© 2000 Nature America Inc. • http://structbio.nature.

com

letters

Crystal structure of a (H2A.Z–H2B) dimer and the (H3–H4)2 tetramer. Moreover,

H2A.Z nucleosomes have an altered surface that includes a
nucleosome core particle metal ion. This altered surface may lead to changes in higher
order structure, and/or could result in the association of spe-
containing the variant cific nuclear proteins with H2A.Z. Finally, incorporation of
H2A.Z and H2A within the same nucleosome is unlikely, due
histone H2A.Z to significant changes in the interface between the two
H2A.Z–H2B dimers.
Robert K. Suto1, Michael J. Clarkson2, The organization of eukaryotic DNA in nucleosomes results in
David J. Tremethick2 and Karolin Luger1 a global repression of transcription that may be overcome by
ATP-dependent chromatin remodeling factors or reversible post-
1Department of Biochemistry and Molecular Biology, Colorado State translational modification of histone tails. In addition to these
University, Fort Collins, Colorado 80523-1870, USA. 2The John Curtin intensely studied pathways for activating chromatin, there are
School of Medical Research, Australian National University, PO Box 334, alternative ways to locally alter chromatin structure. All eukaryot-
Canberra Australian Capital Territory 2601, Australia. ic organisms contain specialized histone protein variants with
distinctly different amino acid sequences and expression patterns
Activation of transcription within chromatin has been corre- compared to their replication-dependent counterparts. These
lated with the incorporation of the essential histone variant replacement histones may create an architecturally distinct chro-
H2A.Z into nucleosomes. H2A.Z and other histone variants matin structure by localized incorporation into nucleosomes.
© 2000 Nature America Inc. • http://structbio.nature.com

may establish structurally distinct chromosomal domains; H2A.Z, an essential histone variant found in all higher eukary-
however, the molecular mechanism by which they function is otes, is highly conserved with ∼90% sequence identity among
largely unknown. Here we report the 2.6 Å crystal structure of different organisms1. This high degree of conservation of his-
a nucleosome core particle containing the histone variant tones among species is observed for all histone genes. However,
H2A.Z. The overall structure is similar to that of the previous- H2A.Z shares only ∼60% sequence identity with major H2A
ly reported 2.8 Å nucleosome structure containing major his- (Fig. 1a), suggesting a unique function. H2A.Z is preferentially
tone proteins. However, distinct localized changes result in associated with actively transcribed chromatin2. The fundamen-
the subtle destabilization of the interaction between the tally important region of H2A.Z required for the development of

a b

c d e

Fig. 1 Overall comparison of H2A.Z-NCP and major-NCP. a, Sequence alignment of Xenopus laevis H2A and mouse H2A.Z. Intervals of 10 amino acids
for H2A (filled circle) and H2A.Z (open circle) are indicated. Differences between the two amino acid sequences are colored red. Regions that are
essential for the function of H2A.Z are boxed. The docking domain is indicated with a dashed line, secondary structure elements of the histone fold
(α1, α2 and α3) and extensions (αN and αC) are indicated. b, Stereo view of a section of the |2Fo - Fc| electron density map, calculated at 2.6 Å and con-
toured at 1.0 σ, showing the coordination of Mn2+ by H2A.Z (Figs 2, 3). c, Superposition of major-NCP and H2A.Z-NCP, viewed down the superhelical
axis. Only 73 bp of the DNA and associated proteins are shown. Regions of protein–DNA interaction are numbered starting from the nucleosomal
dyad. H3 is colored blue, H4 green, H2B red, H2A yellow, H2A.Z gray, and DNA brown. d, Side view of the superimposed nucleosomes in (c) (rotated
by 90° around the y-axis) with parts of the DNA removed for clarity. e, Superposition of H2A and H2A.Z, in a view similar to that in (c). The docking
domain is boxed.

nature structural biology • volume 7 number 12 • december 2000 1121

 © 2000 Nature America Inc. • http://structbio.nature.com

letters

Fig. 2 The docking domain. The docking domains of a,c, H2A.Z and
a b b,d, H2A are shown in two different views. The boxed inset shows the
complete H2A docking domain in the same orientation as in (b), which
displays amino acids 100–119 (H2A numbering). Regions of H3 (blue) and
H4 (green) that interact with the docking domain are indicated by num-
bers in italics; the manganese ion is shown in magenta. Intramolecular
hydrogen bonds are shown in cyan, intermolecular hydrogen bonds in
magenta. (c) and (d) shows the ring-shaped base of the domain (amino
acids 81–107; H2A numbering; circled) rotated by ∼90° around the x-axis.
The deletion in H2A.Z is indicated (∆). Conserved acidic patch residues
are bracketed.

