The “bead-string” morphology of eukaryotic chromatin has been an established view for 40 years [1, 2]. Its basic folding unit is the nucleosome, in which about 150 bp of DNA are wrapped around a protein core consisting of two copies each of H2A, H2B, H3, and H4 histones.
In addition to their packaging role, nucleosomes regulate the accessibility of the DNA to proteins. Chromatin accessibility depends on nucleosome stability through local epigenetic modifications to histones and DNA, which are associated with cell development, differentiation, and cellular dysfunction such as cancer. Thus, exploring nucleosome dynamics is central to our understanding of chromatin structural variations and their role in biology. These aspects may be studied in molecular detail in vitro on mononucleosome samples.
The first X-ray crystallographic determination of the nucleosome structure to atomic resolution  has stimulated the field enormously and has given rise to a large number of follow-up studies to higher resolution, with histones from different origin, and different DNA sequences (for a review, see ). The two DNA sequences from which successful crystal structures have been obtained—the strongly positioning alpha-satellite and the Widom 601 sequence—yielded very similar nucleosome structures . This does not exclude possible strong sequence effects on nucleosomes in other contexts. Histone protein variation seems to influence the crystal structure more strongly, e.g., the dimer–dimer contacts in nucleosomes from recombinant histones seem to be tighter for Xenopus than for yeast histones  and the internucleosomal interactions are different between nucleosomes prepared from yeast or metazoan histones [7, 8].
The structure obtained from crystallography is static and represents only one particular conformation. While nucleosomes in a crystal reconstituted from perfectly identical pieces of a very particular DNA with recombinant histones can be viewed like cultured pearls, nucleosomes in the real genome would be like genuine pearls with a practically infinite variety of DNA sequences together with epigenetic variations of histones. At first glance (e.g., in electron microscopic images), these “natural pearls” all seem to be the same, because these variations affect only the fine structure, not the overall shape. The fine structure of nucleosomes is the key to their function, allowing control of DNA accessibility or chromatin packaging.
The consequence is that for getting access to nucleosome fine structure and dynamics, adequate experimental techniques must be used. There also exists a large body of experimental data on the solution structure of nucleosomes by solution scattering and analytical ultracentrifugation [9-12]. The samples for these studies are often a mixture of naturally diverse nucleosomes isolated from a cell type. Such solution studies show that nucleosomes are highly dynamic under in vivo conditions and that structure fluctuations play an important role in their function.
Structural studies on nucleosomes in solution have been advanced in particular by using Förster resonance energy transfer (FRET) on fluorescently labeled samples. FRET uses a donor and an acceptor fluorophore label on the molecules under study. The donor dye is excited by a laser, the longer wavelength acceptor dye may then be excited by nonradiative energy transfer from the donor. The efficiency E of this transfer increases with decreasing interdye distance R (E ∝ R−6) and, therefore, allows determination of the distance on the nanometer scale . Fluorescent labeling requires the use of a homogeneous population of reconstituted nucleosomes, similar to crystallization. However, two important differences exist: First, not all nucleosomes that can be reconstituted will also crystallize. Second, in a crystal, only particular conformations are selected and a particular ionic environment is present. Nucleosomes in solution, on the other hand, are free to explore all their accessible conformational space and solution conditions can be varied over a much wider range.
Bulk solution FRET as measured by fluorescence emission intensity spectra measures an average over many molecules. While some deeper-going information can be obtained through fluorescence lifetime measurements, distinguishing species of different FRET efficiencies by their lifetimes is not a very selective method: Estimating a distribution of exponential decay times is a known ill-posed problem . For detecting dynamics and conformational heterogeneity, molecules must be observed individually. To this aim, single-pair FRET (spFRET) has been developed, where the sample fluorescence is excited and observed in a high-resolution microscope with confocal optics. At sufficiently low fluorophore concentrations (<0.1 nM), one observes fluorescence bursts from single molecules passing through the laser focus. These bursts can then be analyzed statistically to create a FRET efficiency distribution and to characterize eventual time-dependent conformational changes [15, 16]. We note that lifetime analysis of single molecule bursts can improve the quality of the data considerably .
A number of FRET and spFRET studies on nucleosomes exist: for review, see . With the establishment of these techniques, the literature on FRET and spFRET application to the structure and dynamics of nucleosomes, chromatin, and chromatin-associated proteins has shown significant growth in the last 5 years. In our own work, we measured the average distance between the ends of different length nucleosomal DNA and established that the DNA ends of mononucleosomes in solution do not cross . Furthermore, we characterized the salt- and linker histone H1-dependent structure of mono-  and tri-nucleosomes .
