Sequence-Dependent Variations Associated with H2A/H2B Depletion of Nucleosomes

DNA and nucleosome preparation

Fluorescently labeled dsDNA fragments (∼160 bp in length) were made by polymerase chain reaction techniques (43), using various templates and the appropriate, labeled primers, as described in detail previously (42). The GAL10 fragment (206–365 bp from an EcoRI site (cf. (44)) corresponds to the in vivo position of a TATA-containing nucleosome. The MMTV-B fragment (−70 to −230 bp on the promoter (51)) corresponds to the in vivo position of nucleosome B (47). The sea urchin 5S rDNA fragment is the 5′ EcoRI-Ban II region (49).

Labeled DNA is gel purified then reconstituted into nucleosomes using purified HeLa histone octamers as described previously (42,52). The histone octamers were prepared as described in Yodh et al. (53) and were a generous gift from Dr. J. Yodh. For some studies, the mononucleosome band was eluted from (unstained) gels by excising a gel slice containing the band and placing the slice in TE buffer pH 8 at room temperature for 12–24 h.

Nucleosome response to yNAP-1

Nucleosomes (10 nM concentration) were treated with yNAP-1 protein (10 nM monomer concentration) in TE buffer (10 mM Tris and 1 mM EDTA, pH 8), containing 1 mM dithiothreitol. FRET efficiency was measured (see below) immediately after the addition of the yNAP-1 to the sample (time “0”) and at 15–30 min intervals afterward. The yNAP-1 was a generous gift from Y. Park and K. Luger; it functions as a dimer under these conditions ((19); K. Luger, Dept. of Chemistry and Biochemistry, Colorado State University, personal communication, 2007).

Single-molecule distributions and salt dependence

Single-molecule FRET measurements were conducted using an experimental setup described in Kelbauskas et al. (42). Briefly, a microscope (ECLIPSE TE2000-U, Nikon, Melville, NY) operated in a confocal configuration and equipped with an oil immersion objective lens (100×, numerical aperture = 1.4) was utilized. The excitation source was a continuous wave frequency-doubled Neodymium-doped yttrium vanadate (Nd:YVO₄) laser (Millenia Xs, Coherent, Santa Clara, CA) operating at 532 nm. The photons in Cy3 and Cy5 spectral emission channels were detected with silicon avalanche photodiodes (SPCM-AQR-12, Perkin Elmer, Fremont, CA) using appropriate emission filters (BP570/40 for Cy3 and BP670/40 for Cy5 detection channel; Chroma Technology, Rockingham, VT) placed in front of each detector. FRET efficiency was calculated using the following equation:

(1)

where I_A and I_D are the fluorescence intensities measured in the acceptor and donor channel, respectively, and γ = 1.12 is a factor correcting for the cross talk between the detection channels and any contribution from direct excitation of the acceptor. E_FRET errors were calculated as the mean ± SE using where σ is the standard deviation and N is the number of independent measurements.

The salt-dependence measurements were carried out in TE buffer, pH 8, at increasing concentrations of NaCl, achieved by stepwise addition of 1 μl aliquots of 500 mM NaCl to the sample. After each addition, 5–10 min were allowed for equilibration before measurement.

Fluorescence correlation spectroscopy measurements

FCS measurements were carried out using the experimental setup as described above and samples diluted to ∼200 pM. Correlation curves were measured using a hardware dual-channel digital correlator card with a sampling time of 12.5 ns (Flex2k-12x2; Correlator, Bridgewater, NJ) and the vendor's software. Data analysis and fitting were performed using locally written software based on LabView.

Calculation of conformational dynamics parameters

For molecules labeled with a FRET pair-like Cy3 and Cy5, the emission intensity fluctuations in the donor (Cy3) or acceptor (Cy5) detection channel can be caused by two different kinds of processes (assuming that there are no photophysical effects, like intersystem crossing to the triplet state, and that photobleaching can be neglected): fluctuations due to the diffusion of molecules through the excitation volume; and fluctuations due to conformational transitions that result in changes in the energy transfer efficiency. Applying the formalism described in Magde et al. (54), Berne and Pecora (55), Krichevsky and Bonnet (56), and Bonnet et al. (57) adapted to our specific model system (see Supplementary Material), we were able to determine conformational dynamics parameters of particles. As required, our system exhibits two-state behavior (see below).

Treating nucleosomes with yNAP-1

To obtain an independent test of the possible relationship between H2A/H2B release and FRET response differences among these three types of nucleosomes, we treated all three with the yeast histone chaperone yNAP-1, which has been shown to efficiently remove H2A/H2B from nucleosomes in vitro (19). Substoichiometric concentrations of yNAP-1 (0.5 mol of yNAP-1 dimer per mole of nucleosomes) were used to minimize nonspecific effects, and nucleosome concentrations (∼10 nM) were well above those at which we observe concentration dependent FRET decreases (42). Thus, the substrates for yNAP-1 should be intact nucleosomes. FRET was monitored for up to 3 h at room temperature.