amino acid substitutions between major H2A and H2A.Z are

accommodated with surprisingly small changes in the overall
structure of H2A.Z (Fig. 2e). As a consequence, interaction
between the histone folds of H2A.Z and H2B are virtually
unchanged, as are contacts between the H2A.Z–H2B dimer and
c d the DNA. However, sequence differences do lead to localized
changes in the interaction of the H2A.Z–H2B dimer with the
(H3–H4)2 tetramer, and between the H2A.Z–H2B dimers them-
© 2000 Nature America Inc. • http://structbio.nature.com

selves.

The H2A.Z docking domain

The ladle-shaped docking domain of H2A (residues 81–119, Figs
1a, 2 inset) creates a large interaction surface (2,400Å2) with the
(H3–H4)2 tetramer by guiding the H3αN helix to interact with
the last turn of the nucleosomal DNA, and by forming a short
β-sheet with the C-terminal region of H4 (Figs 1c, 2b). The
adult flies resides in a region that is responsible for maintaining equivalent region of Drosophila H2A.Z is essential in fly develop-
interactions between histone dimers and tetramers, whereas ment, and cannot be replaced with residues from H2A3. The
other regions are functionally exchangeable3. The crystal struc- overall shape of this domain, as well as the network of intermol-
ture of a nucleosome core particle containing H2A.Z (H2A.Z- ecular hydrogen bonds that stabilize the ring-shaped base of the
NCP), and the comparison of this structure with that of the domain, is conserved between the two structures (Figs 1e, 2c,d).
previously published structure of nucleosome core particle con- However, substitution of Gln 104 in H2A with glycine (the corre-
taining major histone proteins (‘major-NCP’)4 provides insight sponding position in H2A.Z is Gly 106) results in the loss of
into the molecular mechanisms by which this histone variant three hydrogen bonds, causing subtle destabilization of the inter-
exerts its essential function. action between the H2A.Z docking domain and H3 (Fig. 2a,b).
The amino acid sequence of H2A.Z is highly conserved
Overall structural comparison between different organisms5. This is also true for residues in the
The structure of H2A.Z-NCP (containing recombinant mouse docking domain that are solvent accessible and define a signifi-
H2A.Z and recombinant Xenoupus laevis H2B, H3 and H4) was cant portion of the surface of the histone octamer. One of these,
determined by molecular replacement, using major-NCP as a His 112, binds a manganese ion that is partially stabilized by His
starting model. Data collection and refinement statistics are sum- 114, whereas the comparable region in major H2A makes no
marized in Table 1; a representative region of the final electron such interaction (Figs 1b, 2a,b, 3). Manganese is included in the
density is shown in Fig. 1b. Differences between H2A.Z and H2A crystallization buffer4; a more likely metal to be found at this site
were clearly visible in the initial electron density map (not shown). in vivo would be copper or zinc6. Potentially, the location of a
The crystal structures of H2A.Z-NCP and major-NCP4 are metal ion at the nucleosomal surface may create a very specific
superimposable with an root mean square (r.m.s.) deviation interaction interface for other nuclear proteins7, such as some
<1 Å, with no major differences in the path of the DNA superhe- isoforms of linker histones (H1), assembly factors, or chromatin
lix nor in protein–DNA interactions (Fig. 1c,d). The numerous remodeling factors (see below).

a b

Fig. 3 The electrostatic potentials of nucleosomal surfaces.

a, H2A.Z-NCP. b, Major-NCP. The electrostatic potential
ranges from +7.0 (blue) to -7.0 (red) kcal mol–1 e–1 (color bar);
shown is the value at the solvent accessible surface (which is
extended from the molecular surface by 1.4 Å, the radius of a
water molecule) mapped onto the molecular surface32.
Molecular surfaces were calculated with the program MSMS31
using a 1.4 Å probe sphere. The manganese ion is shown in
green. The acidic patch that provides an essential crystal con-
tact with the N-terminal tail of histone H4 of a neighboring
nucleosome core particle is indicated by a black arrow.