Under physiological conditions, nucleosomes open by thermal fluctuations that expose DNA sites [22, 23]. These opening events are only transient and sparsely populated, but they allow proteins such as chromatin remodelers, transcription factors, or replication factories to bind and stabilize the open state. Without additional protein factors, characterization of the open state requires to increase its equilibrium fraction artificially, e.g., by thermal denaturation, denaturating agents, or change in salt concentration. We have chosen the latter as a convenient means to destabilize the nucleosome structure and thereby increase the opening probability.
We used spFRET to characterize the stability changes in nucleosomes upon histone acetylation, changes of DNA sequence, and ionic strength . Using salt-induced destabilization in spFRET, we proposed an opening mechanism of nucleosomes reconstituted on the “Widom 601” DNA  and recombinant Xenopus histones . Later, by comparing the salt-induced decrease in FRET for five different donor/acceptor pairs on the DNA, H2B, and H4 histones, we discovered a new intermediate in which the histone core opens at the interface between the H2A/H2B dimer and the (H3H4)2 tetramer, before the first H2A/H2B dimer dissociates from the DNA . We could estimate that this new “butterfly state” is populated to the order of one percent under physiological conditions.
It is now important to establish whether the mechanism found in that work is generally characteristic of nucleosome opening, and how nucleosome stability and opening pathways may vary between nucleosomes from different species and DNA sequences. For instance, earlier biochemical studies on isolated oligo- and mono-nucleosomes revealed differences in salt-induced and thermal unfolding between yeast and higher eukaryotes . To understand these biologically important variations, we studied a larger variety of nucleosomes reconstituted from histones of different origin, and on varying DNA sequences. Our data show conclusively that nucleosome stability is strongly affected by DNA sequence, whereas the origin of the histone proteins only modestly altered nucleosome stability. More importantly, we demonstrate that the pathway of nucleosome dissociation is identical for all nucleosome samples, suggesting that disassembly via the butterfly state is a general mechanism for nucleosome dissociation.
Materials and Methods
Preparation of Labeled Mononucleosomes
Mononucleosomes were assembled on different 170 bp long fragments containing the “Widom 601” , 5S rDNA, or MMTV-B nucleosome positioning sequence. Fluorescently labeled DNA was prepared by PCR [20, 24]. Three different labeling schemes were used to monitor the dissociation pathway as described previously . Donor and acceptor dyes were either placed on the DNA at positions −42 and +52 with respect to the dyad axis (denoted as DNA-dl), or the donor was conjugated to a cysteine introduced at position 122 of histone H2B, and the acceptor dye was placed at positions +15 or +52 with respect to the dyad axis (referred to as H2B-dyad and H2B-int, respectively), see Figure 1. The approximate geometry of the labeled complexes is described in . Before PCR, the quality of the labeled primers was checked on a native polyacrylamide gel. The labeled, PCR-amplified DNA fragments were purified on an ion exchange column (Waters) using HPLC (Unicam); only the fraction with the correct ratio of Alexa594 to Alexa488 absorption was used for subsequent nucleosome reconstitution. We estimated that the fraction of nondouble-labeled DNA after purification was less than 5%.
Recombinant histone proteins of Xenopus laevis were expressed and purified as described previously . Recombinant yeast and mouse histones as well as mutated H2B T112C histones  from Xenopus, mouse, and yeast were purchased from the Protein Purification and Characterization Facility (Department of Biochemistry Colorado State University).
Mutated H2B histones were labeled on the 112C position with Alexa488-maleimide. Labeled and unlabeled H2B histones were mixed at 1:9 ratio to minimize the presence of two donor molecules in one nucleosome. This will create an excess of nucleosomes labeled only at the acceptor DNA resulting in photon bursts only in the acceptor channel; such events did not cause a major problem since the probability of direct excitation of the acceptor at the donor wavelength is quite low, and they were not considered in the data analysis. Recombinant histones were reconstituted into octamers and purified on FPLC .
Purified DNA and octamers were mixed in 2 M NaCl-TE buffer and reconstituted into nucleosomes by gradual salt dialysis down to 5 mM NaCl. The molar ratio between DNA and octamer was optimized between 1:1.6 and 1:2 to avoid aggregation and to minimize excess free DNA. Where needed, nucleosomes were centrifuged at 10,000 rpm (Eppendorf Centrifuge 5417R, corresponding to 10,600g rcf) for 10 min to remove residual aggregates. The quality of nucleosomes was checked on a native 6% polyacrylamide gel (acrylamide:bisacrylamide = 60:1). Nucleosomes were stored in stock solution of 100–200 nM concentration at 4°C for up to 3 weeks.
