REPORTS

30. A. K. Chippindale, J. R. Gibson, W. R. Rice, Proc. Natl. University (H.K. and J.L.); the Finnish Cultural Foundation all co-authors. Authors after the first author are listed in
Acad. Sci. U.S.A. 98, 1671 (2001). and Emil Aaltonen Foundation ( J.L.); and the Centre of alphabetical order. Data have been deposited in the Dryad
31. See supporting material on Science Online. Excellence in Evolutionary Research, University of Repository (doi:10.5061/dryad.6m0f6870).
32. R. M. Calisi, G. E. Bentley, Horm. Behav. 56, 1 Jyväskylä. We thank the staff of the Experimental Animal
(2009). Unit and Konnevesi Research Station, University of Supporting Online Material
33. D. S. Falconer, T. F. C. MacKay, Introduction to Jyväskylä; R. Närä and H. Pietiläinen for logistical www.sciencemag.org/cgi/content/full/334/6058/972/DC1
Quantitative Genetics (Pearson Prentice Hall, London, support; T. Laaksonen, V. Lummaa, and three anonymous Materials and Methods
ed. 4, 1996). reviewers for comments; and C. Soulsbury for statistical Figs. S1 to S3
Acknowledgments: Supported by Academy of Finland grants advice. The authors declare no conflicts of interest. All Table S1
115961, 119200, and 218107 (E.K.), 132190 (T.M.), co-authors designed this study; M.M., E.K., T.M. and H.M. References (34–41)
and 103508 and 108566 (S.C.M.); the Vanamo collected and analyzed the empirical data; J.L. analyzed
Biological Society and Ehrnrooth Foundation (M.M.); the theoretical results with input from all authors; and M.M. 20 May 2011; accepted 29 September 2011
Australian Research Council and Australian National led the preparation of the manuscript with input from 10.1126/science.1208708

cluster is referred to as the P cluster. The P clus-

X-ray Emission Spectroscopy ters serve as electron-transfer sites. Several re-
action intermediates in nitrogenase catalysis have
Evidences a Central Carbon in the recently been observed (3, 4). However, despite
the progress in the experimental and theoretical
Nitrogenase Iron-Molybdenum Cofactor analysis of the FeMoco (4–7), neither the re-
action that occurs at the FeMoco nor the struc-
ture of FeMoco has been fully clarified. In 2002,
Kyle M. Lancaster,1 Michael Roemelt,2 Patrick Ettenhuber,2 Yilin Hu,3 Markus W. Ribbe,3* Einsle et al. identified a light atom in the center
Frank Neese,2,4* Uwe Bergmann,5* Serena DeBeer1,4* of FeMoco that could be attributed to a single,
fully ionized C, N, or O atom (2). No consensus
Nitrogenase is a complex enzyme that catalyzes the reduction of dinitrogen to ammonia. has since emerged concerning the nature of this
Despite insight from structural and biochemical studies, its structure and mechanism await key atom. Study of FeMoco by electron paramag-
full characterization. An iron-molybdenum cofactor (FeMoco) is thought to be the site of dinitrogen netic resonance and related techniques is com-
reduction, but the identity of a central atom in this cofactor remains unknown. Fe Kb x-ray plicated by complex spin-couplings between the
emission spectroscopy (XES) of intact nitrogenase MoFe protein, isolated FeMoco, and the open-shell ions, which are not fully understood.
FeMoco-deficient ∆nifB protein indicates that among the candidate atoms oxygen, nitrogen, Mössbauer spectroscopy suffers from spectral
and carbon, it is carbon that best fits the XES data. The experimental XES is supported by crowding, and neither nuclear resonance vibra-
computational efforts, which show that oxidation and spin states do not affect the assignment tional spectroscopy nor extended x-ray absorption
of the central atom to C4–. Identification of the central atom will drive further studies on its fine structure are sufficiently conclusive (8).
role in catalysis. Herein, we report iron Kb valence-to-core
(V2C) x-ray emission spectroscopy (XES) of
itrogenase (N2ase), found in symbiotic num ion, and nine sulfides (Fig. 1A); this cluster N2ase and demonstrate that these data provide a

