Introduction

In my office is a painting of Stockholm (Figure 1). It is a watercolor painted and given to me by my mother depicting a view from Sodermalm showing parts of Gamla Stan and Kungsholmen and prominently displaying Stadshuset, Stockholm’s City Hall. I guess one could call the painting a cityscape depicting one part and one perspective of that northern city. When I look at the picture I often wonder how the cityscape has changed over the years, decades, centuries and even millennia. Before the founding of the old city on Gamla Stan, the surrounding islets, rivers, valleys and hills would have appeared much different. But even under the urban exterior one can still make out the underlying features intrinsic to the landscape upon which the city was built.

Parallels can be drawn between this cityscape and how it is influenced by the underlying/natural landscape and that of chromatin architecture and underlying DNA sequence (natural landscape). Just as the natural landscape can influence the ultimate form and outcome of the cityscape, so too can DNA sequence influence the position of nucleosomes and chromatin architecture. The extent to which underlying DNA sequence affects nucleosome positioning is currently a topic of considerable discussion and active experimentation. To contribute to the discussion, I will outline a few of the methods, data and arguments that I find compelling and that I believe will ultimately resolve the question of what positions nucleosomes. Many of the experimental methods described below rely on the use of micrococcal nuclease for isolation of nucleosomes, and various high-throughput sequencing methods for mapping of nucleosomes to the genome or analysis of positioning. These reagents and technologies are not without potential biases [previously discussed in (1) and (2)], many of which can be overcome, or at least somewhat decreased, with careful control experiments. Nonetheless, it is important to be cognizant of these limitations. Despite these caveats, I will attempt to sketch a portrait of my current perspective on what influences the landscape of nucleosome positioning and chromatin architecture.

In vitro reconstitution: Invitrosomes

The influence of DNA sequence on nucleosome positioning at its most basic level can be assayed effectively by in vitro nucleosome reconstitution experiments, where octamer and DNA are combined together in a high salt solution. These conditions keep the protein core and DNA from associating, and with subsequent dialysis, the salt concentration is slowly decreased and the protein cores and DNA are allowed to come together and form in vitro reconstituted nucleosomes (invitrosomes) (3). This procedure is thought to allow formation of nucleosomes on those DNA sequences which are most amenable to the extreme bending of the double helix which is necessary for nucleosome formation (4). Thus, when such a reconstitution is performed, the DNA cores of these in vitro nucleosomes contain the sequences which are most favorable (energetically and otherwise) to bending around and/or attractive to the histone core. Invitrosome reconstitution, being done in the presence of only DNA and histone octamer, reveals the contribution of only the constituent components independent of the other chromatin influencing factors normally found in the cell.



Figure 1: Watercolor of Stockholm City Hall, by Becky H. Johnson.

There are important considerations to take into account when designing these reconstitution experiments and evaluating the outcomes of different invitrosome reconstitution assays. These reconstitution experiments can be done using either isolated or recombinant histone octamer and with either isolated-naked or synthesized DNA. Isolated octamer is a complex mixture of histone proteins with all the combinations of histone tail modifications and histone variants present in the source material, whereas recombinant octamer is a homogenous population of identical histone protein cores. Additionally, the density at which the invitrosomes are reconstituted on the DNA is also a major consideration.

Different nucleosome densities enable similar analyses to be done, but at the same time address different fundamental questions. Nucleosomes can be reconstituted at a DNA to protein mass ratio of 1:1 with the goal of forming arrays of nucleosomes on the DNA molecules that mimic the density of nucleosomes found in vivo and recapitulate the regular spacing that is present in cells [e.g., (5)]. It is possible that in this type of experiment, a physiological density of nucleosomes may not be achieved without the help of nucleosome assembly factors or even histone H1, but adding these types of factors can actually alter the positioning of the reconstituted nucleosomes (5). The other extreme is to have an overabundance of DNA molecules such that, on average, only one invitrosome will form on each DNA molecule. The former method, by mimicking the general structure and spacing of arrays that are found in the cell, can obfuscate the underlying contribution that sequence has on nucleosome positioning, while the latter allows no consideration of steric hindrance or constraint on nucleosome positioning due to neighboring nucleosomes which exist in the cell. The preferences revealed by the array reconstitutions may be similar to the type of data from in vivo experiments, while the single invitrosome experiments show the basic biophysical preferences of individual nucleosome formation under no other constraint, but with the possible caveat of not assaying DNA sequences with low affinity for nucleosomes due to the overabundance of DNA relative to octamer.

