Discussion and outlook

When interpreting new results one of the most important questions is whether current models remain valid.

The probability density positions obtained form MFD FRET data- sets without FRET anisotropies were similar to the positions deter- mined using surface based TIRF smFRET data, presented in [87]. This result indicates that neither the uncertainties caused by the experi- mental limitations of TIRF microscopy such as the amount of free

144 New developments and applications

double labeled DNA not bound to a protein nor the uncertainties of the MFD experiments presented here (e.g. γ) did cause severe posi- tioning errors.

However the inclusion of FRET anisotropies into the nano-positioning- system (NPS) analysis of the non template DNA strand in RNA poly- merase II (PolII) caused distinct changes especially of positions nt-12, nt-15 and nt-18 (see figure 4.35).

This new data can be best described by a hairpin formation in the non template DNA strand, a situation completely different to that of the expected dsDNA strand exiting the PolII. Additional experiments further solidified the interpretation of the existence of this hairpin with and without PolII bound to the DNA.

Compared to the model derived from the initial probability den- sities [4] the hairpin detected here is positioned a bit further away from the polymerase even though the general direction of the nt-DNA seems to be similar (figure 4.40A).

On the one hand this is a great result since it is the first time that the importance of FRET anisotropy for distance measurements could be demonstrated on non simulated datasets. Furthermore, the fact that in separate experiments the detected hairpin structure was indeed found to be the dominating population proves the reliability of the NPS method.

On the other hand the results presented here require new experi- ments to be performed on the PolII elongation complex to verify the correctness of the DNA exit path model presented in [4].

To this end a new DNA construct not capable of forming stable hairpins was designed. First NPS experiments yielded probability densities with the correct order of nt-8 being closest to the polymerase followed by nt-13 and nt-19. The relative distances of the observed positions to each other are in good agreement with the distances ex- pected for a dsDNA.

The distance of the probability densities to the polymerase is how- ever about 6-10 basepairs larger than predicted by the previous model. Importantly this shift is too large to be consistent with an elongation complex where the distance of each base to the active center is limited by the length of the DNA (figure 4.40B).

4.3 Using the NPS with MFD data 145

One possible reason for such an outward shift is an error in the γ factor. To understand the problem note that, due to the way they are determined, all obtainedγfactors are connected. Throughout the so called quantum yield network analysis (see chapter 4.3.7) a net- work of dye QYs is optimized together with an instrument constant such that in a perfect scenario all datasets yield identical stoichiome- tries. Since the target stoichiometry is unknown and also optimized throughout the sampling a too small network might allow for a too low or too high target stoichiometry caused by a single impurity in one measurement. This would shift all γ factors simultaneously to- wards higher or lower values respectively. As a result such an offset in the correction factor values would directly offset the FRET effi- ciencies obtained. The problem with this theory is that for the data presented here, even a relatively largeγoffset of e.g. 0.3 would cause a distance shift of only about_∼3 Å (assumingκ2=2/3).

Hence, theγfactor alone is not sufficient to fully explain the ob- served shift away from the polymerase.

Another important factor in the calculation of distances from FRET efficiencies that could also contribute to the observed position shift are the anisotropies. In case of the NPS analysis it is assumed that the dyes reorient faster than their fluorescence lifetime and that the observed anisotropies are an average of all possible orientations.

One indication for a problem with that model describing the dye orientations is that only positive FRET anisotropies with approxi- mately the same value (in most cases about 0.1) were detected. For dyes undergoing fast reorientation within fixed boundaries one would however also assume that conditions exist that cause negative FRET anisotropies (i.e. the two dyes are on average tilted by more than the magic angle).

The absence of such negative FRET anisotropies could for example be explained by assuming that several distinct average orientations exist for each dye position and that they interchange slower than the fluorescence lifetime. If the observed average anisotropies are caused by subpopulations of dyes temporally restricted in different positions instead of by one population re-orientating on timescales faster than the dyes lifetime, the static model applied in NPS would no longer be

146 New developments and applications

valid.

Sub-states stable on the timescale of a burst can for example be detected by an anisotropy probability distribution analysis (PDA). For the data presented here and on a timescale of 1 ms, this anal- ysis however revealed only one anisotropy population for each of the channels (donor, acceptor, FRET). It has to be noted that the absence of sub-states on that time-scale does not rule out the possibility of sub-populations stable for a fewµsor shorter.

Nevertheless, further experiments will be necessary to fortify or disregard the above hypothesis of stable sub-states and if they in fact turn out to be an issue, the development of a different model account- ing for such states might be required.

In order to verify or change the non template DNA exit path model it is necessary to precisely determine the distances in the in the PolII elongation complex. Based on the results presented in this thesis a combination of TIRF microscopy and MFD data might be a good approach to minimize remaining error sources.

Using such a combination would allow for an accurate γ correc- tion of individual molecules by the TIRF experiments resulting in more precise FRET data. This data could then be combined with anisotropy (chapter 4.3.6) and Riso

0 (chapter 4.3.7) values from MFD

experiments that can be filtered from artifacts caused by free dye or DNA not bound to PolII (chapter 4.3.5). In addition the MFD FRET efficiency data can be used to ensure that 100% of the labeled DNA are actually bound to the polymerase and PDA (chapter 4.1.2) might serve as a fallback for samples where this goal can not be achieved and a change in FRET efficiency upon protein binding is detected.

Chapter 5

Nucleosome dynamics and

accessibility

5.1 Overview

Nature has developed a highly effective method to fit the DNA into the small compartment of the nucleus of a cell. The so called chro- matin was first detected by Alexander Flemming at the end of the 19th century when he detected a stainable (chroma = colored) struc- ture in the nucleus of a eucaryotic cell. Oftentimes the vast number of the 10000 fold DNA compaction is highlighted to be the most impres- sive feature of the chromatin. The simple compaction alone would be however rather useless if the second feature even more important for life would be missing.