N and free-living diazotrophs, catalyzes

the reduction of dinitrogen (N2) to am-
monia (NH3) using eight electrons, eight protons,
is referred to as the iron-molybdenum cofactor
(FeMoco) and is thought to be the site of di-
nitrogen activation. For each FeMoco (of which
signature for the presence and identity of the
central atom. Ka and Kb XES monitor the emis-
sion of photons after ionization of a metal 1s
and 16 MgATPs (ATP, adenosine triphosphate) there are two in the a2b2 tetrameric MoFe pro- electron. The Kb1,3 emission line (~7040 to 7070
(1). Industrially, the same reaction is performed tein) there is an additional cluster that consists eV) corresponds to an electric dipole allowed
by the Haber-Bosch process that produces more of eight irons and seven sulfides (Fig. 1B); this 3p → 1s transition. To higher emission energies,
than 100 million tons of NH3 each year, there-
by accounting for ~1.4% of global energy con-
sumption. Understanding how nature activates Fig. 1. The FeMoco (A) and P clus-
the strongest homodinuclear bond in chemistry, ter (B) of nitrogenase (adapted from
the triple bond of N2, is the key for the future the Protein Data Bank: identifica-
tion number 1MIN). Orange, Fe; yel-
design of molecular catalysts.
low, S; light blue, Mo; black, C4 –,
The high-resolution crystal structure of N2ase
N3-, or O2–; dark blue, nitrogen; gray,
determined by Einsle et al. (2) showed that the carbon. For clarity, the homocitrate
active site of the molybdenum-iron (MoFe) pro- and histidine ligands to the Mo have
tein component of N2ase binds a complex cluster been omitted.
consisting of seven iron ions, one molybde-

1
Department of Chemistry and Chemical Biology, Cornell Uni-
versity, Ithaca, NY 14853, USA. 2Institut für Physikalische und
Theoretische Chemie, Universität Bonn, D-53115 Bonn, Germany.
3
Department of Molecular Biology and Biochemistry, University
of California, Irvine, CA 92697, USA. 4Max-Planck-Institut für
Bioanorganische Chemie, Stiftstrasse 34-36, D-45470 Mülheim
an der Ruhr, Germany. 5Linac Coherent Light Source, SLAC
National Accelerator Laboratory, Menlo Park, CA 94025, USA.
*To whom correspondence should be addressed: mribbe@uci.
edu (M.W.R.); frank.neese@mpi-mail.mpg.de (F.N.); bergmann@
slac.stanford.edu (U.B.); serena.debeer@mpi-mail.mpg.de
(S.D.)

974 18 NOVEMBER 2011 VOL 334 SCIENCE www.sciencemag.org

 REPORTS

Fig. 2. (A) Normalized V2C XES spectra of isolated FeMoco (red) and a representative fit to the data (black dashed line). (B) Comparison of the
normalized V2C XES data for FeMoco (red), the MoFe protein (gray), and the ∆nifB MoFe protein (black). (Inset) V2C satellite region for Fe2O3 (red),
Fe3N (blue), and MoFe protein (gray).