Fig. 2 a shows typical response curves for the three types of treated nucleosome samples. FRET efficiencies from 5S samples (solid squares, solid line) are constant, but GAL10 (solid triangles, dashed line) and MMTV-B (solid circles, dotted line) samples show a time-dependent decrease in FRET efficiency. After ∼3 h of incubation, the FRET efficiency in MMTV-B and GAL10 samples has decreased by 10%–15%, whereas that from 5S nucleosomes remains essentially unchanged. In the absence of yNAP-1, the FRET efficiency in such incubated samples remains constant for all three types of samples (cf. GAL10 nucleosomes, open inverted triangles). Thus, yNAP-1 treatment reduces the FRET efficiency of MMTV-B and GAL10 but not 5S nucleosomes. The precise level of the FRET decrease for MMTV-B and GAL10 samples varies by 3%–5% (mean ± SE of two to three experiments; see Materials and Methods), as might be expected under these relatively dilute conditions; but the clear qualitative behavioral difference, decreasing FRET for yNAP-treated MMTV-B or GAL10 but not 5S nucleosome samples, is a constant feature.

To check whether yNAP-1 was actually acting on the 5S nucleosomes, we carried out salt titrations of 5S samples that had been treated as above with yNAP-1. H2A/H2B absence makes the nucleosome more salt sensitive (N. Chan and L. Kelbauska, unpublished observations). Therefore, the salt stability of yNAP-1-treated 5S nucleosomes should be lower than the stability of untreated 5S nucleosomes if yNAP-1 treatment has caused H2A/H2B release. This is the case (Fig. 2 b). The yNAP-1-treated 5S nucleosome samples (open circles, dashed line) are more salt sensitive, i.e., their structure is disrupted (the FRET efficiency decreases) at lower NaCl concentrations than for untreated 5S samples (solid squares, solid line). Therefore, yNAP-1 is acting on 5S nucleosomes even though their FRET efficiencies do not change and that action produces changes (lower salt stabilities) that are consistent with H2A/H2B removal. yNAP-1 treatment also decreases the salt stabilities of GAL10 (Fig. 2 c) and MMTV-B (data not shown) nucleosomes to very similar levels as that of yNAP-treated 5S nucleosomes (cf. Fig. 2 c, “x” curve versus dashed line). Thus, yNAP-1 appears to be acting more or less equally, i.e., depleting H2A/H2B to similar extents, on all three types of nucleosomes.

It is important to note that these three types of nucleosomes have salt stabilities that are much higher at bulk concentrations (42) and similar to the salt stabilities shown for other types of nucleosomes under bulk conditions (22,60). However, stability is lower at these (10 nM) nucleosome concentrations. Thus, lower salt concentrations can produce destabilization (and FRET changes) in these experiments than at bulk nucleosome concentrations. Importantly, the relative stability differences noted between these three, 5S > GAL10 > MMTV-B, are the same at bulk (42) or these lower concentrations (compare “untreated” nucleosomes, Fig. 2, b and c, data not shown). Also, controls to test the effects of salt on fluorescence properties of Cy3 and Cy5 dyes (using singly labeled nucleosomes) have demonstrated that differential salt effects on FRET efficiency are negligible in this system (42).

We also carried out treatments with yNAP-1 present in excess, to try to estimate the extent of FRET loss achievable from MMTV-B and GAL10 nucleosomes. The average (multiple experiments) FRET decreases produced by high yNAP-1 concentrations (2–4 molar yNAP-1 excess) are roughly similar to the decreases observed when nucleosomes are diluted to sub-nM concentrations (data not shown): ∼10% for MMTV-B nucleosomes by dilution or by NAP-1 treatment; and ∼30% (dilution) versus ∼20% (yNAP-1 treatment) for GAL10 nucleosomes. Quantitative similarities are consistent with both treatments causing the same type of change, namely H2A/H2B loss from nucleosomes. In sum, these yNAP-1 experiments provide independent and compelling evidence that effects associated with H2A/H2B release produce differing FRET responses in 5S compared to the two types of promoter nucleosomes, thus strengthening the similar conclusion from nucleosome dilution studies. The results also demonstrate directly that 5S nucleosomes show no FRET change when H2A/H2B depleted.

Single-molecule distributions

Some of our previous studies (42) were carried out at sub-nM concentrations. Based on prior results (50), the yNAP-1 results above, and results presented below, nucleosomes are likely to be H2A/H2B depleted at these concentrations and will thus be referred to as “particles”. H2A/H2B-depleted nucleosomes (particles) are of intrinsic interest (see Introduction), so we turned to single-molecule techniques to compare the properties of the particles produced by dilution of 5S, MMTV-B, and GAL10 nucleosomes.

