High-resolution biophysical analysis of the dynamics of nucleosome formation.

Gene expression is essentially controlled through the spatial and temporal distribution of nucleosomes on the genome. The presence of a nucleosome generally has an inhibitory effect on DNA binding proteins; indeed the binding of TATA binding protein and the whole pol II transcription machinery requires the absence of nucleosomes. There is much debate concerning the differential contribution of various factors such as sequence, remodellers and transcription factors to nucleosome positioning1,2 although in essence the central question concerns those molecular mechanisms that are involved in nucleosome formation, stabilisation and destabilisation. Originally DNA condensation may have been initiated by wrapping of DNA around prototype histones using positioning parameters inherent in the DNA sequence. With the advent of the role of nucleosome shuffling mediated by chromatin remodellers in gene regulation the requirement for strong positioning signals may have been attenuated. A consequence of this idea is that the search in vivo for strong positioning sequences in modern genomes may be fruitless. However it is of considerable interest to try to determine the biophysical parameters of DNA that initiate nucleosome positioning and that probably served in primordial nucleosome binding. Nucleosomes are formed by histone octamers consisting of two heterodimers H2A/H2B and one tetramer, (H3/H4)2 wrapping ~145/147 base pairs of DNA ~1.7 times around them in a left-handed supercoil with an average radius of curvature of 9 nm. The way in which a DNA sequence can intrinsically and specifically modulate its malleability and thus variations in the shape of the double-helix, is thought to be an essential factor in nucleosome formation3,4,5,6. Ground breaking studies to identify SELEX-generated DNA sequences that possessed advantageous parameters for nucleosome formation7,8 lead to the suggestion of a positional code for nucleosome positioning and paved the way for crystallographic studies on reconstituted nucleosomes that provided remarkable insights in particular into the bound DNA shape4,9,10 and DNA-histone interface11. Of note however, is that in none of these seminal works is the atomic structure of free DNA studied. Indeed, because of their length, the detailed structures of free 145/147bp DNAs cannot be revealed by traditional approaches such as X-ray diffraction or NMR. This hampered the comparison between free and bound DNAs, and thus the changes induced upon histone binding remained elusive as well as the mechanism of nucleosome formation and the exact nature of positional signals. In this context, we have applied an approach developed in our group12 based on the measurement of the probability of UV induced cyclobutane dimer formation between adjacent pyrimidines (Y-Y dimer) on the same DNA chain12. This technique of photochemical analysis of structural transitions (PhAST), was applied to naked and bound DNA as a probe of changes in local base structure not only between naked DNA and reconstituted nucleosomes but also at different stages of nucleosome formation. UV induced Y-Y dimer formation has already been used to probe nucleosome core structure either by looking at the periodicity of photoproducts13 or by correlating the rate of Y-Y dimer formation with the degree of, and direction of, bending in nucleosomes14. The location of Y-Y dimers and intensities of photo-induced modifications are themselves affected by external agents that reshape the DNA structure and thus alter the photochemistry. However the incident UV light is in no way hampered by the presence of for example a protein. So it has to be borne in mind that although PhAST is not a footprinting technique, as for example in the case of DNase I or micrococcal nuclease, and thus does not give a precise idea of the contact area of a protein with the DNA, it provides unique information on the local DNA structure at a base level. We reveal photochemical products using a simple primer extension technique coupled to capillary electrophoresis. This confers high-resolution, excellent quantification, application in vitro and in vivo, and the possibility of high-throughput since practically any size DNA sample may be analysed in a fashion analogous to genome sequencing. X-ray derived DNA structures containing a thymine-thymine dimer show that the formation of two C5-C5 and C6-C6 covalent bonds is characterized by marked positive rolls (~+20°, see Supplementary Fig. 1) and low twists (~25°)15. At a very simple level, the probability of inducing Y-Y dimer formation may be therefore modulated by the local architecture of naked DNA, in particular the roll and twist angles. The roll angle measures the rotation between two successive base-pair planes about their long axis (y-axis); the roll is positive when it opens up on the minor groove side of the bases, decreasing the distance between the two C5 or C6 atoms of successive pyrimidines. In both naked and bound DNA16,17 positive rolls are generally associated with low twist, thus minimizing the rotation between two successive base-pair planes about the z-axis and hence reinforcing the proximity between two successive bases. One would thus expect that maximal and minimal probabilities of Y-Y dimer formation are indicative of intrinsic positive roll/low twist and negative roll/high twist respectively, in the targeted DNA. In the nucleosome, examination of the high-resolution X-ray structures confirmed that roll and twist are correlated4. Here, we chose to focus on the roll parameter to interpret the measured probabilities of Y-Y dimer formation in free and nucleosomal DNA. Indeed, this parameter is the major player accounting for DNA curvature in the nucleosome4. In addition, roll values show a spectacular periodicity along the nucleosome DNA, clearly less accentuated in the case of twist9. Of course, if no change in Y-Y dimer formation is observed this does not necessarily mean that the roll angles are not affected, Y-Y dimer formation could also be dependent on other factors; in fact, alterations in local flexibility will also affect the time-averaged probability of trapping a suitable Y-Y structure for Y-Y dimer formation. However, on the whole, for reasons that will be discussed in more detail below we interpret changes in Y-Y dimer formation as being indicative of alterations in roll angles. Accordingly, a decrease in probabilities of Y-Y dimer formation can be produced for three roll angle couples of free/bound DNA: i) positive to negative rolls ii) positive to less positive rolls and iii) negative to more negative rolls. An increase in probability of Y-Y dimer formation also relates to three roll angle couples, substituting negative rolls by positive rolls, negative to less negative rolls and positive to more positive rolls. In an attempt to understand ab initio nucleosome formation at a given sequence from a dynamic point of view, we follow structural changes occurring at the base pair level in DNA, as nucleosomes are formed in vitro under decreasing ionic strength conditions. Despite the lack of corresponding DNA sequences in vivo, the possibilities of some bias in the selection process, and the arguments against the existence per se of positioning sequences in vivo advanced above, we used the “Widom” 601 sequence with the idea that its high affinity for the histones is due to a concentration of intrinsic structural properties favouring nucleosome formation. In addition, we considered a mutant sequence (601.2.4) designed to attenuate the putative positioning sequences of the 601 sequence. Our results provide the first dynamic analysis of nucleosome formation that indicates, in agreement with recent data on nucleosome unwrapping18, that sequence dependent intrinsic properties of DNA strongly impact on nucleosome stability and, more specifically, on the recruitment of (H3/H4)2 that is the first stage of nucleosome formation in vitro. This first stage, that serves as a point of nucleation of DNA bending and defines the dyad axis, is thus critical for determining nucleosome positioning subsequent to H2A/H2B recruitment. Moreover, we propose that hydrophobic interactions could play a non-negligible role in the initial recognition by (H3/H4)2. Ultimately, we discuss the existence of positioning sequences in modern genomes and of putative signals persisting from ancient mechanisms of DNA compaction, and which are maintained during the first stage of nucleosome formation.