1122 nature structural biology • volume 7 number 12 • december 2000

 © 2000 Nature America Inc. • http://structbio.nature.com

letters

a b c

Fig. 4 The L1 loop. a, Superposition of the L1 loops and α-helices of H2A.Z and H2A.Z′ (gray and off-white) and of H2A and H2A′ (yellow and light
yellow), emphasizing the alignment of the C-terminal ends of the H2A.Z α1 helices and the structural conservation of the DNA binding region of the
loop. b,c, Detailed view of the boxed area in (a) shows fundamental differences in the intermolecular interactions between two H2A.Z molecules
and two H2A molecules, respectively. First and last amino acid shown in (b) and (c) are circled.

These subtle differences in the molecular surface extend to None of these significant differences affect the protein–DNA
an acidic patch on the surface of the nucleosome, formed by interaction interface formed by the L1 loop of H2A.Z and the L2
© 2000 Nature America Inc. • http://structbio.nature.com

five amino acids from the H2A docking domain and one from loop of H2B (Fig. 4a). However, the pairing of one H2A.Z and
H2B (H2A: Glu 56, Glu 61, Glu 64, Asp 90, Glu 91 and Glu 92; one H2A chain within the same nucleosome core particle would
H2B: Glu 110; Figs 2d, 3). This region makes an essential crystal lead to steric clashes. Alternatively, if these clashes were avoided
packing contact with the H4 tail from a neighboring nucleo- by main chain and side chain rearrangements, no stabilizing
some8. It is likely that the acidic patch also serves as an interac- interaction could be formed between H2A and H2A.Z.
tion interface for histone tails or other protein factors in vivo. Interestingly, macroH2A, an H2A variant that is primarily asso-
In H2A.Z, these six amino acids are conserved, and the acidic ciated with the inactive X chromosome in mammals9, deviates
patch is extended by replacing Asn 94 in H2A with an aspartate again in this very region10, suggesting that the L1 loop provides a
(the corresponding position in H2A.Z is Asp 97), and H2A Lys mechanism to ensure the incorporation of only one type of H2A
95 with a serine (H2A.Z Ser 98). These substitutions, along histone into one nucleosome. Only histones that are self-inter-
with the deletion of one amino acid after H2A.Z residue 100 acting within a single nucleosome core — that is, H3 and H2A
(Fig. 1a), and the interaction of H2A.Z Lys 101 with the main — are known to have variants11–13.
chain (Fig. 2c), leave a hole at the center of the nucleosome that
is occupied by H2A Lys 95 and Arg 99 in the structure of the Conclusions
major-NCP (Fig. 3). Together, the above changes result in an How H2A.Z and other histone variants exert their specific and
altered surface on the H2A.Z–H2B dimer, and an uninterrupt- diverse function is still largely unknown and subject to wide
ed acidic surface across the face of the H2A.Z-containing his- ranging speculation14–16. The crystal structure presented here
tone octamer. provides a significant step toward understanding the function of
this essential histone variant. The structure allows us to rule out
H2A.Z–H2A.Z interaction that replacement of H2A with H2A.Z leads to major distortions
Another significant deviation between the two structures is in nucleosome structure or to changes in the path of the DNA.
observed at the L1 loop where the two subunits (H2A with H2A′, Protein–DNA interactions are also largely maintained. The
or H2A.Z with H2A.Z′) interact, seemingly holding together the observed structural differences most likely result in a subtle
two gyres of the DNA superhelix (Figs 1d, 4). The path of the α1 destabilization of the interface between the (H2A.Z-H2B) dimer
helix and parts of the loop connecting α1 and α2 of the histone and the (H3-H4)2 tetramer and in changes in molecular surface
fold are different between the two structures. The changes in the features. The latter may modify higher order structure, either via
L1 loop lead to a perfect alignment of the two H2A.Z α1 helical altered interactions between nucleosomes, or between a nucleo-
axes; the same helices are in a parallel orientation in the major- some and its linker histone. These unique surface features may
NCP structure (Fig. 4a). serve to recruit factors involved in chromatin assembly and
Even though the size of the surface area that is buried between remodeling. In this context, the presence of a metal binding site
the two sets of H2A.Z-H2B (or H2A-H2B) dimers are similar is particularly intriguing, since a number of proteins implicated
(∼480 Å2), the character of the interface is dramatically differ- in chromatin remodeling contain a zinc finger-like motif, the
ent. Two intermolecular hydrogen bonds between Glu 41–Asn PHD domain17–20. The crystal structure of the H2A.Z nucleo-
38′ and Glu 41′–Asn 38 in major-NCP (Fig. 4c) are replaced by some allows us to design specific experiments to analyze the role
two main chain interactions and four main chain–side chain of H2A.Z in the cell. In particular, we are in the process of identi-
interactions in H2A.Z-NCP (Fig. 4b). The two-fold noncrystal- fying factors that specifically bind to H2A.Z-nucleosomes.
lographic symmetry of the particle is disrupted by the presence
of asymmetric hydrogen bonds between the main chain of Thr Methods
40 and Thr 41′, and between Ser 42 and Ser 42′. To enable these Nucleosome core particle crystallization. Previously developed
interactions, the peptide bond between amino acids 40 and 41 is protocols21,22 were adapted to express, purify and reconstitute
recombinant mouse H2A.Z with recombinant Xenopus laevis H3, H4
flipped in one of the two H2A.Z molecules (Fig. 4b), and the and H2B, and a 146-base pair (bp) palindromic DNA fragment
loop conformation between residues 38 and 44 differs slightly. derived from human α-satellite DNA4. Crystals of the H2A.Z-contain-
The electron density in this area suggests that both conforma- ing nucleosome core particle were grown from 80–96 mM MnCl2,
tions are present in the crystal. 68–78 mM KCl, and 20 mM potassium cacodylate (pH 6.0) contain-