Single Molecule FRET Experiments
spFRET experiments were performed on a homebuilt confocal system as described in detail elsewhere . A schematic view of the device is shown in Figure 1B. All confocal experiments were performed in 384-well microplates (SensoPlate Plus, Greiner Bio-One) that were passivated with Sigmacote™ as described in . Briefly, the microplates were cleaned by soaking in 1% Hellmanex solution (Hellma) and 100 mM HCl twice for 30 min, with thorough washing with ddH2O in between. After the final acid treatment, microplates were washed and dried under low vacuum. To passivate the surface, each well was filled with Sigmacote™ solution, incubated for 15–20 s and washed with ddH2O. The plates were again dried under low vacuum and sealed with film (Bio-Rad) to avoid exposure to dust and stored for subsequent use.
Nucleosomes were freshly diluted into the experimental buffer (TE buffer, pH 7.5, supplemented with 0.02% Nonidet P40 (Roche Diagnostics), 1 mM ascorbic acid to minimize photobleaching, and NaCl as noted). For DNA-dl samples, 40–50 pM labeled nucleosomes were mixed with unlabeled nucleosomes to a final nucleosome concentration of 250 pM. For the histone-DNA labeled pairs, a total nucleosome concentration of 250 pM was used, containing about 25 pM single donor labeled samples and 100% acceptor labeled DNA. Samples were incubated for 90 min, after which spFRET data were taken for 15–20 min for each sample. spFRET data were analyzed by our own software which filtered the raw data and selected single molecule events from the data stream of a time-correlated-single-photon-counting board (TimeHarp200, PicoQuant). A burst was defined as a group of at least 50 photons with a mutual separation of less than 120 µs. For each burst, several parameters were recorded, including proximity ratio, burst duration, and photon intensity per time. Histograms of the proximity ratio and other parameters were built and further analyzed with IGOR Pro software (WaveMetrics).
Microplate Scanning FRET Analysis
For some samples, salt-dependent dissociation of nucleosomes was independently quantified using microplate scanning FRET (μpsFRET) as described previously . A salt titration series for each nucleosome sample was incubated in the same 384-well microplate used for spFRET; for each set of samples, three fluorescence images were taken on a variable mode scanner (Typhoon 9400, GE Healthcare) corresponding to donor channel (excitation at 488 nm, detection at 500–540 nm), acceptor channel (excitation at 532 nm, detection at 595–625 nm), and transfer channel (excitation at 488 nm, detection at 595–625 nm). Pixel resolution was 100 μm and voltages on the photomultiplier tubes were set to 620 V for the donor channel and 670 V for the transfer and acceptor channel. Images were analyzed with ImageQuant™ software and proximity ratios were calculated for each well as described below.
Estimation of FRET Efficiencies via the Proximity Ratio
Energy transfer was estimated from the sensitized emission of the acceptor upon selective donor excitation  Fluorescence was detected in two spectral windows, yielding signal intensities ID0 for the donor and IT0 for the transfer channel. Depending on the type of the experiment; these represent either the intensity within a region of an image (μpsFRET) or the number of donor and acceptor photons per single molecule burst (spFRET). Intensities were corrected for background from buffer (BD and BT), spectral crosstalk from the donor into the transfer channel (αDT), and direct excitation of the acceptor dye (fdir), yielding corrected intensities
All correction factors were determined in independent control experiments as described in Supporting Information, section S1. The proximity ratio P was calculated as a measure of energy transfer:
The P values were used as a measure for FRET efficiency, but were not converted into absolute efficiency values. This is sufficient for comparative measurements as we did here. While slight variations in the orientation factor κ2 (which is needed for converting FRET efficiencies into distances) might be caused by changing buffer conditions, we assumed that such effects would be very similar in the different samples, thus would not affect comparative measurements.
DNA Sequence Rather than Histone Origin Determines Nucleosome Stability
Nucleosome stability can be conveniently assessed by exposing nucleosomes to elevated ionic strength or reduced overall sample concentration. Analysis of spFRET histograms at equilibrium allows determination of the extent of nucleosome destabilization, assuming that a particular FRET pair can only exist in a closed state with high FRET or an open state with low FRET. Using this assumption, we normalized all data to values between 1 (fully intact nucleosomes at low salt) and 0 (fully dissociated nucleosomes at high salt). The salt titration curves were approximated by a sigmoidal function:
where F is the fraction of intact nucleosomes, c is the salt concentration, c1/2 is the concentration, and k is the slope of the curve at the transition midpoint. Nucleosome stability was quantified by the c1/2 value [27, 33].