Single-molecule population distributions can provide insights into specific features that are hidden in ensemble average results. To eliminate contributions from free DNA and higher molecular weight material that may be present in reconstituted samples (42), the samples were run on gels and the mononucleosome band was eluted (see Materials and Methods). The eluted mononucleosomes were diluted to 0.2 nM concentrations, and single-molecule counting was carried out in solution (Fig. 3). The features described below are also characteristic of single-molecule distributions from noneluted samples (data not shown) and therefore do not result from effects associated with elution. For the sake of discussion, the distributions will be divided into three regions: high (0.8–1.0), low (0–0.2), and midrange (0.2–0.8) FRET efficiency. High efficiency molecules maintain their fluorophores close enough for major energy transfer, whereas in low efficiency molecules, the fluorophores are too far apart for significant energy transfer. Molecules in the midrange of the distribution fall between these two extremes. Based on the R_o value, the distance at which energy transfer is 50% (5–6 nm for Cy3/Cy5 probes, Amersham Biosciences, Buckinghamshire, UK), FRET efficiencies of 0.8–1.0 (high efficiency molecules) or 0–0.2 (low efficiency molecules) would correspond to average probe separations of <4 nm or >7 nm.

Distributions for 5S particles contain a significant population of molecules with efficiencies ≥0.8 (Fig. 3, top panel, solid bars), whereas GAL10 (middle panel, solid bars) and MMTV-B (lower panel, solid bars) distributions show fewer such molecules. Thus, the relative fraction of particles that maintain their FRET labels very close to each other, on average, is larger in 5S than in MMTV-B or GAL10 populations. These population differences are consistent with the higher average FRET efficiencies noted for 5S compared to MMTV-B or GAL10 particles (sub-nM concentrations) or nucleosomes (bulk concentrations) (42).

All three distributions contain significant numbers of molecules with low FRET efficiencies (0–0.2) but the relative fraction of these molecules is higher in the MMTV-B and GAL10 distributions (Fig. 3, solid bars). None of the three distributions shows a major population at any FRET efficiency in the midrange (0.2–0.8). In all cases, the individual frequency values, i.e., the frequency of molecules with a given efficiency, are much lower in this midrange of the distribution than at very high (0.8–1.0) or very low (0– 0.2) FRET efficiencies. A distribution in which molecules tend to have either very low or very high FRET efficiencies suggests that molecules reside mainly in one of two conformational states: a state in which DNA associates with the particle such that the fluorophores are, on average, close enough to produce strong energy transfer (0.8–1.0); and a state in which the fluorophores are, on average, too far apart for significant energy transfer to occur, which results in low (0–0.2) FRET efficiency. Each of these states must be stable for time periods that are long in comparison to the average time it takes a molecule to diffuse across the excitation volume (see also below).

To confirm that the high FRET efficiency molecules reflect energy transfer occurring in structured complexes, single-molecule distributions were obtained for the same samples in 200 mM NaCl, which under these conditions is sufficient to destabilize the particles and abolish energy transfer (see below). This treatment completely removes the high (and intermediate) FRET efficiency molecules from all three distributions, leaving only low efficiency molecules (Fig. 3, all panels, hatched bars). Thus, the high efficiency molecules observed in the low salt distributions (Fig. 3, solid bars) must reflect stable, structured complexes in which Cy3 and Cy5 are on average close enough for very efficient energy transfer. The loss in the intermediate FRET efficiency ranges shows that these low-abundance molecules also are structured complexes. Their lower FRET efficiencies could reflect a less well-folded and/or a more dynamic structure. Intermediate efficiencies could also result from complexes that happened to change conformation (high to low FRET or vice versa) while passing through the beam, although our FCS results (see below) suggest that the probability of such an event is <1–2 per 100 molecules.

There is some sample-to-sample variation in these single-molecule distributions, but the major features, and the differences between 5S and MMTV-B or GAL10 samples, are consistently observed. For example, the fraction of high efficiency molecules (0.8–1.0) in MMTV-B or GAL10 samples never reaches 0.4, whereas the fraction in 5S samples is never below 0.5. Also, frequency values in the midrange of the distribution are always small relative to the values for low and high efficiency molecules, for all three types of samples.

Salt titrations at the single-molecule level

To study the detailed response of these three types of particles to salt treatment at the single-molecule level, single-molecule population distributions were obtained at various NaCl concentrations (cf. Fig. 4 a). Again, we note that destabilization and FRET changes occur at lower salt concentrations in particles at these sub-nM concentrations (L. Kelbauskas and N. Chan, unpublished observations) than under bulk conditions (42).

