DNA bendability and nucleosome positioning in transcriptional regulation

The placement of nucleosomes along genomic DNA is determined by signals that can be specific or degenerate at the level of sequence; the latter signals are harder to find using conventional methods. In recent years, the development of sophisticated machine learning techniques that can extract subtle phased signals has improved our ability to distinguish between various classes of nucleosome-positioning sequences. Our knowledge of the structural mechanics of free DNA also has reached the point where it can be fruitfully incorporated into predictive models. More importantly, the accumulation of high-resolution structures with proteins bound to DNA, and those of nucleosomes in particular, has provided important clues about the role of DNA bending and flexibility in nucleosome positioning. Introduction Eukaryotic DNA is compacted and organized in nucleosome arrays that make up chromosomes (1). The nucleosome core, a basic unit of chromatin, contains two copies each of the core histones H2A, H2B, H3 and H4, and about 146 base pairs of DNA wrapped around the protein octamer (2). Further compaction of DNA by the linker histone H1 is achieved in higher-order structures assembled from repetitive nucleosome cores and linker DNA. Although nucleosomes show no clear binding preference for particular DNA sequences, they are not randomly distributed on DNA. Translational and rotational positioning of nucleosomes along the DNA molecule are in part determined by signal sequences that are often degenerate, and sometimes have periodicity corresponding to the helical repeat of DNA (3-9). Structural properties of DNA, such as intrinsic bending and flexibility, play important roles in DNA recognition by sequence-specific DNA-binding proteins (10-13). A similar role has been proposed for these DNA features in nucleosome positioning (14, 15), and verified using a variety of experimental and theoretical approaches (5, 6, 8, 16-22). There is a statistical preference for rotational positioning of DNA around the histone octamer such that AAA/TTT and AAT/ATT trinucleotides have the DNA minor groove facing the octamer, while the minor groove of GGC/GCC and AGC/GCT faces away from protein (5, 16). These observations were confirmed by designing artificial nucleosome positioning sequences that form nucleosomes significantly better than bulk nucleosomal DNA (6). In addition to static bending properties of DNA, the variation in sequence-dependent bendability (18) can also impart strong rotational and translational orientation to nucleosomal DNA (8, 9). Comparison of nucleosomal DNA with B-DNA oligonucleotides The recent flurry of nucleosome core structures from Richmond and Luger labs (2, 23-25) has revealed a wealth of structural details about protein-DNA interactions (Figure 1). These efforts culminated with a 1.9 Χ structure that provided a high-resolution view of the DNA conformation as well (26, 27). The resolution of this structure is comparable to highresolution structures of oligonucleotides, providing a solid basis for statistical analyses aimed at extracting general principles of conformational variability in nucleosomal DNA. We therefore set out to determine to what degree the nucleosomal DNA angular parameters and deformability compare with those of B-DNA oligonucleotides (Table I), and whether these differences can be rationalized in terms of nucleosome positioning. Many sequencedependent characteristics of DNA are shared between the two data sets, clearly showing that intrinsic conformational properties of the double helix are utilized by the histones much the same way as in specific protein-DNA complexes (13). On the other hand, certain features of nucleosomal DNA are not found as major trends in B-DNA oligonucleotides. Sequence-dependent deformability of DNA is reflected in the dispersion of base-pair parameters. The inspection of Table I reveals that nucleosomal DNA shows greater flexibility than B-DNA in terms of roll and tilt angles, and only slightly lower flexibility for twist (standard deviations from mean values are shown as subscripts in Table I). Higher flexibility of DNA angular parameters was also observed when comparing specific protein-DNA complexes to B-DNA (13). Although reduced variability of B-DNA is likely caused in part by crystal packing effects, it is clear that the entropy of protein-DNA complex formation