Replacement of histone H3 with CENP-A directs
global nucleosome array condensation and
loosening of nucleosome superhelical termini
Tanya Panchenkoa,b,1, Troy C. Sorensenc,1, Christopher L. Woodcockd, Zhong-yuan Kana, Stacey Wooda,
Michael G. Reschc, Karolin Lugerc, S. Walter Englandera,2, Jeffrey C. Hansenc, and Ben E. Blacka,b,2
aDepartment of Biochemistry and Biophysics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA 19104;
and Molecular Biology, University of Pennsylvania, Philadelphia, PA 19104;
University, Ft. Collins, CO 80523; and
bGraduate Group in Cell
cDepartment of Biochemistry and Molecular Biology, Colorado State
dDepartment of Biology, University of Massachusetts, Amherst, MA 01003
Contributed by S. Walter Englander, August 18, 2011 (sent for review July 18, 2011)
Centromere protein A (CENP-A) is a histone H3 variant that marks
centromere location on the chromosome. To study the subunit
structure and folding of human CENP-A-containing chromatin,
we generated a set of nucleosomal arrays with canonical core
histones and another set with CENP-A substituted for H3. At the
level of quaternary structure and assembly, we find that CENP-A
arrays are composed of octameric nucleosomes that assemble in
a stepwise mechanism, recapitulating conventional array assembly
with canonical histones. At intermediate structural resolution, we
find that CENP-A-containing arrays are globally condensed relative
to arrays with the canonical histones. At high structural resolution,
using hydrogen-deuterium exchange coupled to mass spectrome-
try (H/DX-MS), we find that the DNA superhelical termini within
each nucleosome are loosely connected to CENP-A, and we identify
the key amino acid substitution that is largely responsible for this
behavior. Also the C terminus of histone H2A undergoes rapid hy-
drogen exchange relative to canonical arrays and does so in a man-
ner that is independent of nucleosomal array folding. These
findings have implications for understanding CENP-A-containing
nucleosome structure and higher-order chromatin folding at the
hydrogen exchange ∣ epigenetics
The histone H3 variant, CENP-A, is a highly conserved con-
stituent of all eukaryotic centromeres and is the most attractive
candidate for carrying the epigenetic information that specifies
the location of the centromere (3). Recent findings have led
to several fundamentally different proposals for how CENP-A
marks centromere location: (i) CENP-A confers structural and
dynamic changes to octameric nucleosomes (4, 5); (ii) CENP-
A confers alterations of nucleosomal histone stoichiometry (6, 7),
including the incorporation of nonhistone proteins into nucleo-
some-like structures (8); (iii) CENP-A directs the reversal of
handedness of DNA wrapping from left to right (9). The nature
and composition of CENP-A-containing nucleosomes remain
controversial and areas of intense investigation. The available
data regarding their structure, however, could be reconciled by
distinct CENP-A-containing complexes existing over the course
of a cell cycle-coupled maturation program of newly expressed
CENP-A protein that propagates the epigenetic centromere
In the bulk chromatin fiber, nucleosome–nucleosome interac-
tions are central to packaging eukaryotic DNA into the nucleus,
to compacting chromosomes during mitosis, and to organizing
functional subchromosomal domains (11). Although much is
known about how chromatin fibers condense in vitro, the extent
to which the structured helical histone core of the nucleosome is
physically impacted by contact with neighboring nucleosomes in a
folded chromatin fiber is not yet known. Eukaryotic centromeres
he centromere is the control locus that directs the faithful
inheritance of eukaryotic chromosomes at cell division (1, 2).
are composed of lengthy arrays of CENP-A-containing nucleo-
somes. An exception is the budding yeast point centromere
that harbors a single CENP-A-containing nucleosome (12, 13).
Despite the central role of the specialized CENP-A-containing
nucleosomal array in specifying centromere location and direct-
ing chromosome inheritance, the internucleosomal interactions
of nucleosomal arrays in which CENP-A replaces canonical his-
tone H3 are completely unexplored.
To address the subunit structure and folding of CENP-A-
containing nucleosomal arrays, we couple folding measurements
using analytical ultracentrifugation (AUC) with mass spectrome-
try-based hydrogen/deuterium exchange (H/DX-MS). The AUC
studies measure the bulk behavior of the arrays, and we find that
CENP-A-containing arrays are somewhat more condensed upon
folding than canonical arrays. H/DX-MS is an approach capable
of measuring the dynamic behavior of the polypeptide backbone
of each histone in the nucleosome core. Prior H/DX-MS experi-
ments with subnucleosomal ðCENP-A∕H4Þ2heterotetramers
(14) and CENP-A-containing mononucleosomes (4) found that
the CENP-A/H4 interface is substantially rigidified by side-chain
interactions that restrict transient unfolding of the contacting
α-helices (5). In the present study, we demonstrate that the
αN-helices of canonical H3-containing nucleosomes are substan-
tially restricted (50- to 100-fold) at their superhelical termini
upon nucleosome array folding, indicating an unexpected conse-
quence of nucleosome-nucleosome interactions during chroma-
tin folding. Importantly, both the initial rigidity of CENP-A at
its own αN-helix and the rigidity imposed upon chromatin folding
are reduced compared to its conventional counterpart containing
H3, indicating looseness at the nucleosome superhelical termini.
