Histone variant H2A.B-H2B dimers are spontaneously exchanged with canonical H2A-H2B in the nucleosome

H2A.B is an evolutionarily distant histone H2A variant that accumulates on DNA repair sites, DNA replication sites, and actively transcribing regions in genomes. In cells, H2A.B exchanges rapidly in chromatin, but the mechanism has remained enigmatic. In the present study, we found that the H2A.B-H2B dimer incorporated within the nucleosome exchanges with the canonical H2A-H2B dimer without assistance from additional factors, such as histone chaperones and nucleosome remodelers. High-speed atomic force microscopy revealed that the H2A.B nucleosome, but not the canonical H2A nucleosome, transiently forms an intermediate “open conformation”, in which two H2A.B-H2B dimers may be detached from the H3-H4 tetramer and bind to the DNA regions near the entry/exit sites. Mutational analyses revealed that the H2A.B C-terminal region is responsible for the adoption of the open conformation and the H2A.B-H2B exchange in the nucleosome. These findings provide mechanistic insights into the histone exchange of the H2A.B nucleosome. Hirano et al. show that the H2A.B-H2B histone dimer incorporated within the nucleosome exchanges with the canonical H2A-H2B dimer without assistance from histone chaperones or nucleosome remodelers. This study provides insights into how the H2A.B nucleosome exchanges its histones.

I n eukaryotes, chromatin compacts genomic DNA for accommodation within the nucleus. The basic structural unit of chromatin is the nucleosome, which is composed of the nucleosome core particle (NCP) and linker DNAs. In the NCP, two histone H2A-H2B dimers associate with one histone H3-H4 tetramer, forming the histone octamer, and 145-147 base pairs of DNA are tightly bound to its basic surface 1,2 . Histones are stably incorporated into chromatin with a very slow exchange rate in living cells 3 , indicating that the NCP exists as a stable architecture. Therefore, the machineries managing genomic DNA functions, such as transcription, replication, repair, and recombination, must work on the DNA tightly wrapped within the NCP [4][5][6] . To relieve this NCP barrier in the genome, NCPs have versatile structures and dynamics [7][8][9][10][11] .
Z.2, the overall NCP structures are equivalent to those of the canonical NCPs, but the local structures around the H2A.Z L1-loop regions are substantially different 21,22 .
H2A.B (formerly named H2A.Bbd in human) is an evolutionarily distant histone H2A variant that conserves about 50% amino acid identity, as compared to the canonical H2A, and has short C-terminal tail 28 (Fig. 1a). In mammalian cells, H2A.B reportedly accumulates on the transcription start sites [29][30][31] and/ or gene body regions of transcribing genes [31][32][33][34] . H2A.B also transiently assembles at DNA repair and replication sites 35,36 . These findings suggest that H2A.B may briefly form an NCP and protect the genomic DNA from endogenous and exogenous attacks, such as ionizing radiation or nucleases, when chromatin is re-configured after DNA transcription, repair, and replication. Indeed, H2A.B forms the NCP structure with flexible DNA ends in vitro 35,[37][38][39][40][41] . The presence of flexible DNA ends in the H2A.B NCP is also supported by a molecular dynamics simulation study 42 . However, the biochemical properties of the H2A.B NCP have remained poorly understood.
In the present study, we found that the nucleosomal H2A.B-H2B dimer efficiently exchanges with the canonical H2A-  NCP structure. Interestingly, the H2A.B-specific C-terminal tail segment is important for the adoption of the open conformation and the H2A.B-H2B exchange in the NCP. These findings provide important insights into understanding the unusual behavior and function of H2A.B in cells.