Figure 2A shows salt titration curves of nucleosomes reconstituted from double-labeled 601 DNA and histone octamers derived from different organisms. In all samples, we observed a plateau at lower ionic strength, where the fraction of intact nucleosomes remained constant until 600 mM NaCl. Dissociation of nucleosomes is evidenced from the steep decrease in relative FRET population above 700 mM NaCl down to a second plateau. For each nucleosome sample, salt titration experiments were repeated at least three times; Table 1 lists c1/2 values averaged over all independent experiments.
Table 1. c1/2 Values (NaCl concentrations at transition midpoint) for the various histones, DNAs, and FRET pairs studied here
c1/2 values (mM)
Errors are standard errors from the fit of Eq. (3).
700 ± 20
790 ± 20
910 ± 10
620 ± 30
760 ± 30
870 ± 10
590 ± 20
690 ± 10
790 ± 20
430 ± 10
460 ± 10
560 ± 30
530 ± 20
Variation in DNA sequence has a much stronger effect on nucleosome stability than histone origin, as shown in Figure 2B. Recombinant Xenopus laevis octamers were reconstituted onto double-labeled DNA fragments centered on the 601 sequence, the 5S rDNA, and the MMTV-B sequence. In comparison with 601 nucleosomes (c1/2 = 870 ± 10 mM), we found that both MMTV-B and 5S nucleosomes were significantly less stable with measured c1/2 values of 560 ± 30 mM (5S) and 530 ± 20 mM (MMTV-B). This is a reduction by more than 30%, compared with the moderate 5–10% variation in c1/2 that was observed for the variation of histone origin. We performed supplementary μpsFRET analysis of 601, 5S, and MMTV-B nucleosomes to confirm the large variation in c1/2 observed in spFRET measurements, and we could rule out possible artifacts due to improper sample handling (see Supporting Information, section S2).
In summary, our spFRET salt titration data show that the stability of nucleosomes appears to be dominated by the underlying DNA sequence and only to a smaller extent by the organism from which the histones originate.
The Mechanism of Nucleosome Dissociation Does not Depend on DNA Sequence or Histone Origin
We next asked whether the origin of the histone proteins or the nature of the DNA sequence change the underlying mechanism of nucleosome dissociation. To do so, we followed the experimental procedure described in  where we monitored the loss of FRET between different sites of the DNA and histone H2B upon salt-induced dissociation. Again, experiments were performed in 384-well microplates with a total nucleosome concentration of 250 pM and 90 min incubation before data acquisition. Figures 3A–3C compare nucleosomes, in which octamers from yeast, mouse, or Xenopus were reconstituted onto the 601 DNA sequence. Panel (D) shows corresponding data for Xenopus octamers reconstituted onto 5S DNA. c1/2 values for all samples are summarized in Table 1.
While the individual c1/2 values show significant differences due to variations in nucleosome stability, the sequence in which the three constructs dissociate remains the same. FRET between histone H2B and the DNA close to the dyad axis decreases at the lowest NaCl concentration, followed by FRET between histone H2B and DNA that is located close to the H2A/H2B dimer. Loss in FRET between the two labels on the DNA, indicating gross unwrapping of the nucleosome, occurs at rather high ionic strength only. Based on the difference in opening between H2B and dyad-associated DNA on one hand and H2B and dimer-associated DNA on the other hand, we recently proposed the dissociation mechanism illustrated in Figure 3E, where a new intermediate conformation is proposed (shaded). In this “butterfly” conformation, the interface between the histone dimer and tetramer is broken, yet the dimer remains attached to nucleosomal DNA. Our observation that independent of histone origin and DNA sequence all nucleosome samples followed the same pattern suggests that the step-wise opening through the “butterfly state” is a general mechanism for nucleosome disassembly.
Different Nucleosomes Show Similar Conformational Changes During Dissociation
To shed further light on the structure-defining properties of histone origin and DNA sequence, we took a closer look at the sample heterogeneity observed for different salt concentrations. Figure 4 compares spFRET histograms obtained at low, intermediate, and high ionic strength for all nucleosome samples analyzed.