Stepwise Assembly of CENP-A-containing Polynucleosome Arrays.
Conventional nucleosomes and nucleosomal arrays can be as-
sembled by adding the four core histones at the appropriate
molar ratios to defined DNA templates consisting of nucleosome
positioning sequences in order to saturate all of the nucleosome
binding sites. This is followed by salt dialysis assembly from 2 M
to 2.5 mM NaCl (15–17). The stepwise assembly is characterized
by the initial binding of the ðH3∕H4Þ2heterotetramer to DNA to
form a tetrasome as the [NaCl] is lowered to 1 M, followed by
Author contributions: T.P., J.C.H., and B.E.B. designed research; T.P., T.C.S., and C.L.W.
performed research; T.P., T.C.S., Z.-y.K., S.W., M.G.R., K.L., and S.W.E. contributed new
reagents/analytic tools; T.P., T.C.S., C.L.W., S.W.E., J.C.H., and B.E.B. analyzed data; and
T.P., S.W.E., J.C.H., and B.E.B. wrote the paper.
The authors declare no conflict of interest.
1T.P. and T.C.S. contributed equally to this work.
2To whom correspondence may be addressed. E-mail: email@example.com or
This article contains supporting information online at www.pnas.org/lookup/suppl/
16588–16593 ∣ PNAS ∣ October 4, 2011 ∣ vol. 108 ∣ no. 40www.pnas.org/cgi/doi/10.1073/pnas.1113621108
binding of H2A/H2B heterodimers to the tetrasome to complete
nucleosome formation as the [NaCl] is lowered to ≤0.6 M (16).
Whether CENP-A nucleosomes assemble by such a stepwise
mechanism has not been tested.
To determine if CENP-A-containing nucleosomal arrays as-
semble via the same sequential pathway as canonical arrays,
samples that contained ðCENP-A∕H4Þ2tetramers, (H2A/H2B)
dimers, and a linear DNA template containing twelve tandem
copies of the so-called 601 sequence (18, 19) were mixed in 2 M
NaCl and thensuccessively dialyzed against buffer containing 1 M
NaCl, 0.6 M NaCl and 2.5 mM NaCl. At each dialysis step, the
reactions were assayed by AUC (Fig. 1 A–C and Fig. S1 A and B).
AUC has been successfully employed in classic sedimentation
velocity experiments to measure physical properties of nucleoso-
mal arrays including array composition as well as global changes
that accompany intraarray nucleosome–nucleosome interactions
and interarray oligomerization (20). The CENP-A-containing
samples were compared with otherwise identical samples con-
taining canonical ðH3∕H4Þ2in place of ðCENP-A∕H4Þ2, as well
as samples that contained only ðH3∕H4Þ2(Fig. 1 A and B). As
expected, the canonical H3-containing nucleosomal arrays as-
sembled as previously reported, with the ðH3∕H4Þ2tetramers
bound to each DNA repeat at 1 M NaCl to form a 16–20 S
12-mer tetrasomal array (we measured the free 12-mer DNA
template sedimenting at approximately 13 S). This is followed by
dimer binding at 0.6 M NaCl to generate the completely as-
sembled 28–30 S beads-on-a string species also present in low
salt (i.e., 2.5 mM NaCl) (Fig. 1 B and C and Fig. S1 A and B).
Similarly, for the centromeric counterpart complexes, the
ðCENP-A∕H4Þ2tetramer binds to DNA in 1 M NaCl and forms
a 17–20 S tetrasomal array (Fig. 1B and Fig. S1A). In reactions
that contain ðCENP-A∕H4Þ2tetramers as well as H2A/H2B
dimers the H2A/H2B binding is prevented by the presence of
1 M NaCl and we still observe a 17–20 S array (Fig. 1 B and C).
Upon dialysis into 0.6 M NaCl, CENP-A-containing nucleosomal
array assembly is completed as H2A/H2B dimers bind to the
tetrasomes to form a stable 28–30 S nucleosomal array (Fig. 1 B
and C and Fig. S1 A and B). We confirmed by electron micro-
scopy that the 28–30 S CENP-A-containing arrays are saturated
with 12 nucleosomes per input linear DNA fragment (Fig. 1D and
Fig. S1 D and E). These results indicate that CENP-A arrays form
octameric nucleosomes via the same set of intermediate steps as
do conventional nucleosomal arrays.
Protection from H/DX Upon Nucleosome Array Folding. Despite
no predicted gross change in their static structures (5, 21), the
histone fold domains of ðH3∕H4Þ2and ðCENP-A∕H4Þ2undergo
>1000-fold slowing in H/DX rates upon incorporation into nu-
cleosomes (4). Further protection from H/DX upon nucleosome
array folding, however, was previously untested. Nucleosome–
nucleosome interactions are restricted in nucleosomal arrays in
the absence of cations, most notably Mg2þ(22). Addition of
1–2 mM Mg2þleads to nucleosomal array folding mediated by
internucleosomal contacts that condense the structure so signifi-
cantly that the 29 S array now sediments at 40–55 S (22) (Fig. 2 A
and B). Importantly, such folding behavior is observed in
1.25 mM Mg2þ, both with our conventional arrays assembled with
canonical recombinant human histones, and with the centromeric
counterparts containing CENP-A in place of H3 (Fig. 2 A and B).
