to gene copy number. Without the boundary sequences, transgene expression is variable. Interestingly, when they are placed between an enhancer and an adjacent promoter, these elements insulate the promoter from enhancement, suggesting that the boundary elements delimit domains in which enhancers can function (Stief et al., 1989). Similar results in transgenic plants have shown that yeast or plant MAR/SAR elements allow consistent expression levels from reporter transgenes (Allen et al., 1993, 1996).
However, MARs/SARs do not always coincide with boundary/insulator elements. For example, Kellum and Schedl (1991) characterized “specialized chromatin sequences” (scs elements) flanking a Drosophila locus which includes two divergently transcribed HSP70 genes (Kellum and Schedl, 1991). These scs elements map to the boundaries of the ~15-kb “puff” (a region of chromosome decondensation) apparent at this locus on polytene salivary gland chromosomes following heat shock. Like the elements flanking the chicken lysozyme gene, the scs elements are able to insulate a reporter transgene from chromosomal position effects. Similarly, an scs element placed between an enhancer and an adjacent promoter blocks enhancer function. However, the scs elements do not appear to colocalize with MARs/SARs. Instead, a strong MAR/SAR in the region maps to the interval between the promoters of the two HSP70 genes. Interestingly, this MAR/SAR does not block enhancer function when it is placed between an enhancer and a promoter, unlike the scs elements (for reviews, see Eissenberg and Elgin, 1991; Schedl and Grosveld, 1995).
Other domain boundary elements have been shown to play a role in the regulation of complex loci containing multiple genes. These loci include the bithorax homeotic gene complex in Drosophila and the vertebrate β-globin gene locus. In the case of the bithorax locus, multiple boundary elements play a role in restricting the domains responsible for the orchestrated expression or silencing of individual genes within the locus (Hagstrom et al., 1996). In the case of the β-globin locus, five developmental stage–specific globin genes are all dependent on a common locus control region (LCR) located ~75 kb from the most distal gene (Felsenfeld, 1993). The LCR appears to help create a decondensed chromatin structure that establishes the transcriptional competence of the gene cluster early in development. Gene-specific enhancers and promoter elements then appear to dictate which globin gene will be expressed at a given point in development.
Like scs elements and MARs/SARs, and indicative of a distinct chromatin structure, the LCR is characterized by regions of DNase hypersensitivity. Importantly, sequences within the LCR have been shown to have the ability to protect transgenes from position effects in transformed vertebrate cells and Drosophila (Chung et al., 1993) and to block enhancer function when interposed between an enhancer and an adjacent promoter. These findings suggest that many boundary elements work in the same way in most, if not all, higher organisms (Schedl and Grosveld, 1995).
Finally, two good examples of silenced domains are provided by the yeast silent mating-type loci, which contain repressed genes that are otherwise fully functional (Laurenson and Rine, 1992). Mutations in any of a number of genes encoding chromatin proteins are sufficient to derepress the silent mating-type genes. Many of these same mutations also derepress genes silenced due to their proximity to telomeres (Aparicio et al., 1991). Importantly, reporter genes integrated near silent mating-type or sub-telomeric loci are also silenced, indicating that silencing is domain specific rather than gene specific.
Despite the flurry of investigations described above, relatively little is known about the structural or functional organization of chromosomal DNA in plants. On pages 1349–1359 of this issue, Paul and Ferl take one of the first steps in defining the chromosomal environment encompassing specific plant genes. By focusing on DNA rather than MARs/SARs, a conceptual approach pioneered by Razin and colleagues (reviewed in Razin and Gromova, 1995), Paul and Ferl show that limited digestion of nuclei with DNase I preferentially releases large DNA fragments whose average size varies between ~25 and 45 kb in maize, Arabidopsis, and lily. When subjected to Southern blotting, a small but detectable portion of the hybridization signal obtained with probes for genes encoding alcohol dehydrogenase (Adh1) and general regulatory factor (GRF1) corresponds to discrete fragments whose size is different for the two genes. Because random nuclease cleavage would be expected to generate a smear on the gel blots, these observations suggest a nonrandom chromatin organization for Adh1 and GRF1. Indeed, the authors interpret their data to indicate that the two genes are organized on DNA loops that are cleaved from the scaffold/matrix at sites at which the DNA is thought to be more accessible to DNase I.
Support for Paul and Ferl's hypothesis comes from their experiments showing that similar sized fragments are also released by chemicals that facilitate DNA cleavage by Topo II without allowing DNA religation by the same enzyme. These data are significant because Topo II is thought to be enriched at the base of the DNA loops, where it may function to relieve the torsional strains inherent to replication and transcription and/or DNA compaction during mitosis.
Several predictions can be derived from the hypothesis advanced by Paul and Ferl. Most importantly, Topo II and DNase I cleavages sites should colocalize and should map at or near MARs/SARs, as Razin and colleagues have shown in their studies of the Drosophila X chromosome (Iarovaia et al., 1996). Hopefully, subsequent publications from Ferl's group will indicate whether or not this prediction can be supported with experimental data. Interested readers will also look forward to finding out whether there is any relationship between the DNase I and Topo II cleavage sites that generate the putative loops and the DNase I hypersensitive sites and MARs that have been mapped close to the Adh1 gene (Paul et al., 1987; Paul and Ferl, 1993).
The relationship between gene function and chromosome organization is one of several areas where plants, aided by robust genome projects, should provide particularly useful model systems. The order of genes along a chromosome is unlikely to be random or a historical accident but is, instead, likely to be shaped by natural selection. It would be efficient to group genes best suited to similar chromatin environments, such as genes that tend to be highly methylated (e.g., the rRNA genes studied in my laboratory), genes that replicate early or late, and genes that can utilize common LCRs or other regulatory regions. Clearly, understanding how chromosome structure and gene functions are interrelated is a challenging subfield of functional genomics. Paul and Ferl's studies should serve as a useful starting point in an area that is poised for further investigation and that is certain to yield important new discoveries in the near future.