nature structural biology • volume 7 number 12 • december 2000 1123

 © 2000 Nature America Inc. • http://structbio.nature.com

letters

Table 1 Statistics for data collection and refinement Mn; all other histidine side chains were protonated on NE2. Partial
charges were assigned using the AMBER library that includes polar
Data collection statistics hydrogens only27. Electrostatic potentials were calculated on a grid
Space group P212121 with 1 Å spacing using a dielectric constant of 3 for the protein, a
Unit cell dimensions (Å)1 a =105.7, b = 183.2, c = 109.9 dielectric constant of 80 for the surrounding environment, and an
Resolution range 25–2.6 Å ion exclusion radius of 1.4 Å, and assuming monovalent added salt
at 150 mM ionic strength. Figures were prepared with BOBSCRIPT28,
Unique reflections 65,959
MIDAS29, and AVS30.
Completeness (%)2 99.6 (99.9)
Rmerge2,3 0.11 (0.24)
Coordinates. The coordinates have been deposited in the Protein
Refinement statistics Data Bank (accession code 1F66).
Number of protein residues in the final model4 769
Number of base pairs in the DNA 146
Number of water molecules 325 Acknowledgments
We thank T. Earnest at the Advanced Light Source in Berkeley for support and
Number of Mg ions 15
cooperation; W. Schreurs, M. Schnizer and E. Schonbrunn (Colorado State
Total number of atoms in the final model 12,399 University) for technical support; and V. Roberts (the Scripps Research Institute)
R-factor5 / Rfree 0.193 / 0.249 for help with Fig. 3. This work was supported in part by a Searle Scholar Award to
Resolution range 25–2.6 Å K.L., by the Cancer League of Colorado, by the Graduate School of Colorado
R.m.s. deviation from ideality State University, and by the Basil O’Conner Starter Scholar Award from the
Bonds (Å) 0.013 March of Dimes to K.L.
Angles (°) 1.55
© 2000 Nature America Inc. • http://structbio.nature.com

Correspondence and request for materials should be addressed to K.L.