This indicates that CENP-A does not function by locally decon-
densing nucleosomal array structure. Indeed, CENP-A-contain-
ing nucleosome arrays show a small but highly reproducible
shift toward a state that is condensed relative to canonical arrays
(Fig. 2 A and B).
H/DX measures how polypeptide backbone amide protons are
exchanged with deuterons from heavy water in solution. Folded
regions (e.g., α-helices and β-sheets) only exchange upon transi-
ent unfolding events when amide protons lose main chain hydro-
gen bonding. Slow exchange can be achieved by many stabilizing
interactions (23), including, in the case of DNA-binding proteins,
assembly into higher-order complexes with DNA (4, 24, 25).
In order to test the extent to which the polypeptide backbone
dynamics of the structured histone core of the nucleosomal sub-
units are altered by array folding, we monitored H/DX exchange
behavior of folded and unfolded array fibers (time points at 10,
100, 1,000, and 10,000 s at 23 °C) (Fig. 2C). To potentially detect
changes on the most rapidly exchanging regions we also included
an additional 10 s time point at 4 °C because such a reduction
in temperature leads to a nearly 10-fold slowing in chemical ex-
change rates at the amide protons that we can measure (23).
Throughout this time course, both H3- and CENP-A-containing
arrays remain intact and extensively folded as determined by
fining the assembly pathway of CENP-A containing nucleosome arrays. (B)
Average sedimentation coefficients (Saveis defined at 0.5 boundary fraction)
that were measured by AUC of all the different assembly species at varying
NaCl concentrations as outlined in A. ðCENP-A∕H4Þ2assembles onto the DNA
as a tetramer forming a species sedimenting at approximately 19 S, with com-
plete assembly of an approximately 29 S complex corresponding to a 12-mer
octameric nucleosome array that occurs at NaCl concentrations at or below
0.6 M. For corresponding AUC profiles of all 12 conditions shown in B see
Fig. S1 A and B. (C) Representative AUC profiles of CENP-A or H3 containing
nucleosome arrays analyzed at 1 M and 0.6 M NaCl. All sedimentation coef-
ficients have been corrected for temperature and normalized for water.
arrays showing that both form similar “beads-on-a-string” 12-mer arrays.
CENP-A nucleosome array assembly. (A) Experimental scheme for de-
Panchenko et al. PNAS
October 4, 2011
AUC (Fig. 2 A and B). We monitored the H/DX behavior of
overlapping peptides spanning the majority of the folded core
of the nucleosome with approximately 90% coverage of the his-
tone fold domains and approximately 60% coverage of the total
polypeptide length of all histones for both H3- and CENP-A-
containing arrays. There is a striking lack of folding-dependent
protection at most locations throughout either type of nucleo-
some core (Fig. S2). However, at peptides spanning much of
the αN helix of either H3 or CENP-A, we observed additional
protection from H/DX (Fig. 2 D and E and Fig. S2). In the
canonical nucleosome structure (21), the αN helix contacts the
DNA at the superhelical termini (i.e., the DNA entry/exit site)
of the nucleosome.
The αN Helix of CENP-A is Less Protected than That of H3 Upon Folding
of Nucleosomal Arrays. Although peptides spanning the αN helix
of CENP-A are additionally protected upon array folding, the
magnitude of this protection is less pronounced than for the
same region in H3 arrays (Figs. 2 D and E and 3 A–C). The smal-
ler magnitude of protection in this region of CENP-A, relative to
H3, was observed in several time points from three independent
experiments. H/DX data from several peptides that cover the
same region, spanning a portion of the αN helix and following
loop in each complex, can be used to compare the local effect
of nucleosomal array folding. Representative peptides are shown
in Fig. 3 B and C and Fig. S3. While nucleosomal array folding
slows the H/DX of the αN helix of H3 by 50–100 times compared
to unfolded arrays (Fig. 3B, compare −Mg2þto þMg2þ), array
folding only leads to a 5–10-fold slowing of H/DX in the corre-
sponding region of CENP-A (Fig. 3C). We also noted that in
addition to not being as restricted by nucleosomal folding at its
αN helix, this region of CENP-A exchanges approximately 10
times faster than the corresponding region in H3 prior to poly-
nucleosome folding (i.e., its beads-on-a-string; compare −Mg2þ
in Fig. 3 B and C). Indeed, the αN helix of CENP-A is nearly
as flexible in folded chromatin as is the αN helix of H3 in com-
pletely unfolded arrays (compare −Mg2þin Fig. 3B with þMg2þ
in Fig. 3C).