enables, and possibly requires, larger conformational flexibility of DNA. DNA duplex bends towards the minor or major grooves (roll) much more easily than in a direction along the longer base-pair axis (tilt) (14, 28); this bending anisotropy holds true in protein-DNA complexes as well (13). In nucleosomal DNA, there is even greater preference for roll over tilt, in terms of both mean values and the dispersion (Table I). The roll in nucleosomal DNA contributes to smooth bending into either groove and to kinking into the minor groove (12, 27). The latter feature is never seen in B-DNA oligonucleotides and is rarely seen in protein-DNA complexes (28). Kinking into minor groove is observed almost exclusively at CA/TG steps (roll angles are in the range –12 to –21), in marked contrast with protein-DNA complexes where CA/TG step has mostly positive roll (28). However, CA/TG steps have a preference for low-roll/high-twist in a YCAR context (29), and most of nucleosomal CA/TG steps with negative roll values are indeed preceded by a pyrimidine and followed by a purine. Pyrimidine-purine (YR) dimers are the most easily deformed steps both in B-DNA crystals (15, 30, 31) and protein-DNA complexes (13, 28). YR dimers in nucleosomal DNA are even more flexible as judged by standard deviations of roll angles (27), living up to their billing as “flexible hinges” that fit the DNA duplex to the protein surface (13). Though both TA and CA/TG steps have the highest flexibility of all 10 unique dimers in terms of twist and roll values (Table I), most of negative roll angles occur at CA/TG steps, while TA steps have mostly positive roll angles (32). Greater flexibility of nucleosomal DNA is achieved by concerted changes in roll and twist angles. A strong anti-correlation between twist and roll has been observed in earlier analyses (13, 30, 33), and is also present in our subset of B-DNA oligomers shown in Table I. Overall, all dimers in nucleosomal DNA have excellent correlation between twist and roll values (Figure 2), as is the case for specific protein-DNA complexes (12, 13), which supports the notion that sequence-specific constraints of the sugar phosphate backbone contribute primarily to the conformational variability of protein-bound DNA. In B-DNA oligomers, where crystal packing forces induce subtle but noticeable variations in DNA deformability (34), the correlation between twist and roll angles is less prominent (Table I). It is important to note that YR steps have the most significant twist-roll correlation, consistent with their highest flexibility. In particular, roll angles in CA/TG steps cover the span of more than 40 (from -21 to +23). This remarkable variability enables them to be positioned either on the “inside” or on the “outside” of the histone octamer. DNA bending and flexibility are utilized for nucleosome positioning The analysis above shows that nucleosome positioning, for the most part, takes advantage of the intrinsic structural mechanics of the double helix. When DNA is bound by specific transcription factors, most of the free energy contribution comes from specific interactions between protein side-chains and DNA bases. In this case the protein has to accommodate structural properties of only a limited number of nucleotides, and YR dimers have been selected over the course of evolution as most frequent sequence elements to “fit” DNA around the protein because of their unique conformational properties (13, 15, 28). When wrapping DNA around its surface, a sequence-specific DNA-binding protein needs to solve a “local” optimization problem, as its binding site will typically be short. This is achieved efficiently by utilizing only part of the conformational space where YR dimers have positive roll angles (13, 28). In contrast, core histones have to wrap tightly a longer piece of DNA regardless of its sequence. This is a “global” optimization problem and can be solved only by exploiting a wider range of conformational variability of DNA. Part of the solution is similar to sequence-specific DNA-binding proteins in a sense that YR dimers are most flexible and roll is preferred over tilt. However, nucleosomal DNA has certain characteristics that are only partially employed in B-DNA oligomers and specific protein-DNA complexes: (1) higher overall flexibility of all dinucleotides; (2) extremely tight coupling of twist and roll angles; (3) negative roll angles in CA/TG steps. Integrating