Results
Nucleosomal H2A.B exchanges with canonical H2A without additional factors. H2A.B is a rapid-exchanging histone variant in cells 35,43 . To study the structural features of the H2A.B NCP, we reconstituted the NCP with recombinant human histones, H2A.B, H2B, H3, and H4, and obtained its crystals. However, the crystals of the H2A.B NCP generated poor X-ray diffraction data, probably due to the flexible nature of its DNA ends 35,[37][38][39][40][41] . We previously reported that the crystal of the heterotypic NCP with the histone variant, CENP-A, which forms an NCP with flexible DNA ends, diffracted better than that of the homotypic NCP 19,44 . This fact led us to prepare the heterotypic NCP containing one each of H2A.B and H2A in the NCP, to improve the quality of the crystals (Fig. 1b, c). We then performed the X-ray crystallographic analysis (Fig. 1d). Surprisingly, the putative H2A.B/ H2A heterotypic NCP lacked H2A.B and was formed with two canonical H2As (Supplementary Fig. 1 and Table 1), although the NCP sample before crystallization actually formed the heterotypic NCP containing both the H2A.B and H2A proteins (Fig. 1e). This implied that, during the crystallization processes, the nucleosomal H2A.B may be exchanged with the canonical H2A without assistance from additional factors.
Nucleosomal H2A.B-H2B exchanges with free H2A-H2B in solution. We then performed the histone exchange assay to determine whether the nucleosomal H2A.B is actually exchanged with the canonical H2A, without additional factors. In this assay, the H2A.B NCP or the canonical NCP (H2A NCP) was incubated with the purified H2A-H2B dimer, and the resulting NCPs were analyzed by native polyacrylamide gel electrophoresis (PAGE). To confirm the incorporation of the H2A-H2B dimer exogenously added to the NCP, we used a fluorescently labeled H2A-H2B dimer (H2A-H2B fluo ) as the exogenously added H2A-H2B dimer ( Fig. 2a and Supplementary Fig. 2). Since the H2A.B NCP migrates more slowly in 6% native PAGE as compared to the canonical H2A NCP, the H2A-H2B fluo dimer exchange can be detected by the migration change of the resulting NCPs, in addition to the fluorescence signal ( Fig. 2b, Supplementary Fig. 3). As shown in Fig. 2b-d (lanes 5-8), the H2A-H2B fluo dimer was substantially incorporated into the NCP, when the H2A.B NCP was incubated with the H2A-H2B fluo dimer. In contrast, only a trace amount of the H2A-H2B fluo dimer was incorporated into the H2A NCP by spontaneous histone exchange, when the H2A-H2B fluo dimer was added to the reaction mixture ( Fig. 2b-d, lanes 1-4). These results indicated that the H2A.B-H2B dimer in the NCP spontaneously exchanges with the canonical H2A-H2B dimer.
The H2A.B NCP forms an open conformation. To understand the mechanism of H2A.B-H2B dimer exchange in the H2A.B NCP, we performed a high-speed atomic force microscopy (HS-AFM) analysis 45 . This method allows the visualization of the dynamic structural transition of the nucleosome [46][47][48] . We found that 49.4% of the H2A.B NCP existed as the open conformation, in which two small histone complexes (probably H2A.B-H2B dimers) are bound to the DNA and detached from the large histone complex (probably H3-H4 tetramer) at the initial stage of the HS-AFM analysis (Fig. 3a, c). In contrast, in the canonical H2A NCP, only a small proportion (5.3%) of the open conformation was observed at the initial stage (Fig. 3b, d).
We selected NCPs with about 6 nm heights and monitored the NCP dissociation induced by scratching with the HS-AFM probe (Fig. 4). We found that the putative H2A.B-H2B dimers bound to the detached DNA region were continuously observed in the H2A.B NCP (Fig. 4a). This suggested that the H2A.B NCP was transformed into the open conformation ( Fig. 4a and Supplementary movie S1). In contrast, in the H2A NCP, the putative H2A-H2B dimers were rapidly released from the DNA, and an obvious open conformation of the H2A NCP was rarely observed when the H2A-H2B dimer was released ( Fig. 4b and Supplementary movie S2). The dwelling time of the H2A.B-H2B dimer on the nucleosomal DNA was quite long (average 24.6 s), as compared to that of the canonical H2A-H2B dimer (average 4.4 s), during the NCP disruption process ( Supplementary Fig. 4). These results suggested that the H2A.B NCP, but not the H2A NCP, may dynamically adopt the open conformation.