In the canonical H3-containing nucleosome, the αN helix lies
at the DNA entry/exit site (Fig. 3D) (21). Reconstituted CENP-A-
containing nucleosomes wrap approximately 150 bp of DNA in a
left-handed manner (5). The finding that CENP-A nucleosomal
arrays maintain local flexibility after chromatin folding builds on
the earlier observation that topologically constrained DNA mini-
circles containing a single CENP-A nucleosome prefer a more
“open” conformation where the exiting DNA strands do not
cross (26). In our reconstituted arrays on a repetitive and strongly
positioning DNA sequence, digestion with MNase clearly pro-
tects less DNA (Fig. 3E). CENP-A-containing arrays have a clear
pause site at approximately 150 bp (Fig. 3E, 0.2 U MNase), con-
sistent with a heavily populated steady-state species that is a fully
wrapped octameric nucleosome. Further, CENP-A-containing
nucleosomes dynamically and transiently release their DNA
superhelical termini to allow the digestion of an additional one
or two turns of DNA (i.e., 10–20 bp) at either terminus to yield a
fragment of approximately 125 bp upon extensive digestion
(Fig. 3E, 2 U MNase).
Arg49 of H3 Contributes to Rigidity at the Superhelical Nucleosome
Termini in Folded Arrays. One likely contribution to the increased
flexibility in folded CENP-A-containing nucleosomal arrays is
the substitution of a lysine at the position corresponding to Arg49
in histone H3. Arg49 of H3 intercalates into the DNA one half
turn from the superhelical terminus of the nucleosomal DNA
(Fig. 4A) (21). Indeed, with mononucleosomes assembled onto
topologically constrained minicircles, the R49K mutation leads
to a preference for an open DNA entry/exit arrangement (as
opposed to with entering/exiting strands crossed in a closed
arrangement) to nearly the same extent as seen with CENP-A
mononucleosomes (26). The R49K mutation generates increased
local flexibility, measured in six out of six peptides that span at
least some portion of the αN helix, including adjacent peptides
whose amino acid composition is identical to WT H3 (Fig. 4B
and Fig. S4). The increased flexibility in H3 R49K nucleosomes
is more pronounced in folded than unfolded arrays. While this
substitution does not account for the full extent of the flexibility
observed in CENP-A-containing nucleosomal arrays (Fig. 3), the
lysine in place of the DNA intercalating arginine is clearly a major
contributor. Thus, the R49K mutation in H3 generates an inter-
mediate level of flexibility that also corresponds with an inter-
mediate level of sensitivity to digestion of superhelical DNA
termini by MNase (Fig. 4C).
Beyond the αN helix, we also observed a region of H2A
(a.a. 91–134) with 10 out of 11 peptides showing faster exchange
rates in CENP-A containing arrays (approximately 10-fold faster
than in H3-containing nucleosomes; representative peptides are
shown in Fig. S5). This observation was missed in earlier efforts
(4) but was readily detected in the present study where many
more nucleosome-derived peptides could be monitored due to
major technical improvements in peptide resolution. Unlike
the αN helices of H3 and CENP-A, the difference in H2A at this
location is independent of nucleosomal folding (Fig. S5).
AUC profiles showing MgCl2dependent array folding and array stability over
the course of the H/DX experiment. (C) Experimental scheme for determining
differences in protection from H/DX upon nucleosome array folding. Protec-
tion profiles of CENP-A containing nucleosome arrays at 100 s are repre-
sented in D, and protection profiles of H3 containing nucleosome arrays at
100 s are represented in E. Protection upon array folding is calculated for
each peptide in each array type as the difference of percent deuterium con-
tent of the peptide from array in the “unfolded” state minus the percent
deuterium content of the same peptide in the “folded” state. Peptides are
represented by horizontal bars and color coded based on the difference in
protection upon folding.
Protection from H/DX upon nucleosome array folding. (A and B)
www.pnas.org/cgi/doi/10.1073/pnas.1113621108Panchenko et al.
CENP-A in the Nucleosomal Array Context. The measurement of
>100 partially overlapping peptide “probes” that span the major-
ity of each histone subunit in canonical and CENP-A-containing
nucleosomal arrays provide a high-resolution view of these
structures. Previous high-resolution/site-specific studies of the
dynamic and structural impact of CENP-A focused on the sub-
nucleosomal tetramer that it forms with histone H4 (5, 14, 27),
mononucleosomes (4, 26), and the ternary complex that CENP-A
and H4 forms with the centromeric chromatin assembly protein
HJURP (or the HJURP counterpart, Scm3, in yeast) (27–29).
The CENP-A targeting domain (CATD), comprised of the L1
and α2-helix of CENP-A, contributes hydrophobic stitches that
rigidify the interface between CENP-A and H4 (5, 14). This
rigidity is maintained after assembly into mononucleosomes (4).
L1, within the CATD, generates a surface on the face of the
CENP-A nucleosome that is divergent in shape and electrostatic
charge (5) from the corresponding surface on canonical H3-
containing nucleosomes (21). These features structurally and
dynamically distinguish CENP-A mononucleosomes from con-
ventional ones. The essential role of the CATD in centromere
function (30) strongly suggests that these features are key to
distinguishing centromeric chromatin from the rest of the chro-
Beyond individual mononucleosomes, each centromere is
made up of many CENP-A mononucleosome subunits. CENP-
A-containing nucleosomes coalesce on the face of the chromo-
some to define the location of the mitotic kinetochore assembly,
but they are not arranged in a linear context. Instead they are
interspersed with conventional H3-containing nucleosomes (31).