the results from the present analysis with earlier theoretical and experimental data, we propose that nucleosomes are positioned by a combination of static and dynamic signals encoded in DNA. Statistical analysis of nucleosomal DNA cleavage (16) showed that GGC/GCC elements strongly prefer to have the minor groove facing “outside” (positive roll angles). The same trend is seen for the GG/CC step in our analysis, as well as AC/GT, AA/TT and TA steps. On the other hand, the GC step, and to a lesser degree the AT step, show stronger preference for negative roll values. We propose that sequence elements with stronger bending preferences set the initial frame for nucleosome positioning by assuming their preferred conformation. When these sequences are positioned in such a way that satisfies most of their rotational preferences, the rest of DNA is “molded” around the histone octamer by exploiting the conformational variability of DNA. The CA/TG step is most useful in this regard as it can conform both to “inside” and “outside” positions; other DNA sequences are also capable of adopting alternative conformations depending on the context (35). This concept implies that strong nucleosome positioning can be achieved by sequences with properly phased rotational signals, by sequences that contain many flexible elements, and by those that contain favorable combinations of the two. Indeed, all three types of nucleosome positioning sequences have been observed experimentally (6, 19, 21). Preference of nucleosomal positioning in exon and intron DNA Phased structural features can be “hidden signals” for nucleosomal positioning, which are not apparent from looking at the sequence. Using machine learning methods, we have found such a pattern in human DNA which codes for proteins (8). There is a periodicity of about 10 bp for human exon DNA, whilst intron DNA contains a somewhat weaker periodicity signal. However, it is unlikely that the periodicity in the exon DNA is coming from alpha-helix encoding sequences; the functional saturation of the genetic code precludes its involvement in nucleosomal positioning (8, 36). Figure 3 displays a wheel shaped hidden Markov model architecture for human exon DNA (in this case of length 10 nucleotides), where sequences can enter the wheel at any point. The thickness of the arrows from “outside” represents the probability of starting from the corresponding state. After training the emission parameters in the wheel model did show a periodic pattern [^T][AT]G in a clearly recognizable form in states 8, 9 and 10. This wheel with a model of ten nucleotides provided the best hit from several different numbers of states. Furthermore, using the wheel model to estimate the average negative log-likelihood per nucleotide, values specific for various types of exons, introns and intergenic regions were also computed. The ranking of these also strongly indicate that the above described periodic pattern is strongest in exons. The period in the alignments (average distance between state 9 nucleotides) is in the order of 10.1-10.2 bp (8). This value is the periodicity of DNA wrapped around nucleosomes, as discussed above. By looking for periodicity signals in prokaryotic genomes, at the DNA structural level, we found evidence for horizontal DNA transfer from an Archaea to the bacterium Thermotoga maritima (37). Is it possible that such signals exist in DNA that is more compacted, than in regions of highly expressed genes? One possibility is that there might be a difference between DNA which contains protein-coding regions vs. the rest of the DNA. A good test of this is chromosome 1 from C. elegans (~16 Mbp), which has a coding density of about 25% that is, about 25% (~4 Mbp) of the chromosome is transcribed into mRNA. Furthermore, this “coding” region is divided into two almost equal fractions of intron and exon DNA (~2Mbp each). Figure 3 shows the periodicity plot for the whole chromosome, as well as for the exon and intron containing DNA. It is quite clear that, in contrast to the results for human DNA, the dominant contribution to the periodicity is from the DNA containing introns, with the exon DNA having a less strong periodicity. For comparison, the periodicity plot for the 4.6 Mbp E. coli K-12 chromosome, which mostly contains coding DNA, is also shown in Figure 3. Thus, it seems likely, at least for C. elegans chromosome I, that the noncoding DNA is more likely to be wrapped