The H2A.B C-terminal region is responsible for the open conformation adoption and the H2A-H2B exchange activity. The H2A.B variant is smaller than the canonical H2A, because of its shorter C-terminal tail region (Fig. 1a). In addition, the C-terminal amino acid sequence of H2A.B is not conserved among the H2A variants 28 (Fig. 1a). We hypothesized that the Cterminal region of H2A.B may play a role in its specific characteristics, such as the adoption of the open conformation. To test this hypothesis, we prepared the canonical H2A mutant, H2A H2A. B(102-114) , in which the C-terminal region (amino acid residues 98-129) of canonical H2A is replaced with the corresponding H2A.B C-terminal region (amino acid residues 102-114) (Figs. 5a and 1a). The NCP containing the H2A H2A.B(102-114) mutant was reconstituted ( Supplementary Fig. 5). Consistent with our  We next performed a small-angle X-ray scattering (SAXS) analysis. In this method, the apparent NCP volume can be evaluated as the radius of gyration (Rg). As shown in Table 2, the Rg value of the H2A.B NCP was 56.5 ± 0.5 Å, which is substantially larger than that of the canonical NCP (44.9 ± 1.0 Å). Intriguingly, the Rg value of the H2A H2A.B(102-114) NCP was 52.9 ± 0.4 Å, which is also substantially larger than that of the canonical NCP (Table 2). These results supported the idea that the C-terminal region of H2A.B plays a role in forming the open conformation of the NCP in solution.
The detachment of the H2A.B-H2B dimers from the H3-H4 tetramer in the NCP may be important for generating the open conformation. The cryo-EM structure of the H2A.B NCP demonstrated that the H2A.B-H2B dimers associate with the H3-H4 tetramer in the NCP 41 . However, our gel filtration chromatography experiments revealed that, in the absence of DNA, the H2A.B-H2B dimer eluted separately from the H3-H4 tetramer, and did not form a histone octamer under conditions with 2 M NaCl (Fig. 5d, Supplementary Fig. 7a). In contrast, the canonical H2A-H2B dimers associate with an H3-H4 tetramer and form a histone octamer under the same experimental conditions (Fig. 5e, Supplementary Fig. 7b). These differences indicate that the association of the H2A.B-H2B dimer with the H3-H4 tetramer is weaker than that of the H2A-H2B dimer with the H3-H4 tetramer. This is perfectly consistent with the previous reports 37,41 . Interestingly, the H2A H2A.B(102-114) -H2B dimer, like the H2A.B-H2B dimer, eluted separately from the H3-H4 tetramer (Fig. 5f, Supplementary Fig. 7c). These results supported the hypothesis that the weak association between the H2A.B-H2B dimer and the H3-H4 tetramer in the NCP is mediated by the H2A.B C-terminal region, and may be required for the formation of the open conformation of the NCP.
We finally tested whether the H2A.B C-terminal region functions in the H2A-H2B exchange. As expected, the exchange rates of the H2A H2A.B(102-114) -H2B dimers and the wild-type H2A.B-H2B dimers in the NCPs with the H2A-H2B dimers were similar (Fig. 6a-c, Supplementary Fig. 8). Therefore, we concluded that the H2A.B C-terminal region may enhance the H2A-H2B exchange in the NCP, through the adoption of the open conformation (Fig. 6d model).
Previous studies demonstrated that H2A.B is incorporated into chromatin at replication and repair sites, and rapidly exchanged within several minutes in cells 35,43 . The exchangeable H2A.B variant may be utilized as a histone that temporarily protects the naked DNA regions emerging during DNA replication and repair. This rapid removal of the nucleosomal H2A.B may be mediated by the spontaneous H2A. B-H2B exchange with the canonical H2A-H2B. H2A.B also accumulates around the transcription start sites and/or gene bodies of transcribed genes [29][30][31][32][33][34] . The NCP containing the H2A. B-H2B dimers has a tendency to adopt the open conformation (Figs. 3 and 4), in which the nucleosomal DNA may become more accessible to DNA-interacting factors, such as transcription factors and RNA polymerase 54,55 . It will be intriguing  to study the transcription efficiency on the nucleosome containing H2A.B.