Whether or not there is a regular geometry to centromeric
chromatin organization (31–34), it is very likely that internucleo-
somal contacts between CENP-A-containing nucleosomes are
fundamental in organizing centromeric chromatin. Using H/DX-
MS, we have found that a major difference between the subunit
structures of folded CENP-A- and H3-containing nucleosomal
arrays is increased flexibility at the αN helix of CENP-A that
contacts superhelical termini of each nucleosome.
After this work was complete, a crystal structure of an octa-
meric CENP-A-containing nucleosome was reported (35). In the
crystal structure, the terminal 13 bp of DNA are not visible (35),
consistent with our observations of additional flexibility in the
beads-on-a-string configuration (Fig. 3 B and C; −Mg2þ). The
H/DX studies of nucleosome array folding show the difference
at this site between CENP-A- and H3-containing nucleosomes
is much greater in folded than unfolded arrays (Fig. 3 B and C).
This finding greatly extends our understanding of CENP-A
beyond mononucleosomes (4, 5, 35) because it provides the first
view of the dynamics and structure of the CENP-A-containing
Dynamics at the Superhelical Termini of the Nucleosomal Subunits
of Folded Arrays. For both ðH3∕H4Þ2and ðCENP-A∕H4Þ2het-
erotetramers, assembly into nucleosomes causes major H/DX
protection along their polypeptide backbones (approximately
1,000-fold slower exchange throughout the respective histone
fold domains) (4). However, the internucleosomal contacts that
accompany nucleosomal array folding apparently remain fluid
and do not further restrict the backbone dynamics of the bulk of
the octameric histone core of the nucleosome (Fig. 2 D and E
and Fig. S2). The observation that the αN helix of H3-containing
arrays is substantially affected upon folding (50–100-fold slower
H/DX; Fig. 3 B and C and Fig. S3) was not predictable from
earlier crystal structures. In both nucleosome (21) and tetranu-
cleosome (36) static structural models, the local structure of
histone and DNA are not changed (21, 36). However, it is inter-
esting that rigidity is required for linker DNA and up to 10 bp of
Boxed region highlights the sequence of the representative peptides in B and C. (B) H/DX of a representative αN-H3 peptide from nucleosome arrays containing
H3. Top panel shows normalized deuterium levels at each time point for arrays that are either “unfolded” (−Mg2þ) or “folded” (þMg2þ) and bottom panels
show raw peptide data where the dotted blue and pink lines are drawn as guides to visualize differences in H/DX and red arrows indicate peptide centroid
values. (C) H/DX of a representative αN-CENP-A peptide from nucleosome arrays containing CENP-A. Top and bottom panels are displayed as in B. (D) Schematic
representation of the locations where H3 αN helix contacts nucleosomal DNA. (E) MNase digestion of CENP-A or H3 containing nucleosome arrays.
Increased flexibility at the αN helix of CENP-A in nucleosomal arrays. (A) Alignment of H3 and CENP-A sequences in the αN and adjacent loop region.
Panchenko et al.PNAS
October 4, 2011
terminal nucleosome core DNA in order to make an idealized
chromatin fiber model based on the tetranucleosome crystal
structure (36). It is also interesting that in the leading models
for nucleosome higher-order packing, the nucleosome superhe-
lical termini always face the interior of the fiber (37). Our results
suggest that the fiber interior imposes the most rigidity on nucleo-
Our studies also conclusively demonstrate that CENP-A-con-
taining arrays can be readily assembled in a stepwise manner and
form condensed fibers in a manner highly similar to canonical
nucleosomal arrays (Figs. 1 A–C and 2 A and B). Prior to our
present study, the data indicating whether or not CENP-A-con-
taining complexes can assemble nucleosomes in a stepwise man-
ner was not clear. Indeed, it has been reported that after assembly
of the ðCENP-A∕H4Þ2heterotetramer a 147 bp DNA template
(i.e., the number of base pairs that wraps the canonical nucleo-
some core particle) resulted in formation of complexes that
showed native gel migration consistent with two stacked hetero-
tetramers, not a single one as for canonical ðH3∕H4Þ2heterote-
tramers assembled onto the same template (26). However, the
exact nature of the CENP-A “tetrasomes” was not explored
further. Nonetheless, the apparent deviation from the canonical
tetrasome (i.e., ½H3∕H4?2þ 147 bp DNA) behavior (26) was used
to strongly question the relevance of subsequently reconstituted
octamers to bona fide centromeric chromatin (38).
The histone H3 N-terminal tail is known to participate in array
folding (39). We note that the CENP-A N-terminal tail, which
is completely divergent in sequence identity from bulk H3 but
maintains the strongly basic charge, does not preclude the nucleo-
some–nucleosome interactions that lead to formation of exten-
sively folded array structures (Fig. 2 A and B). Indeed, we
always observe that at the same Mg2þconcentration, CENP-A
nucleosomal arrays are somewhat more condensed than canoni-
cal arrays (as judged by the right-shifted sedimentation profiles).