tightly in nucleosome complexes, whilst the coding DNA might have a greater chance to exist in a more open conformation. The strong peak of around 10.2 nucleotides in C. elegans has been previously found using several different methods, and has been localized to introns and intergenic regions (38, 39). Repeats of various sizes have been found in other chromosomes (40). It is likely that these periodic repeats are reflections of global chromatin properties. Nucleosome positioning and transcriptional regulation The idea has been around for more than 20 years – phased nucleosomes can result in more compact chromatin structures, and these regions are less available for the transcriptional machinery (41). The size of an RNA polymerase molecule is about the same size as a nucleosome octamer with DNA wrapped around it, as shown in Figure 1. Thus if these nucleosomes are tightly bound together, in some sort of higher order structure, then there is little chance that the RNA polymerase enzyme can get access to the promoter region, and hence transcription will be repressed. Although this idea is simple, in actuality there are complications. The degree of compaction of the chromatin depends on many things, including modification of the histone protein tails (42, 43) as well as DNA methylation (44), and there are many pathways which can regulate these modifications (45, 46). In this chapter, we are focusing on the structure of DNA, and whether there are some general properties of the double helix which can affect the ability of certain sequences to condense. Based on the sequences of DNA wrapped around trimmed nucleosomal cores, a trinucleotide model was developed for the preference of certain trinucleotides to be in phase with the helical repeat. Depending on where these trinucleotides are located, they will have either the major or minor groove facing away from the histone octamer (16). Trinucleotides were given a number based on the frequency of occurrence; those favoring the major groove were assigned a negative value, whilst if the minor groove faced away from the nucleosome, they were given a positive number. By taking the absolute value of the numbers, one has a measure of the relative propensity of a sequence to be positioned in nucleosomes, which we call “position preference” (9). In human promoters, we have found that several trinucleotides known to have high propensity for major groove compression occur much more frequently in the regions downstream of the transcriptional start point, whilst the upstream regions contain more low-bendability triplets. Within the region downstream of the start point, we find a periodic pattern in sequence and bendability, which is in phase with the DNA helical pitch. The periodic bendability profile shows bending peaks roughly at every 10 bp with stronger bending at 20 bp intervals. These observations suggest that DNA in the region downstream of the transcriptional start point is able to wrap around protein in a manner reminiscent of DNA in a nucleosome. This notion was further supported by the finding that the periodic bendability is caused mainly by the complementary triplet pairs CAG/CTG and GGC/GCC, which previously have been found to correlate with nucleosome positioning. When these values are calculated for individual genes in the E. coli K-12 genome, there is a correlation between low position preference values and highly expressed genes (47). This makes sense from a structural point of view, in that regions that are not preferentially localized in nucleosomes would tend to exist in more open conformations, and hence be more accessible to the RNA polymerase. This works even though in E. coli there are no histones or nucleosomes; however, the E. coli chromosome is compacted roughly 7000 fold, so there must still be a need for some sort of chromatin structure, and the physical chemical properties are likely to be the same. The E. coli chromosome contains clusters of highly expressed genes, localized to certain regions of the chromosome (48). Similarly, based on an extensive series of gene expression experiments in human cells, it has been shown that highly expressed human genes cluster together in distinct regions of the chromosomes (49, 50). It is possible to predict such regions throughout the whole genome using methods such as the nucleosomal position preference. As an example, Figure 5a shows a plot of the nucleosomal position preference along