As NaCl concentration increases, the relative numbers of molecules with high (≥0.8) FRET efficiencies progressively decrease and the numbers of molecules with very low FRET efficiencies progressively increase. At NaCl concentrations >100 mM, the very high efficiency molecules are largely gone and very low efficiency molecules strongly dominate the distribution. Most importantly, there is no significant accumulation of species in the midrange, from 0.2–0.8 FRET efficiency, at any salt concentration; the frequency values in this range remain low throughout the salt titration except for a slight increase in the relative fraction of molecules in the 0.2–0.3 efficiency range at the higher NaCl concentrations. MMTV-B and GAL10 distributions undergo similar profile changes with increasing salt concentrations; high efficiency molecules disappear, low efficiency molecules accumulate, including an enhancement of molecules in the 0.2–0.3 FRET efficiency range at the higher salt concentrations, and there is never a significant accumulation of species in the midrange (data not shown).

The relative fractional loss of high FRET efficiency (0.8–1.0) molecules from the 5S distribution between 0 and 200 mM NaCl is similar to the relative fractional gain of low efficiency molecules (0–0.2) over this range of salt, 0.56 vs. 0.61 (Table 1). The small difference undoubtedly reflects the conversion of molecules from intermediate to low efficiencies as the former are disrupted by salt. A decrease in high efficiency molecules, a parallel increase in low efficiency molecules, and no significant accumulation of molecules with intermediate FRET efficiencies is again indicative of a two-state response; the 5S particle exists mainly in either a folded (high FRET) or unfolded (low FRET) state and equilibrium is shifted toward the low FRET state with increasing NaCl concentration. Qualitatively similar results are obtained with MMTV-B and GAL10 samples (Table 1).

The two-state behavior observed for the 5S particles (sub-nM concentrations) is more consistent with the behavior expected for H2A/H2B-depleted nucleosomes than for intact nucleosomes. The nucleosome has a tripartite organization, two H2A/H2B dimers, and one H3/H4 tetramer (1,2) and is known to dissociate in stages rather than in an all-or-none, two-state process. For example, salt addition causes the preferential release of H2A/H2B from nucleosomes; H3/H4 tetramer-DNA dissociation occurs at distinctly higher salt concentrations (cf. (22)). Thus, if the 5S complexes at these sub-nM concentrations were intact nucleosomes, they would first lose H2A/H2B as salt is added. Based on the FRET response seen with MMTV-B and GAL10 nucleosomes when H2A/H2B are released (by treatment with yNAP-1 or dilution to sub-nM concentrations), this loss should produce an initial partial FRET efficiency decrease for the 5S complexes. A partial FRET decrease in a substantial proportion of molecules ought to be reflected in the 5S single-molecule distributions. Specifically, the high FRET efficiency peak should shift to lower but still significant efficiency values as H2A/H2B are released and then fall to very low values as the depleted particles are completely disrupted by salt. However, there is no evidence of such a peak shift in the 5S data. Thus, this two-state behavior also suggests that 5S complexes are H2A/H2B-depleted at these sub-nM concentrations.

We do see evidence for distinct subpopulations of molecules in other states but only under particular conditions. For example, in the salt studies above, the relative fractional increase of molecules in the 0.2–0.3 FRET efficiency range at higher NaCl concentrations (cf. Fig. 4 a) suggests that a new state (structural or dynamic) might be available at these salt concentrations for all three types of particles. However, it remains a fairly minor component in the total population. As another example, raising the temperature to ∼40°C causes a reversible, 10%–15% FRET efficiency increase in 5S (but not MMTV-B or GAl10) particles at sub-nM concentrations (42), but these particles have a lower salt stability (data not shown). This implies that at ∼40°C, 5S particles are in a state with a lower average donor-acceptor separation, leading to higher FRET, but a lower salt stability than at room temperature.

Particles with FRET efficiencies from 0.9 to 1.0 represent the molecules in each population that show the most efficient energy transfer, i.e., the particles that are able to keep their fluorophores closest to one another, statically and dynamically. How does the salt stability of this specific class of molecules compare for the three types? The data in Fig. 4 b show that these very high FRET GAL10 and 5S particles have similar salt stabilities, whereas MMTV-B very high FRET particles are clearly less salt stable. Thus, 5S populations contain relatively more of these very high FRET particles than GAL10 populations (Fig. 3), but the two types of particles have similar structural stabilities (Fig. 4 b), at least as measured by salt sensitivity. On the other hand, 5S and MMTV-B differ both in the relative numbers of these molecules in the population and in their salt sensitivities. These results suggest that there may be a diversity of sequence-dependent variations in individual nucleosome properties, detectable with a sensitive approach such as this one. Interestingly, the low salt stability demonstrated by these very high FRET MMTV-B particles (Fig. 4 b) is also observed in salt titrations of MMTV-B nucleosomes (42). This suggests that this feature may be implemented largely via the H3/H4 tetramer.

Nucleosome dynamics