It is attractive to speculate that the enhanced propensity to fold
is related to the loosened superhelical termini of CENP-A nu-
cleosomes, perhaps working in conjunction with the specialized
CENP-A N-terminal tail.
In budding yeast, one of a subset of fungal species where
centromere location is determined not by epigenetic informa-
tion but by a specific 125 bp DNA sequence, the CENP-A coun-
terpart, Cse4p, assembles into an octameric nucleosome that
protects only 110–125 bp of DNA with no evidence of further
wrapping (40, 41). We note that at the position corresponding
to Arg49 in histone H3 (where substitution to the lysine corre-
sponding to human CENP-A leads to increased flexibility; Fig. 4
and Fig. S4), Cse4p has substituted a tyrosine that likely prevents
any DNA wrapping by its αN helix. It seems likely that the
Arg → Tyr substitution evolved to prevent full 145 bp DNA wrap-
ping and to accommodate the constraints imposed by a system in
which the centromere is defined by a particular DNA sequence.
In many eukaryotes, including animals, CENP-A-containing
nucleosomes are thought to be arranged in several local clusters
of adjacent nucleosomes per centromere interspersed by clusters
of H3-containing nucleosomes (31). Although the exact numbers
of nucleosomes that make up each cluster is not known, our
12-mer arrays represent a single cluster. Our AUC experiments
show that CENP-A-containing arrays are generally more con-
densed than canonical arrays upon folding, while our H/DX
experiments demonstrate local flexibility at the contact point
between the CENP-A αN-helix and the nucleosomal terminal
DNA. We suggest that both characteristics (Fig. 5) are important
for centromeric chromatin. The condensed nature of the array
may reflect strong self–self interactions that culminate in the
coalescence of discontinuous CENP-A nucleosomes on each cen-
tromere into a discrete focus on the surface of the chromosome
(31). In addition, the condensed nature of the array may be
important to help maintain structural integrity of centromeric
chromatin under mitotic spindle stretching forces. The increased
flexibility at the DNA entry/exit site may accommodate internu-
cleosomal DNA paths unique to the centromere (31, 33, 34) and/
or nucleosomal (or nucleosome-proximal) binding proteins that
require increased local access to the DNA and/or histone core.
Alternatively, it is formally possible that the increased flexibility
at the nucleosomal termini of the human CENP-A protein is a
vestige of a common ancestral orthologue that may have lacked
any stable DNA binding at the superhelical termini (i.e., similar
to the behavior of Cse4 in Saccharomyces cerevisiae) (40, 41).
Rapid Exchange on the C Terminus of Histone H2A in CENP-A-Contain-
ing Nucleosomes. The region of H2A (a.a. 91–134) that undergoes
structure of the human nucleosome particle containing histone H3 (PDB ID
code 2CV5) (44) is shown to highlight the location of Arg49 in relation to the
αN helix and the DNA entry/exit site. (B) Representative peptides from αN
region of H3 containing arrays and mutant H3 R49K arrays. Boxed region
highlights the sequence of the representative peptide. (C) MNase digestion
of the H3 R49K nucleosome containing arrays.
Arg49 of H3 contributes to a more rigidified folded fiber. (A) Crystal
CENP-A- and H3-containing nucleosomal arrays. AUC experiments showed
that CENP-A-containing arrays are generally more condensed than canonical
arrays upon folding (indicated by closer spacing of adjacent CENP-A-contain-
ing nucleosomes). At the same time the H/DX experiments measured local
relative flexibility at the CENP-A αN-helix, indicating that the DNA at the
entry/exit sites is less constrained (indicated by uncrossed internucleosomal
DNA for CENP-A-containing arrays). See the text for a discussion of the
potential implications of these findings.
Summary of physical differences identified in this study between
www.pnas.org/cgi/doi/10.1073/pnas.1113621108Panchenko et al.
rapid exchange in CENP-A-containing nucleosomes is juxta-
posed to the αN-helix of H3 and is exposed on the surface of
the nucleosome in the conventional nucleosome structure (21).
Therefore, the increased exchange rates in H2A (a.a. 91–134)
is possibly related to the flexibility of the αN-helix of CENP-A.
This seems unlikely, however, because there is no correlation
with the αN-helix behavior. The H2A H/DX in this region is
not substantially slowed in H3-containing arrays upon array fold-
ing (Fig. S5) at the same time points where up to a 100-fold
slowing is observed in the αN-helix of H3 (Fig. 3C). Another
possibility is that the increased H2A H/DX in CENP-A-contain-
ing nucleosomes reflects a different orientation of H2A that
accompanies rotation of the H2A/H2B dimers. ðCENP-A∕H4Þ2
heterotetramers prefer a compact state, either in solution or
in crystals, which comes about by rotation at the CENP-A/
CENP-A four-helix bundle (5). Upon nucleosome formation, the
CENP-A/CENP-A interface may rotate to form a nucleosome
of similar shape to the canonical nucleosome (5, 35), or the
H2B/H4 four-helix bundles could rotate H2A/H2B dimers away
from the central axis of the nucleosome to avoid steric clashes (5).