chromosome I of the yeast Schizosaccharomyces pombe (51). A close-up (Figure 5b) shows the marked region with a low position preference (dark green in lane C) corresponds to the SPAC23A1.07 gene, which is a zinc finger protein that can be highly expressed in S. pombe. Other regions, such as highly expressed rRNA operons, consistently exhibit position preference values significantly lower than the chromosomal average in both eukaryotic and prokaryotic genomes (i.e. they are more likely to exclude chromosomes). In general, although different genes are expressed under different conditions, utilization of DNA structural properties can be helpful in finding regions along a chromosome that are potentially highly expressed. Conclusions Nucleosome positioning is governed by various sequence signals, including the differences between coding and non-coding DNA sequences imposed by evolutionary constraints. Expanding the alphabet of position signals beyond simple sequence, for example by using the extended set of DNA structure parameters in connection with machine learningmethods, will further improve our ability to predict the role of nucleosome placement intranscriptional regulation. References1.van Holde KE. Chromatin. New York: Springer-Verlag, 19892.Luger K, Mader AW, Richmond RK et al. Crystal structure of the nucleosome coreparticle at 2.8 a resolution. Nature 1997; 389:251-260.3.Trifonov EN, Sussman JL. The pitch of chromatin DNA is reflected in its nucleotidesequence. Proc Natl Acad Sci U S A 1980; 77:3816-3820.4.Zhurkin VB. Sequence-dependent bending of DNA and phasing of nucleosomes. JBiomol Struct Dyn 1985; 2:785-804.5.Drew HR, Travers AA. DNA bending and its relation to nucleosome positioning. JMol Biol 1985; 186:773-790.6.Shrader TE, Crothers DM. Artificial nucleosome positioning sequences. Proc Natl Acad Sci U S A 1989; 86:7418-7422.7.Ioshikhes I, Bolshoy A, Trifonov EN. Preferred positions of aa and tt dinucleotides inaligned nucleosomal DNA sequences. J Biomol Struct Dyn 1992; 9:1111-1117.8.Baldi P, Brunak S, Chauvin Y et al. Naturally occurring nucleosome positioning signals in human exons and introns. J Mol Biol 1996; 263:503-510.9.Pedersen AG, Baldi P, Chauvin Y et al. DNA structure in human rna polymerase iipromoters. J Mol Biol 1998; 281:663-673.10.Harrington RE, Winicov I. New concepts in protein-DNA recognition: Sequencedirected DNA bending and flexibility. Prog Nucleic Acid Res Mol Biol 1994; 47:195-270.11.Werner MH, Gronenborn AM, Clore GM. Intercalation, DNA kinking, and thecontrol of transcription. Science 1996; 271:778-784. 12.Dickerson RE, Chiu TK. Helix bending as a factor in protein/DNA recognition.Biopolymers 1998; 44:361-403.13.Olson WK, Gorin AA, Lu XJ et al. DNA sequence-dependent deformability deducedfrom protein-DNA crystal complexes. Proc Natl Acad Sci U S A 1998; 95:1116311168.14.Zhurkin VB, Lysov YP, Ivanov VI. Anisotropic flexibility of DNA and thenucleosomal structure. Nucleic Acids Res 1979; 6:1081-1096.15.Ulyanov NB, Zhurkin VB. Sequence-dependent anisotropic flexibility of b-DNA. A conformational study. J Biomol Struct Dyn 1984; 2:361-385.16.Satchwell SC, Drew HR, Travers AA. Sequence periodicities in chicken nucleosomecore DNA. J Mol Biol 1986; 191:659-675. 17.Travers AA, Klug A. The bending of DNA in nucleosomes and its wider implications.Philos Trans R Soc Lond B Biol Sci 1987; 317:537-561.18.Brukner I, Sanchez R, Suck D et al. Sequence-dependent bending propensity of DNAas revealed by dnase i: Parameters for trinucleotides. EMBO J 1995; 14:1812-1818.19.Widlund HR, Cao H, Simonsson S et al. Identification and characterization ofgenomic nucleosome-positioning sequences. J Mol Biol 1997; 267:807-817.20.Cao H, Widlund HR, Simonsson T et al. Tgga repeats impair nucleosome formation. JMol Biol 1998; 281:253-260.21.Widlund HR, Kuduvalli PN, Bengtsson M et al. Nucleosome structural features andintrinsic properties of the tataaacgcc repeat sequence. J Biol Chem 1999; 274:31847-31852. 22.Roychoudhury M, Sitlani A, Lapham J et al. Global structure and mechanicalproperties of a 10-bp nucleosome positioning motif. Proc Natl Acad Sci U S A 2000;97:13608-13613.23.Suto RK, Edayathumangalam RS, White CL et al. Crystal structures of nucleosome core particles in complex with minor groove DNA-binding ligands. J Mol Biol 2003;326:371-380.24.White CL, Suto RK, Luger K. Structure of the yeast nucleosome core particle revealsfundamental changes in internucleosome interactions. EMBO J 2001; 20:5207-5218. 