Other variables in invitrosome experiments are the source of the input DNA and how the DNA fragments used to form invitrosomes are generated. If the input DNA is isolated from a cellular source then it is, in many cases, subject to methylation or other DNA modifications; whereas synthesized DNA fragments could be controlled and uniform in these aspects. Also, it is impractical to reconstitute nucleosomes in vitro on an entire chromosome, thus smaller fragments must be used. The decision on what size DNA fragments to use is not a trivial one, for it will determine the length of arrays possible or artificially limit or influence nucleosome formation by the proximity of the ends of the DNA molecules.

The ends of DNA fragments can be of considerable concern as there is a tendency for nucleosomes to “roll” to the end of short DNA fragments during reconstitution or even when isolating in vivo nucleosomes under high salt conditions (6, 7). Thus the resulting data can be skewed and reveal a picture heavily influenced by DNA ends and not by favored nucleosome positions based on sequences preferences. This is a concern for both array-reconstitution and individual-invitrosome reconstitution experiments. DNA fragments ends would not only influence positioning of individual nucleosomes, but could influence the phasing (regular spacing) of entire in vitro nucleosome arrays thus revealing end preferences and not sequence preferences. It is possible that some of these caveats, at least in the case of individual-invitrosome reconstitution experiments, can be overcome by using DNA molecules with defined ends produced by digestion of genomic DNA to completion with restriction endonucleases. Multiple DNA samples can be produced using different restriction enzymes to allow for reconstitution on fragments that would be too small in any one restriction set. Producing DNA fragments this way with defined ends would allay some of the concerns mentioned above and avoid other real or imagined concerns about random shearing of DNA via sonication or other methods.

Despite all these variations and caveats, invitrosome experiments on both synthesized DNA and whole genomes have revealed definite preferences between histone core and underlying DNA sequences (4, 5, 8). Eukaryotic genomes have been found to harbor more of these sequence features and be more amenable to nucleosome formation than prokaryotic genomes (5), suggesting evolution has worked on these genomes to enhance nucleosome formation. In addition to the DNA characteristics that enhance invitrosome formation, early experiments have revealed sequences that are strongly recalcitrant to nucleosome formation (homopolymeric runs of A’s and T’s) (reviewed in 9). These sequences are now experiencing a renaissance of interest and evaluation by current researchers. Many of these nucleosome positioning and repelling DNA preferences are present and were even first discovered by groups looking at relatively small sets of in vivo nucleosome cores (10-13). Recent analyses of more extensive in vivo nucleosome sets have reinforced and expanded on these early findings (14, 15). What is clear from all the invitrosome data is that when all else is discounted, DNA sequence can position nucleosomes in the test tube, and rules for nucleosome position under these conditions can, in many cases, be defined.

Other powerful in vitro methods exist to assay the interaction of DNA sequences with histone octamer in the formation of nucleosomes [e.g., laser-tweezer single molecule techniques (16, 17) and gel-shift assays on reconstituted nucleosomes (6)]. While these techniques do not directly address nucleosome positioning, they can provide insight into nucleosome assembly on defined DNA sequences and should be used to test any rules of DNA and octamer interaction elucidated by invitrosome experiments.

In vivo analysis

While the preferences of DNA histone interactions can be defined in vitro, the extent to which these rules are followed in living organisms can only be assessed by looking at nucleosome positioning in vivo. In its most general sense, there are two approaches to in vivo analysis of nucleosome positions and positioning. The first is a random shotgun-like approach and the second is a focused locus-based approach. While the term shotgun evokes images of massive sequencing, and locus-based might imply smaller scale and/or individual assays, this is not the distinction I am making with these terms. Rather, shotgun is a more random, unbiased method relying on sequencing of individual, isolated nucleosome core DNA fragments that can be applied to either a few hundred sequences, or a few hundred million. The locus approach differs fundamentally from the former in that it looks at occupancy of genomic loci rather than assaying the position of individual nucleosome cores. Again it can be applied to an individual locus with Southern blots or other such techniques, or used genome-wide with microarray assays. Whereas at high enough density the former (shotgun approach) can be readily converted to the latter (locus approach) with coverage at the resolution of the micrococcal nuclease cutting bias, the latter is not easily converted to the former to elucidate positioning without sophisticated mathematical computations and use of statistical predictions invoking such things as the hidden Markov model, which is beyond this investigator. Thus at their core, and in their unadulterated forms, the two methods assay very different things: absolute position/positioning versus nucleosome occupancy