Our H/DX data suggests that the possibility of whether one or
both of these two states of CENP-A-containing nucleosomes
are substantially populated at centromeres should be the subject
of further careful analysis.
H/DX Reactions. Nucleosome arrays (assembled and characterized by EM,
AUC, and MNase digestion as detailed in SI Methods) were incubated for
30 min at 4°C or 23°C with either 1 × TEN or 1 × TEN with 1.25 mM
MgCl2prior to starting deuterium on-exchange reactions. Deuterium on-ex-
change was carried out by adding 5 μLof the array (containing approximately
3.8 μg of array) to 15 μL of deuterium on-exchange buffer (10 mM Tris,
pD 7.0, 0.25 mM EDTA, 2.5 mM NaCl in D2O; supplemented with 1.25 mM
MgCl2where indicated) so that the final D2O content was 75%. Reactions
were quenched at the indicated time points by withdrawing 20 μL of the
reaction volume, mixing in 30 μL ice cold quench buffer (2.5 M GdHCl,
0.8% formic acid, 10% glycerol), and rapidly freezing in liquid nitrogen
prior to proteolysis and LC-MS steps (detailed in SI Methods).
H/DX Data Analysis. MATLAB-based MS data analysis tool ExMS was used
for data processing (42). Detailed information regarding the ExMS algorithm
is described elsewhere (42) and briefly in SI Methods. The level of H/DX
occurring at each time point is expressed as either the number of deuterons
or the percentage of exchange within each peptide. In each case, corrections
for loss of deuterium label by individual peptides during H/DX-MS analysis
(back exchange) were made through measurement of loss of deuterium from
reference samples (fully deuterated control, FD) that had been deuterated
under denaturing conditions as described elsewhere (43). The loss of deuter-
ium for the FD sample was approximately 10% for most peptides, and the
measured centroid values for several of the αN helix-containing peptides
are shown as part of Fig. S3. Calculation of deuterium loss correction and
other data operations were performed using MATLAB.
ACKNOWLEDGMENTS. We thank D.W. Cleveland for plasmids and M.U. Salman
for assistance with MATLAB code. This work was supported by National
Institutes of Health (NIH) grants GM045916 (J.C.H.) and GM082989 (B.E.B.).
T.P. was supported by NIH Grant GM08275 (University of Pennsylvania Struc-
tural Biology Training Grant). This work is also supported by a Career Award
in the Biomedical Sciences from the Burroughs Wellcome Fund and a
Rita Allen Foundation Scholar Award (B.E.B.).
1. Cleveland DW, Mao Y, Sullivan KF (2003) Centromeres and kinetochores: from
epigenetics to mitotic checkpoint signaling. Cell 112:407–421.
2. Allshire RC, Karpen GH (2008) Epigenetic regulation of centromeric chromatin: old
dogs, new tricks? Nat Rev Genet 9:923–937.
3. Panchenko T, Black BE (2009) The epigenetic basis for centromere identity. Prog Mol
Subcell Biol 48:1–32.
4. Black BE, Brock MA, Bédard S, Woods VL, Jr, Cleveland DW (2007) An epigenetic mark
generated by the incorporation of CENP-A into centromeric nucleosomes. Proc Natl
Acad Sci USA 104:5008–5013.
5. Sekulic N, Bassett EA, Rogers DJ, Black BE (2010) The structure of (CENP-A-H4)(2)
reveals physical features that mark centromeres. Nature 467:347–351.
6. Williams JS, Hayashi T, Yanagida M, Russell P (2009) Fission yeast Scm3 mediates stable
assembly of Cnp1/CENP-A into centromeric chromatin. Mol Cell 33:287–298.
7. Dalal Y, Wang H, Lindsay S, Henikoff S (2007) Tetrameric structure of centromeric
nucleosomes in interphase Drosophila cells. PLoS Biol 5:e218.
8. Mizuguchi G, Xiao H, Wisniewski J, Smith MM, Wu C (2007) Nonhistone Scm3 and
histones CenH3-H4 assemble the core of centromere-specific nucleosomes. Cell
9. Furuyama T, Henikoff S (2009) Centromeric nucleosomes induce positive DNA
supercoils. Cell 138:104–113.
10. Black BE, Cleveland DW (2011) Epigenetic centromere propagation and the nature of
CENP-A nucleosomes. Cell 144:471–479.
11. Szerlong HJ, Hansen JC (2011) Nucleosome distribution and linker DNA: Connecting
nuclear function to dynamic chromatin structure. Biochem Cell Biol 89:24–34.
12. Meluh PB, Yang P, Glowczewski L, Koshland D, Smith MM (1998) Cse4p is a component
of the core centromere of Saccharomyces cerevisiae. Cell 94:607–613.
13. Furuyama S, Biggins S (2007) Centromere identity is specified by a single centromeric
nucleosome in budding yeast. Proc Natl Acad Sci USA 104:14706–14711.
14. Black BE, et al. (2004) Structural determinants for generating centromeric chromatin.
15. Simpson RT, Thoma F, Brubaker JM (1985) Chromatin reconstituted from tandemly
repeated cloned DNA fragments and core histones: A model system for study of
higher order structure. Cell 42:799–808.