25.Suto RK, Clarkson MJ, Tremethick DJ et al. Crystal structure of a nucleosome coreparticle containing the variant histone h2a.Z. Nat Struct Biol 2000; 7:1121-1124.26.Davey CA, Sargent DF, Luger K et al. Solvent mediated interactions in the structureof the nucleosome core particle at 1.9 a resolution. J Mol Biol 2002; 319:1097-1113. 27.Richmond TJ, Davey CA. The structure of DNA in the nucleosome core. Nature2003; 423:145-150.28.Dickerson RE. DNA bending: The prevalence of kinkiness and the virtues ofnormality. Nucleic Acids Res 1998; 26:1906-1926. 29.Yanagi K, Prive GG, Dickerson RE. Analysis of local helix geometry in three b-DNAdecamers and eight dodecamers. J Mol Biol 1991; 217:201-214.30.El Hassan MA, Calladine CR. Two distinct modes of protein-induced bending inDNA. J Mol Biol 1998; 282:331-343. 31.Dickerson RE. Helix structure and molecular recognition by b-DNA. In: Neidle S,editor. Oxford handbook of nucleic acid structure. Oxford: Oxford Sci.; 1999. p. 145-197. 32.Goodsell DS, Kaczor-Grzeskowiak M, Dickerson RE. The crystal structure of c-c-a-t-t-a-a-t-g-g. Implications for bending of b-DNA at t-a steps. J Mol Biol 1994; 239:79-96.33.Gorin AA, Zhurkin VB, Olson WK. B-DNA twisting correlates with base-pairmorphology. J Mol Biol 1995; 247:34-48.34.Dickerson RE, Goodsell DS, Neidle S. "...The tyranny of the lattice..." Proc NatlAcad Sci U S A 1994; 91:3579-3583.35.Dlakic M, Harrington RE. The effects of sequence context on DNA curvature. ProcNatl Acad Sci U S A 1996; 93:3847-3852.36.Levitsky VG, Podkolodnaya OA, Kolchanov NA et al. Nucleosome formationpotential of exons, introns, and alu repeats. Bioinformatics 2001; 17:1062-1064. 37.Worning P, Jensen LJ, Nelson KE et al. Structural analysis of DNA sequence:Evidence for lateral gene transfer in thermotoga maritima. Nucleic Acids Res 2000;28:706-709.38.Widom J. Short-range order in two eukaryotic genomes: Relation to chromosome structure. J Mol Biol 1996; 259:579-588.39.Frontali C, Pizzi E. Similarity in oligonucleotide usage in introns and intergenicregions contributes to long-range correlation in the caenorhabditis elegans genome.Gene 1999; 232:87-95. 40.Fukushima A, Ikemura T, Kinouchi M et al. Periodicity in prokaryotic and eukaryoticgenomes identified by power spectrum analysis. Gene 2002; 300:203-211.41.Wittig S, Wittig B. Function of a trna gene promoter depends on nucleosome position.Nature 1982; 297:31-38. 42.Turner BM. Cellular memory and the histone code. Cell 2002; 111:285-291.43.de Ruijter AJ, van Gennip AH, Caron HN et al. Histone deacetylases (hdacs):Characterization of the classical hdac family. Biochem J 2003; 370:737-749.44.Cunliffe VT. Memory by modification: The influence of chromatin structure on gene expression during vertebrate development. Gene 2003; 305:141-150.45.Geiman TM, Robertson KD. Chromatin remodeling, histone modifications, and DNAmethylation-how does it all fit together? J Cell Biochem 2002; 87:117-125.46.Formosa T. Changing the DNA landscape: Putting a spn on chromatin. Curr Top Microbiol Immunol 2003; 274:171-201.47.Pedersen AG, Jensendagger LJ, Brunak S et al. A DNA structural atlas for escherichiacoli. J Mol Biol 2000; 299:907-930. 48.Ussery D, Schou Larsen T, Trevor Wilkes K et al. Genome organisation andchromatin structure in escherichia coli. Biochimie 2001; 83:201-212.49.Caron H, van Schaik B, van der Mee M et al. The human transcriptome map:Clustering of highly expressed genes in chromosomal domains. Science 2001;291:1289-1292.50.Versteeg R, Van Schaik BD, Van Batenburg MF et al. The human transcriptome mapreveals extremes in gene density, intron length, gc content, and repeat pattern fordomains of highly and weakly expressed genes. Genome Res 2003.51.Wood V, Gwilliam R, Rajandream MA et al. The genome sequence ofschizosaccharomyces pombe. Nature 2002; 415:871-880.52.Berman HM, Olson WK, Beveridge DL et al. The nucleic acid database. A comprehensive relational database of three-dimensional structures of nucleic acids.Biophys J 1992; 63:751-759.53.Lavery R, Sklenar H. The definition of generalized helicoidal parameters and of axiscurvature for irregular nucleic acids. J Biomol Struct Dyn 1988; 6:63-91. 54.El Hassan MA, Calladine CR. The assessment of the geometry of dinucleotide stepsin double-helical DNA; a new local calculation scheme. J Mol Biol 1995; 251:648-664.55.Bhat TN, Bourne P, Feng Z et al. The pdb data uniformity project. Nucleic Acids Res