What is common between these two approaches is an attempt to analyze nucleosome positions/occupancy in a native context, which is accomplished by isolating nucleosomes from chromatin contained within the cell/tissue/organism and analyzing nucleosome positions en masse or by looking at individual loci one at a time or in a more comprehensive manner. Whereas the small scale shotgun approaches have been useful to establish and reinforce the ideas of DNA sequence preferences and periodicities in the DNA cores (11, 13), and the small-scale or single locus based approaches have defined individual examples of increased or decreased nucleosome occupancy or even positioning (18), the recent large genome-wide assays have helped to define nucleosome DNA preferences in vivo (14, 15) while demonstrating a general lack of nucleosome positioning, with the limited amount of observed positioning not due to a DNA nucleosome positioning code, but rather as a result of occupancy resulting from genic features (some of which is influenced but not dictated by underlying sequence) (1, 2, 9, 19-23).

While some might argue that occupancy is sufficient to reveal how and where nucleosomes are being positioned, it is important to not confuse relative occupancy with absolute nucleosome positioning. Because of the nature of the technology, based on hybridization and relative signal intensities, microarrays reveal relative occupancy. The fold enrichment of nucleosome occupancy at any locus is compared to a genome-wide (or array-wide) background level of signal and the occupancy in other regions, and is presented as either enriched (for positioned nucleosomes) or de-enriched (for nucleosome-free regions) based on normalized signal intensities. The fundamental flaw with this approach is that it could distort relatively modest enrichment or de-enrichment to appear as high occupancy or depletion.

Consider a scenario where a genome has a general lack of positioned nucleosomes, with 90% of nucleosomes being randomly positioned. In assaying this chromatin state with microarrays, the uniform signal from nine tenths of the nucleosome cores is subtracted out as background and only the remaining 10% of positioned nucleosomes across the genome is assayed as above background. This would result in areas of the genome with only randomly positioned nucleosomes appearing to be relatively nucleosome free and areas with only 10% fluctuations in nucleosome occupancy possibly appearing as highly enriched. Despite their appearance, these nucleosome-free regions would not be free of nucleosomes, but rather highly occupied by non-positioned nucleosomes, and by the same logic, loci with seemingly highly positioned nucleosomes would in reality be 90% randomly positioned with only a small fraction of nucleosomes actually positioned. By contrast, shotgun approaches using sequencing to define the absolute positioning of millions of individual nucleosomes can be converted to occupancy data that does not suffer from this relative occupancy flaw (2). Therein lies the fundamental difference between relative occupancy as assessed by microarrays or other similar methods, versus absolute positioning and true occupancy or positioning as assayed by ultra-high throughput sequencing.

Evidence, Arguments and Synthesis

In the past half decade technological advances have allowed us to move from the single locus analysis, or the assay of a few hundred nucleosome cores, to comprehensive, genome-wide analysis of positioning, not just occupancy as provided by microarrays. When I first joined Andrew Fire’s lab and started working on the question of nucleosome positioning in the middle of 2004, we thought we would be able to make a major contribution to the question of nucleosome positioning by sequencing and analyzing 30,000-50,000 nucleosome cores. I embarked in optimizing our protocols so we could use a modified SAGE-tag method to assay multiple nucleosomes with a single sequencing reaction using the Sanger method. We realized we did not need to sequence across the whole nucleosome core in order to define a nucleosome position in the genome, but rather only needed a reliable digestion protocol, resulting in a uniform mononucleosome DNA core size, which would allow us to sequence only one end of the core to define its position in the genome. While optimizing and performing qualitative sequencing assays and even dabbling in using microarrays, Andy came to me with the proposition of getting hundreds of thousands of core sequences/positions using the 454 pyrosequencing platform. We made use of this technology and sent off our samples to Branford, CT, resulting in a couple hundred thousand nucleosome positions in C. elegans that we used to produce the first genome-wide glimpse of nucleosome positioning using high-throughput sequencing (1). Soon after, we performed a similar analysis in collaboration with Applied Biosystems to ultimately look at 44 million nucleosome positions in C. elegans (2). We have now, using the SOLID platform, looked at over a billion nucleosome positions in human tissues. The exponential growth in the capacity of sequencing technology has allowed a paradigm shift to occur, moving from anecdotal and local assays to global analysis. What has this global look told us? The same thing that the initial 454 data of a few hundred thousand nucleosome positions did: that nucleosome positioning is relatively rare across the genome of a multicellular organism.