16. Hansen JC, van Holde KE, Lohr D (1991) The mechanism of nucleosome assembly onto
oligomers of the sea urchin 5 S DNA positioning sequence. J Biol Chem 266:4276–4282.
17. Hansen JC, Ausio J, Stanik VH, Van Holde KE (1989) Homogeneous reconstituted
oligonucleosomes, evidence for salt-dependent folding in the absence of histone
H1. Biochemistry 28:9129–9136.
18. Lowary PT, Widom J (1998) New DNA sequence rules for high affinity binding to his-
tone octamer and sequence-directed nucleosome positioning. J Mol Biol 276:19–42.
19. Dorigo B, Schalch T, Bystricky K, Richmond TJ (2003) Chromatin fiber folding: require-
ment for the histone H4 N-terminal tail. J Mol Biol 327:85–96.
20. Hansen JC (2002) Conformational dynamics of the chromatin fiber in solution: Deter-
minants, mechanisms, and functions. Annu Rev Biophys Biomol Struct 31:361–392.
21. Luger K, Mäder AW, Richmond RK, Sargent DF, Richmond TJ (1997) Crystal structure
of the nucleosome core particle at 2.8 Å resolution. Nature 389:251–260.
22. Schwarz PM, Hansen JC (1994) Formation and stability of higher order chromatin
structures. Contributions of the histone octamer. J Biol Chem 269:16284–16289.
23. EnglanderSW (2006) Hydrogenexchange and mass spectrometry: A historical perspec-
tive. J Am Soc Mass Spectrom 17:1481–1489.
24. Kalodimos CG, et al. (2004) Structure and flexibility adaptation in nonspecific and
specific protein–DNA complexes. Science 305:386–389.
25. Hansen JC, et al. (2011) DNA binding restricts the intrinsic conformational flexibility
of methyl CpG binding protein 2 (MeCP2). J Biol Chem 286:18938–18948.
26. Conde e Silva N, et al. (2007) CENP-A-containing nucleosomes: Easier disassembly
versus exclusive centromeric localization. J Mol Biol 370:555–573.
27. Cho U-S, Harrison SC (2011) Recognition of the centromere-specific histone Cse4 by
the chaperone Scm3. Proc Natl Acad Sci USA 108:9367–9371.
28. Hu H, et al. (2011) Structure of a CENP-A-histone H4 heterodimer in complex with
chaperone HJURP. Genes Dev 25:901–906.
29. Zhou Z, et al. (2011) Structural basis for recognition of centromere histone variant
CenH3 by the chaperone Scm3. Nature 472:234–237.
30. Black BE, et al. (2007) Centromere identity maintained by nucleosomes assembled
with histone H3 containing the CENP-A targeting domain. Mol Cell 25:309–322.
31. Blower MD, Sullivan BA, Karpen GH (2002) Conserved organization of centromeric
chromatin in flies and humans. Dev Cell 2:319–330.
32. Zinkowski RP, Meyne J, Brinkley BR (1991) The centromere-kinetochore complex:
A repeat subunit model. J Cell Biol 113:1091–1110.
33. Santaguida S, Musacchio A (2009) The life and miracles of kinetochores. EMBO J
34. Ribeiro SA, et al. (2010) A super-resolution map of the vertebrate kinetochore. Proc
Natl Acad Sci USA 107:10484–10489.
35. Tachiwana H, et al. (2011) Crystal structure of the human centromeric nucleosome
containing CENP-A. Nature 476:232–235.
36. Schalch T, Duda S, Sargent DF, Richmond TJ (2005) X-ray structure of a tetranucleo-
some and its implications for the chromatin fibre. Nature 436:138–141.
37. Dorigo B, et al. (2004) Nucleosome arrays reveal the two-start organization of the
chromatin fiber. Science 306:1571–1573.
38. Henikoff S, Furuyama T (2010) Epigenetic inheritance of centromeres. Cold Spring
Harb Symp Quant Biol 75:51–60.
39. Zheng C, Lu X, Hansen JC, Hayes JJ (2005) Salt-dependent intra- and internucleosomal
interactions of the H3 tail domain in a model oligonucleosomal array. J Biol Chem
40. Kingston IJ, Yung JSY, Singleton MR (2011) Biophysical characterization of the
centromere-specific nucleosome from budding yeast. J Biol Chem 286:4021–4026.
41. Dechassa ML, et al. (2011) Structure and Scm3-mediated assembly of budding yeast
centromeric nucleosomes. Nat Commun 2:313.
42. Kan ZY, Mayne L, Englander SW (2011) ExMS: A data analysis program for HX-MS ex-
periments. J Am Soc Mass Spectrom, in press.
43. Zhang Z, Smith DL (1993) Determination of amide hydrogen exchange by mass spec-
trometry: A new tool for protein structure elucidation. Protein Sci 2:522–531.
44. Tsunaka Y, Kajimura N, Tate S, Morikawa K (2005) Alteration of the nucleosomal DNA
path in the crystal structure of a human nucleosome core particle. Nucleic Acids Res
Panchenko et al.PNAS
October 4, 2011