What makes chromatin so outstanding is the combination of high compaction and maintaining the accessibility of the DNA. This not only allows for transcription but also for a regulation of genes by making some easier accessible than others.

Chromatin can be divided into two classes namely euchromatin which is the less condensed form mainly found in the center of the nucleus and the peripherally located heterochromatin. Both of these forms consist of the same "building block" structure called nucleo-

148 Nucleosome dynamics and accessibility

some.

Nucleosomes are the first step of DNA compaction reducing the DNA extension by a factor of about seven. This is achieved by wrap- ping the DNA on a length of 147 base pairs 1.7 times in a left handed superhelical manner around a protein complex [58].

This complex is stabilized by positively charged residues of four core histones, H2A,H2B,H3 andH4interacting with the negatively charged DNA backbone at 14 histone DNA-Backbone contacts. The 11-16 kD core histones all show a similar combination of highly or- dered α helices well conserved throughout species and rather un- structured N-terminal tails.

The next steps of compaction is associated to the linker histones H1 and H5. The H1 and H5 histones are involved in higher order compaction of chromatin and nucleosome spacing. They are not part of the nucleosome core and do not show good conservation between species.

The chromatin is thought to have a "beads on a string" like struc- ture [93] commonly referred to as11 nm fiber followed by the30 nm fiberstate. In this state neighboring nucleosomes are assumed to inter- act resulting in a more condensed structure. Till today the real struc- ture of this 30 nm fiberis unknown, however two competing models exist.

The solenoid model assumes that consecutive nucleosomes are lo- cated next to each other forming a helical superstructure [105] with only one start. In contrast the zigzag model proposes a structure where consecutive nucleosomes are in different helical stacks of a two start helix [56][112]. Further compaction of the 30 nm fiber is necessary to achieve the 10000 fold compaction. This maximum chro- matin condensation however only occurs during the metaphase and the structures involved are not fully understood yet [82].

A canonical nucleosome is formed by a histone octamere core con- sisting of two H2A/H2B dimers, a (H3/H4)2 tetramer [71] and a

DNA wrapped around it. In addition also not fully assembled nu- cleosomes such as tetrasomes (only (H3/H4)2 [1]) and hexasomes

((H3/H4)2and oneH2A/H2B[20]) are known to exist. Interestingly

5.1 Overview 149

activity of PolII an enzyme responsible for transcription also studied in this work (see chapter 4.3).

In order to maintain and control accessibility of the nucleosomal DNA for transcription, replication and repair proteins, several mech- anisms have evolved. [13] use transcription as an example that in- cludes the most important of these processes all at once. They ar- gue that the AT-rich regions of many PolII promoters are related to a bending of the DNA in a way that destabilizes nucleosomes. This destabilization causes a so called nucleosome free region (NFR) which is approximately 100 base pairs long and located at the transcription start site. The first nucleosome next to this NFR is oftentimes found to contain alternatively spliced H2A.Z instead of canonical H2A while the mayor part of the coding region is usually packed by canonical nucleosomes. Nucleosomes containing H2A.Z are thought to be less stable than canonical nucleosomes and might hence make the NFR more easily accessible. It is stated that the transition to the active state is accompanied by increased histone acetylation, nucleosome movements in the promoter and coding regions and nucleosome loss close to the transcription start site.

Following this example the easiest way of controlling DNA ac- cessibility is related to the intrinsic instability of nucleosomes which allows for a detachment and reattachment of the DNA ends com- monly referred to as breathing or transient DNA unwrapping [67]. Even though no additional proteins are involved in the breathing it- self various pathways to control this process exist.

The two most important ways of controlling the nucleosome sta- bility aiming to change the histone DNA affinity are alternatively spliced histones and post translational modifications mostly methy- lation or acetylation of amino-acids in close proximity to the DNA.

In chapter 5.5 the change in affinity caused by acetylation and methylation of a lysine on position 64 of the H3 histone is investi- gated. The destabilizing effect is found to be negligible for strong positioning sequences such as theWidom 601sequence [70][122] used here [24].

Furthermore the stability of nucleosomes containing a novel alter- natively splicedH2Avariant (H2A.Z.2.2) or an other variant (H2A.Z.2.1)

150 Nucleosome dynamics and accessibility

is compared to the stability of canonicalH2Ain chapter 5.4.

Another way for an organism to guarantee accessibility of genes is active repositioning of nucleosomes on the DNA strand catalyzed by an ATP driven chromatin remodeling factor (remodeler). In the above example the remodeler is required to allow access of the PolII to the actual gene. This repositioning is also required for other pro- cesses e.g. to induce equal distances of nucleosomes on a DNA strand prior to further compaction oftentimes accompanied by gene silenc- ing. An example for such a remodeler is the well studiedACF(ATP- utilizing Chromatin assembly and remodeling Factor) complex which is amongst other functions capable of equally spacing the nucleo- somes on a DNA strand [138][104]. The mammalian ACF is assumed to be a tetramer consisting of two Snf2H (Surcorse non-fermenting 2 homolog) motor proteins and two ACF1 proteins [103]. While Snf2H is responsible for the actual repositioning of the nucleosome the ACF1 is supposed to control direction and speed of the repositioning [39].

In order to understand the individual functions of different re- modeling enzymes it is important to understand the differences be- tween them. To this end, the two ISWI ATPases present in human cells, namely Snf2H (ATPase of the CHRAC and ACF complexes) and Snf2L (ATPase of the NURF complex), were investigated in chap- ter 5.3 and different activity changes upon increasing dilution could be resolved.