In presenting these data for review and publication, a common criticism was: of course you don’t see nucleosome positioning when looking in a multicellular organism with different cells and different tissues. Each tissue will have its own nucleosome density (the number of nucleosomes formed per unit length of DNA), and thus looking across tissues will not show you positioning. This statement demonstrates a glaring lack of comprehension of the fundamental question being addressed by these studies. The real question being asked was whether sequence alone positions nucleosomes in the genome or if positioning is accomplished by other factors. The argument is based on the following lines of reasoning:

1. Each cell (or at least most cells) has identical copies of the genome contained within its nucleus regardless of the cell or tissue type.
2. In the absence of remodeling due to cell-specific activation of genic regions or other factors, nucleosome positioning is dictated by DNA sequence (a nucleosome positioning code).

If statements 1 and 2 are true, then it follows that since the DNA is the same in every cell, the nucleosome positions should be the same in every cell, especially across genomic regions that lack cell- or tissue-specific chromatin remodeling due to factors relating to genic expression. The data are clear in that they show a general paucity of positioned nucleosomes across the genome in our pancellular assay in C. elegans. Therefore, either the entire genome is remodeled in each cell/tissue type, or nucleosome position is not dictated by the underlying sequence. In fact, it is ironic that areas in the genome that do seem to have an increase in positioned nucleosomes, correlate with genic features (transcription start sites, exons, introns and polyadenylation sites) (2, 23-25), further reinforcing the idea that positioning is not a result of the underlying sequence, but rather a result of ATP-dependent chromatin remodeling that is coupled with activation and transcription (or at least initiation of transcription having RNA Polymerase II present and occupying the transcriptional start site) of specific genes in specific tissues. Indeed in C. elegans we observed that regions upstream of the translational start sites of genes, which were de-enriched for nucleosomes, had well positioned nucleosomes flanking these nucleosome-free regions on a genome-wide level, and that these features could be seen on a local level when looking at individual loci containing ubiquitously expressed genes (2).

This leads to the proposed mechanism for nucleosome positioning. Having demonstrated that DNA sequence is not dictating nucleosome positions in a pancellular way, there is good evidence showing that there is increased nucleosome positioning in individual cell types, or in single-celled organisms (19, 23, 26). Again, based on the above arguments, this positioning is not being dictated by the sequence, but is present nonetheless. Thus, if it’s not sequence that is doing the positioning, then what is? Genome-wide nucleosome positioning data demonstrate that the most likely culprit is trans-acting DNA-binding factors, like transcription or chromatin factors such as CTCF (26-28). In fact, whereas in vivo genome-wide data demonstrate nucleosome-free regions at these DNA-binding element sites, in vitro data demonstrate that it is not the sequence of the DNA that is excluding the formation of nucleosomes, as these sequences are actually tremendously enriched in invitrosome occupancy (unpublished data). These data invoke a barrier model for the cause of nucleosome occupancy and positioning (29). This model is very attractive because it provides a ready explanation for the in vivo observations. In this model, promoters, enhancers, transcription factor and other trans-acting factor binding sites would be occupied by their cognate trans-acting factors and thus prevent nucleosomes from forming at these sites. The pattern of binding across the genome would be cell and tissue specific, and remodeling and displacement of nucleosomes previously occupying these sites would be accomplished by ATP-dependent chromatin remodeling factors as directed by the normal function of the cell. After binding of the trans-acting factors, the sites now unavailable to nucleosomes because of steric hindrance would set up a barrier against which nucleosomes previously occupying these sites, and now immediately flanking the barrier, would be well positioned a linker length away. Subsequent nucleosomes radiating out from the barrier position would gradually lose their precise positioning due to subtle variation in linker length, which even occurs among regularly spaced nucleosomes arrays (the default state of the bulk of the genome as observed by laddering of micrococcal nuclease-digested chromatin). Because of the cell- and tissue-specific nature of the binding patterns of these trans-acting barrier factors, nucleosome positioning would be more evident when assayed within a single cell or tissue type, but when assayed in a pancellular manner would be relatively absent, except in loci harboring ubiquitously expressed genes (such as house keeping genes) or with factors that bind in the same place in the majority of cell types or tissues. As we further define the binding sites of trans-acting factors in a cell- and tissue-specific manner and we generate cell- and tissue-specific nucleosome position maps, we will be able to see whether the predictions based on the barrier model hold true and explain the landscapes of chromatin architecture as assayed by nucleosome positions.