H2A.B reportedly exists in spermatogenic cells and sperm 56 . During spermatogenesis, the chromatin architecture drastically changes and most of the nucleosomes are replaced by protamines in sperm 57 . The H2A.B-H2B exchange activity may be important to promote transitions of the chromatin architecture in the testis. A mouse histone H2A variant, H2A.L.2, reportedly functions in the histone replacement process by transition proteins and protamines 26 . Although mouse H2A.L.2 lacks a human homolog, like human H2A.B, it contains a shortened C-terminal region 58 . We determined that the H2A.B C-terminal region is responsible for the H2A.B-H2B exchange activity via the adoption of the open conformation (Figs. 5 and 6). In addition, both mouse H2A.L2 and human H2A.B are retained in sperm 56,57,59-61 . These similarities suggest that, in humans, H2A.B may serve as a counterpart to mouse H2A.L.2, which plays an essential role in spermatogenesis 26 .
In the present study, we found that the H2A.B NCP forms an open conformation, which may be an intermediate structure for the H2A.B-H2B exchange in the NCP (Figs. 3 and 4). In light of this finding, we propose a model for the nucleosomal H2A.B-H2B exchange (Fig. 6d). In this model, the H2A.B NCP (closed conformation) dynamically adopts the open conformation, which may facilitate access to the H3-H4 tetramer (Figs. 6d, (1) to (2)). A conformational change between the closed and open conformations may occur, because the association of the H2A.B-H2B dimer with the H3-H4 tetramer is substantially weaker, as compared to that between the H2A-H2B dimer and the H3-H4 tetramer 37,41 (Fig. 5). This is consistent with the fact that the H2A docking domain (mapped to its C-terminal region), which interacts with the H3-H4 tetramer in the histone octamer, is not conserved in H2A.B 37,39 . In the canonical NCP, the H2A Cterminal residues (P109 and I111), which are not conserved in H2A.B, directly interact with the H3 residues (L48, I51, and Q55). In the open conformation, the H2A-H2B dimers may bind to the H3-H4 tetramer, because the H3-H4 surface becomes accessible in this conformation (Figs. 6d, (3)), and the H2A.B-H2B dimers could be evicted from the nucleosomal DNA (Figs. 6d, (4)). This model is consistent with the previous mutational analysis, in which the mutations of the histone H3 l51 and Q55 residues, located on the binding surface with the H2A C-terminal region in the NCP, enhanced the H2A-H2B exchange rate 62 .
The nucleosome has a stable architecture that often becomes an obstacle for gene functions, such as transcription, DNA replication, DNA recombination, and DNA repair. This negative effect of the nucleosome is utilized to regulate the genomic DNA functions, and may play a central role in the epigenetic regulation of genes in eukaryotes. Histone variants are considered to provide versatility in the nucleosome structures and physical properties, and play important roles in the epigenetic regulation of the genome [6][7][8][9][10][11] . The nucleosomal H2A-H2B exchange revealed in the present study is a quite unique characteristic specific for the H2A.  H3.1, H4, H2A.B, and H2A H2A.B(102-114) , were produced as  Supplementary Fig. 7. The uncropped gel images are shown in Supplementary Fig. 9. recombinant proteins in Escherichia coli cells, and purified by the methods described previously 35,63 . Briefly, proteins were produced in E. coli BL21(DE3) cells as hexa-histidine (His 6 )-tagged proteins. The His 6 -tagged proteins were purified by Ni-NTA agarose (QIAGEN) chromatography under denaturing conditions. Except for H2A.B and H2A H2A.B(102-114) , the His 6 -tag was removed by cleavage with thrombin protease, and the histone proteins were further purified by MonoS cation exchange column chromatography. The His 6 -tags of H2A.B and H2A H2A.B(102-114) were removed after the H2A.B-H2B or H2A H2A.B(102-114) -H2B dimer formation, as described below. The resulting proteins were desalted and lyophilized.