Table 1. V2C XES fit parameters. n/a, not applicable. plex (19). Both the P clusters and the Fe4S4
cubane have very similar XES spectra (figs. S2
∆nifB MoFe protein FeMoco MoFe protein and S3). Thus, the weak 7098.8-eV feature must
Integrated Integrated Integrated be attributed to a S 3s → Fe 1s transition. Under
E (eV) E (eV) E (eV) the plausible assumption that the S 3s contri-
intensity intensity intensity
butions to the P cluster and FeMoco V2C XES
Kb′′ peak 1 7098.8 0.30 7098.8 0.18 7098.8 0.30
spectra are similar, we can model the satellite
Kb′′ peak 2 n/a n/a 7100.2 1.60 7100.5 0.58
region with two features: one fixed at 7098.8 eV
Total Kb′′ integrated intensity n/a 0.30 n/a 1.78 n/a 0.88
(corresponding to the S 3s contributions) and a
Total V2C integrated intensity 7.73 10.45 7.61
second to higher energy (7100.2 eV) with in-
creased intensity (1.6 units) (Fig. 2A and Table 1).
The higher-energy feature is attributed to the
valence-electron transitions into the metal 1s core maxima at ~7108 and ~7100 eV are assigned to presence of the interstitial light atom. Compar-
hole are observed (referred to as the Kb2,5/Kb′′ or ligand np and ns contributions, respectively (12). ison of the energy of the 7100.2-eV satellite fea-
V2C region). These transitions have previously To assess the contribution of the sulfur ligands ture in FeMoco to the O 2s → Fe 1s (~7092 eV)
been assigned as ligand np → metal 1s (Kb2,5, relative to the interstitial atom X, data were also and N 2s → 1s (~7095 eV) transitions observed
~7102 to 7112 eV) and ligand ns → metal 1s (Kb′′ obtained for the ∆nifB MoFe protein (Fig. 2B). in Fe2O3 and Fe3N, respectively (Fig. 2B, inset),
or “satellite,” ~7090 to 7102 eV) transitions (9). This mutant contains only the two P clusters indicates that this feature arises from a ligand
V2C XES studies of Cr and Mn complexes have (17, 18). Based on their structural similarity, it with 5- and 8-eV lower ionization potential than
shown that the Kb′′ features provide a signature can be assumed that the P cluster and the FeMoco O or N, respectively. Therefore, this comparison
for the identity of the directly coordinating lig- have similar sulfur contributions to their XES argues against either N or O 2s contributions and
ands, because energies of the observed features spectra; this assumption is also supported com- strongly supports a C 2s → Fe 1s assignment.
depend primarily on the ligand 2s ionization en- putationally (see below). Data were obtained for Fe V2C XES spectra can be predicted sur-
ergies (10, 11). We recently developed an exper- the intact MoFe protein (containing both clusters) prisingly well within a simple scheme based on
imental and theoretical protocol for the analysis (Fig. 2B). The MoFe protein spectra map well density functional theory (12). To complement
of V2C XES spectra and applied it to mono- onto an average of the spectra of the P cluster the experimental data, we performed detailed
(12–14) and multinuclear (15) iron complexes. (represented by the ∆nifB MoFe protein) and the calculations on the FeMoco, the P cluster, and the
Of particular relevance is a study of a six-iron isolated FeMoco (fig. S1). [Fe4S4(SPh)4]2– model complex. The FeMoco
cluster with a central m6-C4– (15). These data Comparison between the data of the isolated was modeled by a structure containing 152 atoms
show a feature at 7099 eV that is attributed to a FeMoco and that of the P clusters in the ∆nifB obtained from the high-resolution crystal struc-
transition originating from the m6-C4– 2s orbit- MoFe protein allowed us to assess the relative ture of Einsle et al. (2), which incorporates the
al. Computationally, this feature is predicted to contributions of these clusters to the spectra. The key structural and electronic features of the sys-
shift to 7094 eV for a m6-N3– and 7088 eV for a V2C XES data of the P clusters showed only a tem. The oxidation state of this iron-sulfur cluster
m6-O2–. These trends closely parallel previous weak satellite at 7098.8 with 0.30 T 0.03 units of in the resting state of the enzyme has not been
observations for infinite lattice complexes and integrated intensity. In contrast, isolated FeMoco determined unambiguously. Thus, we performed
mononuclear molecular complexes, thus high- exhibited a well-resolved satellite feature to high- calculations for the two oxidation states that have
lighting the general applicability of this method er energy (7100.2 eV) with an approximately been shown to be most likely (20–22). Although
(11, 12, 16). sixfold increase in the integrated intensity of the these two states differ by two units of charge, the
Figure 2A presents the normalized V2C XES satellite feature (1.78 T 0.18 units). To better un- differences in the calculated V2C XES transition
data of the isolated FeMoco of N2ase, together derstand the origin of these satellite features, we energies and intensities are much smaller than
with a representative fit to the data. Based on also compared the data for ∆nifB MoFe protein the experimental resolution (fig. S6). The broken
previous studies, the features observed with to the XES data for a [Fe4S4(SPh)4]2– model com- symmetry approach was used to approximate the

www.sciencemag.org SCIENCE VOL 334 18 NOVEMBER 2011 975

REPORTS

Fig. 3. (A) Comparison of the calculated V2C XES spectra of FeMoco

with an interstitial C4 – (black), N3– (blue), and O2– (red) and of the
spectra of the P clusters (gray). (B) Calculated V2C XES spectra of FeMoco
with an interstitial C4 – (black) and the P clusters (gray). (C) Experimental
difference spectrum of FeMoco with the P clusters (gray), as well as cal-
culated difference spectra of the P clusters with FeMoco containing in-
terstitial C4 – (black), N3– (blue), and O2– (red).