Average B-factors (Å)2 email: [email protected]
Protein 53.3
DNA 99.0 Received 6 July, 2000; accepted 19 September, 2000.
Solvent 64.3
1. Iouzalen, N., Moreau, J. & Mechali, M. Nucleic Acids Res. 24, 3947–3952 (1996).
1Unit cell naming conventions were adopted from the previously pub- 2. Stargell, L.A. et al. Genes Dev. 7, 2641–2651 (1994).
lished structure of major-NCP4. 3. Clarkson, M.J., Wells, J.R., Gibson, F., Saint, R. & Tremethick, D.J. Nature 399,
2Values in the parentheses are for the final shell. 694–697 (1999).
4. Luger, K., Maeder, A.W., Richmond, R.K., Sargent, D.F. & Richmond, T.J. Nature
merge = ΣIh − <Ih> / ΣIh, where <Ih> is the mean of measurements for a
3R
389, 251–259 (1997).
single hkl. 5. Baxevanis, A.D., Arents, G., Moudrianakis, E.N. & Landsman, D. Nucleic Acids Res.
4Resides included in each histone subunit: H3: 36–135, H3′: 33–135, H4:
23, 2685–2691 (1995).
23–102, H4′: 17–102, H2A.Z 16–118, H2A.Z′: 16–122, H2B: 28–122, H2B′: 6. Harding, M. Acta Crystallogr. D D55, 1432–1443 (1999).
28–122. The remainder of the histone tails was too disordered to be 7. Mackay, J.P. & Crossley, M. Trends Biochem. Sci. 23, 1–4 (1998).
included in the final model. 8. Luger, K. & Richmond, T.J. Curr. Opin. Genet. Dev. 8, 140–146 (1998).
5R-factor = ΣF 9. Costanzi, C. & Pehrson, J.R. Nature 393, 599–601 (1998).
obs − Fcalc / Σ Fobs 10. Pehrson, J.R. & Fuji, R.N. Nucleic Acids Res. 26, 2837–2842 (1998).
11. Sullivan, K.F., Hechenberger, M. & Masri, K. J. Cell. Biol. 127, 581–592 (1994).
12. Pehrson, J.R. & Fried, V.A. Science 257, 1398–1400 (1992).
ing 4 mg ml–1 core particle equilibrated by vapor diffusion against 13. Thatcher, T.H. & Gorovsky, M.A. Nucleic Acids Res. 22, 174–179 (1994).
40–48 mM MnCl2, 34–39 mM KCl, and 20 mM potassium cacodylate 14. Wolffe, A.P. Dev. Biol. 157, 224–231 (1993).
15. Li, W., Nagaraja, S., Delcuve, G.P., Hendzel, M.J. & Davie, J.R. Biochem. J. 296,
(pH 6.0). Crystal size was increased by macro seeding. Crystals of 737–744 (1993).
H2A.Z-NCP were grown in the same space group and unit cell as 16. van Daal, A. & Elgin, S.C. Mol. Biol. Cell 3, 593–602 (1992).
major-NCP4. 17. Aasland, R., Gibson, T.J. & Stewart, A.F. Trends Biochem. Sci. 20, 56–59 (1995).
18. Wade, P.A., Jones, P.L., Vermaak, D. & Wolffe, A.P. Curr. Biol. 8, 843–846 (1998).
19. Borrow, J. et al. Nature Genet. 14, 33–41 (1996).
Data collection and refinement. X-ray data was collected at 20. Moosmann, P., Georgiev, O., Le Douarin, B., Bourquin, J.P. & Schaffner, W. Nucleic
beamline 5.0.2 at the Advanced Light Source, Berkeley. Data from Acids Res. 24, 4859–4867 (1996).
21. Luger, K., Rechsteiner, T.J. & Richmond, T.J. Methods Enzymol. 304, 3–19 (1999).
three crystals were processed with Denzo and Scalepack23. 22. Richmond, T.J., Searles, M.A. & Simpson, R.T. J. Mol. Biol. 199, 161–170 (1988).
Molecular replacement (using PDB entry 1AOI as the original search 23. Otwinowski, Z. & Minor, W. Processing of X-ray diffraction data collected in
model) and refinement was done with CNS24, model building was oscillation mode, 307–326 (Academic Press, New York; 1997).
24. Brunger, A.T., Adams, P.D. & Rice, L.M. Structure 5, 325–336 (1997).
done with O25. The entire model was checked using SA-OMIT maps 25. Jones, T.A., Zou, J.Y., Cowan, S.W. & Kjelgaard, M. Acta Cryst. A 47, 110–119
during early stages of model building. (1991).
26. Gilson, M.K., Davis, M.E., Luty, B.A. & McCammon, J.A. J. Phys. Chem. 97,
3591–3600 (1993).
Electrostatic potential. Electrostatic potentials were calculated 27. Weiner, S.J. et al. J. Am. Chem. Soc. 106, 765–784 (1984).
with the program UHBD26, which uses finite difference methods to 28. Esnouf, R.M. J. Mol. Graph. Model 15, 132–143 (1997).
solve the Poisson-Boltzmann equation. Polar hydrogen atoms were 29. Ferrin, T.E., Huang, C.C., Jarvis, L.E. & Langridge, R. J. Mol. Graphics 6, 13–27
(1988).
added to the crystallographic coordinates with the molecular 30. Upson, C. et al. IEEE Comp. Graphics Appl. 9, 30–42 (1989).
graphics program Insight (MSI, Inc., San Diego). His 112 of H2A.Z- 31. Sanner, M.F., Olson, A.J. & Spehner, J.C. Biopolymers 38, 305–320 (1996).
NCP were protonated on ND1 due to ligation of the NE2 atom to 32. Getzoff, E.D. et al. Nature 306, 287–290 (1983).

1124 nature structural biology • volume 7 number 12 • december 2000