The general tendency is that an intermediate FRET peak at low salt concentration (P ≈ 0.35 for double labeled DNA, P ≈ 0.4 for H2B-dyad, P ≈ 0.5 for H2B-int) decreases in amplitude with increasing salt, while a peak close to P = 0 increases. Interestingly, for the 601 DNA samples (Figs. 4A–4C), we always observe a shift of the main FRET peak to slightly higher P values during the transition; a clear shoulder is seen at higher P, suggesting the formation of an intermediate state where the two labels approach each other. This effect is not observed for the 5S samples (Fig. 4D), where the initial P distribution is wider. The DNA-histone FRET pairs (Figs. 4E and 4F) show a rather uniform P peak, which decreases in amplitude with higher salt, but whose mean and width do not vary significantly.
Discussion and Summary
This work further characterizes the nucleosome opening mechanism proposed in our earlier work [26, 27], where we found that dissociation of mononucleosomes proceeds through three steps: First, the dimer-tetramer interface is destabilized and the H2A/H2B dimer moves outward, away from the center position of the DNA at the dyad axis, while still being attached to the outer DNA segment; in the next step, the dimer dissociates completely, leaving a “hexasome” or “tetrasome” structure; finally, the (H3/H4)2 tetramer dissociates, leaving the free DNA. Here, we continued these studies on nucleosomes reconstituted from Xenopus, mouse, and yeast histones on the Widom 601, 5S rDNA, and MMTV-B positioning sequences. We used salt-induced destabilization as a convenient means to create artificially such structural states as might exist during transient opening under physiological conditions. While other destabilization techniques are conceivable, such as temperature- or solvent-induced denaturation, we believe that changing the ion concentration is the least invasive of these.
As a general conclusion, we find that nucleosomes on the 601 sequence, with a 300 mM higher c1/2 value, are much more stable against salt-induced dissociation than on the 5S or MMTV-B sequences. We also see a slightly higher stability of 5S as compared with MMTV-B nucleosomes, the difference in c1/2 being 30 mM. This latter finding is qualitatively similar to the stability difference found by Kelbauskas et al.  by dilution experiments. There, the authors suggested a very pronounced stability increase of 5S over MMTV-B nucleosomes; however, their experiments were done at much lower ionic strength (10 mM Tris, no added salt, no detergent).
Mouse histones formed the most stable nucleosomes under our conditions, followed by Xenopus and yeast. It had already been shown that natively isolated nucleosomes from yeast are less stable than those from metazoa, as seen by salt dissociation analyzed on gels . Our present study suggests that this is not due to variations in post-translational modifications, but due to the differences in the histone sequence itself.
Using known strongly positioning sequences and uniform recombinant histones allowed us to generate compositionally monodisperse nucleosome preparations, so that variations in the spFRET population distribution would most probably stem from structural polydispersity of the sample. The width of the spFRET distribution then indicates the tendency of the particular nucleosome to assume structural substates. A significantly narrower spFRET peak is seen for Xenopus nucleosomes on the 601 sequence at low salt (which has been in vitro-selected for strong positioning) than for the 5S and MMTV-B sequences, of natural origin. A similar behavior was found earlier for end-labeled DNA fragments . We may conclude that the positioning power of the 601 sequence narrows down the accessible conformational space relative to the other two DNAs.
For all DNA and histone combinations studied, the general opening mechanism through the butterfly state could be confirmed. Other than postulated earlier , our data strongly suggests that nucleosome opening precedes histone dimer dissociation.
We also note that earlier spFRET studies on nucleosomes free in solution [24, 34, 35] showed disruption at much lower salt concentrations than presented here, i.e., in the range of 50–100 mM NaCl. Although a systematic study of the solution conditions that contribute to sample stability at the low concentrations necessary for spFRET is still overdue, we note here that the sample conditions in those earlier studies were different: lower nucleosome concentration, untreated sample chamber surfaces, and the absence of detergents could be responsible for the lower stability.
Concluding we can say that spFRET of mononucleosomes labeled at various donor/acceptor positions seems to be a reliable method to characterize the stability and conformational variability of chromatin from different origins. In particular, double-labeled DNAs lend themselves particularly well for systematic studies because they can be prepared rather easily and do not pose the problem of labeling ambiguity, as in the case of a histone-DNA FRET pair.
We thank Karolin Luger and her lab for generously providing some of the recombinant histones, and for highly fruitful and inspiring discussions. J.L. thanks Alexei Onufriev for suggesting the term “butterfly state” during a session at the Biophysical Society Meeting 2013.