effects of magnetic coupling of the various spin ions give rise to two features in the V2C spec- tensity) ions considerably exceed this threshold.
centers in the FeMoco calculations (23). Howev- trum associated with transitions from the ligand In addition, several other studies on O2–and N3–
er, calculations reveal that the ligand-to-metal 2s and 2p orbitals, respectively. These features have shown features at the corresponding energy
crossover region of the predicted V2C XES spec- occur at 7096.1 and 7105.1 eV for N3– and at offsets (9–12). This finding raises interesting ques-
tra is largely unaffected by magnetic coupling 7091.0 and 7104.0 eV for O2–. When a C4 – ion tions about both the role of the central atom and
(fig. S7). This finding is understandable con- is placed in the center of the FeMoco, the two the possible pathways for biosynthesis of such an
sidering the large linewidth of the experimental features are observed at 7100.2 and 7107.9 eV organometallic cluster.
spectra and the rather subtle differences in or- (Fig. 3B). For N3– and C4 –, the higher-energy fea-
bital energies arising from different magnetic ture is not distinguishable from the large peak at
References and Notes
coupling schemes. More importantly, the pre- ~7107 eV that is dominated by transitions orig- 1. Y. L. Hu, M. W. Ribbe, Acc. Chem. Res. 43, 475 (2010).
dicted V2C spectra were highly sensitive to the inating from the S 3p orbitals. 2. O. Einsle et al., Science 297, 1696 (2002).
identity of the interstitial ion. Figure 3A presents Taken together, the experimental and theoret- 3. B. M. Barney et al., Biochemistry 48, 9094 (2009).
the calculated spectra of the FeMoco, assuming ical results support assignment of the interstitial 4. B. M. Hoffman, D. R. Dean, L. C. Seefeldt, Acc. Chem.
interstitial O2–, N3–, and C 4– ions together with species as a C4 –. The calculated position of the 5.
Res. 42, 609 (2009).
D. Lukoyanov et al., Inorg. Chem. 46, 11437 (2007).
the calculated spectrum for the P cluster. As ex- C4 – 2s → Fe 1s peak matches the experimentally 6. F. Neese, Angew. Chem. Int. Ed. 45, 196 (2005).
pected, all four spectra exhibit a relatively strong determined position at 7100.2 eV. Both N3– and 7. T. V. Harris, R. K. Szilagyi, Inorg. Chem. 50, 4811 (2011).
feature at ~7099.3 eV, corresponding to tran- O2– are unlikely, as their respective calculated 8. Y. M. Xiao et al., J. Am. Chem. Soc. 128, 7608 (2006).
9. P. Glatzel, U. Bergmann, Coord. Chem. Rev. 249, 65 (2005).
sitions from S 3s orbitals to the Fe 1s orbitals. spectra show strong features at 7096.1 eV (N3– 2s)
10. G. Smolentsev et al., J. Am. Chem. Soc. 131, 13161 (2009).
The only exception is FeMoco with a central and 7091.0 eV (O2– 2s). In addition, the mea- 11. S. G. Eeckhout et al., J. Anal. At. Spectrom. 24, 215 (2009).
C4– (Fig. 3B), where the maximum is slightly sured spectra do not exhibit any features at lower 12. N. Lee, T. Petrenko, U. Bergmann, F. Neese, S. DeBeer,
shifted to higher energies due to contributions energies than the S 3s peak, whereas such fea- J. Am. Chem. Soc. 132, 9715 (2010).
from C4–-related transitions in the same region. tures have been observed experimentally in other 13. C. J. Pollock, S. DeBeer, J. Am. Chem. Soc. 133, 5594
(2011).
Hence, our presented data, along with analogous N3– and O2– systems (as shown in the inset of 14. K. M. Lancaster, K. D. Finkelstein, S. DeBeer, Inorg.
calculations on [Fe4S4(SPh)4]2– (fig. S8 and Fig. 2B). The assignment is further supported by Chem. 50, 6767 (2011).
S9), support the aforementioned assumption our previous studies that have shown that features 15. M. U. Delgado-Jaime et al., Inorg. Chem. 10.1021/ic201173j
that the S peak in the V2C region appears at the with a calculated intensity of more than 10 to 15 (2011).
16. U. Bergmann, C. R. Horne, T. J. Collins, J. M. Workman,
same position of the spectrum for all measured units of intensity are experimentally observable S. P. Cramer, Chem. Phys. Lett. 302, 119 (1999).
species. (12). These studies also showed that the inte- 17. B. Schmid et al., Science 296, 352 (2002).
Subtraction of the calculated P-cluster spec- grated intensities of experimental and calculated 18. R. M. Allen, R. Chatterjee, P. W. Ludden, V. K. Shah,
trum from the calculated spectrum of the three V2C agree strongly, with a 19% error for crystal- J. Biol. Chem. 270, 26890 (1995).
19. B. A. Averill, T. Herskovitz, R. H. Holm, J. A. Ibers, J. Am.
FeMoco species yields the contributions from the lographic structures. Even considering this error, Chem. Soc. 95, 3523 (1973).
respective interstitial ions (Fig. 3C). Analysis of the calculated low-energy features related to the 20. H. I. Lee, B. J. Hales, B. M. Hoffman, J. Am. Chem. Soc.
the difference spectra reveals that the interstitial N3– (31 units of intensity) and O2– (26 units of in- 119, 11395 (1997).