Finally, the fact that DNA sequence does not dictate nucleosome position is not the same as saying that it does not influence nucleosome positioning. The early in vivo and in vitro evidence clearly demonstrate DNA sequence preferences, that are perfectly compatible with the barrier model. Furthermore, within the confines of the locales dictated by the barrier model, DNA sequence can influence translational positioning (nucleosome placement) with preferred nucleosome translational positions being spaced ten base pairs apart as to maintain a constant rotational setting (orientation of DNA in relation to the histone octamer) enhancing nucleosome formation in these regions [e.g., (1, 2, 5, 21, 30)].

Beyond positioning due to trans-acting factor binding, there are both very old and relatively new invitrosome data suggesting that specific sequences are actually quite recalcitrant to nucleosome formation (reviewed in 9). These nucleosome-excluding sequences would play the role of the barrier, excluding nucleosomes from these loci and setting up highly positioned flanking nucleosomes and arrays as described above. It is interesting to note that these sequence characteristics are enriched in linker sequences (1) and in some promoter elements (31), the latter possibly enabling core transcriptional machinery more ready access to their cognate binding sites. It is likely that use of these nucleosome-excluding sequences will ultimately be the most successful approach in directing nucleosome positioning for therapeutic and experimental use, and that a combination of both nucleosome-excluding/repelling and positioning sequences along with ubiquitous barrier binding sites may ultimately prove to be a genuine nucleosome positioning code. Current studies in my lab are actively engaged in testing these ideas to regulate transgene expression and influence heterochromatinization within the worm, with the aspiration that such studies will be translatable to higher organisms. Another intriguing aspect of this work will be to determine the affect of histone variants and tail modifications on nucleosome formation, and to see if sequence plays a role in the fundamental associations at these levels and with these and other epigenetic modifications.

In conclusion, I’ll return to my watercolor of Stockholm. With current tools and methods, the immutable landscape’s influence on the cityscape of nucleosome positioning is changing perspective. Ultra-high-throughput sequencing and genome-wide invitrosome analysis are now refining our perceptions of what is going on within the cell, refocusing our perspective from a fuzzy impressionistic interpretation to a sharply defined, almost photographic depiction of the nucleosome landscape.

This research was reported by the author in part at Albany 2009: The 16th Conversation (32).