Purification of the NCPs. The NCPs were prepared by the salt dialysis method with the palindromic 146 base-pair α-satellite DNA fragment, as described previously 1,65 . The DNA fragment containing one half of the α-satellite DNA fragment in pGEM-T Easy vector was amplified in the E. coli strain DH5α, and was excised from the plasmid DNA by EcoR V (Takara). The DNA fragment was then dephosphorylated by alkaline phosphatase (Takara), and was further cleaved by EcoR I. The DNA fragment was purified by DEAE-5PW anion-exchange column chromatography (TOSOH). The DNA fragment was self-ligated by T4 DNA ligase (NIPPON GENE), and the resulting DNA fragment was further purified by DEAE-5PW anion-exchange column chromatography (TOSOH).
Determination of the crystal structure. The X-ray diffraction data were collected at the beamline BL41XU (wavelength: 1.00000 Å) at SPring-8 (Harima, Japan). The diffraction data were scaled and processed using the HKL2000 program 66 . To prepare the search model for molecular replacement, the H2A atomic coordinates were removed from the human NCP structure (PDB ID: 5Y0C) 67 . The molecular replacement was performed with the PHASER program 68 . The atomic coordinates were refined using the PHENIX and Coot programs 69,70 .
HS-AFM observations. HS-AFM images of NCPs were obtained with our laboratory-build microscope, as described previously 71 . Briefly, HS-AFM was performed in the tapping mode. Deflections of the cantilever were detected by a two-segmented PIN photodiode, using an infrared laser (0.8 mW, 780 nm) focused through a ×60 objective lens (Nikon, CFI S Plan Fluor ELWD 60x) onto the back side of a cantilever (Olympus, BL-AC10DS-A2) covered with a gold coating. The free oscillation amplitude of the cantilever was~1 nm, and the set-point amplitude was~90% of the free amplitude for feedback control of HS-AFM. An amorphous carbon tip grown by electron beam deposition (EBD) was used as the AFM probe. For HS-AFM observations of NCPs, a mica surface was treated for 5 min with 50 μg/mL poly-L-lysine (mol. wt. 1000~5000, Sigma-Aldrich). HS-AFM observations were performed at room temperature (~25˚C), in a buffer consisting of 20 mM Tris-HCl (pH 7.5), 100 mM KCl, and 0.03% NP-40. HS-AFM observations of each NCP were performed at least three times to confirm the reproducibility. The imaging rate of HS-AFM is 2.5 frames per second for 100 × 80 nm 2 . HS-AFM images were collected using Igor Pro Ver. 8.0.4.2. (WaveMetrics). The images were analyzed using Igor Pro Ver. 8.0.4.2. (WaveMetrics) and ImageJ. SAXS analysis. SAXS was performed with a NANOPIX instrument (RIGAKU) at the Institute of Radiation and Nuclear Science, Kyoto University. To cover the wide qrange, we measured the sample with two sample-to-detector positions: 1330 mm for 0.007 -0.03 Å −1 , and 300 mm for 0.03-0.8 Å −1 , and then combined the measurements. After the standard procedures of transmission correction, buffer scattering subtraction, and conversion to an absolute scale with water scattering, we obtained the scattering profile of the NCPs. The NCP concentrations were 1.49 mg/mL, in 20 mM Tris-HCl (pH 7.5), 50 mM NaCl, and 1 mM dithiothreitol, and the temperature was kept at 20°C.
We first examined the sample structure with the Guinier formula, which is established in the low q-range ðqÞ ¼ 4π λ sin θ 2 ; (λ and θ are the X-ray wavelength and the scattering angle, respectively).
where I(0) and R g are the zero angle scattering intensity and gyration radius, respectively. R g values were calculated with standard error. The observed SAXS intensity was corrected for background scattering, empty cell scattering, buffer scattering, and transmission factors, and subsequently converted to the absolute scale by SAngler (http://pfwww.kek.jp/saxs/SAngler.html). The Guinier analysis was performed with the linear least square method by Igor Pro (7.04).
Statistics and reproducibility. For HS-AFM imaging experiments, the numbers of total counted particles are presented in Figs. 3c, d, 5c, and Supplementary Fig. 4b, c. The histone exchange assay was repeated three times. The octamer formation assay was repeated two times.