976 18 NOVEMBER 2011 VOL 334 SCIENCE www.sciencemag.org

 REPORTS
21. A. E. True, M. J. Nelson, R. A. Venters, W. H. Ormejohnson, the SFB 813; M.W.R. thanks the NIH for funding Supporting Online Material
B. M. Hoffman, J. Am. Chem. Soc. 110, 1935 (1988). (grant R01-GM 67626). Portions of this research were www.sciencemag.org/cgi/content/full/334/6058/974/DC1
22. S. J. Yoo, H. C. Angove, V. Papaefthymiou, B. K. Burgess, carried out at the Stanford Synchrotron Radiation SOM Text
E. Munck, J. Am. Chem. Soc. 122, 4926 (2000). Lightsource (SSRL), a U.S. Department of Energy (DOE), Figs. S1 to S11
23. L. Noodleman, J. Chem. Phys. 74, 5737 (1981). Basic Energy Sciences user facility. The SSRL Structural Tables S1 and S2
Acknowledgments: S.D. thanks Cornell Univ. for financial Molecular Biology program is supported by DOE, References (24–35)
support and the Alfred P. Sloan Foundation for a Biological and Environmental Research, and NIH,
fellowship; F.N. acknowledges financial support from National Center for Research Resources, Biomedical 1 April 2011; accepted 8 September 2011
the Univ. of Bonn, the Max Planck Society, and Technology Program. 10.1126/science.1206445

tion in Sir3 (D205N) that confers increased bind-

Structural Basis of Silencing: ing to nucleosomes in vitro. Expression of the
BAH D205N domain fused to LexA partially
Sir3 BAH Domain in Complex with a restores silencing of mating type loci in a sir3
null background. This domain is able, therefore,
Nucleosome at 3.0 Å Resolution to combine with Sir2 and Sir4 to cause partial si-
lencing when it is attached to an ectopic dimeri-
zation domain (27).
Karim-Jean Armache,1,2 Joseph D. Garlick,1,2 Daniele Canzio,3,4 We report the crystal structure of the complex
Geeta J. Narlikar,3 Robert E. Kingston1,2* of the hypermorphic D205N Sir3 BAH domain
(BAHSir3) and the nucleosome core particle
Gene silencing is essential for regulating cell fate in eukaryotes. Altered chromatin architectures (NCP) at 3.0 Å resolution. Details of complex
contribute to maintaining the silenced state in a variety of species. The silent information regulator reconstitution, crystallization, data collection, and
(Sir) proteins regulate mating type in Saccharomyces cerevisiae. One of these proteins, Sir3, interacts refinement can be found in the supporting online
directly with the nucleosome to help generate silenced domains. We determined the crystal structure material (28). The BAH domain interacts exten-
of a complex of the yeast Sir3 BAH (bromo-associated homology) domain and the nucleosome core sively with each of the four core histones and,
particle at 3.0 angstrom resolution. We see multiple molecular interactions between the protein consequently, the solvent-accessible surface area
surfaces of the nucleosome and the BAH domain that explain numerous genetic mutations. These buried between BAHSir3 and the nucleosome is
interactions are accompanied by structural rearrangements in both the nucleosome and the BAH large (1750 Å2, probe radius 1.4 Å). The structure
domain. The structure explains how covalent modifications on H4K16 and H3K79 regulate formation shows a pseudo-two-fold symmetry, similar to
of a silencing complex that contains the nucleosome as a central component. that seen with the RCC1-nucleosome complex
(29), in that BAHSir3 interacts in a similar manner
ukaryotic cells normally carry the com- somal arrays in vitro (3–5). The involvement of with each of the two opposite faces of the nu-