References and Footnotes

1. S. M. Johnson, F. J. Tan, H. L. McCullough, D. P. Riordan, and A. Z. Fire. Genome Res 16, 1505-1516 (2006).
2. A. Valouev, J. Ichikawa, T. Tonthat, J. Stuart, S. Ranade, H. Peckham, K. Zeng, J. A. Malek, G. Costa, K. McKernan, A. Sidow, A. Fire, and S. M. Johnson. Genome Res 18, 1051-1063 (2008).
3. K. Luger, T. J. Rechsteiner, and T. J. Richmond. Methods Enzymol 304, 3-19 (1999).
4. A. Thåström, P. T. Lowary, H. R. Widlund, H. Cao, M. Kubista, and J. Widom. J Mol Biol 288, 213-229 (1999).
5. Y. Zhang, Z. Moqtaderi, B. P. Rattner, G. Euskirchen, M. Snyder, J. T. Kadonaga, X. S. Liu, and K. Struhl. Nature Structural & Molecular Biology 16, 847-852 (2009).
6. G. Meersseman, S. Pennings, and E. M. Bradbury. EMBO J 11, 2951-2959 (1992).
7. T. Sakaue and K. Yoshikawa. Physical Rev Letters 87, 078105 (2001).
8. N. Kaplan, I. K. Moore, Y. Fondufe-Mittendorf, A. J. Gossett, D. Tillo, Y. Field, E. M. LeProust, T. R. Hughes, J. D. Lieb, J. Widom, and E. Segal. Nature 19, 362-366 (2009).
9. E. Segal and J. Widom. Curr Opin Struct Biol 19, 65-71 (2009).
10. S. Muyldermans and A. A. Travers. J Mol Biol 235, 855-870 (1994).
11. D. C. Satchwell, H. R. Drew, and A. A. Travers. J Mol Biol 191, 659-675 (1986).
12. A. A. Travers and S. V. Muyldermans. J Mol Biol 257, 486-491 (1996).
13. I. Ioshikhes, A. Bolshoy, K. Derenshteyn, M. Borodovsky, and E. N. Trifonov. J Mol Biol 262, 129-139 (1996).
14. F. Salih, B. Salih, and E. N. Trifonov. J Biomol Struct Dyn 26, 273-281 (2008).
15. I. Gabdank, D. Barash, and E. N. Trifonov. J Biomol Struct Dyn 26, 403-412 (2009).
16. B. D. Brower-Toland, C. L. Smit, R. C. Yeh, J. T. Lis, C. L. Peterson, and M. D. Wang. Proc Natl Acad Sci 99, 1960-1965 (2002).
17. S. Mihardja, A. J. Spakowit, Y. Zhang, and C. Bustamante. Proc Natl Acad Sci 103, 15871-15876 (2006).
18. F. Dong, J. C. Hansen, and K. E. van Holde. Proc Natl Acad Sci 87, 5724-5728 (1990).
19. E. Segal, Y. Fondufe-Mittendorf, L. Chen, A. Thåström, Y. Field, I. K. Moore, J. P. Wang, and J. Widom. Nature 443, 750-752 (2006).
20. H. E. Peckham, R. E. Thurman, Y. Fu, J. A. Stamatoyannopoulos, W. S. Noble, K. Struhl, and Z. Weng. Genome Res 17, 1170-1177 (2007).
21. I. Albert, T. N. Mavrich, L. P. Tomsho, J. Qi, S. J. Zanton, S. C. Schuster, and B. F. Pugh. Nature 446, 572-576 (2007).
22. W. Lee, D. Tillo, N. Bray, R. H. Morse, R. W. Davis, T. R. Hughes, and C. Nislow. Nat Genet 39, 1235-1244 (2007).
23. D. E. Schones, K. Cui, S. Cuddapah, T. Y. Roh, A. Barski, Z. Wang, and G. Wei, K. Zhao. Cell 132, 887-898 (2008).
24. T. N, Mavrich, C. Jiang, I. P. Ioshikhes, X. Li, B. J. Venters, S. J. Zanton, L. P. Tomsho, J. Qi, R. L. Glaser, S. C. Schuster, D. S. Gilmour, I. Albert, and B. F. Pugh. Nature 453, 358-362 (2008).
25. N. Spies, C. B. Nielsen, R. A. Padgett, and C. B. Burge. Mol Cell 36, 245-254 (2009).
26. unpublished data.
27. Y. Fu, M. Sinha, C. L. Peterson, and Z. Weng. PLoS Genet 4, e1000138 (2008).
28. S. Cuddapah, R. Jothi, D. E. Schones, T. Y. Roh, K. Cui, and K. Zhao. Genome Res 19, 24-32 (2009).
29. R. Kornberg. Nature 292, 579-580 (1981).
30. P. D. Partensky and G. J. Narlikar. J Mol Biol 391, 12-25 (2009).
31. K. Struhl. Prot Natl Acad Sci 82, 8419-8423 (1985).
32. Abstracts of Albany 2009: 16th Conversation. June 16-20, Albany, New York, USA; S. M. Johnson, A. Valouev, S. Boyd, C. Smith, A. Sidow, A. Fire.. Genome-wide Mapping and Analysis of Nucleosome Positions in Multiple Human Tissues, Abstract #205. J Biomol Struct Dyn 26, 787-927 (2009).