E plete set of genes needed to specify every

cell type. Establishment of a specific cell
fate requires the silencing of genes whose ex-
nucleosomes in the mechanism of silencing was
first indicated by the observation that yeast could
not silence HMLa and HMRa when they con-
cleosome (Fig. 1). We observed 30 residues of
BAHSir3 making contacts predominately with the
core histones rather than nucleosomal DNA, sug-
pression would disrupt that fate. Several diverse tained a mutated form of histone H4 with a gesting that this protein-protein interface is crit-
families of protein complexes maintain silencing; deletion of the N-terminal tail (6). Subsequently, ical to silencing.
however, the mechanisms involved are similar in specific point mutations that affected silencing Interactions with the core histones are medi-
Saccharomyces cerevisiae and in multicellular were found in the N-terminal tails and in the glob- ated through five regions on the surface of BAHSir3.
eukaryotes (1). Regulation of mating type loci in ular portions of core histones (7–14), and deacetyl- These regions map well to contacts inferred from
S. cerevisiae serves as a paradigm for silencing. ation of histone H4 was identified as a hallmark genetic screens (see Figs. 1D and 2B for a sum-
Yeast growing as haploids can adopt two mating of silenced regions (15). Reporter gene expres- mary). The BAH domain interacts with the H4
types, a and a. The genes that are expressed at the sion, restriction enzyme accessibility, and micro- tail, which becomes folded upon binding, and the
MAT loci determine cell fate, whereas genes spec- coccal nuclease susceptibility were used to show regions of histones H3 and H4 that make up the
ifying the opposite fate can be found at the silent that domains of silenced chromatin created by the LRS domain. In addition, BAHSir3 contacts his-
HMLa or HMRa loci (1, 2). The silent information SIR complex are several kb in length (16–21). tone H2B at a position adjacent to the LRS surface
regulator (Sir) proteins are essential for silencing Several aspects of the extensive body of work and the H2A/H2B acidic patch. Of the histone
of HMLa and HMRa, as well as telomeres and on Sir3 interactions with nucleosomes are es- residues contacted by BAHSir3 only one residue
the ribosomal DNA (rDNA) loci (1, 2). pecially relevant to the structural work described (H4V21) varies between the Xenopus laevis his-
The Sir proteins create domains of silenced here. Silencing requires deacetylation of histone tones used here and yeast histones (Fig. 4B and
chromatin. A long-standing hypothesis is that H4 lysine 16 (H4K16); we describe the atomic fig. S3). Both of the histone residues that can be
these proteins form specific repressive architec- contacts in the Sir3 binding pocket for H4K16. covalently modified and participate in the regula-
tures that involve the basic unit of chromatin, the We also describe contacts with H3K79, whose tion of silencing (7–9, 30) (H3K79 and H4K16)
nucleosome. In support of this hypothesis, the methylation has the potential to modulate silenc- are ordered in the structure (Fig. 1B and below).
SIR complex or Sir3 alone can compact nucleo- ing. Many of the mutations in histones that affect Interactions between BAHSir3 and the nucleo-
silencing lie in the LRS (loss of rDNA silencing) some are established through flexible regions,
1
Department of Molecular Biology, Massachusetts General Hos- (11, 12) domain of the nucleosome core, and we which fold upon interaction (Fig. 2 and Fig. 1C).
pital, Boston, MA 02114, USA. 2Department of Genetics, describe numerous contacts between that region The structures of both the BAH domain and the
Harvard Medical School, Boston, MA 02115, USA. 3Department and Sir3. Mutations that affect silencing have NCP alone were determined previously (27, 31, 32),
of Biochemistry and Biophysics, University of California, San been found both at the N terminus and at the allowing comparison to the structure of the com-
Francisco, San Francisco, CA 94158, USA. 4Chemistry and Chem-
ical Biology Graduate Program, University of California, San C-terminal part of Sir3 (22). Most of these mu- plex described here. One striking transition that ac-
Francisco, San Francisco, CA 94158, USA. tations are clustered in the bromo-associated ho- companies assembly of the complex is folding and
*To whom correspondence should be addressed. E-mail: mology (BAH) domain that is found in the N ordering of the histone H4 tail through extended
kingston@molbio.mgh.harvard.edu terminus of Sir3 (23–26). Here, we used a muta- interactions with loops 2 and 4 of BAHSir3 (Fig.

www.sciencemag.org SCIENCE VOL 334 18 NOVEMBER 2011 977